Preview

INTERNATIONAL LABOUR OFFICE

METHODS OF LABOUR
PRODUCTIVITY STATISTICS

Report prepared for the Seventh International Conference
of Labour Statisticians
(Geneva, September 1949)

GENEVA
1951

S T U D I E S AND

REPORTS

New Series, No. 18

PUBLISHED

BY THE INTERNATIONAL LABOUR
GENEVA,

OFFICE

SWITZERLAND

Published in the United Kingdom for the INTERNATIONAL LABOUR OFFICE
by Staples Press Limited, London

PRINTED BY "IMPRIMERIE KUNDIG", GENEVA, SWITZERLAND

CONTENTS
Page
INTRODUCTION

CHAPTER I. Concepts and Definitions
The Two Main Concepts
Man-Hours Expended per Unit of Production
Various Aspects of Labour Productivity
Labour Productivity at the Various Levels of Production . .
Labour Productivity in the Various Branches of Economic
Activity
CHAPTER II. Factors Affecting the Productivity of Labour
General Factors
Organisation and Technical Factors
Human Factors
CHAPTER III. Problems in the Measurement of Labour
Different Kinds of Labour
Output per Man-Hour and Output per Man
Heterogeneity of the Labour Force
Measurement of Labour
CHAPTER IV. Problems in the Measurement of Output
Product Specifications
Definition of a Product and of an Industry
Degree of Integration
Measures Based on Production Processes
Duration of the Production Process
Measurement of Heterogeneous Outputs
CHAPTER V. The Measurement of Labour Productivity
Some Possible Approaches
The Two General Concepts of Labour Productivity
Comparison of Production Indices with Labour Indices . . .
Measures of Labour Productivity in Non-Manufacturing Industries
Labour Productivity of a Nation
CHAPTER VI. Special Enquiries
Past Experience
Direct Productivity Reports in the United States
Conclusions
CHAPTER VII. International Comparisons of Labour Productivity . .
Problems in the Measurement of Labour
Problems in the Measurement of Production
Comparability of Production and Labour Data
Special Problems of International Comparisons
Special Enquiries
CHAPTER VIII. Interpretation of Labour Productivity Data
The Concepts
.
Measurement
The Interpretation of Over-all Measures
Necessity for Detailed Information

6
7
8
9
10
12
16
20
23
27
30
31
36
38
41
43
43
46
47
50
51
52
55
55
57
68
73
75
78
78
81
89
91
92
95
99
101
105
108
108
109
112
115

CONTENTS

CHAPTER IX. General Conclusions
Definitions
Formulae
Man-Hours per Unit of Output
Computations Based on Net Output
Division of Production Indices by Employment Indices . . .
Periods of Computation
Special Enquiries
Publication
I.
II.
III.

APPENDICES
Analysis of Labour Productivity Formulae
Resolution Adopted by the Seventh International
Labour Statisticians
Selected References

Page
118
118
119
119
120
121
121
122
123
124

Conference of
132
134

INTRODUCTION

Mankind has always striven to obtain the maximum of
satisfaction for the minimum of effort.
In this desire for
efficiency or productivity of effort originated the invention of
primitive tools and of the most outstanding machinery of modern
times.
For a given community, higher production implies a higher
standard of living; but, since economic welfare is the sum of
material production and leisure in which to enjoy the fruits of
production, productivity is a far better index of economic welfare
than actual production.
Productivity is a general term; any factor of production may
be studied for its productivity. Labour productivity has received
the greatest attention because of its direct relation to the results
of human effort.
The importance of labour productivity has been summarised
as follows:
If problems of production are approached from the viewpoint of
capita] employing labour, the cheapness of labour relative to capital
is one of the main factors determining methods of production, and
money costs of production are of paramount importance. If, on the
other hand, the view is taken that labour employs capital (and not
the other way round) to increase its standard of life by capitalistic
methods of production, " dear " labour becomes the object of policy
and not one of its determining factors. Here labour ought to be made
dear and capital cheap, and an economics of plenty substituted for an
economics of scarcity. Money costs of production, influenced by monopoly, market imperfections, low living standards in overseas countries,
and a host of other factors, are no longer sovereign. Real costs of production now become important—costs of production in terms of labour
and other real factors of production. Ultimately, real costs of production
can be reduced to labour costs and the cost of what the classical economists called " waiting ". Hence the importance of the concept of the
productivity of labour. The problem of policy consists in increasing
the productivity of labour, each increase in productivity being balanced
against the increase in the cost of waiting called forth. The acceptance
or rejection of hydro-electric schemes, for instance, depends chiefly on
relative wages in coalmining and in industries producing hydro-electric
plant, and also on the customary rate of return on capital in different
industries; it ought to depend on differences in manpower requirements

METHODS OF LABOUR PRODUCTIVITY STATISTICS

between different methods of producing energy, to be balanced against
differences in the amount of waiting required. 1
The advantages of high labour productivity have been pointed
out many times; a United States expert writes:
Productivity, once simply a tool for the economist, is acquiring new
significance to the worker, the business man, the soldier, and the statesman. It has long been understood that the relatively higher standard
of living of the American citizen was the direct fruits of higher productivity in our economy. In effect, it has given us not only bread and
butter, but jam as well.
During most of the last century there has been a high and steady
rate of growth in production per man-hour in the United States. Primarily, this has resulted from the continued acquisition of technical knowledge and its steady application to the jobs we want done. The
cumulative effects have reshaped the lives of all of us. As our production potential has increased, the average worker has been able to
buy steadily increasing quantities of goods. Simultaneously with more
goods, we have been able to afford more leisure time in which to enjoy
them. Many arduous but necessary tasks have been eliminated. The
labour of the home has been greatly reduced. As we have freed ourselves from the pinch of material wants, we have been able to divert
an increasing proportion of our resources to such activities as schools
and social services.2
Similarly, Lenin wrote that " productivity of labour is, in the
end, the most important matter, the essential matter for the
victory of our social system ", and the increase of productivity
of labour has been one of the most important aims of the Soviet
Union's policies.
Since the second world war, practically all countries have put
this question in the foreground of their preoccupations. In France
the question of higher productivity is now considered to be crucial:
To restore its economic situation, France has been striving since the
liberation to increase production. In her efforts to reach this aim, she
has tried to ascertain the factors which limited production; some
factors, such as the coal crisis and restricted imports, have been eliminated or their influence gradually reduced; one of them—and one of the
most important—has not yet been sufficiently brought to light and
studied; it has limited and still limits our production: it is our low
productivity. 3
1
T. B ARNA : " The Productivity of Labour : Its Concept and Measurement ",
in the Bulletin of the Oxford University Institute of Statistics, Vol. 8, No. 7,
July2 1946, p. 205.
E. CLAGUE: "Planning Guides for Industry", in Dun's Review, Oct.
1948, pp. 20-21.

