Preview

INTERNATIONAL LABOUR OFFICE

FAMILY LIVING STUDIES
A Symposium

INTERNAI ¡ONAL INSTITUTE
FOi LABOUR STUDIES

- 5 OCT 1961

LI B R A R Y

GENEVA
1961

STUDIES AND REPORTS
New Series. No. 63

PRINTED BY " LA TRIBUNE DE GENÈVE ", GENEVA, SWITZERLAND

CONTENTS
Page
INTRODUCTION

I
II

A Budget Survey in Sweden, by Dr. Erland von HOFSTEN . . .
The Continuous Budget Survey in the United Kingdom, by R. F.
FOWLER and L. Moss

III

15
36

The Continuous Budget Survey in the U.S.S.R., by S. V. POSTNIKOV

A Budget Survey in the Urban Areas of Greece, by A. B. REISZ

A Farm Household Budget Survey in Germany (Federal Republic),
by Herbert KÖTTER and Joachim LUNZE

International Comparison : Budget Survey Made by the E.C.S.C.,
by Rolf WAGENFÜHR

103

VTT A Survey of Food Consumption in Great Britain, by D . F .
HOLLINGSWORTH and A. H. J. BAINES

VIII
IX

120

A Survey of Underemployment in Puerto Rico, by A. J. JAFFE
A Population Survey in Mysore State, India, by C. CHANDRASEKARAN

X
XI
XII
XIII
XIV

156

A Health Survey in Japan, by Takemune SODA

175

Repeated Surveys of Consumer Finances in the United States,
by James MORGAN

191

A Survey of Living Conditions for Appraising Community
Development Programmes in the Philippines, by B. G. BANTEGUI

207

The Continuous Population and Labour Force Survey in the
United States, by A. Ross ECKLER

221

An Inquiry into Levels of Living in an Area of the Ivory Coast,
by J. L. BOUTILLIER and J. CAUSSE

139

Repeated Surveys of Rural Living Conditions in India

SELECTED REFERENCES

245

. . . .

269'
279'

Introduction
Modern family living studies find their origin in the surveys of poverty
conducted from 1829 to 1879 by Frédéric Le Play, whose monographs
describe in detail the living conditions of a number of workers' families
selected as typical of large groups of workers. Many other sociologists
or statisticians, such as Quételet, Engel, Wright and Mayhew, undertook
similar surveys in the second half of the nineteenth century. Perhaps the
outstanding survey of the type was the monumental inquiry on hving
conditions of the London working class, initiated in 1886 by Charles
Booth, the results of which were published in 1902 in a 17-volume
edition.
The techniques used in these surveys were far from scientific. With
Rowntree's use of interviewers in 1901 and the introduction of sampling
by Bowley in 1912, however, techniques improved greatly.
The public, stirred by the sociologists' findings concerning poverty
among the working classes, soon demanded larger and more complete inquiries than could be conducted with private means as well as completely impartial data, not influenced by the personal views of the inquirers.
The statistical services of central governments were thus led to undertake family hving studies, and by the outbreak of the First World War,
many such studies had been conducted by national administrations.
Though it was initiated to study and publicise the distressing living
conditions of the poor, the family living study soon became a method of
collecting data for many other purposes. The rapid rise in the cost of
living during and after the First World War aroused interest in studies
of the impact of price changes on the consumers' economic situation.
The development of good consumer price indices involved the use of
weights that would properly reflect the consumption expenditure of the
population. This led to further extension of family hving studies in
different countries and for different periods, mainly to secure information on patterns of consumption expenditure. The need for a more
scientific approach to the study of poverty than existed in the earlier,
nineteenth-century efforts, led also to the development of more specialised family living studies, aimed at the measurement of actual food
consumption levels and nutrition deficiencies.

FAMILY LIVING STUDIES

The methods followed in these surveys, however, did not undergo
many changes until the nineteen-thirties. But the emergency agencies
set up in the United States after 1933 to deal with the economic and
social problems of the depression undertook numerous social surveys.
In Jbeir efforts to improve the quality of the survey results, they gave
increasing attention to scientific methods of household interviewing,
sample selection, etc. By 1940, when labour force sample surveys were
started in the United States, interviewing and sampling techniques had
already reached an advanced stage, compared to the crude methods
used in the past.
One of the consequences of the Second World War was, in many
countries, a transformation of consumption habits. Pre-war patterns of
consumption were no longer valid, and a new round of family living
studies was needed to supply the necessary data for the making of new
consumer price indices.
The war and post-war conditions also gave rise to the need for a wide
range of information on all aspects of living conditions, such as nutrition,
health, education and employment. On most of these subjects little
information was available from other sources of statistical data; the
statistical services in many countries had been considerably developed
during the war, however, and they were in a position to meet many of
these needs by the collection of data through various sample surveys.
Post-war conditions also focused attention on the problems of the
underdeveloped countries, where practically no data were available
in most cases. It soon became clear that household surveys offered
the best means of collecting many types of essential information in those
countries of the world that had not yet developed systems for the regular
collection of statistics.
As a result, the number and variety of household surveys has greatly
expanded in the post-war years. These studies are of two general types :
specialised and multi-subject. Examples of each type are described in
the following chapters, including examples of continuous surveys of
both types.
Although many significant advances have been made in household
survey techniques, problems of slow communication, language barriers
and limited resources have restricted the dissemination of information
regarding the newer methods. The objective of this Symposium is to
bring together reports on a representative selection of current surveys of
these types in various parts of the world, to illustrate their range and
flexibility, and to suggest possibilities for the application of this survey
method in solving some of the statistical problems of the newly developing countries.

INTRODUCTION

TYPES OF FAMILY LIVING STUDIES AND THEIR USES

The range of data obtainable through household inquiries is indicated
in the following paragraphs describing various types of specialised
surveys, and multi-subject and continuous surveys.
Family budget surveys are specialised surveys in which the bulk of the
data collected relates to consumption expenditure. Their usefulness is
enhanced if they also include, as is often the case, information on
amount and sources of income, number of income earners per household,
transfer payments and savings.
Budget surveys usually provide detailed data on consumer expenditure for commodities and services needed for the construction or revision
of consumer price indices. The details supplied by such surveys on consumer purchases, together with estimates of the value of items acquired
without monetary payment, are also useful in the computation of national
accounts. They provide the basis for comparisons of consumption
levels at different periods and between different population groups.
Estimates of average amounts of expenditure on different items, of
average values of individual purchases, and of the distribution of purchases over time, as well as between different types of market outlets,
provide a basis for consumer demand analysis and market research.
In the more advanced countries, family budget surveys sometimes provide information on consumers' plans and expectations, amounts and
types of purchases on an instalment payment basis, and ownership
of durable goods and other assets.
In food consumption surveys, information is collected on the types and
quantities of food consumed by the households surveyed. Analyses of
such data indicate the nutrient content of the food consumed for
comparison with given nutritional standards. The data are also helpful
in the preparation of estimates of total consumption. Surveys of this
type require a more elaborate definition of consumption than is used in
the common family budget survey, inasmuch as attention is focused on
quantities and nutritional values rather than monetary outlays.
Health surveys use the household as a reporting unit in order to collect information on health and sanitary conditions through interrogation
and direct observation by well-trained investigators. Data obtained may
include, for example, the number of days of absence from school because
of illness, information on inoculations and vaccinations received by
members of the family, the number of days of hospitalisation, the costs
incurred and the methods of payment, and evidence of the need for
sanitation programmes.

FAMILY LIVING STUDIES

Labour force sample surveys are designed to obtain more complete
and up-to-date information about the employment status of the population than is usually available through population censuses, establishment
surveys or unemployment registers. Surveys of this type are especially
effective for obtaining periodic estimates of unemployment or part-time
employment, and characteristics of the labour force including its distribution by sex and age, occupation, etc.
In underdeveloped countries, where the concepts and definitions of
employment and unemployment used in industrialised countries may not
be applicable, similar surveys have been developed where data such as
the number of hours worked per day, the number of days worked per
year, and the amount of wages or income are collected in order to assess
also the amount of visible or invisible underemployment.
Demographic surveys are used to obtain data on birth, death and
fertility rates in countries or regions where vital records are incomplete
or non-existent. They also furnish the basis for population estimates
during intercensal periods, and in some instances may be useful as test
checks on the accuracy of data obtained from other sources.
Education surveys may provide data on literacy, school enrolment
and attendance, standards of education, books and newspapers read,
and the use of other cultural facilities.
In housing surveys, the field investigator visits the family dwelling to
obtain information on the physical condition of the dwelling unit, the
number of occupants, whether the family rents or owns its living quarters, availability of public facilities such as water, fuel, telephone service,
etc., and other related information.
The value of these data is obvious in connection with estimating the
need for housing programmes, extension and improvement of public
utilities, and studies of the relation between overcrowding and health
and social problems.
These are some of the more common types of specialised family
living studies. Although in each instance the study focuses on a particular problem, some more general information, such as the economic
status of the family, is frequently included in order to facilitate the analysis of the primary data.
In contrast, multi-subject surveys give equal emphasis to a broad
spectrum of data, combining two or more major topics, such as family
characteristics, income, employment, education, housing, nutrition,
health, etc. Where statistical systems are poorly developed and an
immediate need exists for a wide variety of data, the multi-subject
survey has advantages of speed and economy. In addition to the lower

INTRODUCTION

total cost of using the same sample and a single interview to collect data
on several topics, the multi-subject survey also has the advantage that
the different aspects of levels of living can be seen in relation to each
other. At the same time, however, multi-subject surveys involve difficult
problems of sample and questionnaire design, and risks of interviewer
and respondent bias resulting from the lengthy interviews that may be
required if an attempt is made to cover too much ground in a single
interview. Nevertheless, the multi-subject survey is likely to be employed
on an increasing scale as statisticians develop familiarity with the method
and as the public becomes accustomed to responding to such inquiries.
In contrast to the specialised and multi-subject inquiries that are
ad hoc in nature, the continuous surveys involve periodic interviews with
the same or a similar sample of families to obtain information on trends
in unemployment, family income, morbidity, etc. The Current Population Survey in the United States is an example. Another is the continuing
family budget survey conducted in the United Kingdom for the purpose
of checking on the representativeness of the retail price index and to
provide data for use in national accounting. Continuous surveys of
various types have also been in use for several years in some of the
Eastern European countries. Since it may be impracticable to secure in
a single inquiry all the data desired on the various aspects of levels of
living, some continuous surveys have been made to achieve the comprehensive character of multi-subject surveys by the cumulative effect of
a variety of questions, spread over several survey " rounds ".
It will be seen that the uses and objectives of family living studies are
manifold. Family living studies may be used to provide materials for
research into the living conditions and the behaviour patterns of different
groups of the population. Family living studies can also supply the basic
data needed for policy-making in connection with social and economic
planning. Objectives related to policy-making include the establishment
of norms or the determination of needs, in preparation for social and
economic measures, as well as the assessment of the impact of policy
decisions already applied in implementing welfare programmes.
Both highly developed industrial countries and underdeveloped
areas have a need, in connection with their planning programmes, for
national and regional data on a number of demographic, social and
economic characteristics of the population. In the more developed
countries, a large amount of such information can usually be obtained
through censuses or administrative records, but family living studies are
used to fill gaps in the existing information, to provide checks on the
reliability and completeness of the existing data and to obtain estimates
between census dates. In less developed areas, where basic data are

FAMILY LIVING STUDIES

often lacking, family living studies represent the most direct way of
assessing the needs and resources of the population and of obtaining
the background data essential for national and regional planning.

MAIN FEATURES OF MODERN FAMILY LIVING STUDIES AND PRINCIPAL
PROBLEMS

Even though the objectives served by the various types of family
living studies may be different, the general methods followed in these
surveys are similar. Some of the more important aspects of sample
survey techniques are mentioned in this section, primarily to remind the
reader of significant features to be noted in the following chapters.
For detailed treatment of survey problems and methods, the reader is
referred to the works cited at the end of the book.
Whatever the survey objective, certain problems of methodology
must be tackled and solved if the study is to succeed. The most important of these are the coverage, or geographical and social scope of the
survey; the sample design; the method of collecting the data; the period
of reference ; the problem of non-response ; and response errors. Mention
must also be made of the importance of pilot surveys to test the survey
plans and assumptions.
Coverage
The coverage or geographical and social scope of a family living
study is determined by the objectives of the inquiry as well as by the
resources of the agency conducting the survey.
Until recent years, most of the household surveys were limited to a
community or other relatively restricted and sometimes poorly defined
area. The families interviewed were frequently selected without regard to
scientific sampling principles, so that it was impossible to define the
survey coverage with any degree of precision.
With increasing needs for data, demands for accuracy and completeness, and the development of scientific sampling methods, the scope
and coverage of family living surveys has expanded. Especially notable
is the effort to extend the surveys to all sectors of the population. Thus,
for example, the population coverage of family budget surveys has been
expanded in such countries as Austria, France, the Scandinavian countries, the United States and the United Kingdom, where recent studies
have included all sectors of the urban or rural population instead of
covering exclusively wage and salary earners' households. Comprehensive studies have also been conducted in Greece, Italy and Spain.

INTRODUCTION

The size of the operations in the larger surveys has at the same time
increased. The number of households included in the sample in recent
family budget surveys in Greece, Sweden and the United Kingdom
amounts to several thousand, as compared with the few hundred included
in most of the earlier studies.
Sample Design
One of the distinguishing characteristics of modern family living
studies is the application of scientific sampling principles in the selection
of families to be interviewed.
The sampling frame differs according to the degree of statistical
development of the area where the survey is to be undertaken. Population censuses or permanent population records provide an excellent
frame. Such records may, however, be incomplete or out of date,
and it may be necessary, therefore, to draw the sample from other
sources.
Samples drawn from lists of households or of heads of households generally involve a certain amount of bias because of sample
attrition caused by movements of families between the date of the frame
and the date fixed for the interview. In order to avoid, or at least minimise,
such problems, one solution is to select a sample of houses instead of a
sample of households. Another method is to use maps instead of registers
as a sampling frame; in such area sampling, the area to be covered by a
survey (e.g., a country, a county or a city) is divided into a number of
smaller areas, of which a sample is selected at random; within these
areas, either a complete enumeration is carried out or a subsample of
households is taken. This method has been used extensively in the
United States, where there are no convenient general population
lists.
In underdeveloped countries, lack of suitable sampling frames and
background data on which to base sample designs, together with a shortage of experienced staff, often make it difficult to apply random sampling
principles. Nevertheless, the use of quota or purposive sampling
for statistical surveys is generally condemned on the ground that
the results obtained cannot be checked by standard statistical reliability tests, which are only applicable to randomly selected samples.
The determination of sample size is usually governed, on the one hand,
by the statistical requirements for accuracy and detail, and on the other
by the resources available. Lack of trained staff, particularly in the
underdeveloped countries, often limits the survey to a smaller scope
than would otherwise be desirable. Determination of the sample size for

FAMILY LIVING STUDIES

a survey covering several different subjects also presents special problems,
inasmuch as the optimum sample size and distribution for one item of
data may differ from what is appropriate for the others.
Collection of Data
The data are collected from the households through the use of questionnaires or by interviewers.
Both methods have advantages and disadvantages. It is necessary to
have first-hand knowledge of local conditions, habits and attitudes to be
able to choose the appropriate method. It is generally cheaper and
quicker to collect the information by questionnaires than by interviews,
provided that the rate of response is adequate and the questionnaires
are returned within a reasonable time. In family budget surveys the
" questionnaire " is usually the account book. Where ability and willingness to maintain household accounts are not in doubt, the account book
method has been found to be effective, provided proper safeguards are
used to ensure the accuracy of the information.
But households willing to maintain accounts may form a select
group, and maintenance of accounts may in itself influence household
expenditures. Also, household units differ from the establishments and
the institutions that provide information for many types of social and
economic statistics in that they normally do not keep records of their
activities. Many respondents in household surveys have only a vague
notion of statistical procedures and may suspect or misunderstand the
objectives of the study. Finally, in a population where illiteracy is
widespread, the questionnaire method cannot be used.
Personal interviewing is in fact the most common method used in
family living studies. In this approach great importance attaches to the
employment of well-trained interviewers, who can describe the objectives
of the survey in simple and convincing terms and can retain the confidence of the respondents. In the selection of interviewers close attention
must be given not only to their knowledge or previous experience but
also to their personal characteristics and attitudes. Aptitude tests are
frequently used to screen out applicants who may be emotionally or
otherwise unsuited for the purpose. The combination of qualified interviewers, carefully prepared manuals and guides, and a thorough training
programme offers the best guarantee of success in obtaining complete
and accurate responses. In fact, the most successful use of questionnaires
or account books has been observed when the method is combined with
one or more visits by an interviewer to assist the respondent in answering
the questions.

INTRODUCTION

Period of Reference
The choice of the period of reference for which household events are
to be reported is of special importance. In the reporting of household
expenditures, for example, different reference periods are used with
advantage for different goods and services, depending upon the size and
frequency of individual purchases. While it may be essential for most
types of expenditure to obtain information relating to a full year in order
to avoid the impact of seasonal fluctuations, most families have difficulty
in itemising their expenses over so long a period. This problem has been
attacked by various methods, including the use of subsamples, that is, a
different group of families provides information for each week or month,
or of repeated interviews to obtain data for each season, which are then
used to estimate annual totals.
Non-response
Careful attention to questionnaire design, selection and training
of interviewers and appropriate notice to the public regarding the survey
and its objectives are essential for accurate results. Despite the most
careful preparation, however, it is rarely possible to obtain a completed
questionnaire from every unit in the sample.
Non-response includes both outright refusals to supply data and the
lack of information because sample units either could not be located or
for some reason could not provide the information requested. The
effects of non-response may vary according to the types of data sought.
In any event, however, if the percentage of non-response is large, considerable effort is warranted in examining its possible effects. Comparison of the missing units with the original sample in terms of the general
characteristics that were used in establishing the sampling frames may be
useful. If it is found that the non-respondents are widely distributed in
the sample with respect to the significant characteristics it may be safe
to assume that the bias is relatively small. On the other hand, a concentration of refusals in a particular cell or region is a warning that a serious
problem may exist. Likewise, if a high percentage of the sample units
in certain cells could not be located, the validity of the original sampling
factors may be questioned.
The problem of non-response appears to be of less importance in
surveys in underdeveloped regions of the world than in industrialised
countries. In the latter, non-response rates of the order of 10 per cent.
or more are common, while in the former they are often as low as 1 or 2
per cent. This problem should not, however, be too readily dismissed.

FAMILY LIVING STUDIES

The practice of replacing non-respondents with substitutes is generally
undesirable because it reduces the statistical validity of the sample and
thereby casts doubt on the standard statistical tests of the survey results.
There might be a certain amount of hidden non-response, for instance,
when the household member from whom the information was supposed
to be collected was out at the time of interview and the questions were
addressed to others. The results may also be affected by failure to obtain
answers to individual questions although an otherwise complete questionnaire is returned. Families may sometimes co-operate in the survey
reluctantly, and refuse to supply sufficient details to ensure accuracy.
As a result there may be little non-response, but significant response
errors.
Response Errors
In many instances a source of bias more important than that of nonresponse arises from errors in reporting the data.
Response errors arise from a variety of causes including the faulty
memory of the respondent, the conditioning influence of recording
expenses, the conflicting reports that may be secured from different
members of a household, etc.
Various methods are used to correct such biases, depending upon the
nature of the study, the availability of appropriate external data and the
degree of accuracy desired. In some studies, especially those involving
family budget data, internal checks for consistency may be applied. The
technique of interpenetrating samples, as applied in the National Sample
Survey of India, is useful for detecting and estimating errors in interviewing, coding and processing. The method of quality checks or postenumeration surveys developed in the United States and Canada are
also useful. Finally, especially at the processing stages, quality control
techniques can help to improve the results of large-scale surveys.

Pilot Surveys
It is difficult to plan a survey without some prior information on the
subject matter, the population it is to cover, the way people will react to
questions and the length of the interview needed. Pilot surveys, or
small-scale reproductions of the main survey, are frequently used to
provide this information, as well as to test the questionnaire and the
instructions, and as a testing period for the interviewers.

INTRODUCTION

INTERNATIONAL WORK IN THE FIELD OF FAMILY LIVING STUDIES

The International Labour Office has long had an active interest in
family living studies. The subject was approached indirectly by the
Second International Conference of Labour Statisticians in 1925 in
connection with the methods of consumer price indices. The Conference
adopted a resolution pointing out that the most satisfactory method of
determining the weights of the different items and groups for such
indices was through family living studies.
In 1926 the I.L.O. prepared a report on methods of conducting
family budget inquiries for the Third International Conference of Labour
Statisticians. A resolution adopted by the Conference contained recommendations concerning the periodicity of family living studies, the
selection of families, the period during which records should be maintained, the details to be recorded, and the tabulation and presentation
of the results.
Following up the work of the Third Conference, the Office conducted
a programme of analysis and publication of the results of family budget
inquiries soon after their appearance; the analyses were summarised on
various occasions in the International Labour Review and brought
together in comparative tables in the Year Book of Labour Statistics.
Attention was also paid to aspects of levels of living other than
those illustrated by family budget patterns, as evidenced by the report,
The Workers' Standard of Living, issued by the I.L.O. in 1938, and the
comprehensive reports on national projects in the field of food consumption levels and nutrient requirements prepared by different
agencies of the League of Nations in co-operation with the I.L.O.
The materials gathered at the outbreak of the Second World War
relating to family budget surveys were summarised in a series of articles
in the International Labour Review on income, expenditure and consumption habits, and in a comprehensive report, Methods of Family
Living Studies, issued by the I.L.O. in 1940.
Following the Second World War, in 1949, the subject of family
living studies was taken up anew by the Seventh International Conference
of Labour Statisticians, for which the Office prepared a report containing
a study of the scope and timing of family living studies, an analysis of the
application of sound sampling methods to the selection of families, and a
review of the problems involved in the collection of data, the classification of receipts and disbursements, and the analysis and appraisal of
results. Food and dietary analyses, special problems of family living
studies in less developed territories, and farm family living surveys were

FAMILY LIVING STUDIES

also treated. Taking into account the developments in this field since the
Third Conference was held in 1926, the Seventh Conference adopted a
resolution defining the objectives of family living studies and setting new
international standards as regards the organisation of inquiries and the
analysis and presentation of the results.1
Other international agencies also showed increasing interest in
spreading knowledge of the techniques of family living studies. Thus,
the Sub-Commission on Statistical Sampling set up by the Statistical
Commission of the Economic and Social Council emphasised the advantages of the use of adequate methods in all types of family living studies,
and, following a suggestion of the Sub-Commission, the Statistical Office
of the United Nations has since 1949 published a series of papers on the
preparation of sampling survey reports and on sample surveys of current
interest in which the procedures of investigation are described.
In 1953, the United Nations convened, jointly with the International
Labour Organisation and the United Nations Educational, Scientific
and Cultural Organisation, and in co-operation with the Food and
Agriculture Organisation and the World Health Organisation, the Committee on International Definition and Measurement of Standards and
Levels of Living, which laid great stress on the desirability of planning
and conducting family living studies designed to obtain a direct and
comprehensive measurement of actual family living conditions.
The recommendations of this committee as well as those of the
Seventh International Conference of Labour Statisticians led in 1955 to
the convening by the I.L.O. of the Working Group of Experts on Family
Living Studies. This group reviewed the purposes and objectives of
family living studies, with particular reference to living conditions,
consumption and savings, and made recommendations on the design of
household samples, the organisation of the surveys, the methods of collecting data and the special problems of conducting family living studies
in less developed areas.
Another feature of international activities in the field of family living
studies is the help given in this field to underdeveloped countries, particularly within the scope of the technical assistance programme. The I.L.O.
provides experts to these countries on request to advise and assist in the
organisation and carrying out of family living studies. An example of a
survey undertaken with the help of an I.L.O. expert is given in Chapter IV.
International assistance also takes the form of fellowships granted to
statisticians of underdeveloped countries, and of international training
'See I.L.O.: The International Standardisation of Labour Statistics, Studies and
Reports, No. 53 (Geneva, 1959).

INTRODUCTION

courses or meetings such as the Asian Seminar on Labour Statistics
held in 1958 in the Philippines.
The United Nations, the Food and Agriculture Organisation and the
World Health Organisation offer similar assistance as concerns surveys
of demographic characteristics, nutrition, health and related topics.
The present volume is thus one of a series of I.L.O. contributions to
the study and development of family hving studies. The primary purpose
of this Symposium is to draw attention to different types of family living
studies, and to their methods and problems. The case studies have been
chosen so as to illustrate the variety of application of broadly similar
techniques. The aim has been to include surveys from less as well as from
more developed countries; to represent the major regions of the world;
to concentrate on recent studies, and to give preference to surveys carried
out under official auspices, and based on large-scale and representative
samples. It is obviously impossible however, in fifteen chapters, to
include all case studies of value conducted in the world, and many outstanding surveys have had to be omitted.
Six of the chapters in the Symposium describe family budget surveys,
that is studies concerned primarily with family income and expenditure
patterns. These include two classical family budget inquiries in industrially advanced countries, namely an ad hoc survey in Sweden (Chapter I)
and a continuous survey in the United Kingdom (Chapter II); a continuous budget survey in the U.S.S.R. (Chapter III); a family budget
survey in Greece, a less industrialised country (Chapter IV) ; a study of
family budgets in agriculture in the Federal Republic of Germany
(Chapter V); and a regional survey concerned with international comparisons in the countries belonging to the European Coal and Steel
Community (Chapter VI).
A second group of chapters relates to specialised family living studies
other than family budget surveys: a food consumption survey in Great
Britain (Chapter VII), a survey of underemployment in Puerto Rico
(Chapter VIII), a population study in the Mysore State of India (Chapter
IX), a health survey in Japan (Chapter X), and repeated surveys of
consumer finances in the United States (Chapter XI).
The third series of chapters deals with family living studies of a more
comprehensive type, or multi-subject surveys. These studies include a
survey undertaken for the appraisal of community development programmes in the Philippines (Chapter XII), the Current Population
Survey in the United States (Chapter XIII), a household sample survey
of levels of living in a region of the Ivory Coast (Chapter XIV), and
finally the National Sample Survey in India (Chapter XV).

I
A Budget Survey in Sweden
By Dr. Erland von HOFSTEN *
THE 1958 COST-OF-LIVING SURVEY

Purpose
A household expenditure survey is decidedly a multi-purpose survey.
In Sweden the principal purposes for which data from such surveys have
been utilised are the following: for revising the budgets underlying the
different index series of retail prices ; for estimates of " total private consumption " of different items in the national accounts ; for comparisons
of the current economic situation of different population groups, especially in connection with legislation, e.g., regarding child allowances, direct
or indirect taxation, agricultural prices, etc. ; for econometric analysis and
prognosis, such as the computation of income elasticities; and for historical comparisons. Such purposes as market research and other uses on
the commercial side may also be mentioned.
Heretofore, revision of the weighting diagram of the price index was
considered the primary objective of household budget surveys in Sweden.
The cost-of-living index, as it was formerly called, referred to an urban
family of a particular composition, and the only way to obtain a weighting
diagram was to conduct a special survey of households of that type.
That index has been discontinued, however, and there is now only one
index of retail prices in Sweden—the consumer price index—which is
weighted on the basis of total private consumption. Its weighting diagram
is, in principle, revised each year, and this can be done from the estimates
used for the national accounts. Data from a budget survey are still of
importance in this connection, nevertheless, because for many items good
national account estimates can only be derived from budget surveys.
* Division Chief, National Social Welfare Board, in association with Mr. Arne
NÄVERFELT, leader of the 1958 Cost-of-Living Survey for Sweden.

FAMILY LIVING STUDIES

However, the weighting diagram for the index can no longer be considered the main purpose for undertaking a budget survey.
Therefore, the other purposes mentioned above were given chief
emphasis in planning the 1958 Cost-of-Living Survey for Sweden.
An objective of particular importance in connection with the planning
of the 1958 survey was the demand for estimates that could be used for
the interpretation and formulation of Swedish agricultural policy. This
policy is concerned with prices of agricultural products and has formulated certain aims with regard to the level of living and the level of consumption of small-scale farmers in relation to industrial workers. As
statistics of incomes were not suitable for the comparison of such population groups, it was concluded that data from a budget survey could be
utilised effectively in administering the policy.
Organisation
Primary responsibility for the survey was assumed by the Social
Welfare Board, a government agency that had undertaken a number of
earlier surveys of a similar character. The preliminary plans were accepted and necessary funds made available by the Swedish Riksdag in
1958. The field work for the survey was organised in co-operation with
the Survey Research Centre of the Swedish Central Bureau of Statistics,
which provided the field staff. The planning of the survey, as well as the
processing of the data and the analysis of the results, was undertaken by
the Social Welfare Board.
For the organisation of the survey, the permanent consumer expenditure unit of one of the statistical sections of the Social Welfare
Board was enlarged. During the main part of the survey, this unit comprised the following staff: three statisticians with professional training;
Iwo supervisors ; and seven computing and checking clerks. During the
period of data collection, the staff of the Survey Research Centre was
•enlarged by one professional statistician and one supervisor, who were
"wholly occupied with the central supervision of the field work.
Coverage
As one of the main objectives of the survey was to provide countrywide estimates of total private consumption, all types of private households were included, i.e., both urban and rural, both lower and higher
income groups, all kinds of family types, etc. There were, however,
different sampling fractions for different groups (see below). Hospitals,
military installations and other non-household institutions were excluded
from the survey.

BUDGET SURVEY IN SWEDEN

Sampling Frame
Sweden has a double system of household registration. According to
one system, every parish keeps a complete register of its population.
This register is kept up-to-date by the recording of current information
about births, deaths, migration, etc. The other system involves the taking
of a simplified census in October of each year. For each dwelling unit a
special form is obtained; separate forms are used for lodgers. The form
gives for each member of the household the name, date of birth, sex, civil
status, occupation, name and address of employer, whether moved during
the past year, etc. These two systems supplement each other and form the
basis for the lists used for taxation, election rolls, etc. A set of the yearly
census forms from the autumn of 1956 was used as a frame for taking the
sample for the budget survey. Because of the high sampling fractions a
supplementary sampling frame had to be used for agricultural households (see below).
There are possibilities of bias because of errors in the register used as a
frame. For instance, forms may not have been handed in for every
household, there may be some duplication, and the household unit may
have been incorrectly defined (lodgers being included with the main
household, etc.).
However, in some communities comparisons were made between
the number of persons recorded on the census forms and those on the
lists for taxation and elections, which can be considered very accurate.
These comparisons indicated that the number of persons not entered on
the census forms was generally less than 1 per cent, of the population in
the examined area and was never more than 1.8 per cent.
Changes that had occurred between 1956 and the time of the survey
were on the whole not taken into account. Thus, households in newly
built-up areas were not included in the survey. If it was found that the
dwelhng of the originally selected household was occupied by another
household, this new household was included in the survey instead.
The household unit used for the survey was defined as a group of
persons living in the same dwelling and sharing the meals, either in
whole or in part. The elementary unit of the frame did not entirely
coincide with the survey unit. The deviations were, however, small,
and will not be discussed here.
The Sample
Sampling was carried out partly in two and partly in three stages.
Primary sampling units were those included in the master sample which

FAMILY LIVING STUDIES

has been used by the Survey Research Centre for a number of surveys
since the Centre started its activities in 1953. This master sample was
drawn after stratification of the Swedish towns and rural communities,
and its construction implies that samples of individuals or households
can be obtained which are structurally representative for the whole
country, as well as for the urban and rural population separately, etc.1
Three-stage sampling was used only for the big cities, where it was
necessary to effect division into districts. The last stage was the household
as defined in the yearly census.
Because of the multi-purpose character of the survey and the high
number of variables included—each item of expenditure implying one
variable—it was not possible to arrive at a decision about the required
sample size with reference to the level of confidence desired. Estimates
of the variances from the 1952 survey had, however, given a " rule of
thumb " : separate figures should preferably not be given for groups of
households comprising less than 300 units, at least if consumption
figures are required for the complete list of items. Taking into account
financial considerations and the difficulty of organising a very big survey,
the total size of the survey was fixed at about 4,500 households.
With a survey sample of this size, proportionate sampling would not
have given a sufficient number of households in several of the groups
for which there was need to obtain separate estimates, and it was consequently necessary to use different sampling fractions for different
categories of households. The fractions used were as follows :
Type of household

Employee households with children:
Annual income below 11,000 crowns
Annual income over 11,000 but below 18,000 crowns
Annual income 18,000 crowns and above
Agricultural households with:
2-5 hectares cultivated area
5-10
„
„
10-20
„
„
„
20-30
„
„
„
over 30
„
„
„
All other categories

Sampling
fraction

. . .

1:500
1:1,000
1:250
1: 1,000
1:250
1: 100
1: 150
1: 100
1: 1,000

The reasons for the high sampling fractions for agricultural households have been given above. The selection of the farm household sample
was facilitated by a special register of all farms in the country, covering
all areas of cultivated and pasture land and also of forest land owned by
private persons and certain compames and institutions. This register
1
Further details about the master sample are given in T. DALENIUS: Sampling in
Sweden (Stockholm, 1957), pp. 104 ff.

BUDGET SURVEY IN SWEDEN

contains an index card for each farming unit, with information about the
name and address of the farmer, the number of hectares with subdivisions
for cultivated land, pasture and forest land, and the name of the local
administrative unit. The register is maintained by a farmers' organisation
in co-operation with several government agencies. It is used as a membership index by the organisation referred to, and also for statistical surveys.
As the card index is kept up to date, it can also be used for studying
structural changes within agriculture.
Information Covered
Each household in the sample drawn for the 1958 survey was asked
to supply data with detailed specifications about every item of income
and expenditure for a period of one month. The quantities and values
of products from owned farm and business consumed by the households
were also to be recorded. In addition, information was collected about
the size and equipment of the dwelling and about ownership of certain
durables (such as refrigerators, vacuum cleaners, sewing machines,
television sets, cars, etc.).
Consumption can be measured in three different ways, namely
according to the " wearing off " principle, according to the " delivery "
principle, and according to the " payment " principle. The first principle,
which as regards durable goods would imply an assessment of the value
at the beginning and at the end of the period, has theoretical advantages.
For practical reasons it is not possible, however, to apply this principle.
It was therefore decided to adopt the delivery principle, for the reason
that this would be in accordance with the principles used in national
accounting.1 Compared with the payment principle, the main difference
produced by adopting the delivery principle occurs with regard to credit
purchases of durables. According to the principle chosen, the whole
value of a delivered car, a television set, etc., was registered, irrespective
of whether it was paid for in full or only in part during the survey month.
However, for practical reasons, some exceptions had to be made from
this principle with regard to expenditure on such items as house repairs,
newspaper subscriptions, telephone bills, etc.
Regarding the value of the consumption of home produce, added both
on the income and on the expenditure side, only home-grown food,
articles taken from own business and certain smaller items of a similar
type were included. The value of services performed by members for the
household, above all in the form of household work performed by the
1
Cf. O.E.E.C. : A Standardised System of National Accounts, 1958 edition (Paris,
1959).

FAMILY LIVING STUDIES

housewife, was not included. This conventional delimitation is an
important point which would merit further discussion. This type of
supply in kind is still important and in certain respects (home painting
and carpentry as a hobby, etc.) is on the increase, because of shorter
working hours.
Collection of Data
The normal procedure for collecting data in the 1958 survey was the
maintenance of account book diaries by each household member 12 years
old and over (sometimes even 8 years old and over). In case one or all
members of a household failed to keep the diary at all, the investigator
tried to secure an interview at the end of the month. This supplementary
interview was intended to obtain all information which would normally
have been included in the account books, except for the distribution of
the purchases, according to days of the month.
Six types of forms were used in the data collection: one questionnaire
for the introductory interview; four types of account books; and one
questionnaire for the supplementary interview.
The questionnaire for the introductory interview secured information
about the size and composition of the household, the occupation of
every household member, the size and equipment of the dwelling, etc.
Of the four account books, the principal ledger of the household
accounts was to be kept by the household member responsible for most
purchases, i.e., in most cases by the housewife. This book had four
sections : (1) for cash disbursements and receipts ; (2) for credit purchases ;
(3) for consumption of products within the household itself; (4) for
recording the mileage of cars and motor cycles (distinguishing between
home-to-work, business and other transportation). A second type of
account book was to be used by all the other household members except
the children. It resembled the principal ledger but did not contain the
third section. An account booklet that was simpler than the others was
to be completed by children under the age of 17. There was finally a
fourth book, designed for retailers and other business owners for the
recording of products withdrawn for private use.
All the account books were simple. Each entry was to occupy one
separate line, and all expenditure figures were to be entered in the same
column. In addition, the quantities consumed were to be given for certain
food items. In the case of clothing, shoes and other personal belongings,
the member of the household in respect of whom the expenditure was
incurred was to be identified.
The questionnaire used for the supplementary interview was rather
detailed. It consisted of 50 pages and contained information about

BUDGET SURVEY IN SWEDEN

purchases and deliveries of some 400 item groups that were to be used
in tabulating the account book data.
Classifications according to total consumption value during the
reference period were feasible, of course, for all household types. In
order to have a classification by income with reference to a period longer
than a month, information about the assessed income in 1958 was
obtained at the end of the survey year from the local assessment boards.
Information of this type, however, was only collected for certain household categories, namely households consisting of husband and wife
(with or without children but with no other members) as well as persons
living alone. For technical reasons, such information about other
household types could not be secured.

Period of the Survey
Three pilot inquiries were undertaken during 1957. Their purpose
was to test the account books and the instructions to households and
investigators, as well as to serve as a basis for estimating the cost of the
main survey,
The main survey was initiated on 10 January 1958, when one-twelfth
of the households in the sample started to keep records of their income
and expenditure for one month. A second group of households started
their record-keeping on 10 February and so on until the last group
completed their work on 9 January 1959. This arrangement, with a
random distribution of the households over 12 monthly periods, was
introduced in order to level out seasonal differences.

Field Staff
The data collection was entrusted to the Survey Research Centre of
the Swedish Central Bureau of Statistics, which has at its disposal a great
number of field workers, most of them women, working on a part-time
basis. These investigators in most cases have considerable experience of
interview work. About 160 investigators were selected for the budget
survey, and received additional training. This special training started
with a correspondence course, which comprised four circular letters,
each of about 25 mimeographed pages. After having replied to a number
of questions at the end of these letters, the selected investigators attended
training courses in groups of about 30. Five such courses were organised,
each one lasting two full days. The programme consisted of lectures as
well as practical exercises.

FAMILY LIVING STUDIES

Contacting Households
Some days before the beginning of the survey period the investigator
mailed a standard letter signed by himself to notify the households
concerned that they had been randomly selected for the survey. This
letter gave certain information about the purpose and method of the
survey and urged the household to co-operate. It mentioned that the
investigator would make a personal call within a few days.
At his first visit the investigator presented copies of the different
account books to be used by the household members and explained to
them how to use the books. On this occasion he normally also conducted
the introductory interview.
Within a week, as a rule, the investigator renewed his contact with the
household, either through a personal call or by telephone, which is
available in most Swedish households. On this occasion the investigator
checked whether the accounts were properly kept and whether anything
was unclear in the minds of the household members. It was left to the
discretion of the investigator to decide how many more contacts were
required during the survey month.
At the end of the survey month the household returned the account
books to the investigator, who then made a preliminary scrutiny of the
data, and, if required, contacted the household for necessary clarifications or supplementary information. The account books were then
forwarded, for further scrutiny, to both the Survey Research Centre
and the Social Welfare Board. As a result of the central scrutiny of
the account books, reference had to be made back to about 50 per cent.
of the households in order to obtain some further clarifications on the
entries.
The main reason for the rather high proportion of references back
was the demand for details about many items of consumption. For
example, materials had to be specified for fabrics (wool, cotton or
nylon) and household articles (whether made of glass, china or plastic),
etc.; for dental care it had to be stated whether this was provided by a
private dentist or under the public dental care scheme. Even at the risk
of a higher proportion of references back to households, it was considered
preferable to avoid too detailed instructions to the householder in order
not to discourage him.
If for reasons of illness, travelling, inability, etc., the households
were unable to furnish account book data for the whole or for part of
the survey month, or if they refused to undertake the work, data were,
as far as possible, secured by means of the supplementary interview at
the end of the period.

BUDGET SURVEY IN SWEDEN

Rewards
Provided that they delivered satisfactory accounts for the whole
survey month, the households were given a reward, varying from 30
crowns for single person households, 40 crowns for households with
two to four members and an additional payment of 5 crowns for each
member over four.1
The rewards to the households may seem high; nearly 20 per cent.
of the total cost for the survey was made up of" rewards to households ".
However, it was considered necessary to have relatively high rewards
in order to induce uninterested households to co-operate. After all,
the amount of work involved in providing detailed accounts for a month
is considerable.
Response
To reduce non-response, the investigators were instructed to do
whatever they could to encourage the householders in the selected
households to co-operate, if possible by keeping accounts, or at least by
consenting to an interview at the end of the period. As a rule, when one
investigator failed, another investigator, specially suited for this task,
was sent instead. In many cases, when a household had refused at the
first visit, it was possible in this way to obtain their co-operation by a
second visit. Substitute households were not used.
Tabulation
Even account books of the simple type described above imply a
very great number of entries, many of which recur daily (e.g., milk,
bread, newspaper, bus fare, etc.). In order to reduce the material to
manageable proportions with as little manual work as possible, code
" words " of five letters were assigned to each of the about 400 expenditure
items. The code " words " were constructed in such a way as to permit
easy recall. (A corresponding code in English would give eggs as "egggg",
eyeglasses as "optic", hired automobiles as "taxii" and so on.) Inasmuch
as the punch-card operators could learn the code words by heart, they
were able to punch directly on an alphabetical puncher from the account
books.
There was one punch card for four entries. Each card was transcribed with certain information about the household and contained
entries referring to item code, expenditure amount, quantity (for certain
1 crown=ls. 4d. sterling, or 19 U.S. cents.

FAMILY LIVING STUDIES

food items), and the household member to which the expenditure pertained (for items of clothing and other personal effects).
The punch cards were run through an electronic computor, so as to
obtain summary cards for each item. These summary cards could then
be handled in sorters and tabulators of the conventional type.
The starting point for the construction of the item code was a classification used for previous Swedish budget surveys, as well as for the
consumer price index. This classification is mainly functional, i.e.,
expenditure on, for instance, varnish may come under house repair,
under boat, under furniture, etc. Attempts •were made, as far as possible,
to satisfy demands from different users of these statistics, including, in
particular, the national income statisticians. As the list provides detailed
specifications, it is expected that it will lend itself to international comparisons.
The survey data were tabulated separately for different categories of
households. The principal classifications used were the following:
(a) type of household (single men, single women, married couples
without children, married couples with children, mixed households);
(b) number of children;
(c) occupation;
(d) type of community (big cities, other townships and rural communities).
Households with working wives will be shown separately, and there
will also be tabulations according to the age of the head of the household. Special efforts have finally been made to show separate budgets
for different categories of agricultural households (see above).
The household head was defined according to the following rules:
(a) in families with husband and wife : the husband ;
(b) in families with one parent : that parent (exception : if he or she was
above the age of 67, the oldest child was considered the household
head) ;
(c) in other households: generally the oldest household member.
As the total number of consumption items exceeded 300, it was not
feasible to show detailed budgets for the numerous categories of households. To a certain extent, cross-classified data, such as by occupation and number of children, have been provided; in such cases, however,
only the seven major groups of expenditures were given.
Individual consumption of clothing and other personal items have
been tabulated according to age and sex of the individuals. Information

BUDGET SURVEY IN SWEDEN

has been provided about the size and equipment of the dwelling, as well
as about the possession of certain household articles (such as refrigerator,
freeze box, vacuum cleaner, sewing machine, etc.).
The estimates produced are naturally subject to random variation.
Standard errors have been calculated; but for cost reasons, only for a
limited number of estimates, namely for 44 expenditure groups with
reference to over-all figures for all groups of households. For this
purpose the items of expenditure have been divided into four groups,
namely (1) food; (2) non-food items with a high frequency of purchase;
(3) non-food items with a relatively low frequency of purchase; and
(4) non-food items with extremely low frequency of purchase. The
coefficients of variation thus obtained show that for each one of the four
groups of items mentioned the size of the variance closely follows a curve
according to the size of the expenditure. These curves can therefore be
used as rough estimates of the variance regarding items for which the
variances have not been calculated.
The number of consumer units for each household according to the
so-called American calorie scale has been punched into the cards, but
classification according to this factor has not been much used, as a calorie
scale does not have much relevance with regard to most kinds of exexpenditures.
Dissemination of Results
Certain preliminary results pertaining to the first eight monthly
periods were prepared and published in February 1959. Further preliminary results, referring to the whole survey year, were published by the
National Social Welfare Board in the journal Sodala Meddelanden.
The complete survey report, a 400-page printed volume entitled
Hushallens konsumtion 1958 (The Consumption of Households in 1958)
and forming part of the series Sveriges Officiella Statistik (Official
Statistics of Sweden), was ready at the end of 1960.

Costs
The total cost for the survey is estimated at about 800,000 crowns.
This figure does not include the salaries of certain members of the permanent staff of the Welfare Board who worked for the survey or did
work on the regular data processing machines of the Welfare Board.
The above total breaks down as follows: selection of sample, 16,000
crowns; training of investigators, 80,000 crowns; field operations,
, 320,000 crowns; rewards to households, 140,000 crowns; production of

FAMILY LIVING STUDIES

punch cards and operations in electronic computers, 120,000 crowns;
salaries to additional central staff, 124,000 crowns.

BACKGROUND FOR THE CHOICE OF METHOD USED IN THE 1958 SURVEY

The Planning of the 1958 Survey
Large-scale family budget surveys had been undertaken in Sweden in
the years 1914, 1920, 1923 and 1933. Households to be included in these
surveys were mostly of a certain type, i.e., they were preferably to consist
of man, wife and two or three children, and they were chosen among wage
earners or lower-paid salaried employees. A considerable interest was
shown in obtaining a " representative " selection in so far as geographical area and type of community were concerned, but no random element was introduced in the selection of the households, which were
" recruited " through contacts with trade unions and other labour
organisations. The data were obtained through expenditure account
diaries which were kept over a whole year. Many of the households
originally recruited failed to complete their account books. Similar
methods were used during the Second World War, when food consumption was studied for a two-week period each quarter on a small panel.
After the Second World War the necessity for the use of probability
samples was recognised, and such samples were used in the surveys
undertaken in 1947, 1948, 1951 and 1952. In these surveys the main part
of the data was collected through interviews shortly after the end of the
year of reference, but the interviews were supplemented by account
books for the food items for a short period. In the 1952 survey the food
accounts referred to two weeks, and in order to level out seasonal differences they were spread over a whole year following the reference period of
the interviews.
The post-war surveys were all rather small (200-900 households) and
estimates could consequently only be obtained for broad subgroups of
households. The need was therefore felt for a considerably larger survey.
When plans were elaborated for the 1958 survey, a possible revision of
the methods came under discussion.
As mentioned above, the main aim of the 1958 survey was to give
estimates of the expenditure during a calendar year for a great number
of items (variables). For the Swedish population as a whole, such figures
may, for certain items, be estimated more accurately from over-all
statistics. This applies, in particular, to items which are not produced
within the country and to the items on which there is a government

BUDGET SURVEY IN SWEDEN

monopoly (liquors, tobacco, etc.)- For other goods and services, however,
as well as for estimates required for different categories of the population,
the information must be obtained directly from the households.
The 1952 survey had shown that non-response could be reduced to
manageable proportions also with the use of probability sampling at
least as far as data were obtained through interviews. The interviews did
not, however, provide enough detailed information and were no doubt
subject to considerable response errors. For the two-week accounts after
the interview, there was a considerable non-response (37 per cent.). A
further drawback was that the account book data on food referred to a
different year than the information obtained through the interviews.
It was thus felt that the main problem to be solved was to combine a
suitable period of reference with a procedure of data collection in such a
way that both response errors and non-response were kept low. The
arguments, which led to the design chosen, are presented below, considering, in turn, response errors with the use of interviews and account
books.
Response Errors for Different Types of Expenditure
in a Survey Conducted by Interviews
For a discussion of the size of the response errors, the following
breakdown of the expenditure items into five groups seems appropriate :
(1) occasional, big purchases of durables (10 per cent.);
(2) regular expenditure, such as house rent, insurance premiums, etc.
(15 per cent.);
(3) regular small purchases, such as milk, tobacco, daily travel (20 per
cent.) ;
(4) irregular not quite small purchases, such as books, items of clothing,
household utensils, restaurant meals (25 per cent.);
(5) irregular small purchases, such as daily food items, sweets, newspapers
(30 per cent.).
The figures in parentheses give a rough idea of the magnitude of the
different groups in an average Swedish household.
If we now consider the possibilities for a household at one interview
to give estimates of the expenditure for the whole preceding year, such
possibilities will be fairly good for the first three groups of items, which
account for about 45 per cent, of the total expenditure. Thus items in
groups (1) and (2) may be fairly accurately reported on the basis of
receipts or other notes or even from memory. For group (3) a yearly
estimate based on the daily or weekly expenditure will prove satisfactory.

FAMILY LIVING STUDIES

It should be noted, however, that seasonal variations may introduce
problems in this case. A yearly estimate based on the regular consumption pattern during a part of the year, when consumption is lower or
higher than normal, will be biased. As an example, it may be mentioned
that in Sweden even tobacco consumption is subject to a considerable
seasonal variation, being higher in the summer.
However, it is with regard to item groups (4) and (5), totalling about
55 per cent, of the total expenditure, that the most serious problems
occur. Regarding group (4), experience shows that satisfactory estimates
for a whole year can be obtained, if the interview is conducted by a
skilled investigator and the schedule is accompanied by a detailed list
of items, in order to aid the memory of the respondents. Individual
response errors, however, may be high. For certain items there may be a
heavy upward or downward bias in the averages. For most items in
group (5), finally, it is almost impossible to ascertain detailed data
covering a whole year at a single interview. Even an estimate of the total
amount spent on all food items, which represents the best that can be
achieved, may be subject to a serious bias.
The conclusion would seem to be that the best method would be to
split up the interview and to use different reference periods for different
groups of items. On the whole, this method may be expected to lead to
smaller response errors and larger variances. Let us consider the five
groups of items referred to above.
For group (1), durables, the variances will be considerable, even with
a reference period of one year. The variances will increase, and no
substantial reduction of response errors will result, if the reference period
is shortened. For items in groups (2) and (3), both comprising items
purchased regularly, a reduced reference period will presumably lead to
some reduction of response errors, whereas the variances will not increase
much because of the regularity of the expenditure. It is primarily for
items in groups (4) and (5) that a reduction of the reference period will
reduce recall lapse and improve accuracy. However, especially regarding
group (4), the variances will increase considerably.
One could therefore consider using a reference period of, for instance,
one year for groups (1) and (2), one month for group (4), and one week
or even less for groups (3) and (5). A system of different reference
periods for different groups has been used with success in some other
countries. This design, however, has the serious drawback that no
checks of expenditure against incomes (minus savings) can take place.
Such a reconciliation played an important part in the 1947-52 surveys.
The same drawback would be attached to the alternative solution,
whereby the interview was split up completely, by asking one sample of

BUDGET SURVEY IN SWEDEN

households about items in group (1), another about items in group (2),
etc. Such a method would, on the other hand, have the advantage that
samples of a different size could be taken so that the random variation
would be reduced to roughly the same size for each item group. The
interview time for each household would then also be shorter; in the 1952
survey the average interview lasted more than four hours, which is
rather long.1
The method just described would presumably be rather expensive,
and it would be a definite disadvantage not to have data on all kinds of
expenditure for each individual household.
Response Errors in a Survey Conducted with Account Books
for a Probability Sample of Households
It is commonly assumed that account books can only be used for
deliberately recruited households. Experience from the use of account
books for food in the 1952 survey seemed to confirm that to ask a
probability sample of households to keep records even for the short
period of two weeks may result in a non-response of a magnitude that
cannot be tolerated. Nevertheless, it was decided for the 1958 survey
to rely principally upon account books with a period of one month on a
probability sample of households because of two main arguments.
1. With a good staff of investigators, with much stress on the problem
of non-response in the instructions and with good publicity it should be
possible to obtain accounts for a month from a reasonably high proportion of households selected at random.
2. As a last resort, in cases where the account book procedure had
failed, supplementary interviews at the end of the period would secure
most of the information which otherwise would have been included in
the account books.
If account books are employed, response errors due to memory
lapse should on the whole be much fewer than for interviews. Furthermore, details can be obtained for food, etc., which are difficult to retain
in interviews even with a rather short reference period.
For certain items response errors may, nevertheless, be considerable
also when information is based on accounts, A tendency to give too
low figures for liquors may be present, and there is evidence of underreporting of small irregular purchases, such as weekly magazines,
sweets, etc. This is confirmed by preliminary comparisons for such
1
On the other hand, it is felt that in most cases, once the household members have
accepted an interview, they as a rule become so interested in the subject that they
accept even a long interview.

FAMILY LIVING STUDIES

items among all-country estimates based on the 1958 survey and independent estimates which show that the former estimates are generally
lower.
There is one special source of error present when account books
are used, namely the tendency among household members to change
their expenditure pattern once they start keeping account books. Thus,
even if there is no non-response and every single purchase is correctly
entered in the account book, the changed behaviour will produce a bias.
Many households may, for instance—consciously or unconsciously—
think that because they have been selected for a survey, they should
behave in a more conventional or " proper " way than they usually do.
Account-keeping may also tend to make the householders conscious of
the fact that they spend rather much on certain items, and they may
consequently try to reduce such purchases.
Not very much is known about this source of bias. It is believed
that this is the main reason why results from account-book methods
often tend to vary with the length of the period. This was, for instance,
found in the Swedish food studies carried out during the Second World
War. The period used, as a rule, was two weeks, but when on some occasions a four-week period was used instead, it was found that the
expenditure per week was considerably lower, especially for certain
items.
The tendency to change the pattern of expenditure, when starting
to keep accounts, can be expected to be greatest in the beginning, and
thus the conclusion would be that the period for account keeping should
be long. On the other hand, non-response will increase rapidly with the
length of the period, and non-response can be expected to give a serious
bias. A month seems to be about the longest period that can be used
for account bookkeeping.
The only way to prevent households from changing their expenditure
patterns seems to be to stress, when giving instructions to the households,
the reason why they should in no way change their purchasing habits
during the month they take part in the survey. One ought also to let the
households know that the data they give are confidential and that they
need not write their names on the books.
Regarding methods, it might finally be said that another method
which might be advantageous would be the combination of the two
procedures considered above by using account books during one month
for food and related items, whereas information about all other
items would be secured through an interview covering, say, three
months.

BUDGET SURVEY IN SWEDEN

Internal Checks
The possibility of discovering omissions or other response errors in
account books is for obvious reasons limited. An account book may look
very neat and complete, but the householders may nevertheless have
forgotten to note down many entries, or they may have made false
entries by mistake or because of expenses they do not want to disclose.
One way of checking the consistency of the data may be to compare
income and expenditure. In the beginning of the survey this was done
for about 140 households. It was found that for 50 per cent, of the
households, the expenditure was within + 2 per cent, of the income.
For 18 per cent, of the households the deviation was more than i 10
per cent. It is believed that in a good many cases the figure for cash on
hand at the end of the survey month was derived from the account book,
and did not—as it should—come from an actual count of the cash.
The figures just quoted, therefore, do not give very reliable information
about the accuracy of the account books. For this reason, and because
the comparison of income and expenditure proved very laborious, it
was given up after the first few months.
In cases where the households only provided information at the
supplementary interview at the end of the month, a reconciliation of
income and expenditure was made by the interviewer immediately after
completing the questionnaire. If the deviation was more than 10 per
cent, in either direction, certain additional questions were to be put, but
previously given information, in such cases, was not to be deleted by
the investigator.
External Checks
As the Swedish surveys (both in 1952 and in 1958) covered all types
of households, the data may be used for obtaining country-wide estimates,
which can be compared with corresponding estimates derived from other
statistics. The results of such a comparison for the 1958 data are not
yet available, but the corresponding comparison based on the 1952 data
may be of interest and is summarised below.
The independent estimates in Table I are those used for the national
accounts, and are derived from different sources, such as import statistics,
production statistics, etc. These estimates are of varying precision, and
the figures given in column (5) indicate within what per cent, of the
amount shown it is " believed " that the true values are found.
The rough figures in column (5) have thus nothing to do with estimates of random variation, etc., but demonstrate that some of the independent estimates are considered quite good, whereas others are admittedly

FAMILY LIVING STUDIES
TABLE I. COMPARISON OF ESTIMATES
OF CONSUMER EXPENDITURES
Expenditure in million
crowns according to
Item
1952
survey

Percentage
deviation of
Independent (2) from (3)
estimate

Percentage
range of
precision
of (3)

(2)

(3)

(4)

(5)

8,352
527
3,163
3,096
8,333
317

8,541
740
3,099
3,570
8,461
296

- 2
-29
+ 2
-13
- 2
+ 7

4
1
8
12
15

69
96
204
145
34
523
549
561
236
280

90
103
184
199
44
806
410
778
279
420

-23
- 7
+ 11
-27
-23
-35
+34
-28
-15
-33

15
30
40
20
15
9
20
1
10
1

Total . . . 23,471

24,411

(1)

Food
Wine and liquor
Housing l
Clothing and shoes
Sewing machines and vacuum
Doctors

Books, stationery
Postage
Trams, buses and railways . . .
Purchase of automobile
. . . .
Domestic help
Lottery tickets

1
Including heat and electricity.
* Private practitioners and doctors at hospitals but not stay in hospital.

poor. The best independent estimates are for liquors, tobacco and
lottery tickets, and the reason why these estimates are so good is because
all three items are more or less completely state monopolies. Also regarding such items, however, there are certain problems, such as the
overlapping from one year to the next, storage in retail outlets, etc. For
all the three items mentioned the estimates from the budget survey are
about 30 per cent, lower. Swedish experience seems to indicate that for
items of this kind there will always be a serious underestimate in expenditure surveys perhaps because people will not admit—not even to
themselves and so much less to other household members—that their
consumption is as high as it actually is.
However, no inference can be drawn for most other items of expenditure. The situation is unclear ; some independent estimates are very
imprecise, and deviations between these and the estimates derived from
the budget survey cannot therefore lead to the conclusion that there is a
bias in the survey. There is rather some support for the view that, at least
for many items, the budget survey estimates are better than the indepen-

BUDGET SURVEY IN SWEDEN

dent estimates. The larger estimate of total consumption expenditure is,
of course, based on independent statistical sources; yet, excluding
liquor, tobacco and lotteries, Table I shows that the budget survey
estimate was almost the same.
Non-response
In the 1952 survey the non-response had been 14 per cent, for the
interview relating to the preceding year. Less than half of this nonresponse was due to refusals, whereas the remainder referred to deaths,
sickness, incomplete interviews and households that could not be found.
In the 1958 survey, with a different method for data collection, the
following result was obtained :
1. Households keeping accounts (including those households for which
there were interviews for part of the household or part of the
period) : 78 per cent.
2. Households for which information could be secured through a
supplementary interview at the end of the month (for the whole
3. Non-response: 14 per cent.
The number of households in the second category was considerably
smaller than had been anticipated at the planning stage.
A breakdown of the non-response according to its character is given
in Table II.
The frequency of non-response varied considerably among different
geographical units, which indicates that it would be possible to attain a
better result if every investigator were of the same good quality as those
in the districts giving the best results.
TABLE II. DISTRIBUTION O F NON-RESPONSE

Number

Type of non-response

Incomnlete data
Household not found
Refusals due to:
Sickness, death in family
Lack of time
Principles
Old age, invalidity
Total . . .

Per cent.

36
29

6.1
4.9

72
15
412
22

12.3
2.6
70.3
3.8

586

100.0

FAMILY LIVING STUDIES

A breakdown of non-response according to type of community gives
the following picture:
Type of community

Three big cities (Stockholm, Gothenburg, Malmö)
Other urban townships
Rural communities

Per cent.

19
16
10

The following tabulation gives the non-response rate according to the
occupation of the principal household member:
Occupation

Per cent.

Farmers
Other entrepreneurs
Higher salaried employees
Lower salaried employees
Wage earners
Pensioners
Other categories

11
20
25
12
11
16
40

CONCLUSIONS

The survey that has been described above was carried out according
to methods and principles that were in many ways new and untried. Now
that the work is more or less concluded, it might be appropriate to put
the question whether those who have worked with the survey feel satisfied
or whether they, in starting a new survey, would revise the methods and
if so, in what ways.
In its main outlines the approach chosen and the design arrived at
seem to have worked well, and for the purposes given it would not seem
possible to arrive at an entirely different solution that would work better.
However, this does not exclude improvements in specific respects.
The choice of the " delivery principle " has resulted in high
variances for durables and has also, in other respects, proved troublesome by making the account books complicated. In spite of the theoretical arguments in favour of the " delivery principle ", it might consequently have been better to choose the " payment principle ", which
also has the advantage of corresponding well with ordinary household
budgeting habits.
To give participating households a reasonable compensation for their
work was very important in order to keep down non-response. Some
further reduction of the non-response might be possible, provided that
the selection and training of the interviewers can be further improved.
But it is presumably not possible to cut it down much further.

BUDGET SURVEY IN SWEDEN

Account books seem to be the only way of securing accurate enough
information for most items of expenditure. The length of the period
chosen, one month, seems to have been appropriate. It might be advantageous to supplement accounts for all items for the month with
interviews relating to consumer durables for a longer period backwards.
This should have been done, according to the original plan, but no funds
were available for this purpose.
Information about the ownership by the households of durables was
obtained for a few items, such as cars, refrigerators, vacuum cleaners and
sewing machines. The list could easily have been extended, and it seems
that information of that kind would have been of great use.
What shall be included as " consumption in kind " can always be
discussed. There are some arguments in favour of adopting a broader
definition of this concept than was actually applied.
With a more complete use of electronic data processing it would seem
possible to attain results at a more rapid rate and to calculate standard
errors for all estimates without much extra cost.

II
The Continuous Budget Survey
in the United Kingdom
By R. F. FOWLER and L. Moss *
Following a large-scale Household Expenditure Inquiry in the
United Kingdom, carried out by the Ministry of Labour and National
Service and assisted by the Social Survey Division of the Central Office
of Information in 1953-541, a continuing small-scale inquiry known as
the Family Expenditure Survey was started in January 1957.
This small-scale inquiry is continuous in the sense that the field
work never stops, although each household and each area in the sample
are covered only for limited periods. Each co-operating household
provides a detailed written record of all expenditure within a 14-day
period, and some other information relating to longer periods. The
households in the sample are visited by interviewers at a controlled rate,
but there is some variation in the number of households maintaining
records of expenditure at any particular time, depending on the extent
to which the co-operation of the households approached is secured.
In a year, a total of nearly 3,000 households co-operate (out of a sample
of about 5,000 visited) so that the number of households covered each
year is about one-quarter of the number in the 1953-54 Household
Expenditure Inquiry.
* Director of Statistics, Ministry of Labour, and Director, Social Survey Division,
Central Office of Information, respectively. The authors wish to acknowledge the
help given them in the preparation of this article by their colleagues, R. M. HOFSBAUM
and R. TURNER of the Ministry of Labour, and W. F. F. KEMSLEY of the Social Survey

Division, Central Office of Information. They are also indebted to J. J. NICHOLSON
of the Central Statistical Office for his help.
1
See Ministry of Labour and National Service: Report of an Enquiry into Household Expenditure in 1953-54 (London, H.M. Stationery Office, 1957). The Social
Survey is the official government field research organisation which normally undertakes most sample survey inquiries for government departments in the United Kingdom.

