INTERNATIONAL LABOUR OFFICE
EMPLOYMENT, UNEMPLOYMENT
AND LABOUR FORCE STATISTICS
A STUDY OF METHODS
Report prepared for the Sixth International Conference
of Labour Statisticians
(Montreal, 4-12 August 1947)
GENEVA
1948
STUDIES AND REPORTS
New Series, No. 7
(Part 1)
P U B L I S H E D B Y T H E INTERNATIONAL L A B O U R O F F I C E
Geneva, Switzerland
Published in the United Kingdom for the INTERNATIONAL LABOUR
by Staples Press Limited, London
OFFICE
Distributed in the United States by the INTERNATIONAL LABOUR OFFICE,.
Washington Branch, 1825 Jefferson Place, Washington, 6, D.C.
P R I N T E D B Y A T A R , GENEVA
CONTENTS
Page
PREFACE
INTRODUCTION
v
1
PART I
CHAPTER
I : Objectives
6
CHAPTER I I : Definitions
9
I.
II.
III.
IV.
V.
The Employed
The Unemployed
Labour Force
Time Reference
Other Definitions
9
12
14
14
18
CHAPTER I I I : Reporting Systems
I. Employment
II. Unemployment
19
19
23
PART II
CHAPTER IV : Types of Employment Series : National Practice
.
I. Total Employment
II. Employment by Branches of Economic Activity . . .
III. Employment by Geographical Area
IV. Employment by Sex, Age and Hours Worked
28
28
29
31
32
CHAPTER V : Types of Employment Series : General Problems . .
I. Agricultural Employment
II. Non-Agricultural Employment
III. Employment by Industry
IV. Total Employment
V. Labour Force
VI. Employment by Geographic Region
VII. Employment by Age and Sex
V I I I . Employment by Occupations and Hours "Worked . . .
IX. Partial Employment
X. Seasonally Adjusted Series
32
33
34
35
35
36
37
40
41
42
42
CHAPTER V I : Employment : Coverage
I. Status
II. Condition
III. Age
43
43
47
51
CHAPTER V I I : Employment: Sample
I. Establishment Sample Statistics
II. Labour Force Surveys
51
52
61
IV
CONTENTS
Page
CHAPTER V I I I : Establishment List
I. National Lists
II. Non-Reporting
III. Comparisons based on Identical Firms
67
67
68
70
CHAPTER I X : Trend Bias
71
I. National Experiences
II. Adjustment for Bias
71
73
CHAPTER X : Collection of Data
76
PART III
CHAPTER X I : Types of Unemployment
I. National Practice
II. General Problems
CHAPTER X I I :
I.
II.
III.
IV.
Unemployment
Series
78
79
80
: Sample
82
Registration at Employment Exchanges
Trade Union Statistics
Social Insurance Statistics
Labour Force Surveys
.
CHAPTER X I I I : Unemployment : Coverage
I. Employment Exchange Statistics
II. Trade Union Statistics
III. Social Insurance Systems
IV. Labour Force Survey Statistics
82
85
88
95
95
96
96
96
100
P A R T IV
C H A P T E R X I V : Consistency
101
CHAPTER X V : International Comparability
I. Purposes
II. Industrial Classification
III. Occupational Classification
of Data
C H A P T E R X V I : Publication
106
106
108
109
110
I. Promptness of Publication
II. Presentation
CHAPTER X V I I : General Issues
Ill
112
114
PART V
CHAPTER X V I I I : Proposed Resolutions
117
APPENDIX
Resolutions of t h e Second International Conference of Labour
Statisticians, Geneva, April 1925: II. Unemployment
Statistics
129
PREFACE
The Sixth International Conference of Labour Statisticians
was held in Montreal from 4 to 12 August 1947 under the
auspices of the International Labour Organisation and adopted
Resolutions setting international standards for the compilation
and publication of statistics of employment, unemployment
and the labour force, cost of living and industrial injuries.
The present report on employment, unemployment and
labour force statistics, including proposed Resolutions, was
printed and circulated in advance to the Governments. As the
report, although prepared primarily for t h e Conference, may be
of general interest as a study of the subject, it has been reprinted
separately from the account of the proceedings of the Conference
itself. The present report is as submitted to the Sixth International Conference of Labour Statisticians except for a few
minor corrections.
The Resolutions as finally adopted, together with a résumé
of the discussions in the several Committees dealing with the
different topics on the agenda of the Conference will be found
in the General Report of the Conference, which is issued in a
separate publication. 1
1
Studies and Reports, New Series, No. 7, Part 4.
INTRODUCTION
The subjects of employment and the labour force have become
a major focus of economic and social policy in all advanced
countries. In part, this is the result of the heavy demands of
war: in the belligerent countries all industries were combed to
provide the manpower needed in the armed forces and in the
civilian production industries which were necessary for their
support. In consequence of these demands, attention was
drawn to the need for statistical data on employment, unemployment and the labour force. Apart from the influence of
the war, the need for full data on employment and employment
policies has been more and more recognised, because of t h e
growing realisation that employment and employment policies
are of fundamental concern to the economy. The impact of
the depression in the 'thirties and its long duration forced
economists and social planners to consider long-run policies
in terms of employment for the working population. Full
employment has become the objective of those who hope to
prevent economic depression and to raise the level of welfare
of the population as a whole.
All these developments emphasise the need for adequate
and accurate statistics of employment, unemployment and the
labour force. It is of interest in this connection to note the
shift of emphasis from unemployment as a major social problem
— in the period following the First World War — to employment as a major task for sound economic planning.* The
change includes an extension of the concept of the problem to
cover not only employment as such, but also the development
of methods of fitting the available labour force to employment
opportunities in a manpower budget plan.
1
This is reflected in the relative number of statistical series on the
two subjects. In 1924, when the International Labour Office began
the publication of current series on labour topics, only 4 series were
included for employment, as against 19 for unemployment. The 1941
edition of the I.L.O. Year Book of Labour Statistics included 32 countries
in the employment tables as compared with 31 in the unemployment
tables.
2
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
Other aspects of the developing interest in employment may
be touched upon briefly. The trends of economic development
are reflected in employment trends: countries advancing on
the road towards industrialisation can trace their course in
the statistics of the shift in the distribution of the gainfully
occupied from agriculture to manufacturing industries, and in
the increased proportion engaged in commerce and other services. Estimates of employment in different countries are
useful in indicating the trends of business and in forecasting
prosperity and depression.
The measurement of employment, unemployment and
labour force and of their fluctuations is now recognised as
vitally important for the period of reconstruction and for the
post-war world, where problems of full employment and the
reduction of unemployment to a minimum are in the forefront
of interest.
Labour Force Statistics at Previous International Conferences
Employment.
The Second International Conference of Labour Statisticians,
held in April 1925, included the topic of unemployment statistics
on its agenda. One Resolution urged the collection of information on the state of employment, to be "published periodically,
if possible monthly, based on returns made by a representative
number of employers " . 1
Unemployment.
On the subject of unemployment statistics the Second
Conference adopted a series of Resolutions with respect to
standard definitions, procedures and comparability. These
Resolutions are reprinted in the appendix to this report.
The Gainfully Occupied.
At the First International Conference of Labour Statisticians,
held in Geneva in October 1923, the problem of classifying the
1
See International Standardisation of Labour Statistics, Studies and
Reports, Series N (Statistics), No. 25 (Montreal, 1943), p. 68. In addition,
statistics of employment were treated as a necessary element in calculating percentages of unemployment on the basis of the sum of employed
and unemployed.
INTRODUCTION
3
gainfully occupied by industry and occupation was discussed.
The topic was taken up subsequently at the Second and Third
Conferences, and as a result a list was drawn up by the International Labour Office for use in classifying and arranging industries for purposes of presenting labour statistics in a uniform
arrangement.
In this connection mention should be made of the minimum
nomenclature of industries drawn up by the Committee of
Statistical Experts established in pursuance of the International
Convention Relating to Economic Statistics. The minimum
nomenclature was drafted for use in classifying the gainfully
occupied according to industry in censuses of population.*
In this connection also, the work of the International
Conference Relating to Economic Statistics called by the
League of Nations in 1928 should be mentioned. This Conference adopted a Convention which provided for the taking
of statistical surveys or censuses of industrial establishments,
including among other data the "number of persons of each
sex employed therein; and distinguishing, so far as possible,
such persons according to categories of their employment, and
distinguishing adults from young persons, the age at which this
distinction is made being stated. An estimate shall also be
made, if possible, of the number of persons employed in establishments which are not included in the survey. " 2
The Sixth International Conference of Labour Statisticians
In view of the increased importance of employment statistics,
the International Labour Office proposed that the Sixth International Conference of Labour Statisticians should be convened
to consider the means and methods by which adequate statistics
1
L E A G U E O F N A T I O N S , Studies a n d Reports on Statistical Methods,
No. 1: Statistics of the Gainfully Occupied Population.
Definitions and
Classifications recommended by the Committee of Statistical
Experts
(Official, No. C 226 M. 128-1938 (C.E.S.127) Appendix I ; Ser. L.O.N.
Pub. I I , Economic a n d Financial, 1938 II A. 12).
2
L E A G U E OF N A T I O N S , International Conference Relating t o Economic
Statistics, 26 Nov.-14 Dec. 1928 (C.606 (1), M. 184 (1), 1928 II (C.S.O.
73), p . 4 of Article 2, V (A) (1)). In addition, provision was made for
"Statistical series in t h e form of either absolute, figures or relative
figures referring to a period taken as a basis of comparison, at regular
intervals, if possible quarterly or preferably monthly, showing the variations of t h e industrial activity of t h e most representative branches of
production." (Ibid., Article 2, V (C).)
4
EMPLOYMENT, U N E M P L O Y M E N T A N D LABOUR FORCE
of employment and unemployment and labour force might be
developed on a basis of maximum international comparability
in the different countries.
The Governing Body of the International Labour Office, at
its 100th Session in October 1946, decided to convene this
Sixth Conference, and placed the topics of employment and
unemployment statistics on the agenda in the following terms :
(1) Employment and payroll statistics; methods of obtaining an
overall picture of the volume of manpower and employment, as well
as its fluctuations, in the economy as a whole and in the major branches
of economic activity; the methods and uses of payroll statistics which
are obtained in conjunction with employment data.
(2) Unemployment statistics; methods of obtaining statistics of
the numbers and proportions of workers unemployed, and the fluctuations in these, in the economy as a whole and in the different economic
sectors.
A considerable expansion in the statistics of employment
and unemployment took place during the war. Among important recent improvements and statistical techniques has been
the development of the population sample survey. A further
expansion of these statistics is in process or in prospect in many
countries. It will be the task of the Sixth International Conference of Labour Statisticians to further this expansion by
establishing a common ground of definition, procedure and
presentation.
Preparatory Meeting
In connection with the calling of the Conference, the Governing Body of the International Labour Office authorised the
convening of a small preparatory meeting of statistical experts
for a preliminary discussion, with especial reference to the
definition of the topics to be included. At this preparatory
meeting, which was held in Montreal from 4 to 7 March 1947,
drafts were submitted of the various reports on the different
items on the agenda. The report as presented here has been
revised in the light of the discussions at this preparatory meeting,
which also recommended that the questions of employment and
unemployment statistics should be combined in a single
report. In accordance with this recommendation, therefore,
this report includes the discussion of both questions. *
1
The Office wishes to express its thanks to the experts who participated in the preparatory meeting for their most helpful criticisms and
suggestions. It. should be pointed out, however, that the Office assumes
full responsibility for the proposals submitted in this report to the Sixth
International Conference of Labour Statisticians.
INTRODUCTION
5
Sources of Materials
In preparation for the Sixth Conference the Office asked
the Governments of a number of countries to prepare monographs describing the statistical methods used in their statistics
of employment and unemployment. Separate standard outlines were drawn up for the different types of statistics of employment and unemployment, namely, establishment statistics,
industrial censuses, sample surveys, social insurance systems,
unemployment insurance and employment exchange data, the
two last mentioned referring to unemployment statistics.
Monographs were received covering each type of statistics and
presenting answers to a series of questions on the specific
methods used in compiling data. These monographs have
been used in the preparation of this report and the materials
furnished have been invaluable as a basis for drafting the proposals contained in it. The Office wishes to express its thanks
to the different Governments which contributed these monographs.
Payroll Statistics
It was found impossible to treat the subject of payroll
statistics adequately on the basis of the information received.
Furthermore, from many points of view it seemed preferable,
instead of treating payroll statistics as a single independent
source of earnings data, to broaden the inquiry to include the
whole subject of wages and earnings as derived from all sources.
The preparatory meeting, in considering this question, recommended the omission of payroll statistics from the agenda of
the present Conference. The Sixth International Conference
of Labour Statisticians, however, may wish to suggest that the
topic of the statistics of wages and earnings should be placed on
the agenda of a future session, and a proposed Resolution to
this effect is included (No. II).
PART I
I.
OBJECTIVES
Many important purposes may be served by statistics of
employment, unemployment and the labour force. In any
detailed study of international procedures and standards it is
useful to keep them in mind. A brief review of these purposes
is offered in the present chapter.
During the war years extensive and detailed information on
the labour force was collected in most of the belligerent countries.
Such data served to indicate where potential reserves of labour
might be found; how these reserves were tapped and drawn
into the labour market; and how the limited supply of labour
was allocated to various civilian industries. Only on the basis
of such information was it possible to apportion workers between
the armed forces, war production and civilian industry without
destroying the balance of production.
Effective social policy in periods of reconstruction, even
more than in wartime, must be based upon a knowledge of the
number and characteristics of both present workers and those
potentially available. Such data can indicate the numbers
of persons in unemployment or not in the labour force who may
be drawn into productive activity. They can suggest the limits
of available labour supply under specified conditions. They
can point out the actual uses of manpower by industry. Thus
in one country, for example, data on the increasing number of
persons engaged in non-essential construction activities in postwar months emphasised the possibility of using that labour in
other activities. To reconcile manpower needs with supply,
and to do so with as much skill and precision as possible, current
employment, unemployment and labour force figures are
valuable tools.
OBJECTIVES
7
These data serve as the foundation for most evaluations of
the condition of labour. What single fact tells more about the
welfare of labour and the economic health of a nation than the
proportion of the labour force that is unemployed ? Intelligent
and effective collective bargaining rests, among other things,
on a knowledge óf just what the level of employment in a given
industry or region is and what trend it is taking. Problems
of reducing seasonal unemployment, of stabilising the course of
the business cycle and of guaranteeing annual wages require
for their solution data on employment and unemployment, as
well as studies to measure, for example, the frictional minimum
of unemployment or to determine how many workers are covered
by unemployment insurance. In order to ascertain average
earnings or annual earnings, or to compute the income available to the worker's family, it is necessary to know how many
persons are employed, how many members of families are
employed, and the portion of the year for which each is
engaged.
Business men use data on changes in employment and
unemployment to forecast the changes in consumption expenditure. They use such data in computing the productivity
of labour and in working out the pattern of changes in the
demand for labour by their industry. Changes in this demand
in turn affect their hiring policy, their pension policy and, to
some degree, their pricing policy. They use such data, finally,
in measuring labour cost per unit — one determinant of the
competitive position of a firm or an industry.
Employment and labour force statistics are a valued adjunct
in the operation of almost every local and central government
policy. Changes in tax policy, for example, require an estimate
of revenue from alternative sources. But such revenue forecasts must be built upon the data of employment trends if they
are to be soundly based. Agricultural programmes demand
knowledge of the changes in size of the farm labour force and
of migration to the cities, as well as of those changes in farm
income which grow out of variations in the level of national
employment. Education authorities must, plan school construction and school curricula on a knowledge of the shifting geographical distribution of the labour force, and of the trends in the
occupations demanded. Social security authorities obviously
cannot operate without comprehensive information on the
labour force; it is more than a coincidence that these systems
8
EMPLOYMENT, U N E M P L O Y M E N T A N D LABOUR FORCE
have been primary contributors to the development of employment and unemployment data.
Finally, the forecasting of future levels of economic activity
and forthcoming changes in the industrial structure must rest
on a broad base of detailed statistics on all aspects of the labour
force. That forecasting may be in connection with national
employment programmes for reconstruction, such as those now
operative in France, Norway, Czechoslovakia, Great Britain
and other countries, or it may be required for the use of business
and labour in making their plans, where no such national programmes exist. In any event, such forecasting is becoming of
increasing importance. And to be fully effective it must rest
on a comprehensive scheme of current data on employment,
unemployment and the labour force.
The elements of such a comprehensive scheme of data on
employment, unemployment and the labour force may be indicated in broad outline. Though the details of these various
elements will be discussed in full in later chapters, it may be
useful to survey them briefly, not in order to indicate what any
one country should have in the way of statistics on these topics,
but rather to state an ideal system towards which, as resources
and facilities permit and as needs for data suggest, the statistics
of the different countries may be expected to develop.
A comprehensive system of statistics of employment,
unemployment and the labour force will provide useful data
drawn from various sources and developed from the extensive
battery of techniques and methods available in this field.
The data will include both benchmark statistics indicating
the status of the employed and unemployed groups at a given
date, drawn for example from censuses of population, labour
force sample surveys, social insurance data and other sources,
as well as current series indicating movements of employment
and unemployment over the period since the date of the benchmark figures. Such current series may include figures based
upon establishment samples, labour force sample surveys and
trade union data showing current fluctuations, and figures
based upon combinations of these or upon other data.
The methods will include the various refined techniques
which have been developed in the different countries to meet
the specific problems of employment and unemployment
estimates. One of the chief of these is the application of
sampling techniques to the problems of employment and
DEFINITIONS
9
unemployment statistics. Though the development of these
techniques is a task for the specialist, it is important to focus
attention upon the possibilities of such applications to this
field of labour statistics. Not only this method, but other special
methods, such as the special techniques developed for estimating
bias, should be available for use wherever their application
would be helpful.
As a final result, each country should develop its statistics
in these fields, in accordance with its own requirements, in the
direction of providing data drawn from the best sources and
worked up with the aid of the most appropriate techniques.
In this way, not only will each country be able to furnish adequate statistics according to the best methods, but also, by the
adoption of such programmes, international standardisation
and international comparability of these statistics will be
advanced.
II. DEFINITIONS
The adoption of standard definitions is an important and
indispensable preliminary step in the development of standard
methods and techniques in any field of statistical enquiry. A
sufficient groundwork of agreement already exists in the various
countries for standard definitions for the employed, the
unemployed and the labour force. These are set out below.
I. T H E
EMPLOYED
For statistical purposes, the "employed " should include
all persons who work for their own account or in the employ of
others — employers; persons working independently, i.e., who
work for their own account; salaried employees; wage earners;
and unpaid family workers who are engaged in tasks directly
related to the operation of a family enterprise for a minimum
of 15 hours a week, not including time spent in unpaid domestic
work (I, 8 (1)). *
In other words, the employed include all persons in remunerated activities, regardless of age. The remuneration received
1
Throughout the report, the roman numeral refers to the number
of the Resolution, and the arabic numerals to the paragraphs and subparagraphs.
10
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
may be in money or in kind. Employers and self-employed
persons may actually receive no net income whatever in any
given year. In all cases, however, the work must be of such
type and character as is customarily remunerated.
Since remuneration for the housewife is neither customary
nor expected, and since her work for the family is not subject to
appraisal by the labour market, housewives are excluded from
the "employed "- 1
One group of considerable importance, especially in agriculture, deserves special discussion — unpaid family workers.
It is obviously necessary to have some basis for determining
the number of those who are more or less actively concerned
with production for the market, including those who as family
members receive no pay for this work, otherwise the measurement of agricultural productivity and international comparisons
of employment and productivity data are quite impossible.
Three alternatives can be envisaged. First, all persons who
perform any such work could be counted as unpaid family
workers, and classified according to the number of hours worked.
These figures would then be used to determine the total volume
of man-hours for the calculation of productivity. An alternative is to include only persons who perform a minimum
amount of such unpaid labour, for example 15 hours a week,
in tasks directly related to the family enterprise. Such a
definition would apply to women, children and other
supplementary earners on farms as well as to unpaid family
workers in trade and other urban enterprises. This definition
has the considerable advantages of simplicity and definiteness.
It is possible, however, that the significance of a 15-hour minimum might differ from country to country according to the
length of the farm work week. A different percentage of total
man-hours worked in agriculture would probably be excluded
by such a limitation as applied to the United States from that
excluded by such a limitation as applied, for example, to Poland.
A third possibility would be to state the criterion for inclusion
in terms of the average hours per week of a full-time worker, and
1
The group of housewives is not excluded on the ground that the
housewife's work is not " productive" or " useful", but for the practical
reasons cited. If the criterion for inclusion in the " employed " groups
were that the work should be " productive ", it would first be necessary
to set up a satisfactory definition of " productive " and a method for
the application of such a criterion that would not require arbitrary
decisions by the statisticians. In fact, the criterion used is that the
work should be remunerated, not that it should be productive.
DEFINITIONS
11
to include in the labour force, for example, those unpaid family
workers who are engaged in tasks relating to the operation of
the family enterprise for at least half as many hours per week
as the average full-time workers in the same industry. Such a
criterion, of course, would vary from country to country and
from industry to industry and would relate the proportion
excluded in any country to the national averages. Special
studies of national data would be required to determine the
effect of using such definitions. The application of such a
definition, for example, to agriculture in Canada, where the
average hours of work in 1946 were over 65 a week, would mean
that unpaid family workers in agriculture who worked less than,
say, 33 hours a week would be excluded from the labour force. x
In any case, for purposes of international comparisons of productivity, it is desirable to have a complete distribution of all
persons, including the excluded unpaid family workers, according
to hours worked per week.
A second point concerns the limits of age, if any, to be
introduced into the definition. On this question, the simplest
recommendation would be to have no age-limits, in order to
obtain a complete measure of the labour force. In practice,
however, law or custom may set age-limits for employment; it is
usual to set a minimum age-limit for reporting or for tabulating
the returns of the gainfully occupied, corresponding to the age
below which few or no children are employed. This minimum
limit varies from country to country; in Norway it is 15, in Great
Britain the tabulations start at 14, and at the 1940 census in the
United States the lower age-limit was 14. However, the disadvantage of adopting such a minimum limit is that it makes it
difficult to carry out any investigation into questions relating
to child labour which may be concerned precisely with the
extent and importance of the labour of children under these
minimum ages. For such purposes data on employment under
these ages are needed.
At the upper ages, also, it is difficult to justify any definite
limit. In practice, farmers or persons in other occupations
where a considerable freedom of working hours prevails may
work to advanced ages. Data may be needed to throw light
upon the actual extent of employment of old persons, for
1
For other industries such a test, if rigorously applied, would
involve a considerable amount of work—more work than would probably be justified.
12
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
example, with relation to the pensionable age in an old-age
pension plan. In general, therefore, the widest possible coverage
of ages is to be recommended.
Two special groups should be mentioned. The "employed "
should include persons in labour camps if they are free to seek
alternative employment, but should not include persons in such
camps who are not free to make that choice, or persons confined
to prisons or similar institutions (I, 8 (2)).
The "employed " should include persons who are directly
employed by any public authority on emergency public relief
work, but where such persons are employed under conditions
inferior to those of regular public employees engaged in the
same type of work their number should be indicated separately
(I, 8 (3)).
II. T H E UNEMPLOYED
For statistical purposes, the "unemployed " should include
all persons seeking work on a given day who are not employed
but are able to take a job if offered one (I, 9).
The worker must be seeking work to be classified as unemployed. He may be registering at an employment exchange,
visiting prospective employers, answering advertisements or
making similar efforts to secure work. (Workers in communities
where the chief industry is shut down or not hiring should also
be included when they would be looking for work if jobs were
available in their occupation.)
The requirement that the worker must be seeking work in
order to be classified as unemployed excludes persons in such
institutions as prisons and mental hospitals, persons who are
ill and persons in other similar categories. Persons who have
retired from employment are excluded from the unemployed
because they are not seeking work. On the other hand, young
persons who have never before held positions but are seeking
to enter industry are properly included among the unemployed.
The worker must be able to take a job if offered one. This
provision excludes persons specifically prohibited by law or
regulations from taking any employment, e.g., tourists or
immigrants in special categories who are denied the right of
accepting employment in the country to which they have been
admitted. Persons who are ill, and those who are in custody
or detention so that they could not accept work if offered are
DEFINITIONS
13
also excluded under this provision. On the other hand, persons
who are merely subject to handicaps — such as being too old,
too young or too inexperienced, lacking the specific training
needed for a given job, women who have family responsibilities
to fulfil in addition to their work, persons partially crippled
or suffering from other similar disabilities — are not excluded.
All workers are subject to disabilities in getting work, and some
more than others. It is not for the statistician to exclude from
the "unemployed " group any who are seeking work and are
able to take a job if offered one.
Workers may be declared ineligible for unemployment
benefit because they are over a given age, because there is
little prospect of their being re-employed, or because interviewing panels consider them unsuited to ordinary employment. These rulings involve the application of specific provisions of unemployment insurance laws. Such workers will properly be excluded from the total of those who receive benefits.
But if they are seeking work and are able to take it if offered
they belong just as properly in the total of the " unemployed".
Experience during the war demonstrated that the employability of older workers tends to be a function of the level of
employment. Many workers are regarded as " unemployable "
in periods of heavy unemployment who can and do obtain
employment in periods of labour shortage.
A worker is "unemployed" when he is actually seeking
work, even though he demands a rate of pay higher than he is
likely to secure, or is looking for a job in an occupation for which
he is not qualified. The decisions as to the rate of pay he will
accept and the job he wants are made by him. This interpretation of unemployment is linked up with the right of the
worker's free choice in a free economy. These decisions are
parallel to the decisions of the employer in hiring. As the
employer need not take any given worker — regardless of
qualifications and regardless of the wages the worker is willing
to accept — so the worker need not take any given j o b . x Both
can continue to seek what they require regardless of how
unrealistic their aims may seem to others. Neither is held to
any given standard of what is reasonable or "really deserved ",
or of the speed with which a job should be filled or accepted.
Workers are included in the employed group, no matter how
1
This qualification, or indeed the entire discussion, does not apply
where such freedom of choice is limited or abrogated.
14
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
low their productivity and how unemployable they may have
appeared to observers the day before, provided only that they
are employed. Similarly, workers must be included in the
unemployed group no matter how low their productivity may
prove to be or how unreasonable their economic aims may
appear, provided only that they are seeking work and able to
take a job — but not necessarily any given job — if offered one.
It follows from this definition that an unemployment total
which includes young and old workers, those whose chances of
re-employment are excellent and those whose chances are poor,
does not provide enough information about unemployment.
An accurate unemployment total must be supplemented by
particulars of the characteristics of the unemployed. Information must be provided on how many of the unemployed are
young workers attempting to enter the labour market; how many
are older workers, nearing the end of their working Ufe because
of age or changing technologies; how many have been without
work for a week, a month, six months, a year; how many seek
part-time, and how many desire full-time employment. Whatever totals are computed in any country for the insured unemployed, the number of unemployed drawing benefits, or still
other groups, the most satisfactory measure of total unemployment will include all persons seeking gainful work who are not
actually employed, provided they are able to take work if
offered it. The usefulness of this comprehensive measure will
be greatly increased if it is supplemented by distributions
providing information on the characteristics of the unemployed.
III.
LABOUR FORGE
For statistical purposes, the "labour force " should include
all employed and all unemployed ptrsons, as defined above,
including the armed services. The civilian labour force should
include all employed and unemployed persons exclusive of the
armed services (I, 10).
IV.
TIME REFERENCE
One further point remains to be considered with reference to
the definitions of employment, unemployment and the labour
force. To what period or point of time should the definitions
relate; should they refer to the status on a given day, during a
given week, or for some other period ?
DEFINITIONS
15
In deciding this question, three points must be borne in
mind: (1) the correctness of the result in obtaining a true
picture cf the prevalence of employment and unemployment;
(2) the variations in status over any given time; and (3) consistency with other data.
In general, in most countries, both census data and current
data on employment and unemployment refer to the status as
on a particular day.
Censuses are practically always taken
with reference to a critical date, which is chosen with especial
regard to the effects of the choice upon the various results
obtained. With regard to industry and occupation, however,
census data have commonly referred to the " u s u a l " status,
and, in the case of the unemployed, to the industry or occupation
of last employment. In current statistics of unemployment,
the data are commonly referred to a particular date, as, for
example, the end of the month. In employment exchange data,
the actual number of persons on the register of unemployed at
a given time is sought, in preference to the total number of
applicants during a week or month; it is essential to distinguish
between the flow of applications and the numbers existing at
a particular moment, and the latter figure is the significant one
for the percentage of unemployment. 1 Similarly with unemployment insurance data, the number and percentage of workers
unemployed at a particular time show the prevalence of unemployment, while figures showing the whole number of workers
who have received benefits or have been unemployed at any
time during the month would grossly overstate the true percentage of unemployment.