3
PRÉSIDENCE DU CONSEIL, COMMISSARIAT GÉNÉRAL DU PLAN DE MODERNISATION ET D'ÉQUIPEMENT: Programme français pour Vaccroissement de la

productivité (Peb. 1949).

INTRODUCTION

Productivity of labour has been the subject of numerous
studies for many years. One of the first statistical enquiries on
the subject, directed especially to a comparison of the productivity
of labour in hand and machine operations, was carried out as far
back as 1898 by Carroll D. Wright, Commissioner of Labor and
head of the Bureau of Labor Statistics in the United States. 1
The question has, in fact, always aroused considerable interest
in the United States, and the amount of statistical literature
available on the subject in that country is particularly large. The
selected list of references appended to this report gives the main
publications. The studies of David Weintraub and the Works
Progress Administration, Solomon Fabricant and Duane Evans in
the United States, and of Laszlo Rostas in the United Kingdom,
are among the main contributions.
The rehabilitation of shaken economies, the necessity of balancing imports and exports, the development of full employment
policies and the pursuit of high levels of production have shown
labour productivity to be of vital importance to all national and
international bodies. The productivity of labour may be considered
a cornerstone of the economy of the future.
The International Labour Organisation is primarily interested
in this topic, not only from the point of view of the development
of general welfare and higher standards of living, but also as a
problem of fundamental concern to workers.
In the development of productivity policies, detailed information must be sought, mostly of a statistical, nature : information
concerning the level of productivity of labour in undertakings,
industries and countries ; comparisons between plants, regions and
above all countries are prerequisites to any action on this point.
Since the end of the second world war, the productivity of
labour has been studied by many conferences or committees;
examples are the conferences on productivity held in Washington
and the Anglo-American Council on Productivity in the United
Kingdom. In France the Office of the Commissioner-General for the
Industrial Equipment Plan has devoted numerous conferences to
the subject, and an Interministerial Committee on Productivity has
been created. These conferences and committees have all stressed
the importance of obtaining correct and complete information,
and the necessity of developing the measurement and comparison
of labour productivity.
1
C. D. WRIGHT: Hand and Machine Labor, United States Department of
Labor, 13th Annual Report, 1898.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

Even in the United States, which is conspicuously in advance
of other countries in the measurement of productivity of labour,
an official report recently noted that—
(1) the methods of measuring productivity of labour must be
developed beyond their present scope; satisfactory solutions to
the problems of production measures, labour input and price
deflators should be sought;
(2) long-range historical studies should be made of the changes
in productivity of labour and their causes ;
(3) there is urgent need for current and up-to-date data on
productivity of labour, since most of the present data are out of
date when they become available.1
In view of the increasing interest in this problem, the Governing
Body of the I.L.O., at its 104th Session in March 1948, placed the
subject of methods of statistics of productivity of labour on the
agenda of the Seventh International Conference of Labour Statisticians.
The present report, prepared for this Conference, reviews, therefore, the methods of measuring and comparing productivity of
labour.2 As it represents the first report of the International
Labour Office on the statistical aspects of this subject, it is of a
preliminary and informational character, and endeavours rather
to raise problems and analyse them than to prescribe definite
solutions to the considerable difficulties that are to be encountered.
The report is entirely concerned with problems of measurement,
comparison and interpretation of data, and no analysis is attempted
of the effects of high or low productivity of labour, or of its relations
to other aspects of economic or social life.
In particular, economic or social aspects and effects of high
or low productivity of labour are not within the scope of the
present study. Whether better conditions of life are to be sought
through a high standard of material consumption or more leisure,
or whether increases in productivity are obtained, for instance,
through undue fatigue of the workers, are questions outside the
present enquiry. This report is concerned merely with how levels of
productivity can or should be measured and compared.
The report is divided into nine chapters. In Chapter I are
reviewed the varying objectives of the measurement of the pro1
Report of the Joint Committee on the Economic Report on Current Gaps
in our Statistical Knowledge (80th Congress, Second Session, 1948).
2
See Appendix II for the text of the resolution adopted by the Seventh
International Conference of Labour Statisticians.

INTRODUCTION

ductivity of labour, its concepts and possible definitions, and the
different types of labour productivity in the various stages of production or branches of economic activity. Chapter II gives an
account of the many factors which influence labour productivity,
and Chapters III and IV deal with problems concerning the
definition and measurement of labour and of production in relation
to labour. The actual measurement of the productivity of labour
is the subject of Chapter V1; the comparability of production and
labour figures is also discussed in this chapter, and an analysis is
made of productivity or unit labour requirement indices and of
possible formulae. A separate chapter is devoted to the collection
of data through direct enquiries into the productivity of labour
(Chapter VI), in view of the importance and advantages of this
source. The first six chapters thus clear the ground for the
analysis of the special problems involved in international comparisons of labour productivity which are discussed in Chapter VII.
Chapter VIII is devoted to a consideration of the various interpretations that might be given to labour productivity data in
relation to the bases of computation. Some general conclusions
are drawn up in Chapter IX.

Appendix I contains a more formal presentation of the argument of
Chapter V.

CHAPTER I
CONCEPTS AND DEFINITIONS
Output is obtained by the combined input of a number of
factors which all have their importance—equipment, resources,
energy, work, skill, management; these factors may be grouped
under the headings of labour, capital, land and organisation.
The ratio between output and one of these factors of input is generally known as the productivity of the factor considered. One
writer comments:
Some think of productivity as a measure of performance of the
economy as a whole. Others think of productivity in terms of individual
industries or plants. Some business men, in their public orations,
speak as though the whole matter of productivity had to do with the
degree of application of workers to their jobs. At other times, the
concept of productivity is used as though it were a measure of the
degree of efficiency achieved in production. The dictionary merely tells
us that this word stands for the " quality or state of being productive ".
One thing common to all these concepts of productivity is the desire
to portray someone's ability to produce or the rate at which production
is carried on.1
Another writer states :
One can compare any factor to the corresponding production.
Measures of all kinds have been made; production has been compared
to the surface covered or available, to the cost of buildings, to energy
provided or consumed, to invested capital, to the total cost of planning
and even to the average temperature of the premises . . . Production can also be compared to the quantity of raw materials used.
Other things being equal, the results will yield interesting indications
of the degree of waste, the degree of technical efficiency of certain
machines or methods, etc.2
Historically, Government and private research workers have
usually defined and measured productivity as the ratio between
production (or output) and the expenditure of some resource
used in production (or input). The greatest interest has always
1

L. TEPER: " This Thing called Productivity ", in The American Federationist,
Nov. 1948.
2
G. DEURINCK: "Mesure de la productivité", in Organisation Scientifique, Dec. 1948, p. 362.