BUDGET SURVEY IN THE UNITED KINGDOM

The Origin of the Survey
In 1951 the Cost-of-Living Advisory Committee suggested that, in
addition to a large-scale household budget inquiry, "there should be
smaller-scale inquiries at frequent intervals thereafter". In recommending smaller-scale inquiries, the committee said—
Apart from other advantages, such inquiries would show whether the
weights used for the index were becoming unrepresentative of current expenditure patterns and would make it possible to determine at what time a further
new inquiry should be made for the purpose of again revising the weights of
the index to keep them up to date.1
The large-scale household budget inquiry referred to above was
carried out in 1953-54. The information required as a basis for weighting
a new index of retail prices became available for consideration by the
committee in 1955, and the present Index of Retail Prices, based on the
recommendations of the committee, was introduced from the beginning
of 1956.2 In that year plans were made for starting a continuing smallscale inquiry, and the Family Expenditure Survey was put into operation
from the beginning of 1957.
The carrying out of the Family Expenditure Survey is the direct
responsibility of the Ministry of Labour, but because of the multipurpose nature of the survey a co-ordinating committee, on which all
the government departments interested in the material obtained from
the survey are represented under the chairmanship of the Director of
the Central Statistical Office, guides the detailed planning and organisation of the work. This committee is assisted by a small working group
consisting of representatives of the Central Statistical Office, the Ministry
of Labour and the Social Survey. The field work of the survey is carried
out by the Social Survey on behalf of the Ministry of Labour. The
Social Survey also selects the sample of households, undertakes the
coding of the information obtained and makes some preliminary tabulations. The Ministry of Labour is responsible for the punching of machine
cards and for a good deal of the sorting and tabulation work, while
the Central Statistical Office undertakes the work connected with the
1
See Ministry of Labour and National Service : Interim Report of the Cost of
Living Advisory Committee, Cmd. 8328 (London, H.M. Stationery Office, 1951),
paras 10 and 11. This committee, which is appointed by the Minister of Labour
to advise him on all matters concerning the foundations and structure of the Index of
Retail Prices in the United Kingdom, includes representatives of the employers'
organisations, trade unions and other organisations with a particular interest in the
Index, together with statisticians and economists from the universities and from
government service.
!
Idem: Report on Proposals for a New Index of Retail Prices, Cmd. 9710 (London,
H.M. Stationery Office, 1956).

FAMILY LIVING STUDIES

more complex tabulations for the analyses which they require and which
are produced by an electronic computer.
Objectives of the Survey
A continuous small-scale survey was adopted, in preference to a
series of such surveys at frequent intervals, in order that, in addition to
providing a continuing check on the representativeness of the weights
of the Index of Retail Prices, the results might be used for other purposes
for which there was a continuing need by various government departments ; for example, in improving the estimates of consumer expenditure
in the national income and expenditure statistics.
The primary purpose of the Family Expenditure Survey, the one
envisaged in the recommendation of the Cost-of-Living Advisory Committee, is to enable the Minister of Labour to know whether the weights
used for the Index of Retail Prices are becoming unrepresentative of
current expenditure patterns, and to indicate when a revision of the
weights is necessary. To this end, information is being obtained from
the survey on a quarterly basis, so that in each quarter, data relating
to the four (or possibly eight or twelve) previous quarters can be used
to provide average expenditure patterns over a period of 12 (or 24
or 36) months. It will then be possible to compare the average pattern
of expenditure of households of the types covered by the Index with
the pattern of the comparable range of households in 1953-54, after
pricing both expenditure patterns at a common price level.
The needs of the Central Statistical Office fall under three main
heads. In the first place, the results will provide quarterly estimates of
consumers' expenditure under broad headings and annual estimates in
considerable detail. The quarterly estimates, which can be produced
within seven to eight weeks of the end of each quarter, provide one of a
number of short-term indicators of economic trends. The annual
estimates supplement the information that is used at present in compiling
the estimates of consumers' expenditure for the National Income Blue
Books. These estimates are specially useful for items for which the
estimates derivable from other sources are known to be weak. The
extent to which the Family Expenditure Survey will be used in preparing
these estimates in the future will depend on what is known about the
reliability of the results of the survey, as indicated, e.g., by the sampling
errors, as compared with the reliability of the estimates derivable from
other sources.
Secondly, the results of the survey will be analysed to show the
amounts paid in different forms of direct and indirect taxation, and the

BUDGET SURVEY IN THE UNITED KINGDOM

values of the benefits obtained from subsidies, national insurance and
the various social services by households classified by type and by broad
social classes. This analysis will show the effect on the distribution of
income of any particular form of tax or benefit and the net redistribution
effect of all forms of tax and benefit combined. It will also show the
distribution of income at different stages, e.g., before tax and excluding
all benefits, after taking account of all direct taxes (income and surtax
and national insurance contributions) and direct benefits (such as pensions, children's allowances and school meals), and after taking account
of all taxes and benefits. The results will incidentally show the extent to
which the relative position of particular households in the income scale
is changed by each main form of tax and benefit. The analyses described
in this and the next paragraph will not necessarily be undertaken every
year.
The object of the third main analysis to be undertaken by the Central
Statistical Office is to determine the relations between the expenditure
patterns of households of different types (i.e., the proportions spent on
different goods and services) and net household income. These relations,
expressed in the form of Engel curves, will be separately7 estimated for
groups of households classified by composition and social class.1
For the purpose of this analysis and that described in the preceding
paragraph, households will be grouped according to their composition.
(a)
(b)
(c)
(d)
(e)

(f)
(g)
(h)
(i)

(J)
(k)
(I)
(m)
(n)
(o)

1 old-age pensioner
2 old-age pensioners
1 adult (non-pensioner)
1 adult + 1 child
2 adults (non-pensioners) + 0 children
2 adults + 1 child
2 adults + 2 children
2 adults + 3 children
2 adults + 4 or more children
3 adults + 0 children
3 adults + 1 child
3 adults + 2 children
3 adults + 3 or more children
4 adults + 0 children
All other households

1
It is proposed to use the Sigmoid Engel Curve, which was successfully applied
to the results of the inquiry into working-class household expenditure made by the
Ministry of Labour in 1937-38. See J. ArrcmsON and J. A. C. BROWN: " A Synthesis
of Engel Curve Theory ", in Review of Economic Studies (Cambridge), Vol. XXII
(1), No. 57, 1954-55, p. 35.

FAMILY LIVING STUDIES

A " pensioner " household is here defined as one in which at least
75 per cent, of the household income is derived from national insurance
retirement pensions or similar state pensions, and/or national assistance
payments. Other households, including the one- and two-person nonpensioner households, may contain pensioners.
Households will also be grouped for these analyses in two broad
social classes : (1) those in which the head of the household is an employer,
a member of a profession, a teacher, a clerical worker, a manager, an
officer in the armed forces, an officer in the police or fire service, or a
retired unoccupied person who has an income (including pension) of
more than £10 a week and/or an " unearned " income of more than
£5 a week; and (2) those in which the head of the household is a manual
worker, a non-professional person working on his own account, a
shop assistant, a member of the armed forces, the police, the fire service or a retired unoccupied person other than those mentioned above.
The Two-Stage Sampling Design
The sampling design used is governed by the purposes of the survey,
the survey methods and the finances, staff and other resources available.
The main general requirements are that:
(a)

All private households in the United Kingdom x should have an
equal chance of being included in the sample.2

(b)

The field work should be spread over the whole year.

(c)

The sample should be adequate to yield annual and quarterly
estimates of average expenditure per household (i.e., not per head)
on the various groups of items—first priority being given to annual
estimates.

(d)

The initial sample should consist of about 5,000 addresses to be
visited.

(e)

Each household in the sample should be visited by an interviewer
and all spenders invited to keep detailed written records of expenditure for two weeks.

These requirements necessitated the adoption of a two-stage sample
design.
1

In 1957 the survey was limited to Great Britain.
For this purpose, a household means either one person living alone, or a group
of two or more persons, not necessarily related, living together in the sense of wholly,
or partly, sharing meals and other household expenses. There are about 16.9 million
private households in the United Kingdom.
2

BUDGET SURVEY IN THE UNITED KINGDOM

Several calls may have to be made in trying to establish contact with
a household; if contact is established, the co-operation of all spenders
has to be sought before record-keeping begins and then at least two
further visits by the interviewer are necessary. The interval between the
initial and final visits to a co-operating household according to these
procedures is more than two weeks. Therefore, the first-stage or primary
sampling units are reasonably small to minimise interviewers' travelling
time and the number of households in each unit is such that interviewing
continues in each area for a reasonably long period.
Each year 128 primary sampling units in Great Britain are selected;
116 in England and Wales and 12 in Scotland. In England and Wales,
the units are electoral wards, combinations of wards, parishes or combinations of parishes. There are 6,544 such units in England and Wales,
and they are first stratified into urban and rural units. The urban units
are further stratified by " Standard Region ", by conurbation or large
urban areas and other urban areas, and by the " J-index "—an index
based on the proportion of the electorate liable for jury service. The
rural units are further stratified by " Standard Region ". In this way
116 units are selected from the 6,544 with probability proportional to
the size of the parliamentary electorate.
In Scotland, where the factors available for stratification are different,
the primary sampling units are wards, combinations of wards, burghs,
combinations of small burghs, district council areas or groups or subdivisions of district council areas. There are 589 such units in Scotland,
and they are first stratified into urban units and rural units. The urban
units are further stratified into four divisions, namely—Northern, East
Central, West Central and Southern, and by conurbation and other
urban areas. Since jurors are not marked by the suffix " J " in the Scottish Register of Electors, the rateable value of an area per elector is
used as a further means of stratification. The rural units are stratified by
electoral unit only. Out of the 589 first-stage units 12 units are then
selected with probability proportional to the parliamentary electorate.1
In Northern Ireland, the interviewing is carried out by the staff of
the Ministry of Labour and National Service (Northern Ireland), and
the sample design has been modified to take account of the slightly
different circumstances as compared with Great Britain. After stratification into urban areas and rural areas, ten areas are selected from the
total of 83 first-stage units. In each of these ten areas, addresses are
1
The following materials relating to the sampling units discussed above have been
produced by the Social Survey Division of the Central Office of Information: " Some
Useful Data When Sampling the Population of England and Wales " (1956); " The
Proportion of Jurors as an Index of the Economic Status of a District" (1951);
" Some Useful Data When Sampling the Population of Scotland " (1957).

FAMILY LIVING STUDIES

selected from the rating records. One new address in each area is visited
every four weeks, making 130 new addresses each year.
Slight changes in the sampling frame were introduced from the
beginning of 1959 with the object of improving the stability of the sample
from quarter to quarter, by increasing the size and heterogeneity of the
primary units in urban areas. Thus, whereas in 1958 the total population
of 16.9 million private households in the United Kingdom was distributed
among 7,216 primary units of which 138 were selected, in 1959 the
number of primary units in the sampling frame was reduced to 3,060,
but the same number, namely 138, was selected for the sample.
The second stage of sampling is the selection of addresses from the
registers of electors maintained for parliamentary and local government
election purposes. This produces a sample of addresses, some of which
may contain more than one household. At each address selected, the
interviewer has instructions to visit all the households living at the address
up to a maximum of three. At addresses where there are four or more
households, three are chosen at random. The number of households to
be visited in each area is either 36 over a period of six weeks, or 42 over
a period of seven weeks, thus providing interviewers with an economic
load. (See the section below on field-work arrangements.) To avoid
exceeding the quota of households visited in an area, special instructions
are given to interviewers for the deletion, at random, of other addresses
to compensate for the incidence of multi-household addresses.
The selected first-stage sampling units are not all covered at the same
time. They are allocated systematically to the four quarters of the year
in groups of 16; two groups to each quarter. As quarterly analyses of
the results are made, these groups are balanced so far as possible in
respect of the stratification factors. Each group of 16 units consists of
two subgroups ; one of eight units with quotas of 36 households each
and one of eight units with quotas of 42. Interviewers make initial calls
at six fresh addresses per week. Field work begins in one group of 16
first-stage units at the beginning of the quarter and begins in the subgroups of units in the second group six or seven weeks later, according
to the quota size of the subgroup.
Range of Information
In this multi-purpose survey a record is needed of certain limited
aspects of a wide range of activities and forms of behaviour, namely
complete coverage of the financial aspects of most social activities. The
breadth and coverage of the inquiries make it inevitable that only
limited attention can be given to problems of detail but, since some

BUDGET SURVEY IN THE UNITED KINGDOM

information must be detailed to be of use, other information collected
can only be recorded in general terms.
Although the aspect of social behaviour with which the survey is
concerned is limited—attention being given only to its monetary expression—activities in very different fields are involved. The way people
think about these different fields of behaviour is varied. Some are matters
of much greater concern than others. Thus purchases of basic necessities,
such as bread, are not viewed in the same way as the purchases of
luxury items, such as jewellery, or items such as alcohol, which to some
extent, may have a social stigma. Attitudes to the same commodity may
vary from one social group to another, the purchase of a motor car
arousing much more worry and family excitement in the groups which
buy one rarely than in the groups in which it is more customary. Such
matters are relevant in the design of expenditure inquiries because the
way people feel about particular forms of behaviour may affect their
willingness or ability to give accurate information about expenditure in
such fields. If there is a social stigma, expenditure is liable to be understated, whereas if a form of expenditure has a prestige value it may be
overstated. Clearly major expenditures are likely to be more easily
remembered than minor items; they also occur less frequently. Informants may overlook the recording of purchases that have become
habitual, for example of bread, unless special steps are taken to remind
them of such items.
In view of this wide range of attitudes, if records were wanted only
in relation to particular fields or groups of items by themselves a wide
and varied range of methods of inquiry might be thought appropriate.
However, for the purposes of this survey all the data collected from any
household have to be brought together in one set of standard documents.
The range of techniques employed and the amount of attention that can
be given to any single field of expenditure are limited. A compromise
has to be effected between the practical possibilities of a limited set of
recording documents and the special detailed problems of different
fields of expenditure. Furthermore since all household expenditure
recorded in the sample covered has to be aggregated and averaged,
every co-operating household has to use a uniform set of recording
documents.
Questionnaires
Although household budget studies have a fairly long history, the
use of systematic survey techniques in this field is comparatively recent,
and there are as yet relatively Utile data available on the most effective
methods used or on the many theoretical and other problems which

FAMILY LIVING STUDIES

arise. The effectiveness of these surveys is completely dependent first
on the co-operation of the households and secondly on the quality of
the information they give. The design of the recording documents is a
matter of considerable importance, as a well-designed set of forms can
be a constant and useful guide to the informants. Few people are
accustomed to keeping detailed itemised records of their expenditure
or to distinguishing payments relating to one period from those relating
to another. Nevertheless, information obtained in the survey might be
misleading or its value reduced if there were not a considerable degree
of precision in the records of expenditure as indicated in the wording
of the survey documents. The content of the various questionnaire forms
used in this survey is therefore described below, so that their designs
may be seen in relation to each other and to the purposes of the survey.
Account has been taken of experience gained in various experimental
studies * and in the Household Expenditure Inquiry, 1953-54. The
forms generally follow the pattern of those used in that inquiry 2, but
there are various differences, mainly reflecting requirements for additional information, e.g., more detailed information about incomes.
Some modifications and improvements have been made later in the light
of experience gained since 1957.3
The forms, in general, distinguish between expenditure incurred by
the households as a single unit (e.g., on housing, electricity)—form A,
and expenditure incurred by individual household members—forms B
and C. In the former group information is obtained from the member
of the household most able to give an accurate account on behalf of the
whole household and questions are worded accordingly. In the latter
group, information relates to payments (e.g., for tobacco) made personally by the spenders. Income is regarded as an attribute of an individual
and hence information about it is also collected from the individuals;
the total income of the household is obtained by subsequently adding
the individual incomes. Furthermore a distinction is made between
questions relating to regular but infrequent expenditures, e.g., payments
for railway season tickets, and questions relating to those expenditures
which because of their frequency or the detailed approach thought
necessary are best recorded by the informant as soon after the event
as possible. The latter are recorded daily by informants whilst the
1
See W. F. F. KEMSLEY and J. L. NICHOLSON: " Some Experiments in Methods
of Conducting Family Expenditure Surveys ", in Journal of the Royal Statistical
Society (London), Series A, Vol. 123, Part 3, 1960, pp. 307 ff.
2
See Appendix " B " of Report of an Enquiry into Household Expenditure in
1953-54.
3
See W. F. F. KEMSLEY: "Designing a Budget Survey", in Applied Statistics
("London, Royal Statistical Society), Vol. VIII, No. 2, June 1959, pp. 114 fF.

BUDGET SURVEY IN THE UNITED KINGDOM

former have relatively long reference periods and are recorded in
interviews.
Form A is completed by the interviewer, who questions the head of
the household or the housewife or both. It covers information about
the age and sex of members of the household, and their relationship to
the head of the household, and the occupation, industry and working
status (employee, employer or on own account) of each member who is
gainfully occupied, together with a note of those who are retired, and
the types of school or college attended by those members who are undergoing full-time education. This information assists the interviewer to
ensure that all " spenders " x in a household are asked to give details of
their expenditure, and also provides the basis for analysis of the results
according to household size and composition. Information is also obtained about any special circumstances affecting the household during
the period of the survey, e.g., the absence of any members, visitors,
births or deaths. The form covers information about housing, the type
of dwelling (whether rented from a local authority or a private landlord,
furnished or unfurnished, occupied rent free or owner-occupied) particulars of any subletting and the amount of rent, rates (local taxes), water
charges, mortgage payments, house insurance premiums, etc. Other sections include questions on the possession of certain durable goods and
gas, electricity, telephone and hire purchase payments. Finally sections
are provided for noting payments which will be charged as expenses to a
business or organisation, for information about food and fuel supplied
free or at concessionary prices by employers, about school meals, about
milk supplied free at schools and about milk and welfare foods for children supplied free or at reduced prices under various state schemes.
Form B is also completed by the interviewer who questions each
individual spender in the household; in certain circumstances it may be
completed by the individual concerned. The first part of this form covers
certain kinds of regular payments such as for television, radio and driving
licences, motor tax and insurance, other types of insurance (except house
insurance, which is included on form A), railway and similar season
tickets and also expenditure and grants for education and training
courses. The second part of the form covers income, providing separate
headings for various components such as wages and salaries, business
profits, investment income, family allowances, national insurance payments, pensions, annuities, etc., and also for income tax and other
deductions from income such as national insurance contributions and
subscriptions to pension funds.
1

All those over the age of 16, and those under 16 who are working.

FAMILY LIVING STUDIES

Form C is used by the individual concerned for recording all items
of expenditure on seven consecutive days. There are separate pages for
each day and broad headings of expenditure are given. Records are to
be provided of all expenditure within a 14-day period and two of the
forms (one for each seven days) are used for this purpose by each spender
in the household during the survey. Recording is generally on a payments
basis (with certain specified exceptions) irrespective of when the goods
are received or the services rendered (i.e., only payments made during
the recording period must be included). Detailed instructions on how to
maintain the records are given to spenders, with special notes on such
matters as instalment payments and purchases of second-hand goods.
There is also a section for the noting of payments charged as expenses
to a business or organisation, and for the recording of betting payments
and winnings. Finally, spenders are asked to look through a fairly
detailed check list to remind them of any purchases they may have forgotten to record. Spenders are not asked to exclude from form C any
items which they have already reported to the interviewer for recording
on forms A and B, as this might cause confusion. Instead, any such
duplication is allowed for when the material is being coded. Payments of
rent and insurance are examples. The expenditure on rent is taken from
replies to questions on form A and expenditure on life assurance from
replies to questions on form B. Any entries on form C in respect of rent
or life assurance are then deleted, whether or not the payments are
identical with those recorded on forms A or B.
Particular importance is attached to interviewers being provided with
clear definitions of such terms as members of a household, retired
persons, spenders, used in the documents and with all the information
necessary to enable them to explain to informants exactly what is
required.
The Interviewers
The interviewer is the indispensable intermediary between the
statistician who will use the survey data and the household supplying
information. The more detailed the design of the inquiries and the more
specific the requirements, the more important are the interviewers who
explain to informants how to compile the records. They need general
training in interviewing methods, some understanding of sampling
principles and careful briefing in the details of the survey being undertaken. The quality of interviewers can affect the extent to which households co-operate and the accuracy of the information collected.
In the Family Expenditure Survey the field work is done by regular
interviewing staff of the Social Survey, chosen from amongst those

BUDGET SURVEY IN THE UNITED KINGDOM

trained for the whole range of Social Survey work. Their general training
is directly relevant to the requirements of this survey since the sampling
methods, types of questions asked and mode of classifying data on this
survey are comparable with other Social Survey work. The Social
Survey's basic training manual for interviewers, " The Handbook for
Interviewers " (1956), covers much of the ground and many of the
techniques of questioning and definitions used in this survey. In addition,
the interviewers attend briefing sessions on the specific procedures to be
followed and their attention is drawn to points of special importance.
Field-work Arrangements
From the sampling lists of addresses in the selected interviewing areas,
quotas are issued to field workers throughout the year. The quota in
each area consists of 36 or 42 addresses. These are numbered serially.
With certain exceptions all these addresses are visited. The initial or
" placing " visits to seek co-operation and place the forms are made at
a regular rate (about six fresh households a week) throughout a period
of six or seven weeks. Field work then continues in the area for another
two or possibly three weeks until all the co-operating households have
completed their records.
As this programme cannot be followed exactly in practice, there has
to be some flexibility in the field-work arrangements. For example, if a
whole household is away when calls are made during the week, such a
household is not counted as one of the number to be visited in the week
and attempts are made to establish contact in subsequent weeks; at
least four attempts are made in all. Similarly, if one or more members
of a household are away on holiday, although it may be possible to
complete form A the other recording is delayed until all the spenders can
begin to keep records at the same time. On the other hand, as mentioned
earlier, an address will occasionally yield more than one household so
that sometimes the quota of six households visited may be reached by
calls at less than six addresses.
At every selected address, calls are first made to find out who is
living at the address, to explain the purpose of the survey and to seek
the co-operation of the households. The co-operation of all spenders
in the household is necessary. When such co-operation is secured, a
copy of form C is supplied for each spender to cover a first period of
seven days. Calls are made subsequently to ensure that the records are
being kept satisfactorily, to supply forms for the second period and
finally to collect the completed records. The interviewer has to determine
who are the spenders on the basis of information about household

FAMILY LIVING STUDIES

composition obtained from any responsible member of the household.
When co-operation is secured, the interviewer seeks the answers to
the detailed questions on form A from a responsible member of the
household, usually the head of the household or housewife, and on form B
from the individual spenders. The interviewer leaves one form C for
each spender, explaining how it is to be filled in and asks all spenders
to start together as soon as possible. Record keeping may begin on any
day, but preferably not on Saturday or Sunday, when the risk of importing
payments made previously is believed to be highest.
At least one intermediate call is made during the fourteen days to
see how the records are progressing, usually after about five days, to
check over the records and to leave form C for the second week. If only
one intermediate call is made, the collection of records for the first week
is delayed until the final call. Additional intermediate calls, however,
may be made at the interviewer's discretion. Sometimes it is desirable
to make a call on about the eighth day to collect the first week's records,
and see that the second set has been started satisfactorily. At the final
call the records are checked and the interviewer makes sure that forms A
and B are complete.
Provided all the spenders in a household co-operate fully, they are
each paid £1. Pilot studies1 have shown that such payments improve
response rates, and have suggested that this level of payment is about
the optimum level for a survey of this type.
Data Processing Arrangements
When documents are received from the field worker they are
scrutinised for major errors and omissions, classified and counted. The
purpose of this pre-checking is to enable documents to be returned, when
necessary, to interviewers for correction or verification, whilst they are
still in the same area and the event or purchase is still fresh in the
informant's memory. In pre-checking, special attention is always given
to documents from interviewers new to the work to ensure that they
have a thorough grasp of their instructions. Any difficulties are brought
to the attention of supervisors in direct contact with the interviewers.
The pre-checking staff have been trained to handle all sections of the
documents. They examine all the forms for the entire household, giving
special attention to points where experience has shown that difficulties
may occur.
1

See W. F . F . KEMSLEY and J. L. NICHOLSON, op.

cit.

BUDGET SURVEY IN THE UNITED KINGDOM

In the subsequent classification, or coding stage, in which a code
number is allocated to each entry, the work is sectionalised. Different
documents or sections thereof are handled by different coders. Individual
coders can thus specialise on particular sections and work more quickly
than would be possible if they were having to cover a wider field. As
the various pieces of information obtained on forms A and B cover
different periods of time (e.g., month, quarter or year) and have to be
reduced to a common period, preliminary computing is necessary to
convert the figures to a uniform basis.
After coding, documents are checked by highly experienced staff
who once again handle all the documents completed by separate households and are thus in a position to consider the data as a whole. While
the main task of the coding staff is to classify recorded information as a
prerequisite to machine processing and analysis, a very important
secondary task is to check the efficiency with which documents are
completed. In this way a continuing audit of interviewers' efficiency is
carried out.
As soon as coding is completed, a special summary sheet is prepared
for each household analysing the expenditure in various broad item
groups. These summaries are used for compiling manually the preliminary quarterly tabulations mentioned above, so that these are
available to provide early estimates of changes in consumers' expenditure
in advance of the completion of full processing.
The coded information is then punched on machine cards. A separate
card is punched for each entry of expenditure and income; certain
" sorting codes " (used to classify the sample into different categories of
households) are also punched on each card to enable the various required
analyses of expenditure to be made. In a year, over half a million such
cards are punched. Conventional sorting and tabulating machines are
then used to prepare " summary " cards each recording the total expenditure of a household during a week on a particular item.
For a continuing survey organised on a basis of calendar years and
quarters, resources must be available which are at least sufficient *to
complete each stage of the processing routine for data relating to one
period before data relating to the following period are readyKfor that
stage of processing. It is also important that the various analyses of the
data should be available in time to meet the purposes for which they are
required. The information required for use in preparing the National
Income and Expenditure estimates, for example, is required within five
months of the end of the year to which they relate. The complete processing of the results of this multi-purpose survey is technically possible
on conventional mechanical equipment using punched cards but it is

FAMILY LIVING STUDIES

most complicated and lengthy, and electronic computing facilities are
almost essential to reduce the processing time.
The Ministry of Labour has not yet obtained an electronic computer.
At present an electronic computer at a government research establishment is being used to prepare most of the analyses required by the
Central Statistical Office and conventional punched-card machines for
the remainder, chiefly those required by the Ministry of Labour for
retail price index purposes.
Reliability of the Data
The checks on the reliability of the data obtained in the survey that
are made cannot be entirely satisfactory, since, if completely reliable
data about the pattern of consumers' expenditure were available from
independent sources, the survey would be superfluous. Some comparisons can, however, be made with the estimates of consumers'
expenditure derived from various other independent sources, for example,
some used in compiling the National Income and Expenditure Statistics.
These comparisons with independent estimates are fairly encouraging
in so far as the differences are small in most cases where the national
income estimates are thought to be most reliable, and larger in cases
where the national income estimates are thought to be less reliable. One
exception to this general finding is expenditure on alcoholic drink and
tobacco; as is usual in this type of survey, there is apparently marked
under-recording in the family expenditure survey. It is, however, premature to come to firm conclusions from comparisons of this type on
the basis of a few years' results, particularly as the full variance analyses
and standard error calculations on the survey material are not yet
available.
The expenditure recorded by households in the first week of the
record-keeping period is also being compared with that recorded in
the second week, and the inter-week variations studied to determine
whether they are greater than would be expected. For a number of
items the first week's expenditure is significantly greater on average
than that of the second week, indicating that there may be some bias in
the results. As it is not known which week provides the more reliable
data, the average of the two weeks is being used in work relating to the
Index of Retail Prices.
Those who co-operate in surveys of this kind must show considerable
patience and take the trouble to answer long and detailed questionnaires
and to record every item of expenditure incurred during the specified
period. It is not surprising that many of the households included in the

BUDGET SURVEY IN THE UNITED KINGDOM

sample are unwilling to co-operate. The proportion which co-operate
in these surveys is about two-thirds and, as it cannot be assumed that
the co-operating households are necessarily representative of the whole
sample, attempts are made in the analyses undertaken by the Central
Statistical Office to correct the results of the surveys for differences
between the characteristics of co-operating and non-co-operating households. The method applied to the first year's results was to re-weight
the number of households of different types and in different income
ranges so as to reduce the differences between the Inland Revenue
distribution of income-tax units of different types and the corresponding
distribution of income-tax units in co-operating households. The
method was only partially successful. Reconciliation would have been
impossible without using a wide range of coefficients. In the second year,
efforts were made to collect information on a few of the more important
characteristics, viz. the number of persons, type of dwelling and rent,
of households which were unwilling to co-operate in full. This additional
information will be used in re-weighting the data collected from households which co-operate in full.
Special Features of the Survey
The continuity of the survey produces certain features worth particular mention. First, it provides the opportunity to learn continuously
from experience and to improve techniques progressively, e.g., modifications in the sampling design, interviewing technique and forms of
questionnaire. The response rate, that is the proportion of households
which co-operate, is a case in point. By changes and improvements in
the interviewing technique, this rate has been increased gradually over
the three years from a figure somewhat below 60 per cent, to a level
considerably above.
Secondly, it makes it especially important to ensure a steady flow of
work at every stage and to avoid any delays in processing. Thus field
work must be spaced carefully and a regular flow of interviewing and
recording must be arranged. At the processing stages the fact that records
are received continuously means that a steady rate of coding, checking
and computing must be maintained. There is then a steady pressure at
all the stages of the work and this can be sustained only by an appropriate
organisation of these stages. Thirdly, if continuity is associated with a
multi-purpose character, it may involve the work being rather complex
at every stage. The purposes for which the survey is used also make
speed very important in carrying out the heavy programme of processing work.

FAMILY LIVING STUDIES

In the Household Expenditure Inquiry in 1953-54, publicity arrangements made in co-operation with the press and British Broadcasting
Corporation facilitated the task of interviewers and contributed to the
success of the inquiry in which about two-thirds of the households
visited co-operated. For the continuing form of survey, however,
comparable effective arrangements, particularly at the national level,
cannot be made. In consequence, the Family Expenditure Survey which
started in 1957, has had little publicity and greater reliance has had to be
placed on the skill of interviewers in securing co-operation.
The nature of the information required is not the same for all the
various purposes served by the survey. Thus for checking whether the
weights used in the Index of Retail Prices are representative of current
expenditure patterns, expenditure has to be analysed in groups comparable with those used in the Household Expenditure Inquiry, 1953-54,
on which the 91 section weights of the present Index are based. These
sections have to be defined in a way which provides small groups of
commodities, the prices of which can be expected to move roughly
together, and which can therefore be conveniently represented in the
index by corresponding families of price indicators. The " sorting codes "
used at the coding stage must enable the expenditures of those types
of households which are excluded from the coverage of the Index of
Retail Prices to be distinguished on the basis of classification similar to
those used in the 1953-54 inquiry, and a comparable income grouping
must therefore be provided.
For other purposes, however, different analyses of expenditure are
required. Thus for estimates of consumers' expenditure used in national
income evaluation work, the classification must follow the definitions
adopted in the national income estimates, which are not always the same
as those used for the Index of Retail Prices. Additional item codes are
needed for purposes of demand analysis. And, for the analysis of the
redistribution of income, it is necessary to separate, as far as possible,
items which have different rates of indirect tax or subsidy.
As a result of these various requirements, the analysis of expenditure
used in the Family Expenditure Survey is necessarily considerably more
detailed than that which was needed in the Household Expenditure
Inquiry in 1953-54. Items must be distinguished in sufficient detail to
enable any one of the several required groupings to be obtained and some
items have to be coded in more than one way to suit alternative definitions. There are also differences in the sorting codes used for the various
analyses. Thus the definition of income for National Income and
Expenditure purposes differs from that used for the Index of Retail
Prices. Extra details of income received from the various social services

BUDGET SURVEY IN THE UNITED KINGDOM

are also required for the redistribution of income analysis. Thus the
income sorting code apphed is more complicated than that used in the
Household Expenditure Inquiry, 1953-54. Additional classifications of
households are also required, e.g., in terms of the social class of households, for some purposes.
General Observations
Although the results of the first few years of the Family Expenditure
Survey are still being studied, some general observations can be made
about the organisation of this type of survey based on the experience so
far gained.
First, there are definite advantages in designing a continuing survey
to serve a number of different purposes. As a result of the multi-purpose
character of the survey, however, the coding and processing of the
material is made much more complicated than for example in the
Household Expenditure Inquiry, 1953-54.
Secondly, as the need for more up-to-date information is one of the
main reasons for instituting a continuing survey, the results are required
fairly quickly.
Thirdly, the resulting heavy data processing programme and time
limits require electronic computing facilities.
Fourthly, should it become necessary to carry out another large-scale
survey comparable with the Household Expenditure Inquiry, 1953-54,
the techniques developed and experience gained in the Family Expenditure Survey are likely to make its organisation much easier.

Ill
The Continuous Budget Survey
in the U. S. S. R.*
By S. V. POSTNIKOV

The Central Directorate for Statistics of the U.S.S.R. (C.D.S.) began
inquiries into the budgets of wage and salary earners shortly after the
October Revolution. Before the Second World War, 19,200 employees
were covered by C.D.S. inquiries; since then the number has been
increased to 24,200. The main task of the C.D.S. in this field [i.e.,
employees' budget statistics] is the collection and analysis of data on
the income, expenditure and consumption of wage earners in the different branches of industry and of salary earners throughout the U.S.S.R.
and in the different industrial areas.
Budget inquiries are carried out in the following industries: iron
and steel and non-ferrous metal industries, coal mining, iron mining,
the petroleum industry, the chemical industry, the rubber and asbestos
industry, electric power stations, the engineering industries, the metal
trades, the building material industry, the wood and paper industries,
textiles, ready-made clothing industry, and the leather, hide and shoe
industries. In addition, budgets are investigated in construction, rail
transport, state farms, and in the teaching and medical professions.
Methods of Selecting Families
In 1951 the C.D.S. selected a number of employees' families with a
view to conducting investigations into their budgets. The choice was
* The description given in this article is a translation of part of Chapter VI,
written by S. V. Postnikov, in a volume entitled Uchebnoe posobie po otdeVnyn otrasliam statistiki (Manual on Different Branches of Statistics) (edited by M. R. Edelman,
Moscow, Gosstatizdat, 1958). The chapter in question, entitled " Statistics of Budgets
of Wage and Salary Earners and of Collective Farmers ", comprises the following
sections: (1) Budget Surveys in Pre-Revolutionary Russia; (2) General Principles of
Budget Surveys in the U.S.S.R.; (3) Statistics of Budgets of Wage and Salary Earners;
(4) Statistics of Budgets of Collective Farmers; (5) Utilisation of Statistical Data on
Budgets of Wage and Salary Earners and of Collective Farmers. The extract given
here with the permission of the author is taken from the third section of the chapter.

BUDGET SURVEY IN THE U.S.S.R.

made in accordance with the general principles of stratified sampling,
and account was also taken of the need to have the families studied
concentrated, to some extent, in particular areas.
The co-operation of the industrial undertakings simplifies the selection of families for the inquiry, and the welfare organisations of the
undertakings are able to assist in giving the selected families all necessary
instructions. First, before beginning with the selection of families, the
number of budgets to be investigated under the plan are allocated among
the different branches of industry, republics, territories and regions
according to the number of workers in each.
The selection of the employees to be investigated is effected in two
stages; first a choice is made among the undertakings, after which
individual workers are chosen. Undertakings are selected mechanically
from the total number of them in a given industry in the particular
republic, territory or region concerned. For this purpose, lists of all
undertakings are drawn up in the manner shown in Table I.
The size and structure of a worker's budget are determined by his
wages. The inquiry must therefore cover workers of different wage levels.
Accordingly, lists of all undertakings are drawn up, those in which the
average wages of production workers are highest being given first and
those in which they are lowest, last. The interval to be used for selecting
the undertaking in which workers are to be chosen for the purposes of
the inquiry is determined by dividing the total number of production
workers employed by the number of inquiry points needed. Let us
TABLE I. LIST O F U N D E R T A K I N G S F O R THE SELECTION O F P R I M A R Y
SAMPLING UNITS

No.

Name and address
of undertaking

Average wages
of production
workers

Average number
of production
workers
employed

Cumulative
total number
of workers

Number
of inquiry
points

1,200
1,170
1,165
1,130
1,115
1,080
1,065
1,025
1,010
980
950
900
880

2,880
1,360
3,700
120
2,500
1,300
1,080
3,070
1,730
80
2,300
1,000
3,320

2,880
4,240
7,940
8,060
10,560
11,860
12,940
16,010
17,740
17,820
20,120
21,120
24,440

1
2
3
4
5
6
7
8
9
10
11
12
13
1

The figures given are hypothetical.

FAMILY LIVING STUDIES

assume that in the example given above four inquiry points must be
selected. By dividing the total number of workers (24,440) by four we
obtain an interval of 6,110. Thus the inquiry points fall in undertakings
Nos. 1, 3, 7 and 11 in the table. A large undertaking in which the total
number of workers is more than twice the interval thus calculated will
be allocated two inquiry points, or more according to the case. This
method of selection is used in all the branches of industry covered by the
inquiry.
Once the undertakings have been selected, their representativeness
of the entire group is ensured by comparing the average wages of production workers in them with those in all the undertakings concerned.
The next step is to classify all the workers in each undertaking
selected on the basis of their levels of skill and wages. To do this, lists
or card indexes are made in each undertaking of all production workers
classified into two categories (skilled and semi-skilled) and placed in an
order based on their wage levels according to their positions in the wage
scale in force in the works to which they belong. When the list has been
drawn up, intervals are determined by dividing the total number of
workers on the list by the number to be selected; they are then selected
from the list in the same way that undertakings were selected earlier.
When the selection process is completed the representativeness of the
selected workers from the point of view of wages is checked. Thus the
method of selecting workers described above secures proportional
representation of workers in each industry according to their levels of
skill and earnings.
Coverage of the Inquiry
The inquiry into the budgets of employees is designed to secure
information on the entire income of the families concerned according
to origin and of all expenditure according to its purpose, the personal
consumption of these employees, and the origin and amounts of foodstuffs and manufactured goods acquired by them.
The forms used for the budget survey contain approximately 1,000
headings. The inquiries are based on interviews of households; but to
obtain accurate information the findings of each interview are collated
with the information contained in the household account books, receipts
for rent and payments for public utility services, etc. In addition,
information provided by the undertakings and establishments concerned
on wages, bonuses, pensions, taxes and other payments is used.
The families buy foodstuffs every day and industrial products frequently. To obtain exact data on their budgets they keep records of
their daily income and expenditure in the form indicated in Tables II to IV.

BUDGET SURVEY IN THE U.S.S.R.

TABLE II. CASH I N C O M E
Day
and
month

Amount
Received (by whom,
from whom, on what grounds)

Remarks
Roubles

Kopeks

TABLE III. I N C O M E I N K I N D

Day
and
month

Product (specify:
milk, potatoes, eggs,
footwear, firewood,
etc.)

Origin of products
(own cow,
vegetable garden,
gift, workplace,
etc.)

Quantity and unit
of measurement

Remarks

The families also record, under a third heading, the quantity of each
foodstuff remaining unconsumed at the end of the month.
In Table II the family keeps a daily record of all money receipts.
Each item is to be entered on a separate line and the reason for the
payment (wages, pension, bursary, etc.) given. In the case of wages,
the family must indicate not only the amount actuaËy received but also
the gross wage, including deductions for income tax, loan subscriptions
and other payments. If an item of income consists of the proceeds of a
sale the family is required to specify the nature and amount of goods
involved and to state whether they were sold to a state organisation, a
co-operative or a private person.
In Table III all information is given concerning income in kind
(produce obtained from livestock kept by the family; produce from the
family vegetable garden, preserved mushrooms and berries, firewood,
gifts, etc.). Although, generally speaking, income in kind constitutes
only a small percentage of the income of employees, it may account for
a very considerable proportion in some particular families.

FAMILY LIVING STUDIES

Cash expenditure is recorded in the manner indicated in Table IV.
TABLE IV. CASH E X P E N D I T U R E

Day
and
month

Article

Place of purchase or to
whom money paid (state
store, co-operative,
canteen, buffet, kolkhoz,
market, private person)

Quantity
and unit
of measurement

Amount spent
Remarks
Roubles

Kopeks

Every day the members of the family record all their expenditures
in this form, each item being recorded on a separate line. If the money
is spent on goods, the nature of the goods purchased must be indicated
in detail (for instance, rye bread, veal meat, cotton material, etc.),
together with the amount purchased and the place of purchase. The
keeping of these records does not involve any difficulties for the family
itself. The investigator merely has to ensure that they are kept regularly
and without interruption.
To record the data supplied by the undertaking or establishment a
different set of forms is used; these are entitled " Extract from the Wage
Register ", " Feeding in Canteens and Buffets " and " Feeding in Institutions for Children ".
The extract from the wage register is filled in by the accounting
department of the undertaking or establishment concerned. It contains
details of all sums credited to the employee covered by the inquiry, namely
wage or salary, [holiday pay, bonuses paid out of the wage fund,
payments in respect of temporary incapacity to work (sickness, childbirth) and all deductions from the gross wage and from payments in
respect of such incapacity. It also gives the amounts actually drawn,
bonuses received from sources other than the wage fund, lump-sum
social insurance benefits and pension payments. All these data are used
by the investigators infillingout the budget form.
One of the most significant pointers to the levels of living of the
population is the consumption of foodstuffs. To determine the composition of the diet one must obtain information on the quantity of each
product consumed per person in each population group. This task is
compUcated considerably by the fact that employees take some of their

BUDGET SURVEY IN THE U.S.S.R.

meals at home and others in canteens. It is not enough to record all
foodstuffs consumed at home because communal feeding plays an
important part in the feeding of the population. Many employees take
their morning, midday and evening meals at a canteen, a restaurant or
buffet, while their children have their meals in pre-school institutions.
To calculate the quantity of foodstuffs consumed by employees and
members of their families in communal feeding centres the statistical
services use a special form entitled " Feeding in Canteens and Buffets ".
On this form is given a detailed list of the produce consumed at each
communal feeding centre during the month under consideration and the
amount and cost of each item. The information is obtained from the
accounts of the canteen or buffet concerned. At the same time, employees
covered by the inquiry record the sums spent in communal feeding on
their account sheets. To give a very simplified example, suppose that
the sales of a buffet during a month (500 kg. white bread = 1,000 roubles;
80 kg. pork products = 1,000 roubles; 200 kg. milk = 400 roubles, etc.)
brought in a total of 2,960 roubles, and the employees covered by the
inquiry spent a total of 740 roubles at the buffet. If one assumes that
the workers covered by the inquiry purchased the same amount and
range of products as did all the workers who took their meals there
during the same period, the figures show that they consumed onequarter of all the products consumed, that is to say, 125 kg. of wheaten
bread, 20 kg. of pork products and 50 kg. of milk.
The quantity of produce consumed by the employees' children in
institutions for children cannot be calculated by exactly the same method
for children take their meals in such institutions fairly regularly. The
quantity of produce used to feed children and the quantity consumed per
child and per day are recorded on a form entitled " Feeding in Institutions for Children ". The number of days on which each child took one
or more meals in an institution is recorded in the family account sheet.
The proportion of produce consumed by the children of families covered
by the inquiry is then calculated by establishing the ratio between the
total amount of food provided per child and per day and the amount of
food consumed by the children in the families covered by the inquiry.
For instance, one nursery school used 2,250 litres of milk and 3,000
eggs to feed 120 children during a period of 25 days. Ten children
from families covered by the inquiry attended that school and took their
meals on every one of the 25 days. Thus the total number of " feeding
days " spent by these children in that school was 250, or one-twelfth of
the total number of " feeding days " during the same period. Consequently, during the period in question the children in the families
covered by the inquiry consumed 187.5 litres of milk and 250 eggs.

FAMILY LIVING STUDIES

The third set of forms used for the budget inquiry is entitled " Budget
of Worker or Salaried Employee ". On these forms are recorded the
composition of the family, cash incomes, the purchase of manufactured
products other than foodstuffs, the purchase of logs ready for use in the
building of the outer walls of houses, the purchase of houses, cattle,
poultry, etc., expenditure on items other than goods, purchases of
foodstuffs, wines and spirits and expenditure on communal feeding,
the estimated consumption of foodstuffs, purchases of firewood, peat,
hay, straw, wool, leather and hides, etc.
In filling out the questionnaire the first item of information to be
given is the composition of the family concerned. Information on the
size and composition of the family is of particular importance for the
analysis of the budget data, as over-all figures may be worked out per
family, per working person, per person fed converted into adults, or
(most frequently of all) per member of family. It is therefore essential
to ensure that the composition of the family concerned is indicated
exactly. All persons living with the employee and on his household
budget are considered members of his family. Members of the family
who are for the time being living away from home for various reasons
(studies, military service, mission) are treated as being temporarily
absent and are kept on the family register.
Information is given as to the place of employment or other source
of income of each member of the family and the type of work performed
by him—wage or salary earner, member of an industrial artel, member
of a kolkhoz, holder of a student bursary, pensioner. The ages of the
different members of the family and the number of days spent at home
by each are also given. In addition, the number of days on which each
child took meals in institutions for children is also given.
The cash incomes of families are made up of the employees' earnings,
bonuses, student bursaries, pensions, allowances and other benefits paid
by the state. The value of receipts in kind provided for by the law is
also included in cash incomes. This item includes the value of all payments in kind out of the wage fund (supply of housing and public utility
services without charge) and also receipts in kind from sources other
than the wage fund (free distributions of milk, soap, etc.). Grants received
from undertakings, establishments, public organisations and trade unions
to enable employees to stay in health resorts, rest homes, pioneer camps,
open-air schools, etc., grants for the maintenance of children in institutions and grants for communal feeding are also treated as forming part
of cash income. As they are recorded under " income " these items are
also recorded under the appropriate headings in the " expenditure "
section. Lastly, the proceeds of the sale of goods and effects to private

BUDGET SURVEY IN THE U.S.S.R.

persons, money received for services rendered, gifts, etc., are also entered
under the heading of cash income.
Employees spend their cash income on foodstuffs, industrial products
other than foodstuffs, services and recreation, the payment of taxes and
other items. Expenditure is broken down into a fairly considerable
number of items; and expenditure under each item is classified according
to whether the payment was made to a state organisation, a co-operative
or a private person. When expenditure on goods is recorded, the nature
and amount of the goods and the price paid are specified.
When all the data on cash income and outgoings have been recorded,
a balance sheet is drawn up; the quantity on hand at the beginning of
the month, added to the total income, should add up to the same figure
as the total expenditure plus the quantity on hand at the end of the
month. If the two totals are equal the record of cash income and outgoings is considered to be satisfactory. If they are not the same, the family
is interviewed again.
One of the purposes of the budget survey is to obtain data on employees' nutrition. Formerly, only produce purchased was taken into
account, but in employees' families the purchase of foodstuffs and the
expenditure on food are not always identical. In addition to bought
produce, food consumption may include produce from store or from
employees' own production, and produce received as gifts. At the same
time part of the produce bought is used for consumption and gifts, and
part is stored. Thus, in order to be able to calculate the quantity of
each product actually used in feeding the family a table like that shown
in Table V is included in the inquiry form.
TABLE V. F O O D CONSUMPTION O F THE HOUSEHOLD
Quantity consumed
by family

Incomings during month
On
hand
on 1st
of
month

From own
production
inch
home-made

Received
without
payment
(as gifts or
for services
rendered)

Processed
goods

Goods purchased in
state stores,
co-operatives, kolkhoz markets
or from private persons

Used
otherwise
Quantity
than for
on hand at
household
end
consumption of month
(fodder,
sales, gifts,
etc.)

During the
month
under consideration

During the
previous
month

FAMILY LIVING STUDIES

To calculate the quantity of each product consumed, all entries of
foodstuffs during the month—whether home-produced or home-made
or received without payment—and all foodstuffs purchased must be
added to the quantity on hand on the first day of the month; from the
total thus obtained must be deducted outgoings not for feeding the family
and the quantity on hand at the end of the month. To enable the statisticians to undertake a critical assessment of the information obtained,
the table also contains space for data on the food consumption of the
family during the previous month.
Organisation of the Inquiry
The inquiry into the budgets of employees is carried out by statisticians specially detached for this purpose from the statistical agencies in
which they are normally employed. Each statistician investigates
between 20 and 23 family budgets. These people play an important part
in the general organisation of inquiries into budgets, for statistical work
in the sphere of family budgets can only be carried out satisfactorily by
qualified statisticians.
The specialists who carry out this work must, first of all, have a
thorough knowledge of the forms used in the inquiry and the appended
instructions. They regularly improve their knowledge by attending
special courses organised by the statistical services. In order to be able
to check the information given by families, the statisticians inquire into
the sales in local shops, deliveries and sales of industrial products and
prices in shops and kolkhoz markets, and make a careful study of the
way of life of the families covered by the inquiry.
Every month, before they carry out the last interviews, the budget
specialists receive " extracts from wage registers " from the accounts
departments of the undertakings in which the employees concerned
work. The statisticians endeavour to ensure that the family members
keep full records of their income and expenditures. This enables the
statisticians to make a critical general assessment of all the replies
received. Allowance must be made for the fact that, although the people
welcome budget inquiries, in a large budget survey project mistakes will
inevitably be made unless the information provided by households is
checked. Most mistakes are due to misunderstanding the questions,
ignorance or lapses of memory. These factors must be eliminated, or at
least their effects must be reduced to a minimum.
In many cases there are two or more persons with incomes in the
family. On the expenditure side, money is spent by every adult member
of the family—and sometimes even by children. Consequently, steps must

BUDGET SURVEY IN TUE U.S.S.R.

be taken to ascertain whether all the members of the family are at home
during the period covered by the inquiry; members of the family who
regularly spend money away from home must keep records of such
expenditure.
In order to be able to fill in the budget questionnaire the statistician
must know how to put his questions simply and clearly and in a manner
that will not place the person being interviewed in difficulties. For
instance, he must not ask how much the interviewed person has spent
on haberdashery, for this term covers a great variety of objects (needles,
buttons, thread, hooks, ribbons, laces, collars, wallets, umbrellas, etc.).
Obviously the question about " haberdashery " is expressed in overgeneral terms, and the interviewed person will have great difficulty in
answering it without leaving any item out.
To avoid errors due to lapses of memory on the part of the respondent, the subsidiary notes, whose data will be of value in filling in the
questionnaire, must be analysed by the statistician before the actual
interview.
In order to keep the reference periods relatively short the families are
interviewed twice every month. The field work lasts ten days. As the
families spend considerable amounts of money each day, the statistician
would have difficulty interviewing every family during the same period.
In addition, the members of a family if interviewed, for example, on the
25th of the month, will have difficulty remembering, for instance, how
much sugar they bought during a reference period running from the
1st to the 15th of that month—i.e., a period which had ended ten days
before the interview. Consequently, the statistician fixes a half-month
reference period for each family; for one family it may be from the 1st
to the 15th of the month, for another from the 6th to the 21st and for a
third from the 10th to the 25th. The statisticians are free to establish
their own work programme.
It sometimes happens that wrong answers to specific questions in
the form are given deliberately. This usually happens in the case of
questions concerning activities or propensities that the family or society
frown upon.
There is, however, one absolute rule about budget statistics: the
statisticians handling them are strictly forbidden to show the individual
budgets to anybody outside the statistical services or to give information
on those budgets.
Once the workers covered by the inquiry are convinced that the
information on their family budgets will not be communicated to anybody they answer the questions more honestly. A more delicate situation
arises when individual members of a particular family do not wish other

FAMILY LIVING STUDIES

members of the family to know how much they earn or spend. In such
cases the statistician must interview the persons concerned separately
and is not allowed to record the information given in the general family
budget until after he has finished interviewing the family.

Verification of Information
Once the information has been collected a final verification is carried
out, which is both technical and analytical in character. The technical
verification consists first of all of comparisons of corresponding figures
given in the different parts of the questionnaire; after this the totals
are checked. The analytical verification consists of a simple co-ordination of the data obtained, followed by an analysis of an economic
character.
First of all, the statistician goes through the questionnaire to find
inter-related items of information. A number of items in the questionnaire
are inevitably interdependent with other items. If there is a pensioner or a
holder of a student bursary in the family, one may expect to find information on pension or bursary payments; if a working member of the
family has fallen sick, the sickness benefits paid should appear in the
section on income ; if the children of the family are attending a kindergarten or studying, payments to the kindergarten or expenditure on the
purchase of books and school material should appear in the record of
expenditure. If the family owns cattle, there should be information on
the production and consumption of fodder.
In addition, as the inquiry into budgets goes on month by month
without interruption, and as the questionnaires are filled in and analysed
every month, the statisticians can, at their leisure, examine trends in the
individual items by studying the changes in the budgets from month to
month. To do this they use a " control book ", in which the basic data
•obtained by the inquiry—the composition of the family, the principal
items of income, the principal items of cash expenditure, the number of
head of livestock and poultry and the produce from these, produce
obtained from the garden or home-processing, the family consumption of
foodstuffs and the quantity of foodstuffs on hand at the end of the
month—are recorded.
The statistical services are constantly endeavouring to obtain exact
information on budgets. In addition to the measures already mentioned,
•considerable importance is attached to the systematic checking of the
work of the family budget statisticians by officials of the regional statistical offices and the district inspectors of the C.D.S. The object of these

BUDGET SURVEY IN THE U.S.S.R.

checks is to maintain high standards in the replies to the budget inquiry
forms, to show the statisticians correct methods of interviewing, to help
them to become more familiar with the inquiry forms and the accompanying instructions and to help them to take part in the dissemination of
knowledge about budget inquiries among the population at large.
Analysis of Results
The analysis of the budgets of employees includes
(a) the establishment of over-all group totals ; (b) obtaining figures
for communal feeding; and (c) verification of the over-all results.
The statistical services only rarely use information taken from individual budgets in their work. They are concerned with the averages
obtained from the budgets of all the households representing the population groups covered by the inquiry. The data obtained by the budget
inquiries are summarised in tabular form. Before the process of tabulation is begun the data of each individual budget must be assigned to a
particular group. This distribution of statistical data into groups plays
an important part in the statistics of family budgets.
Conforming to the objectives in the field of statistics on employees'
budgets assigned by the government to the Central Directorate of
Statistics of the U.S.S.R., the statistical services have, since 1952, been
classifying workers' households into groups according to level of skill
(skilled or semi-skilled) for each industry covered by the inquiry. In
addition, since 1957, monthly analyses have been made of the budgets
of workers on state farms (sovkhozes).
Family budget inquiries for salary earners are conducted for the
following categories : engineers and technicians in industry, white-collar
workers in industry, teachers, doctors and auxiliary medical staff.
The general rules governing the calculation of the quantity of products consumed by employees and members of their families and provided by communal feeding facilities have been described earlier. To
perform this calculation the directorates of the statistical services draw
up summaries of the forms relating to communal feeding undertakings
and institutions for children (kindergartens, day nurseries).
The summaries of the forms concerning communal feeding facilities
contain information on the products used to provide communal meals
as well as on the total receipts, and information on expenditure on
communal feeding is obtainable from the summaries of the workers'
budgets. The ratio between the total receipts and the amounts paid
by the employees covered by the inquiry gives the quantity of produce used for communal feeding consumed by the families covered.

FAMILV LIVING STUDIES

These summaries are also used as a basis for the calculation of the
amounts of produce consumed in institutions for children by the children
of these employees. In this case the ratio is established on the basis of
the consumption of food per child and per day, in institutions for children,
for all children and for the children of employees covered by the inquiry.
When the summaries of the budgets have been completed they are verified
both technically and analytically.
Use of Results
The programme for the survey and study of the budgets of employees
and the parallel inquiry on the budgets of members of kolkhozes (cooperative farms) cover a substantial part of the life and activities of the
population. They provide information on the volume and composition
of income and on the consumption of the different groups of employees
and members of kolkhozes; they provide, also, the data required to carry
out calculations on a variety of subjects, such as the volume of consumption, production for own use, and various balance sheets of the national
economy. Budget inquiries are necessary for planning, for the administration of the state and for scientific research work. Many and varied uses
can be made of the data to be obtained from budgets of these kinds.

IV
A Budget Survey in the Urban Areas
of Greece
By A. B. REISZ
GENERAL DESCRIPTION

The 1957-58 Household Survey in the Urban Areas of Greece was
undertaken jointly by the Government of Greece and the International
Labour Office. The responsibility for the planning and the execution
of the survey rested with the National Statistical Service of Greece,
advised b v an ex n ert a^^ointed b v the International Labour Office under
the Expanded Programme of Technical Assistance of the United Nations.
An advisory committee, on which were represented, among others,
the Ministry of Co-ordination, the Ministry of Agriculture and the
Ministry of Labour, was charged with the co-ordination of planning.
The responsibility for the food consumption and nutritional aspects of
the survey was shared with the chief nutrition officer of the Ministry of
Co-ordination. Important contributions at the planning stage were
also made by the Food and Agriculture Organisation of the United
Nations.
Objectives
The urgent need for the establishment of an official consumer price
index provided the main impetus for the launching of the Household
Survey. The National Statistical Service considered it essential to obtain
reliable data on consumption patterns in order to provide a factual basis
for the weighting of the proposed new index.
In view of the many statistical gaps which it was thought could be
wholly or partly filled by data obtained from households, it was decided,
after discussions with interested government departments, to broaden
the objectives of the survey far beyond the limited aim of obtaining
weights for the index. Inter alia, it was intended that the results of the
survey should strengthen the basis for national income estimation.
Furthermore, in view of the increasing tempo of economic development,

FAMILY LIVING STUDIES

the government attached very special importance to the provision of
data required for demand analysis, in particular to data showing the
relation of household income and household composition to expenditure
on selected commodities or groups of commodities.
Further important objectives of the survey were to obtain data on
the level of living of the urban population in general and in particular on
the comparative expenditure patterns of households of differing social
and economic characteristics, and to collect data relating to food in
sufficient detail for a comprehensive study of food consumption in the
urban areas of Greece. To meet the needs of development planners, the
survey was also to provide data on the structure of households and on
the relation between the demographic and the economic characteristics
of households and household members.
The decision to adopt such wide objectives was influenced by the
fact that household surveys were considered to be particularly well suited
for the collection of some demographic and economic data. Another
very important consideration was cost; by increasing the fields of inquiry
it was intended to reduce the unit cost of collected information.
Scope of the Survey
The survey was originally intended to embrace the entire geographic
scope of the country, including the rural areas. In view of limitation of
resources and certain unsolved problems connected with data collection
from households in the rural areas, however, it was necessary to restrict
the scope of the survey to the urban areas, defined as towns which
according to the 1951 census had a population of 10,000 or more.
The survey covered all private households, irrespective of size and
composition and of the economic and social status of the head. It was
decided, however, to exclude households which included permanent
members of the armed forces and the police or members of foreign
diplomatic and commercial missions.
Collective households of all kinds, e.g. groups of persons living in
hotels, boarding houses, hospitals, clubs, etc., were excluded. Households with three or more boarders were regarded as boarding houses
and were, therefore, excluded.
Sample Design and Method of Selection
The 1951 Census of Residential Dwellings, supplemented by the
Register of Building Licences, was chosen as the frame for the selection
of the sample.

BUDGET SURVEY IN GREECE

A two-stage sample was adopted, the selection of towns and the
selection of dwellings within the sample towns constituting the first and
the second stages respectively. The households occupying the selected
dwelling units at the time of the visit by the interviewer constituted the
sample of households.
Out of the total of 49 towns with a population exceeding 10,000 at
the 1951 census, 25 were selected to represent the urban areas of Greece.
In addition to the ten largest towns, which were all included, 15 of the
remaining 39 towns were selected with probability proportionate to size.
For the purpose of selecting the sample dwellings, the schedules of
the census of residential Dwellings and the entries in the Register of
Building Licences were numbered and the units were selected at fixed
intervals, starting from a random number. The intervals were selected in
such a way as to ensure that the combined list of dwellings selected from
the census and from the register slightly exceeded the number of dwellings required for the sample in each of the towns. The required number
was obtained from this preliminary list using random elimination.
The size of the total sample was 2,830 households. This figure was
arrived at by using the 1951 Census of Population as basis and applying
a sampling fraction of 1 : 250.
The size of the sample in the ten largest towns was proportionate with
the population and varied between 40 in the case of Candia to 1,400 in
the case of Athens. In the 15 towns selected with probability proportionate to size, a fixed number, namely 46 households, were assigned to
each town. This figure was obtained by dividing equally among the
15 towns the aggregate number of households allotted to all the 39 towns,
other than the ten largest ones.
The Organisation of the Survey
The pilot survey which preceded the main survey was intended to
test the schedules and the instructions, to make sure of the adequacy
of administrative arrangements and, last but not least, to provide an
opportunity for the field staff to gain confidence in their approach to
households, to acquire an intimate knowledge of the schedules and
instructions and to obtain experience in questioning respondents and
recording the obtained data.
The analysis of the results of the pilot survey provided also an
opportunity to develop editing and coding procedures, to test the
suitability of the " item codes " and the " sorting codes ", to evolve
a flexible and efficient method of analysis, and to train staff for the
processing of the main survey.

FAMILY LIVING STUDIES

The field work for the main survey, which was launched in April
1957, extended over a period of a full year. In order to provide data on
seasonal variations in consumption patterns, the sample households, each
of which reported for a short period only, were evenly distributed over
the year. The total of sample households was divided into 52 equal and
random groups, one of which was allocated each week to interviewers.
The staff at headquarters dealing with processing and administration
amounted to eight persons, while a total of 44 interviewers, 28 full-time
and 16 part-time, were engaged on field work.
The Questionnaires Used
Interviewers completed six separate questionnaires for each household: the sample control form; two separate questionnaires relating to
household purchases; the record of personal purchases of spending
members of the household ; a short supplementary questionnaire relating
to current expenses of owners of motor vehicles; and last, the record of
individual cash income of all income recipients.
The purpose of the sample control form was to facilitate the control
of interviewers and of the sample. It contained full details of the sample
dwelling and of the date on which the interviewer was expected to pay
the first visit.
Provision was made for questions to determine the eligibility of the
dwelling and of the household for inclusion in the sample, and about the
willingness of the household to co-operate.
In order to assess the effect of refusals on the composition of the
sample, this form included also a number of questions on the characteristics of the household and of the head of household, to be completed
for those households which refused to co-operate. These questions
related to the size and the composition of the household, to the number of
working members, and to the age, occupational status and employment
status of the head of household.
The Main Household Questionnaire was completed in the course of
daily visits over a period of seven consecutive days. The questionnaire
was made up of seven sections :
(1) The composition and characteristics of the household.
(2) Living accommodation and facilities.
(3) Housing expenses (including fuel and light).
(4) Daily record of all household purchases for seven consecutive days
(irrespective of duplication with other sections of the household
schedules).

BUDGET SURVEY IN GREECE

(5) Purchases of major household equipment, furniture, etc., during a
12-month period preceding the interview.
(6) Expenditure during the past 12 months (or other fixed period) on
education, licences, subscriptions, regular contributions, christenings,
weddings and funerals, etc.
(7) Goods and services received free of payment.
Special importance was attached to the section relating to household
characteristics. Data recorded for each household member included age,
sex, relationship to head and, if applicable, occupational status, occupation, industry in which the occupation was carried out, and finally, employment status, i.e., number of hours worked if employed and length of
unemployment if out of work. These data, in addition to permitting the
classification of households according to significant criteria, were to
provide material for the study of the relation between demographic,
social and economic characteristics of individuals included in the survey.
Under the section on living accommodation and facilities, questions
were included on the type of dwelling, its size and its structural condition,
where the household prepared its meals, available cooking facilities,
means of heating and lighting, the source of drinking water, refuse
disposal, etc. In addition, provision was made for the recording of an
inventory of selected items of household equipment, such as number of
beds and blankets, the possession of refrigerators, radios, etc.
The section on goods and services received free of payment, provided
for the recording of payments in kind from employers, items obtained
without payment from own business or enterprise, the value and quantity
of food obtained for household use from own farm or gathered in fields
and forests by household members, items received by way of gift or
charity and free school meals.
The Supplementary Household Questionnaire was completed by the
interviewer over a period of four consecutive weeks, covering purchases
of clothing and household utensils.
The Record of Personal Expenditure covered expenditure which, it
was feared, could not be obtained with any accuracy from the housewife.
The data were collected for seven consecutive days, as far as possible concurrently with the main household questionnaire. This form, which was
completed for every spending member of the household, comprised under
separate sections: meals and drinks taken outside the home; tobacco and
cigarettes ; reading matter and stationery ; toilet requisites and cosmetics ;
public and hired transport; communications, recreation, personal services; and a miscellaneous section for any other items left out from the
household questionnaires.