In order to take account of variations in the status over a
period of time, the average number or percentage of unemployed
persons on each day over a period óf a week or a month can be
taken. In particular, the method of dividing the total days of
unemployment by the sum of the possible man-days of work
during a week was recommended as a method worthy of
consideration by the Second International Conference of
Labour Statisticians. 2
1
But of course the particular day, as discussed below, should be
chosen so as to reflect typical rather than abnormal conditions — or an
average of daily percentages may be shown : these give the picture of
unemployment at a given moment, or an average percentage, rather
than the flow.
2
See Appendix, p. 129.
16
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
In the 1940 census of the United States, in order to obtain
better data on employment and occupation, the week prior to the
critical date of the census was taken as the time reference for
data on employment and occupation.
The labour force surveys in the United States and Canada
have adopted the enumeration week as the period of time
reference; in order to be counted as unemployed, however, the
individual must have been unemployed throughout the week,
while in order to be counted as employed, any employment, no
matter how short in duration, is sufficient. Under this definition
obviously the number and proportion of " unemployed " are less
than if all those unemployed for one day or more (the definition recommended by the 1925 Conference) were included.1
As to establishment sample statistics, in most countries the
employment data are referred to a specific date, usually the
date of the payroll day, or in some cases the last day of the
month or the last Friday in the month, etc. 2 In other countries,
the data may include all persons on the payroll. It is obvious
that if the employment data relate to the same date, the extent
of double counting, where a person appears on two payrolls
1
The effect on the number of employed, in view of their larger number,
is relatively negligible. In discussing the subject Ducoff and Hagood
give two considerations: the period for reporting census labour force
status should be long enough t o : (1) avoid fluctuations due to holidays,
etc.; (2) avoid bias in .covering only one phase of known short-term
periodic fluctuations, such as would occur if the reporting period
were a given day of the week, for example, Saturday. (Louis
J. DUCOFF and Margaret Jarman HAGOOD : Labour Force Definitions
and Measurement, Social Science Research Council, Bulletin 56, New
York, 1947, p. 17.) It may be noted, however, that fluctuations
due to holidays might affect the employment pattern of a week, but
that if the time reference is to a day a holiday would not be chosen ;
also that, in general, Saturday is not chosen as a date of reference.
Furthermore, if it is sought to obtain unemployment or employment
patterns over a period, this could be done by a procedure of sampling
of different days of the week or by a procedure of averaging, as noted
in the text, where average unemployment over a period of a month
may be measured.
2
In Sweden, as in Australia, what is sought by establishment reports
is the number of workers employed on the last work day of the last pay
period in the month. France specifies the first day of the quarter,
while in the United Kingdom the number on the payroll at a specific
day—31 August, 28 September—is requested, exclusive of those who
have been discharged or who have left during the week. New Zealand
specifies the number on the payroll on the pay day immediately preceding the 15th of the month. The establishment reports of Canada and
the United States ask for the number on the payroll for a period ending
nearest the end of the month for the Canadian and nearest the 15th of
the month for the American.
DEFINITIONS
17
because of a shift in employment during the payroll period, is
greatly reduced or eliminated. On the other hand, if the payroll
data are made up in such a way that an establishment cannot
easily furnish the numbers actually employed on a given day,
while the total numbers on the payroll can be given promptly,
the statistical offices may have to weigh the difficulties of obtaining compliance with their requests for information against the
advantage to be derived from obtaining the data for a particular
date, and may prefer to adapt their statistical demands to
the customary practices followed in making up payrolls.
Where this is done, however, the extent of double counting
remains a problem and should be made the subject of special
studies.
In reaching a decision on this point, the basic definition may
be distinguished from the working definition. For the basic
definition, the time reference should be of a given day. The
reason for this is not only that it is the common practice in most
countries, but also that, so far as unemployment and employment statistics are concerned, it yields an unimpeachable
result; furthermore, the different sources when compared will
be compared on the basis of the same definitions instead
of a series of different definitions. The basic difficulty in the
choice of a week or a month is not the length of the period,
but the errors involved if the condition is imposed that the
status of employment or of unemployment must last throughout the period, or that the status of employment and unemployment must be counted if it appears at any time during
the period. The clear definition of the numbers employed
or unemployed, therefore, requires the time reference to a
given day.
The designation of the particular day to be used of course
requires careful consideration of the various factors that affect
employed and unemployed, in order to avoid holidays, days of
partial activity, periods of exceptional short time, or periods of
exceptional activity.
A further point is that the public commonly refers to employment status as of a particular day, rather than to employment
" a t any time during a period". Calculations of employment
status with reference to a long period do not provide an exact
measure as of a day.
It is therefore to be recommended that in the basic definitions
the reporting of employment, unemployment and labour force
18
EMPLOYMENT, U N E M P L O Y M E N T A N D LABOUR FORCE
status should be referred to the status as of a particular day. 1
However, where it is deemed unwise to insist that the employer,
in making returns of payroll figures, should make up an entirely
new set of figures instead of those based on his customary
procedure, the statistical office may adopt a working definition
different from that recommended above and accept data
including all persons on the payroll in place of data showing
persons actually employed on a particular date. In such cases,
it is recommended that special studies should be made to determine the influence of such procedures upon the accuracy of
the resulting figures.
V. OTHER DEFINITIONS
Other definitions which should be considered include
industry, occupation, and industrial status.
By "industry " is meant the kind of factory, store, or the
place of business or kind of economic activity in which a person
works or exercises his occupation. The classification of a
person according to the industry in which he is employed is
determined on the basis of the nature of the activity of the
establishment in which he works, irrespective of the particular
occupation or kind of work he performs. 2
"Occupation" is the trade, profession or type of work
performed by the individual, irrespective of the industry in
which he exercises it.
" Industrial status " is the position of the individual in
respect of his employment: employer; independent worker on
his own account; salaried employee; wage-earner; and unpaid
family worker. The last-mentioned category has already been
discussed above. In most countries, managers and directors
are classified with employees, but in some with the employers.
In order to be able to make sound comparisons between different
countries, in countries where managers and directors are classified as employers, a special subdivision should show managers
and directors separately. The figures can then be recombined
to show the several industrial status groups on a uniform basis.
1
The Resolution of the Second International Conference of Labour
Statisticians phrased this point as follows: "the necessary and sufficing
condition for being enumerated as unemployed is that the individual
must have been not at work for one day at least." (Resolution II,.
para. 9 (c) (3).)
2
The classification of industries is discussed below in Chapter XV.
REPORTING SYSTEMS
19
IÌI. REPORTING SYSTEMS
At the present time there are two chief systems in use for
reporting the statistics of employment and four for the reporting
of unemployment statistics. Each of these will be discussed in
turn.
I. EMPLOYMENT
Of the two chief systems for reporting on current changes of
employment, both the establishment reporting and the labour
force survey techniques can provide precise and voluminous
data on employment. Both can provide such data promptly.
The choice between the systems, or the decision to use both,
will depend on a combination of circumstances which will
differ from country to country. The conditions of their appropriate use will depend largely on the existing scope of employment statistics, the character of additional data required, and
the relative cost of obtaining such further data by these alternative techniques or a combination of them. The additional
data needed may be (1) estimates of total employment, or
estimates of employment in very detailed industry groups;
(2) information on earnings and hours that is comparable with
such employment data; and (3) comparable data on unemployment and on various aspects of labour force participation. The
choice between reporting systems can be decided only by assessing total requirements for employment, unemployment and
labour force data in terms of cost and personnel.
1. Establishment Reporting
By far the commonest method of securing current information on employment changes is the system of postal questionnaires which are filled out in each establishment and returned
to the statistical agency for co-ordination into estimates of
employment. The establishment enters on the report form
data on employment in the current and previous months. The
statistical agency then computes the percentage change from
these data, applies these percentages to benchmark data and
computes either an estimate of change in employment from the
previous month, an estimate of employment in the current
20
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
month, or both. The series may apply solely to reporting
establishments, as in Canada, or to all establishments in the
industries covered. Such reports are obtained in the Netherlands,
Czechoslovakia, Norway, Denmark, France, Australia, Great
Britain, Switzerland, New Zealand and the United States,
among other countries.
This system of employment reporting was originally
developed in most countries in response to a demand for full
information on employment, earnings, hours of work and labour
turnover in individual manufacturing industries, and it is
admirably adapted to these purposes. It has since been applied
to a variety of other industries, and in some countries the data
obtained have been used for computing estimates of the total
number of non-agricultural employees. The system is effective
in securing full and consistent data on employment, hours and
earnings for detailed industry groups at a very low cost per
return.
Since the returns are filled in by employers, accurate and
detailed information on industry is obtained; this furnishes the
basis for the classification of establishments according to
detailed industry groups and sub-groups. A further advantage
is that an establishment reporting system can be begun on a
relatively limited scale, surveying at first only those industries
which are considered to be of greatest importance, and gradually
expanding in coverage as funds and personnel permit. On the
other hand, this procedure is not so well adapted to securing
data on the self-employed or domestic workers, or on employees
in certain broad industry groups such, for example, as trade and
service, and hence it is not so well adapted for providing
estimates on employees in all industries or on total employment.
2. Labour Force Surveys
A second and more recently developed system of reporting
on employment is that used in the United States and Canada
for surveys of the labour force. These surveys secure comparable, consistent and comprehensive data at one and the same
time on employment, unemployment and the labour force.
Since the survey operates through interviews with one responsible person in each of the samples of households surveyed, it is
obviously possible to obtain information on topics other than
labour force status. Experience has demonstrated that these
REPORTING
SYSTEMS
21
regular enquiries provide an excellent method of securing certain
materials required for economic analysis, and their potentialities
for these purposes are only beginning to be used. For example,
recent special questions asked by enumerators for the U.S.
Monthly Report on the Labor Force dealt with family incomes,
migration of workers since the end of the war, the amount of
work sought by unemployed and part-time workers, employment of veterans, rent paid by veterans and plumbing facilities
in their dwellings. Additional questions have been asked from
time to time on various aspects of labour force participation
and on many other items for which such intermittent enquiries
provide sufficient data at low additional cost.
The actual methods used in operating the survey may be
briefly outlined. The description is based largely on practice in
the United States, Canadian techniques being substantially
the same.
During the week containing the 8th of the month the enumerators call at all the dwelling places picked for enumeration.
One responsible member of the household is interviewed with
regard to the labour force status of all persons in the household
over 14 years of age. Since most interviews are conducted
during the day, this person is usually the housewife.
For purposes of the survey the U.S. Bureau of the Census
maintains 68 local offices to cover the sample areas used, each
with a permanent staff consisting of a supervisor and from
5 to 15 part-time enumerators. Each local office is responsible
for enumerating from 300 to 500 dwelling units, with a total
enumeration of 30,000 households each month.
The presence of a permanent staff leads to a stability in the
method of questioning and recording. Each enumerator is
provided with a manual consisting of nearly 50 pages of detailed
instructions relating to the different questions. Consistency
is further safeguarded by the presence of the permanent supervisors, who immediately review the data reported, and order
re-interviews wherever necessary.
Since the dimensions of the sample have been reduced to
a minimum in order to cut costs, the problem of non-reporting
becomes important. Standard procedure therefore specified
that call backs shall be made until all but three or four per cent.
of the persons chosen have been enumerated. These call backs
are particularly necessary in the case of families without children,
and families in which all the members are at work.
22
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
The households selected for enumeration are surveyed for
a six-month period and then dropped from the sample, thus
preventing the development of an undue burden on the families
selected and hence of resistance to enumeration.
Practice in the past has not emphasised strongly enough the
desirability of trying out the questions first. A recent analysis
of the discrepancy between the agricultural employment estimates of the United States Bureau of Agricultural Economics
and the United States Bureau of the Census led to a change in
the basic question on labour force status. This new definition,
effective in July 1945, produced a fairly severe shift in the labour
force totals, although in percentage terms the change was
relatively trivial.
It is therefore to be recommended that all
new questions for labour force survey schedules should be pretested in different parts of the sample in order to establish
definitely what interviewing.techniques are actually being used
and what patterns of answers are being given by respondents.
It is further recommended that portions of the, sample areas
should be re-enumerated at periodic intervals in order to check
on the interviewing techniques actually used by enumerators
(I, 48 and 49).
3. Social Insurance Systems
Estimates of employment changes have also been made
by deducting from the total number of insured persons the
number of insured unemployed. While conceivably this
procedure can give satisfactory results, it is not likely to give
such adequate data in practice except under certain limiting
conditions. These conditions are the following:
(1) The coverage of the insurance system in terms of particular industrial, age or occupational group must not change
significantly. Otherwise a mere increase in insurance coverage
will appear as an increase in employment.
(2) Estimates of the number of insured persons must be
continuously available on a reasonably accurate basis.
(3) Estimates of the number of insured persons who are
not at work or not unemployed must likewise be available.
Included in this group will be a number of persons who are ill,
on vacation, or absent from work for other reasons. Estimates
of their number must not, however, include persons who were
not actually in the labour force for the day or week surveyed.
It is perhaps suggestive that after using such a system for a
REPORTING SYSTEMS
23
number of years the United Kingdom switched over to establishment reporting for the current measurement of employment.
Provided that an adequate realisation of the potential flaws
in such reporting exists, further exploration may reveal effective
methods of using social insurance data for the current reporting .
of employment. In countries where employers report employment and earnings to the social insurance agency it may be
possible to develop a procedure for sampling such data directly. *
If only a sample of reports were utilised it should then be possible
to compute and to issue promptly employment estimates — an
impossibility at present because of the sheer bulk of the social
insurance returns and the period of time which elapses between
the date to which the return refers and the date at which it is
finally received by the national statistical agency. 2 Since the
files.provide a complete list of all operating establishments —
subject to certain size limitations — the basis for developing
a sound sample exists. 3
II. UNEMPLOYMENT
1. Registration at Employment
Exchanges4
Of all the measures by which statisticians seek to define the
level and changing course of unemployment, few are essentially
easier to obtain than data for registration at employment
exchanges. The technique is simple : a list is made of all persons
who register at the exchanges for jobs. Since for registration
there is no requirement that a worker should be currently
1
Following the technique used in Australia, it would be possible
for t h e employer to fill in a duplicate form with carbons, t h e second copy
to go to t h e unit concerned with current employment statistics.
2
For t h e sample of co-operating employers selected it would be
necessary to require returns to be filed by a specified date. Such a
concession, however, would save these employers t h e trouble involved
in filling out returns for t h e insurance agency and for an employment
statistics reporting agency. An integral element in support of this
technique would be a central master-list of establishments.
3
It is quite possible t h a t the reporting of employment data to an
insurance system differs from t h a t to an independent statistical agency.
However, since employment data may be benchmarked to social insurance data, as they are in the U.S., this would produce a consistency
between benchmarks and current data not otherwise present.
4
Unless otherwise indicated, discussions of the system of reporting
unemployment b y means of registration at employment exchanges refer
t o those which are not p a r t of a social insurance system. Those which
are p a r t of such a system are discussed under t h a t heading.
24
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
unemployed, he may have a job and be looking for another.
Since there is no requirement that he should be insured, he may
be a member of the insurance system or not. It is therefore
never possible to set any specified level of coverage or to assess
the efficiency of this method of reporting.
It is obviously unsatisfactory to include in the unemployment totals persons who have already found work. Where
attendance at the employment exchange is required any less
frequently than once a week it may be expected that an undue
percentage of all registrants will be persons who have already
found employment or have withdrawn from the labour force.
It is therefore recommended that studies should be made of
the number of persons reported as unemployed by employment
exchange registrations, in order to determine the proportion who
are actually employed or are no longer unemployed because
not seeking work (I, 43). Where this proportion runs to any
considerable size, steps should be taken to exclude those
persons from the unemployment totals, either by changing the
registration requirements, or by adapting the statistical procedures, or by both methods.
2. Trade Union Statistics
Trade union unemployment systems and unemployment
estimates have been historically of considerable importance.
Belgian trade union insurance systems were the precursors of
the present national insurance system. British trade union
data on unemployment provided the statistical basis for the
British system of unemployment insurance. Canadian estimates
of unemployment among trade union members constituted one
of the main bases for measuring unemployment in Canada until
a few years ago.
At the present time, unemployment among trade union
members is tabulated by only a few of the national agencies for
labour statistics, and, of the countries surveyed, it is the main
basis for measuring unemployment only in Australia and Norway.
In Australia trade union forms are forwarded by the Commonwealth Statistician to the permanent secretary or other official
concerned. If the union finds work for its members, that official
will have reasonably comprehensive, exact information at his
disposal for filling in the returns. " Some (unions) make provisions
for a reduction in membership fees in respect of unemployed
members " so that fairly accurate reporting can be expected
REPORTING SYSTEMS
25
from those. "Apart from these it is thought that a number of
union secretaries when compiling the return act to a considerable
extent on general impressions as to the state of employment in
the particular industry. It will be noted that the form includes
the note that 'if exact numbers are not available, approximate
or estimated figures will be of value '. " A distinction may or
may not be drawn between those who are unemployed because
of sickness or accident — hence not actually available for work —
and those who are unemployed for other reasons.
Before the war Norway collected statistics of unemployment
from returns submitted by 10 major trade unions. It is anticipated that the expansion of these statistics which is at present
being planned by the Central Norwegian Union of Workers will
provide a wider statistical basis in the future.
3. Unemployment Insurance Systems
The reporting of unemployment in unemployment insurance
systems is in many ways the simplest of all data-collecting
schemes, for the unemployed worker makes it his business to
report his status in order to obtain unemployment benefits.
In Denmark the basic reporting is that of the individual
trade union member who reports his lack of work to a trade
union unemployment fund, these funds being an integral, and
indeed the chief, part of the national insurance system. Uninsured workers report their unemployment to the extent that
they register with the public employment exchanges, these
registration totals being then included in total reported unemployment. The insurance funds report their data to the public
employment exchanges under three headings: reports on individual workers (except unskilled), monthly lists of those registering as unemployed during the month, together with an annual
list of fund members. A summary statement is then drawn
up giving for each fund the number reporting themselves as
unemployed for each day of the week, separately by sex.
In Great Britain the insured worker who becomes unemployed
and wishes to secure employment and/or insurance benefit
turns in his unemployment book at one of the 1,200 local employment exchanges of the Ministry of Labour. x To the insured
1
Two additional reasons for registration mentioned by the British
report are the fact that registration is required in order to avoid payment
of arrears of health insurance contributions and, in most instances, as a
condition for the receipt of poor relief.
26
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
persons who register as unemployed are added other unemployed persons who register for work. Since attendance at
the exchange at least once a week is usually required of all
persons registered for work with the employment exchanges,
it would appear that the total register is not likely to be overloaded with persons who have already found employment or
have for any other reason ceased actively to seek work.
The data collection procedure in the Netherlands is essentially similar. The country is divided into 25 major districts,
subdivided into a number of smaller units, with employment
exchange offices in each of these smaller units. Any resident
of the Netherlands may register for placement. The registration cards of these unemployed workers remain in the active
file, until the workers either (a) are placed, or (£>) notify the
exchange that they have found employment, or (c) do not
extend their registration on a specified day at the end of the
month. Registration is, of course, a prerequisite for the receipt
of benefit.
In Belgium, likewise, the unemployed worker seeking
insurance benefits must register, presenting himself at one of the
communal administration units in the' 2,670 communes. He
must report daily to the unit in order to ensure continued receipt
of his benefit. Those who are not insured — as well as those
who are —•• may register for work with the employment
exchanges. However, as the Belgian report notes, these latter
figures tend to be inaccurate, despite numerous attempts to
remedy their failings. Their shortcoming is common to most
employment exchange statistics and to some social insurance
data as well — the registers are not kept up to date and generally
include a number of workers who have already found work but
have not notified the exchange to that effect.
Some of the limitations of a social insurance system for the
reporting of unemployment are discussed below in Chapter XIII,
but of the considerable value of such data there can be little
question, for social insurance statistics of unemployment are
in a very real sense cost free, being by-products of the operation
of a system installed for other than statistical purposes, and it
therefore becomes possible to expand statistical coverage and
derive additional estimates (as well as to secure the basic estimates) at a remarkably low cost assignable to statistical purposes. Furthermore, the continuing contact of a social insurance
system with individual workers makes it possible to conduct
REPORTING
SYSTEMS
27
a variety of special studies on unemployment problems at low
cost and with little inconvenience to the employee. The
possibilities of using social insurance data for the study of
unemployment problems have only begun to be explored.
4. Labour Force Surveys
Labour force surveys constitute an effective method of
securing a complete estimate of total unemployment — an
estimate not subject to the industrial, age and other limitations
characteristic of most insurance systems and free from the
problems peculiar to statistics of unemployment as reported
by trade union or independent employment exchange systems.*•
The decision to use a labour force survey for reporting on unemployment will depend (a) on whether a comprehensive social
insurance system is in operation, and (£>) on the extent of additional labour force, employment and unemployment data required
but not available from social insurance sources at a reasonable
cost and with reasonable promptness. (The actual technique of the
labour force survey is outlined above in section I, subsection 2
of the present chapter.)
5. General Conclusion
Because of the general problems involved in the use of
employment exchange registration data ortrade union unemployment reports as measures of national unemployment — problems noted above and discussed further in Chapters XII and
XIII — it is recommended that wherever possible estimates of
the total numbers unemployed in any country should be derived
either from the operations of a social insurance system, or from
sample surveys of the labour force, or from both sources (I, 36).
1
Unlike other methods of reporting on unemployment, the population survey secures data through trained officials whose duties are first
and last statistical. Their focus of interest is on statistics and not on the
administration of a placement or benefit-paying mechanism. Furthermore, local supervisors can check on the sufficiency of the data gathered
immediately and not at a long remove, in contrast to the procedure
necessary in the case of statisticians who work with social insurance
data.
PART II
IV. TYPES OF EMPLOYMENT SERIES :
NATIONAL PRACTICE
The various systems which are in current use for the reporting of employment have developed in response to the pressure
of special needs and particular emergencies quite as much as
from systematic planning. A series on employment in heavy
industry may have been established because of its value in
guiding war production. A series on employment on public
works may have developed because of the impact of a depression.
In many countries series on employment in manufacturing
were set up because such employment was considered to be a
sensitive indicator of the course of unemployment. As a
result of this kaleidoscope of forces the various systems of
national employment statistics differ widely in their comprehensiveness and detail. To appreciate the full measure of this
variation and the problems to which it contributes it is desirable
to review the chief types of series and practice in the various
countries. Attention in this chapter is directed successively
to national series on total employment, employment by industry,
by geographical area, by age, by sex, by occupation group.
The next chapter deals with some general considerations relating to these series.
I. TOTAL EMPLOYMENT
A considerable number of countries have a series on total
employment scheduled for development in the near future,
but at the present time only three of the countries for which
monographs were submitted obtain current estimates of total
employment: the United States, Canada, and the United
Kingdom. Both the United States and Canada obtain their
estimates of total employment from current labour force
EMPLOYMENT S E R I E S : NATIONAL PRACTICE
29
surveys, taken at monthly and quarterly intervals respectively.
The estimates include all persons in employment — wageearners, salaried employees, employers, self-employed persons,
unpaid family workers; agricultural and non-agricultural
workers; those in large firms as well as in small ones; and persons
in all branches of economic activity.
The British monthly series for total employment includes almost all persons in employment except domestics and older workers. Data for wage-earning and salaried workers are secured by
means of establishment reports, while self-employed and unpaid
family workers are estimated by a variety of assumptions and
techniques.
Some other countries, such as France, obtain monthly
estimates of total non-agricultural employees, and some, such as
Australia, obtain monthly estimates of private non-agricultural
employees, but the majority of the other countries have as their
most comprehensive current series one for manufacturing alone,
or manufacturing combined with a number of other non agricultural industries.
II. EMPLOYMENT BY BRANCHES OF ECONOMIC ACTIVITY
The most detailed, most comprehensive and most frequently
available reports on employment by industry are those of the
United States. One or another of the national statistical
agencies of that country provides monthly estimates of employment in agriculture, mining, manufacturing, construction,
transportation and public utilities, trade, finance and service,
Government service, and all branches of economic activity
combined.
For agriculture, data are separately available on the number
of wage-earning workers and family workers who were engaged
the equivalent of two or more days in the enquiry week in
agricultural pursuits 1 as well as on the number of persons
working one or more hours in such pursuits during the survey
week. 2
For non-agricultural employment as a whole, monthly
estimates are available on the number of persons working one
or more hours at such pursuits during the survey week, including
wage-earning and salaried workers, unpaid family workers and
own account workers. 2
1
2
U.S. Dept. oí Agriculture.
U.S. Bureau of the Census.
30
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
For manufacturing, and separately for 20 manufacturing
industry groups and 153 industry sub-groups, monthly estimates
of production workers are available from establishment reports,
as well as a total for wage-earning and salaried workers in all
manufacturing. 1
For mining, the number of employees in 7 sub-groups is
available. 1
For construction, the number of employees in residential,
non-residential, farm, public utilities, highway and a variety of
types of Federal, State and local construction are separately
available. 1
For transportation and utilities, monthly totals are tabulated
by the U.S. Bureau of Labor Statistics, using estimates for
detailed sub-groups from the Interstate Commerce Commission,
the Maritime Commission and other agencies.
For trade, finance and service, and Government service, the
number of employees is available. 1
Of the countries surveyed, Canada alone has estimates
(quarterly) of the total number of persons engaged in each
major industry group — agriculture, mining, manufacturing,
etc. Monthly figures are likewise available for employees of
reporting establishments in manufacturing, logging, mining,
communications, transportation, construction and maintenance,
service, trade and finance. Separate details are provided on
employees in 19 manufacturing groups and 24 sub-groups.
Great Britain publishes monthly estimates of total employment, except that of older workers, in each major industry
group — agriculture, mining, Government service, etc. Monthly
estimates of employees other than higher paid salaried workers
are calculated for a number of major industrial groups and for
76 manufacturing sub-groups.
While none of the other countries surveyed secured data
on total employment — whether in the entire country or any
given industry — they obtained estimates of wage-earners
or of employees in considerable detail. France, beginning in
January 1946, began to collect quarterly estimates of employees
in 16 manufacturing groups and a variety of other non-agricultural industrial groups — construction, part of personal service, transportation (except the State railways), trade, amusements, banking and insurance, liberal professions. Beginning
1
U.S. Bureau of Labor Statistics.
EMPLOYMENT S E R I E S : NATIONAL PRACTICE
31
in May 1945, the Netherlands has released a series of quarterly
estimates for employees in some 9,000 industrial concerns.
Czechoslovakia began publishing continuous quarterly data on
employees in October 1945. The data are separately available
for 17 manufacturing groups, utilities and construction. Beginning during the war, Australia has released monthly estimates
of employees in manufacturing as a whole, in mining, trade and
the other major industry groups.
Among the older established series are those of Denmark
(monthly data on hours worked by employees in 8 manufacturing
groups) ; Sweden (monthly indices for employees in 8 manufacturing groups and 53 sub-groups, Government construction and
public works, construction, and small segments of trade and
transport); and Switzerland (quarterly estimates of employees
in construction and 17 manufacturing industry groups). Norway,
which had published monthly employment estimates for a
number of manufacturing groups and other industrial groups
for the period 1935-1946, is in process of developing a new
establishment sample; information on the details of this series
was not available in time for the present report. However,
monthly totals are published for the number of persons insured
against unemployment who are not unemployed.
III.
EMPLOYMENT BY GEOGRAPHICAL AREA
Norway issues estimates of employment separately for cities,
suburban or industrial communities and rural districts.
France publishes excellently detailed estimates not only for
total employment in the non-farm industries surveyed but also
for employment in each of the industry groups for which
national data are issued.
Canada and the United States likewise publish regional
data. The United States Bureau of Labor Statistics releases
monthly estimates of employment in the chief cities of the
country for a whole series of non-agricultural industries, and has
recently initiated similar estimates for each of the 48 States.
Total employment, however, is only available for the nation
as a whole. (Agricultural employment totals for the 48 States —
comparable with neither the Bureau of Labor Statistics figures
nor those of the Census Bureau — are issued by the U.S. Department of Agriculture.)
32
EMPLOYMENT, U N E M P L O Y M E N T A N D LABOUR FORGE
In Canada, estimates of employment in the detailed nonagricultural industry groups are available for each of the provinces and major metropolitan centres, while estimates of total
employment (employees plus self-employed) are available for
each of the provinces.
The reporting of employment in detailed areas, however, is
not yet common practice. Neither Australia nor Denmark
publishes current employment estimates for regions or large
cities. Sweden provides no current employment estimates by
region nor do Great Britain, New Zealand, Czechoslovakia or
the Netherlands. 1
TV. EMPLOYMENT BY SEX, A G E AND HOURS W O R K E D
Separate estimates for the current employment of males
and of females are available for Sweden, Great Britain, Australia,
New Zealand, Canada and the United States but not for Norway,
the Netherlands, France or Denmark.