CONCEPTS AND DEFINITIONS

centred in the relationship between production and labour, the
universal resource, and the term " productivity " is frequently used
without qualification to refer to this ratio.
The most general definition of productivity of labour is therefore the ratio of output to the corresponding input of labour.
This definition is very broad, since it leaves the terms " output "
and " labour " open to a number of interpretations, but it covers all
definitions, whether implicit or explicit, which have been presented
in various studies. For instance, it does not decide between
employment or man-hours as the correct measure of labour.
THE TWO MAIN CONCEPTS

The various definitions so covered can be sorted into two large
groups, which correspond to the alternative answers to the following
question : should the productivity of labour for a group of products,
undertakings or industries be so defined that it remains unchanged
when the productivity of labour of each component is unchanged,
or should the definition be such that even when the productivity
of labour for each component has not varied, the over-all productivity of labour may have changed, because of variations in
the relative importance of the various components—for instance,
the rise in importance of those with a high level of productivity
to the detriment of those with a low one ?
The first concept, that is, an average productivity which will
remain unchanged when each individual productivity is unchanged,
will normally be used when variations in the productive efficiency
of an industry or the combined effect of such variations in a group
of industries is studied. It will also answer such questions as
the following, taken from the report of the United States Works
Progress Administration :
(1) What relative volumes of'labour time are required to produce
a given composite of products at different times ?
(2) What relative volumes of production of a given composite of
products are obtainable at different times with a given amount of
labour time ?
Answers to these questions are of use in estimating (1) employment requirements for different levels of production and (2) future
production
under various conditions of availability and utilisation of
labour.1
1

UNITED

STATES

WORKS

PROGRESS

ADMINISTRATION,

SEARCH P R O J E C T : Production, Employment and Productivity
ing Industries, 1919-1936, May 1939, P a r t I, p . 3.

NATIONAL

in 59

RE-

Manufactur-

METHODS OF LABOUR PRODUCTIVITY STATISTICS

The second concept, that is, an average productivity which may
vary, owing to the changing composition of production, even if
each individual productivity is unchanged, is of much wider scope ;
it takes into account that average productivity increases when
products, undertakings or industries corresponding to a high level
of productivity increase their importance in comparison with
those corresponding to a low level.
This second concept will be used when comparing over-all
changes in the ratio of manufacturing output to total manufacturing
employment, or national output to total employment or total
labour force. It will be shown later that it is often implicitly
used when labour productivity indices, for instance, are computed
by the division of ordinary production indices by employment
or man-hour indices.
Thus, the choice between these two concepts will depend on
the information sought, although it will, of course, always be
preferable—whenever possible—to compute data on both bases
in order to ascertain, for instance, the influence on the observed
changes of the modification in the composition of the production.
However, most of the considerations on the measurement of
production and labour apply to both concepts and can be examined
independently of those concepts; the meaning and interpretation
of the various measures and formulae used will be pointed out in
Chapters V and VIII with regard to their relation to these two
concepts.1
MAN-HOURS EXPENDED PER UNIT OF PRODUCTION

The first approaches to the subject of measuring labour productivity generally referred to the ratio of production to labour.
But in everyday practice, this ratio, while full of interest, has often
proved to be more difficult to use than its reciprocal, termed
" unit labour requirements ", or " man-hours expended per unit
of production ".
In effect, this last measure is being more and more used in place
of the productivity of labour to characterise the same phenomenon.
Its definition runs as follows :
For a single uniform product, the unit labour requirement for any
particular time is the labour consumed per unit of output of the specified
1
Other concepts of productivity of labour may be considered, such as
marginal productivity, i.e., the variation of output for an additional unit
of labour. As these concepts are not within the immediate scope of this
report, they have not been dealt with.

CONCEPTS AND DEFINITIONS

good, or (which amounts to the same thing) the ratio of the total labour
required for the production of a given volume of a homogeneous good
to the given volume of that good.1
The main advantage of data concerning man-hours expended per
unit of product is that they can be added together, whereas labour
productivity data cannot. When absolute figures of productivity of labour are computed in terms of physical units produced per man-hour, for example pounds of cotton yarn spun per
man-hour, or pounds of woven cotton produced per man-hour,
the data cannot be combined with similar figures through direct
addition; new computations must be made. On the contrary,
" unit labour requirements ", expressed in terms of hours worked
in order to produce a defined good, can be directly added or subtracted. The advantage of this characteristic of unit labour requirements is particularly important when comparing the productivity
of labour of two undertakings, only one of which is integrated:
man-hours expended per unit of output can be shown for each
stage of the production, and comparisons can therefore be made
between corresponding stages; with figures stated in terms of
output per unit of labour this is difficult or impossible. Thus, if
comparisons are required between two cotton textile plants, in
one of which the cotton is spun and then woven, whilst in the
other only the weaving stage is performed, if unit labour requirements are computed it will be possible for the first plant to show
separately, in the total unit labour requirements, those needed to
spin the cotton and those needed to weave it, and therefore comparisons will be possible with the data collected for the second plant.2
Moreover, no reference to the efficiency of the workers is
implied in the number of man-hours expended per unit, but rather
the amount of labour required having regard to the possibilities
or drawbacks of the techniques used ; in other words, the influence
of factors other than that of labour alone is obvious.
VARIOUS ASPECTS OF LABOUR PRODUCTIVITY

It was noted in the introduction that this report is confined
to the discussion of methods of measuring productivity, and there1
D. EVANS and I. SIEGEL: " T h e Meaning of Productivity Indexes",
in the Journal of the American Statistical Association, Mar. 1942.
2
The United States Works Progress Administration, in developing formulae for productivity indices, consistently utilised " unit labour requirements " measures in preference to " productivity ". (Cf. Production, Employment and Productivity in 59 Manufacturing Industries, 1919-1936, op. cit., Part I,
Chapter 1, pp. 3 et seq.)