FAMILY LIVING STUDIES

The Record of Motor Vehicle Expenses was a simple questionnaire on
owners' current expenses. The questions covered the annual cost of
circulation tax, driving licence and insurance; expenditure during the
preceding month on maintenance and repair; and petrol and oil purchases
during seven consecutive days. A further question was included on the
approximate mileage run by the vehicle during the preceding year for
private and for business purposes respectively.
In order not to prejudice the success of the rest of the survey, interviewers were instructed to leave the completion of the Record of Cash
Income until all other questionnaires had been completed.
The record, which was completed for every household member with a
cash income in excess of 75 drachmas a week, covered in detail the following sources of cash income: salaries, wages and bonuses; net income from
business or profession ; net income from the sale of agricultural produce;
income from rents, interest and dividends; and social security receipts
and cash gifts. In the case of wages, salaries and bonuses, provision was
made for the recording, in itemised detail, of deductions at source. Space
was provided against each entry for the recording of the period to which
each item of cash income referred.
Selection and Training of Interviewers
The selection of interviewers, with the exception of some from the
permanent staff of the Statistical Service, was based on competitive
examinations held in Athens, Thessaloniki, Patras and a number of
smaller towns. The examinations consisted of written tests and an interview. The examination papers were drawn up to test the intelligence
of the candidates, their ability to follow instructions, their grasp of complex questionnaires and the neatness of presentation. The interview,
conducted by the Appointment Board, aimed at assessing the personality
of the applicants, their likes and dislikes, and especially their attitude to
outdoor pursuits.
In order to test the efficiency of the method of staff selection, an
attempt was made, after completion of the field work, to correlate the
ranking of 15 Athens interviewers in the competitive examination with
the ranking based on the independent assessment of their work judged by
completeness, neatness and accuracy of questionnaires.
The results of the rank correlation were disappointing. While the
correlation showed that the written tests were of some limited use as
means of selection, the marks based on the interview were practically
unrelated to the performance of the interviewers in the field.

BUDGET SURVEY IN GREECE

The rank correlation coefficient for the written tests and for the interview were 0.68 and 0.22 respectively.
A training course of three weeks' duration was organised in Athens
for all interviewing staff and for chiefs of field offices. The course, in
addition to formal lectures, consisted of discussions, mock interviews in
class, and actual visits to households, with subsequent discussion and
review of the problems encountered.
The heavy expenses of such a prolonged training course, while raising
considerably the total cost of the survey, appear to have been, in the light
of results, an exceptionally good investment.
Co-operation of Households
Either a sole member of the household (e.g., the housewife or the
head of the household) or every spending member of the household
might be asked to provide data on spending and income. The choice of
a sole respondent has the great disadvantage that information obtained
might not cover the complete range of expenditure nor all the sources of
income of the household. Only in the case of items destined for use by
the household as such or purchased by herself can the housewife be
relied upon to provide complete records. In order to obtain reliable data
on the aggregate purchases, and in particular on all sources of cash
income of the household, it was decided to obtain data separately from
every spending member of the household.
In some recent household surveys rewards to respondents appear
to have contributed to improved response rates. The possibility of
making some payment in cash or offering some incentive in kind was,
therefore, examined but rejected in view of the many administrative
difficulties involved in such a step under the conditions prevailing in
Greece.
Out of 2,830 households which constituted the sample, 259 households
refused or were unable to co-operate—a little over 9 per cent. The rate
of non-response in Athens was 14 per cent., while in provincial towns
it was under 5 per cent. Contributory factors to the high response rate
were, among others, the intensive publicity both prior and during the
survey and the use of well-trained interviewers for the collection and
recording of the data.
Less than a quarter of the non-co-operating households gave
objective reasons for refusal, such as recent bereavement, illness in the
family, lack of time, etc., while the reasons put forward by the majority
were disagreement with the aims of the survey, fear of tax repercussions
and other subjective reasons.

FAMILY LIVING STUDIES

The households allocated to individual interviewers in Athens
constituted a subsample of all sample households. The differences in
response rates between individual interviewers might be regarded,
therefore, as related to their skill and ability. Variations in the refusal
rate among the Athens interviewers ranged from 2 per cent, to 30 per
cent. In an attempt to relate the skill and ability of the 15 Athens interviewers, based on editors' reports, to the refusal rate, it was found that
the average refusal rate for the five best, the five medium and the five
poorest interviewers was respectively 8.7 per cent., 14.2 per cent, and
18.7 per cent. These figures, together with the high percentage of households that gave subjective reasons for refusing to co-operate, seem to
suggest that the " hard core " of not-interviewable households has not
yet been reached and that it might be possible to raise response rates by
improved selection of field staff.
Processing of the Data
Sorting Codes.
In order to be able to analyse the characteristics of households and
of individuals comprising the sample, and furthermore to facilitate study
of the expenditure patterns of households of differing characteristics.
information was recorded on a great number of characteristics of the
household and its members. To facilitate the mechanical analysis of
these varied data, 40 " sorting codes " were developed, 33 of which were
reproduced on each of the " primary " punched cards. The complete
set of these " sorting codes " provided a description of each household
and individual. These codes for household characteristics included,
among others, the place of residence, the household composition, the
number of income recipients, the cash household income, detailed
characteristics of the dwelling, the extent of hospitality offered, etc.
Codes for individual characteristics included age, sex, relationship to
head, occupational status, occupation, etc.
Item Codes.
In order to facilitate the classification and mechanical aggregation of
the many thousands of different items that were recorded in the schedules,
item codes were developed, permitting the reduction of the number of
separately coded items to manageable proportions. A four-digit code
was used for classification. The first digit denoted the source of the goods
and services, the second digit was used to divide goods and services into
nine main groups, the third digit denoted classes within the group and
the fourth digit was used to identify items.

BUDGET SURVEY IN GREECE

Special importance attaches to the first digit of the code, the classification by source, which constituted an attempt to provide, on an experimental basis, for the separate identification of (1) purchases; (2) goods
and services obtained without payment from farm or business ; (3) payments in kind by employers; (4) gifts and charities in kind; and (5) free
school meals. The results have shown that the coding jointly of farm
produce and of goods and services obtained from non-farm enterprises
was not appropriate, and that a separation of these figures would have
enhanced the value of the results. This planning error was due to the
complete absence of any information, prior to this survey, regarding the
importance in the urban areas of subsistence production of farm produce.
Goods and services obtained free of payment equalled 16 per cent, of
the total value of purchases. Free food amounted to some 10 per cent.
of food purchases, a very much higher figure than anticipated in an
urban survey.
The above averages conceal considerable variations attributable to
differences in the size of town and in income and other characteristics of
households.
Editing.
All decisions involving statistical judgment devolved on the editors,
whose responsibility comprised the critical examination of all completed
schedules to ensure that there were no omissions of relevant data, that
the data furnished were reasonable and consistent, that all exceptional
features were fully commented on by interviewers and that duplications
among various sections of the schedules and of purchases reported by
more than one member were eliminated. The " sorting codes ", that is,
the coding of the characteristics of the household and of individual members, a task of special importance, was also the responsibihty of the
editors. Their work, furthermore, included the conversion of quantities
and values into uniform units.
In order to enable the editor to arrive at a general assessment of the
household, a summary of income and expenditure was prepared, without
attempting, however, a cash reconciliation.
Detailed editing instructions were drawn up which were amended and
supplemented in the course of the work.
Coding.
The work of coding was largely mechanical, consisting of the entry
in the schedules of a code number against each entry relating to goods and
services and to cash income. In order to increase both efficiency and accuracy, the coding of the different sections was allocated to specialist coders.

FAMILY LIVING STUDIES

List Editing.
The correction of errors once the tabulation stage has been reached
would have involved very considerable tabulation delays. List editing was
introduced in order to ensure that as far as practicable errors were detected at an early stage, while the punched cards were still filed separately
for each household. Immediately after completion of the punching and
verification of the punched cards of a particular household, a list was
prepared on the tabulating machine showing, in addition to the complete
household characteristics, the total value and, if applicable, the quantity
of each item against the appropriate code number. These lists were
carefully scrutinised and, if necessary, checked with the questionnaires.
The large majority of errors detected were due to editors' mistakes. The
number of errors by coders were few and errors due to mistakes in the
punching of cards and due to machine failure were negligible.
Tabulation and Analysis of Data.
In order to ensure maximum flexibility, a separate " primary card "
was punched for each entry relating to goods and services obtained and
relating to cash income. Each primary card contained the appropriate
code number of the item, the value and, if applicable, the quantity and
also the complete coded characteristics of the household.
All the primary cards of each individual household were then sorted
according to item code and aggregated. The list of code totals provided
the basis for list editing.
Simultaneously with the mechanical aggregation of the primary cards,
a summary card was automatically punched for all primary cards bearing
the same code number. This summary card, the " item card ", was used
for the tabulations. Provision was made for the automatic punching on
each item card of a figure indicating the number of primary cards that it
replaced. The item card thus provided information not only on the total
value and quantity of entries for an individual household in respect of
every separately coded item but also on the number of recorded entries.
In the case of bread purchases, for example, the item card contained, in
addition to the aggregate value of purchases entered, the number of
replaced primary cards. This made possible the estimation of the frequency as well as the average value and quantity of the purchases of
bread. In order to advance the speedy analysis of the characteristics of
households and of household members independently of the main body
of punched cards, a set of " master household cards " and " master
member cards " was punched incorporating all the codes relating to
household and individual characteristics.

BUDGET SURVEY IN GREECE

Tabulations were prepared presenting weekly averages classified according to principal household characteristics, such as size and composition, cash income, occupation and occupational status of head, etc. In
the case of the most important household characteristics, such as size
and composition of household, the data were further subclassified by
income. In addition, a large number of tables have been compiled
relating to the economic, social and demographic characteristics of
households.
While the tabulated results of the survey have already made an important contribution to Greek statistics, even greater value attaches to
the accumulated file of punched cards. These cards constitute a valuable
and highly flexible depository of economic, social and demographic data,
enabling the Statistical Service to provide the government, at short notice,
with data ranging over a wide field and in the form required for specific
purposes of planning and administration.
Publication of Results
The first seven major tables were circularised in duplicated form to
the members of the Advisory Committee and to government departments within ten months after the completion of the field work. Preliminary reports comprising a brief description of methods and summaries of
the principal results were published in the April 1959 and the July 1959
issues of the Monthly Statistical Bulletin of the National Statistical
Service of Greece.1
Continuous Household Surveys
Encouraged by the results of the 1957-58 Household Survey, the
government decided to carry on with the survey, on a small scale, as part
of the permanent work programme of the Statistical Service. The
methods and procedures adopted for these " continuous surveys ", which
in the years 1958-59 and 1959-60 comprised 300 households, are in substance identical with that of the 1957-58 main survey. The approved plans
for the 1960-61 continuous survey provide for raising to 600 the number
of households in the sample.

1
Monthly Statistical Bulletin (Athens, National Statistical Service of Greece),
Vol. IV, No. 4, Apr. 1959, pp. ix-xxv; and No. 7, July 1959, pp. ix-xxiv.

FAMILY LIVING STUDIES

SOME GENERAL CONSIDERATIONS AFFECTING THE CHOICE OF METHOD OF
COLLECTION AND PERIODS OF REFERENCE IN HOUSEHOLD SURVEYS

Two Methods of Data Collection
As this was the first survey of its kind to be undertaken in Greece, the
population had never before been asked to answer detailed questions
regarding family expenditures and consumption. The methods to be employed for obtaining the data were therefore of crucial importance to the
success of the study. As background for the description of the procedures
finally adopted, it might therefore be useful to review the pertinent
features of the two principal methods of collecting data from households.
These two methods are commonly referred to as the " interview method "
and the " account book method ".
In the interview method, trained investigators obtain and record information on transactions over a certain period in the course of one
single or repeated interviews. Usually fully classified and, in order to
facilitate processing, also pre-coded questionnaires are used.
The account book method provides for the current recording of
transactions over a certain period by the respondents themselves. Specially designed account books are supplied for this purpose.
The choice of the method of data collection in a family budget survey
is affected by a number of factors, among which the effect of the method
on non-response, on accuracy of information, on the processing of the
data and on cost are among the most important.
Non-response.
Owing to the strain of keeping daily records for a protracted period,
the refusal rate is often higher with the use of account books than in the
case of interviews. This has, in many surveys conducted by the account
book method, prevented the use of probability sampling and led to the
application of quota selection or to the recruiting of respondents from a
roster of volunteers. Illiteracy among the population under survey may
also present a serious obstacle to a random selection of respondents with
the use of account books. A skilled interviewer can, on the other hand,
make the interview both interesting and stimulating, thus creating an
atmosphere favourable for getting co-operation from sample households.
Accuracy of Information.
In the case of the interview method, the accuracy of the information
collected depends largely on the ability of the respondents to recall items
of income and expenditure. Unless written records are available, errors

BUDGET SURVEY IN GREECE

due to memory faults may considerably detract from the value of the
results. The impact of such errors depends both on the frequency and
regularity of the transactions to which the questions relate and on the
length of the period for which the data are to be supplied.
Expenditure on items purchased frequently may not be recalled
easily for longer periods than a day or a week. A respondent may therefore be expected to report more accurately daily than, say, monthly
purchases of bread, milk, etc. Any extension of the period of reference
will in this case tend to increase reporting errors. Information on items
purchased in quantity to meet the needs of a week or a month may, on
the other hand, be obtained with reference to a longer period than
information on more frequently purchased items. As regards items
purchased at long intervals, e.g., major household equipment and
furniture, respondents may accurately recollect details referring to considerable periods, including as long as a full year, without serious
reporting errors.
When account books are used, the accuracy of the data depends to a
large extent upon the promptness of the recording. If the recording is
completed on the same day or on the spot, as purchases are made, the
records might be expected to be accurate. If data are recorded hurriedly
and without care for a period of many days at a time, less reliability is to
be expected.. Frequent and careful checking of the respondents' account
books is therefore essential for accuracy.
The length of the period for which each individual household supplies
information by this method may also affect the accuracy of data. Thus,
the keeping of records may change the purchasing habits of the household. As time passes, the very fact of record keeping will then render
reporting households less and less representative of the community
which they have been selected to represent. Other sources of reporting
errors in the case of account books are the " end-of-period effect " for
short periods of reference and fatigue in the case of prolonged periods.
The " end-of-period effect " results from a tendency on the part of respondents to include during the first few days of recording purchases
effected prior to the commencement of the survey, thus overstating their
expenditures. The less scrupulous reporting of purchases due to fatigue
can result in the apparent falling off of purchases after the lapse of some
time.
The accuracy of the information obtained by the interview method
is, again, affected not only by faults of memory but also by the personality of the interviewer and by the technique of probing which, by
influencing the answers of respondents, may introduce a source of
error.

FAMILY LIVING STUDIES

The use of fully classified questionnaires in the case of the interview
method also represents a danger. If the adopted classification appears
artificial to the respondents, the recording of the information following
such a frame can lead to bias. Small and unusual items of consumption,
including commodities that are new in the market, might be omitted.
The discretion given to interviewers is also considerable. Decisions
about the classification of marginal items have to be made on the spot,
with a consequent lack of uniformity of coding.
In keeping account books respondents are, on the other hand, encouraged to enter in full detail every single transaction, which considerably
reduces the risk of systematic omission of specific items.
Processing of Data.
The main advantage of a fully classified questionnaire such as that
used in interviews is attained in the processing stage. Items can be precoded and manual coding is then reduced or entirely eliminated. Account
books, completed by respondents, in addition to lacking uniformity,
might be inconsistent and illegible, requiring a greater volume of checking
and coding. A sizable portion of the account books might actually have
to be rejected, which introduces a danger to the representativeness of
the results.
Cost.
The principal factors influencing the cost of the operations are, in
the case of interviews, the salaries of the interviewers and, for the account
book method, the cost of supervising the respondents, the cost of unusable account books and the higher cost of processing.
In the case of the interview method, the length of the periods for which
individual households supply data does not materially affect cost. Some
additional probing may be necessary on the part of interviewers if
information is collected for a year instead of for a month, but the resulting
increase in interviewing time can well be ignored for practical purposes.
Other Considerations.
Only where account books are used is it practicable to record individual purchases separately and to obtain, in addition to data on the
seasonal variations in the pattern of consumption, basic materials for a
detailed study of purchasing habits, including the frequency and periodicity of purchases, the preference, if any, as to day of purchase, and the
average value (and quantity, if recorded) of separate purchases.
The exceptional flexibility associated with detailed recording might be
one of the principal advantages of the account book method.

BUDGET SURVEY IN GREECE

A Compromise between the Two Methods of Data Collection
Consideration of the comparative merits of the account book and the
interview method of collection raises the question of whether it is practicable to use a compromise method combining the principal advantages
of both.
Entrusting interviewers with the responsibility for recording into
account books the data obtained in the course of a series of daily interviews is a procedure which combines the advantages of separate and
accurate records of daily purchases with that of uniform and carefully
made entries.
The data obtained using the account book method are at best no
better than those obtained by a series of daily interviews. If, for purposes
of supervision, daily visits are made to the households keeping account
books, then the only question is whether the respondent or the interviewer
makes the entries on the schedule. Provided that the period for which the
individual households supply data is not in excess of a few weeks, the cost
of the daily recording of transactions by interviewers need not exceed the
cost of using the account book with entries made by the respondents.
The recording of transactions in the account books by interviewers
may, in fact, lead to some important economies through a lower refusal
rate, the reduction of waste due to carelessly or illegibly completed
account books and the expected improvement in the accuracy of the data
which tends to reduce the unit cost of the returns. Lastly, the cost of
processing may be reduced because account books are completed
uniformly and in agreement with instructions.
Choice Oj Reference Periods
The determination of reference periods for the various types of
information sought requires careful consideration.
As already noted, changes in the length of the period for which
individual households are asked to supply information do not materially
aifect the cost of the interview method. The length of the period has
therefore to be determined with regard to the size of the reporting and
sampling errors that can be expected to follow from the use of a certain
reference period.
For items such as milk, bread, fresh vegetables, etc., which are
purchased in small quantities on a daily basis, the extension of the
period of reference from a week to a month might result in a mere
repetition of a cycle of identical purchases and the samphng error
attributable to day-to-day variations will not be significantly affected.

FAMILY LIVING STUDIES

In contrast, a similar extension of the period of reference for less frequent
purchases, such as clothing and furniture, might be expected to be
attended by a reduction in the sampling error.
In view of the great variations in the frequency and the regularity of
purchases among classes of goods and services, a strong case can be made
for utilising differential periods of reference for separate groups of items.
Costs of collection and of processing represent, in the case of the
account book method, the most important consideration. Any extension
of the period during which individual households record transactions in
account books entails an increase in costs which, given fixed resources,
has to be counteracted by a reduction in the size of the household
sample. The most efficient way of minimising sampling errors may in
this case be to adopt a relatively short reference period but to increase
sample size for frequently purchased items and to extend the period of
reference, at the same time reducing the sample for items purchased at
infrequent intervals.
The difficulties of administration and of processing created by such a
combined differentiation of reference period and sample size may,
however, put undue strain on resources.

METHODS OF COLLECTION AND PERIODS OF REFERENCE ADOPTED IN THE
. URBAN HOUSEHOLD SURVEY IN GREECE

Owing to the lack of prior experience in Greece, the decisions about
methods of collection and periods of reference had to be largely based on
crude data collected on an experimental basis during the early planning
stages from a small sample in the Athens area, chosen from a roster of
volunteer households.
The results of the pilot survey confirmed on the whole the correctness
of these decisions but were regrettably not available until some months
after the launching of the main survey.
The frequency and regularity of purchases by Greek urban families
was found to vary widely among groups of goods and services. The
range of items purchased in very small quantities, on a hand-to-mouth
basis, was much larger than expected, while purchases of some items,
particularly the durable and semi-durable goods, appeared to be less
frequent than in economically more highly developed countries.
For the purpose of selecting methods of collection and periods of
reference, all goods and services were grouped according to frequency
and regularity of purchases into the following categories:

BUDGET SURVEY IN GREECE

(a) " daily purchases ", such as food, household supplies, transport,
recreation, etc. ;
(b) clothing, household utensils, etc. ;
(c) items purchased infrequently, such as furniture, major household
equipment, etc. ;
(d) items in respect of which regular payments were made at fixed but
differing intervals, such as rent, electricity, water, etc.
Daily Purchases.
In order to reduce reporting errors arising from faulty memory and to
raise the standard of recording, the account book method of collection
was adopted for the recording of " daily purchases ". But the task of
eliciting the correct and complete description of transactions and the
responsibility for recording the entries in the " account book " was
entrusted to trained interviewers.
The period of reference chosen for daily purchases was one week; an
extension of the period for which individual households were to keep
records beyond this period was not expected to lead to a reduction in
sarm^lin*7 error commensurate with the additional cost involved in
collecting and processing the data.
Clothing, Household Utensils, etc.
As in the case of " daily purchases ", it was decided to adopt for these
less frequently acquired items a compromise between the account book
and the interview methods. There was no factual evidence to show that
this method was superior to the interview method proper, but the decision
was governed by the need to obtain details relating to each individual
purchase and by the fact that a complete and acceptable item classification required for the drafting of a classified questionnaire and as a basis
for prompting by interviewers was not available.
The questionnaire was therefore designed in the form of a " sectionalised " account book.
On the basis of available data on comparative frequency of purchases
it was decided to adopt four weeks as period of reference for these items.
Recording started simultaneously and continued during the first week
concurrently with the collection of data on " daily purchases " ; in the
remaining period the frequency of visits was reduced to twice a week.
Furniture, Major Equipment, etc.
Because of the infrequency of purchases and the desirability of precluding excessive sampling errors, a full year was adopted as the period

FAMILY LIVING STUDIES

of reference for the information on consumer durable goods and certain
other similar items.
The excessive cost of keeping account books for such a long period
left no alternative but to adopt the interview method of collection. Data
were obtained in respect of purchases made during the 12-month period
preceding the interview.
Recourse had to be taken in this case also to a " sectionalised "
questionnaire instead of using a fully classified questionnaire. In order
to obtain basic data for the study of seasonal variations, provision was
made in the questionnaire for the recording of the date of the transactions in addition to the full description and value of each separate
purchase.
Regularly Occurring Periodic Payments.
The interview method was used for the collection of data relating to
such regular payments as rent, electricity, water, etc. There was no fixed
reference period for these items. Instead, the amount of the last payment,
and the period in respect of which the payment was made, was reported.
Assessment of the Choice of Periods of Reference
In order to assess on a factual basis the correctness of the decisions
on the length of period of reference adopted for the main survey, two
special studies have been undertaken.
The first was based on the pilot survey and restricted to food purchases. It aimed, in general, at clarifying the relation between the results
obtained by the use of a reporting period of three weeks as adopted in
the pilot survey and those obtained on the basis of one week of reporting
as adopted for the main survey.
The results of the pilot survey permitted the calculation of coefficients
of variation with reference periods of different length.
The coefficient of variation of mean food purchases amounted to
5 per cent, for the first week of reporting, to 4.7 per cent, for the first two
weeks combined and to 4.6 per cent, for the full period of three weeks.
Thus, the extension of the period of reference from one to three weeks
did not result in a reduction of the coefficient of variation of the mean
by more than about 10 per cent. On the sole grounds of reducing the
sampling error, there would have been, therefore, no justification for
extending the period of reference beyond the one week adopted for the
main survey.
The means and the standard errors of the means were also calculated
separately for each of the three consecutive weeks covered by the pilot
survey. These figures are shown in Table I.

BUDGET SURVEY IN GREECE
TABLE I. E X P E N D I T U R E O N F O O D
(In drachmas)

Mean expenditure

Standard error
of mean

First week

271

13.5

Second week

254

13.5

Third week

272

13.5

Period

If respondents had either overstated their expenditure during the
first few days of reporting or, because of fatigue, failed to report some
of their purchases after the first week or so of the survey, one would have
expected, in successive weeks, a progressive drop in total purchases.
As there is no evidence of such a trend, the results do not appear to have
been affected by these types of reporting errors.
Comparisons of the coefficients of variations of mean purchases for
selected groups of items were developed on the basis of the main survey.
The results are shown in Table II.
TABLE II. COEFFICIENT O F VARIATION O F T H E
MEAN WEEKLY PURCHASES

Goods and services

Period of
reference

Coefficient of
variation of the mean

(%)
Food
Bread, flour, cereals, etc. . . .
Meat
Fish
Oils and fats
Dairy products and eggs . . .
Pulses, vegetables and fruits
Sugar and confectionery . . .
Other food
Beverages (excl. alcoholic drinks)

Week

2.03
3.35
15.01
3.84
4.14

Alcoholic drinks and tobacco . .
Household supplies and services .
Medical care
Recreation
Transport and communications .
Clothing
Furniture
Education

1.22
1.36
1.87
2.22
2.37
1.63
1.38
2.54
2.86
2.55

Month

3.08

Year

8.64
7.49

FAMILY LIVING STUDIES

The coefficients of variation of the mean were found to be lowest in
the case of those items for which one week was adopted as the period of
reference and highest for the items with a reference period of a full year.
There was, however, one important exception. The coefficient of variation of the mean expenditure on medical care was nearly four times as
large as that of the mean expenditure on transport and communications,
the group with the next highest coefficient of variation whose period of
reference was one week. In view of the high correlation between the
expenditure of households on medical care during the first and subsequent
weeks, an extension of the period of reference would presumably not
affect the sampling error significantly. Accordingly, in the case of expenditure on medical care, there appears to be no other practical means of
reducing the coefficient of variation but to increase the number of
respondents.
While the introduction of differential periods of reference appears
to have reduced differences between coefficients of variation, substantial
differences still exist, e.g., between food and furniture. This raises the
question of whether it would be practicable to take the method a step
further by increasing the differences between periods of reference, and
thus to render coefficients of variation more equal. Regrettably, there
are serious obstacles to such a step. In the case of food purchases there
is evidence of day preferences within a seven-day cycle. The reduction
of the period of reference for food to less than a week would, therefore,
introduce some difficult problems of sampling design as well as of
analysis and administration. As regards furniture, a much greater
knowledge of reporting errors than is presently available would be
required before one could contemplate extending the period of reference
beyond one year. The only way open for further reducing differences
between coefficients of variation appears to be to obtain data on the
consumption of selected items from an enlarged sample of households.

V
A Farm Household Budget Survey
in Germany (Federal Republic)
By Herbert KÖTTER and Joachim LUNZE

BACKGROUND AND PURPOSE

Reports on the Agricultural Situation in Accordance with the
Agricultural Act
Certain elements of backwardness in a°riculture have resulted in an
increasing volume of interest in the problems of the agricultural population in the industrialised countries, both on the part of the public and in
economic and social policy-making circles. In 1955 the so-called " Agricultural Act " was passed in the Federal Republic of Germany, requiring
the government to take measures that would enable agriculture to participate in the progressive development of the national economy. Existing
natural and economic disadvantages are to be compensated and the social
position of persons engaged in agriculture assimilated to that of comparable occupational groups. Since agriculture as a simple entity does
not exist and the situation of farmers varies considerably with external
circumstances and individual behaviour, any deliberate policy of promotion calls for exact determination of the relative situation of agricultural
families. In particular, investigation must turn on—
(a) income from farm production and income of agricultural families ;
(b) agricultural labour input and hours worked ;
(c) the income situation of the so-called " comparable occupational
groups ".
The following report deals with the methods of determining income
and consumption in agricultural families. Determination of hours
worked necessitates separate investigations, which had not yet been
completed at the time of writing.

FAMILY LIVING STUDIES

The main obstacle to the determination of hours of work of persons
engaged in agriculture is that in family farms work for the farm and
work for the private household cannot be easily distinguished in many
cases. Nor is it easy to draw the Une between work on strictly farming
activities and other work done by the same workers. Hours actually
worked can only be determined with a reasonable degree of accuracy
by using work journals.
The study of hours worked had therefore been using work journals
kept throughout the year in 800 farms selected on a representative basis
by the Federal Statistical Office, which show hours worked on an hourby-hour basis for all persons engaged in agricultural activities. Entries
are made at the actual farms and are supervised by consultants. The
study is being conducted under the direction of the Research Association
for Agricultural Policy and Rural Sociology, Bonn.
The elements used in the selection of farms were size of farm, crop
system and size of family. The procedure followed by the Federal
Statistical Office was to take addresses for the individual districts from
the 1950 census results. Some 70 per cent, of the farmers selected in the
first stage expressed their willingness to co-operate. The remaining
30 per cent, were made up by taking alternative farms with similar
characteristics. It can already be said that in some 95 per cent, of the
cases work journals have been properly kept on farms.
Industrial workers living in rural districts were taken as the " comparable occupational group ". The problems and methods involved in
ascertaining " a comparable wage " have been discussed in an article that
appeared in the International Labour Review.1 This particular question
cannot be discussed at any greater length here, but the survey described
below must be seen in the light of the investigations mentioned above.
The Agricultural Act requires the federal government to submit an
annual report on the situation in agriculture forming the basis for action
to be taken.2 This " Green Report " draws on the bookkeeping of some
7,000 farms scaled according to size, crops grown and locality. Private
and public bookkeeping organisations are used for supervision and
analysis of data. The results are compiled by the Ministry of Agriculture
1
See H. KÖTTER: " The Comparison of Industrial and Agricultural Earnings ",
in International Labour Review (Geneva, I.L.O.), Vol. LXXXII, No. 1, July 1960,

pp. 44 ff. See also H. KÖTTER, K. DAHM and B. VAN DEENEN : " Die wirtschaftliche und

soziale Lage von Arbeitnehmern in agrarischindustriellen Mischgemeinden der
Bundesrepublik 1957 ", in Berichte über Landwirtschaft (Hamburg, Bundesministerium
für Ernährung, Landwirtschaft und Forsten), New Series, Band XXXVI, Heft 1,1958,
pp. 1 ff.
a
See Bericht über die Lage der Landwirtschaft (Grüner Bericht) for the years
1957, 1958, 1959 and 1960 ffionn).

FARM HOUSEHOLD SURVEY IN GERMANY

in a report submitted with commentaries to Parliament, which decides
what action should be taken.
The income from farm production is not in itself a full indicator of how
agricultural families live. For this the over-all income, the material and
non-material living standards and patterns of behaviour must also be
taken into account. The Ministry's findings are not sufficient to give an
accurate picture of the social and economic situation of agricultural
families and are regularly supplemented by more intensive sociological
surveys. The general purpose of all this work is to throw light on the
situation of agriculture as part of the general social and economic
development.
One of the institutions entrusted with this form of supplementary
investigation is the Research Association for Agricultural Policy and
Rural Sociology. As the methods of any survey are quite considerably
affected by the institutional framework within which it operates, the
structure and methods of this institution are of some significance.
It is a private body composed of university institutes that have come
together in order to carry out joint research work. Although their
p a l L1VU1U1

Iklt^lWlJ

* (4.1. J—

'VWUVtUlVj

C*.£,it+iJ.l*AJ.j

i i x u i l u g V i i i v u t. W i i U .

LPV/viv

logical institutes all take part—the joint work is carried out on a generally
accepted working plan. Financing is mainly through funds provided by
the Federal Ministry of Food, Agriculture and Forestry, special allocations being made for the establishment of one or two additional
assistants in each institute. These assistants are qualified economists
or agricultural experts and are used exclusively for special tasks of
this nature.
The co-operation of institutes following different specialities and
covering different regions offers the advantage that both the different
objective aspects and the regional variations can be taken into account in
a single joint programme. The Association has a central research office
in Bonn. Once the basic plan has been discussed and agreed upon at a
meeting of the directors of the institutes concerned, the research office
takes over co-ordination of current work and prepares the over-all
conclusions from the results of these different surveys. The fact that this
Association is a permanent scientific institution suggests to some extent
the pattern of organisation of the investigations.
General Character of the Survey
As has already been mentioned, the survey must be seen in the
framework of the annual reports required under the Agricultural Act.
Before the methodology for determining income and consumption of

FAMILY LIVING STUDIES

agricultural families is discussed in greater detail, it should be pointed
out that this represents part of a broadly conceived study by the Association, the results of which were published in 1957 under the title of
" Materials for Determination of the Economic and Social Situation in
Agriculture in Various Regions of the Federal Republic of Germany ".1
The general survey is of course considerably wider than the problem
of determination of income and consumption, to be described below,
and a few preliminary words on the background to the whole question
may not be out of place.
The economy and society of the Federal Republic of Germany are
profoundly influenced by industrialisation. The actual wording of the
Agricultural Act shows clearly that the assessment of the situation of
agricultural families should be based on a comparison of their situation
with that of comparable occupational groups. As rural areas are also
becoming increasingly industrialised or at least have come to house large
numbers of non-agricultural workers, the mass of West German farmers
nowadays live in very close vicinity with tradespeople and industrial
workers living in rural districts. In many areas those engaged in agriculture make use of the opportunity afforded by industrialisation either
to change their occupation or to supplement from outside sources what
they consider an insufficient income from agriculture alone. The opportunity of comparing their own situation every day with that of persons
not engaged in farming has a considerable effect on their satisfaction or
the opposite. Moreover, not only the income levels are compared. The
hours of work, the type of work, or the position enjoyed under the
national system of social insurance are equally important. There are also
intangible factors, such as the particular ways of living, social prestige,
and satisfaction through work responsibility, which Bellerby sums up as
" psychic attractions ", but which are mainly imponderable elements
that can hardly be evaluated.2
While measurement of hours of work frequently encounters considerable difficulties, particularly in the case of small farms, income and
consumption can be evaluated with reasonable accuracy. In judging
the material standard of living of agricultural families in an industrial
society, three fundamental elements have to be borne in mind :
1
Forschungsgesellschaft für Agrarpolitik und Agrarsoziologie: Materialien zur
Feststellung der wirtschaftlichen und sozialen Lage der Landwirtschaft in verschiedenen
Gebieten der Bundesrepublik (Bonn, 1957).
s
J. R. BELLERBY: Agriculture and Industry Relative Income (London, Macmillan,
1956). See also P. VON BLANCKENBURG: " Berufszufriedenheit und Unbehagen in der
Landwirtschaft ", in Soziale Welt (Göttingen), Vol. 9, No. 1, 1958, pp. 32 ff.

FARM HOUSEHOLD SURVEY IN GERMANY

(1) What counts in deciding whether the agricultural population is contented or not is not so much the absolute income level as the relative
place in the income pyramid.
(2) The situation of agricultural families often depends not only on
the income from farm production but also on the availability of
outside sources of income, and this is particularly so for small farms.
(3) Relative satisfaction is achieved if level and composition of income
permit a standard of living equivalent at least to that of an industrial
worker.
Thus, in an industrialised country any study of income and consumption in agricultural families must be so designed as to permit comparison
with the non-agricultural population.
Special Emphasis of the Survey
Although only the methods followed in ascertaining income and
consumption will be discussed here, it will be seen that despite this
deliberate limitation valuable conclusions can be drawn throwing light
on the whole survey.
Notwithstanding certain trends towards uniformity, the agricultural
household reveals various peculiarities calling for special methods of
investigation. Surveys on income and consumption in wage and salary
earners' households have been carried out for some time now in the
Federal Republic of Germany. 1 Information on agricultural families is
much more restricted. If reasonably indicative findings are to be obtained, the following elements must be clearly covered:
(a) individual income elements ;
(b) the number of persons earning incomes ;
(c) the number of persons who have to live from the income earned ;
(d) consumption breakdown.
Particular problems arise as regards the composition of income in
cash and kind, and from the fact that the private budget and the farm
budget frequently coincide in the running of a farm.
Income from farm production in an industrialised society is seldom
identical with the income of farming families, which may be composed
of the following elements :
1
See " Der Verbrauch in Arbeitnehmerhaushalten ", in Wirtschaft und Statistik
(Stuttgart, Statistisches Bundesamt), 9th Year, New Series, Jan. 1957, pp. 49 ff.

FAMILY LIVING STUDIES

(a) farming activities;
(b) non-farming activities;
(c) outside activities;
(d) pensions and social benefits;
(e) capital.
Without precise, methodical division of these constituent elements, the
income elements cannot be determined and family income cannot be
evaluated.
(a) Income from farming, related to the agricultural family, constitutes the gross agricultural income. It represents the income from
the family's work and interest from invested capital, and breaks down
into:
cash surplus or subsidies;
+ estimated value of own farm production consumed, including
imputed rental value of farm;
t changes in value of assets (livestock, stocks, premises, machinery).
After fixing the amount representing interest on capital, the working
income remains as the residual income. However, all that is available
for consumption is the cash surplus less new investments, plus personal
consumption. Thus, consumption income and property growth must be
distinguished.
(b) Non-farming operations include forestry or extraction of natural
resources (e.g., sand or stone quarrying) or technical transformation or
immediate use of agricultural produce (e.g., distilling, taking in paying
guests). Income from non-farming operations is calculated in the same
way as income from farming.
(c)-(e) These headings include sources of income obtained outside
the farm by members of the family living in the same household. The
size of such incomes is determined by the regional economic structure,
conditions of ownership, the size and composition of the family, its
occupational and social structure and the distribution of the manpower
available among farm work, outside activities and work around the house.
For the purposes of the study, determination of the qualitative composition of family income was essential. The whole style of living, the
working economy of the household and the pattern of consumption are
vitally affected by the composition of income from elements of cash and
kind. Comparison of agricultural and non-agricultural income is further
complicated by the fact that the former almost invariably consists of
income earned by several persons or, more correctly, by the family as a

FARM HOUSEHOLD SURVEY IN GERMANY

unit. The point is therefore to find out the full manpower available in
order to relate total income to that figure. It must also be attempted to
distinguish between farm work and work for the private household,
which often overlap.
Absolute income figures give little indication of the possible living
standard. What matters is the number of persons who have to live from
that income. The reference point for income and expenditure was
therefore taken as a calculated unit, the " fully maintained " person, as
described in detail below.
The structure of consumption permits of certain conclusions a
posteriori with regard to actual standards of living, subject to reservation.
For the purposes of comparison, expenditure groups were taken on a
similar basis to that used in surveys relating to the living standards of
wage and salary earners' households.
ORGANISATION AND EXECUTION OF THE SURVEY

Selection of Research Workers
The organisation of the study reflected the structure and working
methods of the Research Association responsible for this work. Eighteen
university institutes each contributed the services of two assistants, all of
whom had full qualifications and had previously taken part in other
surveys conducted by the Association. In view of the very detailed
methods of investigation and evaluation problems that arose, it was
exceedingly valuable to have scientifically trained personnel.
After the basic lines had been decided upon at a discussion of the
directors of the institutes concerned, the central research office prepared
the survey materials. At a seminar lasting several days, the 36 assistants
became acquainted with the special techniques to be followed. It was not
necessary to organise any longer period of training, as they had all had
some experience in such research work. Further co-ordination was the
responsibility of the central office. Although very detailed questionnaires
and investigation schemes were worked out, it was found necessary in
practice to hold several practical discussions during the survey, and it
was thus possible to clarify some points of uncertainty that had arisen.
This teamwork by scientific institutes covering the whole of the Federal
Republic of Germany proved wholly satisfactory.
Extent of the Survey and Selection of Subjects
A total of some 550 farms and families were covered by the survey.
It was not intended to be representative of the whole of the Federal
7

FAMILY LIVING STUDIES

Republic of Germany. The aim was to cover the social and economic
situation of the agricultural population under varying regional conditions. The central concern of this special survey was not so much the
income from farm production as the standard of living of the farm
population. The realisation, gained from previous studies, that the
situation of the agricultural population does not depend exclusively on
agricultural conditions in a narrow sense, but on the sum of all economic
and social factors taken together, was of decisive importance in the
method of selection. Areas with a reasonably uniform structure were
selected first of all, generally consisting of two or three districts (Kreise),
scattered over the whole country.1 The criteria adopted were designed
to cover as far as possible all factors which might be expected to affect
income formation and standard of living of the agricultural population. Particular attention was paid to natural conditions, agrarian
structure and the degree of industrialisation. For purposes of selection,
primary statistics were employed.2 This material also enabled the
sizes and types of farms to be determined for the areas of investigation.
In the final choice of farms no exact mathematical representation was
attempted. What was considered more important was to select sample
units reflecting, in their diversity, all the different features revealed by
the primary statistics. The final result was that in each of 18 areas, from
the north coast to the Alps, some 30 farms were covered. The final
choice was made in close liaison with the competent agricultural schools,
economic adivisory offices of the state agricultural administration and
the occupational associations. Primary emphasis was laid on averagesized and small farms, according to the agrarian structure in western
Germany, but definite part-time holdings were excluded, where the
farmer's main occupation was outside agriculture. Farmers co-operated
on a voluntary basis and were paid no fees. Nevertheless, a sufficient
number were prepared to help.
Since the survey was very much of a model study, in which the first
concern was to see whether and in what way detailed methodical coverage
of income and consumption in agricultural households can be achieved,
the decision not to aim for mathematical representation seems quite
justified. However, if any such typical selection is to correspond reasonably to actual distribution, appropriate primary statistics must be
1
The Kreis is a minor administrative unit. The Federal Republic of Germany has
425 country Kreise and 140 towns independent of these districts.
2
In particular, the Kreis maps of the Institute for Topographical Research, Bad
Godesberg. Cf. also G. MÜLLER: Die Stadt- und Landkreise in der Statistik (Bad
Godesberg, Institut für Raumforschung, 1959).

FARM HOUSEHOLD SURVEY IN GERMANY

available. Then the farms selected can offer a picture of the possible
variants in each of the areas.
Methods and Materials Used in the Investigation
Owing to the complicated pattern of income and structure of consumption in agricultural families, such an investigation requires a
detailed and specific method. Reliable data can only be obtained by
means of observations over at least one year, and trained personnel
must constantly supervise operations. It was found in practice that mere
questioning was quite insufficient, since the respondents were frequently
unsure of the values involved. This uncertainty extends both to income
and to consumption. In particular, there is often only a vague conception
of the value of own produce consumed, which is so characteristic of an
agricultural household. For all the above reasons, observation covering
a whole year was considered indispensable.
For each farm and household the following items of documentation
were kept:
(a) a farm ledger;
(b) a household ledger;
(c) ajournai of cash and material operations;
(d) ajournai of consumption in kind from farm production.
The first and second ledgers were filled out by the research workers in
co-operation with the farmers, whereas the journals were kept in the
farms and regularly checked by the research workers.
(a) The farm ledger contained indications of the size of the farm, the
farming system (crops and cattle rearing), the number of workers,
machinery, etc. It enabled complete reconstruction of the farm's
organisation to be made.
(b) The household ledger indicated the size and composition of the
family, the number of persons not belonging to the family living in the
household, and the economic activities of individual family members
both inside and outside the farm.
(c) As the journals were kept by the farmers themselves, entries had
to be as simple as possible. Cash and material operations were entered in
chronological order and according to quantity and value, but without
any attempt at following accounting methods. There was no separation
of farming expenses from household expenses. Some difficulty arose
when accounts with co-operatives, banks and merchants were dealt with
at long intervals only. In no case was admission to check the relevant
documents refused.

FAMILY LIVING STUDIES

(d) The journal for consumption in kind from the farm's production
noted quantities without valuing them. In addition, the number of
persons fully or partly maintained was noted for each day.
One major obstacle occurred in the fact that most of the farmers
covered had never kept such notes before. It was found, however, that
if the research workers checked entries regularly useful material was
acquired, on the basis of a simple scheme that merely provided for all
possibilities and did not require any complicated reflection. The research
worker responsible in each case undertook all consolidation of accounts,
transfer to accounts, and statistics of farming and bookkeeping operations, evaluation and arrangement of relevant data, such as income in
kind and manpower used.
Inevitably any method contains certain defects. The drawback with
a system of observation lasting over a year is that trends are difficult to
ascertain as there is no possibility of vertical comparison. However, if a
uniform method of assessing accounts relating to a given year is adopted,
no major inaccuracies can occur in a horizontal comparison. Several
research workers observed one fact that seems worth communicating. In
many cases, farmers were considerably more frank with them than with
representatives of schools and the advisory service. The certainty that
the research worker will be leaving the district after a year is very important, as anonymity seems more assured. The particular effect of this
was that information concerning exceptional income, or even debts, was
more willingly vouchsafed.
Other advantages of the method employed are obvious. As a result
of regular checking, it is possible to appreciate causal connections that
remained hidden to a brief survey. By and large, the one-year procedure
seems justifiable, notwithstanding its comparative expensiveness. What
was here involved was in fact a form of basic research, as there were no
previous indications of the values to be obtained. Strict uniformity was
maintained in notes and analyses. Although uniform methods of approach and forms were used, cases occasionally occurred that could not
be covered in the manner adopted. These questions of detail were
cleared up at meetings of the research workers aimed at a general coordination of the study. The frequent exchange of ideas eliminated the
danger of recordmg particular biases, a danger which had already been
considerably reduced thanks to the tight co-ordination. Thus it may be
said that the observations went beyond ordinary bookkeeping in
extent and in attention to detail, especially with respect to consumption
particulars.

FARM HOUSEHOLD SURVEY IN GERMANY

METHODS OF ANALYSIS

Reference Values
Evaluation of manpower poses certain problems, owing to the
considerable dovetailing of household and farm and the variable participation by family members in farming, household and outside activities.
Moreover, schematic evaluation encounters the difficulty that persons
working in family undertakings are frequently not full workers, as in the
case of aged persons, children and physically handicapped persons.
Since income and eventual consumption must be seen in relation to
productivity, the work of these persons must be converted in terms of
the work of full workers. Some general system had to be adopted here,
and the work of persons over 65 years of age was reckoned at 50 per
cent, ofthat of full workers, and the work of persons under 16 at 30 per
cent. Women's work is divided between household and farm, so that
women cannot be assessed on the same basis as full workers either.
The method followed was to deduct from the over-all labour input of
women workers 18 per cent, of a full worker's contribution in respect of
each family member fully maintained in the household. Such work in
respect of persons other than family members living in the household
(agricultural workers) was regarded as farm work. This basic system
was modified by the research workers as a result of their observations
over a whole year.
The reference unit taken for data on income and household expenditure was a person fully maintained, by which was meant a person
living in the household throughout the year and fully maintained. One
peculiarity occurring in such investigations in agricultural households
lies in the treatment of farm personnel living in the farmer's household.
This form of expenditure is regarded as a cost factor. Since expenditure
relating to meals for farm workers cannot be listed separately where
meals are prepared for everyone together, adjustment must be made
accordingly. The expenditure per person fully maintained results from
the quotient of total expenditure (food) divided by number of persons
fully maintained. The product of multiplying expenditure per person
fully maintained by the number of non-family members fully maintained
then gives the sum by which the over-all account must be adjusted.
Evaluation of Income in Kind
Any agricultural income consists of elements in cash and in kind.
Assessment of consumption in kind is therefore important both for

FAMILY LIVING STUDIES

determination of the level of income and for evaluation of consumption.
Determination of value poses particular problems. In the evaluation of
consumption in kind by the household from farm production, either the
particular seasonal farm prices or, in their absence, the average producer
prices at the nearest central wholesale market were adopted. All previous
expenditure (fertiliser, seeds, equipment, etc.) was included as outgoings
in calculating gross income. Other outlays connected with production of
the items self-consumed appeared as household costs. However, the time
spent by the family in that connection was not included, so that the prices
applied to consumption from own farm production (foodstuffs, fuel)
were lower than normal retail prices to the extent of part of the manufacturing costs and all the distribution costs. Rental values were
determined according to rates applying in the locality.1 Actual consumption in kind from farm production in the course of a year is shown
as income. The corresponding item of household expenditure is reckoned
from such consumption + changes in the level of household stocks.
Presentation of Results
In evaluation of results, the purpose of the whole study must be borne
in mind. This called for a double procedure, one aspect of which was to
classify farms according to the individual areas of investigation. As one
of the purposes of the work was to investigate the connection between
farming results, standard of living and economic and social factors of a
regional character, a short preliminary description of the principal
characteristics of the area was made in each case. By means of this
combination of farming data with evaluation of other relevant social
and economic facts, it was possible to evaluate the general situation of
farms and families from the point of view of causal connections.
Despite certain limitations, it was possible also to give average values
and dispersions. The essential point in calculating averages was to take
into account the circumstances of the case and to follow procedures that
would not overlook essential facts. One such essential fact is the variability of results even where the external circumstances are comparable.
Derivation of the mode therefore seemed most indicated. The crop
system and the size of the farm were selected as basic categories ; these
give a good idea of the internal economic size of the farm. For every
category of farms in which the crop system is the same, one can draw up
a system of co-ordinates and plot the total amount of expenditures for
1
Regarding methods of evaluating produce consumption in kind and use of
housing, see also " Die Lebenshaltung von Bauern und Landarbeitern ", in Wirtschaft
und Statistik, op. cit., No. 10, Oct. 1957, pp. 514 ff.

FARM HOUSEHOLD SURVEY IN GERMANY

each household (on the %-axis) in terms of the absolute size of the farm
(on the j-axis, by means of a logarithmic scale). Dispersion or concentration of the values thus plotted leads to positive and negative
correlations providing valuable indications of the various bases for the
determination of income formation. This makes it possible, in principle,
to make certain general statements independently of the regional
structure.

Points of Emphasis in Analysis Illustrating Method Used
The purpose of this discussion of method is not to give the results of
the study. In order to prove its effectiveness, a few points of emphasis
in analysis will be indicated.
Level and Composition of Income as a Typological Criterion.
The social and economic situation of a large proportion of the
agricultural population is not necessarily identical with income from
farm production. A large part of family income may originate elsewhere.
It is easy to go wrong if income from farms is regarded as the only
element deciding possible living standards. The total income of a household may be composed of various elements besides direct income from
farming. The results of observation made throughout the study reveal that
the structure of the different income elements varies very widely. The
fact emerges that an unsatisfactory income from farm production does
not necessarily mean that the household income is insufficient. If there
are sufficient opportunities for outside earning and these are used by
family members, income can be correspondingly increased. The combination of the various sources of income, then, provides the family
with a full basis of existence. The borderline of purely agricultural
activities can be relatively clearly defined. It may be seen what internal
economic size of farm is required in any particular circumstances to
provide a suitable income from agriculture. There are also various
households that must be ranked among the marginal agricultural enterprises, since they are so small that the working capacity of the family
cannot be properly utilised. In such cases no appropriate income can be
achieved. The criterion applied is the so-called comparable wage,
namely the income of wage earners living in rural districts. This comparable wage was established at 4,500 DM a year for full male workers
and 2,700 DM for women. Quite a considerable proportion of the farms
covered did not reach this level of income.

100

FAMILY LIVING STUDIES

Thus, four groups of households working the land may be distinguished :
(1) full agricultural existence (sufficient income from agriculture);
(2) partly agricultural existence (agriculture and additional source of
earnings) ;
(3) marginal agricultural existence (insufficient income from agriculture
alone) ;
(4) households with agricultural side-income (owner of farm engaged
primarily outside agriculture, sufficient income).
This classification makes several things clear. First of all, only a
portion of the 2 million farm units covered by federal statistics should
be considered as engaged in full-time agriculture. The situation of households with additional sources of earnings is frequently influenced more by
the general economic structure of the area than by the particular agricultural conditions, since it is the economic structure that decides what
alternative sources of earnings there are. The varying income situations
of farms have been discussed here in such detail because they affect both
the absolute level of the living standard and the inner structure of consumption. Thus, it is not possible to say how agricultural families live,
but only how certain types of households farming the land behave in
function of particular possibilities.
Cash Income and Living Standards.
The total income of the household per family member fully maintained
shows the income available for providing for the needs of all family
members. When comparison is made with the family incomes of wage
earners and salaried employees, it must be borne in mind that incomes in
agriculture are:
(a) frequently earned by more than one worker, generally two to three ;
(b) not regular in their receipt;
(c) made up on a larger scale of elements in kind, in the form of personal
consumption and capital increase through investment, than for wage
earners and salaried employees.
The last of these factors means that it is impossible to deduce the
probable living standard of a family from its total income. It can be
assessed, up to a certain point, on the basis of cash income per person
fully maintained. The sum of net cash incomes represents the amount
that can actually be consumed or saved by the family. The tabulation

101

FARM HOUSEHOLD SURVEY IN GERMANY

below gives the percentage distribution of households according to
annual cash income per person fully maintained. Two-thirds of all
Income group

Negative
Oto under 1,000 D M
1,000 „ 2,000 D M
2,000 „ 3,000 D M
3,000 D M and above

Percentage

11.2
36.0
30.5
9.1
9.2

households have an income of between 0 and 2,000 DM, and over onethird between 0 and 1,000 DM. Somewhat more than one-tenth have a
negative cash income, which means that the total of all expenditure in
cash exceeds the total of all income in cash. Cash expenditure for private
consumption must therefore be covered in practice from reserves or
credits, and debts therefore increase. Since personal consumption per
person fully maintained does not fluctuate greatly, cash income offers a
convenient reference for living standards of families.
Structure of Expenditure and Living Standards.
The basic idea of examining household expenditure from the point
of view of income is that the range and structure of private expenditure
are largely influenced by whatever means are available. If income groups
are formed according to data relating to farms (size, system of cultivation) or according to social and economic types of farms, these groups
reveal typical structures of expenditure in the household and standards of
living. According to the conception of our modern society, a proper
standard of living covers not only provision with the basic needs in the
form of food, clothes and housing, but also the supply of various items of
higher-level consumption and both material and non-material services.
Such items can generally only be obtained when there is a sufficient supply
of cash. The cash proportion of income is however very slight in many
farms. True, there is no extreme threat to the existence of the persons
concerned, as the biological minimum of existence seems always present.
Household income in cash may be used for direct consumption or for
saving. Comparison with the total of private consumption expenditure
thus shows whether there was enough to satisfy requirements and for
what specific purposes it has been used. Then, subject to reservation,
certain conclusions may be drawn regarding family living standards.
In the analysis, expenditure was therefore divided under convenient
headings in order to make it clear to what extent farms have a standard
of consumption comparable to that of urban families. Separate indications were given for expenditure relating to food, beverages, housing,

102

FAMILY LIVING STUDIES

domestic equipment, clothes, personal care, education, entertainment
and transport. The form of reporting also allowed for further subdivision,
with regard to food for instance. This in turn permits conclusions to be
drawn with regard to nutrition problems.
It is particularly noteworthy that the proportion of expenditure in
cash to expenditure in kind increases as total expenditure rises. The
reason is that the portions in kind are drawn predominantly from food
production from the farm. Irrespective of the size of the farm, the
intensity of farming or the area investigated, the value of such consumption generally stands—except in a few extremes—between the narrow
limits of 400 and 500 DM for individual households. Consumption of
food produced on the farm is not normally expansible beyond this point.
A rise in total consumption consequently reflects almost exclusively
an increased household expenditure in cash. For the average of all
farms, food bought outside represents about a quarter of the total
expenditure. The bigger the farm the more is spent on food, owing to the
demand for higher quality, but the increase becomes relatively smaller.
Higher expenditure in cash for food also affects the daily costs of the
catering unit. It is particularly in areas with healthy farm structures or
better earning prospects that figures well over 2 DM are to be found.
On the whole, there is a clear shift of emphasis to consumption of other
goods and services, particularly the less basic items.
Undoubtedly, many farmers are still ready to accept the disadvantages
that go together with the peculiar character of income formation in
agriculture in order to have their independence and other partly nonmaterial advantages attaching to their occupation. However, there can
be no doubt either that the consumption preferences of many farmers
are increasingly influenced by comparison with the consumption habits
of industrial wage earners, salaried employees and other economically
active groups living in the vicinity.
As has already been explained, the purpose of this methodological
contribution is not to present the results of the study. The aim has been to
show that such a study, restricted to ascertaining income and consumption, can provide information of far-reaching significance. It shows
how far agricultural families are participating in a consumer society
that is steadily expanding.

VI
International Comparison :
Budget Survey Made by the E.C.S.C.
By Rolf

WAGENFÜHR

INTRODUCTION

The Treaty setting up the European Coal and Steel Community
declares in several places that the raising of the standard of living and
the " equalisation through progress " of the economies concerned, or
at least of partial areas of these economies in the process of integration,
constitutes a task of primary importance. Consequently, economic and
statistical research had and still have a double objective:
(a) to evaluate the standard of living at the outset of this integration ;
(b) to ascertain whether or not the sought-after and proclaimed " equalisation through progress " is being achieved.
During the early years of integration, which began in 1952 with
the mining and ironworking industries, and even today, the first of
these objectives has received priority. In this context the importance
will be realised of the survey of family budgets carried out in 1956-57
by the E.C.S.C. on 2,000 working-class families in the six countries
concerned. To our knowledge this survey was one of the first, if not
the first, to take place in several countries simultaneously, under uniform
conditions, and it may be possible to draw some general conclusions
that could be of interest in the international comparison of family
budgets.
Attention should be drawn to the scientific basis of this survey,
which consisted, broadly speaking, in determining the standard of living
of workers in the mining and ironworking industries in the process of
* Director-General, Statistical Office of the European Communities.

104

FAMILY LIVING STUDIES

integration in the six countries. It was clear from the start that there
was not and could not be a single uniform yardstick for measuring the
standard of living; the aim was rather to establish a certain number of
qualitative and quantitative indices which, by mutually complementing
each other, would after a few years enable fairly reliable answers to be
given to the questions posed.
The choice of a starting-point for a study on the standard of living
is contentious. In the concrete example considered here the best solution
was to begin by inquiring into earned income. Coal mining and ironworking (and to a lesser degree, iron mining) are heavily concentrated
industries forming a relatively complete entity which, production
techniques being to an appreciable extent the same in the six countries,
has given rise to more or less uniform grades of workers. Furthermore,
the high level of concentration made it possible to carry out a searching
inquiry in which certain special problems arising out of the use of
sampling methods were avoided. On the basis of a uniform definition
of the branches of industry concerned, a uniform definition of the
term " worker " and a uniform conception of " income ", it has been
practicable, starting in 1953, to obtain data on average annual earned
income in coal mining, ironworking and iron mining in the six countries.1 Since then this inquiry has become a kind of routine statistical
compilation, but obtaining the nominal amounts was naturally only a
first step. To convert the annual amounts expressed in the various
national currencies into comparable amounts of purchasing power it
was necessary to fix special monetary equivalents with respect to consumption 2, which called for a double series of basic inquiries.
On the one hand, through personal visits to some 2,000 shops
note was taken of the prices of more than 200 articles (very precisely
defined). As the description of these articles was quite comprehensive
and the work has always been carried out by the same group of investigators, there is no doubt that this procedure has resulted in a unique
documentation concerning prices, which could be used in international
comparisons. An inquiry of this sort was carried out for the first time
in the autumn of 1954 in collaboration with the statistical offices of
the six countries; it was repeated in the autumn of 1958, and it is planned
1
The term " earned income " does not correspond exactly to the concept of
income as used in the survey, for on the one hand family allowances are added and
on the other hand direct taxes and social security contributions paid by the worker
are deducted. For simplicity's sake the expression " earned income " will continue to
be used in the pages that follow.
2
See "Revenus réels C.E.C.A. 1954-1958", in Informations statistiques, serie
" Statistiques sociales " (Luxembourg, Office statistique des communautés européennes), No. 2, 1960.

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

105

to make a further inquiry in a few years' time, if financial resources
permit and if sufficiently large price fluctuations are noted.
The second problem requiring at least a provisional solution concerned price weighting and price relations. This is where the question
arises of consumption structure, and hence of family budgets ; for it
is in fact the latter that are the most appropriate for obtaining the
maximum of information on the subject, including, as is eminently
desirable, not only data expressed in monetary terms but also details
of quantities consumed. When the results of the 1954 price inquiry
were known it became rapidly clear that the data available on family
budgets was completely inadequate. Statistics had never been compiled
on family budgets in Luxembourg; in Belgium there had been no official
survey since 1927-28, apart from the post-war inquiries of 1947, covering
a period during which reconstruction was in full swing. In France
partial surveys had been carried out in some coal-mining areas, but it
was not possible to use the results of these surveys because there was
no separate data for ironworkers or iron miners. For the Federal
Republic of Germany, Italy, and the Netherlands, budget survey results
(some of them unpublished) were more significant, and quite recent
in date. But very serious difficulties arose out of the differences in the
starting points of these surveys and in the survey design and in the
classifications used; furthermore it was not possible to separate the
categories of workers in the three branches of industry mentioned above.
While in the analysis of the 1954 price inquiry use had to be made
of these incomplete data on budgets, in the absence of other possibilities
—long experience has shown that in international price comparisons
statistical data on prices were much more important than weighting
coefficients—the need for compiling statistics on family budgets in the
six countries on a long-term basis became clearly apparent. These
statistics, conceived and compiled in such a way as to permit comparisons
among the six countries, would provide the basic information for better
weighting of data concerning prices and thus provide a firmer foundation
for the calculation of real earned income.
This method of procedure—establishing statistics on family budgets
to obtain weighting coefficients—represented no radical departure from
the practice followed in the various countries. For the E.C.S.C. the
question was much more vital, for it was necessary to obtain the weights
simultaneously and on a comparable basis for all six countries.
This bringing into focus of the problem led to considerable simplification, which was the more desirable since statistics on family budgets
generally involve heavy financial outlay and the High Authority of the
E.C.S.C. was obliged to pay subsidies to the national statistical offices.

106

FAMILY LIVING STUDIES

Simplification consisted essentially in limiting the range of families
investigated and adapting it to the particular purpose of the inquiry.
The real annual earned income had to be determined, starting with the
amounts representing the average earnings of workers in the three
industries in each of the six countries. It was therefore decided not to
consider all families of workers in these industries, but only those
belonging to the modal income group (with a margin for error of ± 15
to 20 per cent.), the meaning of " income " here being the income earned
by the head of the family in one of the three industries. At the same time
the following rules were laid down :
(a) Only the head of the family should be in employment.
(b) The family should consist of four persons (parents and two children
of not more than 13 to 14 years of age).
(c) The head of the family should be in a position to return to his
home and family every evening.
(d) The worker should have the nationality of the country in which
he was working.
Condition (a) is explained by the special purpose of the survey;
for financial reasons, (b) restricted still more the scope of the survey,
but for practical purposes the " average " family structure was thereby
still kept; (c) and (d) were to rule out exceptional cases entailing a
radical alteration of the consumption structure; these cases were better
reserved for special inquiries later on.
It was within this framework that the survey had to be carried out
with uniformity of concepts, definitions, classifications and, as far as
possible, methods of investigation.
PREPARATION AND ORGANISATION OF THE SURVEY

In the matter of international comparative statistics such as those
compiled by the High Authority, care is always taken to observe the
principle of constant and close collaboration with all those concerned.
The principal means of achieving this is to set up committees of experts,
to be split up as the need arises into subcommittees, working parties
and where possible even into " national subcommittees ", meeting in all
cases under the chairmanship of an official of the High Authority.
The guiding principles of the family budget survey were laid down
by the Committee on Methods of Comparison of Real Wages. Employers, workers, and the competent ministries and statistical offices
of the six countries were represented on this committee. Since, in view
of its size (it had 50 members), this committee could do no more than

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

107

touch on the broad lines of the problem, a Family Budgets Subcommittee
was set up of similar composition but with the number of members
limited to 28. This subcommittee supplied the first concrete formulations
of the problem. Finally, the technical planning of the project was
entrusted to a working party consisting of two members of the statistical
office of each country. The members of this working party set up
" national subcommittees ", in order to have a wider discussion of the
problems arising in their respective countries.
This preparatory stage lasted about six months. It is impossible
here to give a detailed account of the work accomplished; we may,
however, mention briefly some important points.
The comparison aimed at should not confine itself to a comparison
of countries, but should allow for a comparison of areas. In the coalmining industry the areas are delineated quite naturally by the concentration of the coalfields. In the case of the other two branches of industry,
the regional divisions, as used in the statistics on occupational groups,
were taken as a basis. The problem of the regional structure of a limited
number of branches of industry (as was the case here) is naturally much
simpler than in cases where it is necessary to compare complete national
economies. As a result, there were found to be nine coal-mining areas,
six iron-mining areas and thirteen ironworking districts. This regional
comparison was later found to be of enormous value, particularly as it
provided fresh insight into consumption habits.
A regional breakdown would serve no useful purpose unless at least
50 families were studied per region or per industry or per category of
workers. Bearing in mind that from the beginning a relatively homogeneous group of families was studied, attention was thus given to the
so-called elementary rules of taking random samples.
After long discussions preference was given to the method of household account books, to be kept for one year. Since the primary aim
was to determine weighting factors with a view to price comparison,
it was important to obtain very detailed data, which were most likely
to be obtained by the method of day-by-day entries.1
To make up for possible defections, provision was made from the
beginning for a reserve of families, equal to some 25 per cent, of the
total number, to be included among the participants.
The choice of the account book method—and the desire that these
accounts should be kept continuously for 12 months—excluded in
1
Depending on the country, the household account book covered one week, two
weeks or a month. There was a page for each day, on which were to be noted all
expenses, etc. The account book itself contained no printed list of goods, to spare
housewives the trouble of needless searching.