Canada and the United States alone provide full detail on
employment by age groups, securing data for a large number of
age groups from their labour force series. New Zealand secures
semi-annual estimates of the number of those employed who
were over and under 21 years of age. (Several other countries
present estimates of the number of insured persons by age
groups.)
Data on hours are customarily obtained in connection with
the reporting of employment by establishments. Where such
data are utilised, however, it is most commonly in the form of
computed average hours of work per employee in the industry
or industries surveyed. At the present time only France,
Canada and the United States currently publish data on the
number of workers in each hours-worked grouping.
V. TYPES OF EMPLOYMENT SERIES :
GENERAL PRORLEMS
In the previous chapter the main types of series currently
issued in various countries have been enumerated. The considerable differences which may be noted between one country
1
Czechoslovakia gives figures on employment in the entire republic
and in Bohemia, Moravia and Silesia as a whole.
EMPLOYMENT S E R I E S : GENERAL PROBLEMS
33
and another give rise to certain basic questions. What should
be the scope of a national system of employment statistics?
Should that system provide estimates of agricultural employment, of non-agricultural employment, of total employment
and of the labour force? Should it provide employment series
for geographic areas within a country and for the separate age,
sex and occupation groups ?
Each of these main types of series will be discussed in turn,
but one general principle should first be emphasised. The main
purpose of a statistical system is to provide statistical information on the most considerable national issues. In one country
these may relate to the reconstruction of an economy broken
by fascism and war. In another they may relate to the pressing
demands of rising industrialisation, to ameliorating the lot of
the agricultural population. In a third, they may concern the
development of national resources, while in a fourth, the problem
may be how to ensure full employment. In each country,
the statistics of employment must be developed to serve its
appropriate needs. The emphasis in the following sections is
therefore on minimum requirements — requirements which in
any given country may be considerably less than the total
urgently needed for private and national policy.
I. AGRICULTURAL EMPLOYMENT
Every country with a sizable agricultural population will
require a periodic assessment of the labour force devoted to
agricultural production. Countries such as India and China,
which are entering the currents of world industrialisation, must
look forward to sharp changes in the distribution of their labour
force. They will devote an ever-increasing proportion of their
labour power to manufacturing and the service industries, and
an ever-decreasing proportion to agriculture. To measure the
rate at which this major change-over is occurring and to prepare
for the opportunities and problems which it brings in its train,
they must have sufficiently frequent assessments of the numbers
attached to agriculture. At least quinquennial, and better
still annual, assessments of total agricultural employment are
therefore desirable in these countries. Moreover, should substantial seasonal changes in farm employment occur, it is
further desirable that for the years enumerated data should be
34
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
secured for two separate dates — one at the seasonal employment maximum and the other at the seasonal employment
minimum.
Countries which are more industrialised have developed a
vast reservoir of agricultural labour in the form of small tenants
and wage-earning workers. These frequently form part of a
shifting labour force, divided between town and country, moving
with the seasons and changing with the income changes which
the business cycle brings. To focus properly on the quantity of
labour available in both areas, as well as to mark the variations
in the number of independent farmers and wage-earning workers,
annual estimates of agricultural employment are desirable.
Where seasonal shifts in agricultural employment are substantial,
estimates should be made more frequently than once a year
(I, 16 (1)). In the more industrially diversified countries, such
estimates should be made quarterly (I, 16 (2)). In other
countries, at least semi-annual estimates are desirable in order
to measure employment at its seasonal maximum and minimum
(I, 16 (3)).
II. NON-AGRICULTURAL EMPLOYMENT
As there are special problems for the solution of which it is
essential to have current estimates of agricultural employment, so
there are many which require data on non-agricultural employment. Non-agricultural employment is generally less stable than
farm employment, and its movement usually presages changes in
unemployment, in consumer purchases and in the level of
national income. It is for reasons such as these that many
countries now obtain estimates of non-farm employment, and
that all industrially diversified countries should secure at least
quarterly estimates of non-agricultural employment. This
requirement is not satisfied by obtaining estimates of employment in large establishments or in certain groupings of non-farm
industries. The general public is likely to use such limited
estimates as representing changes in non-agricultural employment. If it can be established that employment changes in
large firms or in six important industries do parallel those in all
non-agricultural industry, then no problem arises. But in
point of fact this parallelism is not likely to be found. It is
therefore desirable that countries having summary series now
restricted, for example, to large establishments or to a group
EMPLOYMENT S E R I E S : GENERAL PROBLEMS
35
of industries should expand their coverage in order to secure
fully representative estimates of non-agricultural employment.
III.
EMPLOYMENT BY INDUSTRY
Most countries which obtain any current employment data
whatever have estimates of employment in manufacturing as
a whole as well as in separate manufacturing industries. Many
also have series for mining, construction, trade and the other
major industry groups. So many of the individual industries
have their special problems, so many are subject to special
legislation, so many affect the welfare of particular regions
within a country, that by now detailed employment estimates
are required in most industrially diversified countries. It is
therefore recommended that in the more industrially diversified
countries quarterly or monthly employment series should be
prepared for each major industrial group in which as much as
5 per cent, of a country's total employment is found, as well as
for each industry for which such statistics may be requested
by an inter-Governmental organisation (I, 14).
IV.
TOTAL EMPLOYMENT
While a series for agricultural employment may illustrate
accurately the movement of agricultural employment, and while
a series for a group of non-agricultural industries may represent
truly the changes of employment in that group, neither series
will necessarily have the same fluctuations as total employment.
Agricultural employment may rise while non-agricultural
employment is declining, as happens during the summer months
in many countries. Manufacturing employment may taper
off during December, but trade employment will increase.
Countries which are rapidly becoming industrialised will have
a decline in handicrafts employment at the same time that
manufacturing employment, exclusive of handicrafts, is increasing. In industrially developed countries during the great
depression sharp cuts in manufacturing employment took place
at the same time that agricultural employment ceased its decline
and in some instances actually increased.
To sum up the effect of these related but often contrasting
movements, a series for total employment is requisite. Only
with such a series can a shift in employment that reduces manu-
36
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
facturing and raises agricultural employment by an equal amount
reveal itself as merely a shift and not as a decline. If manufacturing alone is covered, the change appears as a decline. If
agriculture alone is covered, the change appears as a rise. To
secure just appreciation of the trend of its employment, therefore, every country requires data on total employment. For
the more industrially diversified countries, such estimates
should be prepared at least quarterly (I, 13).
V. LABOUR FORCE
A labour force series measures the amount of currently
available manpower, including all persons in employment and
all persons who are unemployed and hence available for work.
Such a series may be developed as the result of efforts to calculate comprehensive employment estimates and to complete
unemployment estimates. Many countries will consider the
use of labour force sample surveys, because such surveys are
an excellent method for providing dependable and consistent
estimates of both total employment and total unemployment.
The chief uses of a labour force series all arise from the
fact that such a series provides a measure of the amount of
immediately available manpower. Such a measure is of prime
importance during wartime and during reconstruction for all
countries with general labour shortages. By the method of
sample surveys or by means of an adaptation of social insurance
data, where such data are sufficiently comprehensive, it is possible to establish the current manpower budget, contrasting
labour needs with labour supply.
No less important is the use of a labour force series for
forecasting available labour supply at future dates. Where
current estimates have been computed on a sufficiently accurate
and sufficiently detailed basis, it is possible to establish the
level of participation in the labour force that has been characteristic of different age-sex groups and to measure changes in
that level. Given such data, it is possible to forecast the future
levels of available manpower with sufficient accuracy for many
purposes. Policy to adjust manpower needs to manpower
demands cannot be effectively worked out in the absence of
such information.
A labour force series makes it possible to measure at once
the effect of substantial changes in conditions and regulations
EMPLOYMENT SERIES: GENERAL PROBLEMS
37
which affect the numbers at work. Given such a series, it is
possible to measure the entrance of extra workers into the
labour market — for example, young workers and women
workers during the war years, or workers subject to regulation
in a number of European countries during the reconstruction
period. It is possible to study the effect of changes in the
school-leaving age or changes in the retirement age as set by
social insurance systems, for by providing data on the rate at
which new workers enter and older workers leave the labour
market, it is possible to secure a comprehensive view of the
chief changes in that market.
In view of these various purposes to which labour force data
can effectively be put, it is recommended that estimates of the
labour force should be prepared at least quarterly in the more
industrially diversified countries (I, 25). In the process of
European reconstruction, for example, the problems of shifting
labour from country to country are so great that estimates of
the actual labour force in the major countries are essential
merely to measure the dimensions of the problem and for use
in evaluating the various immigration, labour mobilisation and
related measures directed toward re-establishing the economy
of Europe. 1 The uses of labour force data, therefore,
may extend considerably beyond the manifold uses which
each country has for estimates of its own labour force.
While the extent of detail which is obtained on the labour
force will depend on national needs and on the relative cost of
securing each additional classification, full use of the potentialities of labour force data will not be made until separate
series are available for the major geographical or administrative
regions, the chief centres of population, each sex, single years
of age for juveniles and 10-year age groups for adults (I, 26).
VI.
EMPLOYMENT BY GEOGRAPHIC REGION
The necessity for separate employment series, not merely for
the nation as a whole but for the major economic and administrative regions, arises because economic development may
proceed at different rates in different regions. While employment is rising by leaps and bounds in one part of a country, it
1
Cf. the discussion on manpower in the Preliminary Report of the
Temporary Sub-Commission on Economic Reconstruction of Devastated
Areas (United Nations, Economic and Social Council, September 1946).
38
EMPLOYMENT, U N E M P L O Y M E N T A N D LABOUR FORCE
may be increasing at a gentler rate in another region, or even
declining. It is not to be supposed that contrasting employment changes between regions occur only in the large countries.
Table I, for example, presents data on employment in the various
Swiss cantons over the war years. While some cantons were
TABLE I .
MANUFACTURING EMPLOYMENT IN SWITZERLAND
1939-1943 »
Employment
Canton
August 3939
Zurich
Berne
Lucerne
Uri
Schwyz
Unterwald-le-Bas .
Unterwald-le-Haut
Glaris
Zoug
Fribourg
. . . .
Soleure
Bale-Ville
. . . .
Bàie-Campagne . .
Schaffhouse . . .
Appenzell Rh.-Ext.
Appenzell Rh.-Int.
St-Gall
Grisons
Argovie
Thurgovie
. . . .
Tessin
Vaud
Valais
Neuchâtel
. . . .
Genève
Total
1
. . . .
Sept. 1943
Per cent.
change in
employment :
1939-1943
73,300
54,288
9,430
1,648
3,744
427
558
6,142
3,888
4,100
28,822
16,885
10,382
9,580
3,533
187
28,077
2,847
36,494
16,803
9,131
15,151
5,674
13,503
13,330
7,529
18,748
16,950
9
24
37
. 89
18
135
13
3
—1
24
17
10
8
4
—8
2
7
32
8
6
14
29
33
39
27
367,924
425,972
16
80,099
67,228
12,897
3,119
4,425
1,005
633
6,300
3,861
5,099
33,842
18,609
11,217
10,005
3,242
190
30,021
3,766
39,509
17,802
10,380
19,496
La Vie Economique (Berne), April 1945. The data artì those of the factory census.
experiencing employment gains of 10, of 39 or even of 135 per
cent., others were actually subject to declines. Table II
demonstrates that a similar pattern existed for Sweden, with
differences not only between districts but between the rural
and urban sections of each district. For small countries as
well as large, therefore, it is evident that employment changes
may vary in degree and even in direction from region to region
within the same country.
EMPLOYMENT S E R I E S : GENERAL PROBLEMS
39
Given these contrasting employment movements in different
parts of a country, separate employment series can be of considerable value. They will, of course, constitute basic data for
orienting national policy in respect of industrialisation or
reducing unemployment. Local employment data possess
TABLE I I .
INDUSTRIAL
EMPLOYMENT
Employment
District
1939
Stockholm city . . .
Stockholm
province
Uppsala
„
Södermanlands
„
Ostergötlands
„
Jönköpings
„
Kronobergs
„
Kalmar
,,
Gotlands
„
Blekinge
„
Kristianstads
„
Malmöhus
„
Hallands
„
Göteborgs
o. Bonus
„
Alvsborgs
„
Skaraborgs
„
Värmlands
,,
Orebro
„
Västmanlands
„
Kopparbergs
„
Gävleborgs
„
Västernorrlands „
Jämtlands
„
Västerbottens
,,
Norrbottens
,,
Sweden
1
1942
IN S W E D E N 1 9 3 9 - 1 9 4 2
1
Per cent, change in employment
1939-1942
AU Sweden
Country
City
56,583
21,399
9,815
21,509
32,137
25,414
10,608
16,464
2,145
11,237
15,341
55,649
9,393
57,643
22,854
10,182
21,848
34,476
27,866
13,516
16,702
1,841
11,277
18,556
55,767
9,645
1.9
6.8
3.7
1.6
7.3
9.6
27.4
1.4
-14.2
0.4
21.0
0.2
2.7
3.8
0.9
- 9.5
4.8
14.9
30.5
- 1.1
-19.8
-10.5
21.4
2.4
8.8
1.9
12.0
6.5
7.1
9.5
4.3
5.2
5.5
- 2.7
10.1
19.5
- 0.4
0.1
55,652
42,889
15,421
21,956
30,769
21,420
25,422
25,307
17,324
3,100
7,585
8,983
49,655
44,696
17,813
21,839
31,688
23,717
26,436
25,939
16,150
3,108
8,593
8,916
-10.8
4.2
15.5
- 0.5
3.0
10.7
4.0
2.7
- 6.8
0.3
13.3
- 0.7
-16.1
6.4
17.3
- 2.7
-35.0
9.9
3.9
1.9
- 7.3
- 8.9
15.0
- 7.7
- 9.7
'2.1
13.5
4.2
88.2
11.8
4.2
4.5
- 3.5
27.5
9.1
35.5
. . . . 563,522
580,773
3.1
0.8
5.0
Statisíisk Àrsbok (Stockholm), 1945, p. 119.
obvious value for trade union officials in forming their evaluation of the state of the labour market and in adapting their
collective bargaining procedure to changing local conditions.
Such data are also required by business men, who must measure
the level of activity in their industry and must forecast those
changes in purchasing power that current changes in employment foreshadow. It is therefore recommended that, as
resources and facilities permit, the chief national employment
series — whether for total employment or for employment in
40
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
major industry groups — should be accompanied by similar
series for the major economic and/or administrative regions
of the country, and for the chief centres of population (1,17,18).
VII.
EMPLOYMENT BY AGE AND SEX
Data by sex will alone measure the increase in the proportion
of women workers to the total of those employed — whether
because of labour shortage, as in the war years, or because
women accept lower wage rates, as in periods of depression.
TABLE III. NON-AGRICULTURAL EMPLOYMENT IN THE
UNITED STATES, AUGUST 1945 - OCTOBER 1946, BY SEX 1
Aggregates
(in thousands)
Date
August 1945
September .
October . .
November. .
December
J a n u a r y 1946
February . .
March . . .
April . . . .
May
. . . .
June . . . .
July
. . . .
August . . .
September .
October . .
1
U.S.
Indices
(August 1945-100.0)
Total
Male
Female
Total
Male
Female
44,470
42,450
42,770
43,310
44,170
44,660
44,700
45,370
46,360
46,440
46,760
48,190
48,830
48,630
48,840
27,700
26,660
27,060
27,750
28,660
29,910
30,140
30,750
31,590
31,960
32,450
33,460
33,940
33,810
33,930
16,770
15,790
15,710
15,560
15,510
14,750
14,560
14,620
14,770
14,480
14,310
14,730
14,890
14,820
14,910
100.0
95.5
96.2
97.4
99.3
100.4
100.5
102.0
104.2
104.4
105.1
108.4
109.8
109.4
109.8
100.0
96.2
97.7
100.2
103.5
108.0
108.8
111.0
114.0
115.4
117.1
120.8
122.5
122.1
122.5
100.0
94.2
93.7
92.8
92.5
88.0
86.8
87.2
88.1
86.3
85.3
87.8
88.8
88.4
88.9
BUREAU OF THE CENSUS:
Monthly
Report on the Labor Force (Washington),
November 1946.
Data by sex, with supplementary information at intervals on
marital status, will throw light on the extent to which women
with and without home obligations are entering the labour
market (I, 19).
In table III, data on non-farm employment by sex are
presented separately for males and females in the United States.
If attention were paid only to the overall trend in employment,
which rose substantially in the year following the end of the war
with Japan, the fact that there was a substantial decline in
female employment during the same period would be overlooked.
EMPLOYMENT S E R I E S : GENERAL PROBLEMS
41
It is of considerable importance to know that returning veterans
found employment partly because of the rise in employment
possibilities, partly because they replaced women in the labour
force. (The concomitant information on the labour force is
a necessary supplement, since this shows that a substitution of
men for women took place within the civilian labour force, the
women not becoming unemployed but leaving the labour force
altogether.)
In order to assess the changing role of women in the labour
market, current information on employment by sex is indispensable and should be provided at intervals as resources and
facilities may permit (I, 17, 18).
A similar recommendation applies to estimates of employment by age, inasmuch as it is urgently necessary to know the
degree to which young workers are not being hired, or older
workers not being retained, during depression years, and the
extent to which child labour is increasing or decreasing (1,17,18).
VIII.
EMPLOYMENT
BY OCCUPATIONS AND
HOURS
WORKED
Nations with acute shortages of labour will find data on
employment by occupations and by hours worked of particular
value, as the increasing number of such series has demonstrated.
Analysis of labour shortages and bottlenecks, measures for
retraining and plans for national economic development rest
on a knowledge of the actual distribution of workers by occupation group. Similarly, information is required on the number
of persons in the separate hours-worked categories as the basis
(1) for determining the extent of short work weeks; (2) for
measuring the extent of increasing part-time work by housewives, children and other supplementary wage earners; and
(3) for assessing the effectiveness of measures which may be
taken to lengthen establishment work weeks. 1
In periods of unemployment such data serve the analogous
purpose of measuring the rise of short time. For countries
faced with such problems it would seem highly desirable to
provide periodic data on employment by occupations and
1
The use of a distribution of persons into the separate hours-worked
categories as the basis for measuring under-employment and short time
has been emphasised above in Chapter II, section II.
42
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
likewise by hours worked. It is therefore recommended that,
as resources and facilities permit, estimates of employment
should be made available at intervals for the principal occupation groups and for groups classified according to the number
of hours worked per week (I, 19).
IX.
PARTIAL EMPLOYMENT
In close connection with the foregoing topic, partial or
short-time employment should be considered. This is also
termed partial unemployment, its description as unemployment being derived usually from unemployment insurance provisions which allow the payment of benefits for "partial unemployment " under certain conditions. The problem of partial
employment, or short time, therefore, is part of the larger
problem of under-employment. It is therefore recommended
that a distribution of persons employed by number of hours
worked should be made in order to provide data to show changes
in short time or partial employment, by reference to the changing proportion of workers falling in the smaller hours-worked
groups. A further step in the measurement of under-employment as a whole would be to ask workers, irrespective of the
number of hours currently worked, whether they desired more
hours of work on the same job at the same rate of pay. It
is here recommended that, as resources and facilities permit,
series on partial unemployment be replaced by figures showing
the distribution of employed persons according to the number of
hours worked in the period covered (I, 45). Periodic studies
should be made of under-employment, as distinguished from
total employment. In particular, consideration should be
given to the possibility of obtaining, in the case of industries
other than agriculture, information on the number of persons
on short time who are seeking full-time employment (I, 46).
X.
SEASONALLY ADJUSTED SERIES
The marked seasonal variations of employment and the
labour force make the interpretation of month - to - month
changes difficult. A seasonal rise in employment may be taken
as evidence that economic recovery is proceeding apace, or
a seasonal decline that the change foreshadows a depression.
Although it is no easy matter for the statistical agency
E M P L O Y M E N T : COVERAGE
43
to establish seasonal adjustment factors, and though these
factors will necessarily have to be revised from time to time,
the use of seasonally adjusted series has distinct value. 1
It is therefore recommended that, as resources and facilities
permit, seasonally adjusted indices of the chief employment
and labour force series should be presented together with the
unadjusted series (I, 33).
VÏ. EMPLOYMENT: COVERAGE
I. STATUS
Who belongs in the statistics of employment ? Are only
wage-earners to be covered or should salaried workers also be
included ? Should an employment series include only those
who are hired or should it widen its coverage to include working
proprietors and own account workers? If wage-earners are
to be included, does this mean wage-earners attached to establishments or those actually at work ?
1. National Practice
As might be expected, there is no agreed set of answers to
these questions. Practice varies sharply from one country to
another. Perhaps the greatest uniformity appears in the coverage of workers according to status groups. The most common
employment series are those which include all employees
(wage-earning plus salaried workers) or those which are restricted mainly to wage-earners. In a few instances, however,
series are available for all groups in the labour force. The series
relating to employees will be considered first.
The British establishment reporting form asks for two
separate totals. One is for "administrative, technical and
clerical staff, including managers, superintendents and works
foremen; research, experimental, development, technical and
design staff". The second is for operatives, including "all
classes of employees on the payroll other than those to be
i n c l u d e d . . . under administrative, etc., staff". The current
monthly estimates based on these reports are intended to
1
For an excellent example of the analysis of changing seasonal
factors see " Seasonal Variations in the Automobile Industry ", in
Canada, The Labour Gazette (Ottawa), March 1947.
i
44
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
measure the movement of employment of wage-earning plus
salaried workers, excluding only a limited number of higher
paid salaried employees — those earning £420 or more.
The United States Bureau of Labor Statistics series for
production workers in manufacturing defines "production and
related workers " a s :
Working foremen and all non-supervisory workers (including lead man
and trainees) engaged in fabricating, processing, assembling, inspection, receiving, storage, handling, packing, warehousing, shipping,
maintenance, repair, janitorial, watchman services, product development, auxiliary production for plant's own use (e.^. power plant) and
record-keeping and other services closely associated with the above
production operations. Exclude supervisory employees (above the
working foreman level) and their clerical staffs and other groups of
. employees named under instruction 5 (see below).
The Bureau of Labor Statistics series for employees in the
major industrial groups includes not only the employees
enumerated above, but also, according to instruction 5, workers
in:
Executive, purchasing, finance, accounting, legal, personnel (including
cafeterias, medical, etc.), professional and technical activities, sales,
sales-delivery, advertising, credit, collection, and installation and
servicing of own products, routine office functions, factory supervision
(above the working foremen level), and other workers not included as
production workers. Include also employees on your payroll engaged
in new construction and major additions or alterations to the plant
who are utilised as a separate work force (force-account construction
workers).
The semi-annual estimates of employment begun by New
Zealand in April of 1946 are similarly comprehensive. The
employer is directed to include:
all part-time and "full-time employees in receipt of wages, salary or
other remuneration " including : persons working for commission
("other than agents not subject to your direct control"); skilled
manual; apprentices; semi-skilled manual; unskilled manual; domestic
and allied; others, including clerical, shop assistants, commercial professional, scientific, executives, etc.
The Australian payroll tax return simply asks for the number
of "all employees whether paid weekly, fortnightly, monthly,
quarterly, annually or at irregular intervals ".
The French quarterly reporting form asks for "Effectifs
Maux (ouvriers, employés, cadres)" i.e. wage-earning and salaried
workers and higher staff combined as a group.
On the other hand, the Swedish series is based on the
number employed, excluding engineers, non-working foremen,
E M P L O Y M E N T : COVERAGE
45
accounting and clerical personnel, and warehousing and related
employees. The Norwegian form in use until June 1946 asked
simply for the number of workers and working foremen. In
Denmark, the request is made for total man-days worked by
production workers, with a separate estimate for non-production
workers, although the latter estimate is apparently not utilised
in the published series.
In Switzerland only production workers are to be included,
these being defined, using the same definition as the accident
statistics reports, as including:
operatives, supervisory personnel, those engaged in testing, and
janitorial services: technicians overseeing production, overseers of work
processes, shophands, foremen, work guides, employees used in testing,
chief of installation gang, controllers, chief machinists, store-keepers
and dispatchers, packers, janitors and charwomen, even where engaged
in offices and stores.
Up to this point, therefore, two main types of series have
been discussed : (a) British, United States, Canadian, Australian,
French and New Zealand series relating to wage-earners plus
salaried workers, and (b) the Danish, Norwegian, Swedish and
Swiss series relating to wage-earners. A third type of series
measures the total number of persons in employment, including
not only wage-earning and salaried workers, but also employers
and self-employed workers and unpaid family workers.
At the present time three countries issue series of the third
type: the United States, Canada and Great Britain. The first
two countries obtain, in addition to their series for employees
(and for wage-earners in manufacturing), a series for all persons
in employment, based on current surveys of the labour force, as
already noted, while Great Britain supplements its current
series for insured persons by a variety of estimates in order to
prepare monthly estimates of total employment, excluding
certain categories which are for the most part numerically
unimportant. 1
2. General Problems
These, then, are some of the differing answers to the basic
question — who should be included in the statistics of employment? Which of these is the most satisfactory answer?
1
Totals for Great Britain exclude certain workers because of their
age and others because their incomes exceed a certain maximum figure;
they exclude also a substantial number of domestic servants.
46
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
The reply is simply that all of these solutions are about
equally satisfactory in practice. At the same time, the widest,
most comprehensive coverage is preferable to any more restricted coverage. The issue may be considered as it relates to
manufacturing. (For most of the countries studied, and for
the considerable number not studied, a series for employment
in manufacturing is the only available and significant series on
employment.) It is of great importance for all countries.
In practice it makes little difference whether the manufacturing series includes only employees or whether it includes
proprietors and unpaid family workers as well, because of
the extremely small numbers in the latter group compared
to the total number of wage-earners. 1 Since the number of
salaried workers also is relatively small, the movement of
the employment series will be only slightly moderated if they
are included, or will be made somewhat more sensitive if they
are excluded.
Is the lack of preference between alternative methods
equally applicable to industries other than manufacturing?
Owing to the growth of chain stores and integrated units the
importance of self-employment in trade has been steadily
diminishing in certain countries and the proportion of employees
to total trade employment has therefore been increasing. The
economic and even the social status of the small proprietor in
trade or service is not likely to be more than one stage above
that of many employees in those industries. Furthermore, the
steady turnover of businesses in these lines contributes a
continuous stream of ex-proprietors to the ranks of wage-earning
workers and the unemployed. The inclusion of these small
proprietors would therefore seem to be appropriate.
What difference would it make if the employment series
covered proprietors and self-employed as well as wage-earning
and salaried workers? Table IV attempts to answer that
question from experience. The decennial changes indicated
for Denmark and the United States are essentially the same
whether or not proprietors are included. The difference for
Canada is somewhat more marked, while the differences in the
United States over the brief period, 1930-33, are noteworthy —
1
This is not the case where handicraft industries predominate, but
India, China and other countries which are becoming industrialised
will find it desirable to enumerate handicraft establishments independently of an employment series.
EMPLOYMENT:
47
COVERAGE
though perhaps not great enough to require a major revision
in coverage.
In general, each country must decide, on the basis of its own
experience and needs, whether to utilise the wider concept of
employment. It is here recommended that countries should
work toward the most comprehensive coverage possible for
any given industry, presenting, as resources and facilities
permit, periodic estimates separately for all significant status
groups in the industries surveyed (I, 19).
TABLE
IV.
EMPLOYMENT
IN
COMMERCE
FOR
SELECTED
COUNTRIES
Wage-earning and
salaried workers
Total
employment
. . . .
129,293
187,481
+45.0
195,783
281,904
+44.0
. . . .
253,188
335,798
+32.6
359,862
453,310
+26.0
. . . .
. . . .
5,932
4,916
6,906
—17.1
+16.4
Country and date
Denmark 1
1930
1940
per cent, change 1930-40
!
2
Canada
1931
1941
per cent, change 1931-41
(thou sands)
8
United States
1930
1933
!
1940
per cent, change 1930-33
1930-40
7,628
6,575
8,842
—13.8
+15.9
1
Figures relate t o t h e gainfully occupied. D a t a taken from I.L.O. Year Book, 1942
a n d !1943-44.
Estimates compare (a) employed wage-earning and salaried workers, a n d (6) employed
wage-eaining a n d salaried workers plus non-wage-earners gainfully occupied. D a t a t a k e n
from Census of Population, 1941: Occupational Trends in Canada, 1901-1941, N o . 0-6.
3
Figures relate t o persons actually employed. D a t a t a k e n from U . S . B U R E A U O F
LABOR STATISTICS: Employment
in Non-Agricultural
Establishments
in the United States,
1929-43, a n d Stanley LEBERGOTT: Estimates of the Non-Agricultural
Self-Employed,
19291940 (Bureau of L a b o r Statistics, 1945).