METHODS OF LABOUR PRODUCTIVITY STATISTICS

fore does not deal with economic or social aspects of the question.
It might, nevertheless, be useful to touch upon some of the aspects
of labour productivity that might bear a relation to its measurement.
The economic usefulness of the output does not affect the level
of productivity. Whether the placing of publicity posters is
economically " useful " or not has no bearing on the fact that
the bill poster can post so many posters (of a given size) per hour
of work. The economic usefulness of the posting may be important, however, in considering at the national level whether
spending time and money on advertising constitutes the best
investment of the nation's resources.
Should a distinction be made between " manual " and " nonmanual " productivity of labour ? A simple example of the first
is the productivity of a piece worker, while the second might be
exemplified by the productivity of a research chemist. There
seems to be no difference in concept between these two types, but
only a difference in the possibilities of measurement. But there is
certainly such a thing as the productivity of labour of the usherette
in a theatre, of a waitress in a restaurant, of an office clerk, of a
telephone operator, of a translator, a librarian, etc., even though
the productivity of such persons is difficult or impossible to measure
because of the difficulty of measuring their production. This
point will be raised again with respect to productivity in nonmanufacturing industries.
LABOUR PRODUCTIVITY AT THE VARIOUS LEVELS OF PRODUCTION

The measurement of productivity becomes increasingly difficult
as one proceeds from the consideration of the production of a
single independent worker to that of an industry including many
plants manufacturing related but different products.
The case of the single independent worker seems simple—•
in most instances, his output and the labour he contributes can
easily be defined. The cutter in a tailor's shop will prepare a
certain number of pieces per hour for the making of overcoats;
the miner cuts a certain weight of coal per day, etc.
Special problems arise for " non-manual " workers, as noted
above. It is difficult to define the production of an office clerk who
is not assigned an entirely routine job; to define the production
of the supervisor of a gang of workers is still more difficult.
It is important to note that even for the simple case of a single

CONCEPTS AND

DEFINITIONS

worker, the productivity of labour is not identical with efficiency ;
it is not solely a measure of the worker's effort. The faster a
sweeper works and the less time he loses, the faster the floor will
be cleaned; but if he sweeps the same floor with a better broom,
he will be able—for the same effort expended—to finish his job
in a shorter time. If he is given a vacuum cleaner he may not go
any faster, but the job accomplished is not entirely comparable
to the previous one; the production is somewhat different in
quality—the floor is cleaner. A worker is thus not the complete
master of his productivity; the tools or machines used, the technique
followed, the quality of raw materials consumed, etc., have in
most cases more influence on his productivity than the effort he
expends. This fact had already been stressed in 1898:
In what is termed the " hand " method of production, machines
have been used. It is true that the machines thus used are generally
of the most simple kind, such as the saw, the hammer, the chisel, the
pick, the shovel and the knitting needle; yet these are no less machines
than the larger or1 more complicated ones used in what is termed the
machine method.
A distinction might also be made between machine-paced and
man-paced work. There is a wide difference in the output of
base paper, which is dependent on the speed of a paper machine,
and the number of bricks laid in a day, which is the result of
conscious control by the individual operator. Wherever there is
a chain assembly or an automatic machine, the standard capacity
of work is known. In the case of the line assembly, the employee
must keep up to this capacity.
Thus, it should be clear that the productivity of a single worker
is that of a unit, " man plus machine ", that is, the productivity
of the worker with a given set of tools or a certain machine, and
therefore does not depend solely on the worker's effort.
The next level is that of the gang of workers. Here the productivity of labour is a consequence of a number of combined
factors. The ability of the supervisor may have as much influence
on output as the individual efficiency of each worker. In most
cases, however, the production remains easily definable, since the
gang is normally assigned a complete job; it might even sometimes
be easier to define the production of the gang than that of each of
its workers.
The measurement of the labour productivity of an entire
workshop will encounter many of the difficulties involved in
1

C. D. WRIGHT: op. cit.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

measures for larger units. Management influence is increasing;
" auxiliary labour " (clerical workers, errand boys, etc.) contributes
to production. Production itself may already be heterogeneous,
and this will raise the problem of addition.
When a large undertaking is considered, heterogeneity of production becomes a major difficulty; an automobile factory may also
produce automobile parts, wheelbarrows, toys, etc., in addition to
automobiles. Management intervention, specialisation and division
of work, the establishment of standards, recruitment policies, the
building up of stocks, etc., introduce many problems which are
ignored at the job level. Hence, the productivity of labour of an
undertaking becomes a complex concept.
At the next level, the industry, the questions to be solved
become extremely difficult. The obtaining of statistical data will
be impeded by the lack of comparability of accounting methods
in different undertakings, and the differences in production, as well
as by the administrative problems of securing data; the varying
composition of the production may have a considerable influence
on the results; thus the statistical problems may appear insuperable, and the solutions adopted will differ according to the purpose
of the enquiry.
LABOUR PRODUCTIVITY IN THE VARIOUS BRANCHES OF
ECONOMIC ACTIVITY

The preceding remarks apply mainly to manufacturing and
mining. But the manufacturing industries in any country at
most represent less than half of the economy as a whole. Even
in the most industrialised countries mining, manufacturing and
the construction industries together do not attain 50 per cent, of
the total: for example, in Belgium these industries occupy 48 per
cent, of the total labour force, in Czechoslovakia 39 per cent.,
in Germany 41 per cent., in the United Kingdom 46 per cent., and
in the United States 33 per cent. In other countries the same
three groups represent altogether about one fourth of the economy,
while in underdeveloped countries they often hardly amount to
one tenth. However, the bulk of the studies on labour productivity
relate to manufacturing, and it would appear that the importance
of this part of the economy has been over-emphasised, at the
expense of agriculture, transport, trade and services. The main
reason, as can easily be shown, lies in the fact that problems of
measurement are much more difficult in these other industries.

CONCEPTS AND DEFINITIONS

In agriculture the measure of output has to be based on annual
production. As in the case of a manufactured product, it is
conceivable to measure productivity—or man-hours expended per
unit—for the various phases of a job (e.g., in the production of
wheat, ploughing, harrowing, sowing, mowing and threshing). The
chief problems, however, for this part of the economy lie in the
measurement of the labour expended: the definition of farm
employment, not to mention the hours worked and especially
the collection of accurate data in this field, has always been unsatisfactory. It should also be noted that the efficiency of the
worker, understood as the influence of his personal effort, may
be of little importance in production which varies mostly according to the quality of the soil, climatic conditions, equipment, etc.
In transport, on the contrary, the measurement of employment
or man-hours does not raise more problems than in manufacturing.
But here " output " is never very obvious : what is the output of a
bus driver, or of a bus company ? Is it the number of persons
transported or the total " man-miles " accomplished ? Disregarding the problem of collection of data, neither of these measures of
production (or any other conceivable one) is very satisfactory.
For example, the total money collected from the persons transported can be compared to employment or man-hours; but it is
obvious that such measures are quite remote from the basic notion
of productivity of labour. Let it be added that here also personal efficiency has little effect on the output; the number and
quality of guards and drivers on a train is independent of the
number of passengers, though, of course, they have an important
bearing upon accidents and the successful performance of transport
activity.
Nevertheless, measures in this field are of great importance:
the competition between rail and road, and the comparative
advantages of each should be, for instance, tested by the number
of man-hours required for the accomplishment of a certain task,
e.g., the transport of a ton or a unit of money's worth of merchandise.
The measurement of productivity of labour in trade and
services is extremely difficult, because these functions are of a
service nature for which there is normally no unit of physical
measurement. It would be of the utmost importance to develop
such measures, but owing to the problems involved very little
has been achieved in this direction.