108

FAMILY LIVING STUDIES

advance, in view of the limited financial resources available, the putting
into practice of the pure random sampling method. Even so, every
effort was made to take into account the element of chance, as may be
seen from the example of Belgium.
In Belgium the undertakings communicated to the National Statistical Institute the names and addresses of workers corresponding closely
to the criteria of selection. From this list 1,835 families were chosen at
random and a letter was sent to each one inviting them to participate in
the survey. Six hundred and eighty replied in the affirmative. These
680 families were then visited by 25 investigators with the object of
obtaining supplementary information with respect to the criteria required.
Four hundred and ninety-two families fulfilled these criteria.
Four hundred of these families were retained for the survey proper,
lasting 12 months, 140 families being considered as " reserve families ".
The same method was applied, in the main, by the statistical offices
of France, Italy and the Netherlands. In the Federal Republic of Germany
the selection was made in a slightly different manner: it was not the
undertakings that supplied the lists of nominations; these were drawn
up on the basis of statements by the families themselves. In the Saar and
Luxembourg use could not be made of any form of chance selection
whatsoever, the number of families available being too small.
It is obvious that the method of selection described above does not
satisfy all requirements. It must nevertheless be borne in mind that this
was the first joint survey of this type in the six countries. On the occasion
of the subsequent survey for the more simple purpose of studying
workers' housing conditions, carried out in 1958 by the interview
method, it was possible for the first time to carry out actual random
sampling in the six countries.
Finally, it v/as important to ensure permanent contact between the
statistical offices charged with carrying out the survey and the families
investigated, taking into account national pecuharities. In the Saar
and Luxembourg it was necessary to be content with contact by letter
with the 170 or so families in each of these two countries. In the Federal
Republic of Germany these written contacts (to which many families
were accustomed as a result of previous surveys) were pursued parallel
with meetings between the families and representatives of the statistical
office, in the course of which doubtful issues were cleared up informally.
The most fruitful system was the investigator system adopted in Belgium
(25 investigators, two visits per month), France (17 investigators, two
visits per month) and Italy (62 investigators, two visits per week). If
such a survey is undertaken again an effort should be made to extend
this system to all the participating countries.

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

109

Broadly speaking, the six-country budget survey lasted from May
1956 to April 1957: for technical reasons the survey could not begin in
Belgium until 4 June 1956, and in Italy until 1 July 1956, but there, too,
it lasted the full 12 months. The results obtained were processed on a
regional basis by the statistical offices and sent to the Statistical Division
of the High Authority, so that use could be made of them at the central
level. The first publication by the High Authority took place in October
1958 in Informations statistiques ; the final results have been published
in the four languages of the Community in the spring of I960.1
SPECIAL FEATURES OF THE SURVEY

Information about the Family
First, certain facts about the composition of the participating families
were recorded. With few exceptions, the families were of the size desired,
permitting consumption to be determined very simply per " unit of
consumption " (approximately 2.4 units of consumption per family).
The age of each member of the family was recorded, as were facts about
the housing accommodations (ownership, presence of garden, poultry,
etc.). The participants were questioned about the presence of visitors
and about the periods of absence of members of the family; from their
replies it would seem that the presence of visitors and the absence of
members of the family cancelled one another to an appreciable degree,
which was not always as predicted by the experts. If another survey
of this kind is undertaken later, undoubtedly this question will no
longer be asked.
Recording of Receipts
Particular attention was paid to the recording of receipts. The
taking into account of receipts was extremely useful. For one thing,
certain details of the structure of family incomes could be obtained
which it would not have been possible to ascertain by any other method;
moreover, the recording of receipts served as a check on expenditure and
consumption. A systematic collection was made of all possible forms
of receipts, which were then arranged according to a standard classification. The difficulties involved in this were relatively slight, since the
problems arising out of definitions and delimitations had already been
solved in respect of the most important item, the earned income, at the
time of the compilation of the statistics on nominal earned income.
1
" Budgets familiaux des ouvriers de la C.E.C.A. 1956-57", in
statistiques, série " Statistiques sociales ", No. 1, 1960.

Informations

110

FAMILY LIVING STUDIES

Basic and Supplementary Income.
A distinction was made between the basic income and supplementary
income of the family, supplementary or " casual " income comprising,
inter alia, social security benefits, gifts and assistance from third parties,
the value of produce from the garden or poultry run, income from real
estate, subletting, occasional earnings by the wife, etc. Subsequent
analysis confirmed the importance of such a distinction, for in the
various countries and industries the proportion of " casual income "
fluctuated between 4 and 15 per cent, of total income.
Income in Cash and in Kind.
Another distinction was between cash income and income in kind.
Since it had been decided to investigate not only expenditure but also
consumption, it was necessary to make provision for corresponding
items on the receipt side.
Income in kind consisted in substance of benefits in kind provided
by the employer (delivery of coal to employees, the use of accommodations); also to be considered were the value of the produce from the
garden or poultry run and the rentable value of the house or dwelling
if owned by the family. Here, too, the results of the survey show the
importance of these questions, particularly for comparisons between
countries. By distinguishing between branches of industry and between
countries it was possible to establish that " income in kind " varied
between 3 and 14.5 per cent, of the total receipts of the families concerned.
Collection of Data on Receipts.
Numerous experts objected, on general principle, to questioning
families about their receipts and income, as this involved " too much
meddling in people's private lives ". Contrary to all expectations, these
objections were easily overcome. True, it was necessary to take account
of differences in conditions (and reactions) in the various countries. The
most important basic datum, the earned income of the head of the
family, was given in some countries by the employees themselves, who
attached to their household account book the pay slip from their firm.
In other countries this information was supplied, with the consent of
those concerned, on a comprehensive basis by the employers at the end
of the survey. This shows once again how important it is to gain the
support of all concerned for collaboration based on mutual trust.
Evaluation of Benefits in Kind.
A difficulty, already well known in connection with national surveys,
arose out of the evaluation of benefits in kind. In France the rental

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

111

value of accommodation was estimated in accordance with the 1948
legislation, while elsewhere it was based on the official surveyor's evaluation, adjusted to take account of price trends. In some cases an estimate
could be made on the basis of the rateable value, and where this method
in its turn proved impracticable it was possible in certain countries to
make use of rent statistics. The other items of income in kind were
calculated according to uniform methods in the six countries by the
statistical offices concerned.
Evaluation of Social Security Benefits in Kind.
Difficulties also came to light in connection with the recording and
estimating of social security benefits in kind, for as everyone knows, the
social security systems of the six E.C.S.C. countries are very different.
Family allowances could be calculated quite easily by referring to the
statutory provisions governing them. In the case of sickness insurance
benefits it was found necessary to use the figures of the employers' and
workers' contributions. 1 This hardly satisfactory solution was necessary
because in certain countries the medical and pharmaceutical expenses
incurred by the sick worker are refunded to him, except for a nominal
charge; while in other countries he does not have to pay or contribute
anything. But the social security funds in the various countries were not
in a position to verify and evaluate their benefits in kind.
Credit Accounts.
The movements of credit accounts, for which detailed provision was
originally made, were cut out at the end of the survey. Only the balance
was recorded.
Recording of Expenditure
With regard to expenditure, it was possible to draw up a standard
classification for all six countries, comprising 14 large categories, 77
groups and 364 items. This classification, inspired by those in use in
the six countries of the Community, was reviewed and analysed throughout the survey and during the processing of the results ; for this purpose
the written method was used, and meetings of the aforementioned
working party of the statistical offices were held from time to time.
1

Thus, expsnditure on medical attention was assumed to be represented by
workers' health insurance contributions and by the amounts paid by families to
sickness insurance schemes outside the national social security systems. For countries
and industries in which workers initially paid for their own medical attention, the
amounts subsequently refunded them were deducted from their expenditure. The
figures needed to establish these refunded amounts were supplied by the social security
authorities.

112

FAMILY LIVING STUDIES

During the processing this classification was shown to be highly useful
and advantageous. Some participating countries retained their national
classification, in addition, so as to ensure continuity with previous
surveys.
In addition to expenditure on consumption properly so called, social
security contributions, apart from sickness fund contributions, and
taxes paid by the workers were recorded separately, and these were
included in total expenditure.
Let us mention briefly three special factors that presented difficulties
because the items involved were recorded under the heading of expenditure : pocket money for the head of the family, the entering of expenditure on holidays and outings and the manner of recording goods bought
on hire-purchase. These difficulties are inherent in national surveys,
consequently, it was even more important for the six-country survey to
draw up common rules, which, it is true, could only be achieved by
compromise.
Pocket Money.
For the recording of pocket-money expenditures, each head of the
family was given a special notebook, which he could send directly to the
office carrying out the survey if he wished to keep the knowledge of its
contents from the rest of the family. These notebooks, which were
used in all of the countries with the exception of France, provided detailed
information on expenditure that enabled the offices to work out a
breakdown of expenditure in cases where only a total figure for pocket
money was given. (In France this task was performed by the investigators
after interviewing the heads of families.) There is no doubt that this
method of procedure made it possible, in this international comparison,
to prevent, at least in part, a gap which is sometimes criticised in the
case of the survey method based on household account books.
Holidays and Outings.
With respect to expenditure on holidays and outings, many families
•detailed such expenses normally, thus avoiding all difficulty in interpretation. Where a lump sum was given for expenditure in this category,
a breakdown into four groups—accommodation, food and drink,
transport, and entertainment—was made at the end of the survey by
the statistical offices.
Hire-purchases.
The problem of how to enter goods bought on hire-purchase gave
rise to laborious discussions at the working-party meetings; but agree-

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

113

ment on a uniform method was finally reached. This consisted in listing
as expenditure only sums actually paid out. The total amount, including
sums still owing, was taken into consideration fully in the calculation
of consumption. The subsequent processing established the value of this
method. (Moreover, 60 to 80 per cent, of the contracts for sale on hirepurchase entered into during the year of the survey were liquidated
within the same year.)
Expenditure on Consumption and Actual Consumption
The fact that the joint six-country survey covered not only expenditure on consumption, but also actual consumption itself, gave this survey
particular importance. The results subsequently obtained permitted it
to be clearly recognised that differences in actual consumption between
countries were considerably smaller than the differences in expenditure
on consumption, both in total amounts and in relative percentages. The
difference between expenditure on consumption and actual consumption
lay primarily in four items: foodstuffs (produce from the garden or
poultry run), heating (supply of coal to employees), housing (free
accommodation) and goods bought on hire-purchase (figuring under
expenditure only in respect of instalments already paid, while in the
case of actual consumption these goods were included to their total
value; in most cases the goods concerned were electrical appliances,
textiles and means of transport). In our own experience an international
comparison, to be worth while, must be based on actual consumption
and not on expenditure—a fact to which far too little importance is
attached in international discussions.1
Finally, it should be mentioned that in respect of the consumption
of foodstuffs and a limited number of other goods, the survey also
covered the quantity consumed.

UTILISATION OF THE RESULTS

Determination of Real Income
As has already been stated above, the results of the family budget
survey were directly and essentially used for the purpose of determining
real earned income. The results served two purposes:
1
In the extreme case—the Italian coal-mining industry—the recorded expenditure
on consumption by 50 miners' families amounted to only 81 per cent, of actual consumption; the highest was achieved by workers in the German ironworking industry
(97.4 per cent.).

114

FAMILY LIVING STUDIES

(1) They enabled a check to be made on the realistic character of
the list of items drawn up for the price inquiry in 1954. It was found
that many products, such as plastic household articles, had come to
occupy a new and important place in the consumption tally, while other
products could be eliminated in view of their slight importance. As a
result, the price inquiry made in the autumn of 1958 could be launched
with an improved list of items.
(2) The tables of expenditure and consumption made it possible
to obtain the weights needed to establish monetary equivalents with
respect to consumption. In this connection one could adopt the traditional procedure of national " baskets " of goods and services (with
possible consecutive cross-Unking of results) or the Bogers-van Ijzeren
method1, which attempts to establish by successive approximations a
common " basket " for the six countries.
The choice of weight is a problem that calls for special study. Consumption statistics give a much wider picture than expenditure statistics.
Furthermore, the monetary equivalents of consumption should be
applied in the conversion of nominal earned income which, while
including certain elements of income in kind (rentable value of accommodation, supply of coal to employees) does not embrace all the elements
involved, nor does it include supplementary income. The first question
to be decided is whether the choice of different systems of weighting
leads to results very far removed from one another.
Price and Consumption Structure
One consideration closely bound up with what has been said concerns
relations between price structure and consumption structure. The prices
on which the on-the-spot inquiries carried out in 1954 and 1958 were
based referred to items decided on in advance and generally held to be
the most popular (only in rare cases was it necessary to modify the standard as to quality during the price inquiry). Now, it is certainly worth
asking oneself if the goods covered by the price inquiry correspond,
from the point of view of quality, to those actually consumed by workers'
families. One cannot say, yet, whether it is possible to compare recorded
price (in a price inquiry) and price calculated on the basis of a survey
of family budgets. The quantitative data available are derived essentially
from consumption estimates (and not from expenditure estimates); the
" average prices " worked out on the basis of the survey involve estimates
of consumption of home production. Here again, it will be necessary to
analyse the interpretations placed on these calculated "average" prices.
1

See " Revenus reels C.E.C.A. 1954-1958 ", op. cit., pp. 60 ff.

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

115

Other Uses
So much for the direct utilisation of the results of the family budget
survey in the determination of real earned income. But the importance
of these results has so far not been made clear. A considerable number
of new investigations are planned, but here we are forced to content
ourselves with a few schematic observations. Take, for instance, elasticity
of demand, " consumption habits ", and, finally, the possible information
making it practicable, starting with the earned income, to calculate the
total income, not only of the worker but of his family, in so far as it is
an economic and consumption unit. As a final point we shall mention
the degree of diversity in the quantities consumed.
The problem of elasticity of demand has been discussed in detail in
literature ; it is hoped that the results of the survey will make it possible
to isolate a group of " fixed needs ", and departing from this point to
discover new methods for determining and comparing real income.
The estimates already available today provide confirmation of the
laws of Engel and Schwabe; however, there are a few slight anomalies
that call for an explanation. The international comparison of data on
family budgets cannot be brought to a conclusion until the monetary
equivalents with reference to consumption equivalents have been established down to the last detail. It is already possible in each country to
make an inter-regional comparison of the results of the family budget
survey, revealing clearly that income on its own does not always form
an adequate basis for the explanation of a certain level of consumption:
the occupation (that is to say, the arduousness of the work) and consumption habits in some cases play an important part. This is why in
spite of the fact that their earnings are higher than those of their colleagues
in the ironworking industry, the iron miners of Luxembourg spend more
on foodstuffs, both in total and in comparative figures. As the income
rises, the composition of the diet does not always swing towards the
dearer items of food (such as meat), and even the relation between size
of income and expenditure on housing does not always correspond to
the law of Schwabe. The research in this field, with the introduction of
qualitative elements into international standard-of-living comparisons,
enters directly into the field of the study of behaviour.
The study of receipts in family budgets constitutes an important aid
in the comparison of real earned income. While it may be true that in
the case of this survey the experts are aware that it is not possible to
arrive at exact conclusions as regards the total amount of income,
particularly since the survey took into consideration only families where
there was only one earned income, it is no less a fact that certain structural

116

FAMILY LIVING STUDIES

elements stand out very clearly (relation between cash income and total
income; relation between basic income and casual income; relation
between earned income and family income). These relations are often
extremely important in cases where the earned income is comparatively
small. The conclusions of the survey on real earnings carried out by
the High Authority, according to which the differences between the
incomes of mineworkers and ironworkers in different countries are
much slighter than is generally acknowledged, and less marked also than
the differences between incomes in regions within a single economy, has
been backed up by the results of the budget survey. This is one of the
most vital subjects for investigation in comparative statistics on family
budgets on an international scale.
This section should not be concluded without a word of warning.
Actual comparison between countries—and even between areas within
a single country—has once again provided evidence of the extraordinary
diversities which must be taken into account in international comparisons
of consumption statistics. The traditional summary groups (for example,
foodstuffs expressed in calories; meat and fish) merely camouflage the
differences.1 In this respect, again, international surveys on family
budgets have a unique value.
BUDGET SURVEYS AND THE STANDARD OF LIVING

Thus we come back to our point of departure,°the original problem
of measuring the standard of living. It is possible that family budget
statistics can contribute to a certain degree towards solving this problem,
but only to a limited extent, so that new and different statistical surveys
will be necessary. In consequence, we propose to outline here some of
the projects of the Statistical Division of the High Authority which either
have already been carried out or are planned for the more or less distant
future. In the case of some plans it is not in fact known whether they
will be able to be put into action. But that is not the point at issue:
it is much more important to show the thread of ideas behind these
investigations.
1
In fact, for a country-by-country comparison, expenditure should be so broken
down as to give the full value to products of importance in the national pattern of
expenditure. The breakdown used in this particular inquiry, for instance, brought out
the special importance of a number of foodstuffs, and it was established that expenditure on and consumption of such items as spaghetti, macaroni, margarine, olive oil,
milk, tomatoes, wine, beer, coffee and tea represented very different proportions of
the total food packet of the families interviewed in the different countries. Thus the
Italian miner's family's annual consumption of spaghetti, macaroni and so on worked
out at approximately 153 kg. as against the Belgian family's 5 kg. and the German's
8 kg. Annual consumption of milk by Dutch miners' families averaged about 598
litres, and by Italians only 119 litres.

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

117

First of all, concerning what we consider the most important aspect
of the standard of living—the earned income—it is clear that the average
figures available (net annual income per worker present or per worker
registered) are far from presenting an ideal statistical picture. On the
one hand, social security benefits are not included in the income of
registered workers. For this reason a special survey was planned,
covering, to begin with, the coal-mining industry in the six countries,
with the object of remedying this defect. In addition—and the chances
of carrying out this project are unfortunately among the most doubtful—
it would be desirable to obtain, besides figures on average income,
special data permitting the determination of the relation between earned
income on the one hand and occupation, the status within that occupation, age, length of service, sex, conjugal status, etc., on the other.
A survey of this kind would have to be carried out by means of sampling,
but this project has not been welcomed with enthusiasm by the industries
directly concerned. A second possibility of comparing incomes in certain
occupations is at present being tested; attention should be drawn here
to the difficulties involved in grouping occupations in a comparable
manner in several countries simultaneously.
In the more limited sphere of family budget statistics it is hoped,
within the framework of the statistical programme of the High Authority,
to investigate once again the position of foreign workers and salaried
employees in the mining and ironworking industries. A few comments
will enable the reader to understand very rapidly why attention is being
concentrated on these two aspects in particular. Estimates of real
earnings are based on national consumption patterns, but when you
get down to it these patterns—as has been repeated on innumerable
occasions—are a fabrication. Taking as an example miners working
abroad (one thinks, of course, of Italian miners in Belgium), an effort
would be made to determine by means of interviews to what extent they
have adjusted to the consumption habits of the country in which they
are working. A survey for the purpose of international comparison will
be made of the families of salaried employees in the mining and ironworking industries, which should throw some light, for the first time, on
the standard of living of this important group. All previous attempts,
in particular the compilation of comparative statistics on salaries in
the six countries, have so far failed to be brought to a successful conclusion.
If a very wide interpretation is given to family budget statistics,
mention may also be made under this heading of other activities of the
Statistical Division of the High Authority: the vast 1958 survey on the
housing conditions of workers in the mining and ironworking industries,

118

FAMILY LIVING STUDIES

the study (so far still in the drawing-board stage) on the international
comparison of miners' pension schemes, and—tentatively envisaged on
a long-term basis—the survey on the organisation of leisure.
The housing conditions survey was the first real random sample
survey carried out by the High Authority with the collaboration of the
statistical offices of the six countries. This sample survey of 40,000 workers was based on uniform methods and definitions. Its purpose was to
supply information of use in connection with the financial aid given
by the High Authority for the construction of workers' housing. This
survey gives information at the same time on the housing conditions
of workers, the distance they have to travel to their workplace, etc. 1
The borderline between the " standard of living " and " living conditions " has undoubtedly been reached thereby—but as was stated at
the beginning of this article, it is necessary to find several yardsticks.
The housing survey yielded at the same time important data on the
structure as regards the age, family status, number of children, length
of service, etc., of employees in the mining and ironworking industries.
Finally, proceeding from a broad concept of the " standard of
living ", it is planned to carry out later an investigation into the organisation of leisure of workers.
FINAL OBSERVATIONS

While it is true that the above remarks concern a specific case—an
appraisal of the standard of living of workers in the mining and ironworking industries in the countries of the E.C.S.C.—it is nevertheless
possible to draw some conclusions of a general nature :
The international comparison of family budgets calls for common
definitions, classifications and methods. The survey has shown that it
is possible to achieve this.
Family budget statistics compiled with a view to international comparison make it possible in the case of certain social categories to make
comparisons of income and income structure—particularly as concerns
the components of the income, supplementary income or income in kind.
The data collected may be more or less of general value, depending
on the case involved.
Still more important are such data on the structure of expenditure
and consumption as it may be possible to obtain.
The problem of social security benefits, particularly benefits in kind,
has not yet been satisfactorily resolved in the case under discussion.
1
The results of this survey have been published in Informations statistiques,
Sixth Year, No. 3, May-June 1959, and No. 6, Nov. 1959.

INTERNATIONAL COMPARISON: E.C.S.C. BUDGET SURVEY

119

Family budgets alone do not suffice to demonstrate the standard
of living; other inquiries, particularly into the prices of articles and
services connected with consumption, and vast surveys on earned
income seem to be the kind of supplementary investigation most urgently
required.
All statistical surveys in the social sphere, like this one, should pay
particular attention to deviations from the averages calculated. Very
often these deviations are more important than the average values.
Finally—and it is at this point that we shall conclude this article—
in international comparisons of family budgets action must be taken
from the beginning to avoid going to extremes as regards the requirements to be complied with: " absolute " identity of methods may in
certain circumstances prove as prejudicial to the utilisation of the
results as " absolute " non-comparability. This is where the sense of
reality of the statistician comes in, and he must strive to find the proper
balance.

VII
A Survey of Food Consumption
in Great Britain
By D. F. HOLLINGSWORTH and A. H. J. BAINES *

THE NATIONAL FOOD SURVEY OF GREAT BRITAIN
MAIN FEATURES

The National Food Survey of Great Britain, established by the
United Kingdom Ministry of Food in July 1940, is a continuous inquiry
into domestic food expenditure and consumption, using weekly records
supplied voluntarily and without payment by representative samples of
private households and entered by the co-operating housewives in
specially designed log-books. It is intended to keep a continuous check
on the consumption and nutrition of all sections of the population, and
so to provide a factual basis for national estimates and for decisions on
food policy. The survey findings have also proved useful in the fields
of econometrics, social medicine and market research, but these uses are
ancillary to its administrative function.
Staffing Arrangements
The field work was carried out, until February 1953, by the London
Press Exchange Ltd., a commercial market research organisation. It was
then transferred, following a decision that surveys sponsored by government departments should be conducted by a central agency, to the Social
Survey Division of the Central Office of Information. These bodies were
successively responsible for recruiting, training and supervising the field
staff, and editing and coding the record books. The field workers do not
have a uniform background or training but are given detailed instructions
on how to conduct the interview and what checks to apply to the records
* Of the Food Science and Atomic Energy Division and the Statistics Division,
respectively, of the Ministry of Agriculture, Fisheries and Food.

FOOD CONSUMPTION IN GREAT BRITAIN

121

provided by housewives; they do not need any detailed knowledge of
nutrition. The punched cards were analysed by the Ministry's own
Hollerith installation until September 1951, and thereafter by the Central
Tabulating Installation of H.M. Stationery Office. The number of
interviewers engaged on the survey has varied from time to time; it is
now 25. Each housewife is visited at least three times during the survey
week—more often, if necessary.
Field of Information Covered
At the first interview the field worker explains to the housewife how
to record details of all food entering the household by purchase or otherwise during the following week. Since the object of the survey is to determine what families, rather than individuals, consume, the informant is
the housewife, who, as the family caterer, is responsible for buying food
or obtaining it, say, from a garden or farm. The information which she
is asked to provide must be within her knowledge. The survey, therefore,
excludes items which other members of the family may buy on impulse,
such as chocolates and sugar confectionery, soft drinks and alcoholic
drinks, and ice-cream andfishand chips obtained to eat outside the home,
as well as such items as vitamin tablets, the consumption of which by one
or more members of the family might distort the general impression of
the nutritional value of the family's food. Food obtained and consumed
outside the home is also excluded, except liquid milk taken through the
School Milk Scheme, under which all school children are entitled to
one-third of a pint (nearly 200 ml.) of milk free of charge on each school
day. In view of the importance of obtaining a complete picture of liquid
milk consumption, housewives are asked to record the quantity of milk
taken at school by members of their families.
Historical Development
Because of the need to obtain information on changes in the diet as a
guide to wartime food policy, the survey was initiated without a preliminary pilot inquiry in July 1940; its methods, however, were based on the
experience of the pre-war surveys made by Sir William Crawford and
Sir Herbert Broadley and the Carnegie Trust. Until January 1950 the
main survey was confined to urban working-class households, which
were considered most vulnerable to wartime and post-war shortages.
During the difficult winter of 1947-48 the resources of the survey were
devoted to an investigation of the diets of manual workers' households
in heavier industries, including agriculture; and from time to time

122

FAMILY LIVING STUDIES

inquiries were made into the diets of special groups, such as urban
middle-class families, slum dwellers and households in rural areas. In
January 1950, the survey was extended to all parts of Great Britain
except the remoter areas of Scotland, and in June 1951, the field work
was considerably simplified, as described below. The only subsequent
methodological change of importance was a revision of the sampling
scheme in 1953 on the transfer of the field work to the Social Survey.
Design of the Sample
The sampling scheme involves three stages : the selection of 5 0 1
parliamentary constituencies, of polling districts within each of these
constituencies, and of households within these polling districts. The
612 constituencies covered are classified into standard regions. Within
regions, constituencies consisting wholly of urban administrative areas
are stratified in England and Wales by the percentage of electors qualified
for jury service 2, in Scotland by the rateable value per person (excluding
industrial properties). The remaining constituencies, containing both
urban and rural areas, are stratified by the proportion of their population
living in rural areas. Thus the 612 constituencies are divided into 50
groups with approximately equal populations, and one constituency is
selected from each group with probability proportional to the size of
its electorate.
Interviews are made in half the constituencies alternately for threeweek periods; a polling district is worked for one ten-day period at a
time. In urban constituencies in England and Wales, polling districts
are stratified by the " juror index " (an index based on the proportion of
the electorate liable for jury service) and four districts per quarter are
selected with probability proportional to the size of the electorate, so
that when 20 addresses per polling district are selected the chances of
any particular household appearing are approximately equalised. In
mixed constituencies, the " percentage rural ", already used at the first
stage, determines how many of the four polling districts should be rural ;
the urban and rural areas of the constituency are then stratified separately
by the juror index for selection of the correct number of each class with
probability proportional to size. No polhng district is excluded from
selection, except in those few mixed constituencies in which the " percentage rural " is less than 12}4 or over 8 7 ^ . In Scotland, polhng
1
From 1950 to 1956, sixty constituencies were surveyed each year; the number
was then reduced for reasons of economy.
2
In England and Wales, liability to serve on a jury depends primarily on occupation of a house or flat exceeding a certain annual value.

FOOD CONSUMPTION IN GREAT BRITAIN

123

districts are selected without stratification, since the juror index is inapplicable and the rateable value per head not readily available for separate
polling districts.
The third stage of sampling consists of the selection with equal probability of about 85 addresses per constituency per quarter from the
electoral registers of the selected districts. Excluding addresses never
called on and those found ineligible (e.g., institutions), a total of 16,051
households were approached in 1957. It was impossible to get an interview with 3,767 of these (23.5 per cent.); 1,504 were interviewed but
refused to participate; 1,738 accepted but did not complete, and illcompleted log-books were rejected at the editing stage. The final sample
of 8,931 households that supplied complete information gave an effective
response rate of 55.6 per cent.1 The non-respondents are likely to include
a comparatively high proportion of households consisting of adults only,
and perhaps of households in the upper ranges of income.
Main Features of the Analysis
The main analyses of survey data are five in number:
(1) By social class, defined in terms of the gross weekly income of the
head of the household, as stated by the housewife or, if necessary,
imputed from other information.
(2) By household composition. The following types of families are
distinguished :
(a) Households of one man and one woman with—
no other (one or both 55 years of age or over) ;
no other (both under 55 years of age) 2 ;
one child (under 15 years of age); two children; three children;
four or more children;
one or more adolescents (15 to 21 years of age); adolescents and
children;
(b) Other households with—
adults only;
one or more adolescents but no children;
one or more children, with or without adolescents.
1

In 1958 the response rate was 60 per cent.
The younger childless couples proved a suitable basis for comparison with
families containing different numbers of children, since few adults in such families
are over 55 years of age.
2

124

FAMILY LIVING STUDIES

(3) By the occupation of the head of the household, using the RegistrarGeneral's classification by occupational status and skill, which have
long been used for the analysis of census and mortality data.
(4) By standard region.
(5) By degree of urbanisation, distinguishing London and other conurbations from the smaller urban areas, and distinguishing semirural areas (those which adjoin towns with a population of 25,000 or
more, or which themselves have a population density exceeding one
person to four acres) from rural areas.
SOMB METHODOLOGICAL ASPECTS AND FINDINGS OF THE SURVEY

Checks on Accuracy of the Data
The range of items and amount of detail covered by a budgetary
survey may affect the results ; so may the payment or non-payment of the
respondents, the closeness of supervision by the investigator and even
the format of the record book.
The survey log-book contains space for the housewife to record on
each day the menus of the meals she has served. This information is
used by the interviewer in checking the record of purchases; if the ingredients of the meals have not been bought during the survey week, the housewife can be asked whether they were taken from her larder. Prices are
checked by comparison with prices paid in local shops. The housewife
has an opportunity to discuss any difficulties with the interviewer who
calls back at least once during the survey week, and oftener if necessary,
making a total of at least three visits in all. There is a final opportunity
to clarify any outstanding points when the log-book is collected. The
general reliability of the National Food Survey results is confirmed by
their consistency from group to group and from year to year, and their
freedom from bias is checked with supply data, in so far as these are
derived from accurate returns independent of the survey.
Error in measuring family food consumption may arise if the housewife changes her purchasing habits because of the survey itself. Some
evidence of this is supplied by the National Food Survey records of the
energy value of foods purchased by elderly women who live alone. In a
study of the diets of such women over the age of 55 (722 in all) who
provided survey records between April 1953 and March 1954, the
writers * found that in each decade of life from 55 onwards the energy
a
A . H. J. BAINES and D. F . HOLLINGSWORTH: " T h e Diets of Elderly Women
Living Alone ", in Proceedings of the Nutrition Society (Cambridge University Press,
London), Vol. 14, 1955, p . 77.

FOOD CONSUMPTION IN GREAT BRITAIN

125

value of food obtained for consumption was some 1,000 calories per
head per day greater than requirements calculated according to the
recommendations of the F.A.O. Committee on Calorie Requirements.1
More recent analyses on similar groups have confirmed the finding.
Wastage of food in the home does not appear to explain the considerable
gap between purchases and requirements and it may be that this relatively
small group tends to buy more food during the week of the survey than
they intend to eat. Subsequent investigations suggest that the major
part of this overbuying is devoted to cheap storable foods, particularly
flour, sugar, potatoes and fats.2
One way of examining distortions of expenditure is to conduct
similar surveys for different lengths of time. Presumably overspending
for one week would not be continued for a month or more; further,
errors of omission might be accentuated in the second week. For both
reasons, one would expect weekly records to give higher figures than
those based on a longer period, though, in the absence of external
criteria, one could not say which is the more accurate. It is not yet possible to examine this point from National Food Survey results, but
Prais 3 has recently made an analysis in these terms of a survey conducted
in Israel. A sample of about 3,000 households was divided at random
into three approximately equal groups, differing only in the length of
time over which they recorded their food expenditure. The first group
kept records of food expenditure for a month, the other two for the first
and second fortnight respectively; all three recorded their expenditure
on non-food items for the whole month. For all food, the two fortnightly groups recorded expenditures whose sum was 5 per cent, greater
than the expenditure of the first group during the whole month, a difference which was significant at the 1 per cent, probability level. The
greatest differences were recorded for sugar and oil, foods that are
comparatively non-perishable.
Comparison between Consumption Levels Estimates from Supply Data
and from the National Food Survey
The records for the National Food Survey can be compared—
1
F.A.O. : Calorie Requirements, Report of the Second Committee on Calorie
Requirements (Rome, 1957).
2
Ministry of Agriculture, Fisheries and Food : Domestic Food Consumption and
Expenditure, 1959 (London, H.M. Stationery Office, 1961), para. 63.
* S. J. PRAIS: " Some Problems in the Measurement of Price Changes with Special
Reference to the Cost of Living ", in Journal of the Royal Statistical Society (London),
Series A, Vol. 121, 1958, pp. 312 flf.

126

FAMILY LIVING STUDIES

(a) with the " Consumption Levels Enquiry " estimates of foods in
their primary form moving into consumption at the retail level,
based on supply data;
(b) with estimates of food consumption based on the total personal
expenditure on food shown in the National Expenditure estimates.1
An attempt has been made to compare these three estimates for 1956 in
terms of their calorie values. The daily calorie value of food at the retail
level, as estimated by the Consumption Levels Enquiry, was about
3,130 calories per head per day. The National Expenditure estimates
yield a figure of about 2,750 calories, but the following should be added
to this to bring it to the retail level : 30 calories for garden produce not
included, about 290 calories for food obtained in catering establishments
and 15 calories for food provided for the armed forces, giving a total of
about 3,090 calories in fair agreement with the Consumption Levels
estimate. The items of food not recorded in the National Food Survey
(chocolate and sugar confectionery, soft drinks and ice-cream and other
foods purchased by individuals for consumption outside the home) were
estimated to yield about 190 calories per head per day. If this is deducted
from and if the 30 calories for garden produce are added to the National
Expenditure estimate of 2,750 calories, the resulting 2,590 calories is in
reasonable agreement with the 2,620 calories recorded by the survey.
The three estimates are thus approximately reconcilable.
Larder Stocks
From 1942 to mid-1951, detailed records were collected not only of
the expenditure on foods bought for consumption in the home, the
quantities of foods bought and the quantities obtained from gardens and
farms, but also of the changes in larder stocks, the estimation of which
required considerable care in weighing stocks at the beginning and end
of the survey week and took up much of the investigator's and housewife's time. The extension of the survey to middle and upper-class homes
made the measurement of stocks even more difficult and time-consuming,
and it appeared that in all classes more food was withdrawn from stock
during the survey week than was put into stock, the value in the highest
income group amounting to 23 per cent, of total food expenditure, com1
A minor difference in coverage is that the survey relates to Great Britain and
the other series to the United Kingdom; but the population of Northern Ireland is
less than 3 per cent, of that of the whole United Kingdom, and its food consumption
would have to be widely different to affect the national averages appreciably. (In fact
the only food known to be consumed in larger amounts in Northern Ireland than in
Great Britain is the potato.)

FOOD CONSUMPTION IN GREAT BRITAIN

127

pared with 11 per cent, for the whole sample. This apparent continual
running-down of larder stocks meant that the survey was distorting the
normal pattern of expenditure on food, probably because :
(a) the survey took up time which might have been spent in shopping;
(b) the housewife's attention was drawn to stocks which she had forgotten, and then proceeded to use up;
(c) there may have been a tendency to postpone shopping until after the
last weighing in order to save trouble.
It seemed, therefore, that the recording of changes in larder stocks
caused an underestimation of normal food expenditure, but probably
not of the total volume of food consumption (though the pattern of consumption may have been distorted, e.g., by the replacement of perishable
foods by canned or more durable cereal foods). In addition, the weighing
of larder stocks appeared to add to the reluctance of housewives in the
higher income groups to participate in the survey.
Accordingly, in June 1951 the survey procedure was simplified- The
recording of larder stock changes was discontinued, except for a few
items of home-produced food that are normally stored in bulk ; the burdert
on the housewife's time was thus approximately halved. The immediate
result of the change was an increase of Is. 6d. in food expenditure per
head beyond what could be attributed to changes in the price level and
normal seasonal changes in the pattern of food purchases. Of this
amount, 3d. was attributed to reduction in bias reflecting an improved
response by housewives and Is. 3d. to the change in technique. The
increase was most pronounced in the higher income groups, in households without children and especially those with only one or two adults.
This seems reasonable, as the ratio of larder stocks to average consumption is probably greatest in such households, and their weekly purchasing
pattern is less regular than in the larger families.
.. .
The effect of this important change in technique was studied further
by comparing the estimated cost of purchases and those stock, withdrawals still recorded in 1951-52, revalued for each month at the prices
current at the same month of 1950-51, with the actual expenditure and
complete stock withdrawals recorded in 1950-51. To eliminate seasonal
variations a quantity index was calculated by comparing the values in
corresponding months of the two years. The average for 1951r52 was
found to be about 2 per cent, below that for 1950-51. This is in accordance with the available information on total supplies of foods for those
years.1 It appears, therefore, that the total quantity of food represented
1

Ministry of Food Bulletin, No. 720 (19 Sep. 1953) and N o . 755 (29 May 1954).

128

FAMILY LIVING STUDIES

by combined purchases and stock withdrawals on the former system was
approximately equivalent to quantities purchased or withdrawn from
bulk stores of home-produced food under the new system.
Wastage
The National Food Survey records the quantity of food entering the
kitchen, not the amount actually consumed. The energy value and nutrient content of the recorded quantities are evaluated by means of appropriate tables of food composition, which make automatic allowance for the
presence of inedible material such as bones, the skins of fruits and
vegetables and the outside leaves of such vegetables as cabbage, but not
for losses of edible material. To make comparisons between groups containing different proportions of adults and children, males and females,
and persons engaged in occupations of varying degrees of activity, the
calculated " intakes " of energy and nutrients are compared with estimates of the nutritional requirements of the groups concerned, on the
assumption that 10 per cent, of all foods, and hence of energy and all
nutrients available for consumption, is not eaten, but is lost through
wastage or spoilage in the kitchen or on the plate or is given to domestic
pets. No doubt the amount of food wasted varies greatly in different
families, and between families of different types. Larger families probably need to exercise greater economy than smaller ones in their use of
food, and it seems possible that the hypothetical 10 per cent, wastage
ought to be graduated from, say, 5 per cent, in the largest families to
20 per cent, in the smallest. No measurements of wastage have been
made on a national scale, and it seems unlikely that such measurements
will ever be possible; to attempt to record wastage is to draw attention to
it, and once wastage has been brought to the housewife's notice she is
likely to reduce it.
The hypothesis of graduated wastage, however, has been examined in
another way. In 1957, a two-way analysis was made by social class and
household composition. Quantities of food obtained for consumption
were converted into energy value and nutrients, and compared with
requirements after the usual 10 per cent, adjustment for wastage. The
results of this comparison for calories are shown in the first half of
Table I.
On the assumption that the energy requirements adopted for adults,
adolescents and children are equally valid, these percentages suggest
that families containing two adults only, whatever their class, resemble
one another much more closely in their apparent over-consumption than
do, for example, small and large families in Class A. If it is assumed that,

129

FOOD CONSUMPTION IN GREAT BRITAIN

irrespective of class, families of two adults without dependants waste
20 per cent, of edible food, those with one child 15 per cent., those with
two children or with adolescents but no child 10 per cent, and the rest
5 per cent., the picture of the energy value of food consumed which
emerges is that shown in the second half of Table I.

TABLE I. ENERGY VALUE OF DIET AS A PERCENTAGE OF ESTIMATED
REQUIREMENTS FOR HOUSEHOLDS OF DIFFERENT COMPOSITION
WITHIN SOCIAL CLASSES (1957)
Households with one male and female adult and
No other
(both
under 55)

(1) Applying uniform allowance for wastage of (per
cent.)

Class A 1
Class B
Classes C and D I . . .

115
117
113

(2) Applying graduated allowance for wastage of
(per cent.)

Class A
Class B
Classes C and D I . . .

102
104
100

Children only

Adolescents
only

Adolescents and
children

110 106 108 100
107 105 101 102
108 100 95 94

108
103
100

99
94
89

4 or
more

104 106
101 105
102 100

114
106
100

106
108
99

108
103
100

105
99
94

1
Class based on weekly income of head of household, or of principal earner if gross weekly income
of head was less than £7.

The extent of wastage is likely to vary with the relative abundance of
food. Table II shows for the whole sample for each year from 1950 to
1958 the energy value of food obtained for household consumption, and
the adequacy of this is assessed by the usual comparison on the assumption that 10 per cent, is wasted. The results suggest that, on the average,
10 per cent, was a fair estimate of wastage in the home until 1953, when
wastage appears to have been higher, though in 1957 there was some
evidence of a reduction. It is of interest that food rationing was in full
operation in 1950-52, was gradually relaxed in 1953 and was finally
brought to an end in May-July 1954, when fats and meat were freed from
the ration.

130

FAMILY LIVING STUDIES

TABLE II. COMPARISON OF ENERGY VALUE OF FOOD
CONSUMPTION WITH ESTIMATED REQUIREMENTS, ASSUMING 10 PER
CENT. WASTAGE. ALL HOUSEHOLDS, 1950-58
Year

1950
1951
1952
1953
1954
1955
1956
1957
1958

. . . .

Energy value of
consumption
(calories)

Energy value of
consumption as a
percentage of estimated
requirements

2,474
2,461
2,447
2,520
2,626
2,641
2,624
2,587
2,595

101
100
99
101
105
105
105
103
104

This gap between intake and estimated requirements did not appear
for all types of households; it was concentrated in the higher income
groups, in small families, and, surprisingly, in households mainly dependent on the old-age pension. Between 1952 and 1956 the energy value of
the diet increased markedly in relation to requirements for Class A,
old-age pensioner households and especially for childless couples, but
only slightly for the larger families and not at all for the largest. In 1957
all groups showed small decreases. No fully adequate explanation has
yet been advanced for these differing trends; rationing seems to have
maintained demand in some groups while restricting it in others, and it
appears likely that changes in the quantities wasted may also be a factor.
Meals Eaten Away from the Home
In any survey of family food consumption, a method must be devised
for treating food bought for the family but eaten by visitors, and food
consumed by individual members of the family away from the home. 1
Since the main purpose of the National Food Survey is to study the pattern of the diet in the home, its records relate to quantities of food
obtained for consumption in the home, which are expressed " per head
per week ", a " head " being defined as a person eating at least sixteen
meals at home during the survey week ; anyone eating fewer is a " visitor ".
In comparing this estimate of consumption with an estimate of nutritional
need, the nutritional requirements of the household are adjusted to allow
1
In the National Food Survey, packed meals such as sandwiches provided by the
housewife for consumption away from home are conveniently treated as part of the
domestic consumption.

131

FOOD CONSUMPTION IN GREAT BRITAIN

for visitors' consumption and for outside consumption by members of
the household. The log-book records the number of members of the
household and of visitors who partake of each meal in the home and the
number and types of meals taken outside the home by each member of
the household. It is assumed that the normal meal pattern is that of four
meals (breakfast, dinner, tea and supper) each day. The relative importance of each meal is assessed according to the following somewhat
arbitrary weightings, which are subject to revision:
TABLE III. RELATIVE IMPORTANCE O F MEALS ACCORDING TO WEIGHT
Weight
Meal

Breakfast
Dinner
Tea
Supper
Total . . .

per day

per week

4
5
3
2

28
35
21
14
98
(say 100)

These weights were adopted to construct an index of meals taken at
home for each household. If the weekly total for each person is taken as
100, each breakfast, dinner, tea or supper eaten out by that person is
represented by a deduction of 4, 5, 3 or 2 respectively, and the remainder
gives the " net balance " of meals eaten at home by that person. Meals
eaten by visitors are given the same weights and are added to the household total, so that a visitor's meal cancels a corresponding meal taken
out by a similar person. Nutritional requirements are calculated in
terms of the net balance for each person. Thus, if it is assumed that the
nutritional value of similar meals eaten inside and away from the home
is the same, it can be said that the nutritional value of food obtained for
consumption at home is being related to the nutritional needs of the
persons when they eat at home ; the remainder of their nutritional needs
is assumed to be met elsewhere.
In 1952, an attempt was made to determine the effect of outside meals
on domestic food consumption and expenditure. The analysis was limited
to certain comparable groups of households in Class C containing at
least one earner, one group taking some meals out and the other none.
One pair of samples consisted of families of one man and one woman
only, another of families of one man and one woman with two children.
The two samples of each pair were matched as closely as possible in

132

FAMILY LIVING STUDIES

respect of other relevant variables, such as age distribution, family
income and nutrient requirements, so that differences in their domestic
diet could be attributed to outside meals. Table IV gives the results of
this comparison. For each series the variations between the percentage
differences in expenditure and in the nutrient content of the diet are
within the sampling error; the results therefore suggest that for these
households the reduction in expenditure and in the nutrient value of the
domestic diet when outside meals were taken was of the same order as the
proportion of outside meals. These calculations provide some support
for the assumption that the outside meal has the same nutritive value as
the corresponding meal at home.
TABLE IV. EFFECT OF OUTSIDE MEALS IN SAMPLES OF HOUSEHOLDS
IN CLASS C (1952)
Households of one man
and one woman

Proportion of meals out
Domestic food expenditure per head per week
Nutrient value of food obtained for consumption
per head per day:
Calories
Calories per penny
Protein (g)
. . .
Calcium (mg) . .
Riboflavin (mg)
No. of households

. . .

No
meal
out

Some
meals
out

6-9%

313d. 292d.

2,918 2,706
65
65
96
88
1,250 1,140
2.0
1.8
279

211

Percentage
difference

6.6

7.3
8.5
9.1
9.6

Households of one man,
one woman and two
children
No
meal
out

Some
meals
out

8.9%

215d.

197d.

2,337 2,105
76
75
71
64
980
910
1.5
1.4
122

Percentage
difference

8.3

9.9
8.4
7.0
6.2

122

An assumption more difficult to justify is that four meals are eaten in
all households each day and that their weights are as given above. No
National Food Survey data have yet been analysed in terms of meal
patterns, but some interesting information can be obtained from a
market research inquiry1, conducted in August 1955 and February 1956
into the food habits of adults. Table V shows the percentage of the
samples taking the various meals stated.
1
The Foods We Eat, ed. Geoffrey C. WARREN (London, Cassell and Co. Ltd.,
1958).

133

FOOD CONSUMPTION IN GREAT BRITAIN

TABLE V. PERCENTAGE O F TOTAL SAMPLE T A K I N G VARIOUS MEALS,
1955 A N D 1956
August 1955
Meals

Early morning tea
Breakfast
" Elevenses "
Midday meal
Afternoon tea
Evening meal
Late supper

. . . .

February 1956

Men
2,142

Women
2,415

Men
2,142

Women
2,415

40
93
48
95
45
92
76

47
91
50
98
50
92
78

43
96
55
98
56
91
81

49
94
57
98
58
89
82

The inquiry also provides information on what was eaten at each of
these meals. About half the informants had a cooked course at breakfast; a quarter to a third had something to eat with their mid-morning
and mid-afternoon tea or other beverage, and over half with their late
supper beverage. The midday meal appeared on the average to be
slightly heavier than the evening meal. These findings suggest that a
pattern of three main meals and two or three subsidiary ones represents
current British habits more closely than the survey division into four
meals a day. Nevertheless, a change in the present survey procedure
would have little effect on the nutritional results, since all classes take
the great majority of their meals at home: for example, in 1952, nearly
half the households surveyed had no meal out and another 20 per cent.
had less than 5 per cent, of their meals out. Only 3 per cent, took more
than 20 per cent, of their meals outside the home.
Effect of Household Size on Food Expenditure
The much lower food expenditure per person found in large households is obviously due in part to their lower incomes per head and the
larger proportion of children found therein; but an important third
factor operates, namely, the economy which is effected in catering for
large numbers, largely no doubt through avoidance of wastage. To
throw some light on this factor, the food expenditure in each household
of size 2 to 8 persons surveyed in 1956 for which the family income was
declared has been adjusted to that appropriate to a family income of
£4 per head per week, using an income elasticity of 0.3. A further adjustment to a " standard family " was then made for differences in the proportions of men, women, adolescents, children aged 5 to 14 inclusive,
children aged 1 to 4 and infants under 1 year, using least-squares estimates
of the expenditure attributable to each type of person within each house-

134

FAMILY LIVING STUDIES

hold size. The standard proportions adopted were those found in
the whole sample, namely: men 29 per cent., women 32 per cent.,
adolescents 7 per cent., children (5 to 14) 21 per cent., children (1 to 4)
9 per cent., infants under 1 year 2 per cent. The results, given in Table VI,
are subject to sampling fluctuations arising from small numbers and the
correlation between numbers of men and women in households of given
size; nevertheless it is clear that while the adjustments for income differences are substantial, the additional adjustments for demographic
differences are quite small.
TABLE VI. E F F E C T S O F H O U S E H O L D SIZE ON D O M E S T I C

FOOD

EX-

P E N D I T U R E CORRECTED F O R D I F F E R E N C E S I N F A M I L Y I N C O M E PER

Food expenditure per
head per week (shillings)

Food expenditure adjusted to constant income per head (shillings)

1.89
2.21
2.28
2.40
2.53
2.75
2.55

0.05
0.61
1.44
2.14
2.93
3.38
4.37

36.0
30.1
25.7
23.0
20.9
19.7
17.6

31.7
29.4
26.6
24.8
23.5
22.1
21.2

per
household

Approximate standard
error of increment
(shillings)

No. of children per
household

1,307
2,155
1,882
883
373
150
73

per
head

Increment for each
additional person
(shillings)

No. of adults per
household

2
3
4
5
6
7
8

No. of households

No. of persons

H E A D A N D F A M I L Y COMPOSITION (1956)

31.4
29.3
26.7
25.3
24.7
23.2
21.5

62.9
87.8
107.0
126.6
148.0
162.7
172.4

24.9
19.2
19.6
21.4
14.7
9.6

0.7
0.8
1.2
2.0
3.5
6.0

Food expenditure
further adjusted
for difference in
household composition
(shillings)

The adjusted estimates of food expenditure given in Table VI should
be regarded as preliminary results, but they provide interesting information on the effect of household size when income per head and demographic structure are both held constant. Thus the expenditure attributable to the first two persons in the household (adjusted to the average
family composition) was about 31.4 shillings each; that associated with
the third person was significantly lower at 24.9 shillings and that for the
fourth was again significantly lower at 19.2 shillings; thereafter the
increment appears to level off and the standard errors become large.
Brown 1 has discussed the use of the survey results in obtaining
equivalent-adult scales (in the economic, not the nutritional, sense) to
describe the dependence of the consumption of different commodities
and nutrients on the composition of the household. It is necessary to
1
J. A. C. BROWN: " Economics, Nutrition and Family Budgets " , in The Proceedings of the Nutrition Society, Vol. 14, 1955, p. 63.

FOOD CONSUMPTION IN GREAT BRITAIN

135

distinguish between the scale specific to the commodity and the scale
appropriate to income. It had previously been assumed that the measure
of size of the family on either scale was the weighted sum of the numbers
of each type of person in the family, but Brown has pointed out that this
is an over-simplification and has introduced the more general concept
of " equivalent-standard-households ", to allow for the economies of
scale found in the food expenditure recorded by larger households.
Conversion of Consumption Quantities into Calories and Nutrients
Evaluation of the energy value and nutrient content of the average
diet of each household group is carried out mechanically by means of a
punched card system which utilises a Hollerith type 550 electronic
calculator. The nutrient content per unit quantity of each of the 120 food
groups in the classification, suitably adjusted to allow for inedible waste
and losses of thiamine and vitamin C in cooking, is punched by hand on
to a pack of 120 " unit master " cards; separate packs are prepared in
respect of each quarter of the year, so as to allow for seasonal changes in
the energy and nutrient content of certain foods. For each group of
households for which nutrient evaluations are required, the average consumption per person per week of each of the 120 food groups is then
automatically reproduced from the results of prior calculations on to a
further pack of 120 " nutrient average " cards. These packs of nutrient
average cards are pooled and sorted to food-code order, and are then
collated and merged with the unit master cards; they are then passed
through the electronic calculator which multiplies the consumption
quantities by the appropriate nutrient conversion factors, converts the
product into units per person per day, and punches the results on a
designated field of the nutrient average cards. These cards are then
re-sorted and tabulated to show the average intake of each nutrient in
each type of household and the contribution to that intake from each
category of food.
The restraints of machine tabulation and computation mean that the
survey classification of foods must be rigid. All foods are grouped into
about 120 categories, to almost all of which separate nutrient conversion
factors are applied. With so small a number of categories the nutrient
analysis for many of them must be weighted according to the best information available, to take account, say, of the various cuts of meat,
measured together as " carcass meat—beef and veal ". Ideally, there
should be individual factors for each kind of meat purchased by housewives. For the whole sample, these restraints are probably not important :
nutrient analyses are obtained for " average " foods and the weightings

136

FAMILY LIVING STUDIES

adopted are for " average " food consumption. Such average nutrient
analyses may not, however, be strictly applicable to the foods bought by
a group with exceptional food habits. For example, elderly women
living alone appear to prefer mutton to beef, and if they prefer the
cheapest cuts of mutton, such as breast of mutton, to other types, the
nutrient content of their mutton consumption will not be well described
by the average factors devised for " carcass meat—mutton and lamb ".
It is of the greatest importance to obtain accurate analyses of those
foods that contribute most to the nutritional value of the diet. As an
indication of this, Table VII gives the sources of calories in the diets of
all households surveyed in 1958.
TABLE VII. ENERGY VALUE OF FOOD OBTAINED FOR DOMESTIC
CONSUMPTION, ALL HOUSEHOLDS, 1958
Calories per head
per day

Percentage of
total

. . . .

315
381
24
50
386
322
184
49
587
268
29

12
15
1
2
15
12
7
2
23
10
1

Total . . .

2,595

100

Food

Milk and cheese
Meat
Fish
Eggs
Fats
Sugar and preserves
Vegetables
Bread and flour
Other cereals
Other foods

This suggests that, using energy value as a guide, the most important
foods for which accurate nutritional analyses are required are bread and
flour, with meats and fats running second. The first requirement is fully
met in the United Kingdom by a scheme of sampling of flour as produced
and of analysis of samples in the government laboratory. The much more
difficult subject of meat is not covered in such a way, but it is ripe for
study, particularly in relation to the amount of meat fat wasted. Neither
the visible fats (butter, margarine and cooking fats) nor any other food
group presents such a difficult problem as meat, the importance of which
increases as attention is turned from calories to protein, fat and some
other nutrients.
Assessment of Nutritional Requirements
Consideration of the problems associated with the measurement of
food consumption and its conversion to nutritional value is incomplete

FOOD CONSUMPTION IN GREAT BRITAIN

137

without some appreciation of the problems associated with the measurement of nutritional requirements. The precision with which such requirements can be estimated depends on the accuracy of the scales of nutritional requirements used and on the way in which they can be applied.
In the National Food Survey, the housewife is asked to record in the
log-book the sex and age of members of the household and the occupations of working members. From these facts an assessment of requirements of calories, protein, calcium, iron and some vitamins, using as
a basis the recommendations of the Committee on Nutrition of the
British Medical Association (1950), is made on the assumption that
occupation determines activity. No adjustment is made, except in old
age, for the decrease in activity with increasing age, though such an
adjustment would be possible if the recommendations of the Food and
Agriculture Organisation Committee on Calorie Requirements were
adopted. 1 It would not, however, be practicable to apply the recommendations of that Committee on adjustments for body weight. No
records of body weight are obtained in the National Food Survey, and
it is difficult to envisage such an additional burden being placed on
housewives and investigators. Thus the scales of nutritional requirements must be devised to fit adults of both sexes of " average " weight 2 ,
and children and adolescents of both sexes and various age groups.
Finally, as already explained, adjustments can be made for the proportions of meals eaten away from home and the proportions of edible
foods assumed to be wasted. Some of the comparisons between intake
and requirements made on this basis have already been discussed.
The energy requirement of the British population, calculated according to the recommendations of the British Medical Association, is about
2,400 calories per head per day at the physiological level if allowance is
made for different degrees of activity in adults. As the total supplies of
food available in 1956 were equivalent to about 3,100 calories per head
per day, this implies that wastage (including food fed to animals) was of
the order of 700 calories per head per day or more than a fifth of the food
supply. The energy requirement of the households which took part in
the survey in 1956 was 2,250 calories per head per day using the same
scale of allowances. They obtained food amounting to 2,620 calories
and this suggests a domestic wastage of over 350 calories, or about a
seventh. These large gaps between supplies and physiological requirements cannot yet be satisfactorily explained, but their occurrence in all
1

Calorie Requirements, op. cit.
Data are available for the population of Great Britain in 1943. See W. F. F.
KENSLEY: "Weight and Height of a Population in 1943", in Annals of Eugenics
(London, University College), Vol. 15, 1950-51, p. 161.
2

138

FAMILY LIVING STUDIES

well-developed countries is confirmed by comparisons between estimates
of the calorie value of food supplies in F.A.O. Food Balance Sheets and
of calories requirements according to F.A.O. recommendations and, their
existence was noted as long ago as 1952 in the Second World Food
Survey of F.A.O. in the following words :
The significance of average figures which exceed estimated requirements
also calls for special comment. Such excess does not mean that the average
person in the countries concerned consumes more food than he or she requires,
though there may be certain individuals in the population who are overfed and
show a tendency towards overweight or obesity. In well-fed and prosperous
countries a considerable " gap " almost always exists between estimated per
caput supplies at the retail level and calorie requirements as estimated according to any recognised system, and also between the former and the directly
observed consumption of samples of the population. The reasons for this
" gap " are not fully understood.1
In the United Kingdom the gap between the total supply estimates
at the retail level and domestic consumption recorded by the National
Food Survey can be bridged. That between either estimate of food
consumption and physiological requirements cannot, unless the wastage
actually occurring in food processing and distribution and in the home
is much higher than has as yet been recorded.

F.A.O.: Second World Food Survey (Rome, 1952), pp. 12 and 13.

Vili
A Survey of Underemployment in
Puerto Rico
By

A. J. JAFFE *

In the economically less developed countries information about the
number of underemployed persons, their personal characteristics, and
the sectors of the economy in which they work is necessary in order to
formulate plans for economic development. Underdeveloped countries
commonly have very few wholly unemployed; to survive, people have
to make a living somehow, and as a rule few breadwinners are found to
be without any work at all. On the other hand, it is obvious that very
many are not fully or satisfactorily employed.
It is apparent to even the most casual observer that in many countries many of the people are working much less than they seem capable
of doing. Furthermore, in underdeveloped countries in particular, many
of the people who appear to be working hard and for long hours nevertheless seem to be poverty stricken. Hence, it is necessary to distinguish
between more or less full employment and underemployment in current
statistics if they are to have any relevance for purposes of planning for
economic and social improvement.1
Puerto Rico's experience in measuring underemployment since 1952,
one of the longest, is of particular interest because certain aspects of
both visible and invisible underemployment are measured in the regular
sample surveys. Moreover, the definitions used, by and large, are those
contained in the " Resolution concerning the Measurement of Underemployment " adopted by the Ninth International Conference of Labour
Statisticians (Geneva, April-May 1957), which are worth repeating here.
* Director, Manpower and Population Program, Bureau of Applied Social
Research, Columbia University, New York. Formerly Consultant to the Department
of Labor, Bureau of Labor Statistics, Puerto Rico.
1
For further development of these points see UNITED NATIONS: Handbook of
Population Census Methods (3 vols.), Vol. II, Economic Characteristics of the Population,
Studies in Methods, Series F, No. 5, Rev. 1, Sales No.: 58.XVII.6 (New York, 1958).

140

FAMILY LIVING STUDIES

Underemployment exists when persons in employment who are not working
full time would be able and willing to do more work than they are actually
performing, or when the income or productivity of persons in employment
would be raised if they worked under improved conditions of production or
transferred to another occupation, account being taken of their occupational
skills. Underemployment appears in various forms, some of which can be
measured with reasonable accuracy by means of statistical inquiries. The
following major categories of underemployment may be distinguished:
(a) visible underemployment, which involves shorter than normal periods of
work and which is characteristic of persons involuntarily working part
time;
(b) invisible underemployment, which is characteristic of persons whose
working time is not abnormally reduced but whose earnings are abnormally low or whose jobs do not permit full use of their capacities or
skills (sometimes called " disguised " underemployment), or who are
employed in establishments or economic units whose productivity is
abnormally low (sometimes called " potential " underemployment).1
Several other countries have also made or are making underemployment surveys, but a world-wide evaluation of such surveys is beyond
the scope of the present paper. This paper describes how the recurrent
statistics of underemployment were developed in Puerto Rico within
the framework of the regular quarterly sample survey of the labour
force, what concepts lie behind the procedures in use and how the
practical procedures were worked out; a brief résumé of the Japanese
inquiries is then given. There follows a comparison of the procedures
used, in the light of the different social and economic situations in the
two countries, and some general findings with respect to underemployment measurement in sample surveys conclude the article.

ORGANISATION OF THE PUERTO RICAN LABOUR FORCE SURVEY

Although data on underemployment were not collected prior to
1952, the Bureau of Labor Statistics of the Puerto Rico Department of
Labor began in 1946 a statistical survey of the labour force throughout
the country. The sampling scheme is similar to that employed in the
United States for the monthly current population survey.2
Information is collected by a permanent field staff of enumerators
who visit the sample households and inquire about the work activities
of the members of the household during the preceding week.3 Statistical
•I.L.O.: International Standardisation of Labour Statistics, Studies and Reports,
New Series, No. 53 (Geneva, 1959). p. 49.
2
See Chapter XIII.
* It should be noted that the enumerators are used for collecting other statistics
during the parts of the month in which they were not interviewing households for
labour force information.

UNDEREMPLOYMENT TN PUERTO RICO

141

data are obtained on levels of employment and unemployment among
civilians 14 years of age and over (excluding inmates of institutions).
The decision made to attempt the collection of statistics on underemployment implied making subdivisions of the category " employed " :
those underemployed and those fully employed. It was very desirable
that any changes made in procedures should not result in discontinuity
of the labour force statistics collected since 1946. Hence, the first major
decision made, and one which influenced all future decisions, was to
keep unchanged the original set of questions on labour force activity
and the definitions of employed persons and of the unemployed.1
The sequence of original questions designed to provide basic labour
force data is as follows:
(1) During the last week what was this person's main activity? If the
answer is " working for pay or profit " or " unable to work ", no
further questions are asked. Persons who reply " working " are
classified as employed; those " unable to work " as not in the labour
force.
(2) If any other answer is given to question (1), the person is then
asked : " In addition, did you do any work for pay or profit during
this week? " If the person answers " yes ", he is classified as employed.
(3) A person who answers " no " to question (2) and had not already
indicated (in the answer to the first question) that he was seeking
employment is then asked : " Did you look for work last week ? "
If he answers " yes " he is classified as unemployed (as are persons
who replied to the first question that they were looking for work,
and who also replied to the second question that they had not
worked at all during this week).
(4) Persons answering " no " (to the third question) are then asked :
" Did you have a job from which you were temporarily absent?"
If the answer is " yes " the person is classified as employed, but not
at work. If the answer is " no " the person is classified as not in the
labour force.
All persons who worked during the reference week are asked the
number of hours worked, their present occupation and status (" class
of worker ") and the industry in which they are currently engaged.
For the purpose of obtaining statistics of underemployment a number
of questions were designed and added to the original questionnaire.
1
Also retained were the basic questions on personal characteristics: age, sex,
relationship to head of household, marital status, place of residence (farm or nonfarm).