II. CONDITION
Given the groups to be included, under what conditions are
they to be covered ? Are those who are sick, on vacation,
on leave for purposes of military service or because of lack of
work, or absent for still other reasons, to be included ? The
solutions adopted in the various countries differ, although the
differences may often be of a formal rather than of a substantive
character.
48
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
1. National Practice
France simply asks for "e//ecfi/s totaux", or the numbers
employed, leaving to the option of the employer who is to be
included. The result is probably similar to that in Australia
(where the "number of employees on payroll " is specified), in
Switzerland (which asks for number occupied on the payroll
date) or in Denmark (which specifies the number of man-hours;
since the interpretation is left to the employer, it may well vary
from firm to firm). As the Australian Government has indicated, "persons on sick or recreation leave may or may not be
included in the returns depending upon the individual employer's
interpretation . . . (while) persons on strike near the end of a
month often are not counted as employed by their employers".
Other countries are more specific in their directives. Sweden
specifically requests separate estimates for the number of workers
(a) at work on pay-day; (b) on payroll but not employed because
of lack of work; (c) on vacation; (d) absent for other reasons
(sickness, leave for military service, etc.). The Swedish employment series includes those at work as well as those on vacation
and those absent for other reasons, including sickness and leave
for military service. New Zealand requires the inclusion of
"all persons on the payroll for that day, whether absent on paid
leave or on compensation or sick leave. Do noi include persons
who left your employment before that payday ".
Great Britain adds still more qualifications, emphasising
that employment estimates "should relate to persons on the
payroll, including those temporarily absent from work through
sickness, holidays or other causes.
Persons on the payroll
who were temporarily stood off for the whole or part of the week
should be included... but persons discharged, or who left during
the pay-week, should be excluded ". The United States is
equally specific, requesting data for "all full and part-time
production workers on your payroll who worked or received
pay for any part of the period reported. Include pensioners,
unpaid family workers, members of the armed forces and others
carried. on your active rolls who did not work or receive pay
during the period."
2. General Problems
In general, it will be seen that the countries fall into two
groups, one leaving the inclusiveness of the term "number
employed " to the decision of each individual employer, the
E M P L O Y M E N T : COVERAGE
49
other defining the exclusions and inclusions in some detail. It
would seem highly desirable to specify clearly the groups which
it is desired to include. Discussions with representative
employers and employers' associations might throw light upon
which set of qualifications is most practical, and the precise
problems involved in the specific decisions. Given this information, the statistical agency can then proceed to establish the
desirable procedures.
It is therefore recommended that on both the reporting form
and in attached directions there should be provided a definite
list of the major groups to be included and the major groups to
be excluded, specifying at the least the treatment of those (a) on
paid vacation; (b) on paid holidays; (c) on paid sick leave;
(d) on unpaid sick, holiday or vacation leave; (e) on pension;
(/) on temporary furlough because of lack of work; (g) on indefinite furlough because of lack of work; (h) on strike; (i) on
military service.
Granted that specific decisions should be made for the major
groups, what should those decisions be ? For the extremes the
case is clear; those on strike, those on indefinite furlough for
more than a payroll period, those on pension, are all to be
excluded. The basis for exclusions is simply that they are not
at work or remunerated. 1 It would seem similarly correct
to exclude those on military service and on extended leave
of any sort.
But what of those persons on sick or vacation leave ? The
procedure followed in the United States is to include these
workers in employment totals if they are on paid leave, but not
if the leave is unpaid. New Zealand pursues a similar course.
On the other hand, in the Danish figures on man-hours worked
all workers on sick or vacation leave are excluded. Which
procedure is preferable ?
In reaching a decision on this point, the effect of these
different definitions on economic and social analyses should be
considered. On the one hand, to include those on paid vacations as "employed " seems reasonable from the point of view
of their continuing to receive pay. There is no cutting down
of the flow of income to these workers.
1
Persons on indefinite furlough cannot be considered as attached to
the establishment in any real sense. Those retired on pensions must
likewise be excluded, since their connection with the current productive
activity of the establishment is tenuous.
50
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
For estimates of average weekly earnings the base should
be the number of persons actually at work or receiving pay.
During vacation periods payrolls may tend to drop in many
industries, but this drop is most reasonably evidenced by the
movement of payrolls and not by that of average earnings. To
the extent that vacations are unpaid, estimates of payrolls
will properly indicate a decline. But estimates of average
earnings — applicable to persons who remain at work or who
receive pay — should not be so affected.
On the other hand, calculations of productivity per manhour should be based on hours of actual work rather than on
hours paid for. If workers are included in the employment total
when their vacations are paid but not when their vacations
are unpaid, then productivity per man-hour, other things remaining the same, would tend to be lower in plants with paid vacations than in plants with unpaid vacations. And as the number
of plants giving paid vacations increased in proportion to the
total number of plants, productivity would tend to decrease.
Such anomalies could be avoided if productivity estimates were
computed on the basis of the number actually at work or
hours of actual work.
For many purposes, however, it is of interest to know the
number of persons attached to the industry in the sense of
being employed. Normally, a person is regarded as employed
in an industry if he has a job, even though at any given date he
may be sick, or recovering from a minor injury, or on vacation,
whether paid or not ; in such cases he regards himself as having
a job, and is not looking for work elsewhere. In this sense, therefore, and with this practical interpretation of being employed,
it would appear best for most purposes to adhere to the definition
that corresponds to "employment" rather than with the
narrower one that corresponds to " a t work ". With this
definition, this series indicates the trends of employment,
showing the changes that are taking place in the personnel
attached to or employed in the different industries.
It is recommended that the statistics of employment in
establishments should include all persons attached to the establishment, regardless of age, and regardless of whether or not
these persons are actually at work on the date to which the
statistics relate (I, 28).
For more distinct and specific purposes, however, and for
more accurate measurement of the actual man-days or man-
E M P L O Y M E N T : SAMPLE
51
hours of employment, data on actual time at work may be
required. There is need for both types of data. As resources
and facilities permit, it is recommended that periodic studies of
employment as reported by establishments should be made to
determine the number of persons who were at work and who
were not at work on the date to which the studies relate (I, 34).
III.
ÍAGE
The coverage of employment series with respect to age is
almost identical in all countries surveyed. Where the series
are tied to population census or industrial census benchmarks
they are usually all inclusive. However, those which are tied
to social insurance data may be less comprehensive. In the
United Kingdom, for example, the benchmarks are those for
males aged 14 to 65 and females aged 14 to 60 — although the
series used for extrapolation include persons of all ages. The
defects in the use of an age limitation are suggested in a study
of the British lace industry in which it is pointed out that the
decline in the number of workers in the industry between 1935
and 1939 as indicated in the employment series was not consistent with the change in production which occurred in the
same period. "The increasing age of the workers probably
accounts for a good deal of the apparent decline in the numbers
employed; workers over 64 are not included in the Ministry of
Labour figures. " x Because of the need to relate employment
to production, labour turnover, earnings and similar data it is
highly desirable that no upper age limitation should be made.
It is therefore recommended that employment data should
include all persons in employment regardless of their age, and
that both benchmark and movement figures should be consistent
with this recommendation (I, 28).
VII.
EMPLOYMENT: SAMPLE
Statistics of employment on a current basis could conceivably
be secured by a complete canvass of establishments or households. Such canvasses are in fact required periodically in
1
H. A. SILVERMAN, ed. : Studies in Industrial Organisation, Nuffield
College Studies in Reconstruction (1946), p. 93.
52
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
order to provide complete benchmark data. 1 But conducting
such surveys on a monthly or quarterly basis would be an enormously expensive and difficult task. As a result, all countries
which collect employment statistics have conie to rely on
sample surveys for the derivation of current data on employment. There are two main systems for collecting such data:
establishment sample statistics and the labour force survey.
Each of these will be discussed in turn.
I. ESTABLISHMENT SAMPLE STATISTICS
In the establishment sample system the basic assumption
is that changes in employment in a sample number of firms will
truly reflect changes in employment in all firms. In the United
States, for example, the changes of employment in 34,000
manufacturing establishments are taken to stand for the changes
of employment in 184,000 manufacturing establishments.
Changes of employment in 49,000 trade establishments stand
for changes in 1,770,000 trade establishments.
In order that the employment figure estimated from such
a sample may be the same as that which would be derived from
a complete enumeration, it is necessary to establish accurate
procedures under two main headings. The first relates to
estimates of change. The second relates to the complete
benchmark figures to which the estimates of change are applied
in order to yield up-to-date estimates of employment in the
industry or group of industries surveyed.
1. Estimates of Employment Change
What constitutes a proper establishment sample for securing
estimates of changes in employment ? The sample should have
the same general characteristics as the whole universe of
establishments it represents, and have them in their due proportions. What does such a definition mean in practice ?
In the first place, it means that the sample is effective
not in proportion to its size but in proportion to its balance
and representativeness. Obviously where a sample covers
95 per cent, of all employment in an industry that industry will
1
Benchmark data are figures on total employment in the group
under consideration, at a date—usually a census date—when a fairly
complete enumeration was made.
E M P L O Y M E N T : SAMPLE
53
be represented by full and fair statistics. Just as clearly, it
is doubtful whether an industry with a mere 5 per cent, of its
employment included will be satisfactorily represented. But
what of the more typical case — the area between these
extremes ? Is 40 per cent, coverage adequate ? Is 50, 60, 65 ?
Is it possible to state a satisfactory minimum percentage
coverage ?
The criteria for an adequate establishment sample must be
sought not in terms of the percentage of employment covered,
but with reference to the character, balance and distribution of
the sample. If the sample has the same characteristics as the
universe it represents, and in their due proportions, it will be
a satisfactory sample — provided, of course, the numbers are
sufficiently large in each sub-group so that a satisfactory stability
of reporting is secured.
The factors which determine variations in total employment
or industrial employment will differ enough from country to
country, so that no complete enumeration of such factors can
be made. For the purpose of developing an establishment
sample, each country must make a systematic analysis for itself
of the factors associated with variations of employment in its
different regions. (The period chosen for study must be sufficiently long to allow seasonal, cyclical and long-run changes
to manifest themselves.) Such an analysis might show that
the variations in employment in a given region are related to
(a) its industrial composition, (ft) the size distribution of its
firms, and (c) the distribution of firms as between urban and
rural territory. In such a case the national sample would be
properly stratified if it gave suitable representation to different
industries, to different sizes of firms and to firms located in
urban and rural areas. Should employment changes have
differed in different regions during the period studied — after
account had been taken of the three factors mentioned —
separate representation should be given to the chief regions also.
In practice most countries give separate representation to
the chief industries. Where employment in food manufacturing
is more stable than that in jewellery manufacturing, for example,
then each industry must be separately represented in a.sample
intended to measure the changes in manufacturing employment
as a whole. If the pattern of employment in slaughtering and
meat-packing differs from that in commercial bakeries, then
representation must be given to each group.
54
EMPLOYMENT,
UNEMPLOYMENT AND LABOUR
FORCE
Within any given industry group changes in employment
in large establishments will usually differ from those in small
establishments. Since in practice most countries have tended
to secure returns- from the larger establishments, because their
co-operation is more readily secured and the collection of data
simpler, this point has often received insufficient attention. But
the point is none the less vital for industries where great disparities in the size of firms exist and where large numbers of
persons are employed in both large and small firms. Thus, in
the construction industry employment patterns in large firms
are probably more stable than those in small firms, yet large
numbers of persons are employed in small firms and separate
representation must be given to such firms. The same consideration usually applies to trade, service and agriculture as well as
to construction. Even within the manufacturing industry, as
studies in France, Canada and the United States have shown,
employment trends in large firms are found to differ from those
in small firms.x In some industries changes in employment
are positively related, in others negatively related, to the size
of the firm. In setting up an establishment sample for any
given country, therefore, it is necessary t o determine whether
there is any relation between the size of firms and changes in
employment in the chief industrial groups and to set up the
sample accordingly.
Finally, attention must be given to proper representation
of the different geographical areas. The trend growth of
American manufacturing is greater in the South than in the
New England States. Employment in construction for public
utilities during recent years has been rising a t a greater rate in
Valais than the Swiss average, but at a lesser rate in NeucMtel.
A graphic demonstration of geographic differentials in changes
in employment within an industry is given by the figures in
table V for changes in employment in France from April 1939
to 1946. Thus, over the seven years, employment in food
manufacturing rose 14 per cent, in Poitiers but declined 29 per
1
'' L'Enquête sur l'activité économique et les salaires au 1 e r avril 1946 "
in Revue française du Travail, August - September 1946, p p . 494-495;
J. P E R L M A N : Hourly Earnings of Employees in Large and Small Enterprises, Temporary National Economic Committee Monograph No. 14
(1940), p . 82.
Employment,
Cf. also NATIONAL B U R E A U OF ECONOMIC R E S E A R C H :
Hours and Earnings in Prosperity and Depression, p p . 32-34
(New York, 1923); D O M I N I O N B U R E A U O F STATISTICS: An Estimate
Total Employment on October 1, 1943, in Industries
Monthly Survey of Employment (Ottawa, 1944).
Reporting
of
to the
TABLE V.
I N D I C E S OF EMPLOYMENT IN F R A N C E B Y I N D U S T R Y AND BY REGION IN APRIL
1946
(April 1939 = 100)
. * e
ig
CA
§
Industry
&
Food
Chemicals
Rubber
Paper
Printing
Textile
Apparel
Leather
Lumber
Metallurgy
Metal working
. . . .
Precious metals . . . .
Stone, clay, glass . . .
Quarries
Construction
Transport exe. S.N.C.F.
Commerce
Amusements
Finance
Liberal professions
. .
Total
. . .
TS
O
5
ï
a
g«
* .A
¿'•s
S3 0>
SP a
•-I
a
3
0
G
Ö
u
V
BU
C
<
8
3
'3
•a
o
a,
«
«
CD
EO
3
O
"3
o
3
S1
ao
S
«
en
u
a
o
li
oc
u
3
o
«I
o
la
au
B
4-»
3
to
H
86 87 78 88 85 101 85 90 91 84 114 85 90 89 97 96 93 88 71 88
100 115 103 92 104 103 85 80 76 92 108 81 — 109 91 106 95 104 74 95
103 69 84 — — — 100 91 — — —
93 — .— — 123 112 — — 102
75 76 85 65 56 64 64 72 95 85 93 109 81 86 92 93 92 91 71 79
93 89 79 61 76 —
89 90 97 99 75 97 78 108 99 102 145 97 97 92
80 72 76 77 73 78 77 81 79 81 90 94 81 72 60 81 74 96 68 78
81 83 82 70 89 73 72 77 77 74 64 85 98 87 84 83 93 72 89 81
91 72 107 84 97 97 106 90 91 89 87 91 84 76 88 98 76 110 70 91
110 130 114 103 100 101 109 111 118 121 119 105 122 102 108 114 119 91 89 110
97 — 100 80 69 96 92 83 — — — 199 128 — 106 92 — —
73 89
97 100 98 83 92 70 102 97 98 96 98 98 115 109 112 111 109 99 85 98
99
99 92 109 — — — — — — —. — — — — — 104 — —
112 84 94 74 89 80 88 70 94 114 104 87 103 116 111 104 81 101 77 98
79 — — — —
86 100
107 — 110 70 — — —
75 — —
97 — —
103 115 96 107 129 157 110 135 96 118 101 81 105 80 118 100 91 91 121 115
81 79
56 —
99 — —
96 81 91 — 109 75 89 82 96 88 92 — —
73 75 88 72 69 75 78 76 79 75 88 83 .— 81 69 85 95 100 79 77
101 104
90
100 113
103 113108 92 96 103 105 126 122 112 107 103 91 105 99 106 127 131 112 106
99 115 119 — — 102 81 102 — 124 — — — 116 100 — 114 — — 105
•
—
•
92 93 96 76 91 90 91 95 96 97 99 89 106 93 95 99 104 98 84
M
g
t-1
O
g
M
Z
H
en
>
•tí
F
m
56
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
cent, in Strasbourg. Such contrasts suggest the need for
including in the employment sample not merely every major
industry but also each section of the country in determining
the sample for each country.
To sum up, it is recommended that in setting up the sample
of establishment for employment reports, proper representation
should be given to major industry groups, geographical areas,
and size of establishment groups (I, 29 (1)).
2. Benchmark Data
Having established the constituent groups to be represented
in the sample it is necessary to determine the proportions in
which each group is to be represented. This is essentially a
problem of securing benchmark estimates of employment. On
the basis of these estimates of employment, the data on changes
in employment in particular sub-groups can then be combined
into an employment series. The methods used and the frequency with which data on employment are adjusted to benchmark estimates of employment will necessarily differ from
country to country.
As an example of the problems involved, the methods of
adjusting to industrial benchmarks used in a number of countries
may be discussed. Table VI gives the percentage of employTABLE
VI.
STATES, AND
TOTAL
ESTIMATED
IN REPORTING
EMPLOYMENT
Number of
employees in
reporting
establishments
Industry
Metal mining :
Quarrying and non-metallic mining
Crude petroleum producing . . . .
Street railways and buses
Wholesale trade
Retail trade
Power laundries
Cleaning and dyeing
Brokerage
IN
THE
UNITED
ESTABLISHMENTS, F E B R U A R Y
. . . .
10,159,000
63,700
239,000
77,400
35,600
35,900
365,096
194,148
153,998
294.341
915,998
139,337
72,996
15,813
14,323
136,742 '
372,299
1944
Estimated percentage of total
employees covered
by Bureau of Labor
Statistics sample
75
91
65
87
67
39
80
89
66
20
25
40
29
21
23
33
52
57
E M P L O Y M E N T : SAMPLE
ment in each of the major industry groups surveyed by the
United States Bureau of Labor Statistics which was included
in the establishment sample of February 1944. It can be
readily seen that the coverage varies so widely from industry to
industry that an adjustment to benchmark data is essential.
Otherwise, for example, manufacturing would be heavily overrepresented, while trade and finance would be grossly underrepresented.
In the years before the war, the Bureau of Labor Statistics
secured benchmark data for the constituent elements of its
manufacturing series from the biennial Census of Manufactures.
Benchmark data for the other industries covered are obtained
from the operations of the social insurance system. The computation of employment estimates based upon the establishment sample therefore reduces itself to two elements: first,
securing estimates of the percentage change in employment in
identical establishments; and secondly, applying these estimates
to the benchmark data to arrive at the series finally desired.
In marked contrast to this procedure is that of the Swedish
Social Board. The Social Board's percentage of coverage for
each industry is indicated in table VII. From that table it
TABLE V i i .
TOTAL
ESTIMATED EMPLOYMENT
IN REPORTING ESTABLISHMENTS,
Industry
Mining and metal .
E a r t h and stone .
Wood
Paper and printing
Food
Textile and apparel
Leather and rubber
Chemical
.
.
.
.
.
IN S W E D E N
AND
1944
N u m b e r of
wage-earners
(census)
Number of
wage-earners
in reporting
establishments
Percentage
of total employees
covered b y sample
234,496
49,356
60,477
55,187
53,017
83,741
25,147
22,893
163,055
23,955
29,547
36,171
27,846
67,177
18,882
11,749
70
48
49
66
52
80
75
51
i
can be readily seen that the variations are wide. In 1944 about
half the food industry employment was covered, about twothirds of paper and printing, and three-fourths of leather and
rubber industry employment. Within the earth and stone
group, the coverage of the eight constituent industries ranged
from 12 to 45, 56, 62 and 100 per cent. This variation, though
58
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
wide, is not unusual. But the Swedish figures are not tied to
any benchmark data. "The index of the statistics of employment is based exclusively upon statistics of establishments
which report in successive months. No correction of the chain
index number is made to agree with the Board of Trade industry
statistics or with figures derived from the number of workers
in the census or other source. "
For Canada the distribution of wage-earners in non-agricultural industries as indicated by the population census may
be compared with that indicated by the monthly survey of
employment sample. Because the monthly survey attempts to
depict the situation as reported by establishments ordinarily
employing 15 persons and over in nine non-agricultural industries
— excluding employment in most service industries — the differences between the two distributions, as indicated in table VIII,
are striking. Because the survey covers only a small segment of
TABLE
VIII.
DISTRIBUTION OF CANADIAN
BY INDUSTRY, J U N E 1 9 4 1
WAGE-EARNERS
l
Wage-earners: June 1941
Branches of economic activity
Number
Census
Logging
Per cent, distribution
Population census
(June 2)
Monthly survey
(June 1)
77,459
85,847
3.1
3.4
3.0
5.2
914,753
166,600
36.5
6.6
53.7
8.3
242,417
340,271
80,139
605,922
9.6
13.5
3.2
24.1
13.5
10.1
3.8
2.4
i
l
|
2,513,408
100.0
100.0
|
Manufacturing and
Construction
. . . .
Transportation and
communication . .
Trade
Finance
Total non-agricultural
1
Data provided by the Dominion Bureau oí Statistics.
the service group, while the census covers the entire group,
manufacturing has 56 per cent, of the total weight in the survey
as compared to only 34 per cent, for the census non-agricultural
EMPLOYMENT:
SAMPLE
59
group. For service the weights were 2.5 and 25.8 respectively.
(Even if domestic servants were excluded, the disparity would
still be great, since they constituted only 6 per cent, of the nonfarm total.)
These contrasts between the actual coverage of the monthly
survey and the distribution of all non-agricultural wage-earners
are significant, because the movement of employment in large
firms in the 9 industries is frequently taken as representative
of the movement of employment in all firms in these industries
and of all firms in non-agricultural industries.
What effect does this difference in industrial distribution
have on the trend in employment ?
A partial statement of this effect is given in a special Dominion Bureau of Statistics study, An Estimate of Total Employment on October 1, 1943 in Industries Reporting to the Monthly
Survey of Employment (1944). This study is restricted to that
part of service covered by the monthly survey, and is benchmarked to population census results; the change in wageearning employment in Canada from 2 June 1941 to 1 October
1943 is estimated at 19 per cent. But the change actually
shown by the establishment reports was 22 per cent. The
difference between a rise of 19 percentage points and one
of 22 points is 15 per cent. — a not insubstantial figure to
result merely from differences in weighting. Inasmuch as
the Estimate of Total Employment allocates hand trades to
manufacturing and makes no allowance for the bulk of
persons employed in service — where the rise in employment was probably less than average — the difference, was
actually greater.
A serious problem in the adjustment to benchmark data is
the relative infrequency with which they become available. One
reason why the very carefully constructed data of the Canadian
monthly survey are not related to other benchmark data is,
according to the Canadian report, precisely this infrequency.
"The decennial census of population collects statistics from the
individual showing the industry to which he belongs, but. the
census enumerations are far apart, while the data therefrom
can obviously relate only to a point of time. In a country
where the seasonal fluctuations in industry are widespread and
marked and where there are frequent transfers from the ' ownaccount ' or employer class to the wage-earner class, a more
continuous experience would be required to determine the
B
60
EMPLOYMENT, UNEMPLOYMENT AND LABOUR
FORCE
distribution which should be used as the basis for establishing
weights. "
This difficulty, however, should not be over-emphasised. In
most countries social insurance data can be utilised for more
frequent benchmarks, these being available annually or, in
some countries, quarterly. But even the use of decennial
population census results should not be relinquished without
careful analysis. For every employment series must use benchmarks — either explicit or implicit.
This is evident from the fact that although neither the
Swedish estimates nor the non-manufacturing estimates for
Canada are benchmarked to census data, they are none the less
linked to a distribution at a basic date. That distribution is,
however, not a comprehensive one obtained from a census
or social security enumeration, but merely the distribution
of the sample at the basic date. The question, therefore,
is not whether to use benchmarks, but which benchmarks to
use.
Once the question is put in that form the answer becomes
clear. In general, any systematic evaluation — whether of
the social insurance system or a census —- is preferable for the
purpose of establishing benchmarks to the use of the actual
distribution of employment in the sample at the basic period.
The distribution of employment in the sample represents the
final result of a combination of factors, including cost considerations and administrative requirements of efficiency and simplicity. Such considerations are almost certain to produce a
distribution of employment different from that shown by a
systematic enumeration.
For those countries which do utilise benchmark data such
materials are used with varying frequency. In the United
States quarterly benchmark data are provided, after some time
lag, by the operation of the social insurance system. These
data are used as soon as available. For Canada, as already
noted, general benchmarks are not used, but the data provided
by the annual census of manufactures are used for manufacturing. In Denmark, as the Danish report notes, "the monthly
index number of employment is revised once a year on the basis
of the annual manufacturing census." In Great Britain an
annual revision is made in accordance with the number of
employed as indicated by the July exchange of unemployment
books.
E M P L O Y M E N T : SAMPLE
61
To sum up, it may be recommended that employment series
should be adjusted to comprehensive benchmark data derived
from census enumerations or social insurance records as such
data become available (I, 27).
II.
LABOUR FORCE SURVEYS
The basic sample design for current United States population
census surveys of labour force activity was developed by Stevens
and Stock for the United States Work Projects Administration,
drawing on basic work by Fisher, Neyman and Mahalanobis.
This was further extended by Hansen and Hurwitz after the
series was taken over by the United States Bureau of the Census
in 1942. Essentially the same procedures were adopted by
Canada when that country established current surveys of labour
force participation in November 1945.
The United States sample discussed below is one of areas,
not of population quotas. 1 Because of administrative and
financial limitations the basic units for sampling were restricted
to 68 areas. The 3,000 counties in the United States were
grouped into 68 strata, one unit was chosen from each stratum,
and household enquiries were then instituted within these units.
The choice of the county as the original unit is based on
several considerations. Most counties are fairly small units, and
can readily be enumerated from a single local survey office.
Counties are well-established administrative units, located in
both urban and rural regions, and a good deal of published information is available on them. Each county usually includes
persons with a wide range of occupations, and each generally
includes urban as well as rural farm and rural non-farm territory.
Furthermore, wide variations in size and source of income are
present, as well as in racial composition and nationality origin.
1
The objections to quota sampling as an alternative to area sampling
are discussed in " On Sampling in Market Surveys", by Philip M. HAUSER
and Morris H. HANSEN, in The Journal of Marketing (July 1944).
These objections centre on the fact that the relative size of each
quota of persons to be included is necessarily based on sources which may
have become unrepresentative in the course of time owing to heavy
geographic shifts in population, national income, etc. Secondly, the
freedom given to the enumerator to use any persons who satisfy the
specified minimum characteristics has frequently resulted in enumerators
choosing people easy to enumerate. Hence the implicit assumption
that given certain characteristics we have a true sampling because all
other things remain equal, is nullified, because all other things do not
remain equal.
62
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
Two complementary considerations, however, made it
advisable to regroup the counties before setting up the primary
sampling units. The initial reason for making such a regrouping was to produce a somewhat wider variety of occupations
and incomes than existed in certain counties. By making
the basic units more heterogeneous this adjustment decreased
the sampling variance. Thus it was found that for estimating
the number of male workers in the United States at a given date
the use of enlarged units cut sampling errors by 48 per cent.,
while a reduction of 26 per cent, was achieved for the estimates
of female workers. 1 Theoretically the variance between
counties could be reduced by at least 50 per cent, if a given set
of counties were regrouped into half as many units. 2
There is, however, a limit to this progress, and it is a limit
set by cost considerations. The primary units had to be kept
small enough so that all work within the areas could be personally
directed by the local supervisor with the minimum amount
of travelling. The compromise between the contrasting directives of accuracy and cost led to the use of 2,000 combinedcounty units in place of the 3,000 actual United States counties. 3
As already indicated, only 68 of the 2,000 primary units
could be included in the national sample because of administrative and cost limitations. The 2,000 units were therefore
grouped into 68 sets on the basis of similarity of units. The strata
thus established did not, of course, contain identical units. Nor
did they contain units which were selected in accordance with
rigid and undeviating criteria. Both objective and more
intuitive considerations were utilised, the selection reducing
sampling errors to the extent that the units within a stratum
were made more and more like each other, without at the same
time introducing any bias.
The first step in establishing the 68 strata was to group the
2,000 primary sampling units in 4 sub-universes. The
first group included the 12 largest metropolitan areas. Group
2 included all primary units which had cities of more than
x
" On the Theory of Sampling from Finite Populations", by Morris
H. HANSEN and William N. HURWITZ, in The Annals of Mathematical
Statistics (December 1943), p. 351.
2
Ibid., pp. 337-38.
3
Under generous cost limitations, Hansen and Hurwitz note, primary
sampling units would not be used at all, sub-sampling units being selected
independently throughout the stratum. Ibid., p. 338.
E M P L O Y M E N T : SAMPLE
63
50,000 population in 1930 (except those already designated for
group 1). Group 3 consisted of those units which had more
than 75 per cent, of the population residing in non-farm areas
in 1940, plus certain other units with a high in-migration rate
between 1940 and 1943. Group 4 consisted of all other units —
chiefly, of course, farm units — and included three-quarters of
the 1940 farm population.