METHODS OF LABOUR PRODUCTIVITY

STATISTICS

The measurement of labour in the field of distribution is also
complicated by the fact that an occupational classification might
be thought preferable here to an industrial one: it may be urged
that the " output " of distribution should be compared to the
occupational group of sales persons, that is, chiefly those employed
in wholesale and retail trade plus sales personnel employed in other
industries.
It would also seem important to know something about
Government productivity in order to measure average productivity
for the economy as a whole, to understand employment trends
in the service industries, and to estimate rationally the budget
requirements of the various Government activities and services.
Here, however, adequate statistical data are lacking. Such
measurements are possible where the output produced by Government is similar to that produced in some private industries in
which physical productivity can be measured; where units of
output cannot be counted—as is true in much public employment—
no satisfactory method has yet been put forward.
In most of these cases the difficulties mentioned arise from the
fact that the products or the production are not measurable in
the present state of knowledge. It is relatively easy to determine
the production of a piece worker at his machine or of an automobile
plant; it is not feasible to determine the " production " of a civil
servant drafting laws, or the production of a Government department as a whole; it is still more difficult to judge of the productive
value of negative conclusions reached by a research chemist, or
of the presumably tremendous productivity of the discoverer of
atomic energy.
Disregarding all such extreme examples, it may be asserted
that trade and service industries, as well as Government activities
(national defence, public administration, the judicial system, etc.)
clearly contribute to the economy, though the contribution cannot
be measured in terms which are comparable with those used in
measuring the contribution of sectors of the economy which produce physical commodities.
Finally, certain methods of measurement have been used in an
attempt to evaluate the total economic activity within the country,
usually in monetary terms, and their use in conjunction with
employment and man-hour statistics provides valuable information.
For example, in recent years gross national product statistics have
been divided by measures of man-hour input to show the trend
of gross national product per man-hour. Although this relation-

CONCEPTS AND DEFINITIONS

ship is of importance, it is essentially a monetary measure, and as
an indicator of productivity must be interpreted with care: such
measures correspond to the second type of labour productivity
concepts as described at the beginning of this chapter. They will
be considered in more detail in the following chapters.

CHAPTER II
FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR
It will be recalled t h a t productivity of labour has been broadly
defined as the ratio of output to the corresponding input of labour.
There is normally a strong tendency, when considering labour
productivity data, to enquire into the factors responsible for the
variations observed, or to ascribe these variations to the influence
of certain factors.
Management will have a tendency to claim t h a t increases in
productivity are a result of its sole action, or to consider labour as
responsible for any decreases; on the other hand, labour will normally reject the idea that a decline in productivity might be due
to a slow-down of effort, but will consider t h a t higher wages should
be an immediate consequence of increased productivity; technicians
will tend to ascribe all variations to mechanical and technical
improvements.
In order to ascertain the actual meaning of labour productivity data, and to determine whether any measurements can
be devised to show the separate influence of one or another of the
various factors, this chapter is devoted to the factors affecting
productivity of labour.
Although at first thought it might be expected t h a t variations
in the productivity of " labour " must be due to " labour ", if the
definition is examined carefully it should be clear t h a t any factor
affecting output or labour may have an influence on labour
productivity. Only when the quantity of labour contributed to
the output is the sole factor which has changed will the productivity
of labour vary in relation to labour alone ; this case is rare. In most
comparisons many factors other than labour have varied and
therefore account for part of the differences observed.
This observation has been made by all research workers who
have dealt with productivity increases; two quotations will suffice:
Increases in the output of a worker should not be identified with
harder work, or a lack of increase in productivity with slackness on the
part of workers. This may be one of the many determining factors,

FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR

and in certain circumstances it may be an important one but, in the
main, changes in the output of a worker are measurements of general
efficiency. They show the combined effects of a large number of separate
though inter-related influences, such as technical improvements, managerial efficiency, the flow of material and components, the relative
contributions of plants at different levels of efficiency, as well as the
skill and effort of workers. When it is found, therefore, that the output
of a worker in a man-year is increasing, this increase may be attributed
to any or to the joint effect of all these factors.1
Patently the data on labour productivity do not measure the specific
contribution of any single factor in production, be it labour, capital
equipment, management or any other component in the whole process ;
they register instead the combined results of all the distinct but interrelated influences that determine output per man-hour. Yet equally
patently the long-term gains in efficiency stem mainly from the advance
of science and technology and its application to production. Predominant have been the contributions of manufactured power; of varied
improvements in equipment, ranging from the steady accumulation of
small changes to the introduction of highly specialised and semiautomatic or automatic machines; of parallel improvements in the
organisation of the production process, from more efficient layout to
assembly-line production; and so on. Other factors which, if they
have had less potent influence upon productivity within the long-run
pattern, have none the less played their continuing part at any given
time, include the scale of operations, the level of plant efficiency, the
location of the shop, the quality and flow of materials, the skill and
attitude of the work force, and so on.2
While the action of all the various factors is often blurred in
the over-all picture shown by productivity data, certain factors
are sometimes brought to light under the influence of special conditions; this was the case in Germany immediately after the end
of the second world war:
Average efficiency of labour is about half as high as before the war.
This condition has a number of causes. Shortage of food and insufficient
transportation have reduced the physical fitness of individual workers.
The average quality of the working population is reduced by the selective
effects of war losses. Because of bad living conditions absenteeism is
very high. Lack of capital equipment and industrial integration and
the great amount of work going into repair are all forces increasing the
use of labour per unit of production. Furthermore, in an interpretation
of a possible future extension of production, enterprises are pursuing a
policy of storing labour.3
The following list illustrates the number and variety of the
important factors influencing labour productivity; they have been
grouped under three main titles, in order to facilitate analysis.
1
2
3

L. ROSTAS: " Output a Head ", in The Times, 10 Nov. 1948.
Harvard Business Review, Vol. XXVII, No. 3, May 1949.
OFFICE OF MILITARY GOVERNMENT FOR GERMANY, United States Zone:
The Population of the United Slates Zone of Germany, Part II, Nov. 1947.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