142

FAMILY LIVING STUDIES

It was thought that there would not be enough cases in the original
sample of 6,000 households to provide reliable statistics if the employed
group were subdivided into the fully employed and the underemployed
and each were cross-classified by personal and job characteristics. In
addition, it was desired to obtain many more cross-classifications of the
employed and the unemployed than before. Accordingly, it was decided
that at least twice as many households should be sampled.
Doubling the number of interviews to be conducted could mean
doubling the field costs of a survey; it was felt, however, that the total
annual costs of the labour force survey should remain substantially
unchanged.
A study made of the uses to which the monthly data were put revealed
that relatively little emphasis was placed on the monthly feature. Very
little short-run, anti-cyclical action could be taken. Instead, Puerto
Rican government officials were essentially concerned with planning
for long-run economic development, and for this purpose statistics
collected less frequently but providing a wealth of data useful for development planning were necessary. It was decided, therefore, to shift the survey to a quarterly basis (January, April, July and October), and to double
the number of households in the sample so that the total annual costs
of the survey and the number of enumerators needed remained substantially unchanged. The sample now includes about 13,000 households.
One essential feature of the labour force survey data was that they
referred to labour market activities as of one week, and to retain comparability with past data this feature had to be continued.1
Each enumerator now had about twice his previous work load and
was allowed twice as much time. Since the field work now spread over a
longer period, the reference week became a " floating " week and it
extended over about three calendar weeks.
The data on underemployment are issued by the Puerto Rico Bureau
of Labor Statistics from time to time in the series Special Reports on
the Labor Force ; the information on employment and unemployment is
issued regularly in the Quarterly Report : Employment and Unemployment in Puerto Rico.
THE APPROACH TO UNDEREMPLOYMENT MEASUREMENT IN
PUERTO RICO

In Puerto Rico, as in other underdeveloped areas, underemployment
takes two forms—visible and invisible. One can see people who actually
1
If the reference period had been changed to a month, for example, this would
have resulted in shifting some persons from the unemployed to the employed category.

See UNITED NATIONS: Handbook of Population Methods, op. cit., pp. 23, 24.

UNDEREMPLOYMENT IN PUERTO RICO

143

work only a few hours per day, or part of a week, or perhaps only part
of a year—the visible underemployed. In addition, there are many
people who are constantly on the job yet obviously are working at such
low productivity levels that they are producing and earning very little—
the invisible underemployed. Street pedlars, many of the personal
service workers, and subsistence farmers, among others, have this
characteristic.
In deciding how underemployment might be measured in Puerto
Rico 1, the following considerations were kept in mind :
Since the measurement of underemployment was to be attempted
within the framework of the quarterly labour force survey, the period
of reference would be only one week; information for a longer reference
period, one year for instance, had best be obtained by some other means.
The desires of the worker and his attitude towards work are crucial
considerations. Does or does he not desire to do more work than he is
now doing (presumably for more income) ? Ideally, each person should
be asked to weigh all the factors influencing him, and to decide, on
balance, whether he really wishes to have more work.
It might have been presumed that an alternative to asking him
whether he wants more work would be to ask him whether he wants to
earn more money. In Puerto Rico it was so obvious that everyone would
reply " yes " to such a query that this line of questioning was not even
attempted. Instead, it was decided to handle the question of " wanting
more work " by means of hours worked per week. Since the labour
force survey already obtains such information it is possible to build
additional questions on underemployment around " hours worked per
week ".
Measuring underemployment by reference to reported hours worked
per week might do for those clearly suffering from visible underemployment, such as employees on part time. However, from the information
already collected in the labour force survey it was known that the selfemployed and unpaid family workers often reported having worked
long hours per week: 60, 70 and sometimes even 80 hours or more per
week were reported. Yet when one observed them at work, so many
were clearly working at such low productivity levels that they would be
considered as suffering from invisible underemployment. Hours worked
reported by self-employed persons and unpaid family workers were
1
See Puerto Rico Department of Labor, Bureau of Labor Statistics: "Results
of Tests on Measuring Underemployment in Puerto Rico, June and July 1952 ",
in Technical Report on the Labor Force No. 2 (San Juan, Puerto Rico, 1953). See also
F. S. BERDECÍA and A. J. JAFFE : " The Concept and Measurement of Underemployment ", in Monthly Labor Review (Washington), Mar. 1955, Vol. 78, No. 3, pp.
283 ff.

144

FAMILY LIVING STUDIES

therefore not always relevant for the purpose of deciding whether underemployment was present.
These observations led to an early decision to try to measure underemployment separately for employees and for the combined group of
self-employed and unpaid family workers. With respect to employees,
visible underemployment would be measured, and for the self-employed
and unpaid family workers a combination of both visible and invisible
underemployment would be measured. Nevertheless, for both groups
a key item would be the respondent's evaluation of the situation, i.e.,
whether he did or did not want more work. However, " more work "
might ultimately be defined in operational terms.
It was felt that a person who really wishes to have more work ought
to take some steps, however limited, to obtain such additional work.
In Puerto Rico, however, as in many other underdeveloped areas where
there are few job opportunities, workers often do not actively seek
employment because they know that no jobs are available.1 Hence, any
attempt to consider job-seeking in connection with underemployment
must somehow take this into consideration.
On the basis of already available information about the agricultural
population it was believed that farm labourers who were employees
and who, in Puerto Rico, all work for money wages, could be handled
in the same manner that might be chosen for non-agricultural employees.
The self-employed farmers and their unpaid family workers, however,
seemed to pose problems quite different from those of the self-employed
and family workers in non-agricultural industries.2 For example, these
people often had no idea of how much time they might actually work
in a week, since time had almost no relevance for them. Also, they had
even fewer opportunities for supplementary employment than did the
urban self-employed. Accordingly, it was decided to measure underemployment among the self-employed and unpaid family workers in
agriculture by a different method than might be used for agricultural
employees or persons in non-agricultural industries.
Finally, it was deemed absolutely necessary that the survey procedures
be kept as simple as possible ; it would be impracticable to add very greatly
to the existing questions in the labour force survey. Therefore, in devising
the procedures and the additional questions to be included in the labour
1
See, for example, Simon ROTTENBERG and Nora SIFFLEET: Unemployment in
Antigua, Labor Relations Institute, College of Social Sciences, University of Puerto
Rico (Rio Piedras, Puerto Rico, 1952).
2
Actually, there are so few unpaid family workers in non-agricultural industries
in Puerto Rico that they can be disregarded for all practical purposes. The great
majority of unpaid family workers are engaged in agriculture.

UNDEREMPLOYMENT IN PUERTO RICO

145

force survey the fundamental rule followed was to adopt the simplest
approach, one that did not do violence to the basic concept of underemployment.
The information on number of hours worked per week was included
in the final determination of underemployment (as will be seen below)
for the reason that some arbitrary limit of " enough " work had to be
adopted, in view of the state of the Puerto Rican economy. Considering
the large number of people who were totally unemployed together with
the large number who clearly worked less than a full week (5 days at
8 hours per day), the government could not feel obligated to find more
work for everyone who desired more work. Instead, the government's
obligations, at least during the earlier period of economic development,
could be extended only to those most in need of work, or additional
work—the totally unemployed together with those so much underemployed that they approached being unemployed.
Tests were conducted during the spring and summer of 1952. The
first pre-tests consisted simply of lengthy interviews of a number of respondents in both urban and rural areas. On the basis of this pre-testing
it was possible to select a small number of crucial questions. These
questions were included in the June and July 1952 labour force surveys.
Analysis of the results confirmed confidence that the procedures for
measuring underemployment were valid and practical, and hence these
questions have been continued quarterly, beginning with October 1952.
The sequence of supplementary questions asked of all employed persons
at work 1 is as follows :
(a) In addition to working last week, did this person also look for work
last week? Yes or No.
(b) What was the main reason why this person did not work more
hours last week?
(c) Did this person desire to work more hours last week? Yes or No.
Answers to question (a) yield a direct measure of desiring more
work and seeking whatever additional work opportunities the economy
may have to offer. Answers to questions (b) and (c) give indirect
measures of such seeking. For example, a person who replies that he
" could not get any more work " is implying seeking, and if he further
replies that he wanted more work, he is classified as underemployed.
On the other hand, a person who replies, for example, that he " stayed
1
Persons who worked one or more hours during the reference week (except
unpaid family workers who are counted as employed only if they worked 15 hours
or more during the week).

146

FAMILY LIVING STUDIES

home to paint his house " is indicating that he made no efforts to obtain
more work, and he is classified as fully employed.
The intent of question (b) is to differentiate between those people
for whom the economy was unable to supply more work but who were
desirous of having more work, and those people who, whether more
work had been available or not, did not choose to work more. The
purpose of question (c) is to serve as a check on the previous answers.
For a person to be considered as underemployed, it is not enough for
him to reply that he had not worked more hours because more work
was not available; he must also say that he had wanted to work more
hours.
In the surveys of June and July 1952, additional questions were
addressed to those persons whose replies to earlier questions indicated
that they were engaged in agriculture as self-employed or unpaid family
workers, as follows:
(d) How many cuerdas1 of land did this person really work?
(e) What is the principal crop?
(f) Does the family consume or sell the major part of the farm's produce?
Results of the Surveys of June and July 1952
The results obtained are shown in Tables I to III below. The figures
refer to persons in the sample households only, not to estimated numbers
in the whole country.
Employees.
Of the male employees who had worked under 35 hours during the
reference week and who replied that they had looked for work in addition to having worked (thereby indicating that they had really wanted
more work), about 90 per cent, also said that they had been unable to
obtain more work and had desired to have more work (line 14 of Table I).
Among women, a little over half replied the same way. Since there was
a great preponderance of men among those employees who had worked
under 35 hours and had also looked for work, it was decided, for practical
reasons, to use the operational definition that would be suitable for
men. Hence, it was decided to classify as underemployed all employees
who worked under 35 hours and said that they had also looked for work.
As was noted above, under conditions of job scarcity, it is to be expected that many persons do not actively seek work because they believe
no work is available. Hence, employees who worked less than 35 hours
1

A cuerda is just short of one acre.

147

UNDEREMPLOYMENT IN PUERTO RICO

and did not report having sought work, but who reported that they
could not obtain more work and had wanted more work, were also
classified as underemployed.
It was mentioned previously that, in identifying the underemployed,
an upper limit on the number of hours worked per week had to be
imposed. It can be argued that this upper limit should be 30 or 35 hours;
there was general agreement that in Puerto Rico under 30 hours was
definitely " too little " and 40 hours was definitely " enough ", all things
considered.
The available data (see Table I) indicate that among employees who
worked 30 to 34 hours, about the same proportion felt themselves to be
in need of additional work as among those who had worked under
30 hours. However, few of those who had worked 35 hours or more
reported desire for more work. Accordingly, for Puerto Rico, 35 hours
seems to be a suitable and practical cut-off point for employees.
TABLE I. D E G R E E O F JOB SEEKING: EMPLOYEES, J U N E A N D JULY 1952
Hours worked in survey week
Men
1-29
hrs.

1. Total persons . . . . 2,142
2. Looked for work in
addition to working . 564
3. Wanted more work . 541
4. Unable to get more
work
502
5. Other reasons for not
working more hours .
39
6. Did not want more
23
7. Did not look for work 1,578
8. Unable to get more
work
1,129
9. Wanted more work . 1,043
10. Did not want more
work
86
11. Other reasons for not
wanting more work . 449
12. Underemployed: No.
(lines 2+9)
1,607
13. Underemployed: %
(line 12H-line 1) . . .
14. Line 4 as % of line 2 .
15. Line 9 as % of line 8 .

75
89
92

Women

30-34
hrs.

1-34
hrs.

35 hrs. 1-29
or more hrs.

30-34
hrs.

1-34
hrs.

35 hrs.
or more

967

3,109

4,609

882

186

1,068

1,627

195
187

759
728

181
132

118
103

20
19

138
122

38
15

172

674

127

8
31
49
15
1
772 2,350 4,428 764 166

16
23
930 1,589

504
450
54
268

1,633
1,493

2,840
685

165
92

58
39

223
131

1,095
70

140 2,155

1,025

599

108

707

494

866 210
Percentages

269

108

32
1

25
49
59

7
18
6

717

645 2,252

67
88
89

72
89
91

1,588

19
70
24

24
50
56

Source: Puerto Rico Department of Labor, Bureau of Labor Statistics: " Results of Tests on Measuring Under-employment in Puerto Rico, June and July 1952 ", op. cit., Table 1.
1
Percentage not computed (too few cases).

148

FAMILY LIVING STUDIES

Self-Employed and Unpaid Family Workers in Non-Agricultural Industries.
For many of these people, time worked per week has little relevance;
indeed, a very long work week may simply indicate low productivity,
as when a pedlar spends a long time waiting for customers. It was
decided to use the same criteria for measuring underemployment among
them as among employees, except that no upper limit on hours worked
would be imposed (see Table II). Hence, the results for this group
combine both visible and invisible underemployment. If a person had
" worked ", i.e., had been on the job, many hours during the week
(40 or 50 or even more) and still said that he wanted more hours of work,
the supposition is that his work was of such low productivity that he
could be. considered as suffering from invisible underemployment. In
the pre-test, when talking to these persons and noting their earnings
and how they spent their time during the day, it was clear that in many
cases invisible underemployment was being expressed in terms of wanting
more work.
TABLE II. DEGREE OF JOB SEEKING: SELF-EMPLOYED AND UNPAID
FAMILY WORKERS IN NON-AGRICULTURAL INDUSTRIES, JUNE
AND JULY 1952
Hours worked in survey week
Women

Men

1. Total persons . . . .
2. Looked for work in
addition to working .
3. Wanted more work
4. Unable to get more
work
5. Other reasons for not
working more hours .
6. Did not want more
work
7. Did not look for work
8. Unable to get more
9. Wanted more work
10. Did not want more

1-29
hrs.

30-34
hrs.

1-34
hrs.

35 hrs.
or more

1-29
hrs.

30-34
hrs.

1-34
hrs.

253

118

371

1,078

371

434

52
49

23
22

75
71

37
21

48
45

3
3

51
48

3
201

1
95

4
296

16
1,041

3
323

0
60

3
383

115
85

46
28

161
113

441
65

64
43

9
3

73
46

376

600

259

310

11. Other reasons for not
wanting more work .
12. Underemployed: No.
(lines 2+9)

135

137

188

102
91
Percentages

13. Underemployed: %
(line 12-i-line 1) . . .
14. Line 4 as % of line 2 .
15. Line 9 as % of line 8 .

54
88
74

51
91
70

9
51
15

22
63
63

25
60
67

Source: " Results of Tests on Measuring Under-employment in Puerto Rico, June and July 1952 '
op. cit., Table 4.
1
The category of women working 35 hours or more yielded too few cases for separate analysis.
* Percentage not computed (too few cases).

UNDEREMPLOYMENT IN PUERTO RICO

149

Self-Employed and Unpaid Family Workers in Commercial Agriculture.
During the pre-test, it appeared that many of these resembled the
self-employed and unpaid family workers in non-agricultural industries
in so far as their labour market activities were concerned. Accordingly,
it was decided to measure underemployment among them in the same
manner as among the self-employed and unpaid family workers in nonagricultural industries.
Self-Employed and Unpaid Family Workers in Subsistence Agriculture.
Self-employed and unpaid family workers in agriculture were asked
three additional questions about their farms in the surveys of June and
July 1952. The answers (summarised in Table III) were used to distinguish commercial from subsistence (or quasi-subsistence) farms. It was
found that the great majority (90 per cent.) of males, self-employed or
unpaid family workers, engaged on farms where the major part of the
farm produce was consumed by the family, were on small farms (under
10 cuerdas) producing minor crops (vegetables and starch crops, a few
chickens, and a few banana and plantain plants). On the other hand,
among similar workers on farms where the major part of the produce
was sold, only 18 per cent, were engaged on farms of under 10 cuerdas
growing minor crops. The remainder, over 80 per cent., were on farms
growing commercial crops such as sugar cane, tobacco, coffee, pineapples
and citrus crops, or had dairy farms. Furthermore, these were generally
larger farms than those where the family reported consuming the major
part of their produce.
It was believed to be impracticable to include, in each quarterly
survey of the labour force, questions on size of farm and type of crop
raised. Since the single and simple question asking whether the farmer
and his family consumed, or sold, the major part of the farm's produce,
gave results which correlated highly with size of farm and type of crop,
it was decided to ask only this one question in future surveys. All selfemployed and unpaid family workers in agriculture who reply that they
consume the major part of their farm produce are therefore classified
as underemployed.
In Puerto Rico, observations confirm that subsistence or quasisubsistence farm workers fit the definition of invisible underemployed.
They produce little, i.e., have low levels of productivity, and, unless some
member of the family is employed outside the home, they live in poverty.
Furthermore, in Puerto Rico there are very few such farmers ; commercial
crops are grown mainly on large plantations ; the great majority of persons engaged in agriculture are employees. Therefore, even if the pro-

150

FAMILY LIVING STUDIES

cedure used is not the best for ascertaining underemployment among
the self-employed and unpaid family workers in agriculture, it was felt
that it would suffice in Puerto Rico. It has the great virtue of simplicity.
TABLE III. SELECTED CHARACTERISTICS OF MALES IN AGRICULTURE:
SELF-EMPLOYED AND UNPAID FAMILY WORKERS, JUNE
AND JULY 1952

Total
males

Reporting consume major part
of produce
Type of crop:
Minor crops 1
All other crops a . . . .
Reporting sell major part of
produce
Type of crop :
Minor crops 1
All other crops 2 . . . .

Size of holding
where employed

Proportion
engaged on
Under
5 to
10 or more holdings under
10 cuerdas
5 cuerdas 9 cuerdas cuerdas

%
90

395

299

372
23

281
18

55
1

36
4

694

189

153

352

168
526

93
96

34
119

41
311

76
41

Source: " Results of Tests on Measuring Under-employment in Puerto Rico, June and July 1952 ",
op. cit., Table 2.
1
Miscellaneous vegetables and fruits of a type that can be eaten by the household, chickens, etc.
1
Mainly sugar cane, tobacco, coffee, pineapples and dairy products; these are all products that must
be sold by the farmer.
* Percentage not computed (too few cases).

Other Statistics
Puerto Rican statisticians collect other data which in a certain sense
serve as indicators of visible underemployment. In January of each year,
members of the sample households in the labour force survey are asked
about their work experience during the preceding year. Each civilian
14 years of age or over is asked whether he has done any work for pay or
profit during the preceding year. If the answer is " yes ", he is asked the
approximate number of weeks worked, and the average number of hours
worked per week. Those persons who reported having worked 40 or
more weeks and 30 or more hours per week, are classified as " steadily "
employed ; all others who worked any amount of time are classified as
" intermittently " employed.1 The procedures are kept as simple as
possible. No effort is made to determine whether the worker wanted
more work than he obtained. If it were deemed practical to ask more
questions, this approach might be elaborated upon and the desires of the
person taken into consideration.
^ e e , for example, Department of Labor, Bureau of Labor Statistics: "Work
Experience in Puerto Rico: Calendar Year 1958", Special Report on the Labor
Force, No. 23 (San Juan, 1959).

UNDEREMPLOYMENT IN PUERTO RICO

151

MEASURING UNDEREMPLOYMENT IN JAPAN

The basic labour force survey schedule design and sampling procedures used in Japan are quite similar to those used in Puerto Rico.
Information is collected monthly by enumerators who visit some 11,000
households.1
Underemployment or, to use the term frequently used in Japan,
" hidden unemployment ", is touched upon in the monthly survey via a
question addressed to those who worked during the week : " Did you
want to do more work... in addition to the work you did during the
survey week? " If the answer is " yes ", the person is asked to report
how many more hours he wanted to work, and whether or not he looked
for such additional work.
Twice a year supplementary questions aimed at determining underemployment are asked of a subsample of the regular monthly labour
force sample. In addition, special employment-status surveys are
conducted once every three years, using the same questions ; the first was
conducted in July 1956. These special surveys cover 1 per cent, of the
population and permit much more detailed analyses than are possible
from the small number of cases obtained in the semi-annual subsamples.
The semi-annual and triennial surveys include about 30 questions
bearing on underemployment (as compared with seven in the Puerto
Rican surveys) and the majority call for answers much more than a
simple " yes " or " no ". One approach to measuring invisible underemployment is by asking the question " Do you want to continue your
present work ? " If the person wants to continue he is not considered
as underemployed. If not, he is asked additional questions aimed at
determining whether his job is of such a nature that he should be classed
as underemployed.
Another approach is to ask the worker whether he changed jobs
during the previous year. If he did so, he is asked additional questions
aimed at determining whether he was previously underemployed or
unemployed.
All employed persons (plus certain categories of the unemployed)
are asked the number of hours worked per week, or, if not working
1

See the following English language publications of the Office of the Prime
Minister, Bureau of Statistics (Tokyo): Bureau of Statistics—Its Organisation and
Activities, 1958; Labor Force Survey of Japan, Mar. 1956; Employment Status
Survey as of July 1, 1956, Mar. 1958; Report on the Special Survey of the Labor
Force Survey—Oct. '55, Mar. '56, Oct. '56, Mar. '57 and Oct. '57, Sep. 1958;
Monthly Report on the Labor Force Survey ; Report on the Revised Figures of the
Labor Force Survey, Nov. 1957. See also the following English language publications
of the Ministry of Labor (Tokyo): Monthly Labor Statistics and Research Bulletin;
Japan Labor Year Book (annual).

152

FAMILY LIVING STUDIES

regularly, the number of days worked per week, per month or per year.
This line of questioning reveals visible underemployment. Next, questions on earnings and income are asked; these (in theory at least) can
be used for measuring invisible underemployment. Finally, persons not
at work at the time of the survey are asked several questions to find out
whether they had worked previously in the year and, if so, whether they
had been previously underemployed.
Obviously, this rather long series of questions and the attendant
complicated and detailed cross-tabulations of the answers, which need
to be made for analytical purposes, are costly in terms of both personnel
and money. Probably not many countries can afford such surveys.
Nevertheless, the approach to measuring invisible underemployment by
inquiring about the desire to change jobs seems especially intriguing.

COMPARISON OF PUERTO RICAN AND JAPANESE PROCEDURES AND
GENERAL CONCLUSION

Comparison of the procedures used for measuring underemployment
in Puerto Rico and Japan, particularly in the light of the social, economic
and labour market conditions in these two areas, ought to suggest some
general findings which should be relevant for other countries.
An important fact which emerges is that the approach to underemployment measurement is conditioned by the structure of the economy
and the organisation of the different economic sectors.
Agriculture in Japan is conducted largely by the self-employed, mostly
through small farms operated by the proprietors, their wives and children.
Over two-thirds of the farms are under one hectare and less than one in
ten is over two hectares in size. Agriculture is the largest industry,
having in 1957 about 39 per cent, of all occupied persons; there are
comparatively few farm employees. Puerto Rico has a significantly
smaller proportion engaged in agriculture and most of those engaged
are hired workers. There are comparatively few subsistence farmers in
either country, in the sense of farm families living almost completely
outside of the money economy.
A long work week is customary in Japan and a short week in
Puerto Rico. In Japan the average is some 48 hours per week. To the
extent that persons actually do work so many hours per week, it would
seem that they would not wish to have much additional work; they do
not have time for a second job. This may explain why so few report that
they want additional work. In short, there is little visible underemployment in Japan. Consequently more emphasis has been placed on

UNDEREMPLOYMENT IN PUERTO RICO

153

measuring invisible underemployment among wage and salary earners.
Of particular interest is the inquiry about change of job and desire to
change job.
The Japanese worker does not seem to think in terms of work, or
of pay, by the hour. Minimum wages are thought of in terms of a month
rather than an hour, as in Puerto Rico. Apparently it makes comparatively little difference to the Japanese worker how much time within the
month he spends on the job, provided he earns at least the minimum
specified. This may be related to the reported long work week and it
tends to negate efforts to measure underemployment in terms of time.
In their attempts at measuring underemployment, both Japan and
Puerto Rico stress the importance of the worker's personal desires.
It is not appropriate for another person to judge whether the worker
should, or should not, want more work or different work.
One of the problems of using earnings data to derive an indicator
of underemployment stems directly from the fact that the desires of the
workers cannot be taken into account. At first sight, it would appear
to be possible to decide on an acceptable minimum level of earnings in a
particular society and then to classify everyone who earns less than the
specified amount as underemployed. This approach, however, fails to
take into consideration underemployment that may be associated with
earnings above the specified amount. Such data were collected in Puerto
Rico, but it was not found to be a practical way of ascertaining underemployment.
Another problem with earnings data involves their validity. The
replies of employees were generally reasonable in the light of other
information available on wages and salaries, but the replies of the selfemployed were questionable. Accordingly, even if earnings data were
used for employees it would not seem possible to use this basis to measure
underemployment among the self-employed and the unpaid family
workers.
Furthermore, and this was most troublesome, it was virtually impossible to arrive at any cut-off point for amount of earnings (below which
underemployment could be assumed to be present) or to decide whether
earnings should be measured in terms of hourly, daily, weekly or monthly
rates. Some workers may receive relatively high hourly rates, but work
only a short time per week. Others may receive much less per hour but
work many more hours per week. Because of this unsolved conceptual
problem and the difficulties of collecting adequate data, efforts at using
earnings data for determining underemployment were abandoned in
Puerto Rico.
Cultural and social characteristics of the people and the organisation

154

FAMILY LIVING STUDIES

of the labour market have to be taken into consideration in deciding
on procedures to measure underemployment.
The Japanese are highly literate, more so than the Puerto Ricans.
Literate persons are more accustomed to dealing with questionnaires,
have a better understanding of the questions and are better equipped to
give answers; as a rule the Japanese answer freely all the questions
asked by the labour force enumerator. In these circumstances, it is
practicable to ask many questions, probing into various aspects of
labour force behaviour including underemployment.
Under certain conditions visible underemployment may be as
difficult to measure—or perhaps more so—than invisible underemployment. This seems to be the case in Japan where comparatively little
public attention is given to the matter of earnings per hour worked and
all workers keep busy (or at least think that they are keeping busy) most
of the time. The comparatively easier problem of measuring visible
underemployment in Puerto Rico as contrasted with Japan, seems to
stem directly from this factor. By paying more attention to time, and
keeping closer track of the amount of time spent on actual work for pay
or profit, the Puerto Rican employee is in a better position than is the
Japanese worker to tell just how much time he has really worked and
whether or not he wants to work more hours.
Customary hiring procedures can affect underemployment measurement. Different procedures are needed where a worker is customarily
hired by the hour from those required where he is customarily hired by
the month or year.
The society's view as to what constitutes full employment for an
individual is also a factor. In the United States or Australia, for example,
over 40 hours work per week in most occupations and industries constitutes overtime; hence, in these countries about 40 hours per week
would mean a full-time job. Similar legal provisions do not exist in most
of the less developed countries, but there is quite frequently in a given
part of a country, or a specific industry, a customary work week which
constitutes a full-time job.
The general state of the employment market sometimes affects the
problem of finding a suitable criterion for defining underemployment.
If the supply of labour is very great relative to the demand, almost any
job might be considered a good job and three or four days work per
week might be considered a reasonably full work week, everything
considered.
In the last analysis, the procedures adopted for measuring underemployment by means of household interviews must be tailored to actual
conditions in the country. Underemployment among urban employees

UNDEREMPLOYMENT IN PUERTO RICO

155

differs in many respects from underemployment among agricultural
employees and among the self-employed and unpaid family workers in
agriculture and in other industries. Furthermore, even within agriculture
there are significant differences in the measurement problems between
those engaged in commercial agriculture and those engaged largely in
non-commercial agriculture.
Finally, it should be noted that the household interviewing procedures, at best, supply only approximate answers. Questions on an
interview schedule cannot be worded in such a manner as to fit exactly
into any concept that may be adopted for measuring underemployment.
Indeed, the nature of underemployment (however defined) could hardly
permit obtaining answers from the general public that would have a
high degree of precision. This is not to say that a satisfactory working
compromise cannot be achieved. In the case of the Puerto Rico survey,
a workable compromise was reached within the framework of the
quarterly sample survey of the labour force. Within this framework,
and the cost limits imposed, there were few opportunities to try out
entirely new procedures. Another country making an entirely fresh
start and intending to measure underemployment only once a year, or
perhaps even less often, might work out quite different procedures and
use a larger battery of questions, as in the case of the triennial Employment Status Survey of Japan. As a general rule, however, the simplest
techniques possible have to be used, otherwise the cost of such mass
collections becomes prohibitive, in both money and manpower. How
simple the techniques can be depends on the structure of the labour
market, other socio-economic factors, the abilities and willingness of the
people to reply to the questions asked, and the frequency with which
the surveys are conducted.
In conclusion, it may be said that there is no uniquely correct measure
of underemployment, either visible or invisible, which holds for any one
country or which can be applied internationally. The best that can be
hoped for realistically is to devise a procedure for any particular country
that can be used to measure the direction of change over time and to
compare regions of the country at a particular time.

IX
A Population Survey in Mysore State,
India
By C. CHANDRASEKARAN *

GENERAL DESCRIPTION

The Mysore study was undertaken jointly in 1951-52 by the Government of India and the United Nations as an experiment in the use of
sample surveys of households to obtain basic demographic data necessary
for the understanding of changes in population size and characteristics
and their interrelations with economic and social developments in
countries where the relevant statistics are not readily available. In the
different states of India, of which Mysore is one, population censuses
have been conducted every ten years since 1881, and data on the size
and chief characteristics of the population as of different census dates
are available. The available birth and death statistics, however, are poor
in quality, with the result that even the levels of birth and death rates
prevailing in different parts of the country have been in doubt, placing
serious limitations to a deeper understanding of the factors underlying
population change.
The insufficiency of information becomes more and more evident as
the government's programme of planned action for economic and
social development begins to get into its stride. Such a programme
should be planned with reference, among other things, to detailed
estimates of population, including estimates for various age groups,
and numbers of workers available for productive employment in different
areas of the country and segments of the economy. It should take into
account the number of children and young persons who would become
available for education and training, and the schooling and other
facilities required if the potential human resources are to be fully utilised.
* Director, Demographic Training and Research Centre, Chembur, Bombay.

POPULATION SURVEY IN MYSORE STATE

157

In addition, the possible effect of changing economic and social
conditions on the important determinants of population growth, such
as birth and death rates, must be known. There is a special need for
information on those factors affecting the number of births that may be
amenable to influence through measures of population policy.
The Mysore study made an attempt to provide important information
needed for policy-making that could not be obtained from Indian
censuses or vital statistics. Certain topics on which information could
be obtained from other sources were also included in order to gain
experience in the application of new concepts and methods of analysis.

Areas Selected
An important element in the plan of the Mysore study was the
selection of areas for study which would represent contrasting degrees
and forms of economic and social development. The state of Mysore
was divided for this purpose into several strata that, with regard
to such criteria as the extent of urbanisation, the condition of agriculture
and the degree of development of public health activities, appeared to
to be relatively homogeneous.
Not all the zones thus set up could be included in the survey for
budgetary reasons. The following areas were selected for study on the
ground that they could be expected to show marked contrasts in their
demographic characteristics.
Rural Areas—
Zone I : Rural hills, with large-scale anti-malarial operations.
Zone II: Rural hills, without large-scale anti-malarial operations.
Zone III : Rural plains, tank-irrigated areas.
Urban Areas—
Medium-sized towns with population from 10,000 to 25,000. Bangalore City.
The taluqs, which are small administrative divisions (with an average
population of approximately 100,000), were used to delimit the rural
zones in the state. The taluqs with high elevations in the rural areas
were combined to form the " hills " and, in keeping with the geography
of Mysore State, the " hilly " zones were situated along its western
border.
All census units with populations below 5,000 in the 1951 census
were to be regarded as villages and as making up the rural zones. By
the time the survey was designed, a list of villages based on the 1951
a

158

FAMILY LIVING STUDIES

census had not become available. Only some preliminary figures of this
census giving a list of municipalities with their populations and some
important characteristics had been published. The various difficulties
encountered in the formation of the rural zones and the procedures
adopted to overcome these difficulties are discussed below.
Bangalore City was divided into five strata on the basis of the 1941
and 1951 census data, which were designed with a view to ensuring the
inclusion of relatively large numbers of persons from the minority
religious groups for fertility studies. The city has 50 municipal divisions,
for each of which literacy data from the 1951 census were available
when the survey was being planned. Since the 1951 data on religion
had not yet been compiled, the 1941 census figures were used instead.
Municipal divisions with 35 per cent, or more Muslims formed one
stratum, those with 35 per cent, or more Christians formed another and
those with 35 per cent, or more Scheduled Caste Hindus formed the
third. The remaining municipal divisions were divided into two strata ;
the divisions where the percentage of literates among males aged five
years and over was 60 and above formed one stratum, while the residual
divisions constituted the other.

Sampling Design
The field work in the Mysore study included three main operations,
designated as the Village Survey, the Household Survey and the Fertility
and Attitude Survey. These required samples of villages, of household
and of married couples respectively. Since the Household Survey was
the main operation, the sampling plan was designed primarily to meet
the needs of that survey. The villages selected as primary sampling
units in the three rural zones, according to the sampling design described
below, constituted the sample used in the Village Survey. The selection
of couples for the Fertility and Attitude Survey was made from those in
the households included in the Household Survey.
In planning the sample of households, it was necessary to decide,
first, how many households were to be interviewed, and second, how
these households were to be selected.
An essential criterion in fixing the number of households was that the
survey results should provide an estimate of the annual birth rate in
each area with a standard error in the range of 6 to 8 per cent, of the
birth rate. With standard errors held down to this magnitude, it was
expected that sizable differences observed among the birth rates for the
different areas could be established as statistically significant.

POPULATION SURVEY IN MYSORE STATE

159

A two-stage sampling scheme was adopted. The first stage involved
the selection of certain localities (villages in the rural zones, census blocks
in the urban areas) within each study area, and the second, the selection
of a certain proportion of households within each of these localities.
The scheme for the different areas was as follows:
For each of the three rural zones—
(i)

selection of a constant proportion (1 in a) of villages, at random, from each of six strata established by dividing the
villages by population size, i.e., under 100, 100-299, 300-499,
500-999, 1,000-1,999 and 2,000-4,999 ;
(ii) selection of a constant proportion (1 in b) of households,
at random, in each village to be surveyed.
For the medium-sized towns of 10,000 to 25,000 inhabitants taken as a
group—
(i)

selection of a constant proportion (1 in a) of census blocks, at
random, in each of the 15 towns lying within the surveyed areas;
(ii) selection of a constant proportion (1 in b) of households, at
random, in each selected block.
For each of the five strata of Bangalore City—
(i)

selection of a proportion (1 in a) of census blocks, at random,
from those belonging to the given stratum;

(ii) selection of a constant proportion (1 in b) of households, at
random, in each selected block.
In determining the sampling fractions (or the value of the constants
a and b) for the rural zones and the medium-sized towns, a guiding
principle was that the cost of the field work should be at the minimum
compatible with the prescribed standard sampling error. Enlargement
of the number of localities would be expected to result in a reductionof the standard error of the estimates. On the other hand, the larger
the number of locahties, the greater would be the cost, particularly in therural zones where certain cost factors would vary considerably with thenumber of locahties (e.g., travel between villages, housing and field
staff for work in each village, getting in touch with local officials and
obtaining their co-operation and preliminary field work such as listing
of households).
For Bangalore City, a variation in the sample design was introduce d.
The number rather than the proportion of the blocks selected from
each of the five strata was kept constant. The chief object of this pro-

160

FAMILY LIVING STUDIES

cedure was to ensure the representation of a relatively higher proportion
of households of the minority groups in the sample with a view to
providing an adequate basis for revealing any differences in the demographic patterns in the various groups concerned. This stipulation
permitted at the same time the investigation of " interviewer differences "
by arranging the field work in the form of 5 X 5 Latin squares—the
five strata, a set of five investigators and the five " intervals " into which
the survey period was divided being combined to form the Latin squares.
Such a design made it possible also to equalise the effects of " recall
lapse " in the data obtained from the different strata.
Data from previous surveys conducted in India and results of pretests of the present survey were used in the consideration of cost-variance
functions and infixingthe sampling fractions. These werefixedin such a
way that one-third of the households in the villages selected in rural
areas and one out of six of the households in the census blocks selected
from the medium-sized towns or the strata of Bangalore City were
included for survey. The total number of households to be surveyed was
divided among the different zones as follows: Rural areas—1,136 in
Zone I, 1,567 in Zone II, 2,220 in Zone III, Urban areas—1,273 in
medium-sized towns, 4,548 in Bangalore City.
The Sampling Frame
The sampling frames for the selection of census blocks from urban
areas were provided directly by the 1951 census. The sampling frames
used for the selection of villages from the rural zones were made up by
combining the information available from the 1951 and 1941 censuses,
and using the list of " municipalities " with populations of 5,000 or over in
the 1951 census, which had been published by the time the study was
being designed. This group of municipalities was taken out of the 1941
census list of villages and municipalities, and the remaining villages1
were used as the universe of sampling for the different rural zones. This
universe differed from that ideally required in two respects: it contained
some extraneous villages, i.e., those which were populated in 1941 but
became uninhabited by 1951, and it also failed to include villages which
had sprung up between 1941 and 1951.2
1
Including seven villages that had populations of 5,000 and over in 1951 but were
not " municipalities " and had therefore not been included in the published list.
2
The former could be removed during the course of the survey. The magnitude of
the difference, due to the latter, which remained uncorrected, could only be estimated
when the 1951 census data had been fully tabulated; eight persons per 1,000 in Zones
I and II and three persons per 1,000 in Zone III were found to be excluded in this way
from the universe used for sampling.

POPULATION SURVEY IN MYSORE STATE

161

In order to separate villages belonging to Zone I (rural hills, with
anti-malarial operations) and Zone II (rural hills, without anti-malarial
operations) in the census list, the Director of Public Health provided a
list of villages (or hamlets which could be considered as satellites of the
main village) in which the anti-malarial programme had been carried
out since 1946. A large number of places in the list supplied by the
Director of Public Health could be identified easily in the village list.
Where names of hamlets had been provided it was necessary to make
reference to the amildar (taluq official) to supply the name of the main
village to which it belonged. In some instances the amildar was not able
to supply the information; and such cases could not be included in
Zone I. In Zone I were thus included all villages of the rural hills for
which there was definite information that the entire village or a part of it
(main village or hamlet) had been included in the anti-malarial programme. All other villages in the list for the rural hills were taken as
belonging to Zone II.
The villages in the three rural zones were further classified into the
six strata of population size described earlier, on the basis of their 1941
populations.
The frames for the selection of households from selected census
blocks and villages had to be prepared by a fresh listing of households.
Those provided by the 1951 census had become out of date by the time
the survey was undertaken but helped in the field operations for demarcating boundaries of villages and census blocks.
Range of Information Obtained
The Village Survey was carried out in a sample of 194 villages obtained
by the procedures already described. Information on the characteristics
of these villages was obtained by interviewing one or two well-informed
persons in each village and consulting the official records. The characteristics recorded included :
(a) transportation and communication facilities (road and railway
communications, postal and telegraph services) ;
(b) agricultural conditions, including area cultivated, crops grown,
extent of irrigation, land tenure, type of farming, agricultural
techniques and marketing of agricultural produce ;
(c) non-agricultural employment opportunities (nearness of industrial
establishments or other development projects);
(d) social services and amenities, including libraries, radio, cinema,
schools, medical and health facilities, and sanitation.

162

FAMILY LIVING STUDIES

The bulk of the data obtained by the Village Survey was factual and
could be ascertained readily from the officers or other well-informed
persons in the village, or by looking into the official records which were
carefully maintained by the administrative department concerned. In
many instances, the interviewers who were entrusted with the task of
obtaining information for the Village Survey could themselves check on
the accuracy of the data given. In fact, the items of information included
in the Village Survey were so selected, taking into account the availability
of routine official statistics, that there should not be much difficulty in
obtaining the requisite data. As a rule, it was found that the data required
for each village could be obtained with the required accuracy in about two
days. Since the information obtained by the Village Survey was mainly
intended to divide the villages into broad categories for studying their
relations with demographic characteristics, the level of accuracy desired
was not high.
This information, together with the information obtained from the
Household Survey, was used to indicate the degree of economic and
social development and " modernisation " of rural communities in the
three zones, and to study its relation to population trends.
For the Household Survey, a sample of 10,744 households was drawn
at random from the villages selected asfirst-stageunits in the three rural
zones, from the census blocks selected in the medium-sized " towns ",
and from the five strata in Bangalore City. The data were obtained by
interviews with members of each household. The principal subjects of
investigation were the following:
(a) number of births and deaths during a specified period preceding the
date of the survey;
(b) number of children born to each woman during her married life,
and such related factors as age at marriage, incidence of widowhood
and frequency of remarriage;
(c) basic demographic characteristics of the household and its members
at the time of survey, including religion, caste, literacy and education
level, occupation, industry and status (as employer, employee, etc.),
unemployment and part-time employment, type of housing and
domestic lighting (as indicators of economic status of the household)
and, in the case of persons cultivating land, tenure and size of land
holdings.
The Fertility and Attitude Survey covered 1,403 young married
women (between 18 and 33 years of age) and 1,084 husbands belonging
to the households included in the Household Survey in rural Zone III

POPULATION SURVEY IN MYSORE STATE

163

and four of thefivestrata in Bangalore City. The subjects of investigation
included :
(a) detailed data on pregnancy history of married women;
(b) opinions and attitudes with regard to size of family, age at marriage,
the desire to have or not to have more children, the reasons for such
desires, and related questions;
(c) the knowledge and practice of methods of family limitation.
In this survey, information was obtained by interviews with the
individuals concerned. A group of specially trained female interviewers
was used to interview the wives, and a group of specially trained male
interviewers to interview the husbands. The interviews for this survey
were more time-consuming and more delicate than those of the Household Survey and it was therefore necessary to restrict the sample for the
Fertility and Attitude Survey both in size and geographical coverage.
Pre-testing
An important step in the development of the study was the pretesting of the various schedules and procedures which were to be applied
in collecting information. A preliminary testing of the Household
Survey procedures was done in Serai-Julena village (Delhi State) in
March 1951. This village was selected because it was near the office in
New Delhi where the schedule designing was done and had also been
studied earlier by the Indian National Sample Survey. A list of households was readily available and the field trial could begin without
elaborate preliminaries. Extensive pre-testing of the schedule and survey
procedures was begun in Mysore State in July 1951. Two villages of
different size and certain census blocks in one medium-sized town and in
Bangalore City were selected for the pre-test so as to provide an idea of
the difficulties to be met under various working conditions. The pre-test
also furnished information on the time required for various phases of
the work.
Time Schedule for the Main Survey
The collection of data in the main survey began in the middle of
December 1951 and was completed by September 1952. The time
schedule for the survey was so arranged that the zones in which it was
easiest to work were taken up first. Rural Zones I and II were considered
most difficult to survey because of the hilly terrain and want of adequate
communication facilities. These zones were covered after the survey
staff had gained enough experience in Zone III (rural plains) and in the

164

FAMILY LIVING STUDIES

towns or Bangalore City. The Household Survey, along with the Village
Survey in the three rural zones, was carried out first and was followed by
the Fertility and Attitude Survey in rural Zone III and Bangalore City.
Recruitment and Training of Field Staff
The Mysore study paid particular attention to the recruitment and
training of field staff and to field organisation.
The field staff engaged in the Mysore study consisted essentially of
interviewers and inspectors. The interviewers were responsible for
visiting households to obtain data. The inspectors supervised the
work of the interviewers, checked the accuracy of the data collected
and, in addition, independently obtained the data for the Village
Survey. They also provided liaison service between the central office
of the study in Bangalore and the local authorities of the state
government.
The field staff were recruited in batches in order to make it easy to
give them intensive training. The first batch was chosen from among
those who had worked in the Agricultural Labour Enquiry of the Government of India which was concluded just prior to the Mysore study.
This group participated in a pre-test of the Household Survey and, with
this experience, was of much help in training the groups subsequently
recruited by means of advertisements.
The personality of the candidate, his ability to obtain the co-operation
of the public, his acquaintance with rural conditions, familiarity with
local languages and his readiness to cope with difficulties which might
arise while working in such areas were given consideration in the recruitment of the interviewers and inspectors. There was no difficulty in
obtaining personnel that could satisfy these requirements, for about
600 persons applied tofillapproximately 40 positions as interviewers and
inspectors.
The majority of the candidates selected were graduates in economics,
statistics or agricultural and other natural sciences. Age and experience
served as the main criteria for deciding whether a candidate selected
should be appointed as an interviewer or inspector. The average age of
interviewers was 26 years and that of inspectors 32 years.
Two important principles were followed in planning the training
programme. First, in view of the limited facilities available, the group
trained at any one time was not to exceed 15 in number. Second, the
inspectors and scrutiny assistants (who really were part of office staff)
were also to receive all the training imparted to the interviewers, in order
to be able to discharge satisfactorily their supervisory and scrutinising

POPULATION SURVEY IN MYSORE STATE

165

duties. This training was carried out in villages situated in the Ramanagaram Health Centre area. As this Centre maintains vital statistics and
other demographic data, it was possible to assess readily the quality of
the work of the trainees. The training was given in two stages. The first
stage, which lasted for about three to four weeks, was intended to
acquaint the trainees with the schedules and the manuals of instructions
for filling them. This training was conducted in the atmosphere of a
seminar and the trainees were encouraged to build up complicated field
situations to illustrate and test the concepts and definitions given in the
manual. The second stage \ which lasted from a week to ten days,
consisted of field training arranged in three phases. In the first phase, the
trainees observed experienced interviewers at work; in the second, they
worked under close supervision of experienced staff; and, in the final
stage, the interviewers and inspectors were asked to work as they would
when on their own. All aspects of the field work to be undertaken by the
interviewers and inspectors during the survey were included in this
training programme. Before the field training was concluded, the filled-in
schedules were scrutinised and discussed in the presence of the trainees.
Field Organisation
Survey teams for work in the three rural zones and the medium-sized
towns were made up of one inspector and two interviewers. Teams of
one inspector and three interviewers worked in Bangalore City, because
the volume of work for the inspectors was less in Bangalore City than in
the rural areas. (The Village Survey, for instance, did not extend to
urban areas.)
The central oifice in Bangalore City kept in close touch with the work
of each survey team. Each interviewer had to communicate with the
Bangalore office immediately on completing the listing of households in
a village and had to send a duplicate copy of the list, along with an
indication of the households to be included in the sample for the survey.
The interviewer was also required to maintain a daily diary on a prescribed form and to despatch it to the Bangalore office on two specified
days each week. The inspector's daily diary was to be dispatched once
a week. The inspector had also to report to the central office any (necessary) change in the programme scheduled for his survey party. Each
filled-in Household Schedule and Village Schedule received from the
inspectors was examined by the scrutiny staff and discrepancies, if any,
were referred back to the inspectors in the field for clarification and
1
At this stage, the scrutiny assistants were made to work as interviewers. Training
in scrutiny work was given to them in the central office.

166

FAMILY LIVING STUDIES

correction. Survey work within Bangalore City was organised along the
same Unes as in the rural areas except that, as the central office was
within easy reach, the interviewers and inspectors visited the office for
resolving any difficulties that arose during the course of their field work.
Public Co-operation
The field work of the Mysore study was begun with some misgivings
about the extent of co-operation that could be expected from the public.
As the 1951 Population Census had been conducted only a few months
earlier, it was doubtful whether the public would be willing to submit to
more interviews within so short a time. Besides, the schedules used,
particularly the Fertility and Attitude Schedule, tried to obtain data on
items that are usually considered personal, thus providing an occasion
for mischievous persons to start unfavourable rumours. Actual experience, however, belied these fears.
The extent of public co-operation can be gauged from the small
percentage of persons who refused to be interviewed. In the case of the
Household Schedule, 1 per cent, of households in the rural zones and
2 per cent, in Bangalore City could not be interwieved for all reasons
taken together. These included not only refusals but also such reasons as
" households temporarily away ". In regard to the Fertility and Attitude
Survey also, frank refusals to be interviewed amounted to about 1 per
cent. only. Specific questions on the schedule could not in some cases be
answered by as many as 10 per cent, of the persons interviewed but these
were for reasons mostly other than that the questions were " too personal ".
The interviewers were in general able to obtain better co-operation
from the public in rural areas than in towns and Bangalore City. Part of
this difference was due to the fact that conditions in the villages favoured
the establishment of an excellent rapport between the field staif and the
public. Often, the interviewers were required to reside in the villages for
the entire duration of the survey and had ample opportunities for getting
well acquainted with the villagers.
Tabulation and Analysis of Data
The detailed analysis to be undertaken made it necessary to carry out
numerous tabulations of the data obtained in the Household Survey.
The Indian Statistical Institute, Calcutta, which has a wide range of
equipment for mechanical tabulation, assisted in this task. The tabulations made it possible not only to get a cross-sectional picture of the

POPULATION SURVEY IN MYSORE STATE

167

demographic characteristics of the population surveyed but also of the
changes that had been taking place in the community. Trends in age at
marriage and in literacy and education were, for instance, studied by crosstabulating the age at first marriage, or literacy or education status,
according to the present age of the individuals. Such cross-tabulations
can at times serve as substitutes for repeated surveys and are likely to
prove more economical. However, they call for a careful selection of the
items on which information is obtained and involve a detailed tabulation.

USE OF THE SAMPLE SURVEY IN OBTAINING DEMOGRAPHIC DATA

An important consideration in the adoption of the sample survey
technique for producing demographic data is the quality of the information obtained. The purpose of analysing the relations of social and economic factors to mortality and fertility can only be served if reasonably
accurate reports on such basic factors as births, deaths and the number
of children born are available. Experience in India and elsewhere had
shown that data on births and deaths obtained by household interviews
may be affected by large errors. In one survey carried out in India, not
more than 70 per cent, of births and 50 per cent, of deaths during the last
12 months were recorded by interviewers. In the United States, the
National Health Survey, a household sampling inquiry carried out in
1935-36, yielded death rates that were only 58 per cent, of the registered
figures. Attempts to determine the number of children born to women
during their married lives, in the censuses and household samphng
surveys of many countries, have commonly yielded incomplete statistics.
In the Mysore study, special efforts were made to minimise errors in the
reports. As information on the basic demographic characteristics was
obtained mainly on the Household Schedule, the following discussion
pertains for the most part to this schedule.
Some principles of the schedule design are explained below, followed
by a discussion of special aspects in regard to the collection of data on
births, deaths and number of children born.
Special Features of Schedule Design
Formulation of Questions.
It is not an uncommon feature of schedules for household surveys and
population censuses, in India and elsewhere, that most items of information to be obtained are indicated merely by brief titles, the detailed
definitions and explanations being relegated to the interviewers' instruc-

168

FAMILY LIVING STUDIES

tion manuals. This feature of schedule design, which is sometimes
justified on the score that the interviewer corps is highly trained and
experienced, is not always conducive to accurate reporting on any but the
simplest and most straightforward items. Although it would evidently
not be practicable to give all necessary definitions and explanations
relevant to every item on the schedule itself, an advantage is gained by
presenting there the major highlights of the definitions so that the field
worker is kept constantly reminded of them. This was done in the
Mysore study. In addition, the topics were indicated in the form of
questions rather than bare titles. The questionnaire thereby helped to
standardise the procedures whereby the field workers obtained the information, and tended to keep their attention focused consistently on the
essential points of the inquiry, thus reducing errors and variation of
responses recorded. As illustrative of this procedure, the Household
Schedule contained the following two questions in regard to marital
status : " Has the person ever been married ? " and if the answer was
positive " Is the person now married, widowed or separated ? ". By framing
the questions in this manner two important points were brought out: that
the answers should refer to marital status at the date of the survey, and
that only those who had never been married were to be regarded as
single or unmarried.
An important source of error in the case of some data obtained by
interviews is the failure of the respondent to call to mind all the facts
necessary to provide an accurate answer. In regard to items which may
require such marshalling of facts, the accuracy of response may be
improved if the inquiry is put in the form of a series of questions relating
to those different facts. The application of this principle can be illustrated by the inquiry into the number of children born alive to each
woman. It was important to make sure that children who had died or
left the family would not be overlooked. Hence the inquiry took the
form of three questions, calling separately for the number of a woman's
children living with her, the number living elsewhere, and the number
born alive but now dead. When the three questions had been answered,
the interviewer was to present to the respondent, for corroboration, the
resulting totals of children still living and of all children born alive.
Comparison of Responses to Different Questions.
Accuracy may also be considerably improved by requesting the same
information at different stages of the interview, comparing the responses,
and seeking an explanation for any discrepancies. This principle was
applied extensively in the Household Schedule design. During the
course of the interview, the field worker was required to make specific

POPULATION SURVEY IN MYSORE STATE

169

systematic checks between responses to separate questions, and to record
the results of those checks on the schedule. In case any discrepancy was
revealed, additional explanations were asked as necessary to clarify the
situation, and the previous entries were corrected if found to be in error.
For example, the difference between the present ages of husband and
wife was to be compared with the difference in their ages at first marriage
as reported in response to a separate question on the latter point. Lack
of agreement might be explained by one of the partners having previously
been married to someone else, or by the times of the year at which the
two partners had their birthdays. Such explanations, if applicable, were
noted by the field worker. A discrepancy that could not be so explained
called for investigation and correction of the returns for present age or
age at marriage. Some of the other checks that were made are mentioned
below in the description of measures taken to improve the accuracy of
reporting data on births and deaths which occurred after a specific date.
Some ancillary questions as further explained below were finally included
in the Household Schedule expressly for the purpose of providing material
for such checks.
Use of Coded Entries.
For the most part, the Household Schedule called for entries in the
form of numerical codes which, for ready reference, were printed on the
form. In the case of items for which a large number of codes would be
required, or the coding of which would require more time and consideration than the interviewer could afford, descriptive entries were made
and later coded by the inspectors. Such items included vocational
education, occupation and home enterprises.
The Reporting on Births, Deaths and Number of Children Born
Questions on Deaths since 15 September 1950.
A large proportion of the deaths which occur among members of
households fall into two categories : deaths of married persons resulting
in widowhood, and deaths of children born to women in the households.
The combination of topics investigated in the Household Survey made it
possible to introduce checks which would minimise omissions of these
two types of deaths in the answers to questions on deaths which occurred
since 15 September 1950.1
1
Births and deaths which occurred after this date were to be recorded in the Household Schedule. This date was expected to serve efficiently as a benchmark, as a wellknown festival fell on that date.

170

FAMILY LIVING STUDIES

When the question on marital status revealed the presence of a widow
or widower in the household, the interviewer was required to ask whether
the death of the spouse had occurred after 15 September 1950. If so, the
interviewer was required to make sure that this death was not overlooked
when the answers were given to questions on deaths since that date in
the corresponding section of the schedule, which was intended to cover
all deaths that occurred since September 1950.
Regarding the deaths of children, it has already been mentioned that
the series of questions relating to the number of children born to each
woman during her married life included the question " How many
children born alive are now dead?". In the event that one or more
children were found to be dead, the ancillary question was asked : " Did
any of her children die after 15 September 1950?". Here again the
response could be used to correct omissions in the listing of deaths in the
corresponding part of the Household Schedule. Furthermore, for every
live birth reported in answer to the questions on births since 15 September
1950 the respondent was asked: " Is the child now alive? ". This was
helpful in bringing to attention a category of deaths that are often overlooked, i.e., deaths in early infancy.
Another category of deaths that are likely to escape coverage in
household inquiries consists of those deaths that result in the dissolution
or disappearance of households. Examples are the death of a man or
woman living alone, the death of both husband and wife within the
period to which the questions refer, and the death of a household head
whose dependants thereupon go to live with other relatives or establish
new households. Such deaths would relate to households that had
occupied the selected dwelling units at some time since 15 September 1950
but were no longer present at the survey date or to households that had
shared the dwelling units at some time since that date. For the purpose
of ensuring accurate reporting of deaths of this type, a fairly detailed
section on the occupancy of the dwelling unit since 15 September 1950
and a further question on the duration of residence of each individual
since that date were included in the schedule. This information was
referred to both at the data collection stage and when the inspector
carried out his check offilled-inschedules to ensure the coverage of such
deaths.
Special provisions must also be made to assure coverage of deaths in
institutions of persons who are not connected with ordinary households.
Among the persons who die in institutions such as hospitals, maternity
and nursing homes, there are likely to be many who have no local
residence or no residence whatever, other than the institution, and
whose deaths would not be revealed by a survey limited to households.

POPULATION SURVEY IN MYSORE STATE

171

In the Mysore study a special schedule was designed for recording the
deaths of such persons in the types of institutions mentioned. 1
Questions on Births since 15 September 1950.
In order to obtain as complete information on births during the
specified period as possible it would have been desirable to ask every
woman in the household whether or not she had delivered a child during
that period. This procedure could not be followed in the Mysore study
without substantially increasing the cost of the field work. The data
were obtained as a rule by interviewing one responsible member of each
household, and not all members. However, certain other measures were
taken to minimise omissions in the reporting of births.
In the first place, advantage was taken of the responses to the questions on number of children born to each woman during her married
life. For every woman who was reported to have delivered at least one
live or stillborn child, the additional question asked was " Were any of
these children born (alive or dead) after 15 September 1950 ? ". The answer
to the latter question was used as an aid in assuring that all children
born since that date were listed in the section of the schedule relating
to births. The answer was also used as a check upon the fisting of
children now members of the household and upon the recording of their
dates of birth.
Secondly, whenever a maternal death was reported in answer to the
questions on deaths since 15 September 1950, the field worker was
required to verify that the birth involved in that maternal death (and, if
the child also died, its death) was not overlooked.
Thirdly, special provisions were made for the reporting of births to
visitors, to members of other households sharing the dwelling unit, or to
members of different households having previously occupied the same
dwelling unit.
Finally, births to women living in residential institutions such as jails,
schools, hostels, etc., and in non-residential institutions such as hospitals,
maternity homes, etc., were recorded in the special schedules developed
for that purpose in a manner similar to that described above for deaths.
Questions on Numbers of Children Born to Women during Their Married
Lives.
A device used to improve the quality of reports as to the number of
children born to each woman during her married Ufe has already been
mentioned, namely the use of three separate questions on the number of
1
Deaths in residential institutions such as schools, jails, etc., were recorded on a
schedule similar to the general Household Schedule.

172

FAMILY LIVING STUDIES

children still living with the woman, the number living elsewhere, and the
number born alive who had died. In connection with the third of these
questions, an effort was made to avoid the error of including still births
or abortions. This was done by asking for the age of the child at death,
and, in the event that death was reported to have occurred very shortly
after birth, making the appropriate inquiries to ascertain whether the
baby had shown signs of life at birth.
Another device was to compare the reported number of children born
to the woman and still living with her, with the listing of members of the
household by their relationship to the household head or closest relative
in the household. It was usually easy to identify, in the list of household
members, those who were the offspring of the woman in question,
although some difficulties arose, for instance, when the children were
listed by relationship to the father, and the latter had married more than
once. In the process of identifying each woman's children in the household, and making appropriate inquiries, the field worker was able to
obtain checks on the accuracy, both of the reported number of children
born to each woman, and of the household listing. He was required also
to record an explanation of any apparent discrepancies that might
remain so as to facilitate the subsequent scrutiny of the schedules by the
inspector and the office staff.
Accuracy and Reliability of the Survey Data
Completeness of Reporting of Births and Deaths.
If household surveys can provide accurate estimates of birth and death
rates, they serve an important purpose in India and other countries
where vital statistics are lacking or grossly deficient. Steps were therefore
taken for the evaluation of the accuracy and reliability of the data
obtained in the study.
In the first instance, the reports of births and deaths obtained in the
Household Survey for the villages covered by the survey were checked
with the local registers of vital statistics. For this purpose, the following
steps were taken :
For the villages covered by the survey, the births and deaths shown
on the local registration records as having occurred since 15 September
1950 were allocated to the households to which they belonged, and those
belonging to the households included in the sample for the Household
Survey were abstracted together with such particulars as would be useful
for identification.
The resulting list of registered births and deaths belonging to the
sample households were compared with the entries of births and deaths

POPULATION SURVEY IN MYSORE STATE

173

on the household schedules, in order to determine whether or not each
registered event had been recorded by the interviewer and vice versa.
In case of births and deaths found only on the Household Schedule,
a further check was made to ensure that they did not refer to vital events
which occurred prior to 15 September 1950, by comparing these data with
registration records for about six months prior to that date. 1
A cross-check of the Household Survey data with those from the
Fertility and Attitude Survey was made in the areas covered by the latter
survey. The births and deaths reported in the Fertility and Attitude
Survey as having occurred between 15 September 1950 and the date of the
Household Survey were cross-checked with the lists on the Household
Schedules, and, so far as possible, discrepancies were investigated by
fresh visits to the households concerned. This cross-check could be
undertaken in Zone III (rural plains) and in the four strata of Bangalore
City in which the Fertility and Attitude Survey was carried out.
The result of these checks showed that for births in rural areas which
occurred to residents not more than 15 months before the survey took
place, the percentage recorded in the Household Survey was 96, but for
those at a longer interval it was only 86. The corresponding figures for
deaths were 92 and 82 per cent. There was evidence of more deficient
reporting of births and deaths among visitors in the sample households,
and among the members of households that had previously occupied the
dwelling units surveyed but had moved away before the survey date. No
significant relation was shown between the completeness of recording
and the population size of the village, sex, religion or, in the case of
births, the age of mothers. As regards the age of deceased persons the
recording of deaths in the age group 15-44 years was found to be less
complete than for older and younger age groups. As already mentioned,
the survey did not provide direct checks on the accuracy of information
on births and deaths obtained in the medium-sized towns and in Bangalore City. Indirect evidence was however to the effect that, in regard to
births, the degree of accuracy would not be much inferior to that obtained
in rural areas but that, in regard to deaths, the position could be worse.
Accuracy of Data on Number of Children Born.
The accuracy of the data obtained in the Household Survey on the
number of children born to married women during their married life was
1
A cross-check of the data of births and deaths obtained in the Household Survey
for Bangalore City and the " medium-sized towns " with those of local registration
records could not be undertaken in view of the difficulty of abstracting from the
registration records the vital events pertaining to the sample households. Such abstracting would not only have involved a considerable amount of work but was not
likely to be satisfactory because of the lack of careful recording of the addresses of
households to which the vital events referred.

174

FAMILY LIVING STUDIES

also assessed by cross-checking the information with that obtained in
the Fertility and Attitude Survey. Since the latter survey covered only
a subsample of women between the ages of 18 and 33 years, the crosscheck provided information on the accuracy of reporting for only this
restricted group of women. Subject to that limitation, it was found that
the aggregate number of children reported in the Household Survey as
born alive was deficient by not more than 1 per cent, both in Bangalore
City and in the rural plains (Zone III).
Interviewer Differences.
The Latin square design in which the field survey in Bangalore City
had been arranged made it possible to study the interviewer differences in
regard to a number of characteristics obtained in the Household Survey,
such as resident population, resident male population, resident female
population and numbers of births and deaths among residents. Such a
study was done by using the Analysis of Variance technique. A significant difference (at the 5 per cent, limit) among the interviewers was found
in respect of only one characteristic, i.e., the number of vacant dwelhng
units, and that, too, in only one of the four Latin squares in which the
entire field work in Bangalore City had been arranged. It could, therefore, be concluded that there was a high degree of consistency in the data
obtained by different interviewers in the Household Survey in Bangalore
City.
Final Observations
In conclusion, the experience of the Mysore study has unequivocally
demonstrated the potentialities of a sample survey of households for
obtaining much needed information on a number of demographic
characteristics. Further studies in the field will, it is hoped, lead to the
standardisation of procedures appropriate to varying economic and
cultural settings.