Because of the considerable economic importance of the
chief metropolitan areas, each of these was constituted as one
of the 68 strata, thus determining the first 13 strata. x The
primary units in group 2 were combined into 18 strata according
to a combination of factors: geographic location; estimated
migration to or from the area between 1940 and 1943; proportion of labour force in manufacturing; proportion of labour force
in textiles, iron and steel, air craft and shipbuilding ; proportion
of population non-white.
(It is clear that these criteria tend to overlap. Thus, a
selection of units according to geographic location — whether
South or West — will automatically produce a rough selection
according to the percentage of non-white, since the South
contains a high non-white percentage while the West does not.)
In group 2, for example, the North-east strata include Providence, having low migration, high manufacturing and textiles;
Hartford, having high migration, high manufacturing and aircraft; Norfolk plus Princess Anne as representing Atlantic Coast
ports with shipbuilding. The North Central strata included
the combined county Adams-Allen (Fort Wayne) for high
migration, high manufacturing, etc. To the original 13 strata
in group 1, therefore, Providence, Hartford and a total of 18
additional strata were added to represent group 2.
Group 3, consisting of the large urban aggregations not
already dealt with, was classified along the same lines as group 2.
Portsmouth, Ohio, thus represented North-east and North
Central, high manufacturing, with iron and steel; El Dorado,
Arkansas representing South, low migration, high non-white,
etc. This group contributed an additional 11 strata.
Group 4 was stratified according to types of farming, representing general types and specific crops, together with some
attention to the percentage of non-white. Fond du Lac and
1
Washington, D.C., was added to the 12 major metropolitan centres
in group 1 because of its size and unusual industrial composition.
64
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
Sheboygan counties in Wisconsin to represent region 1, very
high dairy; Fremont, Nebraska, to represent region 2, cash corn,
small grains, intensive feeding, etc.
Given the 68 combined-counties which represent the basic
strata the next step was to select primary units from each stratum. The selection here was necessarily a random one, thus
obviating the biases which might arise from a conscious attempt
to select the most representative units. (Because the primary
sampling units did not contain equal numbers of people it would
be obviously unsatisfactory business to have the probability of
selecting unit A, with 100,000 people, no greater than the probability of selecting unit B, with 10,000 people. Therefore,
in order to give the primary units chosen their suitable importance in determining the average for the stratum which they
represented, the selection was designed so that the probability
of choosing unit A, with 100,000 people, was made 10 times as
great as the probability of choosing unit B, which had only
10,000 people.)
Having established first the basic strata and then the primary
sampling units within those strata, the next step is to determine
which households should be enumerated within the primary
sampling units. This selection is made on a random basis from
two sets of materials. For cities and towns the enumeration
is based on a set of fire insurance maps which are available for
most urban areas, including their suburban fringes. These
maps show each individual structure, drawn to scale, in these
areas, and indicate the principal use of each structure —
dwelling place, store, factory, etc. There are indications that
these maps tend to be at least a year behind current construction,
a particular problem when heavy migration and construction
is under way. On the other hand, the frequency with which
insurance is taken out on new construction tends to lessen
errors which arise from this source.
For farm and open country areas the selection is based on a
master sample developed in conjunction with the Department
of Agriculture and Iowa State College. Relying on aerial
photographs and other materials, small areas are designated as
sampling units, these areas containing 4 or 5 farms and a
maximum of 10 dwellings.
After an initial selection has been made of city or village
blocks in urban areas, and of county segments, each containing
about 50 dwelling units, in rural areas, an enumeration is made
EMPLOYMENT:
SAMPLE
65
of households in the specified blocks. This enumeration,
formerly made by field canvass, in now based on the insurance
maps already mentioned. (In practice the enumeration is designed
to include enough households for several periods, each household
being in the sample for no more than 6 months. It is necessary
therefore to bring these enumerations up to date before use
by entering new dwelling-places on the basis of enumerators'
information.)
The households are then selected from the list on the basis
of sampling ratios, or percentage of dwellings to be enumerated.
These ratios are based on the 1940 population distribution and
applied so that, for example, a ratio of 2 per cent, chosen for
a given area on the basis of the 1940 population would mean
a 2 per cent, sample of the 1947 dwellings. x Because the ratio
is not fixed in terms of persons or households, but rather of
dwelling units, this means that where 1947 occupancy rates are
higher than those of 1940, or where persons per room are greater,
or where the number of dwellings has increased, the sample
percentage — being based on dwellings — will reflect this
increase. Thus the sample automatically reflects any change
in population and labour force for the specified area in addition
to providing details on employment status. Therefore the
labour force survey automatically provides a new benchmark at
the time of each enumeration. The use of this technique avoids
the problem of adjustment of employment benchmarks which
arises in the use of the establishment survey.
The remaining steps required to produce estimates of labour
force are simple. Each of the totals secured for the primary
sampling units is multiplied by the sampling ratio originally used
to get that total. By cumulating the inflated totals, a set of
estimated population and labour force totals by age and sex
for the United States is secured. The ratio of labour force to
population in each age-sex group is next computed and these
ratios are then applied to entirely independent estimates of
the United States population in the age-sex group. Similar
estimates for all the other data secured in the labour force
enumerations — major occupation, hours worked, duration of
unemployment, etc., — are made by the application of ratios.
1
Usually one household in six is interviewed per city block, averaging three per block; in open country neighbourhoods about one in
seven is chosen, averaging six to seven per open country enumeration
segment.
66
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
The reason for this step is clear. Independent estimates of
the United States population tend to be based on surer methods
of estimate, since the recording of birth and death rates (as
well as of emigration and immigration) is now accurate enough
to permit very close estimates to be obtained of the population
in each age-sex group. Hence, the application of ratios will
then produce exact labour force figures. 1
The accuracy of the procedures outlined above is contingent
upon the careful design of the sample involved as well as upon
the maintenance of an effective field staff with a careful check
on the adequacy of their enquiry procedure. Experience has
confirmed the main lines of sample design sufficiently well so
that the procedure developed has been essentially carried over
intact into the Canadian inquiry. In broad outline it can
presumably be applied in other countries as well.
Unfortunately, available data do not permit any discussion
of sampling as it is carried on in the U.S.S.R. 2 , but the techniques developed by Mahalanobis and others for recent surveys
in India are obviously only a stage away from those required
for current labour force surveys. 3
The course of improvement for current population census
surveys is probably, so far as sample design is concerned, to
develop the stratification so that detailed estimates of employment in industry can be reliably computed. The gross discrepancies for the trend of United States estimates of service
employment from 1941 to 1943, and the divergences between
the estimates of construction employment made by the Dominion Bureau of Statistics Labour Force Unit and by that
agency's Employment Statistics Branch suggest the need for
such a step. As discussed in Chapter XIV, on consistency, this
means co-operation with the establishment reporting agency to
analyse biases in housewife reporting on industry and occupation. But it also means development of a sample design
which will precisely represent industries as well.
1
In point of fact the delay with which the armed services provided
estimates of the age composition of the army and navy during recent
years led to the use of incorrect assumptions on the age composition of
the civilian population, and hence of the civilian labour force. This
fact explains the major portion of the discrepancies between the movement of the Bureau of the Census and the Bureau of Labor Statistics
estimates of non-agricultural employees, discussed elsewhere in this study.
2
J. GOLDMANN: Statistical Review of Czechoslovakia (March 1946).
8
"A sample survey of after-effects of the Bengal Famine of 1943",
by MAHALANOBIS, MUKERJEE and GHOSH, in Sankhya (July, 1946).
ESTABLISHMENT LIST
67
It is therefore recommended that an analysis should be
made of the relations between the distribution of (a) total nonfarm employment and (b) employment in each major non-farm
industry, as between the various primary sampling units. Such
an analysis can make possible changes in the sample design to
yield reliable estimates of employment in major industry
groups. l
VIII. ESTABLISHMENT LIST
It is now necessary to give more detailed consideration to
the manner in which an establishment list is selected. Attention will first be given to the various national methods and then
to more general problems and conclusions.
I. NATIONAL LISTS
In Australia and Great Britain, where wide-ranging social
insurance systems are in operation, the establishment list is
based on the social insurance rolls. All Australian establishments subject to the payroll tax — and this includes virtually
all non-farm establishments with a payroll greater than £20
per week — supply employment data. 2 In the United States
" t h e list of reporting establishments has been compiled mainly
from telephone directories, trade directories, lists furnished by
Chambers of Commerce, boards of trade, retail associations, and
lists of Government contractors ". Factories with a value of
product less than $5,000 are excluded, as they are in the biennial
census of manufactures, while a general limitation to firms
employing 3 or more persons applies in non-manufacturing.
The Canadian mailing list was originally "based largely upon
the lists of firms maintained in the Annual Census of Industry
1
Sampling of living arrangements in tents and shacks—which are
sure to be ignored on fire insurance maps and sure to be inconveniently
located for purposes of enumeration—is particularly to be recommended
in these days of housing shortages. In the year 1943 some 10 per cent.
of the dwellings constructed in the United States were valued at less than
$500. (Monthly Labor Review, November 1943.)
2
Earnings data for manufacturing, given in the production census
for 1940/41, suggest that manufacturing firms employing less than 4 or
5 employees would be excluded.
68
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
and other branches of the Dominion Bureau of Statistics ".
Among the sources used to keep this list up to date, are "newspapers, trade journals, publications of employers' associations,
boards of trade, chambers of commerce, industrial development
departments of large corporations, provincial departments, etc. "
Sweden's Social Board relies on a list drawn from the Board
of Trade's tabulation of establishments reporting in the latest
manufacturing census, this list being revised from time to time.
The size limitation used in the factory census —firms employing
at least 10 workers — is followed in the current returns, with
most firms employing " a t least 10-25 workers ".
In France the list is basically that of establishments which
reported in the 1936 census, excluding in general firms with less
than 10 employees. Since this list is out of date, as the French
report notes — and, since in addition, some of the local lists
were destroyed during the war — the statistical service has been
forced to place considerable reliance on the local labour inspectors
for a list of the larger establishments in each region.
II.
NON-REPORTING
Given the list of establishments, one main problem confronts
the statistical agency: how can employment changes best be
computed from the returns of these establishments ? The
solutions which the different statistical agencies have worked
out will be outlined briefly.
First, the issue of non-reporting may be considered. With
few exceptions the reporting of employment by establishments
is not compulsory. Hence the extent to which questionnaires
are filled up and returned is dependent largely on the attitude
of the persons reporting. Needless to say, complete success is
never attained. The United States practice is to ignore late
returns if they do not represent a significant proportion of
employment in the industry. If they do, or if they relate to
some very large concerns, revised figures are issued.
In Canadian practice, the employment of non-reporting
firms believed to be in operation is estimated " a t the level of
the preceding month, except in periods of seasonal slackness,
when their preceding month's reports will be reduced in accordance with the seasonal curtailment shown by such establishments in the same month of earlier years, and in accordance
ESTABLISHMENT LIST
69
with the current movement of employers in the same industry
and area. Where employment in a given industry shows
seasonal expansion, however, the figures for a firm whose
returns are in arrears are not adjusted upward, but will be
carried at the level of the preceding report. The correct data
furnished by the firm will then be used as the 'last month's'
figures in the next tabulation, thus providing the revised
figures frequently shown in the monthly surveys".
In contrast to these procedures is that of France, where the
non-response rate is very high. The French report indicates that
the local labour inspectors mail statistical forms to an estimated
70 per cent, of the larger firms in their areas. Since about
a third of the establishments questioned in a recent survey did
not reply, it would appear that only about 45 per cent, of the
estimated total of the firms in the groups to be covered actually
replied. 1
The difficulty with non-reporting of this magnitude is that
the firms which do not report are hkely to have a different
employment experience from those who do. Taking into
account the replies received will imply that the excluded firms
have the same experience. The resultant bias in the results
will be insignificant when the non-reporting percentage is small,
but it may be serious when the percentage is as high as it is in
France.
If there is any substantial percentage of non-reporting, it
may no longer be valid to assume (1) that the movement of
employment in the sample represents that in the universe
of establishments, and (2) that the movement of employment
in reporting firms is the same as that of non-reporting firms in
the sample. The firms which do not report are more likely to
be small, to have declining employment, and to be going out
of business, than the firms which do report. Furthermore, the
non-reports will tend to cluster in certain lines of industry, if
for no other reason than that the characteristic size of firm
and interest in employment data will differ from industry to
industry. It is therefore essential for each statistical agency to
measure the overall level of non-reporting at least semi-annually
in respect not only of the quantity of employment omitted but
also of the distribution of firms in different industry, size and
geographical groups.
1
Revue française du travail, May 1946, p. 175.
70
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
III.
COMPARISONS BASED ON IDENTICAL FIRMS
One further point must be insisted on. The entire procedure
of computing employment changes from establishment reports
is based on comparisons of employment, as reported by an
identical set of establishments for two successive months. The
sample will, of course, change from time to time as firms are
added or dropped. But the comparison between two successive
months (or quarters) must be between an identical group of
establishments. Otherwise a change in the numbers employed
in the establishments in the sample would reflect the changes in
the reporting establishments as well as the change in employment itself.
Where non-reporting is relatively small, this requirement
can be readily met. But where non-reporting is large, it ceases
to be complied with. If there is a large percentage of nonreporting, each quarter's returns may actually be a substantially
new entity, having only an attenuated relationship to those of
the previous month. To illustrate the point, let it be assumed
that the employment experience of firms which reported in
February, Sample A, and that of firms which reported in March,
Sample B, was as indicated in the table below.
INDICES OF EMPLOYMENT
January 1947
February 1947
March 1947
Sample A
100
90
95
Sample B
100
95
100
Each sample shows a decline from January to February,
and each shows a 5 per cent, gain from February to March.
But if the statistical agency reports employment in February,
using Sample A, at 90, and employment in March, using
Sample B, at 100, then a rise of 11 per cent, is shown for the
period from February to March. Either sample alone indicates
a 5 per cent, increase. But by comparing changes in nonidentical firms an 11 per cent, rise in employment, or twice the
true figure, would be inferred.
In practice the errors created by non-reporting are likely
to be less substantial. But of the possibility of such errors
developing there can be little question. Because non-reporting
is a factor even in the best-developed samples, where coverage
is most carefully designed and the fluctuations in the number of
establishments are trivial, it is vital that comparisons be made
only between identical establishments.
T R E N D BIAS
71
It is therefore recommended that all statistical agencies
adhere to the practice of comparing identical establishments. It
is further recommended that at least semi-annually the employment experience of reporting and non-reporting establishments
should be compared so that the adjustments necessary to prevent
the growth of significant non-reporting biases could be made.
(1,29(2)).
IX. TREND BIAS
The present chapter is devoted to a discussion of the bias
which frequently develops in employment series based on establishment reports because a due proportion of newly established
firms is not continuously introduced into the establishment
sample. The manner in which this "trend bias " operates is
essentially as follows.
Let it be assumed that in January of any given year the
sample includes a representative group of all existing establishments. By February the sample will be somewhat less representative because it fails to include any establishments which
began operation in January. Its representativeness will continue to decrease as the months pass. Some new firms are
always being organised in all but the most stable and unchanging
of economies. Hence the trend bias will usually take the form
of a downward bias, making the estimate of employment at
any time too low. The manner in which the bias operates is
described below in Section I, using data from several countries
that publish full statistical materials. Methods of removing
the bias are then outlined in Section II.
I. NATIONAL EXPERIENCES
In Denmark the establishment list is made up at the beginning of each year and is based largely on the list of firms reporting to the annual census of manufactures. According to the
Department of Statistics, the 1944 survey covered 56 per cent.
of all man-hours worked in the larger Danish firms.1 The
1
The census of factories enumeration, which is a starting point for
the monthly index list, excludes establishments with less than 5 wageearners. A further exclusion of small firms is made for purposes of the
monthly index.
72
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
technique followed is to send a set of 12 forms at the beginning
of the year to each one of the firms selected, one form to be
returned each month. Since the list is not increased to include
any firms which begin business between January of one year
and January of the next, it is inevitable that the original
coverage figure — 56 per cent. — should decrease continuously
during the course of the year.
For example, the number of firms in the Danish clothing
industry that employed 5 or more wage-earners rose more than
8 per cent, in a recent year. 1 The number in iron and metal
industries as well as in technical and chemical industries rose
more than 5 per cent. To take account of this continual growth
the Danish series are adjusted annually to figures obtained from
the annual census of manufactures ; hence, an annual adjustment will usually be sufficient.
For Australia the returns of the annual production census
enable a ready comparison to be made between the results of
payroll tax reporting and those indicated by still more comprehensive returns. The trend of employment from July 1941
— when payroll tax statistics begin — through two years of
war production to June 1943 is indicated in the following
table. 2
INDICES OF MANUFACTURING EMPLOYMENT
IN AUSTRALIA
Date
July 1941
J u n e 1943
Production census
100.0
110.8
Payroll tax returns
100.0
100.3
It will be seen that according to the payroll tax statistics
factory employment had not gained at all during two years of
heavy munitions production and forced draft output. But
production census results, available much later, indicate that
a substantial rise did, in fact, take place. The conclusions
for current manpower policy to be drawn from one estimate
must certainly have differed from those to be inferred from the
other.
Finally, the results of a similar comparison between factory
1
Since the net number of clothing firms rose from 795 to 852, it
seems reasonable to assume t h a t the gross increase in new firms was
greater.
2
Production, 1941-1942, P a r t I, p. 31 ; Quarterly Summary of Australian Statistics (June 1943), p. 51.
73
T R E N D BIAS
census and current employment reports for Switzerland may
be presented. 1
INDICES OF MANUFACTURING EMPLOYMENT
IN SWITZERLAND
Year
Factory census
Quarterly index
1939
1940
1941
1942
1943
100
—
118
117
116
100
107
109
110
110
II. ADJUSTMENT FOR BIAS
The trend bias which these current reports reveal cannot
safely be ignored. That bias is not a constant figure or percentage. On the contrary, it is cumulative. Therefore, total
employment as estimated from establishment reports will
depart further and further from the true total. Since the rate at
which new firms come into existence, and the rate at which they
hire workers, are related to the course of the business cycle to the
season of the year, and to the rate of growth of the economy,
this cumulative error does not develop in accordance with any
simple formula. The basis fora correct adjustment of the trend
bias must therefore be specifically studied for each country.
One method of adjustment is that used in Canada. " Newlyformed establishments which employ 15 or more employees are
added to our mailing list to receive monthly questionnaires as
soon as we learn of their existence. Those commencing since
the basic period are tabulated with 'zero' base figures, their
employment and payrolls being regarded as complete accessions
in the industry and area. " Thus, for 1926 (the base year) the
1
Entwicklung des Beschäftigtenstandes nach der Industrieberichterstattung seit Kriegsausbruch (January 1945), pp. 1, 16; International
Labour Review (March-April 1946), p. 290.
Because the current reports for Switzerland include employment
in construction, while the census reports do not, it was necessary to
exclude that group, This was done by extrapolating the 1940 population
census totals for employment in building and for employment in all
industries covered by the current reports; deducting the estimated
building series from the total series; and computing an index for the
balance—an index then reasonably comparable with that of the census.
Analysis of the monthly results indicated that the results would have
been substantially the same had employees absent on military service
been included in deriving the movement of the monthly data—as they
are included in the factory census. Finally, the fact that the census of
factories includes salaried workers, unlike the current reports, suggests
that a comparison for wage-earners alone would have indicated a still
greater discrepancy, the employment of salaried workers usually being
more stable than that of wage-earners.
74
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
employment will be set at 0, for the previous month at 0,
and for the given month, as reported. 1
One technique for trend bias adjustment is utilised in
the United States. Until recently the U.S. Bureau of Labor
Statistics adjusted its estimates of manufacturing employment
to totals given in the biennial censuses of manufactures. Employment totals for other industries were adjusted to the data from
the censuses of trade (1929, 1933, 1935, 1939), of construction
(1929, 1939) and so forth. The adjustments were usually made
long after the census dates, because data were not immediately
available. Beginning in 1939, adjustments were made whenever
the tabulations of the social insurance system, with its wide coverage, became available, though here too there was a long interval
between the date of the figures and the time of adjustment.
Since the beginning of 1946, however, the United States has
utilised a method of continuous adjustment for the manufacturing employment series. 2 The procedure is an empirical one,
founded on the fact that between 1934 and 1944 the manufacturing indices fell below benchmark data at an almost uniform
rate of 2 percentage points a year. Had an adjustment of
0.2 per cent, been made monthly over the entire period virtually
no other correction would have been required. On the basis
of this relationship the Bureau currently adjusts the figures for
each major manufacturing industry group and the total for all
manufacturing each month before they are released. No similar
correction is made in the non-manufacturing industries at
present, nor is one in prospect — the materials for such an
adjustment being largely lacking.
While there is considerable reason to recommend this method
of current adjustment for trend bias, it should be applied to
all series and on a more systematic basis than hitherto. Though
the main line of development must be separately worked out
for each country, certain general suggestions can be offered here.
1
This procedure must be sharply distinguished from that utilised
for firms which have been in existence for some time but have only
recently begun reporting to the Bureau. In that event " their basic
averages are, obtained either from sources in the Bureau, or by correspondence. If it is impossible to obtain the basic averages by these
means... (they) are calculated... on the assumption that the firm's
employment trends will have conformed to those of other establishments
in the same industry and the same area in the given period".
2
See Robert B. STEFFES: Trend Correction in the Manufacturing
Employment Series of the Bureau of Labor Statistics (Bureau of Labor
Statistics, March 4, 1946).
TREND
BIAS
/O
An initial step for each country must be to examine the
actual development of the trend bias, determining its magnitude,
direction and level of variation. Should a sufficiently clear
pattern be found, it may then be possible and desirable to adjust
current data on an empirical basis similar to that used in the
United States.
In general, more detailed and complex methods of trend
adjustment are to be avoided on grounds of expense. But
where an empirical adjustment is impossible or inadequate,
and where a continuously adjusted sample is required, consideration might be given to a technique based on returns made to
social insurance authorities. (1) From an analysis of the
records over a period of years the separate rates at which
new firms in each of the major social insurance regions took
on workers could be determined. Given these rates, the basis
for weighting the data from each region could be worked
out. (2) The next step would be to determine the size of
sample — 2 per cent., 5 per cent., 10 per cent. — required to
produce stable results in measuring employment in new firms.
(3) Given this size, say 5 per cent., a 5 per cent, sample of employment in ail new establishments reporting to the social insurance
authority would be made each quarter — or, if feasible, each
month. The totals for this sample in each area would be combined by using the weights already derived, and the result multiplied by 20 to produce an estimated total increase in employment resulting from the rise of new firms. (4) Given the percentage of employment in the sample at the last benchmark
date — say 79 per cent. — the same percentage would be
applied to the total increase in order to determine the quantity
of employment to be added to the sample total for the given
month. The new total would then include a due proportion
of employment in new firms.
This method of correction for trend bias may merit consideration. However, the nearest method of adjustment would be to
design the sample of establishments so that it automatically registers this (and other) changes. Some aspects of this solution have
already been discussed in Chapter III, on Reporting Systems.
Whatever the method of adjustment utilised, it is quite
clear that no employment series based on an establishment
sample is free from a cumulative trend bias unless the sample
is specifically adjusted to that end. It is therefore recommended
that employment series based on establishment reports should
6
76
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
be adjusted to benchmark data from census enumerations or
from social insurance records as soon as possible after such
data become available (I, 27), Consideration should be given
to the possibility of a more frequent and prompt adjustment
by the development of special techniques. Such techniques
may in the first instance be empirical ones, merely extrapolating
an adjustment for bias that has proved satisfactory in the past.
As resources and facilities permit, periodic studies should be
made of the nature and extent of the trend bias which develops,
between the dates of adjustment for benchmarks, in employment series based on reports from identical establishments
(I, 31 (1)). On the basis of such studies, methods should be
developed to include in the sample a due proportion of persons
employed in new establishments (I, 31 (2)).
X. COLLECTION OF DATA
The entire structure of employment statistics reported by
establishments rests on the questionnaire. The accuracy with
which it is completed will determine in no small measure the
accuracy and completeness of these data. Certain aspects of
the process of filling in returns and securing satisfactory
co-operation with reporting establishments therefore deserve
special attention.
In most countries the customary method for securing the
co-operation of establishments whose employment data are
desired is simply to write requesting such co-operation. If the
establishment agrees, then the usual procedure is to accept
the returns at their face value — subject to certain checks for
obvious inconsistency or error. This procedure can usefully
be supplemented both from the viewpoint of securing more
satisfactory statistics and from the viewpoint of public relations.
From the statistical point of view it is essential to see that
the interpretation of the questions by the person who actually
fills up the form is the same as that desired by the agency. To
achieve this it may be desirable for representatives of the
statistical agency to visit a sample of new establishments,
including all large ones, to verify procedures and answer questions. Such a procedure, however, is relatively expensive and can
usefully be supplemented by providing newly reporting firms with
a booklet explaining what figures are desired and emphasising
graphically the commonest shortcomings in the returns.
TREND
77
BIAS
From the viewpoint of public relations, it is useful to establish
periodic contacts with establishments which report employment
data. These contacts will assist the statistical service to make
the data on employment most useful. They will also indicate
what further data are valuable. Many statistical offices furnish
addressed post-free envelopes for returning the filled-in forms,
thus reducing the rate of non-reporting.
The experience of certain countries has shown that employment data as reported to authorities with tax powers — notably
social insurance authorities — are not the same as those reported
to the more narrowly statistical agencies. It is recommended
that studies should be made comparing differences in the
reporting of employment data to the various agencies which
issue employment series, in order to determine how serious a
margin of inconsistency may exist. As resources and facilities
permit, employment series from establishments should be
adjusted if necessary to take account of differences in the
employment figures reported to different statistical authorities
and of any tendency to include in the reports data which are
more inclusive or less inclusive than those requested (I, 35).
Agencies securing reports from establishments have commonly sought returns for individual establishments. This
has been necessary for a correct allocation of the employment
of large multi-product firms. For example, the Danish report
emphasises that "in the case of firms with more than one
establishment a report is as a rule sent in for each establishment;
this is done in all cases where the single establishments fall
under different industries ". The importance of this point can
be illustrated by a recent study which indicated that 15 of the
largest firms in the United States each operated establishments
in 10 or more separate industry groups.* Data for Great
Britain tell a similar story of diversity. 2 It is therefore recommended that where at all possible employment returns should
be secured for individual establishments, and particular attention should be given to large firms operating in several industries
(I, 29).
1
Willard T H O R P and Walter C R O W D E R : The Structure
of
Industry
(1941), Temporary National Economic Committee Monograph No. 27,
p. 593.
2
" The S t r u c t u r e of British I n d u s t r y " , b y H . L E A K a n d A. M A I Z E L S ,
in Journal of the Royal Statistical Society (1945), p p . 146, 158-159.
PART IH
XI.
TYPES OF UNEMPLOYMENT SERIES
The most important series on unemployment is a measure of
aggregate unemployment. This series should be prepared monthly
in the more industrially developed countries (I, 20). This series,
as discussed below, in Chapter XIII, should be as comprehensive
a measure of the total volume of unemployment as may be
possible. In the presentation of national estimates of unemployment primary emphasis, however, should be placed on
percentages rather than on absolute figures (I, 38).
The reasons for this are, of course, well known. The size
of the labour force is steadily increasing. Since the quantity of
frictional unemployment is likely to increase concomitantly,
the absolute number of unemployed will also increase. But
this would not necessarily point to any growth in the severity of
the unemployment problem. The use of unemployment percentages prevents such misinterpretations. (A similar problem
arises when unemployment totals are derived from a system of
social insurance or from trade union statistics the coverage of
which is widening, and the solution is again to use percentages
rather than aggregates.)
For countries relying on statistics of trade union unemployment the percentage will be calculated as the ratio of unemployed
numbers to the total membership of all reporting unions. 1 For
those with data from unemployment insurance systems it will
be the ratio of the unemployed insured to the total of insured
1
To secure a more representative percentage for unemployment in
all industries, percentages may be computed for each reporting group,
and these may then be weighted in accordance with the actual or estimated
distribution of total employment in covered occupations or industries.
TYPES OF UNEMPLOYMENT SERIES
79
persons. For countries with data from labour force surveys
chief emphasis should be given to the percentage of the civilian
labour force that is unemployed (I, 39). 1
Estimates of the proportion unemployed are only a beginning — for many practical purposes, estimates of unemployment
by age, by region and by sex are also of considerable value.
Such series spotlight the groups which are hardest hit by
unemployment. They suggest where action is required, and
they provide the basis for deciding what action is best to reduce
unemployment.
I. NATIONAL PRACTICE
In addition to their limitations for the measurement of
national unemployment, data from employment exchanges
which are not included in the administration of social insurance
systems, and trade union data, may not furnish a satisfactory
basis for determining which age-groups are most affected by
unemployment, which regions have the most critical problems,
or which occupations are dominant in the unemployed group. 2
On the other hand, social insurance data lend themselves to
very satisfactory classifications and cross classifications of data
on the unemployment of the insured population.