General Factors
Climate.
Geographical distribution of raw materials.
Fiscal and credit policies.
General organisation of the labour market.
Proportion of the labour force to the total population, degree of
unemployment, of labour shortage and of labour turnover.
Technical centres and information concerning new techniques.
Commercial organisation and size of market.
General scientific and technical research.
Variations in the composition of the output.
Influence of low-efficiency plants and their varying proportion in
total output.
Organisation and Technical Factors
Degree of integration.
Percentage of capacity used.
Size and stability of production.
Quality of raw materials.
Adequate and even flow of materials.
Subdivision of operations.
Balancing of equipment.
Multiple machine systems.
Control devices.
Quality of output.
Rationalisation and standardisation of work and material.
Layout and location of the plant.
Maintenance and engineering services: safety, light, sound, ventilation, air conditioning, telephone, etc.
Availability, fitness and accessibility of tools.
Wear and tear of machines and tools.
Amount of machinery (or power) available per worker.
Proportion of maintenance labour to operating labour.
Length and distribution of working hours.
Selection of personnel.
Human Factors
Labour-management relations.
Social and psychological conditions of work.
Wage incentives.
Adaptability to, and liking for, the job.
Physical fatigue.
Composition (age, sex, skill and training) of the labour force.
Organisation of the spirit of emulation in production.
Trade union practices.
Tables I and II give some numerical indications of the changes
in labour productivity attributed to various factors. Table I
refers to experience in various countries, table II to increases in
productivity in the U.S.S.R. in the period following the first
world war.

FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR
TABLE I.

EXAMPLES OF INCREASES IN OUTPUT DUE
TO NON-MECHANICAL FACTORS

Method

Place, industry or operation

Arrangement of premises
and workplaces

German chemical factory

Shape of benches

Swiss food factory

Improved lighting

United States
13 cases
3 cases
1 case
1 case
Bricklaying i
Assembling parts of textile machinery
Manufacture of tin boxes
Shirt factory

Systematic arrangement of
tools
Adjusting stand to height
of worker

Apple packing (United States)

Selection of workers

Pig charging (Bethlehem Steel Works)
Inspection of ball bearings
Miscellaneous

Training of workers

Swiss experiments
Ohio State University studies (average)
Manufacture of tins
Nailing machine
Biscuit packing
Baking operations
Shelling almonds
Addressing

Rationalisation of movements

Ohio State University studies:
Folding cloth
Dipping chocolate
Packing chocolate
Attaching labels
Filing forks
Assembling carburettors
Rope pleating
French experiments:
Typesetting
Type correcting
Rhythmic motion (machine)

. . . .

Chain work

France:
Fitting brake levers
Inspecting brake levers
Bicycle factory
Germany :
Fitting brakes
Repairing locomotives
Motor manufacture
Manufacture of women's underwear .
Manufacture of electrical appliances .
Czechoslovakia: packing webs
Sweden: Manufacture of cooking utensils.

Rest pauses

Manufacture of pressed cardboard . . .
Assembling bicycle chain
Operating stamping machine
Manufacture of sewing machines . . .
Krupp Factory (Austria)

Incentives

United States: Manufacture of tin tubs.
Germany: Manufacture of tin tubs. . .
Czechoslovakia: Bata boot factory . . .
France: Metal industries

Source: UNITED STATES DEPARTMENT OF LABOR, BUREAD OF LABOR STATISTICS:

Monthly Labor Review, Vol. 37, Nov. 1933, pp. 1,029 et seq.

i Gilbreth's examples (cited by F. W. TAYLOR: Principies of Industrial Management).

METHODS OF LABOUR PRODUCTIVITY STATISTICS
TABLE II.

FACTORS CONTRIBUTING TO AN INCREASE
IN THE PRODUCTIVITY OF LABOUR

Industry

Percentage
increase in
product- Most important factors contributing to increase
ivity of
labour

Anthracite

20.5

(1) Increase ot the mechanical process
for obtaining raw material by 16 per
cent.
(2) Increase of delivery by mechanical
means.
(3) Various organisation
improvements.

Machine construction

21.9

(1) Growth of supply to factories.
(2) Specialisation of production.
(3) Introduction of new equipment and
organisation of a conveying system.
(4) Improvement of the intercorporative and corporative planning of
work.

Textiles

19.7

(1) Concentration of work.
(2) Better utilisation of machines.
(3) Organisation methods (output per
person).

Basic chemicals

28.5

(1) Utilisation of new equipment.
(2) Increase of supply.

Rubber

37.6

(1) Shortening of working day.
(2) Change of assortments.
(3) Introduction of new equipment.

Match

31.9

(1) Operating of new factories.

Electroteehnical

33.4

(1) Increase of supply.
(2) Improvement in quality of working
staff.
(3) Introduction of new equipment.

Source: S. M. KINGSBURY and M. FAIKCHILD: Employment and Unemployment in
Pre-War and Soviet Russia, report to the World Social Congress, United States, 1931, p. 65.

Some illustrations are given below of the influence of some of
the factors listed on page 18.
GENERAL FACTORS

(a) The general influence of extreme heat or cold, high humidity
or excessive drought on labour is well known, and explains in
large part the concentration of highly productive countries in the
temperate zones.

FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR

In the equatorial and tropical climates, hot and more or less
humid, intensity of work is necessarily lower than in the temperate
or cold climates. In the latter, the worker can furnish average
work of 8 kilogrammeters per second, but much less in the tropics.
The adverse conditions of equatorial and tropical climates are
reflected in physical abilities: capacity for attention, correctness of
interpretation of instructions, speed of reaction, etc.
(b) In a study of the causes of the relatively low productivity
in France, the Office of the Commissioner-General for the Plan
pointed out certain general factors that exert an unfavourable
influence ; productivity of labour is understood here from the point of
view of the best utilisation of the nation's manpower. The remarks
made by this Office would of course apply to many other countries.
Reference is made in the study to the drawbacks of the fiscal
system, and to the distribution of credit on the basis of financial
criteria, in which the degree of organisation and the ability to
produce economically are seldom taken into account. Industrial
organisations it is stated, contribute insufficiently to the organisation of markets, especially in the field of commercial information ;
the result is a waste of effort and time in the search for customers
or suppliers, and in the search for products by consumers. The
lack of liaison between scientific research and industry is prejudicial
both to research and to the practical application of new processes
and techniques. There is also a general fear that the expenses of
research into new processes, or of " indirect " workers planning
the execution of the work, might be useless, but it is noted that in
those countries where productivity is high the proportion of such
indirect workers is high—and is increasing. The lack of technical
centres in a country where small and average-sized undertakings
are predominant is said to render difficult the exchange of technical
information or the spreading of new techniques. Methods of
modern organisation are not widely taught, and are therefore
seldom put into practice. This is confirmed by the observations
of a United States writer:
Basic scientific research and technology are at least as far advanced
in Europe as in the United States, but the application of technology to
industrial methods has not progressed so far. In short, America has
more " know-how ". During the past few decades, industrial methods,
factory organisation, and manufacturing equipment have developed
along different lines, and as a result productivity has increased
in the
United States at a greater rate than in other countries.1
1
E . C L A G U E : Productivity,
Employment,
and Living Standards,
report
prepared for the Conference on Productivity, Milwaukee, 4 J u n e 1949, p . 6.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