X
A Health Survey in Japan
By Takemune SODA, M.D. *

The relevance of health in the daily life of the population hardly needs
stressing. This relation was fully recognised by the Committee on International Definition and Measurement of Standards and Levels of
Living convened by the United Nations, jointly with the I.L.O. and
U.N.E.S.C.O. and in co-operation with F.A.O. and W.H.O., in 1953,
when it recommended that " health, including demographic conditions "
should be included as an important component of the level of living.1
However, appraisal of health conditions of a country or of a community is far from easy. Progress has been made only through sustained
research and experimentation over decades. Projects formulated in this
field from time to time have often remained unimplemented ; there has,
nevertheless, been a gradual accumulation of experience in estimating the
quantum of illness based on a cross-section of the population.
Since it is almost impossible to have recourse to detailed clinical and
laboratory examinations by professional physicians for the purpose of
exploring the prevalence of illnesses and injuries among a cross-section
of the population of a country, or even among the population in selected
areas, a method had to be developed for collecting this information by
means of household interviews by lay interviewers.

BACKGROUND OF HEALTH SURVEYS IN JAPAN

In 1938 a health survey based on household interviews by lay interviewers was carried out in the Takinogawa district (Ku) of Tokyo. This
* Vice-Director, Institute of Public Health, Tokyo. The author wishes to express
his thanks to the staff of the Division of Health and Welfare Statistics, Ministry of
Health and Welfare, Japan, and the Department of Epidemiology, Institute of Public
Health, Tokyo, for their valuable assistance in preparing this article.
1

UNITED NATIONS: Report on International Definition and Measurement of

Standards and Levels of Living, United Nations Publications, Sales No. 1954.IV.5
(New York, 1954).

176

FAMILY LIVING STUDIES

survey was inspired by the sickness survey of the Committee on the Cost
of Medical Care in the United States.
After the Second World War, in order to prepare for an organised
medical care service as part of a comprehensive social security programme in Japan, the Ministry of Health and Welfare introduced a
series of statistical surveys, including the National Health Survey and
the Basic Statistical Survey of the Health and Welfare Administration,
as well as the Annual Patient Census of Hospitals and Clinics, and
monthly statistics of the number of in-patients and out-patients (Monthly
Hospital Report).
The Basic Statistical Survey of the Health and Welfare Administration
is an annual survey of a fairly large sample (190,000 households and
840,000 persons) to ascertain sickness and injury at the time of interview,
besides other items related to health and welfare. It aims at obtaining an
over-all picture of the relation between the sickness among the population
and their living conditions. However, this Basic Survey does not permit
detailed analysis of morbidity data relating to the occurrence of disease
over a period of time, classification of diseases and injuries, etc.
The Annual Patient Census ascertains the total number of patients
receiving medical care in all hospitals and clinics at a certain date,
together with more detailed information on disease as diagnosed by
physicians. The Monthly Hospital Report gives the total number of
patient-days every month throughout the year in all hospitals in Japan
(both in-patients and out-patients). These two series give no indication,
however, of the prevalence of diseases or injuries not treated by physicians.
The National Health Survey will be the main subject of the following
discussion.
OUTLINE OF THE NATIONAL HEALTH SURVEY

Objectives and Sponsoring Agencies
The National Health Survey was started in 1948 and has been carried
out for one month, in November or October of each year, since 1952.1
The survey is conducted by the Division of Health and Welfare
Statistics in the Ministry of Health and Welfare, under the National
Health Survey Regulations based upon the Statistics Law. Other
1
The decision to choose these months for the survey took account of the seasonal
variation of disease and of the conditions favouring the field work of data collection.
•October is the month during which the number of cases of sickness is usually very
close to the annual average. In November, most of the farm population can spare
time to co-operate in the health survey after the rice harvest.

HEALTH SURVEY IN JAPAN

177

bureaux and sections in the Ministry co-operate in carrying out the survey,
as does the Statistical Bureau in the Prime Minister's office. The prefectural governments are asked to co-operate in the survey, but most of the
expenses are covered by the central government. A consulting body, the
Advisory Council on Health and Welfare Statistics, has contributed
much to the development of the survey, especially in the early
stages.
The National Health Survey is conducted for the purpose of keeping
the medical care service system under regular review with a view to its
possible improvement, and aims to obtain essential information on—
(a) the volume and nature of diseases and injuries among the population ;
(b) the method of treatment of diseases and injuries;
(c) the cost of treatment and the method of payment.
Coverage and Sampling Procedures
The National Health Survey extends to the entire population, including male and female, young and old, and to the entire territory of Japan.
This coverage is ensured by means of a random sample of the population
of Japan.
The frame from which the sample is drawn for both the Basic Statistical Survey of the Health and Welfare Administration and the National
Health Survey is given by the latest list available each year of National
Census Tracts covering the whole country. The sample is drawn primarily for the Basic Statistical Survey of the Health and Welfare Administration by a stratified area sampling method from the list given by the census
tracts. Stratification of the communities is effected according to size of
population, and a sampling fraction of 1: 100 applied to all strata.
The sample used for the National Health Survey is derived as a
subsample of the list of areas sampled for the Basic Statistical Survey
simply by including a sampled proportion of 1: 18 of these areas. All
the persons living in the sample areas finally selected are included in the
National Health Survey, and the ratio between the sample population
for this survey and the total population is thus eventually about. 1:1,800.
The actual size of the sample for the National Health Survey was 12,041
households and 50,030 persons.1
The method of sampling has been changed several times since the
initiation of the survey in the interests of greater accuracy, economy
and convenience.
1

The ratios and magnitudes given in this paragraph are those for the 1956 survey.

178

FAMILY LIVING STUDIES

Information Collected
For the National Health Survey, two kinds of schedules are used,
namely the Household Schedule and the Individual Case Schedule, in
addition to certain supplementary forms. Important items of information covered by these schedules are as follows:
Household Schedule—
(a)
(b)
(c)
(d)
(e)
(f)

identification items for area, household and household members;
number of illnesses and injuries during survey period;
whether or not household received public assistance;
type of social insurance coverage of the family;
economic and social status of household ;
family expenditure in cash for survey month.

Individual Case Schedule (for each patient and each case of sickness or
injury)—
(a) identification numbers of area, household and patient;
(b) name of disease or injury ;
(c) date of onset of complaint;
(d) date of end of complaint (if this fell within survey period);
(e) number of days of sickness during survey period ;
(f) whether complaint was treated or not treated;
(g) whether patient died, was cured, or was still sick at end of survey
period ;
(h) date on which treatment began;
(i) number of days for which treatment received ;
(j) number of physician visits, at doctor's office or at patient's
home;
(k) method of payment for treatment;
(I) amount of expenditure for treatment;
(m) method of treatment (hospitalisation, doctor's office call, dentist,
home visit by doctor, home drug, etc.) ;
(n) indirect expenditure occasioned by treatment of complaint.
The information is collected by lay, but trained, interviewers who
visit each household in the sample twice a week for the interviewing of
one person, usually the housewife of the family, who is assigned beforehand as responsible for giving such information on the health of all the
members of the household. As an aid in this survey, the interviewer
makes a request for diaries to be kept individually for all household
members.
When the interviewer finds, at one of his home visits, that a member
of the household is sick or injured, he records the necessary information

HEALTH SURVEY IN JAPAN

179

in a notebook, to which reference is later made in preparing the Individual Case Schedule after the end of the survey period.
Pilot Survey and Pre-testing
Prior to the first main survey in 1948, a pilot survey was carried out
in Fukuoka Prefecture. This was very useful for appraising and reviewing
the survey methods, especially for testing the practical ability of the
interviewers.
For the second and succeeding rounds, the previous rounds served as
a guide in developing more effective and convenient procedures.
The initial rounds were not expected to give accurate results, but
mainly to serve for the development of methods and techniques for the
later rounds.
Organisation of the Field Work
The field staff employed in the survey has worked out at one interviewer per 10-15 sample households.1 The interviewers are usually
selected from among the Health Centre personnel, mostly public health
nurses or health statisticians, sometimes clerks or statisticians in the
town and village health offices.
The Ministry of Health and Welfare holds regional training courses
on the detailed survey procedures for the staff of prefectural health
departments and health centres in the survey areas. The prefectural
health departments have responsibility for instructing or training the
field investigators and for supervising their work during the survey
period.
Every year an " Interviewer's Handbook for the National Health
Survey " is prepared and distributed to each interviewer, providing
detailed instructions on arranging interviews, on noting of sickness data
at the time of interview, on filling out the two kinds of survey schedules,
and on other aspects of the field work.
In most cases, pencils, towels, glasses or cups are distributed to the
participating families as tokens of gratitude for co-operating in the
survey. The cost of such rewards is about 100 yens per family.
Interviewers first visit the sample households during the week preceding the survey in order to notify them and give explanations, especially
as regards the maintenance of health diaries by all the members in the
household. Further visits are made on the first day of the survey period,
on the day after the end of the survey period, and twice a week in
between.
1

In 1956 there were 879 interviewers for 12,041 households.

180

FAMILY LIVING STUDIES

As the National Health Survey is enforceable under the Statistics
Law, the response rates of the sample families are very satisfactory; there
are practically no cases of non-response.
Tabulation of Data and Dissemination of Results
Tabulation is made in the Division of Health and Welfare Statistics
of the Ministry of Health and Welfare. A special classification of diseases
and injuries is used, which is a modification of and can be compared
easily with List C (Special List of 50 Causes for Tabulation of Morbidity
for Social Security Purposes) of the W.H.O. Manual of the International
Statistical Classification of Diseases, Injuries, and Causes of Death.1
The Division of Health and Welfare Statistics publishes the detailed
results of the National Health Survey each year in Japanese in " Report
on the National Health Survey " ; very simple summary tables are contained in an annual publication in English, " A Brief Report on Public
Health Administration in Japan ".
Use of Results
Rates and other relative measures computed from the results of the
survey can serve as useful tools for analysis. Table I shows a selection
of important rates derived from the results of the 1956 round of the
National Health Survey.
METHODOLOGICAL PROBLEMS

Definitions
In general, the definitions of terms used for the National Health
Survey were designed to fit the aims of the survey and to have practical
application within the limits set by the survey procedures. The terms
sickness and injury were defined to include any unusual disturbance of
physical and mental well-being—
(1) treated by physician, dentist, acupuncturist, moxacauterist, or other
therapeutist, or treated with medicines, drugs or other medical
materials ;
(2) (if (1) is not applicable) a case in which the patient in confined to bed
or incapacitated for work for two days or more;
but excluding the following cases:
1

(Geneva, 1957), Vol. 1, p p . 383 ff.

181

HEALTH SURVEY IN JAPAN

(a) normal pregnancy, delivery, Puerperium, and menstruation, not
receiving medical attention;
(b) handicapped conditions (blindness, deafness, dumbness, impairment
of trunk or extremities and mental disturbance) with symptoms
stabilised and fixed, receiving no medical attention 1 ;
(c) myopia, hypermetropia and astigmatism.
TABLE I. SELECTED RATES OBTAINED FROM THE NATIONAL
HEALTH SURVEY (NOVEMBER 1956)
Indigent
families
receiving
public
assistance

Families
Families not
enrolled in enrolled in
social health social health
insurance
insurance

Rates

All
families

Patient rate per 1,000 persons per
month

182.0

263.4

192.1

149.6

Total case rate of sickness per
1,000 persons per month . . .

205.5

289.6

218.0

164.4

Carried-over case rate (from previous month) per 1,000 persons

41.6

151.9

42.0

30.7

New case rate per 1,000 persons
per month

163.9

137.7

176.0

133.7

Sick days per person per month .

1.91

5.27

2.00

1.45

Number of sicknesses per 1,000 persons per day

63.7

175.7

66.7

48.3

Average duration index of sickness

11.7

38.2

11.4

10.9

Cases started medical care by physician per 1,000 persons per
per month

70.6

74.3

83.1

37.1

Hospital admission rate per 1,000
persons per month

2.8

9.8

3.0

1.9

Number of physician visits per
1,000 persons per month . . .

445.1

1,029.5

518.9

246.7

Information on the number of patients refers to the number of
persons who experienced any sickness or injury during the survey
period. This number, expressed as the rate per 1,000 of the total population interviewed, yields the patient rate, referred to in Table I.
The number of cases of sickness and injury is not equal to the number
of patients, because some patients experience several spells of sickness or
injury during a specified period of time. Moreover, the diagnosis of a
1
Special surveys on handicapped persons of various kinds and mental diseases
have been made independently of the National Health Survey in Japan.

182

FAMILY LIVING STUDIES

disease and the determination of its duration present difficulties. It is
not easy, for instance, to decide whether a patient is suffering from
several independent diseases at a certain time or from only one disease
with several complications. It is also difficult to ascertain whether a
patient has been attacked by a new disease after recovering from the
previous one or is experiencing a recurrence of the same disease during
the course of a morbid process.
The number of cases experienced during the survey period per 1,000
persons in the survey population is the total case rate as referred to in
Table I.
The carried-over case rate is the number of cases per 1,000 which
commenced before and were continuing at the beginning of the survey
month. The terms " presence " or " existence " rate are also applied to
the total.
The average duration index is a means of expressing the average continuous duration of sickness, obtained by multiplying the number of
sicknesses per 1,000 persons per day by the number of days in the month,
and dividing the product by the new case rate.
The number of physician visits per 1,000 includes patients' visits to
the doctor's or dentist's office as well as doctors' visits to patients'
homes, but excludes medical consultation or visits to patients in hospitals.
The total number of days on which a patient suffered from a disease
or injury during the survey period is called the number of sick days.
The total of these figures divided by the average number of persons at
risk is called sick days per person during the survey period. This figure
divided again by the number of days of the survey period gives the
number of sicknesses per 1,000 persons per day as shown in the table.
Though the sick days may be counted excluding either the first or
the last day of a sickness, the Japanese health survey includes both in
order to avoid confusion among lay interviewers. The data presented
in Table I were calculated on this basis.
Incidence and Prevalence
Incidence and prevalence are essential concepts in a sickness survey,
although the terms are not always easy to apply in practice.
The concept of incidence relates to the number of new cases occurring
during a specified length of time.
The concept of prevalence, on the other hand, relates to the number
of cases existing and found at a specified point in time. The rate of this
number in proportion to the population investigated is often called

HEALTH SURVEY IN JAPAN

183

simply the " prevalence rate ". The W.H.O. Expert Committee on
Health Statistics in 1958 recommended the term " point prevalence rate "
for this concept. Their recommendations for the use of the terminology
" incidence " and " prevalence " are as follows 1 :
the term " incidence be used to describe the measurement of frequency of
illnesses commencing during a defined period... ;
the term "period prevalence" (or, more specifically, annual, monthly,
weekly, prevalence) be used to describe the measurement of frequency of illnesses in existence at any time during a defined period (i.e., a year, a month,
a week, etc.) . . . ;
the term " point prevalence " (or, more specifically, the prevalence at a
stated point in time) be used to describe the measurement of frequency of
illnesses in existence at a particular point in time... .
A daily census taken during the survey period would provide a direct
means of calculating the average prevalence rate. The rate can also be
computed by the following formula :
Average yy
(point)
,
.'
prevalence rate
during the survey

Sick days per
f.erson d u n n g
the survey
Days of the
survey period

Incidence rate
during the survey
—Days of the
survey period

Though the last term, i.e., the incidence rate per person per day,
is rather small as compared with the other terms, yet it sometimes
amounts to more than 10 per cent, of either of them, and cannot be
neglected in a detailed study of morbidity.2
Besides the above-mentioned terms, there are other important and
useful concepts and terms in a sickness survey, such as disability, number
of disabling diseases, days of disability, average duration of a disease,
medical care received, number of doctors' visits, etc.

COMPLETENESS AND ACCURACY OF SICKNESS SURVEYS BY LAY
INTERVIEWERS

Health surveys by lay interviewers inevitably suffer from lack of
completeness and accuracy in the information on the presence and
occurrence of diseases and injuries and in their classification.
1
W.H.O.: Expert Committee on Health Statistics, Sixth Report, World Health
Organisation Technical Report Series, No. 164 (Geneva, 1959), pp. 6 ff.
2
For further discussion of the relation between incidence and prevalence rates,
as well as sickness days or duration of sickness, see paper presented by the author
at the International Population Conference, Vienna, 28 Aug.-3 Sep. 1959. Takemune
SODA: " Main Features of Findings Obtained in Sickness Surveys of Japan ", in
International Population Conference, Union internationale pour l'étude scientifique
de la population (Vienna, 1959), pp. 519 ff.

184

FAMILY LIVING STUDIES

Notwithstanding this drawback, sickness or health surveys on a
probabiUty sample covering wide areas of a country and using lay
interviewers have a number of advantages ; for example, the representativeness of the data for a large group of population. This cannot easily
be obtained by other types of sickness surveys with more accurate procedures, such as those involving examination by professional doctors.
The nature, the degree and the causes of inaccuracy in information
obtained by lay interviewers in a sickness survey deserves to be studied
in more detail with a view to finding means of increasing the accuracy,
adjusting the data obtained or explaining the data correctly.
Total Volume of Sickness
In the Japanese health surveys, three types of prevalence rates can be
computed year by year, as shown in Table II: the (point) prevalence
rates of sickness per 1,000 population at the beginning and at the end
of the survey period and the average prevalence rate during the survey
period on the basis of monthly (or yearly) data by the formula given
above.
TABLE II. PREVALENCE RATES OF SICKNESS AND INJURIES
Prevalence rate of sickness per 1,000 persons
Survey period

15 NOV.-14 Dec. 1948

At the
beginning
of survey
period

. . . .

22.4 (66.7)

At the
end of
survey
period

Average
during
survey
period

29.1 (86.6)

33.6

Sep. 1949

28.9 (77.9)

32.9 (88.7)

37.1

Feb. 1950

28.3 (79.1)

30.3 (84.6)

35.8

June 1950

1A.1 (81.2)

28.5 (93.7)

30.4

40.3 (103.9)

21.Q (69.6)

38.8

May 1951-April 1952

. . . .

Nov. 1952

31.7 (79.4)

39.9

Nov. 1953

46.1 (73.1)

50.5 (80.0)

63.1

Oct. 1954

43.8 (73.2)

46.3 (77.4)

59.8

Nov. 1955

37.9 (71.4)

42.2 (79.5)

53.1

Nov. 1956

41.6 (71.4)

48.1 (82 J)

58.3

1
Italic figures in parentheses represent the ratio ( %) of the rate shown to the average prevalence
rate for the whole survey period.

HEALTH SURVEY IN JAPAN

185

An examination of the annual change of each prevalence rate shows
a progressive increase in the initial years of the survey, until by and large
a fairly stable level was attained.
The fact that more cases of diseases and injuries have been reported
in recent rounds is probably due to the improvement of survey methods,
the establishment of a legal organisation to carry out the survey, and the
increased training and experience of interviewers.
Of the three types of prevalence rates, the average prevalence rate
during the survey period is found to be generally the largest; the prevalence of diseases and the injuries at the end of the survey period comes
next, and the prevalence rate at the beginning of the survey period is
found to be the smallest almost every year.
One-day data are apt to give smaller figures of prevalence rate than
the average computed from data covering a longer time period. Even
in one-day surveys, the longer the preparatory stage, the more complete
the information collected on the prevalence of diseases. The problem of
the significance of the length of a survey period in a sickness survey is
discussed below.
Though the National Health Survey has given differing prevalence
rates of sickness according to the survey method, the relations between
the different types of rates have remained relatively constant since 1953,
and allow to some extent the conversion of one type of rate to others.
Though the average prevalence rate computed from data for the whole
month gave the largest figure of prevalence, a fairly large number of
cases of diseases and injuries still remain unreported.
The National Health Survey shows that the average number of consultations made by in-patients and out-patients with physicians and
dentists per 1,000 persons per day during the month of November 1956
was 2.06, while such figure reported from hospitals and clinics in the
Patient Census 1 on 14 November 1956 was 3.15, the former figure being
only two-thirds of the latter.
Classification of Diseases and Injuries
The problem of the bias which is apt to arise in the reporting of
various types of diseases and injuries is illustrated in Table III.
This table gives a comparison of the numbers of consultations with
physicians and dentists according to selected diseases or disease groups
1
The Patient Census referred to above lists the total number of in-patients and
out-patients who visited physicians' offices, had physicians' visits at their home, or
were being hospitalised on a designated day. The census is made annually on sample
institutions (hospitals, clinics and dental clinics) selected from the complete list of
medical institutions, compiled anew each year upon the hospital and clinics reporting.
The completeness of enumeration is considered to be highly satisfactory.

186

FAMILY LIVING STUDIES

TABLE HI. MEDICAL CONSULTATION RATES *• BY SELECTED DISEASES
F O R T H E NATIONAL HEALTH SURVEY (NOVEMBER 1956") A N D THE
PATIENT CENSUS (14 NOVEMBER 1956)
Medical consultation rate
Selected diseases

National
Health
Survey
0»)

Patient
Census

(a)
7b)

Proportion of total
visits for different
types of d'.seases
x

100

(b)

National
Health
Survey

Patient
Census

All diseases

2,062

3,145

65.6

100.00

Tuberculosis

295

406

72.7

14.31

12.91

13.0

0.15

0.73

Malignant neoplasms

75.0

0.73

0.64

Mental, psychoneurotic, and personality disorder

38.9

1.36

2.29

Diseases of eye

108

85.2

4.46

3.43

Diseases of ear and mastoid process

102

73.5

3.64

3.24

Diseases of nose and sinusitis

98.4

3.06

2.03

Diseases of female genital organs

41.8

1.12

1.75

Diseases of the skin and cellular
tissue

117

231

50.6

5.67

7.34

Accidents, poisoning and
lence

166

179

92.7

8.05

5.69

Venereal diseases

vio-

Number of consultations with a physician or dentist per 100,000 persons per day.

as obtained by the National Health Survey with corresponding data
collected by the Patient Census.
As a whole, the number of consultations reported by the National
Health Survey is less than that enumerated by the Patient Census, and
the percentage of the former to the latter is 65.6, i.e., about two-thirds.
In the National Health Survey, tuberculosis, malignant neoplasms,
diseases of sensory organs, and accidents are on the average, more
completely reported than other diseases, while venereal diseases, psychoses
and diseases of genital organs, especially of women, which people tend
to conceal and diseases of the skin, most of which are minor diseases, are
less completely reported than others.
The percentage distributions by disease or disease group of the medical
consultations according to whether these took place at the patient's home,
at the doctor's cabinet or in a hospital or clinic, show little variation.

HEALTH SURVEY IN JAPAN

187

Length of Survey Period
Most of the sickness surveys in Japan have been conducted to cover
a period of one month each time. It is quite clear that one-month surveys
do not give any information on the seasonal variation of the volume
and structure of prevalent diseases and injuries, on repeated attacks to a
person during a longer period, on the duration of each disease, especially
of a chronic disease, etc.
In order to obtain such additional information, a one-year health
survey was conducted from May 1951 to April 1952, on the same sample
population. The result of this survey was not satisfactory, however.
In spite of a more complete enumeration of sickness in the first few
months of the survey, the completeness of enumeration tended to decrease month by month afterwards, probably due to the declining
interest and enthusiasm, by the sample population as well as by the
investigators.
The author is rather of the opinion that information on seasonal
variation of sickness throughout a year may be more rehably and easily
secured by repeated monthly surveys with the sample population
changed or rotated as in the United Kingdom and the United States
surveys. In Japan, there is a great interest in the feasibility of a sickness
survey of a shorter period of time, in a one-day survey particularly. If such
an inquiry gives fairly reliable data, it may enable us to repeat the surveys
several times during a year with less monetary expenditure and efforts.
Table IV presents a comparison of the proportions of each disease
or disease group among the total number of all diseases and injuries,
based upon one-day data and those based on monthly data.
The reporting of each disease for a one-day (cross-section) reference
is, in general, less complete than for a monthly reference period, but the
proportion of each disease to the total does not so vary though it is
recognised as a general tendency, in the one-day survey, for more serious
or prolonged diseases to be more completely enumerated than minor
and transitory diseases and injuries.
On the other hand, the tendency for some diseases (such as venereal
diseases and psychoses) to be neglected intentionally by the respondents,
as already illustrated in Table III, remains unchanged, in both the oneday survey and the monthly survey alike.
In this connection it is necessary to bear in mind the difference
between a survey made by one-day interviews to obtain health information for the preceding month or year—that is a survey of retrospective
nature, and the survey made by one-day interviews to get health information of the sample population at the time of interview, or at a certain

TABLE IV. COMPARISON O F P O I N T PREVALENCE RATES F R O M ONE-DAY A N D
DISEASE G R O U P (NOVEMBER 1956)
Point prevalence rate 1

One-day prevalen
of the

Selected diseases and injuries

At the
beginning
of survey
period

At the
end of
survey
period

Average
prevalence
during
survey
period

Day at the
beginning
of survey
period

All diseases and injuries
Tuberculosis
Venereal diseases
Other infective and parasitic diseases . . .
Malignant neoplasms
Mental, psychoneurotic, and personality disorder
Diseases of eye
Diseases of ear and mastoid process . . .
Diseases of nose and sinusitis
Other diseases of respiratory system
. . . .
Diseases of digestive system
Diseases of female genital organs
Diseases of the skin and cellular tissue . .
Diseases of the bones and organs of movement
Symptom, senility, and ill-defined conditions
Accidents, poisoning and violence
. . . .

4,158
546
8
156
28

4,805
572
8
132
30

5,829
570
8
198
33

71.3
95.8
100.0
78.8
84.8

48
140
102
76
464
768
46
216
168
154
236

54
172
110
100
558
865
60
270
186
142
298

55
207
128
107
911
1,104
60
325
243
213
399

87.3
67.6
79.7
71.0
50.9
69.6
76.7
66.5
69.1
72.3
59.1

Number of cases of sickness or injury per 100,000 persons.

Da
e
s
p

1
1

HEALTH SURVEY IN JAPAN

189

time within 24 hours before the interview. The former procedure is
usually adopted in the United Kingdom and in the United States, while
the so-called one-day survey which was carried out in Japan belongs
to the latter category.
Memory Factors and the Use of the Health Diary
Another reason why the one-day survey is preferred in Japan is the
scepticism as regards the memory of the informant, who often fails to
give detailed information on the condition of his health long before the
interview. It seems necessary, therefore, that if a survey has to rely on
the memory of the informants, the period to be recalled should be as
short as possible, the diseases and injuries under the survey must be
limited to more distinct or serious ones, while the investigator must
exercise great skill in interviewing, with well-designed questionnaires.
Even the Japanese sickness surveys as a whole depend upon the
memory of informants extending for a few days or a week before the
interview, in spite of the fact that interviews during the survey period
are frequent. However, the first contacts with the households are made
at the time of or even before the start of the survey period, as the demographic and socio-economic items on the survey schedules are filled in.
Follow-up interviews are aided by the records in health diaries which
all the family members have been asked to keep every day during the
survey period.
The method of recording sickness and injuries existing at the time
of the interview or at a certain time within 24 hours before the interview,
which is partly or entirely adopted in the National Health Survey and
the Basic Statistical Survey for Health and Welfare Administration in
Japan, has a definite advantage over other methods in that it involves
no dependence on memory.

CONCLUSIONS

The methodological problems have been discussed here on the basis
of the " prevalence " of sickness and injuries, but not with reference to
" incidence ". In the main, however, the situation is almost the same
in both cases, and it is therefore not necessary to discuss the special
features of the incidence statistics of diseases and injuries.
The experience in Japan has clearly demonstrated the usefulness of
sickness surveys by lay investigators, even though the methodology is
still incomplete and has to be further improved.
13

190

FAMILY LIVING STUDIES

At the present stage of development of the household sickness survey
by lay interviewers, it is almost impossible to compare the national
data of one country with those of others. However, by adopting uniform
methodology, definitions of terms, interview procedure, etc., there are
grounds for believing that data on the prevalence or incidence of diseases
and injuries can be made to serve as one indicator of the health component
of the level of living, at least for selected population groups.
The author feels that the methodology of household surveys of health
conditions should be developed to obtain more useful (uniform, correct
and complete) data for specific purposes, especially in the light of
parallel studies with a mass examination by professional physicians using
modem medical techniques, and also with reference to various kinds of
socio-economic data. This development could be realised only by the
accumulation of scientific and practical experience in various countries
under conditions of close international co-operation.

XI
Repeated Surveys of Consumer Finances
in the United States
By James MORGAN *
Organisation of the Surveys
In the United States, annual data have been collected since 1947
through personal interviews with the heads of some 2,800 to 3,600
spending units on the financial conditions, plans and recent activity of
consumers. Interviewing is done in January-March of each year, with
questions largely covering current financial balances and incomes and
major expenditures during the previous year. The spending units that
are visited in the surveys are selected by area probability sampling
methods to represent the population in private dwelling units in continental United States.
Until 1960, these surveys of consumer finances were conducted for
the Board of Governors of the Federal Reserve System by the Survey
Research Center of the University of Michigan. This rather unusual
feature of the surveys—the collaboration of a public agency, the Federal
Reserve System, with a university-connected research group of an interdisciplinary character, the Survey Research Center—has facilitated
flexibility and experimentation while maintaining a focus on major
financial aspects of the consumers' situations. From the beginning a
basic core of important financial and demographic data have been
obtained. Special topics or features that differ from year to year have
been added, and there has also been a continual improvement in survey
techniques. Since 1960, the survey has been conducted by the Center
in consultation with an inter-university executive board of economists,
and withfinancialsupport from the Ford Foundation and from business
and trade associations. In addition, the Ford Foundation grant has
allowed for summer workshops, preparation of a library of microfilms
* Program Director, Survey Research Center, University of Michigan.

192

FAMILY LIVING STUDIES

and I.B.M. cards, and other means by which the data could be made
more easily accessible to research workers. A booklet describing the
archival resources of the programme, including an extensive bibliography, has been prepared. In the first volume of an annual publication
the Center presents the major findings of the 1960 Survey of Consumer
Finances, as well as findings of other economic surveys.
From the beginning, several professional people have been assigned
to the Survey of Consumer Finances in the Economic Behavior Program of the Survey Research Center. Moreover, the professional
people in the Field Section, Sampling Section, Coding Section and in
Data Processing Sections of the Center have also contributed to the
survey. With the specialisation and division of labour at the Center it
is difficult to speak in terms of numbers of people, since most people are
involved in more than one project. In addition, several people from the
research staff at the Federal Reserve Board in Washington were concerned
with the surveys.
The articles on the surveys that appeared in the Federal Reserve
Bulletin were written by the Washington staff in consultation with the
Michigan staff. The Federal Bureau of the Budget approves the questionnaires. The main findings are published each year in a series of articles
in the Federal Reserve Bulletin, but as more and more people have
learned how to use them, the data have been made available to a growing
number of researchers.
Purposes of Financial Surveys
Even in the United States, where a large volume of recorded data
exists, some kinds of information can be obtained only from the decisionmakers themselves, whether consumers or businessmen. Even simple
questions of" how many ", e.g., how many people have any life insurance,
or more than one car, cannot always be answered from available records.
Survey data can answer two further types of questions as well.
First, they can answer the question " who ", e.g., what kinds of people
use instalment credit ? Are debtors people with no assets and low
income ? Whose income has been going up ? Which families have
been affected by unemployment ?
Secondly, properly designed surveys can answer questions of " why ".
In many cases the answers to the " who " questions also allow one to
infer why. The availability of some attitudinal information, as well
as demographic and other financial information, makes such analyses
more fruitful. We have found that instalment debt is most frequently
incurred by young parents in the middle-income groups. We also find

CONSUMER FINANCES IN THE UNITED STATES

193

that older people are more likely to have a negative attitude towards
the use of credit.
Surveys can thus provide answers to questions of " how many ",
" who ", and " why ". They can do this best for activities and conditions
that are relatively common and important to the individual. Survey
data can tell us much more about car buying or instalment debt, for
instance, than about the buying of common stock, which is too rare to
show up in more than a few cases in a sample of the population as a
whole, or about details of small expenditures which are difficult for the
respondent to remember. Hence, the surveys of consumer finances have
concentrated on items of importance to large groups of consumers, with
other financial, demographic, and attitudinal data as explanatory
variables.
In addition, repeated surveys with similar techniques conducted
over periods of changing incomes, prices and levels of employment, can
throw a great deal of light on the differential impact of major events on
different groups of consumers, and their reactions to these changing
circumstances in their attitudes and behaviour.
As the time span covered by the surveys increases, more and more
use can be made of " cohort analysis "—examining changes in the behaviour of people born during the same period (e.g., 1905-14) who will be,
e.g., 35 to 44 in 1949, and 45 to 54 in 1959. A major problem faced by
demographers as to whether behaviour is age-specific or cohort-specific
needs to be solved for many types of economic behaviour, e.g., use of
instalment credit and ownership of automobiles.
Samplingl
People living in private dwelling units in continental United States
comprise the population covered. Thus, the sample frame excludes
persons living in institutions, of whom there are probably one and a
half million or so in the United States, as well as persons in " quasihouseholds "—large rooming houses, Y.M.C.A.s, residential clubs,
dormitories, hotels and other similar places—of whom there are probably
another 2 million. It also excludes all persons living in dwellings on
military reservations.
In keeping with the varied purposes and objectives of the surveys,
they are based on an area probability sample of private households,
stratified, multi-stage and with interlocking controls in the selection of
the 66 primary sampling units (PSUs).
1
For a more detailed discussion of the sample, see: Leslie KISH and Irene HESS,
" The SRC National Sample of Dwellings " (mimeographed) available at the Center.

194

FAMILY LIVING STUDIES

The basic sampling frame is the occupied private dwelling units of
the United States. No lists of people or families or of dwellings as such
are available. The sample is therefore an area probability sample, with
dwellings sampled by sampling geographic areas, and with spending
units attached each to some dwelling. There are 66 primary sampling
units: the 12 largest metropolitan areas represent themselves in the
sample, while each of the 54 other PSUs is selected with probabilities
proportional to population, from a stratum relatively homogeneous as
to size of largest city, region, major type of farming, percentage of
Democratic vote in the North and percentage of the population that is
non-white in the South. Beyond stratification, the selection of the
particular single primary sampling unit (county or county group) to
represent each stratum is controlled to improve the distribution among
the states and the degree of urbanisation.1
The secondary selection stage involves selecting three to six areas
within each primary sampling area, using five substrata in the 12 metropolitan areas, five to eight substrata in other PSUs involving a central
city of 50,000 or more, and three substrata in all other places. The urban
places often fall in with certainty once the primary unit is selected, as
does the residual open country area. The smaller places (towns, etc.)
are sampled with stratification controls on the basis of geographic
location, size, and rate of population growth.
At the next stage, one can distinguish places where city telephone
directories are used and where they are not used. In the former, groups
of five, 10 or 20 lines are selected from the directory and checked in the
field. The resulting sets of addresses may be clustered into segments of
four in the larger cities, or sampled for individual addresses elsewhere.
Where no directory is available, chunks of area are selected, estimated
to contain 20 to 30 dwellings each. These are then checked in the field,
and segments of four dwellings each subselected for a survey. In highly
populated areas without directories, the block-listing technique is still
used, where whole blocks are listed, but in order to improve the efficiency
with which new construction is sampled, whole sections called " block
supplements " are checked to look for new construction, with a higher
initial probability and hence a much lower proportion selected if construction is found.
The 66 primary sampling points change very little from year to year.
Some early changes were made in the sample design so that in addition
to national estimates, the four regions of the country (Northeast, North1

See Roe GOODMAN and Leslie KISH: " Controlled Selection—A Technique in
Probability Sampling ", in Journal of the American Statistical Association (Washington,
D.C.), Vol. 45, Sep. 1950, pp. 350-372.

CONSUMER FINANCES IN THE UNITED STATES

195

central, South, and West) could be analysed separately. More recently,
some adjustments in primary sampling points have taken account of
newer census data on population distributions. Changes are made on a
controlled probability basis.
An important advantage of sample surveys is that one can in advance
decide on the amount of accuracy required, and then on the sample size
required to reduce sampling errors to the desired level or to a desired
level per dollar of costs. In complex financial surveys, however, there
are also response errors which are likely to increase as one increases
the sample and is unable to use a small, highly trained and well-controlled
group of interviewers.
Samples of around 3,000 were chosen as the best compromise providing acceptable sampling and response errors.1 With year-to-year
similarity in method, such a sample is large enough to discover important
changes, yet data from different years can be combined if the situation
does not change substantially.
In each survey, up to and including the 1958 survey, different sampling
fractions were used according to the value of the dwellings. The smallest
sampling fraction was used in selecting low-rent or low-value dwellings,
a larger fraction for medium-value dwellings and the largest fraction for
higher-valued dwellings.
The data were then weighted to adjust for differences in both sampling
and response rates as between these three groups, using a two-digit
weight on each punch card. There were, in addition, complex procedures
for assigning some missing information on the basis of other characteristics of the spending unit. If both income and liquid asset information
were lacking, the interview was not used; otherwise, income, or any one
of four components of liquid assets, or personal debt, mortgage debt,
or house value, may be assigned. These assignments applied, however,
only to a very small fraction of the cases.
In the 1959 and 1960 surveys, a uniform sampling fraction was used,
but the data are still weighted to adjust for differences in response rates.
The use of assignments has been continued and expanded to include
contingencies such as whether or not a car is owned.
In 1961 and 1962, addresses used in the 1960 survey will be revisited,
dropping half of the lower income addresses. This provides a cross1

In the Survey of Consumer Finances, proportions around 50 per cent, based
on the whole sample which differ by 3 per cent, or more from one year to the next
are significant. For proportions around 10 per cent., a difference from one year to
the next of around 2 per cent, is significant. By " significant " we mean that there is
less than one chance in 20 that the difference could have arisen from chance when
there was really no change. Further reductions in sampling errors require expenditures
to increase the sample, with less and less added precision per dollar.

196

FAMILY LIVING STUDIES

section each year, since those who move in are interviewed. It also provides reinterviews with those who do not move away. The 1961 survey
concentrates on earnings history and longer range expectations for the
future. The third interview in 1962 will provide net worth and thereby
saving over a two-year period.
Data Collection by Interviewing
Once an address has been selected, it may prove to have more than
one dwelling unit x and a dwelling unit may contain more than one
spending unit.
The spending unit which is the basic unit for interviewing and for
much of the tabulation is defined as a group of people living together,
related, and pooling their income for major items of expense. All
secondary units at the selected dwelling units are interviewed. In addition
to primary families (or individuals) there are some 10 per cent, of dwellings with " related secondary units " (with separate income and finances),
and some 3 y2 per cent, with " unrelated secondary units " (roomers,
boarders, etc.). For some tables, however, the primary and the related
secondary units are pooled to put the data on a family basis. For each
unit, the head of the unit is the designated respondent, but if he is not
available, and the wife or other adult is familiar with the family finances,
an interview is taken with that member. The head is interviewed in
more than 80 per cent, of the cases.
It is felt that for some questions, such as those relating to automobiles,
assets, debts and buying plans, the spending unit is a better interviewing
and tabulating unit than the family, both conceptually and practically.
For housing and income much of the tabulation has been on a family
basis, but even here it is felt that the spending unit is a better unit for
interviewing.
The Center at Michigan maintains a national staff of interviewers,
of whom about 165 take part in each survey of consumer finances.
They are selected and trained by a staff of travelling supervisors, and
paid on an hourly basis. They also do interviewing on numerous other
surveys conducted by the Survey Research Center, all of which require
careful and uniform use of the fixed-question, open-answer technique.
The time for each interview in the Survey of Consumer Finances averages
about an hour.
Interviewer turnover is relatively low, with 70 per cent, of those
interviewing in one survey of consumer finances also interviewing on
1
A dwelling unit is defined, as in the United States Census, as a unit with either a
separate entrance or separate cooking equipment or both.

CONSUMER FINANCES IN THE UNITED STATES

197

the next one, and nearly half interviewing on four successive surveys.
The continuity of primary sample units and of interviewers makes for
greater reliability of estimates of year-to-year change.
Pre-tests are conducted each year to develop the new sections of the
questionnaire and ensure that the total length is still reasonable. Ambiguous questions and sequences confusing to the interviewer and
respondent are altered or eliminated as a result of such pre-test experience.
Response rates, adjusted for different sampling fractions and for
estimated secondary units in dwellings where no one was contacted,
have ranged from 81 per cent, to 88 per cent. Reinterviews, necessitating
the use of names, and complex financial checks, appear to reduce the
response rate. About half of the non-response results from inability of
the interviewer to contact anyone at the dwelling unit. Various studies
of non-response have indicated that it is somewhat higher in the upperincome groups and among secondary spending units, and adjustments
are made by weighting for some of these biases. Response rates would
be higher except for the fact that each spending unit must be interviewed
separately, and that the standards for accepting the interview (complete
income or liquid assets data or both) are high. The non-response is
higher among secondary spending units, but it is felt that data on their
financial affairs elicited from the primary units at these addresses would
not be sufficiently reliable.
A letter, and in recent years a booklet, is sent to " Head of Household " at each sample address in advance, explaining the purpose and
importance of the survey, and a small " thank-you " card is left with the
respondent by the interviewer, containing the University of Michigan
seal and the name of the interviewer. Respondents who desire them are
later sent reprints, formerly of the Federal Reserve Bulletin articles and
now of a summary report, but they receive no other recompense.
Questionnaire Content
In the actual interviewing, the fixed-question, open-answer form has
been used, except of course where simple facts were asked. At the same
time, a balance was sought between extremely detailed questioning on
fewer topics, and broader investigation of a wider variety of topics.
Attention was focused on major facts or past behaviour that the respondent could be expected to remember. Generally, questions refer
to the situation at time of interview, to activity during the past 12 months,
or to expectations for the next 12 months.
Hence, the major consumer activities studied in the surveys are
important discretionary outlays such as expenditures on durable goods

198

FAMILY LIVING STUDIES

and cars, additions to bank deposits or bond holdings, rather than
expenditures on food and clothing. Even in the case of expenditures on
durables, no attempt was made to be exhaustive, and reliance was placed
on a few questions to secure approximate amounts rather than very
detailed information.
The questionnaire is the result of 14 years of development during
which pre-tests, split-half experiments \ validity checks, and changes
from year to year on format and question wording were used in an
attempt to secure the best instrument. Inevitably, some omission of
detail must be allowed in order to secure the broad coverage of topics.
The present questionnaire has the following general structure:
(a) Composition of the household, determination of spending units
and of changes in family composition over the past year.
(b) Attitudinal questions on recent financial experience and expectations, financial optimism. These serve also as an easy introduction
to the rest of the questionnaire.
(c) A section on housing, house value, mortgage debt and payments,
and plans to move, buy a house, or make additions or repairs.
(d) A section on cars—ownership, purchases in the last year, debt and
payments, plans to purchase.
(e) A section on household appliances and furniture (past and planned
purchases, debts, and payments).
(f) A section on debts not already reported, and rates of repayment.
(g) Occupation of head, recent employment experience, earners other
than the head, and spending unit income by source and recipient.
(This is a rather detailed and extensive income schedule, with
separate brief sections for farmers and owners of businesses. It
concludes with questions on recent changes in spending unit income.)
(h) Liquid assets (bonds and bank accounts) at time of interview and a
year earlier, with details on who owns and number of accounts.
(i) A final section of demographic information : age, education, marital
status, etc.
It can be seen that the questionnaire builds up to the more difficult
and sensitive information. There are also usually a few attitudinal
questions interspersed to relieve the monotony and prevent fatigue of
the respondent.
1

Split-half refers to using one form or set of questions with half the respondents,
and another with the other half. Differences have usually been not significant, and
are not as useful as detailed validity checks.

CONSUMER FINANCES IN THE UNITED STATES

199

A unique characteristic of these surveys is the combination of financial
data, demographic data, recent experience, plans, and attitudes, all for
the same people. Other data based on larger samples are available for
some of these items, but nowhere else are they all available for joint
analysis.
Coding and Analysis
In the tabulation of the data, a complex editing process first (a)
checks out the consistency of reported transactions, e.g., between total
price, down payment, trade-in, and amount borrowed, (b) makes
estimates of income taxes on the basis of taxable income and exemptions,
and (c) adds up components of debts, assets and income, checking for
possible duplication, etc.
During the coding operation, a decreasing proportion of interviews
are check-coded both to improve accuracy and to ensure uniform coding
procedures. (At the start, all the schedules are check-coded, later only
a fraction. In total, about 10 per cent, are check-coded.) For some of
the attitudes and buying plans, all the previous year's questionnaires are
recoded to ensure against possible changes in interpretations of doubtful
answers from year to year. If this year's coders are coding differently
from last year's coders, a comparison of the original and recoded data
will reveal the problem. A measure of coding reliability is also provided
by these data.
Cards are coded also for the non-response cases, and a careful check
of a machine listing against a master sample control list is made to
ensure that every address in the sample is accounted for. In estimating
response rates, an allowance is made for secondary units which might
have existed at dwellings where no one was contacted. After the interviews and worksheets are coded and punched on tabulating cards, an
elaborate set of" consistency check " operations is conducted to eliminate
remaining inconsistencies.
Efficient techniques have been developed for producing weighted
tables, weighted means that take account of cases where amounts are
not ascertained or exceed the capacity of the field on the punch card,
medians, etc. A programme for the electronic equipment produces
weighted cross-runs on as many as nine pairs of variables, giving weights,
card counts and percentages within groups characterised by one control
variable, the other control variable, and both jointly. In 1960, the Center
started using tape and larger capacity equipment, but this will probably
require prior assignment of all missing information for simplicity and
for efficiency.

200

FAMILY LIVING STUDIES

The analysis has of course varied with the changing emphasis from
year to year, but there has been a continuous core of basic data published
each year in the Federal Reserve Bulletin, and since 1960 by the Center,
and a much larger set of basic tables prepared each year both as an aid
in the interpretation of the published tables and as a service to analysts.
Reliability of Data
In addition to these improvements in coding and analysis procedures,
there has been a gradual evolution of techniques both in collecting and
in processing the data. There have also been methodological checks of
various sorts as to the adequacy of the data. Detailed case-by-case
checks have been made on house value, income, some items of debt,
and savings accounts. Reinterviews provide information on consistency
of report on the same facts a year later, e.g., liquid assets, or first-year's
income. Studies have been made of the importance of extra calls to
reduce the number of " not-at-homes ", and of the problem of " missed
dwellings " never listed by interviewers.
There are various ways to assess the reliability of survey data. Some
of these—internal consistency, difficulties and resistances reported by
interviewers, evidences of misunderstanding of questions—are difficult
to quantify. The most revealing, but most expensive, kinds of checks
are those where outside information on the individual unit is available.
Checks of this sort revealed, for example, that the income data were
rather good but could be improved by a more direct approach to income
of extra earners. This change was instituted in the 1954 survey. Independent appraisals of house value compared with owners' estimates
revealed some discrepancies but no apparent bias. Checks on car debt
comparing outside data and interview reports showed failures to report
existing debts in a small but appreciable fraction of the cases. Checks
on savings accounts, though subject to special problems and the use of
interview techniques not sufficiently similar to the surveys of consumer
finances, revealed a substantial number of response errors, with some
tendency towards over-all under-reporting.
Reinterview comparisons of reports at the time and a year later on
liquid asset holdings and income in the year prior to the first interview
revealed some tendency towards memory errors that reduced the reported
amount of change in assets or income.
Where comparisons with other surveys based on probability samples
have been possible, the results have generally been good.
The most common, and most difficult and treacherous, method of
assessing survey data is to estimate population aggregates and compare

CONSUMER FINANCES IN THE UNITED STATES

201

these with aggregate estimates derived from other sources. One must
adjust for conceptual differences, for differences in the populations
covered, etc. For example, as noted earlier, the survey of consumer
finances excludes not only the institutional population, but also all
persons living in dwellings on military reservations and all " quasihouseholds ". The " spending units " to which the data obtained in the
survey refer are not identical with what the United States census calls
" families ".*
Even more important, sample surveys cannot be expected to cover
adequately the very top tail of the income and asset distribution—the
few people who account for substantial fractions of the aggregate income
or asset holdings. Since it is quite possible that some of these very rich
people may not live in ordinary private dwelling units in the United
States but in residential clubs or hotels, this problem is interwoven
with the problem of population coverage.
Aggregate checks show reasonably good agreement on income,
expenditures on durables, additions and repairs, and mortgage debt,
but substantial discrepancies in liquid assets and non-mortgage debt.
In the case of debt, survey estimated aggregates of loans for repair and
modernisation of homes or for the purchase of automobiles appear to
be about three-quarters of outside aggregate estimates. Other smaller
loans appear to be about half the outside estimates. Survey estimated
aggregates of liquid assets have averaged around 60 per cent, of outside
estimates, though aggregates estimated directly by different organisations
also disagreed with one another.
Surveys of consumer finances are clearly not useful for estimating
aggregates of such skewed distribution as assets, debts and income.
Even if the differences between survey and aggregate data result from
survey deficiencies, however, the survey data may still be useful. The
fact that a sample cannot represent adequately the very few extremely
wealthy people does not affect distributional data. Response errors may
affect numerous individual reports but leave over-all patterns and even
relationships reasonably correct. The data are probably still useful for
the bulk of consumers, and in an approximate way. Even if one can
only identify most of those with substantial liquid assets, those with
moderate amounts, and those with little or none, it is still useful to have
1
Even when related " spending units " living together are combined into
" families ", such groups are not identical with the census " family ". Only related
secondary " spending units " consisting of a married couple or a parent with one or
more children are called " sub-families " by the census, such units of other composition
are considered in the census as part of the main " family " The census categories of
" unrelated individuals " and " secondary families " include a substantial number of
people living in " quasi-households ".

202

FAMILY LIVING STUDIES

this information in studying expenditures or income, or the impact of
recent changes in income.
It is an unfortunate fact that often in the past surveys have been
evaluated on the basis of how well their " blow-ups " agreed with aggregate estimates derived from other sources, without sufficient attention to
the adequacy of such aggregates, the precision of the comparison, or the
usefulness of proximate information, i.e., whether imperfect data are
better than none at all. A great deal needs to be done, both in checking
the sources of aggregate data and in using individual validity checks,
to provide more evidence on the sources and distribution of errors in
survey data and to enable the development of improved techniques.
(Such survey validity work is now going on under the general direction
of Professor Robert Ferber of the University of Illinois, both at Illinois
and at the Survey Research Center at Michigan.)
Dissemination of Results
In addition to the published articles, and the unpublished extra
tables, there has been a substantial volume of special analysis and
tabulation done for government agencies, business firms, and research
economists. In addition, a growing number of articles and dissertations
have made use of the data, usually in new, more complex analyses.
Special Emphases and Improvement in Methods
The surveys from 1947 to 1952 collected data not only on income,
assets, purchases of durables and debts, but also on changes in assets
and debts sufficient to provide estimates of saving. In the 1949 survey
some 600 non-farm spending units from the previous survey were
reinterviewed. In 1953 about 1,000 spending units from the 1952 survey
were reinterviewed, and in 1958 some 1,000 units from the 1957 survey
were reinterviewed. In 1961 and in 1962 there will also be reinterviews.
These subsamples, while omitting those who moved or were unavailable
either time, provide useful data both on change and on memory errors
of respondents.
Special emphasis was given in 1952 to refrigerators and television
sets, in 1953 to net worth, in 1954 to fixed payments of consumers and
attitudes towards debt, in 1955 to housing conditions and attitudes, in
1956 to debt, in 1957 to handling and uses of financial assets and in
1958 to various methodological experiments and reinterviews. In 1961,
special emphasis will be given to family history and long-range plans
and, in 1962, to saving over two years for those who had not moved.

CONSUMER FINANCES IN THE UNITED STATES

203

While data on total personal saving have not been collected since
1952, questions on such contractual items as life insurance premiums
have been asked in some years since then, and in the 1955, 1958 and
1959 surveys questions were asked about changes in liquid assets holdings.
In addition, one can estimate for any reinterview, changes in both liquid
assets and debts to provide an estimate of " liquid saving ". Thus, for
1957 one can combine the two reports of reinterview units on liquid
assets and debts in early 1957 and in early 1958 to estimate " liquid
saving ". The same can be done for 1948 and 1952.
In the fall of 1949 some of the respondents of the 1949 survey were
reinterviewed, and in the fall of 1952 a questionnaire was mailed to
respondents of the 1952 survey, as an experiment. The mail questionnaire produced substantial differences in the response rates of the subgroups, and an average response rate of about 50 per cent. The answers
to questions on attitudes and plans also appeared to be more noncommittal than the responses obtained in personal interviews.
To supplement the surveys of consumerfinances,the Survey Research
Center started in 1951 a series of " interim surveys " financed from a
variety of sources, and securing in more detail, though from smaller
samples, data on changes in consumer purchases, plans, expectations
and optimism. By the end of 1958, 16 such interim surveys had been
conducted.
In addition, the Survey Research Center conducted a series of surveys
with a panel, in which the same families were interviewed five times over
a three-year period on their attitudes, plans, finances and past purchases
and saving. The analysis of this study is still progressing, but some
articles on it have appeared.
Substantial improvements were made in the 1954 Survey of Consumer
Finances in questions to elicit better data on income of extra earners.
In 1956 and subsequently, improvements in sampling and field operations
resulted in a more complete coverage of dwellings and a more efficient
procedure for including newly constructed homes in the sample. Among
other improvements, the " half-open interval " method of bringing lists
of addresses up to date was replaced by a technique that called for
checking whole designated areas for dwellings which were newly constructed or were missing from the listings for other reasons. The previous
procedure of checking only between the interview address and the next
listed address, while theoretically unbiased, proved insufficiently precise.
It was discovered also that while the listing procedures were theoretically
correct, changing to different sampling techniques reduced the problem
of " missed dwelling units " to the vanishing point. New sampling
procedures introduced in 1956 were:

204

FAMILY LIVING STUDIES

(a) the use of city directory street address listings as a sampling frame,
which is supplemented by an area sample to include addresses not
appearing in the directory;
(b) interviewing at all dwellings within small geographic areas (segments)
which, for the most part, contain about four dwellings each.
Some Uses of the Data
The objectives of the survey have thus been extensive and varied,
but in general they have been concerned with the financial status of
consumers, consumer attitudes relevant for assessment of consumer
behaviour, and changes in both attitudes and finances of consumers
under the impact of changing events. Such information was and is
important as a background for decisions on monetary and fiscal policy,
as well as for research on consumer behaviour. Immediately after the
Second World War the consumer's potential contribution to inflation
and his plans for the use of savings accumulated during the war were
of prime concern. In later years the impact of the growing consumer
debt and of continuing inflation were some of the major problems
studied.
Through survey data, analysts were able to relate debts to income
and assets, financial condition to attitudes and plans and expectations.
The data showed, for instance, that while awareness of inflation and
expectation of future price increases gradually pervaded the mass of
consumers, they continued to believe in savings bonds and savings
accounts, and only a few in the upper-income groups began to talk
about common stock or to purchase it. At the same time, the dramatic
increases in aggregate debt turned out to result in large part from increased acceptance of the use of instalment credit, even in the uppermiddle income groups, so that little increase in " burden " or in the
frequency of real difficulty in repayment occurred. That is, the ratio of
average debt to average income went up, while the relative frequency of
high ratios of debt to income did not increase. The increased frequency
of working wives, and of two cars in the family, was documented and
analysed.
Since the basic content of the questionnaires has remained the same,
a growing number of analyses have made use of combined " within-year
and between-year " data to investigate simultaneously cross-section
differences and changes as economic events occur. As noted above, the
highly trained interviewing staff has a low turnover, and except for minor
changes the primary sampling points remain the same from year to
year. Hence, the trends are more stable and data for different years

CONSUMER FINANCES IN THE UNITED STATES

205

more nearly comparable than for any other cross-section data of this
complexity.
Data have been collected on ownership, purchase and financing of
cars, houses and appliances. Numerous analyses have been made of
these data, making use of explanatory variables such as income, age,
attitudes, recent income changes, expectations, size of city, asset and
debt position. Occasionally, norms derived from these surveys have
been carried over and used in the analysis of other survey data collected
by the Survey Research Center, containing more information on attitudes
and changes.
In addition to examining consumer activity in buying, incurring debt,
etc., much attention has been given to describing and " explaining "
the current economic situations of consumers. Characteristics of those
with debts, or with assets, or with low or high incomes or with serious
unemployment experience have been investigated.
Reinterview analyses have been used in part to interpret the plans
and expectations reported at the beginning of a year. Investigations
revealed that substantial proportions of those planning to buy even
such a large item as a car failed to do so, their place being taken
by a small proportion (but roughly equally large number) of those
who did not plan to buy, but did so after all. Unexpected events
seemed to cause at least part of each of these types of deviations from
plans.
In view of the possibilities inherent in these surveys for investigations
of who does what, and why, a final characteristic of these surveys needs
to be emphasised. The basic data are available to non-commercial
researchers for little more than the cost of reproducing cards and codes.
The small samples, running around 3,000, make analysis based on
individual cases easily possible for anyone with access to suitable tabulating equipment.
While the individual cards carry two-digit weights to adjust for variations in sampling and response rates, much analysis has been done,
also, using unweighted data. It has been argued that so long as one is
using in a multivariate analysis most of the variables (income, spending
unit type) associated with differences in sampling or response rates, the
possible biases are minimised. Indeed, since the high-income units with
larger variances in financial data are over-sampled, there may be some
advantages to this procedure.
For researchers with more easily specified needs, special tabulations
or even whole analyses have been prepared. The data have been made
available to government agencies through the Federal Reserve Board
staff, and to other users by the Survey Research Center. Tabulating
14

206

FAMILY LIVING STUDIES

cards from various surveys are in use at some six major universities in
addition to the University of Michigan.
In summary, the surveys of consumer finances constitute an available
body of basic material about consumer finances and behaviour using
relatively stable but developing techniques, during a period long enough
to reveal trends, impacts of economic events, and changes in behaviour
over time of the same age cohorts.

XII
A Survey of Living Conditions for
Appraising Community Development
Programmes in the Philippines
By B. G. BANTEGUI *

GENERAL FEATURES

In the early part of 1958 the Presidential Assistant on Community
Development (P.A.C.D.) engaged the Office of Statistical Co-ordination
and Standards of the National Economic Council to undertake an
evaluation survey to provide a basis for appraising the results of the
Philippine Community Development Programme.
Realising the importance of accurate data as a basis for the development programme, the Presidential Assistant deemed it essential to launch
inquiries into living conditions in selected areas covered by the P.A.C.D.
programme in order to determine the need for specific measures for the
improvement of rural living, as well as to establish a baseline from which
the results of such measures could be evaluated in the future.
These inquiries could with advantage be organised as sample surveys
of households and barrios (political subdivisions of the smallest type)
covered by the community development programme and conducted
simultaneously with parallel sample surveys referring to all types of
rural communities.
It was therefore decided that the P.A.C.D. Evaluation Survey—as
this statistical project is commonly called—would be administered by
the organisation set up for the regular Philippine Statistical Survey of
Households (P.S.S.H.) and in operation since 1956 .* This organisation
* Director, Office of Statistical Co-ordination and Standards, National Economic
Council.
1
The Philippine Statistical Survey of Households is a national sample survey
conducted quarterly. The survey was aimed originally at reaching a representative
cross-section of Philippine households numbering 6,500 distributed in 300 barrios,
(footnote continued overleaf)

208

FAMILY LIVING STUDIES

had a highly trained and experienced staff for gathering the type of information required and was also in a position to provide the parallel series
of figures needed for comparative purposes. In order not to hamper the
regular activities of the P.S.S.H., a separate central staff of 26 officials,
of whom 7 were professional, 4 intermediate and 15 auxiliary statistical
personnel, was set up for the P.A.C.D. Survey. Eleven key officers of
the P.S.S.H. were in addition detailed to work with the P.A.C.D. central
staff for varying periods of time. Field operations of both surveys were
synchronised with the Chief of Survey of the P.S.S.H. acting concurrently as Head of the Operations Section for the P.A.C.D. Survey, and
the Director of the Office of Statistical Co-ordination and Standards
(O.S.C.A.S.), serving as the Project Director for the two investigations.
The Sampling Frame and Design
The Philippine Census of Population in 1948 provided the statistical
frame for both the P.A.C.D. Evaluation Survey and the P.S.S.H. According to the census there were 17,603 barrios and 1,256 poblaciones
(centres of population of the municipalities) in the rural areas of the
Philippines in 1948. The total rural population according to the census
was 15.4 million, of which 12.2 million lived in barrios and the remainder
in poblaciones.
The design for the P.A.C.D. Survey originally provided for a sample
of 6,000 households out of the 375,000 households in the area covered by
the development programme. Actually about 5,000 households were
interviewed.
First, all households in all development areas were listed. Then
considering each province covered by the development programme as a
stratum, and the groups of barrios and groups of poblaciones as separate
substrata within the province, sample households were selected systematically, by the use of a random start and a sampling fraction of about
one in 80 within each substratum.
The national rural sample of the Philippine Statistical Survey of
Households that was used as the control sample for the evaluation
survey consisted of 3,239 households randomly selected throughout the
country. A multi-stage sample design was used for this survey.
150 poblaciones and 58 provincial capitals and cities. Sample households randomly
selected in barrios ana poblaciones (national rural sample) represented rural conditions
while those selected in provincial capitals and cities (national urban sample) represented the urban conditions.
In the administrative organisation of the Philippines, provinces are divided into
municipalities and each municipality, in addition to having rural centres known as
barrios, has one población in which are located the municipal buildings, church,
school, etc. In addition, the central districts of chartered cities and of provincial
capitals are also identified as poblaciones.

LIVING CONDITIONS IN THE PHILIPPINES

209

First the country was divided into ten regions according to such
factors as economic activities, climate, crops and ethnic origin of the
population. Then the municipalities in each region were stratified
according to their population densities. Thirty strata of approximately
equal population were thus formed. Five municipalities were randomly
selected with probability proportional to size of population and with
replacement in each stratum. In all, 300 barrios, two from each of the
150 sample municipalities, were selected with equal probability and
without replacement. All households in each selected barrio were listed
by the field staff. Sample households were then selected from these lists
by means of a random start as a specified proportion of the total number
of households in each barrio.
Organisation of the Survey
The results of the sample surveys undertaken in the Philippines since
1956, together with pre-tests conducted specifically for the Evaluation
Survey, provided the basis for a number of decisions relating to the
organisation of the main survey.1
Pre-tests were made in two areas by interviewing 32 households in
four municipalities in each area. The results of these tests led to a number
of revisions of the household schedule and indicated that, on the average,
two hours' interview time would be needed per household.
Field work in the main survey started on 20 January 1958 with the
listing of households and was scheduled to end on 31 March 1958.
However, a small number of forms were not received until 31 May 1958.
The data obtained in the survey related to the period of one year or the
52 weeks preceding the date of each interview.
Household interviews in the development areas were conducted by
218 barrio community development workers assisted by other community development personnel. Household interviews in the P.S.S.H.
areas were conducted by 165 field workers on the regular staff of the
P.S.S.H.
The interviewers worked under the supervision of ten regional
supervisors from the P.S.S.H. staff. They were selected among the
residents of the municipalities where they were to work on the basis of
1
From May 1956 to October 1957, five rounds of the Philippine Statistical Survey
of Households (P.S.S.H.) were undertaken. Other surveys conducted in the Philippines during the same period included the annual crop and livestock survey, the
nutrition and dietary survey (1957), capital formation in agriculture (1957), cost of
production of selected crops (1957) and other minor surveys. All of these surveys,
except the crop and livestock survey, were carried out under the Philippine Statistical
Survey Project with the Director of the O.S.C.A.S. as Project Director.