Denmark, for example, provides monthly data for detailed
industry groups, for a large number of regions, and for the
separate unemployment insurance funds. It also provides
annually a model of detail, giving not only the information
mentioned above, but the number of months of employment
1
An additional percentage t h e usefulness of which should be investigated further would be t h a t for the proportion of employees who are
without work. In most industrialised countries there is a long-term
tendency at work for the proportion of t h e self-employed to decrease,
largely because of the decline in the proportion of agricultural employment. If, for example, with a constant labour force this decline occurred, then a greater proportion of workers would be found in industry, trade
and other urban employments. If the unemployment rate among these
new additions to t h e urban employee group were t h e same as t h a t of
existing urban employees, this would mean a rise in the percentage of
t h e labour force unemployed, since the percentage of unemployment
among t h e self-employed—if it exists at all—is smaller t h a n t h a t among
employees. Hence, even though in each industry t h e percentage of
unemployed remains constant, the national unemployment percentage
would rise on account of the shift of persons from agriculture to industry.
2
Australia publishes d a t a on unemployment of union members by
industrial categories, and New Zealand releases registration data by
region and type of worker (fully employable, semi-employable, juveniles,
aged 60 years or over) as well as by industry.
80
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
in the year for each insured worker according to age, sex, and
industry, as well as still other cross classifications.
In Great Britain distributions of the unemployed are published monthly, by region, cross classified by duration of unemployment (under 2 weeks; 2 to 8 weeks; over 8 weeks; temporarily
stopped) by region, cross classified by age (men 18-65; boys
14-17; women 18-60; girls 14-17) by industry, cross classified
by sex. At quarterly intervals further regional details are
released.
Belgium issues weekly data on unemployment by regions,
cross classified by sex. Separate data are issued for total and
partial unemployment. The unemployed in each category are
subdivided each month into industrial-occupational groupings,
specifying separately agriculture, mining, textiles, employees,
common labour, etc. Switzerland separates the wholly unemployed from those in partially unemployed (a) according to
canton, and (b) according to industry (clothing and leather;
construction and wood-working, textiles, etc., and industry not
specified).*
In the United States data are published for unemployment
by sex. National data on age of worker, duration of unemployment and major occupational groups are available, though not
published. (Data on unemployment insurance beneficiaries by
region are issued by the Social Security Board.)
Canada releases figures for unemployment as estimated from
labour force surveys, by sex; by region (5 provinces); by age
(14-19, 20-24, 25-44, 45-64, 65 and over); by duration of unemployment (under 1 month, 1-3, 4-6, 7-12, 13-18, 19 and over);
by most recent industry (manufacturing, construction, etc.).
II.
GENERAL PROBLEMS
These, then, are some of the cross classifications of data
which provide supplementary information on unemployment.
Are they sufficient ? And what are the requirements of an adequate system for classifying the unemployed ? In general the
unemployment data for a country should be such as to throw
light upon the major questions on the extent, trend, characteristics and causes of unemployment which are put to those in
charge of labour statistics.
1
About one-third of all insurance benefits for total unemployment
in the year 1943 was paid to persons in unspecifled industries. La Vie
économique (August 1944).
T Y P E S OF UNEMPLOYMENT SERIES
81
How large is the unemployed group ? How serious is the
unemployment problem? Is the total inflated by workers only
temporarily in the labour market who will withdraw as soon as
regular workers are unemployed ? How many of the unemployed are women? Is there a growing "hard core" of unemployed who have been without work for months, even years ? Are
younger workers wasting more time in unemployment before they
find their first job ? Is the older worker finding his hold on the
labour market progressively more tenuous? Do particular
cities or regions suffer from unemployment more than the nation
as a whole ? In connection with planning work programmes for
the unemployed, or programmes for moving workers out of
depressed areas, what are the available skills of the unemployed ?
For appropriate action, whether public or private, whether
operating through direct relief, work relief or general Government. expenditures, prompt answers to these questions are
required.
To serve these various purposes a series on the total volume
of unemployment is essential. It is no less necessary to provide
series on unemployment by major geographic region, by sex,
for the chief age-groups (I, 22).
As resources and facilities permit, separate series should be
calculated for unemployment in additional geographical regions,
the chief centres of population and for more detailed age intervals (single years for juveniles; 10-year age-groups for adults
(I, 23). Where such information is of significant utility for
labour retraining and shifting, data on industry of last employment and major occupation group should likewise be gathered
(I, 21). For most purposes it is far more important to have the
series outlined above than it is to have series which are precisely consistent one with the other. Therefore some series, for
example, may be gathered by means of a regular labour force
survey, while other series may be obtained less frequently or
from social insurance tabulations. Where pools of unemployment exist in local regions it is particularly important that
programmes for public action should be designed in such a
way as to alleviate the actual local problems. For such areas
therefore, and as resources and facilities permit, estimates of
unemployment cross classified by age, sex and duration,of
unemployment should be made available (I, 24).
82
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
XII. UNEMPLOYMENT: SAMPLE
Unemployment statistics are intended to measure the volume
and the trend of unemployment. To serve these purposes effectively, these statistics should be secured by a comprehensive
enumeration or by a reasonably accurate sample. For the
very few countries whose social insurance systems provide a
complete and current measure of national unemployment, the
problems of sampling with reference to unemployment data
are of secondary interest. But such data are not available in
most countries. To provide proper perspective on the trend
and level of unemployment in countries which do not have a
complete and current measure of national unemployment, it is
necessary to observe the basic principles of sample design so
that the partial totals secured will yield the same indications
as more complete totals. The appropriate techniques of sample
design and analysis will differ from system to system — registration at employment exchanges, trade union reporting, social
insurance systems and labour force surveys. Each will be
discussed in turn.
I. REGISTRATION AT EMPLOYMENT EXCHANGES
1
Few methods of measuring unemployment are so difficult to
assess for adequacy of sampling procedure as is the method of
registration at employment exchanges, there being so little
published information which bears on the subject. However,
certain conclusions may be ventured.
1. Registrations usually include more than those who are
unemployed. They also cover those who have jobs but are
looking for better ones. In some instances, they may even
include persons who are not in the labour force but who would
take work if they found a job which suited their specifications.
1
The following discussion relates to the use of registration data from
employment exchanges where the data are not linked to the operations
of a social insurance system.
U N E M P L O Y M E N T : SAMPLE
83
2. The coverage of exchange figures tends to be limited
geographically. Insufficient representation is given in the
sample to unemployment in rural regions and outlying urban
areas, where it is difficult to reach registration offices.
3. Because workers in the different occupations and industries do not register at the same rate, the exchange registration
sample tends to be biased occupationally and industrially.
Few professional and managerial workers resort to employment exchanges, since most vacancies in positions of that nature
are not registered with exchanges. At the other extreme of
employment stability, few domestic servants and few casual
labourers, whether urban or rural, are placed by exchanges.
Longshoremen and marine workers, who are particularly
subject to unemployment, usually find work either through the
shape up or through union hiring halls. (A connection may
sometimes exist between these and the exchange, as is the case,
for example, in Belgium.) These practices of course may vary
from country to country and may change over a period of time.
4. Administrative limitations tend to lessen the value of
employment exchange figures as a measure of unemployment.
The employment service files become loaded with inactive
cases — persons no longer looking for work, persons who have
found work but have not notified the agency to that effect.
Weeding out these cases is usually a slow and arduous task.
5. Studies on worker placement reiterate the obvious fact
that many factory workers seek work by direct application to
factory employment offices, by contacts through friends already
employed, or through their unions.
The extent of these placements outside the exchanges has
been effectively emphasised by the Belgian report. The report
notes that " t h e statistics of those seeking work as tabulated
by the placement agencies do not warrant the same measure of
confidence as those of the communal (unemployment insurance)
administrations". The chief difficulty "is to keep up to date
the list of those seeking work"; "the figures secured are often
higher than they are in reality, the registrations of those seeking
work who have already found employment by their own means
not having been eliminated". This will not only inflate the
level of unemployment estimates but, since the amount of such
inflation is likely to vary from time to time as efforts are made
to bring the files up to date, their trends will also be distorted.
84
EMPLOYMENT, UNEMPLOYMENT AND LABOUR
FORCE
6. Finally, the adequacy with which the movement of
employment exchange registrations will represent the movement
of total unemployment is impaired in proportion to the very
success with which employment exchanges accomplish their
end. For as the exchanges achieve a securer place in the confidence of employer and employee progressively more unemployed
workers will resort to the exchanges in their search for work.
This increasing coverage will tend to appear as an increased
number of unemployed, but the larger numbers will actually
reflect merely the increase in coverage and not a higher level of
unemployment.
A direct indication that employment exchange registrations
do not adequately measure the changing course of unemployment is given in table I, where the movement of registration
estimates in a number of countries is compared with that of
TABLE I.
INDICES
COUNTRIES
AS
O F UNEMPLOYMENT
ESTIMATED
BY
TRENDS
EMPLOYMENT
REGISTRATION AND B Y OTHER MEANS
Canada
Year
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
Estimates
100
319
413
597
604
487
451
402
315
380
361
263
United States
Great Britain
Bureau
UnLabor Regis- employRegis- ofStatisment
trants
trants
tics
insurance
estimates
100
220
465
501
546
593
560
601
590
702
692
670
100
84
71
97
87
73
100
103
69
99
84
71
IN
100
153
204
212
191
161
148
125
Registrants
100
158
216
226
207
178
167
144
SELECTED
EXCHANGE
X
Netherlands
Unemployment
insurance *
Registrants
100
163
171
177
201
201
161
150
100
196
234
241
278
300
267
256
1
Data from: I.L.O. Year-Boolt of Labour Statistics, 1943-1944; Mimsfri; of Labour
Gazette, issues of November, 1929-1937 (London); U.S. BUREAU OF LABOR STATISTICS:
Preliminary Estimates of Labor Force, Employment and Unemployment, 1920-19Ì0
(Washington, 1945).
* Data refer to the percentage of insured workers who were unemployed. Indices of
the aggregate number of insured unemployed would be less accurate indicators of the course
of unemployment and would show a still greater contrast.
UNEMPLOYMENT: SAMPLE
85
more accurate estimates, generally those of the insurance
system. The change in unemployment between 1935 and 1936
indicated by employment exchange registration data in both
Canada and United States was upward, while overall estimates
indicated a downward movement. In the Netherlands an
upward movement indicated by employment exchange figures
contrasts with no change according to unemployment insurance
data. Between November 1945 and June 1946 "unemployment in Canada as measured by the number of unplaced
applicants reporting to the national employment service "
showed a very sizable increase, while a direct census of unemployment showed a no less striking fall in unemployment over the
same period. 1
These, therefore, represent the major failings of unemployment estimates based upon employment exchange registrations
where such exchanges are not related to the operation of a social
insurance system. For countries relying on such data as their
main measure of unemployment it is highly desirable that
periodic studies should be made of the relation between the total
number unemployed and the number of unemployed as registered at the employment exchanges (I, 42). Studies should also
be made of the extent to which statistics of the registered
unemployed derived from the operations of employment
exchanges actually include persons who are employed — either
because of failure to include them at the time of registration, or
because of difficulties in keeping the records of the exchanges
up to date (I, 43).
In any case, it is recommended that estimates of the total
numbers unemployed in any country should, wherever possible,
be based on figures derived from sample surveys of the labour
force, from the operations of an unemployment insurance or
other social security system, or from both sources (I, 36).
II. TRADE UNION STATISTICS
In discussing the statistics of unemployment among trade
union members attention will be focused first on the accuracy of
1
CANADA, DEPARTMENT OF LABOUR: The Labour Gazette (August
1946) ; CANADA, DOMINION BUREAU OF STATISTICS : Labour Force Bulletin
(July 1946).
Cf. also the heavy contrast in U.S. figures for September 1945J u n e 1946; U.S. BUREAU OF THE CENSUS: Monthly Report on the Labor
Force (July 1946); and U.S. SOCIAL SECURITY B O A R D : Social Securitg
Bulletin (May 1946).
86
EMPLOYMENT, U N E M P L O Y M E N T A N D LABOUR FORCE
the benchmark data and then on the competence of the movement series used to bring the figures up to date.
The first and most general comment to be made on the basic
data for these statistics is that they are non-existent. In prewar Norway, for example, reports were issued on unemployment among members of 10 major trade unions. Since, as
is customary, "the trade unions are organised on a mixed
occupational and industrial basis " it is hardly possible to make
any accurate allocation of members to the separate industries.
All that is known regarding their distribution is that the returns
represent unemployment among a quarter to a third of all
union members. Whether this sample group can be taken to
represent all union members and, which is more important,
whether it can be taken to represent all workers, is a point which
remains in doubt.
To put this more concretely the Australian data may be
considered. The Australian report makes an approximate
distribution of trade union members in 1943 into 15 industry
groups. Estimated ratios of insured workers to total employment for four industries in 1943 are as follows1 :
Building
Engineering and metal
Food and drink
Wood, furniture
over
„
„
„
90 per cent.
69 „ „
38 „ „
40 „ „
The sharp variations in coverage will inevitably bias the overall
unemployment percentage. A greater weight than is merited
will be given to the rates in the first two industries, and less than
is merited to the other two. But since the rates differ from
industry to industry, this means that the rate for all industries
will be biased. The bias will, of course, be intensified by the
failure to measure the effect of unemployment among noncovered industries and the failure to reflect adequately unemployment among the less skilled, non-union workers, even in the
specified industries. Hence it is that the Australian Government is careful to point out that "trade union percentages do
1
The report presents average membership of reporting unions, in
1943, total union membership on 31 December 1943, and total employment in June 1943. Because of these differences and problems of
industrial allocation, union membership in building is shown as greater
than total employment. No attempt is therefore made to draw conclusions on the precise figures, but merely on their relative magnitudes.
UNEMPLOYMENT:
SAMPLE
87
not always agree particularly closely with the overall percentage
of unemployment found in censuses of the whole population ".
But more than an inaccuracy in level will exist. The pattern
of unemployment change will likewise be distorted. During
the recovery decade from June 1932 to June 1943, the unemployment rates as reported by trade union members fell by 96 per
cent, in construction, 98 per cent, in engineering and 98 per cent.
in wood and furniture, but only by 82 per cent, in food and drink
manufacture. Hence an over-representation in the first two
groups will make the unemployment decline appear to have been
more precipitous than in fact it was.
An additional distortion of movement will be produced by
the fact that union members tend to be concentrated in skilled
occupations to a greater extent than is the entire labour force.
But skilled workers tend to have a more certain tenure of employment than unskilled workers. They certainly have a particular
pattern of employment and unemployment over the business
cycle. That pattern differs from that of new entrants to the
labour force as it does from that of older workers not attached to
pension schemes, who may remain without work for years before
they are forced into retirement. As a result of such variations
the movement of unemployment rates among union members
will not give a true picture of the movement of unemployment
among all workers.
To guard against these difficulties it is recommended that in
any country in which, through lack of adequate data derived
from social security operations or labour force surveys, national
unemployment estimates may have to be based on trade union
returns, such returns should wherever possible be weighted in
accordance with the numbers of persons employed in the
industries to which they relate (I, 41).
For countries which must concomitantly expand their
employment reporting greatly this recommendation points to
the advisability of considering the adoption of a labour force
survey, since such a survey will provide consistent and comparable estimates of both employment and unemployment. Where
a social insurance system is in existence or in prospect, the
decision must necessarily be made with reference to the relative
cost of securing not merely unemployment estimates, but the
entire body of statistics on employment, unemployment, labour
force, hours and earnings, and to the relative accuracy of such
information when secured.
88
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
III. SOCIAL INSURANCE STATISTICS
1. National Methods
The various national systems for social insurance, as is to
be expected, differ widely both in the extent and in the nature of
their coverage. For Switzerland the number insured against
unemployment in September 1941 was just under 525,000, while
the total number of non-farm employees in December of the
same year, as reported by the population census, was about
TABLE
II.
TOTAL
EMPLOYEES
AND I N S U R E D
EMPLOYEES
I N BELGIUM, 3 1 DECEMBER 1 9 3 9 , B Y I N D U S T R Y 1
Employees
Industry
Mines . . .
Quarries . .
Metals . . .
Ceramics . .
Glass . . .
Chemicals
Food . . . .
Textiles. . .
Apparel. . .
Construction
Lumber. . .
Leather . .
Tobacco . .
Paper . . .
Printing . .
Total
Insured
170,357
38,039
330,072
39,412
29,741
68,322
81,197
240,091
84,014
165,255
104,949
44,890
14,840
18,044
23,228
61,511
20,141
137,072
16,202
11,359
19,255
17,250
134,170
8,119
56,665
40,387
16,924
7,663
6,215
10,627
Percentage Percentage Percentage
of all
of all
ot all salaried wage-earning
workers
employees
wo rK.crs
iusured
insured
insured
36.1
52.9
41.5
41.1
38.2
28.2
21.2
55.9
9.7
34.3
38.5
37.7
51.6
34.4
45.7
14.4
13.8
13.7
11.8
12.3
9.2
9.1
22.1
3.9
10.0
13.1
10.1
11.0
9.5
15.0
37.4
55.8
45.1
43.3
40.9
33.5
23.2
58.0
10.0
35.2
39.4
40.0
56.1
37.6
50.3
1
Henri Fuss: L'organisation de l'assurance obligatoire contre le chômage (1937), Premier
rapport, pp. 33-34.
1,300,000.1 The overall coverage figure of 40 per cent, suggests,
therefore, that coverage was by no means so great that an
analysis of the sample can be dispensed with in evaluating the
adequacy of Swiss unemployment statistics.
Belgium excludes some mine and all marine workers as well
as those employed on State railways. Pursuant to the Decree
of December 1944 it specifically includes all employees under a
contract of service, except the above groups, and includes agricultural workers, domestics, and those receiving only tips, etc.2
1
La Vie économique (November 1942); reply of the Swiss Government submitting data for the I.L.O. Year Book.
2
Moniteur belge, 30 December 1944.
UNEMPLOYMENT:
89
SAMPLE
The scope of the new regulations is obviously wider than that
of the pre-war rules and unfortunately the only data available
for analysis apply to pre-war years.
Referring to the last decennial census, that of 1930, table II
presents coverage estimates for wage-earners, for salaried workers
and for employees as a whole. Wide variations in coverage are
found. While 52 per cent, of the tobacco workers were insured,
only 21 per cent, of the food workers were covered. While 56 per
cent, of the textile workers were insured, only 10 per cent, of
workers in the closely related apparel industry were covered.
TABLE H I .
AVERAGE
ANNUAL
UNEMPLOYMENT
P E R INSURED
WORKER IN BELGIUM, 1921-36 ( I N DAYS) X
Industry
Average dnys of unemployment
per insureá worker per year
Mines
Quarries
Metals
Glass
Chemicals
Food
Leather
Printing
1
11.4
27.0
28.8
38.0
40.8
33.0
14.1
35.0
26.7
37.0
34.4
33.0
34.3
28.8
17.0
Henri Fuss: Op. cit., Deuxième rapport, pp. 9-10.
Reference to the average number of days lost each year per
worker during the period 1921-36 (table III) emphasises how
sharp are the differences in unemployment rates from industry
to industry.
The absence of correct weighting must therefore have meant
that the level of unemployment published for any given time
was not the true level, and unless workers entered the insurance
funds at identical rates from all industries an incorrect picture
of the movement of unemployment would also be given.
These conclusions are reinforced by data from the 1937
Belgian census of unemployment. Unfortunately these data
have not been put in juxtaposition with the numbers of those
insured at the same date. But it is possible to compute the
90
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
ratio of the unemployed workers who were not affiliated with
an insurance fund to those who were. * Variations in these ratios
will then tend to indicate differences in the level of affiliation
from industry to industry.
The overall ratio of all wage-earners is 2.2 unaffiliated to
1 affiliated. The ratio for food is 1.2; tobacco, 2.8; textiles, 3.7;
and apparel, 0.3. Thirty-eight per cent, of those unemployed
in 1937 were not connected with the insurance system, and hence
were excluded from its statistics. 2 Since the 1937 estimates
are based on a comprehensive census, they reinforce the inference
to be drawn from the 1930 figures; variations in insurance coverage from industry to industry were wide, and were certain to
have an effect of distorting the true unemployment rate indicated
for industry as a whole.
in addition to the problem of industrial coverage there is
that of age coverage. While only 10 per cent, of the affiliated
workers in 1937 were under 25 years of age, some 33 per cent, of
the unaffiliated workers were. 2 Such a differential, like that
for industries, will necessarily affect the competency of the data
on the insured unemployed to represent data for all unemployed.
Furthermore, variations in the representation of wage-earning
as against salaried workers will affect the results, since each
group has characteristically different unemployment rates; so
too will differences in the proportion of each sex insured.
In all probability the differential weighting of non-farm
industries would have little effect in biasing the total rate at any
one time. This may be equally true for each of the single other
factors mentioned — wage-earning or salaried worker composition, sex or age representation. But the possible effect of the
combination of these factors suggests the advisability of an
analysis of the adequacy of the sample used.
The nature of the Danish sample is more capable of assessment, data being available both from the population census and
from the statistics of insured workers. Table IV outlines the
industrial coverage of the unemployment insurance system
in 1940. The table is designed to demonstrate not only the
considerable percentage variations by industry — 13.3, 61.5,
71.7, 27.0 — but also the variations in coverage of men as compared to women. Variations in the coverage of wage-earning
1
Royaume de Belgique: Recensement économique et social au 27 février
1937, Vol. VI, p . 171.
2
Ibid., p . 51.
TABLE I V .
GAINFULLY OCCUPIED AND I N S U R E D E M P L O Y E E S
Industry
Employees
(census of population)
Male
All industries
Agriculture, etc
Food and delicatessen industries . .
Textile and clothing industries . .
Building industry, road and earthwork
Timber industry
Leather industry
Stone, earthenware and glass industries
Iron and metal industries
Chemical and technical industries .
Paper and printing industries . . .
Cleaning and maintenance industries
Transport
Commerce and sales
Other services
Other
Female
Total
Employees
(unemployment insurance)
Male
915,649 307,358 1,223,007 436,166
228,565 32,206 260,771 34,728
56,823 20,591
77,414 31,778
17,030 49,289
66,319 11,128
211,336 58,714 270,050 126,048
1,436
28,316 18,325
26,880
4,477
12,985
8,508
4,136
16,313
103,238
10,972
16,710
5,923
51,850
112,090
31,302
18,109
1,716
8,780
4,692
9,376
10,565
1,594
75,391
24,189
4,342
18,029
112,018
15,664
26,086
16,488
53,444
187,481
55,491
22,451
I N DENMARK, 1 9 4 0
Female
Total
86,836 523,002
29
34,757
15,822 47,600
29,302 40,430
286 126,334
1,024 19,349
3,282
7,418
14,687
72,648
10,148
9,151
1,289
42,284
40,096
1,343
7,615
2.497
5,936
3,164
13,216
16,030
80,263
12,645
15,087
4,453
42,284
53,312
19,720
3,320
23,040
1
Percentage of
employees insured
Male
Female
Total
47.6
15.2
55.9
65.3
59.6
68.2
48.6
28.3
0.09
76.8
59.4
0.5
71.3
73.3
42.8
13.3
61.5
61.0
46.8
68.3
57.1
90.0
70.4
92.5
54.8
21.8
81.6
35.8
78.3
86.7
53.2
63.3
29.9
88.9
71.7
80.7
57.8
27.0
79.1
28.4
17.5
1
Two changes have been made in the table as submitted by the Danish Government. The Government table included wage-earners in the population
census data and wage plus salary earners in the unemployment insurance data. The present table includes wage plus salary earners in both. Unemployment insurance data include health services, railways, post and telephone in the construction industry, according to the Government report, unlike
the population census data.
In the present table, therefore, a similar re-allocation has been made in the census data. The additional census data were
taken from Denmark, Slatistisk Aarbog, 1945, table 12.
tO
92
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
versus salaried workers are not reviewed here, nor the effect of
excluding workers under 18 and over 60. The general effect of
these variations, however, is certain to give a different aspect
to unemployment figures than is given, for example, by Canadian
unemployment estimates, which relate to workers in all industries, of all ages, including not only employees but self-employed
TABLE V.
NUMBER OF EMPLOYEES IN GREAT BRITAIN AS
INDICATED B Y UNEMPLOYMENT INSURANCE REGISTRATION AND
EMPLOYMENT ESTIMATES, M I D - 1 9 3 9 , B Y I N D U S T R Y 1
Industry
Mining and quarrying
National government
Local government. .
Gas, water, electricity
Transport, sliipping .
Manufacturing . . .
Building and civil enDistributive t r a d e s ,
commerce . . . .
Domestic service . .
Total
. . .
Percentage distribution 1
of employees
Number of
employed
insured
workers
(July 1939)
(June 1939)
687,122
859,492
200,956
414,856
208,777
629,876
6,539,700
640,300
865,143
539,000
846,000
238,370
1,145,457
6,556,030
4.7
5.9
1.4
2.9
1.4
4.3
45.3
3.7
5.0
3.1
4.9
1.4
6.7
38.2
1,192,349
1,142,320
8.2
6.7
3,748,573
3,614,184
1,590,000
25.9
21.0
9.3
14,481,701
17,176,804
100.0
100.0
Estimated
total
employees
Insurance
data
Estimates
1
1
* Data on the insured population were taken from the Ministry of Labour Gazette
(December 1939) ; on the unemployed, from the Ministry of Labour Gazette (August 1939);
on manpower, from the Monthly Digest of Statistics (January 1947).
To provide a reasonable degree of comparability, the distribution of employed persons
according to the manpower estimates was reduced by a deduction of the estimated number
of self-employed persons. This was done by assuming that the ratio of self-employment
to total employment in each major industry group was the same in 1946 as in the census
of 1931. Data from the 1931 census volume on Industries then made it possible to deduct
the estimated numbers of self-employed in 1946. There was probably a greater percentage
of self-employment in 1946 because of the differential age of self-employed and employee
and the consequent effect on calls-up for the Services. Per contra, self-employed persons
undoubtedly left construction, trade and service for war industry. It is therefore assumed
that these influences cancelled out—an improbable conclusion, but acceptable for the present purpose. The domestic service estimate is that of FRANKEL, based on health insurance
registrations, and is exclusive of domestics covered by unemployment insurance (Journal
of the Hoyal Statistical Society, 1945, p. 412). A related estimate of non-covered domestic
service is that of KALDOB, which runs to 1,200,000 for June 1939 (U.S. Strategic Bombing
Survey, The Effects of Strategic Bombing on the German War Economy, 1945, p. 215).
Cf. also, The Manchester School (September 1946, p. 34).
persons as well. In addition, as the Danish Government
emphasised in reply to a specific query: " t h e registration of
unemployed uninsured persons cannot be considered as a satisfactory measure of the unemployment among persons not insured
against unemployment; neither from an occupational nor from
a geographical point of view can this group be considered
as representative of the uninsured unemployed worker".
U N E M P L O Y M E N T : SAMPLE
93
According to the British report the latest data available on
the distribution of all employees are for the year 1931, the
registration data for 1939 being apparently inadequate. Therefore no very satisfactory analysis can be made of even the industrial composition of the British sample. However, a rough
comparison can be made between persons in insured employment in June 1939 and an estimate for total employees on the
same date (table V). The distribution of the sample is much
the same as that of the parent population, though the difference
for manufacturing and domestic service is noticeable.
The insurance system excludes from coverage most domestics,
part-time workers (workers employed less than 30 hours),
older workers (men over 65 and women over 60), and a substantial number of those permanently employed by governmental
authorities. Lesser exclusions include non-manual workers
earning more than £420 a year, professional nurses and certain
classes of casual labourers. Self-employed and own account
workers are, of course, also excluded.
2. General Problems]
To assess the overall sufficiency of unemployment insurance
reporting as a method for getting statistics on unemployment,
it must be borne in mind that any unemployment figure tends
to assume a character by reference to other unemployment
figures. An unemployment rate of 8 per cent, may be considered
tolerable or cause for sudden alarm according to whether the
customary rate over the previous decade was 3 per cent, or
20 per cent. In the former instance the problem has grown
more serious. In the latter, its amelioration is evident. If,
therefore, a constant level of exclusion — by age, industry, etc.—
could be maintained, then even an obviously biased sample
would give a result which would be satisfactory for many purposes.
Where coverage of the social insurance system is full
and wide-ranging — as in Britain, for example — it closely
approaches such a constant level of exclusion. In such a case,
departures from a constant level of exclusion will have only the
most trivial effects on the national record of unemployment. On
the other hand, coverage which is clearly much inferior may
produce a tangible and dominant bias. This bias in turn will
94
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
give a distorted perspective of the course of unemployment, because there is no single standard rate of recruitment of workers
to all industrial, age and geographical groups. Some industries
are still growing, in the full flush of youth, others are at the
peak of maturity, and still others are declining. So too are
industrial areas. Migrations from farm to city may be heavy,
and shifts from uninsured to insured industries, and back again,
may be constant. Entrance to and retirement from the labour
force alone will give many a country an essentially new labour
force every two or three decades. And each of these influences
will bear continuously upon the nature of the unemployment
insurance sample, changing its coverage and eventually distorting the comparability of today's unemployment estimates with
those of yesterday.