210

FAMILY LIVING STUDIES

training and experience in statistical work and were required to be high
school graduates and proficient in the dialect of the locality to which
they were assigned.
All of the interviewers were given a four-day training course under
the direction of the regional supervisors. This included training in listing
and enumeration procedures, definitions of the questions on the schedules,
and supervised practice interviewing.
Co-operation of the Households
All interviews were conducted with a responsible adult member of
the household, even if this meant return visits by the interviewers.
Under no circumstances was a substitute used. No reward or payment
to respondents was made because the households were generally cooperative. The nature, objectives and method of household selection
were explained to the respondents as a way of gaining their cooperation. Since very few farm households keep records of farm
operations, including income and expenses, most replies were based
on memory.
No interviews were obtained from 1 to 2 per cent, of all sample households. For the most part this was due to the household having moved to
another place or to the fact that no responsible member of the family was
available for the interview. There were no reports of refusals. The use
of appropriate follow-up questions provided an internal check on the
consistency of the answers to some of the questions.
Tabulation and Analysis of the Results
The results were tabulated on punch card machines. The preparation
of the tabulating plan and design of the punch cards was carried out
hand in hand with the preparation of the questionnaires. The proposed
tables were submitted to a technical committee representing the several
interested organisations.
The results of the survey were prepared in the form of two sets of
74 tables each, one set for each of the two samples used.
The analysis of the information consisted of three parts. First, an
attempt was made to describe living conditions in the P.A.C.D. areas.
Secondly, living conditions in the P.S.S.H. areas were described in similar
detail. Finally, a comparison of living conditions in the areas covered by
community development programmes and those obtaining in other rural
areas of the Philippines was made.

LIVING CONDITIONS IN THE PHILIPPINES

211

Cost of the Survey
The total cost of the P.A.C.D. Evaluation Survey amounted to
P. 105.612.84.1 This consisted of the following: personal services,
P. 56,054.95; travelling expenses, P. 19,189.15; supplies and materials,
P. 9,869.84; communication, transportation, office rental, furniture and
equipment etc., P.15,198.10; other charges, including cost of printing,
P. 5,300.80.
The above survey costs did not include the salaries of the ten regional
supervisors and the 11 officials of the P.S.S.H. and a few other persons
on the staff of the Office of the Presidential Assistant on Community
Development who assisted in the survey operations for varying periods.

DESIGN OF THE QUESTIONNAIRES

The Philippine Community Development Programme recognises the
barrio as the significant geographical unit. The projects under the programme aim both at providing better facilities for rural living at the community level and at stimulating the utilisation by the population in the
barrios of such facilities.
Information required for appraising the impact of the programme on
the economic and physical well-being of the households in the areas
covered by the programme was therefore collected in the P.A.C.D.
Evaluation Survey by means of two types of schedules :
(1) a barrio schedule for the recording of information about the
community as a whole, and
(2) a household schedule for the recording of the characteristics of a
sample of households and individuals in each barrio.
The Household Schedule
The household schedule was by far the more detailed and required
the largest amount of interviewing time.
Information was obtained through this schedule by means of household interviews on the following aspects of family living conditions :
(a) demographic characteristics—age, sex, marital status and relationship to the head of the household;
(b) education—school attendance, and attendance in extension courses,
educational level attained and literacy;
1

One Philippine peso = 50 U.S. cents, or 3s. 6d. sterling.

212

FAMILY LIVING STUDIES

(c) economic activities of persons ten years old and over—main activities, principal occupation, industry, cash earnings and number of
weeks worked during the past year;
(d) housing—type of tenure of dwelling, number and total floor area
of rooms occupied, number of households per dwelling, mode of
garbage and waste disposal, toilet facilities, lighting facilities and
sources of water supply for drinking and for home use ;
(e) health and sickness—person or agency consulted in the case of
serious illness and number of persons ten years old and over,
chronically ill or physically disabled;
(f) agriculture—farm tenure, area of cultivable idle portion of land,
farm practices such as use of fertilisers, better seeds, pest control,
irrigation systems, agricultural co-operatives, credit facilities, transportation and marketing facilities and farm equipment;
(g) income—personal and family income from crops, livestock and
poultry (including data on the number of livestock and poultry
raised by the family), as well as from work other than farming;
(h) services and sources of information—person or agency consulted
on information about crops, livestock, poultry or farm products,
on health, sanitation and hygiene or related services, on national
and international affairs; attendance at demonstrations of new
planting methods, use of fertilisers, control of plant pests and
diseases; receipt of free medicine, visits by government health
officers and participation in barrio community projects.
A number of considerations were kept in mind in the design of the
schedules. Among the main points considered were: the questionnaires
should be as self-explanatory as possible; the questions should be
arranged in logical sequence to permit easy recall by respondents as well
as checks of the validity of responses; and the questions and the explanatory phrases, examples, codes and simple instructions printed on the
schedule should help to eliminate difficulties of interpretation.
In the following paragraphs some illustrations of the definitions
used in the schedules are given as examples of the enumeration procedures
adopted in the survey.
Demographic Characteristics.
In order to ensure uniformity in making entries for the individual
members of the households, instructions were given that the name of the
head of the household was to be entered first, followed by his wife and the
children in order of their ages. If a son or daughter who was married or
had been married was a member of a sample household, alone or with

LIVING CONDITIONS IN THE PHILIPPINES

213

his or her family, the name of that person was entered after the name of
the youngest unmarried child of the head followed by the members of
the family of the married son or daughter. Other relatives, servants and
boarders were then listed in that order. Usually the head of the household was the father. However, if he was already dead or was not a
member of the household, the mother was designated as head of the
household. The interviewer was instructed to obtain the age in years of
the persons one year old and over. For infants below one year old on
the day of the interviewer's visit, the exact date of birth was entered.
Education.
The question : " Had this person attended school at any time since
June 1957? " was asked for all persons 6 years old and over. By school
attendance was meant attendance in a grade or year included in the
regular system of education, that is, primary, intermediate or high
school (whether general or vocational high school) or college, at any
time since June 1957. It did not include attendance in kindergarten
schools, schools of dressmaking, tailoring, etc., or in education extension
courses for adults.
The question: "Why is this person not attending school?" was
asked for every person 6 to 16 years old whose entry in the preceding
column was " No. ". The interviewer was instructed to use the following words and phrases to convey the reason opposite each case: " Can't
afford "—the family could not afford to send the child to school because
of lack of money for tuition fee, clothing, etc. ; " Sickly "—the child
did not not go to school because he was sickly; " 111 "—the child failed
to enrol because he was ill during the opening of classes; " Disabled "—
the child was suffering from a more or less permanent disability or
illness; " Lazy "—the child was too lazy to go to school; " School too
far "—there were no schools accessible to the child; " Needed at home "
—the child could not attend school because nobody else would do
housekeeping at home; " Needed in work "—the family considered
working more important than schooling for the child, even though the
family could afford to send him to school.
A question on the highest grade completed in school was asked for
every person 6 years old and over. The question on literacy (the ability
to read and write a simple message in any one language or dialect) was
asked for all persons 10 years old and over. The interviewer was instructed to enter " Yes " for a person who was able to both read and
write simple messages in any one language or dialect either in alphabet
or characters. An entry of " No " was made for a person who could
read and write only his name. An entry of " Yes " was made for a person

214

FAMILY LIVING STUDIES

who was blind and who was then able to read and write by the method
especially adopted for the blind at the time of the interview.
Economic Activities.
A series of sorting questions was used to ascertain the main activities
of persons 10 years old and over during the past year. The first two of
these had the purpose of ascertaining the economically active and
economically inactive population, and to separate the persons 10 years
old and over whose main activity was " working " during the past year
and those whose main activity was other than working but did some
work during the past year.
With regard to the first question (" What was this person doing most
of the time during the past year?") the interviewer was instructed to
enter one of the following codes on the schedule :
W — Working at a job or on own farm, business, profession or trade
WP — Unpaid work on family farm or enterprise
H
S

— Doing housework at own home
— Attending school

N —• Doing nothing
RT — Retired or too old to work
DA — Disabled or unable to work
OT — Other (specify)
The interviewer was instructed, in asking the question, to add " working ", " housekeeping " or " attending school " (whichever seemed most
applicable to the person) plus the phrase " or something else". Thus,
for an adult male, the question asked was what was this person doing
most of the time during the past year, " working or something else ? ".
For an adult woman, the interviewer added to the question the phrase
" housekeeping or something else ", and for persons of school age
" attending school or something else ". Then if the answer given
was " working ", the next question asked was whether the person was
working on a job or own farm, business, profession or trade, or working
without pay on family farm or enterprise run by another member of the
family.
If an entry of " W " or " WP " was made in answer to the first
question, or if the answer was " Yes " to the supplementary question
asked for all other persons (" Did this person do any work during the
last year? "), information was also obtained on cash earnings, principal
occupation, kind of business or industry and number of weeks worked.

LIVING CONDITIONS IN THE PHILIPPINES

215

The question on earnings in cash during the past year related to all
money received for work or activity. This included salaries, wages,
commissions including tips; sales of farm products raised by the
person, of fish caught, firewood cut, etc.; net profit or fees derived
from the operation of own business or from practice of profession
or trade.
In this connection, the term " working " did not include such activities
as housekeeping at home, building or repairing own house or own home
yard or fence, voluntary work for somebody else or for the community,
gambling, begging, etc.
The interviewer was instructed not to include money received as
interest or dividend from investment, retirement, pension or gambling;
sales of share of farm products received as landlord, begging, etc. In
filhng in the schedule, the interviewer was directed not to give the actual
cash earnings but to enter a code referring to the income class represented by the reported amount.
The question on occupation, that is, the occupation at which the
person worked longest during the past year was to be answered by a description of the specific job or occupation performed by the person in the
establishment, office, farm, etc., such as farmer, farm labourer, fisherman,
basket weaver, carpenter, typist, lawyer, etc. The interviewer was
cautioned not to accept such responses as employee, factory worker,
mechanic, businessman, labourer, official, etc., that do not adequately
describe the specific job or occupation.
In reporting the kind of business or industry, the respondent was
asked to describe specifically and adequately the nature and characteristics of the business or the place where the work was being performed.
The interviewer was not permitted to accept such vague terms as farm,
store, retail store, mine, factory, transportation company, repair shop,
office, government, etc. The respondent was asked to specify the kind
of business or industry, like corn farm, " sari-sari " (variety) store, gold
mine, rope factory, bus transportation, auto repair shop, etc. The
interviewer was instructed not to accept company or firm names. Some
people conducted their businesses in their homes. These businesses were
reported in the same manner as regular establishments such as tailor
shop, auto repair shop, dental clinic, dressmaking shop, etc., were
reported.
The answer to the question on the number of weeks worked during
the past year was to include the weeks in which the person worked for
salary or wages or was on leave of absence with pay. The interviewer
was instructed to count as one week each calendar week in which the
person worked, whether full time or part time.

216

FAMILY LIVING STUDIES

Agriculture.
As a means of determining whether or not a household operated a
farm, the question was asked: " Does the head or any member of this
household related to the head by blood, affinity or adoption operate a
farm ? " The term " farm ", for purposes of the survey, was defined as
any parcel or parcels of land at least 1,000 sq.m. in area (one-tenth of a
hectare) used for the raising of crops, vegetables, fruits and/or other
agricultural products or of livestock and poultry. If a negative answer
to the question was given, the interviewer was instructed to ascertain if
any member of the household had been engaged in farm operations
during the past 12 months or, if not, whether he was a farmer—by
usual or subsidiary occupation—but unable to operate the farm the last
12 months for one reason or another.
In the case of farm households the question also asked was : whether
the farm was wholly owned by the head or by members of the household
related to the head, etc. This served to identify " full owners ", " part
owners " and tenants. Farm tenants were further asked how the landowner was compensated for the use of his lands, thus permitting a
distinction between " cash tenants ", " crop-share tenants ", etc.
Income.
One of the items on the schedule that was difficult to enumerate
correctly was family income. " Family income " as defined in the survey
was income that accrued to members of the household during the 52
weeks preceding the day of interview. Income of guests, boarders and
household help was excluded. The income of the family was classified
according to source into two kinds, namely: income from work on farm
(income from crops, livestock, poultry, etc.) and from work other than
farming (income from salaries, wages, self-employment); and income
from other sources, consisting of income as landlord, income rental of
other properties, interest and dividend from investments, pensions,
annuities, etc.
In reporting family income, the interviewer was instructed to make
separate entries for " cash " and " not-in-cash " income from the abovementioned sources of income. For purposes of the survey, " cash
income " from crops, livestock, etc., was defined as cash received by the
family from any quantity sold for cash out of the produce or harvest
during the preceding 52 weeks, after deducting market expenses, while
" not-in-cash " income meant imputed value of the remainder. In the
case of income from work other than farming, cash income meant cash
salaries and wages or cash net profits or fees from self-employment.
" Not-in-cash " income was here defined as the imputed value of articles

LIVING CONDITIONS IN THE PHILIPPINES

217

or commodities received as salaries and wages (including value of meals
received, board and lodging, clothing, etc., supplied by the employer)
or received in self-employment.
A special procedure was used in ascertaining " cash " and " not-incash " income from crops, fruits, etc. : the respondent was first asked to
list his principal crops during the past year. Then the interviewer asked,
for each principal crop, the amount of money received from the sale
minus marketing expenses for the past year, and the value at harvest
time of the quantity not sold for cash. The interviewer noted down the
" cash " and " not-in-cash " income for each principal crop. After
asking about each principal crop, the same set of questions was asked for
the minor crops. Similar procedures were employed in obtaining income
from livestock, poultry, and forest products.
The Barrio Schedule
The barrio schedule aimed, as already noted, at recording the general
characteristics and availability of specific community facilities in the
barrios covered by the Community Development Programme. It
included information on the total population, the total number of households and the distribution by usual occupation of the heads of the households as given by the barrio listing sheets and, in addition, a series of
dates obtained by interrogation of the barrio lieutenant, or, if he was not
available, a responsible resident of the barrio, such as a teacher, a member of the barrio council, etc.
Thus, the availability of motor vehicles, shops, medical personnel of
different kinds, and various types of agricultural equipment such as
tractors, threshers, pumps for irrigation, sprays, etc., was recorded. The
inventory of organisations in the barrio such as churches by denominations and rural improvement schools was made. The functioning of
markets, postal services, health centres, schools and the number of
teachers and students in the latter was also registered.
The schedule, in addition, provided for reporting the principal
industries, number engaged, products made and sources of material.
Questions were also included on irrigation, on the extent of flooding of
the village during the rainy season, type of communication between the
barrio and the población and the principal means of transportation used
over this route.
To complete the inventory of the barrio, questions were included
regarding the membership in the farmers' co-operative marketing association, the number of registered voters, number of households with radio
sets and principal dialect spoken in the barrio.

218

FAMILY LIVING STUDIES

MEASURING THE EFFECTS OF THE COMMUNITY DEVELOPMENT PROGRAMME

Community development in the Philippines has been defined as " a
movement to promote better living with the active support and upon the
initiative of the community; and a process to involve people in the
solution of their own problems ". The P.A.C.D. office which, since
January 1956, has responsibility for the planning and implementation
of the programme as well as for the co-ordination of all government
efforts at rural improvement seeks to achieve these aims by: (1) developing a desire among the population of the barrios for an improvement in
their living conditions ; (2) raising their income through better agricultural
practices and the introduction and improvement of crafts and industries
to provide more employment; (3) providing education and health facilities as well as instituting local organisations for self-help projects ; and
(4) assisting the people in citizen education and other nation-building
projects.
Individual development projects under the P.A.C.D. programme are
accorded priority with a view to the generating effect of each project,
i.e., whether or not the project has a tendency to promote other projects.
They include the following types of activities : (a) projects contributing
to increased production and income ; (b) self-help public improvement
projects; (c) projects for the improvement of health and sanitation;
(d) education projects; and (e) social welfare administration and
related activities.
The way in which the P.A.C.D. Evaluation Survey can throw light
on the effects of Community Development projects of these different
types is illustrated below. Since, at this early stage the impact of the
programme on the development of community facilities can be more
readily appreciated than its effects in relation to family living conditions,
most of the illustrations apply to the analysis of the barrio schedule. The
usefulness of the household schedule will however be evident in subsequent rounds of the survey.
Increased production and income. Information on farm tenure and
farm practices, personal and family income and sources of water supply
given by the household schedule; information on agricultural facilities
provided by the barrio schedule.
Self-help public improvement projects. Information on type of route
connecting the barrio with población or highway, means of transportation by that route and existence of irrigation systems, reading centres,
public markets, public playgrounds or recreation areas, etc., as given by
the barrio schedule.

LIVING CONDITIONS IN THE PHILIPPINES

219

Improvement of health and sanitation. Information on the persons
or agencies usually consulted in case of sickness, number of persons
chronically ill or physically disabled and housing characteristics given by
the Household Schedule; information on the number of dentists, physicians, nurses and licensed midwives, public toilets, portion of the
barrio usually flooded during the rainy season, etc., according to the
barrio schedule.
Education projects and related activities. Information on barrio
schools, enrolment and number of teachers obtained through the barrio
schedule.
Social welfare administration. Information on the existence of such
barrio organisations as the parent-teacher association, barrio council,
rural improvement club, athletic club and related associations, religious
organisations and churches of various faiths, also according to the
barrio schedule.
In order to measure in the future the changes in rural living that can
be objectively ascertained to be the result of the P.A.C.D. programme, it
will be necessary to compare: (a) rural living in the same areas covered
by the P.A.C.D. programme over a specified period, say once every two
years; and (b) rural living in the P.A.C.D. areas with that of the rest of
the country over the same period. The 1958 Evaluation Survey was
designed to provide benchmarks for such comparisons by obtaining data
for two sets of samples—a sample of households in the areas covered by
the P.A.C.D. programme and another sample of households in the rural
areas of the Philippines as a whole, representing the conditions obtaining
in all barrios of the country. It is worth while citing at this point also
some additional advantages of the design of the sample adopted for the
P.A.C.D. areas.
1. The size of the sample was such that it would be possible to obtain
fairly reliable provincial estimates of selected characteristics of households and barrios in the provinces where the P.A.C.D. programme
covered a large number (7,000 or more) of households, thus facihtating
comparisons of the results of the P.A.C.D. programme in selected provinces over time.
2. The sample design was fairly simple and flexible and the size could
be increased without much difficulty when necessary. Since the P.A.C.D.
programme was expected to cover more areas as the programme expanded, it was considered advantageous to have a sample that would
allow the taking in of more sample households when necessary.
3. The construction of the frame of the survey provided an actual
count of the people in households in all the P.A.C.D. areas, thus permit-

220

FAMILY LIVING STUDIES

ting also the measurement of actual changes in the population of these
areas.
The P.A.C.D. Evaluation Survey in 1958, described in this paper, was
the first round of a series of surveys planned to be conducted every two
years. Subsequent rounds would obtain data on specific aspects of
rural living in the Philippines. An attempt will be made to find whether
the people have become more productive and whether conditions and
attitudes conducive to further development have been attained. This will
require looking into such evidence as the expansion of the output of
goods and services; the growth of capital formation; and the attitudes
towards government agencies and officials engaged in community
development work.
By following up the initial P.A.C.D. Evaluation Survey with periodic
sample surveys, the rate of change of specific items such as average
agricultural production per hectare, family production and consumption,
land tenure patterns, attitudes towards change, indigenous leadership
patterns, etc., in the P.A.C.D. barrios covered by the community development programmes can be determined. This rate of change can then be
compared to the corresponding rate of change in the control areas by the
use of appropriate significance tests to provide a reliable basis for
assessing the success of the planned undertaking.

XIII
The Continuous Population
and Labour Force Survey in the
United States
By A. Ross

ECKLER

GENERAL FEATURES

The Current Population Survey conducted by the United States
Bureau of the Census since 1942 is doubtless the oldest continuous
national survey of the population of an entire country. The succession of
names applied to this survey tells the story of the evolution of its objectives. It was initiated as the Monthly Report of Unemployment by the
Works Progress Administration in January 1940, when a direct and
reliable measure of the volume of unemployment was a major national
need. When it was transferred to the Bureau of the Census in August
1942, unemployment had ceased to be a major concern and the emphasis
soon shifted to measuring changes in employment and the size and
characteristics of various sources of potential workers. During the war
years the survey was typically referred to as the MRLF, based upon the
initials of the " Monthly Report on the Labor Force ", the name of the
publication that summarised the figures on employment and unemployment. The operation of the survey during the war years indicated its
great flexibility in furnishing timely information on various matters
relating to population and households. In 1947, therefore, the title of
the survey was changed to the Current Population Survey.1
* Deputy Director, U.S. Bureau of the Census. The author wishes to acknowledge
the assistance of a number of members of the Bureau's staff in the preparation and
review of this chapter, in particular that of Dr. H. P. MILLER, R. B. PEARL and Mr.
Joseph STEINBERG.
1

The CPS was the only official monthly household survey in the United States
until 1957, when the Health Household-Interview Survey was initiated as a part of the
National Health Survey, a continuing inquiry to secure information about health
conditions in the population. This survey is carried out by the Census Bureau acting
as agent for the Public Health Service.
15

222

FAMILY LIVING STUDIES

Objectives and Co-ordinating Mechanism
Despite changes in emphasis from time to time, the collection and
publication of information on the size and characteristics of the employed
and the unemployed have been of central importance throughout the
life of the Current Population Survey. Much of the information collected
every month, such as age, sex, family composition, labour force status,
occupation and industry, is basic for family living studies and for many
other analytical purposes. Through the use of a few supplementary
questions in addition to the basic questions, it is often possible to shed a
good deal of light on important issues or problems. In addition, it may
be feasible to concentrate attention upon a particular subgroup of the
population, such as the low-income group, the older worker, the parttime worker, or the population in a particular area, and thereby to contribute information on a particular problem group or area. A number
of illustrative specialised studies are described briefly in a later section of
this article.
Some of the requests for the use of the survey involve objectives that
are mutually inconsistent, and hence choices must be made. The Bureau
of the Census, in its role as a general-purpose statistical agency, receives
guidance from the Office of Statistical Standards of the Bureau of the
Budget, which has the responsibility of co-ordinating the statistical
programmes of the government. Inter-agency committees under the direction of the Budget Bureau are called upon when necessary to review
difficult procedural problems as well as policy questions.

Time Schedule
Because of the key importance of the monthly data supplied by the
Current Population Survey, the time schedule for collection and processing is very tight. The reference period for the labour force data is the
week including the twelfth of the month, and virtually all of the interviewing is carried out during the following week. Within about 15 days
after the end of the reference week, the schedules have been processed
in the central office in Washington and the final estimates have been
compiled by the electronic computers. The results are often available,
therefore, by the end of the month in which the interviewing was
performed. In some specialised inquiries, the data collected may apply
to the preceding year, the preceding quarter, or to some other
period in the past. For comparatively long inquiries the timing is
more flexible.

POPULATION AND LABOUR FORCE IN THE UNITED STATES

223

Coverage
The Current Population Survey applies to the civilian non-institutional population of the United States. Once a year, when information
on families and other data are compiled, it also covers inmates of penal
institutions, homes for the aged, infirm and needy, mental institutions,
tuberculosis sanatoriums and similar places. Employees of such places,
however, are regularly included in the monthly survey. The sample is so
designed as to be useful for measuring not only employment and unemployment, but many other economic and social characteristics of the
population.
Size and Distribution of the Sample
Over the past 20 years the sample design has been repeatedly modified
as the result of continuing research. In 1940 it covered about 20,000
households in 50 counties. Since May 1956, the survey has been extended
to 330 sample areas, comprising 638 counties, from which a total of
42,000 dwelling units and other living quarters are drawn. Completed
interviews are obtained each month from about 35,000 households.1
Thus, over two decades, the sample of households has been nearly
doubled, while the number of sample areas has been increased about
tenfold.
Types of Questionnaires
The Current Population Survey inquiries consist of three general
types : (1) a control card, filled out when a family is initially brought into
the panel and then kept up to date each month that the family remains in
the sample ; (2) the basic monthly questions to measure the number and
characteristics of the employed and unemployed; and (3) special questions
or questionnaires, some of them on a periodic basis as quarterly or
annually, and some on a one-time basis.
The control card covers a few general items (e.g., location, type of
quarters, availability of telephone, etc.) for the household as a whole and
several other items, such as age, sex, marital status, race, and relationship to head, for members of the household. Some of these items are
transcribed each month to the monthly questionnaire. This includes a
total of 11 different questions relating to employment status, hours
worked, description of job or business, and related items for persons 14
1
The difference, about 7,000 dwelling units, includes those in which occupants are
absent or cannot be interviewed for other reasons (about 1/5 of the cases) and units
which are found to be vacant, converted or demolished, or occupied by persons whose
principal residence is elsewhere, etc.

224

FAMILY LIVING STUDIES

years old and over. In general only a part of these need to be applied
to a particular individual to determine his attachment, if any, to the
labour force.
The special (supplemental) inquiries vary widely in type and number
of questions. In some cases they involve the addition of only two or
three questions to the regular monthly inquiry that may be placed on the
same form, but in other cases a separate form is required. Supplements
such as those pertaining to holders of two or more jobs and to work
experience during the preceding year are closely related to the labourforce aspects of the CPS and are usually conducted annually. Other
annual inquiries on family size and composition, school attendance,
population mobility, and income furnish current information on subjects
included in the Decennial Census schedules. Periodic, but less frequent,
studies (every two or three years) are made in fields in which changes are
less rapid, such as educational attainment and fertility. Still other
supplements are rather specialised one-time surveys, which furnish a
considerable amount of information.
Pre-testing of Survey Procedures
Careful testing always precedes the undertaking of any substantial
new inquiry as part of the CPS or any significant revision of questions or
procedures. The initial step may be a small-scale test consisting of, for
example, 25 to 50 interviews conducted by members of the professional
staff with households outside the CPS sample, to obtain some preliminary impressions of public reaction and feasibility of the proposed
inquiry. As a next step, the proposed questions may be added to the
regular monthly survey in one or more areas—generally totalling several ,
hundred interviews—in order to try them out under actual operating
conditions. If the new inquiry is considered potentially troublesome
from a public relations standpoint, this large-scale test may be conducted
separately, using former sample households in order not to affect responses by households from which labour-force data are currently
collected. Although these tests are generally too limited to provide
conclusive statistical evidence, they are valuable aids in reaching
decisions concerning new proposals or in modifying the questions or
procedures.
More intensive preparation may be required for any proposed change
in the basic employment status questions. A major change in the questionnaire made over a decade ago was preceded by some pre-tests on a
national scale and a complete overlap enumeration—first on the old,
then on the new, basis—at the point of transition.

POPULATION AND LABOUR FORCE IN THE UNITED STATES

225

Staff
The central office staff directly and continuously concerned with the
CPS is relatively small, consisting of 8 to 10 statisticians and a clerical
force of about 15 people, augmented at peak periods each month to
meet a rapid time schedule.1 The statisticians, with the assistance of
specialists from other parts of the Bureau of the Census, and in some
cases from other federal agencies, are responsible for the definitions
used, the preparation of questionnaires and instructions, determination
of sample design and estimation techniques, the planning of processing
and tabulation procedures, and the review and analysis of the results.
The clerical staff codes certain items of information such as occupation
and industry and later reviews the tabulated results and puts them in
summary form for publication. This staff also computes measures of
reliability and quality of the data. The small central office staff of about
25 persons is able to call upon specialised divisions of the Bureau for a
number of services. For example, one division provides maps for the
sampling work in the field, another prepares punch cards, the information
on which is later transferred to magnetic tape, and a third carries out
detailed tabulations on high speed electronic computers. The total
number of man-years performed in the Washington office on various
phases of the CPS is about 75. This does not include services paid for
by other agencies.
The field staff is organised to provide a nucleus of personnel in 17
regional offices, each one directed by a regional supervisor who has a
staff of programme assistants. They supervise a total of about 750 parttime enumerators including about 125 who usually work on surveys
other than the CPS. The 17 regional supervisors and their assistants are
responsible for selecting and training enumerators, for review of their
work each month, and for periodic direct observation and checking of
performance. The regional supervisor is required to accompany at least
one enumerator in the field each quarter. Detailed instructions are
furnished by the central office on the selection of the sample areas and on
the procedures for the training of enumerators and reviewing their work.
The enumerators are mostly women, many of them with household
responsibilities which prevent them from taking full-time jobs but who
can nevertheless work part time. All of them are required to take a
qualifying examination which tests their ability to understand instruc1
Since July 1959, as part of an inter-agency transfer of functions, the Bureau of
Labor Statistics of the U.S. Department of Labor has been assigned responsibility
for analysis and publication of the current labour force statistics although the Bureau
of the Census continues to collect, process, and tabulate the data. Since that date,
therefore, the central offbe staff has been subdivided between these two agencies.

226

FAMILY LIVING STUDIES

tions and use maps. It has been the experience of the Bureau that
enumerators selected in accordance with this test are able to carry out a
wide variety of inquiries, even though they do not have a full technical
understanding of all the subjects they cover.
Collection of Data
Most of the information collected in the CPS is obtained by means of
direct interrogation of respondents. For some supplementary inquiries,
however, a questionnaire is left with the household with the request that
it be filled out and either held for the enumerator or mailed in to the
local field office. Revisits are made to households not sending in their
questionnaires and to those for which entries are missing. Telephone
inquiries are sometimes used for completing partially filled schedules, for
contacting respondents who are not found at home, and for conducting
complete interviews in place of personal visits where it is clearly more
efficient to do so, and where previous permission has been obtained from
the respondent.
An important feature of the data collection process is that families in
geographic areas included in the panel are interviewed for a series of four
months, are omitted from the panel for the next eight months, and then
are returned for a second period of four months. This rotation plan
provides month-to-month comparisons of households in identical
geographical areas for 75 per cent, of the total panel and year-to-year
comparisons of households in identical geographic areas for one-half of
the panel. Basic information on household characteristics is collected
only in the first interview with each family and is subsequently brought
up to date to reflect births, deaths, marriages, or complete turnover in the
occupancy of a dwelling unit. Co-operation over the full series of interviews is extremely good with less than 1 per cent, of refusals on an
average, despite the fact that responses are not mandatory and the
households included in the panel receive no compensation of any kind
for the time required by the enumerator. The co-operation extends to the
supplemental inquiries as well, even though some of them take fifteen to
twenty minutes and include questions that might be regarded as personal,
such as income and number of children born.
Tabulation of Data
The data are generally collected in the field on a document-sensing
schedule on which most of the entries of the enumerator take the form of
marks in the proper spaces. For certain items such as occupation and

POPULATION AND LABOUR FORCE IN THE UNITED STATES

227

industry, enumerators enter descriptions in some detail which are
classified by coders in the central office. The information on the schedule,
after the process of review and coding, is converted to punch card form
by the use of a mechanical document reproducer, which depends upon
the completion of electronic circuits through the marks which enumerators or coders have made on the schedules. The information on the
punch cards is transferred to magnetic tape by means of a card-to-tape
converter. Next the tape is processed through a high-speed electronic
computer which equally well meets the fast time schedule of a regular
monthly series as the needs for complex tabulations occasionally involved
in the special supplemental surveys.
Analysis and Publication of Results
The results of the CPS, like those of other series collected and tabulated by the Bureau of the Census, are analysed more or less intensively,
according to the nature of the material. Until July 1959, the monthly
statistics on employment and unemployment were analysed briefly and
issued in the " Monthly Report on the Labor Force " (Census Series P. 57).
Since that time, with the transfer of functions previously mentioned, a
summary of these data is issued by the Bureau of Labor Statistics in a
publication carrying the same title, with more detailed statistics issued
somewhat later in the monthly bulletin of that agency, " Employment
and Earnings ". Somewhat more detailed statistics on special aspects of
the labour force, formerly issued in Census Series P. 50, are now published
by the Bureau of Labor Statistics. Other reports pertaining to the
demographic and economic aspects of the population, such as family
formation, and population mobility, are analysed in moderate detail by
the census staff and published in a series of releases on population
characteristics (CS. P.20) or in a series of releases on income statistics
(CS. P.60). Since the analytical work done by the Census Bureau is
comparatively limited in character even in the case of data relating most
closely to the Census of Population, extensive additional analysis is often
performed by other agencies of the government as well as by research
specialists and organisations.
In the case of supplements for other agencies, the analysis and publication of the results are typically left to the agency requesting and paying
for the work. In all cases, however, the Census Bureau reserves the right
to publish the results of the surveys when it believes that such action
would make available material of public interest.

228

FAMILY LIVING STUDIES

Cost ofthe Survey
The annual cost of the Current Population Survey programme, with
the 330-area and 35,000-household sample, is approximately $1,900,000.
A rough distribution of the annual budget, by major function, is
shown in the following tabulation:
Field collection budget:

Direct enumeration (including travel allowances)
Field office clerical costs (preparation of assignments for enumerators, review and edit of returns, transmittals to Washington and
general record keeping)
Quality control programmes (training, observation of enumerators
and recheck programme)
General field supervision and administration

580,000
170,000
320,000
240,000
1,310,000

Washington central office budget:
Preparatory work for survey (questionnaire design, preparation of
field instructions and training materials, map preparation,
sampling procedures and related research activities) . . . .
Processing and tabulation of results
Analysis and publication of labour force and other related economic
data
Demographic research activities and analysis and publication of
demographic data from CPS

180,000
225,000
100,000
85,000
590,000

Total annual budget . . . $1,900,000

SAMPLE DESIGN

The current sample for the CPS (like its predecessors) was planned
in such a way as to hold to reasonable levels the size ofthe sampling and
non-sampling components of errors as well as the total cost ofthe survey.
The plan took into account judgments as to the amount of precision
needed for some of the widely used figures, such as the volume of unemployment, and as to the total cost that would be reasonable when
balanced against the cost of other important government statistics.
The judgments brought to bear upon this problem included those of
members of the Census Bureau staff, of advisory committees, and of
Commerce Department and Budget Bureau staffs responsible for reviewing and approving budget requests. Prior to the latest expansion of the
Current Population Survey sample in 1956, there had been important
expressions of opinion that the sample should be increased in size.
Among those expressing such views were the committee of business

POPULATION AND LABOUR FORCE IN THE UNITED STATES

229

statisticians that reviewed the entire programme of the Bureau of the
Census in the latter part of 1953. It was determined from reviewing the
costs of alternative procedures that an increase in funds, which would
bring the annual cost of the Current Population Survey to about
$1,900,000, would significantly reduce the sampling variability of the
results and make possible the publication of regional information and
certain other details not previously publishable. At the expanded level
of the sample, when unemployment is at a level of 2,500,000, the standard
error of estimate is 75,000. There are about 10 chances out of 20 that the
results of a complete count would be between 2,350,000 and 2,650,000,
i.e., within a band of two standard errors above and below the figure
provided by the sample. Since government and other decisions that might
be made in response to changes in the level of unemployment do not
ordinarily depend upon a greater degree of precision than that furnished
by the new sample of 35,000 households, it was concluded that the
additional cost of a larger sample would not be justified.
The size of the errors due to sampling and response variability has
been kept within reasonable bounds by means of the wide geographical
dispersion of the sample and through direct supervisory contacts for
initial training, observation, quality check, and periodic face-to-face
refresher training sessions. The total cost is kept within reasonable
limits by having supervisory contacts made only when needed, by having
the enumerators on a part-time basis, and by establishing performance
standards governing the amount of work expected of all types of office
and field employees.
Selection of Areas
Establishment of Primary Sampling Units.
The 3,103 counties and independent cities of the United States were
divided into 1,891 primary sampling units (PSU), consisting usually of a
county or a group of contiguous counties. A separate PSU was set up
for each standard metropolitan area.1 Except in the case of the metropolitan areas, the PSUs were formed by combining dissimilar counties,
whenever possible. A typical PSU, for example, might include urban and
rural residents of both high and low economic levels and would thus
provide, to the extent feasible, a variety of occupations and industries.
Greater heterogeneity could have been accomplished by including more
counties, but this would have involved a disproportionate increase in
travel cost.
1
A metropolitan area consists of one or more counties including a city of at least
50,000: all of the counties in each such group of two or more counties having been
determined to have enough economic inter-relations to justify considering them as an
integrated unit.

230

FAMILY LIVING STUDIES

Selection of Sample of Primary Sampling Units.
Except for the 88 large standard metropolitan areas and four special
cases, the PSUs were put into 238 groups of approximately equal
population. The groups were so established that each included PSUs as
much alike as possible according to such characteristics as geographic
region, population density, rate of growth from 1940 to 1950, racial
composition, principal industry, type of agriculture, etc.
The 88 standard metropolitan areas and the four special cases were
automatically included in the sample. From each of the other 238 groups,
one PSU was selected in a random manner for inclusion in the sample in
such a way that the probability of selection of any one PSU was
proportionate to its 1950 population.
Selection of Households
For the country as a whole and consequently for each group of PSUs,
an over-all sampling ratio of about 1 in 1,480 was used in March 1959.
The sampling ratio used in each particular sample PSU depends on the
relation of its population to that of the group from which it was drawn.
Thus, in a sample area which was one-tenth of its group, the use of a
within-PSU sampling ratio of 1 in 148 achieves the desired ratio of 1 in
1,480.
Within each of the 330 PSUs, the number of households to be enumerated each month is determined by the application of a samphng ratio
rather than through the assignment of a fixed quota. The samphng ratio is
applied to small areas which have been selected with known probability
and designated on available maps. The application of this samphng
ratio is supposed to provide about six households each on the average
where there have been no shifts in the population. Where the population
is expanding, the number of households per area is likely to exceed six,
but where it is declining, the number per area found by the enumerators
may be substantially less than six on the average. The sample thus
properly reflects the changing distribution of the population and avoids
the distortion which would result from the application of fixed quotas of
households or persons.
To apply the samphng ratio within a PSU, several stages of sampling
were used. Use was made of enumeration districts, areas of about 800
population each, which were the assignments for individual enumerators
in the 1950 Censuses of Population and Housing. A sample of these
districts was selected, with the probability of selection of any one of these
proportionate to its 1950 population The selected districts were then
subdivided into segments, that is, small geographic areas with well-defined

POPULATION AND LABOUR FORCE IN THE UNITED STATES

231

boundaries including an average expected total of about six dwelling
units or other living quarters. Where roads, streams, and other natural
features were not available to subdivide an enumeration district into
enough subdivisions, some of the resulting segments were several times
the desired average size. For each subdivided enumeration district, one
segment was designated for the sample, with the probability of selection
of a given segment proportionate to the estimated number of dwelling
units within it. For the nation as a whole, approximately 6,000 segments
are in the sample in any given month. Where available advance information indicates that a selected segment contains about six households, all
units within the segment boundaries are to be included in the sample.
In cases where the advance information indicates a considerably larger
number of units, a field listing is made of all living quarters in the segment
and a systematic sample drawn so as to contain about six households.
Special mapping materials are used to subdivide enumeration districts
into segments and to determine in advance the approximate number of
units in each. In the larger urban places, information concerning the
number of units in each block according to the 1950 Housing Census,
together with the use of very detailed fire insurance maps, make it
possible to identify portions of blocks expected to have about six dwelling
units. In other urban areas, as well as in rural areas, maps of enumeration districts are used. Special procedures are necessary in areas with
large amounts of new construction. When such areas are identified, they
are mapped and subdivided, by means of field canvass as required, in
order to avoid segments of undue size. In order to ensure an unbiased
sample those segments selected from the regular 1950 census materials
that fall within the new construction areas are excluded.
In some rural areas, where the amount of information on available
maps is insufficient to permit setting up segments of the proper size, the
Census Bureau has used its own staff to carry out further field mapping
operations to give locally determinable boundaries for small segments.
This additional operation makes it possible for a large proportion of
rural segments to be of the same average size (averaging six units) as
those in urban areas, where mapping materials are generally more
precise. In such cases, the special listing operation described above is
not necessary.
Rotation of Sample
The accuracy of measurement of month-to-month changes is increased and the cost of field operations is reduced, when identical households are interviewed over a period of months. Experience has indicated,
however, that the co-operation decreases after a period of time, and a

232

FAMILY LIVING STUDIES

period of eight months has been selected as the total time for any given
family to remain in the panel. The system of rotation (four months in,
eight months out, four months in) gives good measures of month-tomonth and year-to-year changes without having an unreasonably long
period of inclusion for any particular household.

ESTIMATION PROCEDURE

Two stages of ratio estimates and a " composite estimate " are used.
It is possible to apply this rather complicated procedure rapidly and
automatically because of the availability of high-speed electronic computers. The principal steps are as follows :
Adjustment for Households Not Interviewed
The weights for all interviewed households are adjusted in the process
of estimation to the extent needed to account for the 3 to 5 per cent, of
occupied dwelling units from which no interview was obtained because of
absence, impassable roads, refusals, or unavailability for other reasons.
This adjustment is made separately by groups of PSUs and, within these,
by a number of characteristics.
Ratio Estimates
The distribution of the population selected for the sample usually
differs somewhat, by chance, from that of the nation as a whole in such
basic characteristics as age, colour, sex, and farm/non-farm residence.
These particular population characteristics are closely correlated with
labour force participation and other principal measurements made from
the sample. Therefore, some of the sample estimates can be improved
substantially if the original returns are so weighted as to make the
sample population approximate the known distribution of the entire
population with respect to these selected characteristics. Such weighting
is accomplished through two stages of the ratio estimates.
The first stage takes into account differences with respect to the
distribution by colour and residence (urban, rural farm, rural non-farm)
between (1) the 1950 actual population in four major regions, and (2) the
population in 1950 as estimated on the basis of the sample PSUs.
Therefore, in the first stage of estimation, special weights are applied
to the individuals in the separate colour and residence categories so as
to give a continuing adjustment to those groups reflecting the 1950

POPULATION AND LABOUR FORCE IN THE UNITED STATES

233

relation between the actual population of the nation and an estimate of
it based upon the PSUs.
The second stage of ratio estimates takes account of current differences between the population distribution of the sample and the latest
estimated distribution for the nation as a whole by age, colour and sex.
Within each age-sex-colour group the CPS sample returns, after they
have been weighted in accordance with the plan for the first stage of
ratio estimates as described above, are in effect used to obtain a percentage distribution by employment status and other characteristics.
These percentage distributions are applied to the independent population
estimates for each age-sex-colour group, in order to obtain the absolute
numbers used in part in the composite estimate described below.
Composite Estimate
The last step in the preparation of estimates makes use of a composite
estimate of each item, in the form of a weighted average of two estimates
for a given month. The first estimate is the result of the two stages of
ratio estimates described above. The second estimate is the preceding
month's final estimate adjusted by an estimate of change between the
two months based upon households included in the sample in both
months (75 per cent, of the total). These two estimates might be given
various weights to produce the most reliable estimate of level; in the
case of the labour force estimates the two are given equal weights, a
procedure resulting in some gain in reliability for practically all items.
For some items the gain in reliability in using the composite procedure is substantial. The gains are greatest in estimates of month-tomonth change, although they are significant also for estimates of level
in a given month or of change over longer periods.
Illustration of Method Used in Ratio Estimates and
Composite Estimate
This illustration of the way the estimation procedure works is based
on simple numbers (rather than actual data) to show how the procedures
described above are carried out.
Let us assume a non-interview rate in a given set of sample PSUs
for urban white households as 3 per cent, in March 1959. Further let
us assume that the first-stage ratio estimate factor for a given sample
PSU for urban white is 1: 032, reflecting the fact that in this region the
1950 number of urban white persons was 3.2 per cent, above the number
in a regional estimate based upon the sample PSUs. Then, if there were 50

234

FAMILY LIVING STUDIES

white urban persons in the sample in this sample area the combined
effect of the non-interview adjustment, the first stage ratio estimate and
the over-all weight would be to give an estimate of 78,659 (50 x 1.03
x 1.032x1,480).
The second stage of ratio estimation involves adding up these estimates for the various age-sex-colour classes of persons. One of the groups
thus available is the number of white males aged 50 to 54 in March 1959.
When this total is compared with the corresponding class from the latest
estimated distribution for the nation as a whole, we may find that the
latter is 1.8 per cent, greater than the number after the first stage of ratio
estimates. The ratio of 1.018 is then used as an additional factor to
obtain the final weight assigned to each person in the group of white
males 50 to 54 years old. Thus, if there were 150,000 unemployed white
males aged 50 to 54 estimated after the first stage of ratio estimation,
then the estimate after the second stage would be 152,700 (150,000 X
1.018).
To obtain the final or composite estimate of unemployed white
males aged 50 to 54 in March, suppose we have the following estimates :
February final composite estimate
175,000
February estimate based on segments common to March . . . . 180,000
March estimate based on segments common to February . . . . 150,000
March estimate after two stages of ratio estimate
152,700
Then, the final composite estimate for March of unemployed white
males aged 50 to 54 would be 148,850, obtained from: i/2(175,000-30,000),
the figure 30,000 representing net change between February and March
based on common segments, plus l/¿ (152,700).

CONTROLLING QUALITY OF RETURNS

The policy of the Bureau of the Census is to provide, as consistently
as possible, measures of the accuracy of all censuses and surveys which
it conducts. In the reports on the Current Population Survey a technical
section on reliability of data is regularly included, containing either a
statistical table or some illustrations of the range of sampling variability
in the figures. The availability of high-speed electronic equipment enables
the Bureau to provide considerably more information of this type than
was possible earlier. Statements are also included to caution readers
about response variability, reporting errors and other biases for which
specific measurements are not available.
In the following paragraphs particular attention will be given to the
problem of measuring and controlling errors that arise in connection

POPULATION AND LABOUR FORCE IN THE UNITED STATES

235

with the field work. There are, of course, other types of errors that affect
results. Among these, three in particular may be noted. First, from
3 to 5 per cent, of the households are not actually enumerated in a typical
month, and a certain amount of error arises from assuming for these
households characteristics similar to those obtained for the interviewed
households. Second, the independent estimates of the population to
which the monthly results are adjusted in the process of estimation are
not precise. Third, despite the use of quality control procedures in all
phases of the office processing, a certain number of errors remain in the
final tabulations. The effects of these errors are believed not to be large.
It is probably surprising to many users of census statistics in the
United States that decennial census figures on the labour force, the
number employed and the number unemployed, are not used as benchmarks for the Current Population Survey. On the contrary, matching
studies between the Census and the CPS and other detailed comparisons
have led to general agreement by technicians and users alike that the
monthly survey statistics on the labour force—in spite of appreciable
sampling variability—are more reliable for national totals than are the
comparable figures from the population censuses. The greater accuracy
in the sample survey results is generally reflected, in the case of labour
force statistics, in a more nearly complete identification of part-time and
occasional workers and job seekers such as housewives, students, and
older persons on the verge of retirement.
These differences arise mainly from the fact that a small, well-trained
group of interviewers can apply complicated definitions and concepts
more reliably than the large group of one-time enumerators recruited
for a decennial census. The effects of response bias on certain inquiries
in a census can easily outweigh the sampling variability in a fairly large
sample survey.
Measures of Variability
Methods are available for measuring the effect of both sampling and
response variability. In the Current Population Survey, where the
probability of selection of each member of the population is known,
the measure of sampling variability indicates the range of difference that
may be expected because only a sample of the population is surveyed.
A measure of response variability indicates the range of difference to be
expected as a result of the compensating types of errors arising from
the practices of different interviewers. This tends to become small when
a sufficiently large enumeration staff is employed.
In practice these two sources of error, sampling and response variability, are estimated jointly from the results of the survey. Response

236

FAMILY LIVING STUDIES

variability is measured in the CPS in the non-self-representing PSUs
(i.e., all the PSUs except the 88 standard metropolitan areas and four
special cases) as part of our over-all measure of variability. The model
which is used in creating estimates of variability measures sampling
variability and response variability jointly. In self-representing areas
randomisation of assignments between pairs of enumerators would
permit us to measure response variability for this set of PSUs by treating
each enumerator's assignment as a clustered result.
The computations do not however measure the effect of response
bias, that is, the systematic errors of response which may take place in a
sample survey as well as in a complete census, and which may, in fact,
be larger in the census. The reinterview programme, discussed later,
has furnished some indications of the direction and approximate magnitude of some of the biases in the survey. Since the measurement of
response bias is so difficult, it is not possible to predict how soon techniques will be available to produce satisfactory estimates of this type of
error.
Programmes to Ensure and Control Quality in Field Operations
Because of the potentially serious impact of response errors and
biases on the statistical results, about one-third of the field collection
budget for the Current Population Survey is devoted to various procedures designed to ensure and control quality. The major procedures
include training, observation, reinterview and inspection of returns.
Training.
New interviewers are given extensive training during the .first three'
months of their employment. The programme includes some classroom
training and home study, but emphasis is given to training on the job
while being accompanied by a supervisor. Experienced interviewers
receive an average of three hours of home study work just preceding
each monthly enumeration, including test exercises which are graded by
the field supervisor and discussed further, if required. Interviewers are
brought together about four times a year, for group training sessions, for
joint discussion of problems, exchange of views, and special demonstrations. The group sessions are frequently held at the time of difficult or
extensive supplementary inquiries.
Observation.
Observation by a supervisor during the course of an actual interview
is valuable not only as a training device for new interviewers, but as a

POPULATION AND LABOUR FORCE IN THE UNITED STATES

237

means of maintaining and raising the standards of experienced interviewers. In such cases, the supervisor evaluates the skill of the interviewer
in phrasing questions or interpreting replies. The supervisor observes the
enumerator's adherence to procedures, efficiency in planning his work,
including itinerary and revisits and his courtesy and tact in answering
questions raised by the respondents. The standard programme requires
supervisors to observe experienced interviewers about twice a year, for a
period of one-half to one day on each occasion. Provision is made for
additional observations of interviewers whose performance has been
substandard. Rating forms are completed by the supervisor in the course
of these observations both to serve as a record of performance and to
identify weaknesses which should be rectified through additional training.
Reinterviewing.
The reinterview programme which has been in effect for several years
requires the supervisor periodically to reinterview a subsample of households already enumerated by an interviewer. A third of the households
in an enumerator's monthly workload are reinterviewed on an average
of four times a year, on an irregular schedule so as to avoid advance
knowledge of when the check will occur. The reinterviews generally
occur in the week following the completion of the original interviews,
but the questions refer back to the original enumeration week. The
supervisor conducting the reinterviews has the original information in
his possession for 80 per cent, of the reinterview assignments, but does not
refer to it until the reinterview is completed. At that time, he compares
the results of the first and second interviews and discusses any differences
with the respondent to determine which report is more nearly accurate
and, if possible, the reasons for the differences. For the other 20 per
cent., his task is completed when he has finished the reinterview. Errors
in the original report are brought to the attention of the interviewer and
additional training is given where required. A special rating is also
computed for each interviewer on the basis of the check, and one for
each supervisor based on comparison of differences between the 80 per
cent, and 20 per cent, of the reinterview results. The results of these
various checks lead to special training programmes as needed, and in
some cases to the replacement of interviewers who, after special attention
and training, remain unable to meet the prescribed standards of quality.
Over-all results are also prepared for administrative use, but they
are not regarded as sufficiently precise for publication as specific measures
of response error. The results provide, however, a valuable indication
of response error, even though they also include differences in the interviewer and respondent in the two interviews, as well as differences in time
16

238

FAMILY LIVING STUDIES

and other elements. The results help to indicate the kinds of bias that exist
and guide the preparation of training materials.
The reinterview programme includes not only the checking of responses to the regular and supplementary questions included in the survey,
but also an investigation of deficiencies ÌD the coverage of living quarters
and population.
Inspection of Returns.
A careful review of the completed survey questionnaires for completeness and consistency is made in field offices, especially for the newer
interviewers. This review on regular monthly items is omitted for those
interviewers whose work over a period of time has met or exceeded standards, although some check of the entries for supplementary questions
included in particular months is made even in those cases. All questionnaires are given a review on tabulation equipment in Washington, and
any deficiencies found are reported to the field offices. In the main, the
rate of omissions and inconsistencies has been exceptionally low for
regular monthly items (generally a fraction of 1 per cent.) and fairly
small also for most supplementary questions. As a result, it has generally
been possible to assign values to this small group of unknown cases on
the basis of known distributions, without risking any material bias in
the results.
Some Conclusions from the Control Programme
The results of the programme of reinterview and various other detailed inquiries are consistent with many other efforts to validate censuses and surveys pertaining to the social and economic status of the
population. On the whole, the biases as well as the net effects of response
variability are comparatively small except for small areas where the
canvassing is performed by a very small number of enumerators, but the
gross differences due to response variation may be of considerable
magnitude. In the monthly survey the size of the variations is doubtless
partly attributable to the considerable fluctuations in the labour
force as women, students and various marginal workers move in and
out in response to seasonal and other influences. For many persons,
attachment to the labour force is not a fixed fact, but an attitude that
may vary considerably according to small variations in the way the
questions are asked or the combination of circumstances at the time of
interview.
The biases observed in the data are rather well catalogued by type,
even if they are not yet well measured. Strong evidence exists of a slight
understatement in the size of both components of the labour force, the

POPULATION AND LABOUR FORCE IN THE UNITED STATES

239

employed and the unemployed. The omissions occur mainly in the same
groups which are much more seriously under-represented in decennial
census results—housewives, students, and older men who work or seek
jobs intermittently. In the case of hours' worked, there is a clear tendency
to report too many persons as working the typical week of 40 hours and
to report too few working shorter or longer hours. The understatement
of part-time work seems to result mainly from failures of respondents
to recall short periods lost because of illness or taken off for personal
reasons, rather than from under-reporting of reductions in hours due to
economic factors. A part of the understatement in hours of work above
the average appears to be attributable to failure to report hours worked
on extra jobs at night or on week-ends.
An interesting phenomenon for which no concrete explanation has
been found is known as the " first month bias " in unemployment.
Tabulations of Current Population Survey data have repeatedly shown
that the unemployment rate in households being interviewed for the
first time is slightly but significantly higher than that in households in
their second or later months of enumeration. Reinterview results have
thus far failed to show whether the " first month " rate is an overstatement or the subsequent rate an understatement.

CURRENT POPULATION SURVEY SUPPLEMENTS AND SPECIAL INQUIRIES
RELATING TO SOME ASPECTS OF FAMILY LIVING

Although the primary purpose of the Current Population Survey
is the collection of monthly labour force data, the survey has been used,
from its inception, to produce data on a variety of subjects. In fact,
social scientists have come to realise, that for national figures the CPS is
in many respects a social survey for the United States.
Some of the studies based on information regularly collected in the
CPS were mentioned in the first section of this article. A number of
the supplementary inquiries, particularly those on migration, personal
and family income, and household composition probably relate as much
to the analysis of family living conditions as they do to an understanding
of activity in the labour market. For example, the CPS income report,
published annually since 1945, provides a fairly complete picture of
sources of family income. This report gives annual data on family
income in relation to size of family; number of earners; age, sex, colour,
occupation, and industry of the family head; employment status of the
wife; sources of income; and numerous other factors which would
appropriately be considered an integral part of any family living study.

240

FAMILY LIVING STUDIES

Similarly, the annual CPS report on household composition provides
valuable data on household formation, family living arrangements, and
other factors related to family living. In general, the use of detailed
cross-tabulations to show relations among such matters as employment
status of wife, number of children, family income, occupation of head,
education, and many other factors can throw much light upon patterns of
family living.
It would be possible to use the monthly survey more extensively
than in the past to show dynamics of change in family living conditions.
Through the availability of three-fourths of the sample for month-tomonth comparisons and one-half of the sample for year-to-year comparisons, considerable knowledge can be obtained of gross changes from
period to period and of the much smaller net changes that result from the
off-setting of considerable parts of the gross movements. It might prove
possible also to spread the collection of income and expenditure data
over more considerable periods of time, thereby reducing the burden
imposed upon a household in any single visit and possibly increasing
the reliability of replies because of shorter periods over which income
and expenditure need to be recalled.
Some special studies have been added to the CPS from time to time.
The purpose of these studies has almost invariably been to provide
current data on an important social or economic problem, usually to a
government agency, but on occasion to a private research organisation.
The following are some of the more important CPS supplements that are
related to family living even though the underlying purposes were often
completely different from those served by family living studies.
Study of Low Income in New York State
In December 1955, the Governor of New York State appointed a
committee to survey the characteristics of low-income groups in the
state and to suggest proposals to improve their economic status. This
programme was designed to identify the population at substandard
levels of living, to investigate the causes of their low income, and to lay
the bases for a long-range co-ordinated effort to improve conditions.
State sources such as public assistance agency records and unemployment insurance records did not provide a picture of the entire population.
This information could only be obtained from a household survey, and
the Bureau of the Census was requested by the state to conduct such a
survey throughout the state. The 3,500 households included in the
Bureau's regular monthly survey in New York State provided the nucleus
of the survey. Information was obtained for these and 3,000 additional

POPULATION AND LABOUR FORCE IN THE UNITED STATES

241

households on work experience, detailed sources of income, education,
housing, nativity, current debt status, and family composition. New
York City, acting as a co-sponsor of the survey, provided funds for the
collection of similar data for certain ethnic groups within the city known
to contain a substantial percentage of low-income families. For this
purpose a supplementary sample of 1,800 households was enumerated
in predominantly Puerto Rican and coloured neighbourhoods.
Study of Consumer Purchases, Credit, and Debt
In connection with a comprehensive report on consumer instalment
credit, at a time when such credit was increasing rapidly, the Federal
Reserve Board in 1956 sponsored a household survey in which the CPS
was used to obtain information on the characteristics of borrowers. A
number of questions relating to family income, cash and credit purchases
of new and used cars and homes, home ownership, and current debt
status were added to the regular CPS schedule. In this way it was
possible to obtain credit and related information from a representative
sample of about 26,000 households. The use of the CPS schedule
automatically made it possible to relate the new information on credit
use and debt status to the detailed demographic and economic characteristics regularly collected in the CPS. It was therefore possible to provide
data on purchases of various items classified by such factors as type of
household, age of head, income, colour, employment status, number of
workers in the family, region, residence, and degree of urbanisation.
The flexibility of the CPS as an analytical tool for current problems
is clearly indicated by the speed with which the results of this survey
were produced. The results were submitted to the Federal Reserve Board,
ready for analysis, in September 1956, about six weeks after the households were interviewed. Thus, within this brief period, information on a
vital national problem was collected, processed, and submitted to the
sponsor in a form suitable for analysis.
Hospitalisation Survey
One of the perennial problems that confront the United States Public
Health Service is the establishment of a valid standard of needs in planning
for general hospitals. In 1947, uniform standards for the number of beds
needed came into general use with the development of state-wide hospital
plans as a result of federal legislation providing for grant assistance in
planning and construction. In the subsequent decade, population growth
and changes in the care and prevention of disease modified requirements

242

FAMILY LIVING STUDIES

for facilities needed to ensure adequate care. It was concluded that the
new levels of need could not be ascertained from studies of hospital
records alone, since these records cover only persons who used hospitals.
Therefore, as a first step in redefining the standards of need for general
hospitals, the United States Public Health Service contracted for the use
of the CPS to survey the level of use of general hospitals. For this
purpose, differentials in level of use were regarded as more important
than actual totals.
In September 1956, 27,000 CPS households (three-fourths of the
entire sample), consisting of about 85,000 persons, provided histories of
the hospitalisation and out-patient care received during the previous
12-month period. For each person in the sample, information was
obtained on the frequency and duration of medical care, and on the name
and location of hospitals or related facilities in which such care was
received. Personal characteristics such as residence, age, sex, race,
veteran status, and occupation were available from the basic CPS
schedules, and additional information was obtained for each household
on family income, hospital insurance coverage, and methods of payment for hospital care. Although these statistics were requested by the
Public Health Service for a special purpose, the applicability of the information to studies of family living is apparent.
Experimental Surveys of Consumer Buying Plans
The economic recession of 1957-58 once more brought to the attention of many economic analysts the potentially great usefulness of
information on the short-run buying plans of consumers of durable goods.
Although the government had available a considerable number of economic indicators regarding changes in employment, unemployment, and
business and government expenditure, it had inadequate information on
the plans of consumers to buy durable goods. The Council of Economic
Advisers recommended to the Bureau of the Budget that such information
be collected on an experimental basis. The Budget Bureau suggested
that the Federal Reserve Board be the sponsor of the survey and that
the Bureau of the Census use the CPS for the purpose.
During the latter part of 1958 extensive testing and experimentation
were conducted in 1,300 CPS households in the Chicago area. The
purpose of these tests was to develop questions, methods, and procedures
for collecting, processing, and tabulating information on consumer
buying plans. On the basis of these tests, quarterly surveys of consumer
buying plans were conducted in 1959 and 1960 as supplements to CPS.
In January, April, July, and October 1959, 18,000 households (one-half

POPULATION AND LABOUR FORCE IN THE UNITED STATES

243

of the entire CPS sample) were asked about purchases and plans to purchase (within the next 3, 6 or 12 months) new and used automobiles,
refrigerators, washing machines, and television sets. In the case of automobiles, additional questions were asked about method of payment, debt
status, whether major repairs were required in last 6 months, and satisfaction with the present vehicle. The survey also included questions about
family income, changes in income and financial status during the past
year, and expected income changes during the coming year.
In the 1960 surveys the questionnaire was expanded to include
information on purchases and plans to purchase homes, air-conditioners,
clothes dryers, dish washers, and hi-fi equipment. The results of these
surveys are now published quarterly in the Federal Reserve Bulletin, and
the surveys will continue, using the present format, during 1961.
One approach to validation of survey data on consumer buying
intentions is to compare the expressed intentions with subsequent
purchases by the same families. The sample was selected in such a way
that one-half of the households in one quarterly survey were included
in the subsequent quarterly survey. Thus, for example, 9,000 of the households that were asked about their 3-month buying plans in January
were also interviewed in April and asked if they purchased any of the
items in the preceding quarter.
Survey of Aged Persons
The past few years have witnessed a great expansion of interest in
(and information about) problems of the aged. This increase is attributable at least in part to the fact that the aged are becoming of increasing
importance in the population of the United States. In 1900, only 4 in
every 100 were 65 years old or over. Although the rise in the proportion
of aged persons in the population appears to be slowing down as a result
of the sustained high birth rate, the proportion will nevertheless
probably approach 9 or 10 per cent, by 1975.
There are numerous manifestations of increased interest in the
problems of the aged. To turn to one example of many, the Institute
of Industrial Relations of the University of California received a grant
from the Rockefeller Foundation to conduct a research project on the
social and economic consequences of an aging population. 1 Although
many sources of information were used in completing this project, a
major segment of the data came from the CPS.
1

The results of this study, based to a large extent on CPS data, were published

in a monograph by P. O. STEINER and Robert DORFMAN : The Economic Status of

the Aged (Berkeley, University of California Press, 1957).

244

FAMILY LIVING STUDIES

Following the completion of the April 1952 CPS, a 60 per cent.
subsample of persons 65 years old and over were revisited for further
information on a number of subjects, including previous work experience,
living arrangements, means of support, housing accommodation,
savings and assets, medical expenses, and various other items. Thus,
the reinterview survey was designed to provide a fairly complete picture
of the living arrangements of the aged population in the United States.

SUMMARY

In summary, the Bureau of the Census regards the Current Population
Survey as a highly flexible tool which has been applied to serve many
purposes under a great variety of conditions. It has proved highly
useful during periods of large-scale unemployment. On the other hand,
it has met equally important but different needs also in a period when
the pressure upon the nation's human resources was so great that unemployment dropped to extremely low levels and millions of persons
were drawn into the labour force either temporarily or for considerable
periods of time.
Many methodological improvements have been introduced into the
survey since its outset, and consideration is continuously being given
to other possible advances. Some experimentation has been conducted
in the phrasing and arrangement of questions in the eventual hope of
improving the classification of occasional workers and the reporting on
such items as hours worked, duration of unemployment, and occupation
and industry. Introduction of a composite estimation procedure for
year-to-year changes has been studied in the interest of increasing the
reliability of such annual series as income and farm population figures.
The many supplements which have been based upon this survey have
repeatedly demonstrated the wide range of inquiries that are appropriate
for a well-designed sample of households. Some of them have been
closely related to family living studies, while others have been of value
primarily for measuring market conditions, pressure upon financial
resources, and related matters. The survey has also served as a laboratory
for experimentation in connection with many technical problems of
survey design and response research. Despite the fact that it has been in
operation for about twenty years, it seems safe to conclude that many
important advances lie ahead, and that the opportunities for progress
and exploitation of this resource are fully as great today as they have
ever been.