Superimposed on these factors are those which relate to the
changing dimensions of the social insurance system. As the
Belgian report notes, the data obtained from unemployment
insurance operations "did not, until a fairly recent date, inspire
any great confidence ", a substantial change occurring with the
advance in coverage when the National Office of Placement and
Unemployment was set up some years ago. Coverage was
further extended when, after the war, it was decided that the
right to receipt of benefits would no longer require a qualifying
year's membership of the funds, and when considerable extensions in industrial coverage were made.
The Swiss unemployment funds increased their coverage
of workers markedly in the early years of the depression.
Significant changes have occurred in the British data too, when,
successively, the general scheme for non-agricultural industries
was introduced in 1920, older workers were excluded in 1928,
and agricultural workers included in 1936.
Because of important changes like these in the scope of
social insurance systems, it is difficult to use their results as
measures of change in the severity of unemployment as the
systems improve their coverage. When a system achieves
complete or practically complete coverage, this difficulty will,
of course, vanish.
It is none the less true that improvements in reporting can
be accomplished. If at periodic intervals an assessment of the
total labour force is made, it then becomes possible to assess
the changing representativeness of the sample. Given this
information, it is even possible to compute a synthetic but
UNEMPLOYMENT:
COVERAGE
95
superior unemployment series. This would be derived by
weighting unemployment percentages not by the sample distribution but by the total distribution of workers.
IV. LABOUR FORCE SURVEYS
In most countries the sample currently used for reporting
unemployment excludes certain major groups from its coverage.
Since the characteristics of these excluded groups often differ
from those of the groups included, and since the proportion of
the excluded groups to the total will vary from year to year,
periodic complete enumerations of the labour force are required
to check on the adequacy of the sample employed. It is therefore recommended that a population census should be taken at
least every ten years in order to provide labour force data
necessary to improve the accuracy of the system used for sampling unemployment as well as for other purposes (I, 12). The
census, together with other surveys, should be used to construct
an adjusted index of total unemployment where this is feasible
and where the existing indices are biased by substantial exclusions.
XIII. UNEMPLOYMENT: COVERAGE
In the previous chapter, considerable space was devoted to
measures of changes in unemployment, the users of such statistics frequently being concerned only with changes in unemployment. But the volume of wasted manpower as such is often
the point of concern.
For example, a dearth of manpower is now faced by many
countries which are in the midst of reconstruction. It may
therefore be asked by how much actual unemployment exceeds
an estimated factional minimum. An answer to this question
is prerequisite before it can be stated, for example, that no more
manpower is now available, or is likely to become available, for
making goods or providing services for the home market.
A measure of the total volume of unemployment is likewise
required when two countries are to be compared in regard to
the time lost in unemployment. A comparison of unemployment in Canada, where all persons in the labour force are
96
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORGE
surveyed, with unemployment in Australia, where reports cover
only trade union members, will be vitiated by differences in
measurement. And finally, in order to ascertain whether the
unemployment problem in a particular country has grown in
intensity, a comparison between different years will be difficult
and inconclusive where the extent of coverage has changed.
(This would tend to apply, for example, to a comparison between
British unemployment in 1927, when agricultural workers were
not insured and older workers were, and unemployment in
1939, when the figures covered agricultural workers but excluded
older workers.)
I. EMPLOYMENT EXCHANGE STATISTICS
l
It is fairly clear that employment exchange statistics cannot
provide a satisfactory estimate of total unemployment at any
given time. Not only do the figures include some persons who
may actually be at work, but also they exclude unemployed
persons who have not registered with the exchanges. The net
result of these opposed biases cannot be accurately determined
except at the times when a complete census is made. For a
current measure of the total volume of unemployment, statistics
based on registrations of employment exchanges are therefore
generally unsatisfactory.
II. TRADE UNION STATISTICS
Statistics of unemployment among trade union members may
furnish reasonably accurate indications of the trend in unemployment, but they can obviously provide no adequate measure
of the total volume of unemployment at any given time.
III.
SOCIAL INSURANCE SYSTEMS
Statistics of unemployment among persons insured against
unemployment tend to fall short of giving a complete measure
of the number of unemployed persons, for several reasons.
(Whether the sum of these exclusions is of substantial importance in any given country will, of course, require special study.)
Among the chief exclusions or omissions are the following:
1
As above, this section refers to statistics of employment exchanges
which are not an integral part of a social insurance system. For a
discussion of the statistics of the social insurance systems, see below.
U N E M P L O Y M E N T : COVERAGE
97
1. Most social insurance systems exclude certain occupational
or industrial groups from coverage, among them such important
occupations as domestic service and such important industries
as shipping.
2. A second group of persons who are frequently excluded
from the social insurance system and hence not included in the
estimates of the total volume of unemployment are the longterm unemployed, unemployed older workers or classified
unemployables. It is, of course, wholly proper for such groups
to be excluded from the numbers of insured unemployed, but
they should nevertheless be included in the estimates of the
total volume of unemployment, provided they satisfy the basic
criteria of unemployment — namely, that they are without
work, are seeking it, and are able to take a job if offered one.
3. A more important deficiency is the de jacio omission of
insured persons who do not register as unemployed, even though
they may actually be unemployed. Persons who have exhausted
their benefits or who are not yet eligible for benefits, and those
who expect to get a job before they could begin receiving
benefits, are likely to be in this group of persons omitted from
the unemployment total as reported by social insurance systems.
4. The numbers registered as unemployed under social
insurance systems are likely to fail to cover two other groups.
The first of these consists of new entrants to the labour market
who have not yet obtained their first job — more specifically
their first job in insured employment. The second includes
older workers above the insurable age who are none the less
attempting to find work.
5. A more serious aspect of these limitations, however, is
that they tend to make the reported unemployment total a
different proportion of the true total in depression years than it
is in those of prosperity. Hence the comparison of these data
over a period is complicated by differences in the degrees of
completeness with which unemployment is measured. During
depression years the entrance of married women into the labour
market is a characteristic phenomenon, as is the re-entrance of
older workers. Because these workers are clearly not eligible
for unemployment benefit, having had too brief a period of
contributions, they are less likely to register as unemployed.
They are also less likely to be members of the insurance system,
since they have had no recent job in insured employment. The
98
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
rising tide of such entrants to the labour market will be partially
concealed by the exodus from the labour force of those married
women and older workers who find themselves blocked in finding
employment. But the net result of these factors will be to
understate the quantum of unemployment.
This bias during depression is contrasted with an opposite
tendency at work during years of prosperity. As was clearly
demonstrated during the war, employment in manufacturing,
transportation and other industries covered by unemployment
insurance becomes a greater proportion of total employment
during high prosperity. But at the same time the proportion
in such industries as trade and service, where insurance coverage
is restricted, will decrease. Since a larger proportion of those
employed are then included in the insurance system a larger
proportion of total unemployed will be recorded. Hence
unemployment estimates based on the operation of social
insurance systems have a double tendency. The first is to
produce a noticeable underestimate of unemployment in depression years. The second is towards a more complete statement
of unemployment in times of prosperity.
6. Because insurance systems frequently exclude farm
labourers, domestic servants and other workers, these systems
tend to understate the unemployment of persons who are
actually members. None of the excluded groups constitute
a wholly independent work force. Each includes workers who
were once employed in covered industries and are no longer able
to find work in them. These workers find temporary employment in domestic service or seek work on the land during depression periods.
Failure to cover these groups is then a failure to acknowledge
the unemployment of those workers who are most subject to
debilitating unemployment just at the time when they experience it. The more competent, more successful, more fully
employed workers are likely to remain in employment. (They
will at least remain attached to the industry where they have
been employed.) But the marginal worker, most subject to
unemployment even in prosperity, and certainly so in depression
years, is more apt to be forced away from the industry. After
seeking work for some time in textiles or in the iron and steel
industry, the marginal worker may shift to casual labour, to
dock work, to petty self-employment in trade or personal service. He may even migrate to a rural area. Hence unemploy-
UNEMPLOYMENT: COVERAGE
99
ment as reported by insurance systems with less than complete
coverage will always be less than the true volume of unemployment. In estimating unemployment in the industries which
they cover, they include the experience of unemployment-prone
workers only so long as they remain in these industries, but
ignore their unemployment after they have been forced into
other industries.
That biases exist in the estimates, however, does not imply
that they are substantial. There is probably a general tendency
for figures based on the unemployment returns of even the most
comprehensive unemployment insurance system to understate
the total volume of unemployment at any time. There is probably
a further tendency at work for the underestimate to be greater
in years of depression than in years of prosperity. But exactly
how important these various tendencies may be in distorting
any given unemployment series cannot be determined in the
absence of special studies.
Comprehensive studies on the nature of the labour force,
employment trends, and the actual administration of the
insurance agency, are needed to determine the importance of
these omissions. It is therefore recommended that in every
country which relies on unemployment insurance statistics for
its chief measure of unemployment, periodical studies should
be made of the relation between the total number unemployed,
on the one hand, and the number unemployed as reported by
the insurance system, on the other (I, 40 (1)). Particular
attention should be given to including in the estimates of total
unemployment such groups as the following, which may be
excluded from or incompletely covered by the unemployment
insurance statistics : older workers ; those who have been unemployed for long periods of time; those not applying for benefits,
even though they are unemployed; new workers who have not
yet found their first jobs; supplementary workers who enter
the labour force at the peak or in the trough of the business
cycle; and industries and occupations specifically excluded
(I, 40 (2)).
As resources and facilities permit, it is recommended that
periodic estimates of the total volume of unemployment, in
contrast to the volume of insured unemployment, should be
computed on the basis of these assessments and published in
addition to the customary series on the insured unemployed
(I, 44).
100
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
IV.
LABOUR FORCE SURVEY STATISTICS
As organised in the United States and Canada, the labour
force survey measures unemployment without restrictions or
exclusions affecting industry, occupation, age or status group.
However, according to the definition adopted, only those are
included as unemployed who have been without work for at least
the inclusive period from Monday to Saturday in the enumeration week. Hence, a certain proportion of those out of work
are omitted solely because their period of unemployment up to
Saturday had lasted less than one week. In order to throw light
upon the importance of these exclusions in relation to the total
volume of unemployment, it is therefore recommended that
special studies should be made of the proportion of those out
of work who are thus excluded. *
1
An idea of the importance of these exclusions can be formed by
taking five times the average daily number of new cases of unemployment. Data for the United States over the period June 1941-May 1942
indicated that the daily accessions to unemployment—assuming a
300-day work year—averaged 48,000 in that year, that is, indicating
some 240,000 persons omitted from the total. This figure would of
course vary with the volume of unemployment as shown in the daily
accessions. A percentage estimate of omissions is therefore of greater
significance. During the same period the total number of unemployed
averaged 4,255,000, indicating therefore an understatement of total
unemployed by some 5.6 per cent. However, this percentage figure
would understate the " normal " percentage, since, during this period,
the size of the unemployed group was rapidly diminishing and the
number of new accessions was not sufficient to maintain its strength.
A true comparison to obtain a normal percentage of omissions is to
compare the estimated omissions based on five times the daily accessions
to the total number of unemployed some three months later, to allow
for the average duration of unemployment and the time lag between
the occurrence of new cases of unemployment and the total volume to
which they contribute; this comparison would raise the figure to 7.0 per
cent, as an average percentage of omissions. During periods of rapidly
increasing unemployment the percentage of omissions would be greater.
In terms of unemployment as a percentage of the civilian labour force,
the correction noted above for the period June 1941-May 1942 would
raise the figure from 7.0 per cent, to 7.4 per cent, of the civilian labour
force unemployed.
PART IV
XIV. CONSISTENCY
As the demands laid upon the national statistical services
for employment materials have multiplied, a variety of statistical
series have been developed to meet these demands. The result,
particularly in countries possessing the most active and energetic
statistical services, has been the development of series of which
some may be in fact or appearance at odds with one another. It
is proposed to discuss this problem as it affects two countries
where statistical data are most copious and most accurately
calculated, the United States and Canada.
For the United States, estimates of employees in non-agricultural establishments are available from both the Bureau
of Labor Statistics and the Bureau of Foreign and Domestic
Commerce. A comparison of the data for recent years showed
substantial differences in the level of employment estimated
for major industry groups by the two agencies until recent
revisions have resulted in substantially identical figures. For
example, the estimates of employment in a number of industry
groups by one agency were from 5 to 10 per cent, greater than
those estimated by the other. However, recent revisions have
resulted in substantially identical figures. 1
A sharp difference is found between the trend of non-agricultural employment as estimated by the Bureau of Labor
Statistics and that based on estimates of the Bureau of the
1
For one industry the estimate of one agency was 50 per cent.
greater than that of the other. The contrasting movement of the
unrevised figures may be observed in data presented in the Survey of
Current Business, for November 1945, and the Monthly Labor Review
for May 1945.
102
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
Census' Monthly Report on the Labor Force. The movement of
the two series over a recent 12-month period was as follows 1:
B.L.S.
100.0
101.3
July 1945.
July 1946.
B.C.
100.0
107.0
Since the Bureau of Census figures used for this comparison
represent wage-earning or salaried workers in non-agricultural
employment and the Bureau of Labor Statistics figures represent
wage-earning or salaried workers in non-agricultural establishments, the contrast is of particular interest.
The Canadian data exhibit a parallel contrast. Table I
presents indices of employment as estimated by the Dominion
Bureau of Statistics', monthly labour force survey and by its
TABLE I. INDICES OF EMPLOYMENT IN CANADA AS ESTIMATED
BY THE LABOUR FORCE (L.F.) AND MONTHLY EMPLOYMENT
(M.E.) SURVEYS: 1946
1 Nov.1 Dec.
1946
Industry
1 March 1 June
1 Sept. Average
M.E.
L.F.
100
100
101
105
103
101
106
111
M.E.
L.F.
100
100
57
39
57
69
96
123
M.E.
L.F.
100
100
103
94
102
89
103
92
M.E.
L.F.
100
100
104
102
110
106
111
109
M.E.
L.F.
100
100
144
170
169
187
167
170
6. Trade and finance . . .
M.E.
L.F.
100
100
104
109
107
115
113
117
T o t a l : Industries 1-6 .
M.E.
L.F.
100
100
102
106
105
108
110
115
1. Manufacturing
2. Logging, fishing and t r a p ping
4. Transportation and com-
1
Bureau of Labor Statistics estimates from B.L.S., Employment and
Payrolls, February 1947. Bureau of the Census estimates from the
Labor Force Bulletin, April 1947. The data used for computing both
B.L.S. and B.C. indices exclude domestic servants.
CONSISTENCY
103
Employment and Payroll Statistics Branch. 1 Since both series
relate to changes in wage-earning and salaried workers employed,
they should show parallel movements. But the rise in manufacturing employment between March and June 1946, for example,
was 5 per cent, in one series and 1 per cent, in the other. For
mining over the same period employment rose 3 per cent.
according to one series and fell 6 per cent, according to the other.
Most striking, perhaps, is the change in employment in construction over the same period, rising by 44 per cent, according to
one series and by 70 per cent, according to the other.
The divergences between the changes in employment indicated by the establishment sample and labour force surveys arise
from a number of causes. One major factor may be differences
in coverage.
For example, the Canadian Monthly Survey of Employment is based on reports from firms with 15 or more employees,
while the labour force surveys cover all persons in employment
regardless of size of firm, Hence the two series are likely to
show different results for such industries as construction, trade
or service in which a substantial proportion of persons employed
are in small firms.
These sharp contrasts are to be attributed partly to the fact
that one series covers a wider employment field than the other,
and partly to differences in sources of data : one series proceeds
from establishment enquiries, the other from direct queries
addressed to members of households in their homes.
Data from establishment samples and population surveys
frequently differ substantially. In labour force surveys the
information is usually obtained from housewives. The tendency of housewives to forget incidental employment and to
overlook the employment of supplementary wage-earners will
probably persist. In this respect, results secured by questioning
the housewife may be inferior to data obtained from establishments — though it should be added that payroll returns would
omit incidental employment performed outside the establishment. But experience has indicated that the effect of these
differences in approach in producing differences of estimate
can be minimised.
A further element in the improvement of population census
survey data which is necessary in order to make them consistent
1
Data kindly supplied by the Dominion Bureau of Statistics (unpublished).
104
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
with establishment figures is a change in the sample design,
where necessary, to produce estimates of the employment of
wage-earning and salaried workers in major industry groups.
The total number of persons employed, inclusive of employers
and persons working on their own account as well as wageearning and salaried workers, classified by industry, is currently
reported by the Canadian labour force survey, but not the
number of wage-earning and salaried workers employed. The
total number of persons employed, classified by industry, is
available from the United States survey but is not released.
Since the sample was originally designed to produce a satisfactory
total for employment, emphasis was placed on securing data
from a variety of industries and not on the accurate representation of each major industry group. It is therefore possible
that changes in the sample will be necessary before data for
such groups are sufficiently trustworthy for release. x
The improvements which must be made in establishment
sample data have already been touched upon. The movement of employment must be analysed with reference to the
critical factors which affect it, for example, distribution by
industries, and by size-of-firm groups. Adjustments for trend
bias must be made, as necessary. And finally, the double
counting which arises when a worker is reported by two establishments for a given payroll date, having worked part of the
period in each, must be examined. (From the analysis of labour
turnover it may be possible to construct a distribution of workers
by length of stay on the job when the turnover rate is at a
given level. Given such a distribution, the turnover rate will
afford the means for adjusting for this possible double counting.
Labour force surveys of multiple employment may serve this
purpose equally well.)
These adjustments should eliminate most of the differences
between establishment sample and population survey data 2,
and estimates of the number of non-farm wage-earners from
both establishment and population census reports would then,
for practical purposes, be identical.
1
For example, employment in the service group fell sharply from
1941 to 1942, but immediately regained its 1941 level in 1943.
2
Such remaining sources of differences as the use of the week containing the 8th of the month rather than the payroll period ending nearest
the 15th; slightly different treatment of persons "employed " but "not
at work "; different likelihood of inclusion for persons who worked only
a few days in the schedule week—these and other remaining sources
of difference appear to be relatively unimportant.
CONSISTENCY
105
What is the point in a reconciliation which, after arduous
efforts, would apparently only succeed in reconciling a total
based on an establishment sample with an originally more satisfactory figure based on a population survey? The answer is
that if the only series required were for total employment or
wage-earner employment no such reconciliation would be
required. But in point of fact a mass of employment estimates
for detailed manufacturing industries is needed — estimates
which must be comparable with data on earnings, output and
turnover. At the present time it would require a prohibitive
expenditure to arrive at accurate estimates of employment in
detailed industries via the population survey technique. There
is even some question whether such an achievement is possible
so long as the housewife is the chief source of information.
To provide estimates of total employment and unemployment
as well as estimates of employment in detailed industries at a
reasonable cost, both systems appear to be necessary.
In order to secure consistency, a number of suggestions for
improving the different systems have already been discussed.
As a final suggestion a semi-annual check into the relationship
between the two series along the following lines might be
desirable. When the schedules for the labour force survey are
being filled up, additional questions should require the name and
address of the firms where wage-earning and salaried workers
are employed to be specified. 1 If transcripts of these schedules
are promptly made available to the agency collecting the establishment data, a direct check of establishment payrolls could
be instituted to determine the actual industry and occupation
of the workers concerned. This check would not only furnish
a test of the accuracy of replies by housewives on industry and
occupation, but would also provide a statistical bridge for connecting data on the employment of wage-earning and salaried
workers in non-farm pursuits obtained from establishments
with the data from population surveys.
By means of such a bridge, the detailed industry estimates
of the establishment reports could be linked to the controlling
overall totals of the population surveys. Once this has been
1
To avoid the development of undue resistance, such questions
might be asked only of families which are being dropped from the survey
in the current months. An alternative, though less satisfactory, procedure
would be to ask whether or not each adult was insured under t h e unemployment insurance system. Those t h a t were could then be checked
from the unemployment insurance files.
106
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
done, a full range of employment series could be issued, each
consistent with all the others. This would mark a substantial
advance over the present situation, in which the series are too
often non-comparable, and sometimes even appear contradictory.
As resources and facilities permit, it is therefore recommended
that, in countries where overlapping employment series are
available from different sources, periodic surveys should be
made of a sample of returns from each source, in order to eliminate any significant double counting or inconsistencies that may
exist, and to lay the foundation for a unified set of employment
series (I, 32).
XV. INTERNATIONAL COMPARABILITY OF DATA
I.
PURPOSES
As the economies of the world have become more and more
closely linked, the need for comparable data on the labour force
and employment in the various countries has become more and
more evident.
International comparability is essential for sound comparisons of persons employed in different industries in the different
countries, not only to measure relative employment levels, but
also to provide a rough measure of the relative importance of the
different industries in each country. How large a proportion of
the wage-earners in the major industrial countries are employed
in the steel industry, in textiles, in coal mining ? What are the
relative proportions of employment in road as against rail
transport ? Sound answers to such questions require comparable
data.
Such data are also needed for making comparisons between
productivity and unit labour cost in the various countries. It is
impossible to compare the efficiency of the cotton textile, the
coal or the steel industry in one country with that in others, or
to compare relative labour costs, without employment data
that are actually comparable. Otherwise the differences
revealed by a comparison may testify not to any real differences
in productivity or cost but to differences in the basic employment statistics.
INTERNATIONAL COMPARABILITY OF DATA
107
To measure the relative severity of unemployment in the
different nations it is necessary to have comparable unemployment estimates. During the great depression, for example, it
was impossible to tell whether country A was affected more than
country B, simply because the coverage of the unemployment
figures was usually different and the definitions not comparable.
The availability of comparable data will preclude such crude
international comparisons as those made in a recent study
which computed the percentage unemployed in various countries
by simply relating the number of reported unemployed, whatever the coverage, to the total number of gainfully occupied,
whatever the definition. 1 The availability of comparable
data will likewise make it possible to set up soundly based
indices of the trend in world unemployment.
While it is, of course, recognised that each country compiling
statistics of employment, unemployment and the labour force
is primarily concerned with its own problems, similar methods
and techniques of statistical analysis of the problems of full
employment, for example, should be applicable in the different
countries. An exchange of information and the utilisation of
available statistical resources, for example, in sampling techniques, would tend to promote the development of useful statistics
on employment and labour force in the different countries.
*
The points of chief concern for purposes of international
comparisons of employment, unemployment and labour force
statistics are common definitions, common classifications, and
common methods.
With common definitions and common classifications, a long
step will have been taken towards making possible the collection
of comparable data in different countries. Data collected on
the basis of uniform definitions should not only serve the purposes of the country itself but should also enable other countries
or international bodies to use the data. The analysis of bias
would make it possible to interpret the results of various series
in different countries in terms of full coverage, and thus lead
to the establishment of synthetic series of a high degree of
comparability.
1
Anuario
Estadístico
de España,
1943, p. 1469.
8
108
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
So far as definitions are concerned, agreement upon the fundamental purposes served by employment and unemployment
statistics should lead to agreement in principle on the basic
definitions. These have been discussed in an earlier chapter.
Where the working definitions developed in specific fields of
employment, unemployment and labour force statistics diner
from these basic definitions, consideration can be given to the
question of estimating the biases arising from these divergences.
As a result it should be possible to make estimates of employment and unemployment in each country on the basis of standard
definitions.
With regard to classification, the details to be furnished in
regard to sex, age, marital condition, and industrial status,
have been discussed in previous chapters in relation to various
specific proposals. It remains to consider certain problems
of standard industrial and occupational classification.
II.
INDUSTRIAL CLASSIFICATION
The most widely accepted international classification of
industries is that which was recommended by the Committee
of Statistical Experts established under the International Convention relating to Economic Statistics. 1 That classification,
published in 1938, provided a nomenclature of 94 industries,
arranged under 8 major divisions and 45 main industries. In
some countries, marked advances and changes in national
classifications were made for the various 1940 and 1941 population censuses and for special registrations during the war years.
The subject of the classification of industries is to be reviewed
by the Statistical Commission of the United Nations ; it is hoped
that their deliberations will eventuate in a standard international
list. To improve the international comparability of employment and labour force data, pending the adoption of such an
international list, it may be recommended that the countries.
should follow provisionally the list proposed by the Committee
of Statistical Experts and should provide such supplementary
details in their national classifications as to permit re-grouping
of the data in accordance with the standard list.
1
LEAGUE OF NATIONS, Studies and Reports on Statistical Methods,
No. 1 : Statistics of the Gainfully-Occupied Population.
Definitions and
classifications recommended by the Committee of Statistical Experts
(Official No. C.226M.128.1938 (C.E.S.127), Appendix I, Ser. L.O.N.
P u b . II Economic and Financial 1938, II.A.12).
INTERNATIONAL COMPARABILITY OF DATA
109
Countries with detailed classifications already established
are likely to be reluctant to make major changes in classification.
However, in practice, the countries with the most developed
systems of their own will have least difficulty in reclassifying
their data into whatever classifications are necessary to produce international comparability. It is therefore recommended
that in the more industrially diversified countries consideration
should be given to the desirability of presenting data on employment, unemployment and the labour force in at least as great
industrial detail as that specified by the international classification.
III.
OCCUPATIONAL CLASSIFICATION
At the present time no internationally accepted standard list
of occupations exists. In many countries occupations are
grouped into those requiring special skill or training, those presumed to be peculiar to specified industries (i.e. manufacturing
occupations, trade occupations) and general groups, such as
clerical workers and labourers. Professional and semi-professional workers may be shown in special groups. Since the
manpower shortages which have developed in many countries
have required detailed information on the occupational backgrounds of persons in the labour force, the time seems ripe for
proposing the consideration of a standard international classification of occupations. The difficulties in the way of a standard set of occupational classifications are great. For detailed
comparisons acceptable definitions in terms of job analyses
and job specifications may be necessary, in order that the
same term should mean approximately the same occupation
under different industrial practices in different countries. Broad
groupings may be needed. The purposes of such classification
both for national and for international uses must be served.
In view of the fact that population censuses will be taken in
most countries in 1950 or 1951 the drafting of a standard
classification in time for use in these censuses would be of
great value. The Conference may wish to consider whether to
recommend that the subject of international occupational
classifications should be placed on the agenda of a future
international Conference (III).
Pending the adoption of standard classifications the most
effective procedure to improve international comparability
110
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
would appear to be for each country to provide sufficient detail
according to industry, age, industrial status and occupation in
its own statistics so that the data can be recombined into the
standard international classifications.
* * *
The methods to be followed in collecting and analysing
employment, unemployment and labour force statistics have
been discussed in the previous chapters. The adoption by the
different countries of the best methods available with reference
to each type of data will naturally tend to the establishment of
international comparability.
A further important step is the provision of analytical
studies of methods and techniques followed in the different
countries. This would include a survey of existing definitions
and methods to furnish the basic information on the divergences
that exist in the practices of the different countries. It would
also include the evaluation and appraisal of these divergences
in relation to standard definitions and techniques, as well as
an appraisal of biases and margins of error as they affect the
statistics of employment and unemployment in each country.
On the basis of such studies, international comparability of
employment and unemployment data should be placed upon a
much firmer foundation. Each series could then be utilised in
the full knowledge of its probable margin of error or bias.
As a final result, it should be possible to take the statistics
of the different countries as established upon uniform definitions and following uniform classifications, subject to such corrections as might be indicated by these analytical studies, and
combine the figures from different countries to set up soundly
based indices to show world trends in employment, unemployment and the labour force.
XVI. PUBLICATION
Not the least of the many elements which constitute a
sufficient statistical system is that of effective publication, and
the first aspect of effective publication to be considered is the
speed with which data are released once they have been gathered.
PUBLICATION
I.
PROMPTNESS
OF
111
PUBLICATION
Whatever the frequency with which data are compiled, their
value will be seriously compromised if they are published a long
time after the date to which they relate. Quarterly statistics
published within a month of the date to which they refer will
be of greater value to most users of employment statistics than
monthly statistics published after a delay of several months.
Business men must make their plans for engaging labour and
must forecast prospective market changes on the basis of the
most recent information on employment variations. Belated
information may be of little more value than none. A "reasonably approximate figure provided promptly is often more useful
than a more accurate figure available some weeks or months
later ". l Unions, too, require quickly available measures of
changes in the labour market. Such changes do not merely
affect the welfare of their members. They will also affect the
relations of each union with management, since different wage
rate and bargaining policies are appropriate to different levels
of employment and unemployment. Government agencies have
an equally vital need for accurate data promptly issued, to
throw light upon the effects of changes in national policy on
employment and unemployment.
An illustration of prompt publication of employment data
is found in the Danish figures, the estimate for total man-hours
worked in manufacturing being published about 3 weeks after
the date to which it refers. In Great Britain and the United
States detailed estimates of employment in manufacturing subgroups and in other industry groups are issued about 4 weeks
after the date to which they relate. Sweden issues a full range
of detail about 5 weeks after the date to which the estimates
refer. In Australia the period is from 6 weeks to two months,
while in Canada the period is about 6 weeks for labour force
detail and two months for wage-earner data. The employment
series which was recently organised in France appears about
three months after the date of reference.
Since most users of the statistics of employment and unemployment require prompt publication of data almost as much as
they require the data themselves, it is therefore recommended
1
U.S. CHAMBER OF COMMERCE, Committee on Business Statistics:
Public Interest as a Criterion for the Collection and Dissemination of
Statistics by Government Agencies (1946), p . 10.
112
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
as desirable that the key totals in current series on employment, unemployment and labour force should be issued within
one month of the date to which they refer; if necessary these
data may be issued in the first instance on a provisional basis
(I, 50).
The wider use of provisional figures is to be recommended.
Denmark, for example, customarily presents unemployment
estimates on a preliminary basis, revising these estimates at a
later date. Experience over a period of years indicates that
the preliminary figures are in fact substantially identical with
the final ones. It is obviously not feasible to provide full data
for a finely detailed set of industries or areas immediately, but
neither is it necessary. What is essential is that the key totals
should be presented promptly, so that business, labour and
Government can at all times have as accurate a perspective as
possible of the changes in employment and unemployment.
The method to be used for securing preliminary figures will
vary with the reporting system and national procedures. For
employment data from establishment returns it may be possible
to establish a list of firms whose employment, taken as a whole,
moves with national employment. Alternatively, it may suffice
to set a cut-off date for a preliminary tabulation. For social
insurance data a full knowledge of the industrial and sex composition of returns from the various districts, and in particular the
districts which report earliest, will be needed. For labour force
survey data national totals may be released after the preliminary tabulation of the sample returns is made but before the
adjustments to independent population estimates are completed.
The adoption of such a procedure will of course depend on the
extent of change in the unemployment percentage usually produced by such adjustments.
In general it may suffice to release preliminary estimates in
the form of estimates of percentage change from the previous
month, thus minimising any inconsistency between the aggregates as reported in preliminary and final results.
II.
PRESENTATION
In the presentation of labour market data attention should
be given to a number of factors bearing on the effective use of
these statistics.
PUBLICATION
113
1. Data as published should be accompanied by a statement
of their scope (I, 51 (1)). Employment series based on data
from large firms only or from a limited group of industries
should be so described; such series do not necessarily have the
same movement as that of total non-agricultural employment or
total employment. These different series should not be confused or used interchangeably. Where the chief unemployment series in any country relates to the insured unemployed,
current statistics in that series should be accompanied by a
brief statement indicating what percentage of the unemployed
were not covered by the series at the date when the most recent
comparison was made between the number of insured unemployed and the total number of unemployed (I, 51 (2)).
2. In each country the statistical authorities should provide
a brief guide to the major benchmarks and series on employment, unemployment and the labour force. This guide would
not be intended for the use of specialists in employment statistics, but rather for the use of the general public. It would
therefore enumerate the series, indicate their frequency and
place of publication, and describe how the basic data for each
series are derived and combined. It would, furthermore, indicate the chief uses to which each series might properly be put
and point out pitfalls arising from improper use. Thus, certain
employment series would be indicated as most suitable for productivity computations and others for combination with
unemployment data to measure labour force (I, 53).
3. In continuation of this basic publication, an annual
publication should be issued containing the major series on
employment, unemployment and the labour force for current
months and for selected benchmark dates. References should
be given to the original sources, in which more detailed
information may be found, and a discussion should be included
on the major changes in employment and unemployment
which have taken place since the previous issue (I, 55). Each
country should present in a basic publication all available
historical estimates of employment, unemployment and labour
force, with full particulars of the sources for the different
series of these methods used in compiling them, and of their uses
and limitations (I, 54).
4. Where more than one series is published which can be used
as an indication of the level or course of employment or unemployment, each series should be accompanied by a statement
114
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
indicating its chief differences from the others and by an indication of the most appropriate uses for each (I, 52). As examples
may be cited the employment exchange, trade union and labour
force survey series on unemployment in Canada, and the Bureau
of Labor Statistics, Bureau of Foreign and Domestic Commerce
and Bureau of the Census series on employment in the United
States.
5. A historical perspective of employment and unemployment trends is invaluable. Figures on employment or unemployment take on meaning when contrasted with the figures for
earlier months or years. Such questions as "What is the frictional minimum of unemployment ? Is unemployment greater
today than it was after the last war? " can only be answered
by reference to historical data. Because of the considerable
changes in calculating techniques, social insurance coverage and
the scope of establishment reporting that have taken place in
most countries, it is often difficult to secure comparable estimates which can alone throw light on these questions. As
resources and facilities permit, it is therefore recommended
that comparable historical estimates should be prepared for the
chief series on employment, unemployment and labour force,
these to extend over a span of years which will include both the
peak and trough of the business cycle, extending back to 1929
and, if at all possible, to 1919 (I, 56).
XVII. GENERAL ISSUES
In addition to the various aspects of employment and unemployment statistics already reviewed, certain general issues
should be considered.
In the first place, data complementary to the statistics of
employment and unemployment are required. Reference has
already been made to the point that employment data must be
analysed in the light of historical trends. Current changes in
employment, unemployment and the labour force must be
placed in relation to the background data of population structure and industrial development. There must be available
estimates of hours, earnings and labour turnover, of prices, of
production and productivity. Only in the context of current
information on these related matters can employment and
GENERAL ISSUES
115
unemployment estimates be properly interpreted. It is surely
not too much to recommend that a population census should be
taken every ten years, and a census covering the major branches
of economic activity at least every five years (I, 12).
A second point to be emphasised is the basic limitation of
statistical averages and global totals. Estimates of total
employment or hours worked or unemployment are essential,
but they are not sufficient. Current information on distributions as well as totals is requisite. What, for example, is the
distribution of workers by the age when they enter employment and the age when they leave i t ? What is the distribution
of workers by the number of jobs held during a given year, and
by the duration of each job — according, separately, to age,
occupational group and sex? What is the distribution of
workers according to the number and type of industries in
which they worked during the previous year? What is the
distribution of workers by earnings, both hourly and annual ?
What proportion of total man-hours worked each year is
contributed by each of the major age and sex groups ?
Answers to these questions, and to many more, require data
on distributions of workers rather than simple summary averages. But the forecasting of business markets, the study of
labour shortages, manpower problems — these and many other
studies can be effectively prosecuted only if employment and
labour force data are available at intervals for distributions
of workers.
A third point worth emphasising is the desirability of a
periodic review of procedures. This is desirable in order to
ensure that the statistics of employment will continuously
improve in accuracy, will expand in coverage to meet the
demands laid upon the statistical agency, and will decline in
unit cost of preparation. Such a review might suggest the introduction of new statistical and administrative procedures,
adapting the statistical materials to the changed needs of those
who use them, cutting down some series, expanding others;
it might be made by the statisticians engaged in the work of
preparing such statistics or by a committee of specially qualified
experts. In the latter case, the recommendations, because of
the disinterest of the experts, might actually win greater support for extensions and improvements in the statistical programme than could be obtained by the statisticians primarily
concerned with the work. In providing for such review all
116
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
these possibilities should be taken into consideration, with a
view to the best development of the statistical programme as a
whole.
A final recommendation is this. It is of the utmost importance to build up a statistical staff with ability, scope for action
and a keen interest in making these statistics applicable to
the problems of today and tomorrow. Given such a staff to
seek out the best in the flow of new statistical and administrative
techniques and to provide more accurate, more promptly
available and more detailed results within given cost limits, it
can be expected that the statistical agencies will be ready with
the necessary data on employment and unemployment to meet
the new problems set by the progress of economic development.
PART V
XVIII. PROPOSED RESOLUTIONS
I
The Sixth International Conference of Labour Statisticians,
Having been convened by the Governing Body of the International Labour Office, and
Having met at Montreal from 4 August 1947 to
, and
Realising the importance of an adequate statistical basis for
the analysis of economic and social problems of employment
and unemployment and in particular for the provision of the
information necessary to the formulation and application of
policies designed to maintain full employment and to promote
economic development, and
Having considered the problems raised by the lack of international comparability of statistics of employment and unemployment, and
Considering, in order to promote the improvement of statistics in each country as well as their international comparability,
international standards covering the statistics of employment,
unemployment and the labour force, based on the best methods
available as tested in the experience of the different countries,
should be set by international agreement as a general objective
for the development of statistics on these topics,
Adopts, this
day of
1947, the following
Resolution :
GENERAL OBJECTIVES
1. A comprehensive system of employment, unemployment and labour force statistics should draw upon all the
resources available from a wide range of sources of data and an
extensive battery of statistical methods and techniques.
118
EMPLOYMENT, U N E M P L O Y M E N T A N D LABOUR FORCE
2. In general, employment, unemployment and labour force
statistics
(a) should be based on standard definitions;
(ft) should cover
(i) all branches of economic activity;
(ii) all persons, employed and unemployed, who
have jobs or who are seeking work, irrespective
of age;and
(iii) all industrial status groups; and
(c) should provide both benchmark data and series to
show current changes.
Employment Statistics
3. (1) For benchmark data comprehensive sources should
include population censuses with their classifications of the
labour force and the gainfully occupied, the censuses of industrial
production with data on persons employed in different industries,
social security registration and unemployment insurance records,
general population registrations as well as labour force sample
surveys.
(2) For series to indicate changes over time, comprehensive
sources should include establishment sample statistics, labour
force sample surveys, the population censuses and statistics
based on social security registrations, on unemployment or other
social insurance records.
Unemployment
Statistics
4. (1) In addition to the population census benchmark
data, the principal sources should include unemployment
insurance records covering the population subject to insurance,
and labour force sample surveys covering a sample of the entire
population.
(2) Where these are not available, the sources should include
trade union unemployment records or labour exchange figures
on unplaced applicants.
Labour Force Data
5. In addition to the population census materials, comprehensive sources should include labour force sample surveys
and other types of data, such as social security and social
insurance registration records.
PROPOSED
RESOLUTIONS
119
6. The battery of techniques and methods should include
in particular the methods of sampling as adapted and applied
in the collection of labour force materials and data for establishment series, as well as methods of appraising sources of error
and calculating adjustments for departures from recommended
standards and definitions.
PRINCIPAL
DEFINITIONS
7. To ensure comparability of data and series on employment, unemployment and the labour force within a country as
well as between countries, standard definitions and standard
classifications relating to employment, unemployment and the
labour force should be adopted as set forth below and in the
Schedule appended to this Resolution.
8. (1) The "employed " should include all persons who work
for their own account or in the employ of others. The "employed " should thus include employers; persons who are selfemployed (i.e. who work for their own account); salaried
employees; wage-earners; and unpaid family workers who are
engaged in tasks directly related to the operation of a family
enterprise for a minimum of 15 hours a week not including
hours spent in unpaid domestic work.
(2) The "employed " should include persons in labour
camps if they are free to seek alternative employment; but
should not include persons in such camps who are not free to
make that choice, or persons confined to prisons or similar
institutions.
(3) Persons who are directly employed by any public authority on emergency public relief work should also be included in
the statistics of the employed, but where such persons are
employed under conditions inferior to those of regular public
employees engaged in the same type of work their number
should be indicated separately.
9. The " unemployed " should include all persons seeking
work on a given day who are not actually employed but are
able to take a job if offered one.
10. (1) The "labour force " should include all employed and
all unemployed persons, as defined above, together with the
armed services.
(2) The "civilian labour force " should include all employed
and unemployed persons exclusive of the armed services.
120
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
11. "Industrial classification", "occupation" and "industrial status " should be defined as in the Schedule to this Resolution.
TYPES OF STATISTICS
A. Benchmark Data
12. A population census should be taken at least every ten
years and a census covering the major branches of economic
activity at least every five years, in order to provide, among
other things, basic information essential to the development of
adequate statistics of employment, unemployment and the
labour force.
B. Current Series
Employment.
13. Series showing the total workers employed should be
prepared at least quarterly in the more industrially diversified
countries.
14. Employment series should be prepared for each major
industrial group in which as much as 5 per cent, of a country's
total employment is found, as well as for each industry for which
such statistics may be requested by an inter-governmental
organisation.
15. In the more industrially diversified countries, consideration should be given to the desirability of preparing separate
series for each of the major industry groups and sub-groups
specified in the international standard classification of industries. 1
16. (1) Where seasonal changes in agricultural employment are substantial, estimates of agricultural employment
should be made more frequently than once a year.
(2) In the more industrially diversified countries, such
estimates should be made quarterly.
(3) In other countries, at least semi-annual estimates are
desirable in order to measure agricultural employment at its
seasonal maximum and minimum.
1
In referring t o " t h e international standard classification of
industries " t h e Conference has in mind for t h e present t h e minimum
nomenclature of industries recommended by the Committee of Statistical
Experts of t h e League of Nations in 1938; b u t if a revised international
standard classification should be recommended by t h e United Nations
Statistical Commission, it would favour t h e use of such revised standard
classification.
PROPOSED
RESOLUTIONS
121
17. From one or another of the current series issued in any
given country, information should be available on employment
for the major geographical regions, for each sex and for the
chief age groups.
18. As resources and facilities permit, separate series on
employment should be made available for —
(a) additional geographical or administrative regions;
(b) the chief centres of population ;
(c) each sex;
(d) age intervals as follows : single years of age for juveniles
and 10-year age groups for adults.
19. As resources and facilities permit, estimates of employment should be made available at intervals for —
(a) the principal marital status groups by sex;
(b) the principal occupation groups;
(c) groups classified according to the number of hours
worked per week ;
(d) the principal industrial status groups.
Unemployment.
20. Series showing the total numbers unemployed should be
prepared monthly in the more industrially diversified countries.
21. Where such information is of significant utility, the
unemployed should be classified by industry of last employment
and by major occupation group.
22. From one or another of the current series issued in any
given country information should be available on unemployment
for the major geographical regions, for each sex, and for the
chief age groups.
23. As resources and facilities permit, separate series on
unemployment should be made available for —
(a) additional geographical or administrative regions;
(b) the chief centres of population;
(c) each sex;
(d) age intervals as follows: single years for juveniles, and
10-year age groups for adults.
24. As resources and facilities permit, estimates of unemployment cross-classified by age, sex and duration of unemployment
should be provided for those economic regions in which unemployment is most severe.
122
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
Labour Force.
25. Series showing the total labour force should be prepared
at least quarterly in the more industrially diversified countries.
26. As resources and facilities permit, separate series on
the labour force should be made available for —
(a) the major geographical or administrative regions;
(b) the chief centres of population;
(c) each sex;
(d) age intervals as follows: single years of age for juveniles
and 10-year age groups for adults.
METHODS AND TECHNIQUES
Employment
27. Employment series should be adjusted to comprehensive
benchmark data derived from census enumeration or social
insurance records as such data become available.
28. Statistics of employment in establishments should
include all persons attached to the establishment, regardless
of age and regardless of whether or not these persons are actually
at work on the date to which the statistics relate.
29. (1) In the case of employment series based on reports
from industrial establishments, particular attention should be
given, in determining the sample of establishments from which
employment reports are to be obtained, to securing proper
representation of major industry groups as well as geographical
areas and sizes of establishments.
(2) Special attention should also be given to the correction
of any errors resulting from failure of establishments to make
reports; to securing reports by establishments rather than by
firms; and to the comparison of returns from identical establishments at successive dates.
30. From time to time each country should provide such
estimates of employment in particular industries or groups
of industries as may be needed to make possible a recombination
of its employment data into the international standard classification of industries.
31. (1) As resources and facilities permit, periodic studies
should be made of the nature and extent of the trend bias which
develops, between the dates of adjustments to benchmarks, in
employment series based on reports from identical establishments.
PROPOSED RESOLUTIONS
123
(2) On the basis of such studies, methods should be developed
to include in the sample a due proportion of persons employed
in new establishments.
32. As resources and facilities permit, in countries where
employment series are available from different sources, periodic
surveys should be made of a sample of returns from each source
in order to eliminate any significant double counting or inconsistencies which may exist and to lay the foundation for establishing a unified set of employment series.
33. As resources and facilities permit, seasonally adjusted
indices for the chief labour force and employment series should
be presented currently in addition to the unadjusted series.
34. As resources and facilities permit, periodic studies of
employment as reported by establishments should be made to
determine the number of persons who were at work and the
number not at work on the date to which the studies relate.
35. As resources and facilities permit, employment series
based on reports from establishments should be adjusted if
necessary to take account of differences in the employment
figures reported to different statistical authorities and of any
tendency to include in the reports data which are more inclusive
or less inclusive than those requested.
Unemployment
36. Estimates of the total numbers unemployed in any
country should wherever possible be based on figures derived
either from the operations of an unemployment insurance or
other social security system, or from sample surveys of the labour
force, or from both sources.
37. Where the definition of the unemployed utilised in any
country differs from that recommended above, estimates should
be made periodically of the difference between the number
unemployed according to the definition in use and the number
unemployed as defined above.
38. In the presentation of national estimates of unemployment, primary emphasis should be placed on percentages
rather than on absolute figures.
39. (1) The percentage of unemployment should be computed
by dividing the number of unemployed in a group by the sum
of employed plus unemployed in the same group, e.g. the number
9
124
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
of unemployed wage-earners and salaried employees by the sum
of the employed plus unemployed wage-earners and salaried
employees.
(2) In countries with complete statistics of the numbers
employed in all industrial status groups, the chief percentage
of unemployment should be computed by dividing the number
of unemployed earners by the total civilian labour force.
40. (1) In every country which relies on unemployment
insurance statistics for its chief measure of unemployment,
periodic studies should be made of the relation between the
total number unemployed on the one hand and the number
unemployed as reported by the insurance system on the other.
(2) Particular attention should be given to including in the
estimates of total unemployment such groups as the following
which may be excluded from or inadequately covered by the
unemployment insurance statistics:
(a) older workers;
(b) those who have been unemployed for long periods of
time;
(c) those not applying for benefits, even though they are
unemployed ;
(d) new workers who have not yet found their first jobs;
(e) supplementary workers who enter the labour force in the
peak or trough of the business cycle; and
(/) industries or occupations which are specifically excluded.
41. In any country in which, through lack of adequate data
derived from social security operations or labour force surveys,
national unemployment estimates may have to be based on
trade unions returns, such returns should wherever possible be
weighted in accordance with the numbers employed in the industries to which they relate.
42. In countries which use statistics derived from the operations of employment exchanges as their chief measure or one
of their chief measures on unemployment and in which such
employment exchanges are not closely linked to the operation
of unemployment insurance or the system of social security,
studies should be made at intervals to determine the relationship between the total number unemployed and the number of
unemployed registered at the employment exchanges. •
PROPOSED RESOLUTIONS
125
43. Studies should also be made of the extent to which
statistics of the registered unemployed derived from the operations of employment exchanges actually include persons who
are employed — either because of failure to exclude them at the
time of registration or because of difficulties in keeping the
records of the exchanges up to date.
44. As resources and facilities permit, in countries where
statistics of the numbers of insured unemployed constitute the
main measure of unemployment and where substantial differences may exist between the total number of unemployed and
the insured unemployed, surveys of these differences should be
used as a basis for constructing an adjusted current index of
total unemployment.
45. As resources and facilities permit, series on partial
unemployment should be replaced by figures showing the distribution of employed persons according to the number of hours
worked in the period covered (week or fortnight).
46. (1) Periodic studies should be made of under-employment, as distinguished from total unemployment.
(2) In particular, consideration should be given to the possibility of obtaining, in the case of industries other than agriculture, information as to the number of persons on short time who
are seeking full-time employment.
Labour Force
47. Since the method of labour force sample surveys has
proved to be an effective, accurate and reliable technique for
obtaining data on employment, unemployment and labour
force, as well as on other topics when desired, its use is recommended for serious consideration.
48. All new questions for labour force survey schedules
should be pre-tested in different parts of the sample in order to
establish definitely what interviewing techniques are actually
being used by enumerators and what patterns of answers are
being given by respondents.
49. Portions of the sample areas used in labour force surveys
should be re-enumerated at periodic intervals in order to check
on the interviewing techniques actually used by enumerators.
PUBLICATION
50. The key totals in current employment, unemployment
and labour force series should be issued, if necessary on a
126
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
provisional basis in the first instance, within one month after
the date to which they refer or as soon as possible thereafter.
51. (1) Data as published should be accompanied by a
statement of their scope.
(2) Where the chief unemployment series in any country
relates to the insured unemployed, current statistics in that
series should be accompanied by a brief statement indicating
the percentage of the unemployed who were not covered by the
series at the date of the most recent comparison between the
number of the insured unemployed and the total number
unemployed.
52. In any country which publishes more than one series
which can be used as an indicator of the level or trend of the
numbers employed, or of the numbers unemployed, each series
should be accompanied by a statement indicating its chief
differences from the others and the most appropriate uses for
each. Particular care should be taken to explain any discrepancies between the various series.
53. Each country should provide a brief guide to the major
benchmarks and series on employment, unemployment and
labour force for the use of the general public.
54. Each country should present in a basic technical publication all available historical estimates of employment, unemployment and labour force, together with full particulars of the
sources for the different series of the methods used in compiling
them, and of their uses and limitations.
55. Each country should issue, in continuation of the basic
volume referred to in the preceding paragraph, an annual
publication containing the major series on employment, unemployment and the labour force for current months and for
selected benchmark dates, together with a discussion of the
major changes which have taken place since the previous issue,
and references to the original sources in which more detailed
information may be found.
56. As resources and facilities permit, comparable historical
estimates should be prepared for the chief employment, unemployment and labour force series, extending back to 1929 and,
if possible, to 1919.
Schedule
1. "Industry" is the kind of economic activity or kind of
factory, store or other place of business in which a person works
PROPOSED RESOLUTIONS
127
or exercises his occupation; the classification of a person according to the industry in which he is employed is determined on
the basis of the nature of the activity of the establishment,
irrespective of his occupation.
2. The "industrial classification " to be recommended is
the international standard classification. At present the only
classification having international recognition is that proposed
by the Committee of Statistical Experts of the League of Nations
in 1938. In the event of a revised international standard
classification being recommended by the United Nations Statistical Commission, the use of such revised standard classification
is to be favoured.
3. " Occupation" is the trade, profession or type of work
performed by the individual, irrespective of the industry in
which he exercises it.
4. " Industrial status " is the position of the individual in
respect of his employment: employer, independent worker on
his own account (self-employed); salaried employee; wageearner; and unpaid family worker. In cases where managers
and directors are classified as employers, a special sub-division
should show managers and directors separately.
II
The Sixth International Conference of Labour Statisticians,
Recognising that the methods and practices which it has
recommended in its Resolution concerning employment statistics
will, when applied, facilitate the compilation of accurate statistics of payrolls, but being strongly of the opinion that fuller and
more comprehensive consideration should be given at the
earliest possible moment to the formulation of international
standards for statistics of workers' earnings;
Requests the Governing Body of the International Labour
Office to place on the agenda of an early session of the International Conference of Labour Statisticians the subject of statistics of workers' earnings with special reference to payrolls and
aggregate wages and salaries.
Ill
The Sixth International Conference of Labour Statisticians,
Noting the desirability of establishing an international
standard classification of occupations which would permit
128
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
comparisons between the numbers engaged, the rates of remuneration, the hours of work and other conditions in the same or
similar occupations in different countries, and noting in particular the need for such a standard classification for use in connection with the national censuses recommended by United
Nations Statistical and Population Commissions;
Requests the Governing Body of the International Labour
Office to place on the agenda of an early session of the International Conference of Labour Statisticians the subject of an international standard classification of occupations.
IV
The Sixth International Conference of Labour Statisticians,
While satisfied that the Resolution which it has adopted on
statistics of the labour force, employment and unemployment,
will provide sound basic guidance for the improvement and
standardisation of such statistics for some years to come, being
confident that further advances will be made from time to time
in the methods and techniques of these statistics;
Requests that the Governing Body of the International
Labour Office should place on the agenda of a future session
of the International Conference of Labour Statisticians the
further consideration of international standards for these statistics as soon as such advances, or the experience gained in
applying the standards formulated at the present session, would
appear to make such further consideration desirable, and
that the Governing Body should direct the International Labour
Office to continue their studies of these statistics and to publish
the results of such studies from time to time.
APPENDIX
Resolutions of the Second International Conference
of Labour Statisticians, Geneva, April 1925 1
II.
UNEMPLOYMENT STATISTICS
(1) In countries in which a widespread system of unemployment
insurance exists the information obtained from the working of such
a system forms the best basis for unemployment statistics.
(2) These statistics should furnish the following information as
a minimum :
(a) Annually, the total number of workpeople insured against unemployment, which should be related so far as practicable to the total
number of workers.
(Í») Monthly, the total number of unemployed on a given day in receipt
of benefit, and the total number of insured workers unemployed
(whether on benefit or not) on the same day.
(c) Monthly, the percentage that on a given day the total number of
insured unemployed (whether on benefit or not) forms of the
estimated number of insured workpeople.
(d) Annually, the total amount paid during the year in benefits.
A method according to which, as in the Netherlands, a percentage
is calculated between the number of days of unemployment during
one week and the total number of days of work which might have been
performed is one which is worthy of attention.
(3) Where statistics based on unemployment insurance, compulsory
or voluntary, are not available, it is desirable to obtain from workers'
organisations the following information:
(a) Monthly, the total number of unemployed on a given day and the
percentage they form of the total membership covered by the
enquiry.
(b) Annually, the number of workers covered by the enquiry as a
percentage of the total number of workers in the corresponding
industries or occupations.
Even when statistics based on unemployment insurance become
available it is desirable, for purposes of comparison, to continue trade
union statistics as long as they are reliable.
(4) The statistics derived from public employment offices should give :
(a) The number of workpeople registered on a given day of the month
as seeking work, with the total number of vacancies remaining
unfilled on the same day.
(6) The number of workers' applications registered; of vacancies
notified ; and of vacancies filled, during the month.
(5) Employment exchange statistics should be compiled so that
as far as possible unskilled workers are distinguished from other classes
of workers.
(6) Information as to the state of employment should also be
published periodically, if possible monthly, based on returns made by
a representative number of employers.
1
Studies and Reports, Series N (Statistics), No. 8 (Geneva, 1925),
pp. 70-72.
130
EMPLOYMENT, UNEMPLOYMENT AND LABOUR FORCE
(7) In countries in which satisfactory information concerning
unemployment cannot be obtained by the means indicated above, it is
desirable that an attempt should be made to obtain at the general
population census, or at an industrial or occupational census, information as to the amount of unemployment, or that special enquiries
relating to the whole population or to an adequate sample thereof
should be made from time to time with a view to ascertaining the number and condition of the unemployed.
(8) The statistics indicated in the preceding resolutions should
distinguish males and females and should, so far as possible, give
figures for separate occupations in the case of statistics arising out
of the activities of employment exchanges, and for separate occupations
or industries as may be most convenient in the case of trade union and
insurance statistics and in the case of special industrial or occupational enquiries. The classifications used should be based on the
classifications adopted in the general population census in so far as
these are applicable to employment exchange operations.
(9) It is desirable, for comparing statistics of unemployment
internationally:
(a) that precise and detailed information should be published and
kept up to date as to the methods adopted in compiling the various
statistics of unemployment; in particular, each country should
indicate any legislative or administrative changes affecting the
value of its unemployment statistics;
(Ô) that copies of all forms and questionnaires used in collecting the
various statistics should be forwarded to the International Labour
Office;
(c) that in each country an enquiry should be made to determine,
as exactly as possible, the representative value of the unemployment statistics in relation to the "ideal statistics" which would
give, at any date, the total number of unemployed in relation to
the total number of workers. For this purpose it is agreed :
(i) that the ideal population "field" to which the statistics should
relate should be all persons whose normal means of livelihood is employment under contract of service, as well as
those persons not hitherto wage-earners who seek to become so ;
(ii) that the unemployment measured should exclude that due to
sickness, invalidity, participation in trade disputes, or voluntary absence from work, and should be limited to unemployment
due to lack of employment or work while in employment;
(iii) that the necessary and sufficing condition for being enumerated
as unemployed is that the individual must have been not at
work for one day at least.
(10) It is desirable that the different statistics of unemployment
(insurance, trade union, and employment exchanges) should be presented together, preferably in graphic form, so that they might be
co-ordinated and checked one with another and as clear and correct
an idea as possible of the fluctuations in unemployment obtained.
(11) Statistics of short-time employment should, if possible, be
given separately from those of whole-time unemployment.