XIV
An Inquiry into Levels of Living in an
Area of the Ivory Coast
By J. L. BOUTILLIER and

J. CAUSSE *

Until recently a few monographs by individual research workers
were all that existed in the way of studies on the living conditions of
families in rural areas in tropical and equatorial Africa. These monographs covered only particular aspects of the subject, as each specialist
concentrated on the problems of special interest to him: the nutrition
expert concentrated on analysing food intake, the agronomist on systems
of cultivating the land and the sociologist on housing or family structure.
Moreover, as financial resources and qualified personnel were limited,
the number of families studied was always very small, and the period of
observation was frequently not long enough to give the findings a sufficiently representative character.
The first large-scale inquiries employing statistical methods were
concerned with population or agriculture ; however, it soon became clear
that other phenomena, such as levels of living, could be examined by
similar techniques. The selection of random samples and the local
recruitment and training of indigenous investigators were experimented
with in the study of a variety of economic problems, such as production,
goods produced for own personal consumption and family budgets.
Moreover, in view of the fact that political developments were awakening
in the young nations of west Africa an awareness of the prerequisites
for economic development, studies of this kind could be of considerable
immediate interest.
Various statistical inquiries have been carried out since 1950 in a
number of west African countries, the most important being Nigeria,
Ghana, the Ivory Coast, Senegal and Sudan. Some special features of
such inquiries in African territories will be described here, using for
* Of the Overseas Scientific and Technical Research Office and the National
Institute for Statistics and Economic Research, France, respectively.

246

FAMILY LIVING STUDIES

illustration the examples drawn from the study carried out in the subdivision of Bongouanou (Ivory Coast) in 1955-56, the most striking feature
of which was that it covered nearly all the possible indicators of levels of
living.
BRIEF DESCRIPTION OF THE BONGOUANOU INQUIRY

Organisation
The responsibility for carrying out this inquiry, which was undertaken at the request of the Ivory Coast Government, was entrusted to
two bodies, the Ivory Coast Statistical Service and the Overseas Technical
and Scientific Research Office. In view of the multi-purpose nature of
the inquiry and the scale of the resources required, the collaboration of
other bodies was arranged. First, in view of the experimental character
of the inquiries, the Food and Agriculture Organisation of the United
Nations was asked for assistance; it agreed to participate by providing
a statistical expert for five months while the methods of investigation
were being worked out. In addition, financial assistance and staff were
provided by the African Food and Nutrition Research Office
(O.R.A.N.A.), the Federal Statistical Service in Dakar, the General
Mobile Health and Preventive Medicine Service, the Health Service and
the Agricultural Service of the Ivory Coast, the National Institute of
Statistics and Economic Research (I.N.S.E.E.) and the Overseas Statistical
Service in Paris.
The inquiry was directed by an economist and a statistician, assisted
by a technician and a woman welfare worker. In addition, a physician
took part in the clinical investigations, working in the field for two
months. The results were compiled and analysed by the same team,
assisted by five clerks trained on the job.
Objectives
The principal objective of the inquiry was to obtain all the indicators
needed to determine the levels of living of populations living in the
selected subdivision. However, in view of the special importance of
certain problems in that particular region, and even in the country at
large, particular attention was given to the demographic situation,
health, nutrition, housing, agriculture and family budgets.
Area Studied
The area of inquiry, which covered a full year (1 April 1955-31 March
1956), was the Bongouanou subdivision of the Dimbokro district. It is

LEVELS OF LIVING IN THE IVORY COAST

247

situated in the southeastern part of the Ivory Coast, astride the fourth
degree of longitude West, and between the latitudes 6° 30' to 7° North (of
the equator). It has an area of 3,000 square kilometres and a population
of approximately 65,000. The population density is thus 22 per square
kilometre—a relatively high level for tropical (and equatorial) Africa.
The indigenous population, which makes up approximately threequarters of the present population, belongs to the Agni ethnic group,
which is related to the Akan group. The remainder of the population
is made up of immigrants who entered the area relatively recently; this
group, of very mixed ethnic composition, came mostly from countries
to the north (Sudan, Upper Volta). The entire region lies within the
forested zone in the southern part of the Ivory Coast. This forest
environment and favourable climatic conditions (two rainy seasons and
a rainfall of 72 inches per year) have facilitated the rapid development
of tree-crop farming (cocoa and coffee), which is now carried on side
by side with the cultivation of traditional subsistence crops.
The profitability of the cash crops and the ease with which land can be
obtained have gradually, over the last 30 years, led all the farmers to take
up this cultivation. It is a remarkable fact that in the Bongouanou
subdivision nearly all the cocoa and coffee produced is grown on
African farms; there is only one European-owned plantation. Another
remarkable characteristic of this expanding economy is the fact that the
African planters employ wage labourers from among the immigrants.
The Sample Design
A two-stage sample design was used. In the first stage, villages were
selected by the equal probability method, the resultant sample being
adjusted by Yates's method to obtain proper representation according
to tribe and size. The sampling unit chosen at the second stage was the
" concession " or " rectangle ". The sociological patterns of the Agni
group are very similar to those of the Ashanti group (social organisation
based on lineage with matrilineal descent predominating). Thus a
" rectangle " consists of one or more related households; it is a unit of
habitation and community Ufe, comprising an average of 25 persons.
This unit was chosen, rather than a household, owing to the impossibility
of estimating the food consumption of individual households, for within
the rectangle the latter often do their cooking or take their meals together.
To ascertain seasonal variations in food consumption and family
expenditure—which might be considerable—the year was divided into
three four-month periods, and one-third of the sample was replaced in
each period. The sample was made up as follows:

248

FAMILY LIVING STUDIES

(a) For the Agni population the sample comprised 162 rectangles,
distributed among nine villages and having a total number of 4,100 inhabitants. The sampling fractions were, at the first stage, 20 per cent., at
the second stage 40 per cent., over-all 8 per cent.
(b) For the foreign population, investigations were undertaken in
60 concessions, distributed among 13 dyoulakro (alien quarters in Agni
villages) and containing 720 inhabitants. 1 The sampling fractions were :
at the first stage 25 per cent., at the second stage 35 per cent., over-all
9 per cent.
Collection of Data
As over 95 per cent, of the population was illiterate, no written
material was given to respondents, and information could be obtained
only by questioning, measurement or direct observations. In view of the
relatively advanced level of development reached by the economy under
study, direct factual observations were collected on a day-to-day basis ;
the respondents were not interrogated about past events. Wherever
practicable, and if finances of the inquiry allowed, actual measurements
were made. All information on food consumption was obtained by
weighing; and information on areas under cultivation was obtained with
a compass and surveyor's chain, while yield and production were measured with a steelyard. In nearly every field, however, there were certain
items of information that could not be obtained by direct observations,
and in such cases questioning was resorted to. Examples are data about
the age and cost of buildings, and the dates of arrival and reasons for
immigration of the immigrant population.
In view of the large number of subjects to be covered by such a
study in a country in which no previous information on levels of living
was available, over a score of questionnaires had to be drawn up. They
contained a large number of questions and space for factual observations,
such as weighings, areas, etc.
Pilot Study—The Recruitment and Training of Investigators
The pilot investigation and the amendment and try-outs of the draft
questionnaires took place concurrently with the training of the investigators in one of the villages in the subdivision covered by the inquiry
where board and lodging had been made available for the supervisors
and the 40 investigators in training by a particularly helpful chief. The
1
This part of the inquiry covered only aliens living in Agni villages. The rest of
the immigrant population, mostly alien agricultural labourers, was studied at the
same time as the Agni families for whom they were working.

LEVELS OF LIVING IN THE IVORY COAST

249

theoretical training of the investigators (purpose of the inquiry, general
description of questionnaires, etc.) was reduced to a minimum. The
questionnaires took final shape only after numerous amendments had
been made on the basis of repeated try-outs carried out among the
families in the village. This six-week period of isolation in the actual
surroundings in which the inquiry was to take place, and in direct and
continuous contact with the population, made it possible to adapt the
methods of investigation, the sampling plan and the questionnaires to
local conditions.
At the end of the training course and the pilot investigation, which
lasted six weeks, some 30 investigators were selected out of the 40
trainees. Most of them were at the level of the primary elementary
school certificate (although this was not required) and belonged to the
same ethnic group as the respondents. They were recruited either from
among the investigators who had just completed an agricultural investigation in the Bouaké region or from among candidates presented by
persons of some local standing (the chief of the subdivision, the doctor,
village chief, etc.) after rapid tests designed to eliminate candidates
intellectually unsuited for that kind of work.
Supervision of Investigators
The work of the investigators was supervised at two levels. Firstly,
three agricultural instructors and the health service technician made a
preliminary check of the information collected in the field. They were
also responsible for passing on instructions, dealing with administrative
questions and public relations (with the population) ; they also collected
the completed forms. Subsequently, at a higher level, all completed
questionnaires were meticulously examined, tested and checked in
various ways for reasonableness and consistency, either at the centre or
in the field. In addition, the tabulation of the information was begun as
soon as the questionnaires had been checked; this made it easier to
detect anomalies and persistent errors made by particular investigators.
Co-operation of the Public
No compensation was paid to the respondents except where samples
of produce (maize, rice, groundnuts, coffee, etc.) were taken during
sampling investigations of crop yields.
Generally speaking, the investigators were warmly welcomed by the
population. Their acceptance was facilitated by the fact that the inquiry
into family budgets and food consumption was carried out simultaneously
with the medical observations, which were accompanied by the provision

250

FAMILY LIVING STUDIES

of treatment. In most of the cases where information was refused, the
hostility was dispelled quite easily by a visit from one of the supervisors.
In fact, only 4 per cent, of the families refused to give information, and
these were families who provided information in the first round but were
unwilling to co-operate a second or third time.
Quality of Data Collected
The information collected was in general extremely satisfactory,
largely because the multiplicity of questionnaires made a considerable
amount of checking and cross-checking possible. However, some
returns (especially on food and agriculture) had to be collected again,
particularly during the early stages of the inquiry. Two investigators who
proved unsatisfactory had to be removed a few weeks after the inquiry
was begun. In certain doubtful cases, the particular items of information
were obtained again by a different investigator.
Data Processing and Publication of Results
As was pointed out earlier, the greater part of the tabulation of the
information obtained was effected in an office set up for the purpose at
Bongouanou itself. Only at the last stage of the compilation of the
information on foodstuffs were the mechanical facilities of the Statistical
Service in Abidjan made use of; the final analysis of the results was
begun in Bongouanou and completed in Paris.
Recognising that the findings in certain fields of particular interest to
the technical services in the territory should be made available as quickly
as possible, seven preliminary reports were issued in mimeographed form
while the inquiry was actually going on. The complete results were
published in 1958 and I960.1

CRITERIA FOR JUDGING LEVELS OF LIVING

The choice of criteria for the determination—and, in certain cases, the
comparison—of levels of living in tropical and equatorial Africa is
fraught with delicate and complex problems. On the one hand, appearances are often deceptive, while on the other hand, a partial study of the
situation may lead to wrong interpretations.
1

SERVICE DE LA STATISTIQUE ET DE LA MÉCANOGRAPHIE DE LA COTE D'IVOIRE:

Enquête nutrition-niveau de vie, subdivision de Bongouanou, 1955-1956 (Paris, 1958)
and J. L. BOUTILLIER: Bongouanou, Cote d'Ivoire, in " L'homme d'outre-mer " collection (Paris, Berger-Levrault, 1960).

LEVELS OF LIVING IN THE IVORY COAST

251

Appearances are certainly deceptive. A person travelling hurriedly
northwards from the coastal region of the Ivory Coast passes first of all
through the villages of the forest zone; the huts have rusty sheet-iron
roofs, the streets are deeply rutted and during the rainy season men and
beasts have to make their way through deep mud. The impression left
with the traveller is most unfavourable; these villages inevitably remind
him of the wretched suburban districts of the large tropical cities. On the
other hand, as soon as he enters the savannah the contrast is striking;
the huts have pretty thatched roofs, huts and outbuildings are kept in
good condition, the streets and avenues are so clean that one might
think that they are kept continuously swept; in short, the villages look
attractive and comfortable. The reality of the situation is quite different ;
for example, the average money income per head is over three times
greater in the forest zone than that in the savannah zone.
One example will suffice to show that the study of a single aspect of the
levels of living of a population group about which so little is known is
totally inadequate. A clinical examination of the Agni children studied
in the course of the Bongouanou inquiry showed them to be in a satisfactory state of health; on the other hand, the population inquiry showed
the infant mortality rate to be very high. These two factors considered
separately point to diametrically opposed conclusions. Examples of this
kind are certainly not rare in Africa, where the law of natural selection
operates mercilessly. The level of sanitation is so appalling that the
infant mortality rate is very high (about 200 per 1,000), and the incidence
of death among children up to the age of ten years is also very high, as
they are weakened by endemic diseases such as malaria and intestinal
parasitic diseases. Only the strongest children with the highest resistance
are likely to survive.
In view of these considérations, the organisers of the inquiry decided
to undertake the study of living conditions at three separate levels—
biological, economic and sociological.
In addition, part of the inquiry was given a practical slant at the
request of some of the bodies providing financial assistance. The sources
of information taken into consideration at each of the three levels is
briefly indicated below. In some cases the existing information sufficed;
in others the mission of inquiry had the task of obtaining the principal
information necessary.
Biological Level
The inquiry into biological conditions was conducted in four
different fields, namely—

252

FAMILY LIVING STUDIES

Climatology.
Observations were regularly made by the agricultural service.
Demography.
The inaccuracy of the administrative censuses and the extremely
summary character of the registers of births, marriages and deaths,
were such that the mission was practically forced to obtain the different
population and vital statistics required by random sampling methods.
A house-by-house census was carried out in the villages selected at
the first stage of sampling; subsequently the women in the sample were
questioned and the births, marriages and deaths were registered. This
gave the conventional data on the population situation and trends
(sex, age, activity, fertility, birth rate, death rate, growth, etc.).
Health and Sanitary Conditions.
Some of the information was collected through the intermediary of
the doctors of the health service. Reliable information on the state of
health of the population was desired. However, as for both material and
psychological reasons it was impossible to make a systematic study of a
sample representative of every age and sex category in the population,
the mission had to confine itself to clinical examinations and related
laboratory analyses covering only children up to 14 years of age. There
were a number of reasons for following this procedure. The analysis of
the death rate had shown that persons in this age group were particularly
vulnerable to certain afflictions (malnutrition, endemic diseases, etc.).
Besides, mass'clinical study is very difficult to organise; where, for these
people, doctors go hand in hand with treatment and sickness, it is difficult
to induce any substantial proportion to undergo medical examinations.
Attendance was fairly irregular even among the children, whose time was
less taken up by agricultural work; attendance among adults would have
been even lower.
Finally, direct observation of the diet, although an onerous task, was
an extremely important one; firstly, from the biological standpoint,
because, taken together with the clinical investigations, it revealed the
nutritional balance of the population, and secondly from the economic
standpoint, as it enabled estimates to be made of the subsistence production.
As regards attitudes towards health problems, the investigation into
family budgets showed the relative importance which the inhabitants of
the region attach to the two principal types of care available in the area—
native medicine on the one hand, and modern bought medicines (particularly antibiotics) on the other. Although the use of antibiotics has

LEVELS OF LIVING IN THE IVORY COAST

253

increased strikingly and sometimes gives rise to abuses (an injection of
penicillin tends to be considered a universal cure-all), the Agnis frequently try the resources of traditional medicine, dispensed by the
village witch doctor, before applying to the nearest dispensary or the
subdivision hospital.
Food and Nutritional Balance.
The mission investigators weighed all the foodstuffs used in the meals
taken by sample families (second-stage sample units) for seven or
fourteen consecutive days, using Roberval-type scales. The information
thus obtained was supplemented by questionnaires and measurements to
ascertain food consumption at other than meal times. Lastly, food
samples were taken and sent to the O.R.A.N.A. laboratories in Dakar
for analysis, so that the F.A.O. food tables could be adjusted for local use.
Economic Level
The large extent of production for own consumption—all consumption is " personal " including purchased items—in underdeveloped areas
should be stressed here. Unfortunately, the scope of consumption
studies is all too often confined to food purchased. In the underdeveloped
countries, however, where often as much as 90 per cent, or more of the
population is rural in character and engages mainly in agriculture,
stock-rearing, fishing, or the picking of fruits and berries, the family
group itself produces most of what it consumes.
The importance of the subsistence sector relative to that of the
monetary (or market) sector is probably one of the best possible measures
of the level of economic development in a particular region. The more
primitive an economy, the higher is the ratio of consumption of own
production to that of goods acquired in the market. This is illustrated by
the following figures of the proportions observed in different regions
during the course of recent inquiries : 72.5 per cent, of the entire consumption of the households of farmers in France 1 was purchased; in the
Senegal valley 2 —a region in the Sahelian zone with a particularly backward economy— the proportion was 42 per cent. ; and in the Bongouanou
zone it was 58 per cent. ; thus this area has reached an intermediate evell
of economic development.
1
1.N.S.E.E. : " Une enquête sur les dépenses des ménages des exploitants agricoles en 1952 ", in Bulletin mensuel de statistique (Paris, Presses universitaires de
France), nouvelle série, supplément July-Sep. 1954, pp. 45 ff.
* J. L. BounLLiER and J. CAUSSE: " Les budgets familiaux, mission socio-économique du fleuve Sénégal " (Dakar, 1958).

254

FAMILY LIVING STUDIES

Subsistence Sector.
Production for own consumption consists of foodstuffs produced on
the family farm and a variety of goods and services traditionally provided
within the group. The latter include the building of houses and fences,
the manufacture of implements for turning the soil and household
objects such as pots or calabashes, the collection of firewood for cooking
and, in some cases, certain types of economic activity carried on by
women.
The estimation of the value of the different items gives rise to thorny
problems, any solution of which inevitably involves a substantial arbitrary
element. Firewood is a significant example. In rural areas in the forested
zone, there are considerable quantities of dead wood available; to some
extent it is available to all comers rather in the same way as water.
Anybody who wants some has only to cut it up and carry it off to the
village. Wood is not a commodity traded in, and thus it has no commercial value; in these circumstances it is extremely difficult to make any
estimate of its worth. Naturally, the situation is very different in the
urbanised centres.
In the zone under study there are practically no handicrafts, and
building workers from other regions are often called in when houses have
to be built; thus all items produced for personal consumption other than
foodstuffs could be ignored. Moreover, this step was particularly
necessary as the amount of foodstuffs produced for own consumption
could easily be calculated from the findings of the food inquiry. The
only difficulty was the estimation of its money value.
Housing.
Housing is also an extremely significant pointer to levels of living.
As the Agnis in the Bongouanou area seem interested in improving their
dwellings, and as experiments with community developments have
recently been begun in the Bongouanou and other regions of the Ivory
Coast as well, this particular subject was carefully investigated. The
inquiry concerned, in particular, the types of buildings, the age of
existing buildings, the built-in surface area and cost, together with the
sociological background and the evolution of types of housing in relation
to the general development of the Agni society.
Agriculture.
It should be recalled that the investigation covered agriculture and
customary methods of cultivation, areas cultivated, yields and types of
crop in production for own consumption and production for the market.
Two features particularly struck the investigators. First, the relatively

LEVELS OF LIVING IN THE IVORY COAST

255

low occupancy of land at the present time is suggestive of particularly
promising developments of levels of living in the future. Also, as the
system of land tenure is veryflexibleand the population density relatively
low (less than 20 persons per square kilometre), land is readily available.
As it is rich forest land, suitable for the cultivation of cash crops, this
should favour economic development.
Consequently, an increase in the population may be accompanied by
an equal increase in the area under cultivation; this means, other things
being equal, that the level of living will be maintained. The amount of
land under cultivation is at present increasing even more rapidly than the
population; thus levels of living standards can steadily increase.
The Market Sector.
Income and consumption. The observation of exchanges for cash was
one of the principal objectives of the inquiry. All purchases, items of
income and gifts of various kinds were recorded for the families in the
sample at the same time that their food consumption was weighed. As
the investigators were continually " around the backyards ", family
expenditure, ranging from exceptional items, such as the payment of
dowries, to the most ordinary items, such as the purchase of a few
tomatoes, brought home by the women in bowls which they carry on
their heads, was recorded as accurately as possible.
Income and investment. The data collected in the agricultural sphere,
and particularly the measurements made of the areas offieldsused for the
cultivation of subsistence crops and cash crops, served for classifying the
households by income group. In this connection it should be mentioned
that the cadastral register which was kept at the time by the agricultural
service in the Ivory Coast was too inaccurate and incomplete to provide
adequate data.
In addition, an examination of family budgets where household
enterprises are also covered yields a wealth of information, as it reveals
the level reached in agricultural techniques and the manner in which the
planter makes combined use of the different factors of production
available to him.
Thus about 30 per cent, of a planter's cash outgoings serve to pay
wages of labourers he employs to tend the plantations he already has and
to clear virgin ground for new cultivation. In striking contrast to this
figure is the small proportion (less than 2 per cent.) he spends on equipment (mainly axes, matchets and hoes). These findings indicate that
traditional farming patterns are still being followed, and that this type of
agriculture is extensive in character.

256

FAMILY LIVING STUDIES

Employment.
The information obtained from farm households and the inquiry
among immigrants reveals the exceptional importance of wage-earning
labour in the cocoa-producing regions.
Equipment.
The relevant information was provided by the technical services of the
Ivory Coast (roads, schools, etc.).
Gifts and Services.
The daily records of food consumption and income and outgoings
showed the importance of free services and gifts of all kinds. Within a
given family line, young people are numerous. Such persons are not
fully independent in the economic sense; frequently they work as assistants to other members of the family, such as an elder brother or a stepuncle (on the mother's side). This dependence is manifested in concrete
form in the fact that a certain quantity of labour is supplied to the
economic unit, and in return some money and goods (mainly cooked
dishes) are received by the family helper from the owner of the plantation.
A related phenomenon clearly visible within each family line (the basic
unit) is solidarity; this spirit is a guarantee that the oldest or the sick
members of the family will enjoy a decent living. The head of each
family line or branch receives contributions in the form of labour from
the members of his family group, and is thus in a position to discharge
his maintenance obligations and also to provide for exceptional expenditure entailed by the reception of guests, caring for the poorest members
of the group, etc. Thus there is a kind of social security system in the
Agni society which constitutes a very novel compromise between security
and freedom for each member of the family group.
Social Level
At the social level consideration was given to a number of more
specific factors, and particularly those governing the development of
Agni society; thus polygamy, customary law, the inheritance system, and
schooling, were all analysed separately. A special study was made of the
latent tensions existing within this rural society undergoing rapid economic development. To this end all disputes submitted to the customary
tribunals were systematically analysed. One of the subjects studied was
that of the problems of inheritance under the matriUneal system; it often
happens that farms inherited by the son of the sister of the deceased are
claimed by the sons of the deceased. Mention must also be made of the

LEVELS OF LIVING IN THE IVORY COAST

257

problems arising from relations between Agnis and aliens: labour disputes between Agni employers, planters and immigrant labourers, lawsuits about land tenure between Agni landowners and foreign farmers,
conflicting rights in cases of inter-racial marriages.

SPECIAL FEATURES OF THE STUDY

The rapid increase in the cocoa and coffee-plantation areas which
has taken place during the last 30 years has raised incomes substantially
and has also brought in its wake two developments which have had
radical effects on traditional social patterns, namely large-scale immigration of persons from other regions and a more or less seasonal exodus
from the villages to numerous farming camps. These factors had to be
taken into account in the study of levels of living.
Immigrant Population
Migration into the region is primarily due to two factors. Firstly, the
increase in the labour requirements of the African planters; and secondly,
the greater specialisation of economic activity, in particular the development of handicrafts and tertiary activities such as commerce.
A few figures will suffice to show the scale of the migratory movements. Today, nearly 30 per cent, of the total population is from areas
outside the subdivision, the remaining 70 per cent, being indigenous Agnis.
Most of the migrants are between 15 and 40 years of age, and threequarters of them are adult males in the most active age groups. They
make up over two-fifths of the economically active population. Their
presence is required in every sector of the economy, particularly in cashcrop plantations. On the other hand, they receive over three-fifths of the
total income of the subdivision.
At the time the study was undertaken migration had reached such
proportions that a separate study of the living conditions of the immigrant population, to be conducted at the same time as the investigation
of the indigenous population, proved necessary. In view of the differences
of ethnic background, religion and ways of living existing between the
indigenous and the immigrant population as well as between the different
groups of immigrants (among whom representatives of over 40 ethnic
groups were to be found) separate investigations had to be conducted in a
number of fields. As, from the purely statistical standpoint, the heterogeneous character of a group under study considerably increases the
standard error, and as, moreover, there could be no question of studying

258

FAMILY LIVING STUDIES

the members of each ethnic group separately, the population was divided
into separate groups on the basis of two criteria, namely membership
of the Agni or a migrant group and the type of housing. The groups
were :
(1) Agni population
(2) Immigrant population
(a) aliens living in dyoulakros;
(b) alien labourers living in Agni farm camps ;
(c) alien planters living in isolated camps.
The sampling plans used for the Agni population and for the first two
groups of the alien population have already been described. The last
group was not covered: it contained an estimated population of only
3,000 persons, so scattered that it was impossible to work out a satisfactory sampling basis and keep within a reasonable cost limit. Practical
difficulties also arose in the establishment of a sampling plan covering
the other migrant groups. In view of the almost complete absence of
census information about them (according to the official censuses
there were 2,000, whereas in actual fact there were between 15,000
and 20,000), a detailed census of all the larger dyoulakros had to be
carried out.
Villages and Farm Camps
As the areas under cultivation increase, land becomes more difficult to
find in the neighbourhood of the villages, and planters have to go farther
away to find new forest areas for clearing. In view of the long distances
they have to cover, usually on foot but occasionally by bicycle (plantations are frequently 12 to 15 kilometres away from the villages) the Agni
planters have put up roughly built camps on the more distant plantations
so as not to have to travel from the villages to the plantation and back
every day. This tendency to live in camps has become more marked
during the past few years ; cases where planters and their families live out
in camps for several weeks at a time are not infrequent. The latter tend
to develop into proper farms, in which a few rooms are set aside for the
labourers, who live there on an even more settled basis than the planters.
Thus, there are two radically different sides to Agni life ; on the one hand,
there is the ordered life of the village, in which the basic events of social
Ufe (festivals, funerals, etc.) and economic life (markets, etc.) take place;
while, on the other hand, part of their life is centred on the farm camp
and is adapted to the pattern of agricultural work. The extreme dispersal

LEVELS OF LIVING IN THE IVORY COAST

259

of these farm camps—the sample survey indicated that there were about
2,500 of them—is the price the village pays in order to be able to remain
a coherent unit and enjoy certain benefits accruing from the general
increase in incomes in the area, such as transport facilities, schools,
dispensaries and stores.
The dual nature of Agni life as described above raised practical
problems regarding the organisation of the inquiry and theoretical
problems as well. It frequently happened that some or all of the members
of sample families in the village were absent at the time of the inquiry.
Moreover, in view of the differences between living conditions in the
village and those in the camps, the two had to be studied separately.
The most striking differences observed were in the field of diet.
When living in camp, an Agni family group is almost completely autonomous and makes practically no money purchases or sales. It only
consumes what it can grow in the fields or what it can catch by hunting
or fishing. Consequently, consumption patterns are much less varied
than in the villages, where the Agnis frequently fall back on the market,
even for food items, such as smoked fish, meat and vegetables. In view
of this absence of variety and the difficulty of checking the information
obtained, only a subsample of the families (rectangles) selected at the
second stage was used for the purpose of study of living conditions in the
associated camps.
In each survey period, one out of every two of the sample rectangles
was systematically selected for the subsample every four months. The
investigator assigned to the village in question was required to obtain—
in principle simultaneously—the standard information for a period of
seven or fourteen days on the consumption of foodstuffs (weighings) and
family budgets (sales and purchases for cash) for all persons living in the
village sample rectangle at that time, and similar information for three
consecutive days regarding the persons living in the camp or camps owned
by planters in the selected subsample rectangles, i.e., Agni members of
the families of the planters concerned and immigrant agricultural
labourers working on the plantations.
By this procedure, supplementary data on the way of life of the
Agnis during part of the year and of a substantial group of the immigrant
population (labourers living in camps) were obtained. The period for
the collection of information in each camp was reduced to three consecutive days, as compared with seven or fourteen days in the case of
similar investigations in the villages, in view of the cost of obtaining this
information (travelling, verifications) and the lack of variety in the
consumption patterns followed in the camps, to which reference was
made earlier.

260

FAMILY LIVING STUDIES

PRACTICAL ASPECTS OF THE FIELD INQUIRY

Efficacy of Multi-purpose Inquiries
Although the simultaneous consideration of the many different
aspects of levels of living requires a large survey staff and involves considerable financial expenditure, it has many advantages as regards both
methodology and efficacy. The following are illustrations of these
advantages.
In Africa, family units are extremely difficult to define exactly.
Obviously this poses a number of problems. For instance, it is difficult
to relate an item of information such as areas cultivated to a specific
number of persons. These problems are avoided by a multi-purpose
inquiry taking into account every aspect of family life, ranging from the
composition of the concession, and the groups which form for the taking
of meals, to the study of farm enterprises and household budgets.
In addition, it is recognised that information obtained in one field is
a valuable complement to that obtained in another, and certain correlations may be made such as that between expenditure on foodstuffs and
the value of the diet to illustrate the probable effect of improvements in
the levels of living of a population on food consumption patterns.
As the different questionnaires were completed simultaneously the
accuracy of the information could more easily be checked. For instance,
it is always extremely difficult—for psychological reasons—to induce a
peasant to state how much his farm brings in, but this was overcome by
getting the necessary elements (areas under crop, yield, selling price per
unit of weight), through observations and questions.
A further practical example of the efficacy of multi-purpose inquiries
can be seen in the organisation of the work of the investigators in the
field. For instance, as an investigator had to be with the family under
study almost continuously to make the necessary weighings for the
purposes of the dietary inquiry, he could be asked, when there were no
weighings to be done, to obtain other data, including information on
family budgets and housing. He might go along with the family, when
it went off for agricultural work, and measure the fields. This system has
two advantages. First of all, as the investigator and the respondents are
practically living together, friendly relations develop, thus facilitating
the task. Secondly, the cost of obtaining each item of information is
substantially reduced.
Choice of Basic Unit of Investigation
As was stated earlier, the choice of the basic unit of investigation to
be studied was extremely difficult to make. By the end of the first week

LEVELS OF LIVING IN THE IVORY COAST

261

of the training course it had become clear that earlier conceptions of the
family group would have to be radically modified for the purposes of the
study of consumption. In particular, the conventional concept of
" household ", in which meals are shared by the husband, his wife or
wives and his children, proved inapplicable to the Bongouanou region.
First of all, very few kitchens are operated in isolation. It is a common
practice for meals to be prepared in common by the womenfolk of
different households on the concession in four or five different kitchens.
The meals are shared by groups of persons who are generally related to
one another; but there are nearly always some who do not belong to
the household. There are, however, no hard-and-fast rules governing the
composition of the group. It was thus impossible, by questioning the
head of a concession, to find out the exact composition of his concession
and which kitchen provided meals for which group. As, moreover, in
A gni societies the men take their meals separately from the women and
children, " basic consumption groups ", that is to say, groups of kitchens
catering for the same people, had to be established. At this point a new
difficulty arose, namely that of determining these elementary consumption
groups in advance in order to be able to draw up a detailed list of them
and to select a satisfactory sample. As this problem was approached
it gave place to another, arising from the fact that the consumption
groups turned out to be extremely unstable in composition. For instance,
in one unit there were three kitchens—A, B and C—catering for the
same group of persons. In the middle of the week, kitchen A and a
number of members of that group vanished completely; anew kitchen,
D, appeared from nowhere and a new group attached itself to the
old one.
The nature of fife in agriculture makes it necessary for the Agnis to
be continually coming and going between the farm camps and the
villages. Thus the problem was one of finding a suitable unit which
could be studied at both ends—that is to say in the village and in the
camps. The solution was finally found by means of a more detailed
investigation of housing patterns. The Agni village is made up of a
number of isolated units consisting of a rectangular courtyard, surrounded by houses. There is a striking parallel here with, for instance,
Arab-type houses, which are built around a completely enclosed courtyard where all of family fife is carried on. However, each courtyard
contains not one house, but several side by side, with a row of kitchens
which also face on to the courtyard. Thus, considered from the housing
standpoint, the rectangle should not be confused either with the administrative unit of the " cour ", with aboussoua (the Agni family in the
broad sense), or with the family line. It is usually a spontaneously

262

FAMILY LIVING STUDIES

formed community of close or distant relatives, without any particular
hierarchy, held together by neighbourly ties (living in close proximity).
Choice of Period of Investigation
The period of investigation must be determined in the light of a
number of factors. In view of the heavy burden which the continuous
presence of the investigator weighing produce and food and asking
innumerable questions involves for the persons being studied, the period
of investigation should not be too long otherwise they will become tired
of it. On the other hand, the period should not be too short, as the
accuracy of the observations may thereby be adversely affected. The
investigator has to be well acquainted with the family he is studying and
know something about all its members and their ways of living. In
certain types of inquiries the optimum period appeared to be four days ;
however, in order to allow for the possibility of there being a weekly
consumption pattern (in Agni society there were two rest days : Wednesday and Sunday, the one being the traditional animisi rest day, and the
other the Christian rest day, as family celebrations take place on those
days, the food consumed is likely to be different from that consumed on
other days), the investigation period selected was the week, or in certain
cases, the fortnight.1
It should be noted in this connection that the Bongouanou experiment
revealed that the respondents accepted a single protracted period of
investigation more willingly than a number of separate shorter ones at
different times of the year. When the sample was reviewed at the beginning of each period a number of difficulties arose with families who were
unwilling to be studied for a second or third time. This was a manifestation of a kind of egalitarian prejudice existing among Agni peasants.
They found it difficult to understand why some families should be studied
only once while others were studied two or three times during the year.
Another manifestation of the same attitude was the fact that in several
villages the inquiry had to be started by obtaining information from the
village chief, even if his concession did not form part of the sample (in
such cases the information obtained was not used) ; the reason was that,
on the one hand, the chief found it difficult to understand why his
concession should be ignored in the inquiry, while on the other hand the
villages found it difficult to understand why the chief was not setting
a good example by being the first to accept the investigators in his
home.
1
For one-third of the sample, that is to say, the rectangles studied during the
three periods of the year.

LEVELS OF LIVING IN THE IVORY COAST

263

Problems of Training Indigenous Investigators
One observation must first be made which is very encouraging for
inquiries of this type : very considerable difficulties were expected, but
in fact there were only minor disappointments. The cutting down of the
theoretical side of training and concentration on field exercises during the
training period enabled the investigators to acquire the techniques of
collecting information—admittedly very rudimentary ones—rapidly.
When investigators were trained for the agricultural inquiry conducted
in Bouaké (Ivory Coast), they were nearly all able, after four days'
training, to measure a field accurately with chain and compass ; similarly,
at Bongouanou after four days they had learnt how to use Roberval-type
scales, and steelyards, and could carry out investigations of food consumption in a satisfactory way. The training had to last four weeks,
owing to the large number of items of information required and questionnaires used and to the continual changes that were made in the
inquiry plan in the light of the field exercises carried out.
One delicate problem was the replacement of investigators who
proved unsatisfactory, or fell sick. In fact, it was settled in an extremely
simple fashion. Each new recruit was sent out on a practical field
exercise accompanied by an experienced investigator; the questionnaires
were completed in duplicate under the control of the latter, and, naturally,
under the close supervision of the directors of the inquiry, who gave
appropriate instructions accordingly. It was observed that training of this
type rapidly produced fully satisfactory investigators.
Problems of Settling Investigators in the Villages
The efficacy of the inquiry was due to a considerable extent to the
friendly relations between the investigators and the families, which could
not have developed if the investigators had not remained continuously in
the villages for a long period—in fact, for a full year. This was, in fact,
the end in view when a sampling plan was selected involving, for the
second and third survey periods, replacements of some sample families
in the same sample villages rather than replacement of sample villages.
Various instructions were given to the investigators to make their
acceptance in the villages easier. Any disputes their presence in the
villages gave rise to were to be referred to the village chief, who could
deal with them according to custom. They were required to provide their
own board and lodging so as not to become a charge on the villages;
in order not to give rise to any change in the diet of the persons under
study during the inquiry period, they were forbidden to take any meals

264

FAMILY LIVING STUDIES

with the families actually investigated. It should be added that the
innate hospitality of the inhabitants of the area and the appreciation
shown by the Agnis at the interest being taken in their living conditions
and their problems greatly facilitated the task of the investigators.

ANALYSIS AND PRESENTATION OF THE RESULTS

The unusual characteristics of the area and the lack of previous
knowledge thereof, gave rise to difficulties, not only in the preparation of
the general plan, but also when the time came to draw up and present the
findings. Some examples may be given of the great benefits of co-operation between the field investigators in the different areas and between
the investigators and the technical services using the results of these
inquiries.
Analysis of Diet
One example can be found in the analysis of the diet. The only food
composition tables available were those published by the F.A.O.
Although comprehensive, they had to be adapted in respect of a number
of items in the light of local conditions. First of all, some of the items
consumed are not mentioned in the tables at all; alternatively, the tables
are not sufficiently specific on the different variations of a given item.
Palm oil and palm seeds, which are widely consumed throughout the
whole of the forest zone of equatorial and tropical Africa, are examples.
There were also a number of types of game and fruits and vegetables
not specifically mentioned in the F.A.O. tables, and equivalents had to
be found for them.
In addition, factors such as the water content of particular foodstuffs
may vary considerably from one region to another on account of climatic
conditions.1 Moreover, some of the coefficients given in the F.A.O. tables
may be modified by local factors. Among these the habits of housewives
are particularly important, for on them depends the amount of waste. It
would have been a particularly serious mistake not to allow for these
factors in the Bongouanou region, for the foodstuffs involved include
tubers and plantains, from which the population under consideration
obtained over 80 per cent, of the calories it consumed. The water
content of these vegetables was found to be 63 per cent, as compared
with the 72 per cent, given in the F.A.O. table; use of the latter would
1
Analyses were carried out by the O.R.A.N.A. laboratory in Dakar, to determine the water content of certain items of produce and the composition of certain
foodstuffs such as palm wine and various vegetables.

LEVELS OF LIVING IN THE IVORY COAST

265

have resulted in underestimation equal to about 34 per cent. Lastly,
the amount of waste, which was estimated from systematic observations
covering over 30 tons of yams, was substantially different from that given
in the tables. Moreover, as these coefficients vary not only from season
to season but also according to the variety of vegetables, it would be
wrong merely to apply coefficients calculated for a particular season or
area. The use of separate coefficients calculated for each season may thus
be necessary unless one is prepared to use an average coefficient, which
will not show seasonal variations even though it may be valid for the
year taken as a whole.
As it is both impracticable andfinanciallyimpossible to carry out a
large number of analyses to ascertain the different coefficients of composition of foodstuffs with sufficient accuracy, the fact cannot be overstressed that when the results are worked out errors of similar size to,
or even greater than, the sampling errors may easily be made.
The calculation of theoretical requirements is used as a reference
standard for the estimation of dietary insufficiencies, and, where they
exist, surpluses of individual items. This concept of a dietary standard
necessary to maintain satisfactory nutritional standards* is still far from
well defined as far as groups living in tropical and equatorial Africa are
concerned. Work has been done in various countries enabling the calorie
requirements of a standard reference individual living in the temperate
zone, and defined according to a number of criteria such as sex, weight
and activity, to be calculated ; on the other hand, the theoretical requirements of persons in the tropics are calculated solely in accordance with
hypotheses on the basis of which certain corrective coefficients are
applied to the standards valid for persons in temperate areas. The
coefficients adopted are based on F.A.O. recommendations 2, and also,
to a certain extent, on experiments carried out in west Africa by the
O.R.A.N.A. Factors of correction for climatic conditions and levels of
activity were easy to calculate; on the other hand, to work out a factor
for body sizes adult men and women living in the sample villages had
to be weighed. Here again, the extreme importance of making allowances
for this factor, which is a function of certain well-defined ethnic characteristics, must be emphasised.
The average weight of Agni men proved to be 55 kilogrammes and
that of women 48; the corresponding weights for the reference individuals
were 65 and 55 kilogrammes respectively. The theoretical calorie
1

J. TREMOLIERES, Y. SERVILLE and F. VINIT: " Etude sur la ration alimentaire

type à préconiser pour le Français ", in Bulletin de l'Institut national d'hygiène (Paris),
tome 7, Oct.-Dec. 1952, pp. 767-825.
2
F.A.O.: Calorie Requirements, F.A.O. Nutritional Studies, No. 5 (Rome, 1950).

266

FAMILY LIVING STUDIES

requirements of a sedentary adult would thus be 2,050 calories instead
of 2,300.
It is unfortunately very difficult to go into greater detail and compare
theoretical requirements and the diet of certain population groups that
are likely to be worse off than the population as a whole. For instance,
it is to be suspected that, even in regions in which the composition of
the diet is satisfactory, pregnant women, nursing mothers and children
will not necessarily be receiving a diet corresponding to their requirements.
As the Agnis are accustomed to living in communities, the artificial
isolation of members of particular categories might often influence the
behaviour of consumers. The only attempt made in the Bongouanou to
investigate behaviour of this kind was a qualitative investigation of the
food consumption of young children under two years of age—that being
the age at which they are weaned.
Analysis of Family Budgets
The presentation of the findings on family budgets also gave rise to
difficulties peculiar to the particularly complex rural groups in this part
of Africa. The main difficulties occurred during the analysis of the
information, when the problem arose of allocating the expenditure and
receipts of the different persons living in the rectangle among clearly
defined economic units. There is no difficulty about combining in a
single household budget all transactions carried out by a planter, his
wife or wives and his children; however, the situation becomes more
confused when transactions carried out by an isolated individual member
of a collateral branch, who is either too young or too old to be the head
of a household or even to count as a separate household (for instance, a
young brother who is still working as helper in the family enterprise) has
to be included as well. For instance, the expenditure of a widow who
did not have full economic dependence was included in the budget of
the household head who had inherited her husband's estate.
Every case was dealt with according to information collected on the
spot ; but in certain particularly complex situations where the budgets of
several households overlapped with one another, certain arbitrary
measures had to be taken to distribute the expenditure and receipts and
to enable thefindingsto be presented in a comprehensible manner. These
difficulties did not, however, affect the structures of family budgets; at
the most they proved that some results had been presented in a way that
did not correspond exactly to the realities of African life. For instance,
there is a certain solidarity among the households belonging to the same

LEVELS OF LIVING IN THE IVORY COAST

267

matrilineal line, and consequently, as a result of the customary system of
rights and obligations, their budgets may overlap ; however, the presentday trend is for each household to be an independent economic unit, its
income and expenditure consisting entirely of the income and expenditure
of its own members.
Problems of Classification
Other problems in the presentation of the results arose from difficulties of nomenclature. When making comparisons between the findings of inquiries in different regions or, for example, when using such
information for national accounts, the great variety of forms of presentation becomes apparent. Moreover, the classifications used often differ;
some of them are extremely detailed, while others are extremely condensed, and it is rarely possible to convert the data from one classification
to another by regrouping the figures or breaking them down.
For example, in some inquiries expenditure on drink is considered
separately, while in others it is combined with expenditure on foodstuffs.
Similarly, a breakdown of consumption into locally produced and/or
imported items—required for the calculation of the propensity to
import, which is an extremely important factor in development problems—is not always carried out. Admittedly, consumption patterns
differ very considerably from country to country, particularly among the
countries of tropical Africa; however, some standardisation of the
results would be desirable.

CONCLUSIONS

While it is true that some levels of riving indicators, such as the
expectation of life at birth and the proportion of children attending
school, are of indubitable interest in tropical and equatorial Africa,
elsewhere there are other indicators, the importance of which varies
from region to region.
It has been stressed several times in the course of this article that
special attention should be given to production for own domestic consumption, which makes up as much as 90 per cent, of total consumption
in certain parts of this area. Moreover, in so far as all countries are
engaged in what one might call an " economic development race ", it is
essential to provide data, i.e., indicators, which reflect, in these economic
characteristics on which future development depends, not only the
present level but also the rate of change. Lastly, as economic develop-

268

FAMILY LIVING STUDIES

ment goes hand in hand with social and cultural developments, which
give rise to radical changes in institutions and patterns of behaviour, the
establishment of indicators for this field should be an objective in any
study of levels of living, especially in tropical and equatorial Africa.

XV
Repeated Surveys
of Rural Living Conditions in India *
The National Sample Survey of India (NSS) was initiated in 1950, and
has been in operation continuously since that time. The general objective
of the plan was to create a repetitive survey programme through which
a wide variety of essential statistical information could be obtained,
which would provide national and regional estimates of demographic
characteristics, income, consumption, agricultural production, etc.
The use of sampling methods had been under consideration for some
time before the survey programme was adopted. As early as 1934,
Professors Bowley and Robertson had recommended the use of a village
sample for a proposed rural survey, which, however, was not implemented.1 After 10 years of operation, the National Sample Survey is one
of the unique statistical operations in the world—a project on a vast
scale covering a population largely rural and scattered over a territory
of 1.27 million square miles.
GENERAL DESCRIPTION OF THE

1950-51 SURVEY

The Sample
The original plan was to assemble detailed maps for the entire
country and to select sample units (of one or more villages each) directly
from the maps. It was soon discovered, however, that suitable maps
were available for only part of the country.
* Professor P. C. Mahalanobis, Director of the Indian Statistical Institute, had
originally agreed to write this chapter. Because of unforeseen circumstances, however,
he has found it impossible to do so. In order that readers of the Symposium may have
a general view of this important survey project, the I.L.O. Statistical Division has
prepared the present summary. This summary has drawn extensively on The National
Sample Survey, General Report No. J on The First Round (Department of Economic
Affairs, Ministry of Finance, Government of India, Dec. 1952), prepared by the
Indian Statistical Institute.
1
A. L. BOWLEY and D. H. ROBERTSON: A Scheme for an Economic Census of India
(New Delhi, Government of India Press, 1934).

270

FAMILY LIVING STUDIES

Attention was then given both to the collection of lists of villages
and to the geographical area and population figures of individual villages.
Even this was not an easy matter, and after references had been made to
all the states it was found that in many of them both population and area
figures of individual villages were not available. Either population or
area figures were available for 78.6 per cent, of the total area of India.
For 15.8 per cent, of the area neither the population nor the area figures
were available, but only the names of the villages could be secured.
For the remaining 5.6 per cent, of the area even the village names were
not available, and lists of villages had to be prepared by the field staff at
a later stage. The situation is shown in summary form for the whole of
India in Table I.
TABLE I. TYPE OF INFORMATION AVAILABLE BY
AREA AND POPULATION
Type of information available for individual localities

Population (1941 census) figures but no area figures
Area figures but no population figures
Neither population nor area figures but the latter can be
measured from thana or tehsil maps
Neither population nor area figures nor maps; only
village names available
Not even village names nor thana or tehsil maps available

Percentage to total
of India
Area

Population

56.7
14.2

58.6
21.7

7.7

4.9

15.8
5.6

12.2
2.6

Source: The National Sample Survey, General Report No. J, p. 47.

As the material available varied according to region it became
necessary to adopt different types of " frames " and methods of selecting
the sample units in the différents parts of the country. The probability
of being included in the sample thus differed from region to region,
which made the design of the survey somewhat complicated and less
efficient in the first round. Varying sampling fractions also complicated
the calculations (this was set right in the fourth round). One important
reason for using varying sampling fractions was the desire on the part of
the field staff to keep the work load of each investigator in each village
the same.
The sample units were selected in two stages. First, the villages were
selected after suitable stratification. Within each sample village all or a
subsample of 80 households, whichever was less, were stratified into
agricultural and non-agricultural classes. Sample households were then
selected at random from each of these strata. The data were collected
in a number of phases. A larger sample of households was interviewed

271

RURAL LIVING CONDITIONS IN INDIA

for general economic information and smaller subsamples were then
interviewed for more intensive investigation.
Formation of Strata.
All the states in India (except Jammu and Kashmir, and Sikkim)
were divided into 160 strata and each stratum was formed on the basis
of geographical contiguity and topographical homogeneity. In those
strata for which population figures of individual villages were available,
four substrata were formed on the basis of the size of the population of
the villages. The four classes were villages with populations of:
(1)
1 to 499
(3) 1,000 to 1,999
(2) 500 to 999
(4) 2,000 and above
Such substrata, however, could be formed in only 32 (out of 160) strata,
and these 32 substrata were then divided into 128 ultimate strata. There
were thus 256 strata altogether on the basis of which the sample villages
were selected.
Random Selection of Sample Villages.
The procedure of sample selection varied depending on the nature of
information available for the villages; and five different procedures for
selection were followed, as shown in Table II below:
TABLE II. COVERAGE OF DIFFERENT METHODS OF
SELECTION OF SAMPLE VILLAGES
Method of selection

No. of
strata

No. of
sample
villages

Probability of selection equal
. . . .
Probability of selection proportional to
area in strata without substratification
Probability of selection proportional to
population in strata without substratification
Probability of selection proportional to
population in strata with substrati-

398

230,764

20.34

475

247,794

21.84

120

103,738

9.15

Area in
sq. miles

Percentage
of total
area

330

242,793

21.40

Probability of selection proportional to
area in strata with substratification

510

309,248

27.27

Total . . .

256

1,833

1,134,337

100.00

Allocation of Sample Villages.
The allocation of sample villages in each stratum was in proportion
to the rural population of the stratum. Altogether 1,833 sample villages

272

FAMILY LIVING STUDIES

were selected for the whole of India. In each stratum the number of
sample villages was made a multiple of three in order that the allocation
of the samples between the two types of questionnaires used in this
round could be made in the ratio of 2: 1 approximately.
The villages spread over all the strata were thus divided into two
groups, the first main group consisting of 1,189 villages, in which a set of
schedules prepared in the Indian Statistical Institute was employed for
the data collection; and the second group of 644 villages, in which
another set of schedules prepared by the Gokhale Institute of Politics
and Economics was used.
The Survey Plan
The first round of the National Sample Survey was started in October
1950 and completed in March 1951.
For the field work, India was divided into 16 parts, each called a
block. A block consisted either of a single large state like Uttar Pradesh
or of a number of smaller states grouped together, as for example,
Assam, Tripura and Manipur. The 1,833 villages were then assigned
to different blocks in proportion to their 1941 populations. Each of
these blocks was further subdivided into smaller areas (not necessarily
of equal size) in such a way that a multiple of three villages (usually
six villages) could be assigned to different areas, keeping the population
proportion the same.
A list was made by the investigators of all the households living in a
selected village, and sample households were drawn at random after
suitable occupational stratification. The required information was then
collected through interviews of a member or members of a sample
household. Since the questions asked referred to personal habits,
economic activities, incomes and expenditures, the interviewers had to
do their work with tact and patience. Books containing detailed instructions on how to approach the respondents and what was meant by
the various questions in the schedules were supplied to each investigator.
The unwillingness of the households to furnish information was negligible, but there was no way to ensure that all the particulars given by the
households were correct, especially when the investigators were at the
job for the first time.
There were naturally many difficulties in carrying out a survey
covering the whole rural area of India. Although the information could
be recorded in English, the interview with the householder had to be
conducted in the vernacular language of the informant. This meant
that only those investigators could be employed who spoke the vernacular
language of the locality in which they were to work. In actual fact, more

RURAL LIVING CONDITIONS IN INDIA

273

than 15 local languages had to be used in collecting the information and
great care was taken that the terms used in the different languages
conveyed exactly the same meaning. Another serious difficulty was the
occurrence of local systems of weights and measures. In the first round
more than 140 local systems of measurements of weight were discovered
and the quantities had to be converted into standard weights and
measures.
Nature of Information Collected in the First Round
The schedules (prepared by the Indian Statistical Institute) that were
used in 1,189 villages may be grouped into four types, namely (1) village
schedules, (2) household schedules for general particulars on demographic
and economic conditions, (3) household schedules for detailed information on household enterprises, and (4) household schedules for detailed
information on consumer expenditure. A brief description of the
schedules is given below.
Village Schedules.
Schedule 0 : for listing of all households of the sample villages.
Schedule 0.1: for collecting information regarding the occupation
of 80 households selected by a random process from the complete Hst of
households of a village.
Schedules 4 and 5: for collecting information on the utilisation of
lands on a sample basis and in regard to yield of crops.
Schedule 6 : for collecting prices of certain selected commodities in
sample villages such as cereals, pulses, oil, milk, vegetables, spices, fuel,
tobacco, etc., and rates of daily wages of various types of skilled and
unskilled (male and female) workers, particularly in relation to agricultural operations.
Household Schedules for General Particulars.
Schedule 1 : for collecting information on the size of the family,
age, sex, marital status, economic and employment status, and occupational particulars of individual members of the sample households;
information on the holding and use of land under various categories;
particulars on livestock, real assets, loans and savings, and housing
conditions.
Household Schedules for Detailed Information on Household Enterprises.
Schedule 2.1 : for collecting information on agriculture and animal
husbandry, particulars such as acreage and production of different crops ;
expenditure on materials and operations in agriculture and animal

274

FAMILY LIVING STUDIES

husbandry; source offinance,nature of disposal of produce and the value
of livestock products; and an account of livestock, agricultural implements and accessories.
Schedule 2.2 : for particulars on industry, crafts and trade, including
fixed capital, machinery and tools, fuel, power, raw materials, quantity
and value of production, type of labour used, source of finance and
income from industrial servicings.
Schedule 2.3 : for particulars on incomes and expenditures in services,
profession, and financial operations of individual earners in sample
households.
Household Schedules for Detailed Information on Consumer Expenditure.
Schedule 3.1 : for particulars on consumption of food under various
groups of items, beverages, intoxicants, fuel and light, in value and
wherever possible in quantity.
Schedule 3.2 : for particulars on expenditure on clothes with breakdowns into types ; household articles of various categories ; medicines,
education, amusements, ceremonials, and services of various kinds.
The period to which most of the questions in all these schedules
referred was one year, from July 1949 to June 1950.
Coverage of Units
It should be noted that the information on consumption and enterprises in rural areas was collected by interviewing the households. In
consequence, the information collected in the survey was confined to
information which could be reached through the households living in
villages.
Proprietary enterprises and those under partnership were covered
but non-household enterprises were excluded from this survey. Consequently, all establishments were excluded which were operated by
(1) joint stock companies; (2) government agencies, central, state or
local; (3) other institutions which are distinct from households such as,
co-operatives, or research and educational institutions not operated by
households.
All mines, factories, plantations, shops, banks, insurance companies,
wholesale and retail trading companies, transport operations, warehouses and godowns, etc., run by agencies of the above three categories
were excluded. Particular care should, therefore, be taken to note that
data collected through this survey will give the results for only a part of
the total economic operations in rural areas. For instance, the value
added in production calculated on the basis of the figures collected in

RURAL LIVING CONDITIONS IN INDIA

275

the survey will cover only that part relating to household enterprises
and will exclude the part relating to non-household enterprises.
It should also be noted that the information was collected only from
households residing in the villages during the time of the survey. If, for
instance, the members of a household living in a village had enterprises
of some kind in urban areas, the enterprise particulars in regard to those
urban establishments were covered in the survey. On the other hand, a
household enterprise operating in rural areas but owned by a household
living in a town was not included in this survey.
Lastly, in regard to particulars on services (as collected in schedule 2.3)
the survey was not confined to household enterprises only but included
employment in any establishment. Since the sample households were
selected from those living in villages, the particulars collected were
naturally limited to those employees or professional workers who were
living in villages.

DEVELOPMENTS SINCE THE FIRST ROUND

Many developments have occurred since the first round of the survey.
The second round, which again covered only the rural areas, was started
in April and completed by the end of July 1951. The number of sample
villages was 1,160 and only one single set of schedules was used to collect
information relating to the consumption and crop production for ten
sample households selected at random from each sample village. One
important change was the reduction of the reference period from one
year to one week or one month for many articles of consumption.
In the third round, which was started in August and completed in
November 1951, the most important change was the extension of the
survey to cover the four big cities of Calcutta, Bombay, Madras and
Delhi, and 50 towns scattered over the whole country. The total number
of sample villages was 920, and 12 sample households were selected at
random from each sample village; and the number of households in the
urban area varied between 40 and 336 depending on the population of
the town or city. The schedules were extensively revised on the basis of
the experience gained in the first round of the survey.
A rapid survey of the winter and spring crops grown in 920 sample
villages was made in December 1951. In each village, 20 sample clusters
each offiveplots were selected with probability proportional to the total
area of the cluster; and the field investigators examined each cluster
separately and made a record of the crops grown on it. This survey
supplies useful information not only about agricultural production

276

FAMILY LIVING STUDIES

but also about developing suitable methods for direct crop surveys
on an all-India scale.
The fourth round of the economic survey started in April 1952 and
was completed in October 1952. It covered both rural and urban areas.
The design for urban areas remained broadly the same as in the third
round but the design for the rural areas was completely changed.
The whole country was divided into 240 strata, and two sample tehsils
were selected in each stratum, and two sample villages in each tehsil
giving a total sample of 960 villages selected in a suitable random manner.
An appropriate number of sample households was selected at random
in each village in such a way that the over-all sampling fraction was
equal.
An attempt was made in this round to make a precise comparison of
results based on two different periods of reference, namely, one week and
one month, for the consumption of foodgrains and other important
commodities. In both rural and urban areas, half the households were
investigated with the week as the period of reference and the other half
with the month as the period of reference, and two investigators worked
simultaneously with the two different forms. In this way it was hoped
to secure critical data relating to the period of reference.
In rounds one through seven, special emphasis was placed on expenditure data, and for this purpose a reference period of one year was used.
From the second round onwards a shorter reference period—a week or
a month—was adopted for the most frequently purchased items. In the
fourth and fifth rounds a systematic check of the effect of the length of
the reference period on the information collected from households was
made by adopting a week as reference period for half the household
sample and a month for the other half in the recording of food, fuel and
light, tobacco, drugs and intoxicants. In the sixth round a week was
used as the reference period for these items. Information on services,
education, medical expenses and certain other items was obtained with
reference to the month preceding the date of inquiry in all rounds and
data on clothing, other durable and semi-durable articles, rents and taxes
for a full year up to the sixth round. In the seventh round the reference
period for all items was one month.
Wherever possible the information on consumer expenditures has
been supplemented by particulars on quantities of various items consumed
and on cash purchases as distinct from values of receipts and disposals of
food grown by the households in the sample.
Emphasis was shifted in the seventh and following rounds to household enterprise and to employment and unemployment characteristics
of the individuals forming the households under survey. Data of this

RURAL LIVING CONDITIONS IN INDIA

277

nature had been obtained from the beginning of the NSS but the sample
was considered too small to permit valid conclusions. In the seventh
round separate lists of households were prepared for different occupations
and larger samples were selected for special interviews concerning
(a) household transport operations, (b) small-scale manufacture, and
(c) household retail trade. In the eighth round an extensive survey was
organised of land holdings in the rural sector through the use of a special
household schedule in addition to the current inquiries concerning
household consumer expenditure, production enterprises, land utilisation
and prices.
Concurrently with these activities a programme was launched
relating to the survey of employment and unemployment conditions.
This was initiated by several special surveys conducted in the interval
between the sixth and seventh rounds, including a study of the employment characteristics of a sample of persons in the " live register " of the
New Delhi Employment Exchange, a Sample Survey of Employment
in Calcutta, and a Preliminary Survey of Urban Unemployment in
23 towns. Materials from the fourth to the seventh rounds on selected
characteristics of the economically active population were analysed in a
special report and a comprehensive survey of employment, unemployment and disposition of labour time was included in the ninth and tenth
rounds of the NSS.
All through the earlier rounds of the NSS general demographic
characteristics of the households as age at marriage, number of children
born, birth spacing, infant deaths and sex ratios were recorded through
the household schedule. Materials derived from the second and the
fourth rounds were analysed in a report on couple fertility. In later
rounds more attention has been devoted to the possibility of obtaining
estimates of birth and death rates through the sample survey.

PUBLICATION OF RESULTS

The results of the National Sample Survey are published in a series
of reports issued by the Ministry of Finance, and, since 1958, the Cabinet
Secretariat, Government of India. Some of the reports have been
reprinted in Sankhya, the Indian Journal of Statistics (Statistical Publishing Society, Calcutta), and special survey reports and data in tabular
form are supplied from time to time to different agencies within the
Government of India.
Data on consumer expenditures given by the first round were
analysed in General Report No. 1 on the First Round (October 1950-

278

FAMILY LIVING STUDIES

March 1951), issued by the Ministry of Finance in December 1952.
Corresponding data collected by the Gokhale Institute' of Politics and
Economics were published separately. Tables with notes on the second
and third rounds were published in December 1953 and January 1954
(Nos. 2 and 3). Reports on consumer expenditure on the basis of the
information collected in the fourth round and a comparative study of the
findings in the second to the seventh rounds (covering the period April
1951-March 1954) were issued in 1959 (Nos. 18 and 20).
Data on some demographic characteristics of the population according to the second and fourth rounds were published in December 1955
in the Report on Couple Fertility (No. 7).
A series of reports on the results of the seventh to tenth rounds,
including the First Report on Land Holdings (No. 10), the Report on
Employment and Unemployment (No. 16) and the Preliminary Report on
Housing Conditions (No. 26) have been issued since 1958.

Selected References
UNITED NATIONS : Enquiries into Household Standards of Living in Less-Developed
Areas, ST/SOA/1, Sales No. 1950.IV.7 (New York, 1950).
— Report on International Definition and Measurement of Standards and Levels
of Living, E/CN.3/179, Sales No. 1954.IV.5 (New York, 1954).
I.L.O.: "Chief Characteristics of Recent Family Living Studies" (Geneva,
1955), Preliminary edition, roneoed.
— "Bulletin on Family Budget Surveys, 1950-1960" (Geneva, 1961), roneoed.
F.A.O. : Dietary Surveys : Their Techniques and Interpretation,
Studies, No. 4 (Washington, D.C., 1949).

Nutritional

— Fact-Finding with Rural People, A Guide to Effective Social Survey, prepared
by HSIN-PAO YANG, Agricultural Development Paper No. 52 (Rome,
1955; 2nd edition, 1957; 3rd edition, 1960).
— Review of Food Consumption Surveys (Rome, 1958).
W.H.O.: Measurement of Levels of Health, Technical Reports Series, No. 137
(Geneva, 1957).
COCHRAN, W. G.: Sampling Techniques (New York, John Wiley and Sons,
1953).
DALENIUS, T. : Sampling in Sweden, Contributions to the Methods and Theories
of Sample Survey Practice (Stockholm, Aimqvist, 1957).
DEMTNG, W. E. : Some Theory of Sampling (New York, John Wiley and Sons,
1950).
FESTINGER, Leon and KATZ, Daniel, eds. : Research Methods in the Behavioral
Sciences (London, Staples Press, 1954).
GOODE, W. J. and HATT, P. K.: Methods in Social Research (New York,
' McGraw-Hill, 1952).
HANSEN, M. H., HURWITZ, W. N . and MADOW, W. G. : Sample

Survey

Methods and Theory : Vol. I, Methods and Applications ; Vol. II, Theory
(New York, John Wiley and Sons, 1953).
HYMAN, H. H. : Survey Design and Analysis : Principles, Cases and Procedures
(Illinois, Free Press, 1955).
JAHODA, Marie and others: Research Methods in Social Relations: Vol. I,
Basic Processes ; Vol. II, Selected Techniques (New York, The Dryden
Press, 1951).
KELLERER, Hans: Theorie und Technik des Stichprobenverfahrens,
Statistische Gesellschaft (Munich, 1953).

Deutsche

MADGE, J.: The Tools of Social Science (London, Longmans Green, 1953).
MOSER, C. A. : Survey Methods in Social Investigation (London, Heinemann,
1958).
PARTEN, M. B. : Surveys, Polls and Samples : Practical Procedures (New
York, Harper and Bros., 1950).

280

FAMILY LIVING STUDIES

PRAIS, S. J. and HOUTHAKKER, H . S. : The Analysis of Family Budgets (Cam-

bridge, Cambridge University Press, 1955).
SUKHATME, P. V.: Sampling Theory of Surveys, with Applications
Iowa State College Press, 1954).

(Ames,

THIONET, Pierre : Application des méthodes de sondage aux enquêtes statistiques
(Paris, Presses Universitaires de France, 1953).
YATES, Frank: Sampling Methods for Censuses and Surveys (London, Griffin
and Co., 1949).