(c) The changing composition of production constitutes one
of the most important of the general factors influencing productivity
data computed in such a way t h a t they reflect these changes.
It is very clear that during the war period a violent alteration in the
production pattern did occur. It is also clear that the direction of the
shift was towards the production of items where gross value added per
worker is high. This shift in the composition of production in part
explains the increase in gross national product per man-hour. For this
reason, it is dangerous to label the change as an increase in productivity.
To some extent the shifts in the production pattern have already
reversed themselves, and it is almost certain that gross national product
per worker will decline from the war-time peak. *
(d) The varying proportions of the output of low-efficiency
plants to the total has a similar influence.
In a study of the changes between 1939 and 1946 in factory
man-hours expended per unit of production in a number of leather
tanneries in the United States, most firms reported that unit
man-hour requirements increased or decreased less than 10 per cent.
in this seven-year period ; however, a number of firms experienced
increases as high as 47 per cent, and declines of more than 50 per
cent. If performance in the less efficient plants was improved
to the level of some of the better (not the best) firms, the average
productivity level for the industry would increase considerably.
Such an improvement would require no new discoveries, but simply
involve the adoption of techniques already in use elsewhere in the
industry.
In the United Kingdom it was reported in 1946 that—
there is an enormous range of technical efficiency (as measured by
output per man-shift) between good and bad coal mines. In December
1944 the over-all average output per man-shift was 20 cwt. But 9 per
cent, of the total output was produced in mines with an output per manshift of less than 15 cwt.; on the other hand, 23 per cent, of the total
coal was produced in mines with an output per man-shift of over 25 cwt.,
and 7 per cent, in mines with an output per man-shift of over 30 cwt.
Thus the inefficiency of the British coal mining industry largely takes
the form of a long weak tail. Cutting out the least efficient mines, total
output would decrease by about one fifth and output per man shift
would rise by over 12 per cent.3

1
D. EVANS: " Recent Productivity Trends and their Implications ", in the
Journal of the American Statistical Association, Vol. XLII, June 1947.
2
J. JEWKES: " I S Britain's Industry Inefficient?", in The Manchester
School, Vol. XIV, No. 1, Jan. 1946.

FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR

ORGANISATION

AND T E C H N I C A L

FACTORS

(a) Many general statements have pointed out that one of the
most important of the factors determining the level of labour
productivity is the extent to which capacity is fully utilised.
The rate of activity at which a particular industry is operating,
—e.g., 25, 50, 60 or 100 per cent, of capacity—would seem to be the most
important of the non-technological determinants of labour productivity. 1
In any industry, and particularly in seasonal ones, monthly
productivity changes may be dominated by changes in the degree of
capacity utilisation.3
The extent to which capacity is utilised is rarely known. The
degree of under-utilisation of capacity is sometimes hidden; for
example, even accurate statistics of the actual number of hours
worked do not reveal that form of underemployment which consists
in assigning to each worker a smaller number of looms when
orders are scarce—a practice which results in a serious slowdown
of labour productivity. This practice is common in the weaving
industry.
When it is known t h a t the utilisation of capacity varies within
wide limits, statements on labour productivity should be made
with due caution. During the reconversion period, in the United
States it was observed that—
until full capacity operations are reached, a large amount of labour
will be needed without any corresponding large output of finished goods.
No comparisons should properly be made with pre-war performance in
these industries until normal utilisation of productive capacity is
attained. 3
The occupational composition of the labour force varies when
the proportion of the productive capacity which is utilised varies.
Thus a larger proportion of workers to the total may devote its
time to non-productive tasks (such as maintenance and repairs)
during the period of under-capacity operations than during the
periods of capacity production. This factor alone will affect the
output per man-hour quotient.
The influence of the degree of utilisation of capacity has some1
A. WUBNIG: " The Measurement of the Technological Factors in Labor
Productivity ", in the Journal of the American Statistical Association,
Vol.2 XXXIV, June 1939.
" The Meaning of Productivity Indexes ", loc. cit.
8
" Recent Productivity Trends and their Implications ". loc. cit.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

times been measured. A study made in the United States* showed
that in the brick manufacturing industry, unit labour requirements
increased steadily (from approximately 100 to 120 man-hours for
1,000 bricks produced) between 1925 and 1933. Computations were
made to correct these figures by taking into account the steady
decrease in capacity utilisation (from 80 per cent, to 10 per cent.);
the recorded increase in unit labour requirements was practically
cancelled by this correction.
Steel production affords another example:
Modern steel mills are designed for high-speed volume production,
and when they are not operating close to capacity they require excessive
amounts of labour per ton of output; the unit labour requirement per
gross ton of finished steel products at a rate of operating equal to 20 to
25 per cent, of capacity is 46 man-hours as compared with 34 man-hours
when 55 to 60 per cent, of capacity is utilised.2
(b) The size of the plant (or the volume of production) is
another fundamental factor. For instance, when comparing
labour productivity in the steel industry in the United
States and the United Kingdom, the scale of production in the
two countries must be taken into consideration. With an output
of steel approximately eight times larger, it is possible to obtain
a much higher productivity in the United States because of production in much larger units.
In the study mentioned above 1, comparisons are made of labour
productivity in the sugar industry according to the size of the
TABLE III. MAN-HOURS PER TON OF BEET SLICED, 1 9 3 3 - 1 9 3 5
Old plant (built 1890-1909)

New plant (built 1910-1929)

1.85

1.35

Dally capacity of plant
Less than 1,400 tons 1,400 tons and above Less than 1,400 tons 1,400 tons and above

2.3

1.5

1.3

i The plants work 24 hours a day ; the average slicing capacity is therefore a measure
of their size.
1
D . W E I N T R A U B : " Some Measures of Changing Labor P r o d u c t i v i t y and
their Uses in Economic Analysis ", in t h e Journal of the American
Statistical
Association, Vol. 33, Mar. 1938.
2

F A B R I C A N T : op.

cit.

22.

FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR

plants. The comparison of plants with a daily slicing capacity of
less than 1,400 tons of beet with those slicing more than 1,400 tons
shows the combined difference in size and age of the plants
(table III).
(c) It is generally recognised that one of the main factors
explaining the high productivity of the United States worker in
manufacturing is the fact that he has more mechanical equipment
to assist him. The advantage of mechanised processes is well
known. The same study 1 shows the differences in labour productivity in mines where coal is loaded by hand and those where the
loading is mechanised:
TABLE IV.

ILLINOIS COAL MINES: INDEX OF MAN-HOURS EXPENDED

PER TON EXTRACTED (1923 = 100)

All mines
Hand-loading mines

1923

1935

100
100
100

81
100
70

i Mines which had mechanised loading of 30 per cent, or more between 1923 and 1935.

(d) Many undertakings, especially small ones, make a number
of different articles. Constant adjustment is therefore necessary,
and the workers are compelled to change their work frequently,
to the great loss of productivity. Here is seen another advantage
of specialisation and standardisation, which promote mechanisation
and chain production, thereby increasing labour productivity.
(e) The organisation of maintenance, safety, sound, heat,
ventilation, etc., is also an important influence in labour productivity. Considerable losses could often be prevented in undertakings
by a better organisation of safety.
When it is possible to determine and let the workers assimilate the
main occupational and safety reflexes connected with the task, then
we shall have the best labour and output conditions. The number of
mistakes and oversights, the number of wrecked machines and damaged
tools—and also the number of accidents to workers—will fall to a few
hundredths of what they are at present. The result will be a saving of
hands, of hours of work, and of insurance and treatment costs, fewer
crippled, fewer unfortunates, and a saving of raw materials. In short,
everyone will be better olì.
3

" Some Measures of Changing Labour Productivity and their Uses in
Economic Analysis ", loc. cit.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

How many eye injuries are due to the fact that thousands of workers
refuse to wear their safety goggles ? Oversight, neglect, indifference ?
When the wearing of goggles is made an automatic reaction, eye injuries
may become almost a thing of the past; there are undertakings in which
this has been and is still being proved.
Here, then, is a possibility of effecting a huge increase in individual
and collective output. 1
The following examples show increases in labour productivity
due to maintenance improvements:
The noise level was reduced from 45 decibels to 35 decibels among
a group of workers in an insurance office who were engaged in a variety
of machine operations. Although no other changes were made in the
office, a 12 per cent, increase in output followed the reduction in the noise
intensity. This improvement was so great that the officials were
inclined to attribute a portion of it to added skill from practice, although
the workers were experienced at the time the change was made.
Lowering the noise level from 50 to 35 decibels in the telephone
operating room of a telegraph company resulted in a 42 per cent, reduction in errors and a 3 per cent, reduction in the cost per message.2
(f) A further example of the influence of organisation factors
is that of the length of the working day or week :
A widely prevalent reason for reduction in the hours of labour
—at least during the earlier part of the century when the work week was
very long—has been the experience, or at least the presumption, that the
average worker can actually produce about as much in 10 hours as in a
12-hour day, even with all other conditions of production unchanged.
Classic examples often cited are the experiences of the Engis chemical
factory near Liège in 1892, the Salford iron works near Manchester in 1893,
and the Zeiss optical plant in Jena in 1900. In all of these plants
moderate reductions in hours per day were accompanied by increases in
output per man-hour, and in two by increases in output per worker.3
An article analysing the results of an enquiry into the relation
between hours of work and output contains the following comment :
If the work is light and if the worker is in complete control over the
operating pace, the eight-hour day and the 40-hour week appear to be
more efficient, in terms of maximum output per scheduled hour of work,
than longer schedules of daily and weekly hours.4
1
F. BILLON : " Psychological and Human Aspects of Vocational
Training ", in the International Labour Review, Vol. LIX, No. 5, May 1949,
pp. 491-492.
.2 " Effect of Noise upon Efficiency ", in the Monthly Labor Review,
June8 1930.
4

S. FABRICANT: op. cit., p. 13.
M. D. KOSSORIS: " Hours of Work

Review, July 1947.

and Output", in the Monthly Labor

FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR

HUMAN FACTORS

There seems little need to insist on the fact t h a t labour productivity depends a great deal on the skill of the worker, and
therefore is modified by better training and the average skill of
the labour force. For instance, in an enquiry into the concrete road
building industry in the United States, the rise in productivity
from 1919 to 1930 was attributed to the introduction and increasing
use of labour-saving machinery, and to the increased skill of
workers in.the operation of this new machinery.
The systems of remuneration of work may have a considerable
influence on the progress of productivity. While wage incentives
tend to increase productivity, the reduction of the differences
between the wages of unskilled and skilled workers (such as resulted
from many inflation measures during the second world war) renders
unattractive the necessary effort for higher qualifications, and will
thus hinder productivity increases.
More attention might be given to general tendencies developed
b y the fostering of the spirit of emulation or b y trade union practices. In the Union of Soviet Socialist Republics, the stakhanovist
and other movements to stimulate output have been considerably
developed :
The term " stakhanovist " is applied in the Union of Soviet Socialist
Republics to " shock " workers who have completely mastered the
technique of their trade, are able to make the most efficient use of their
tools and equipment, work rationally and without waste, and regularly
exceed the prescribed standards of output. The extent to which a
worker must exceed these standards in order to have the right to be
called a stakhanovist is fixed for each undertaking with reference to its
particular technical conditions.
In some branches of industry the best stakhanovist workers—those
whose output is double the prescribed standard or more—are given the
title " master of socialist work ". 1
On the other hand, trade unions have often been charged in
certain countries with advocating restrictive practices—impeding
the introduction of technological improvements in processes and
machinery, and placing restrictions upon job content, incentive
systems, discipline, the allocation of work, the use of prefabricated
products, recruitment and dismissal, and sometimes promotion.
However, the authors of a recent study of productivity and
1

B . L. M A R K U S : " T h e S t a k h a n o v Movement and t h e Increased
P r o d u c t i v i t y of L a b o u r in the U.S.S.R.", in t h e International Labour
Review,
Vol. X X X I V , N o . 1, J u l y 1936, footnote on p. 1 1 .

METHODS OF LABOUR PRODUCTIVITY STATISTICS

labour relations state t h a t in spite of this long-standing identification of output restrictions with trade union policy, day-to-day
observation and systematic study have revealed that the reactions
which give rise to restrictive practices are not the result primarily
of trade union organisation and collective bargaining. These
reactions are rather basic human responses appearing among all
workers, non-union as well as union. The report continues:
It is important to realise, however, that unionisation does introduce
its special influences. If the patterns of restrictive behaviour exist
among workers before their entry into a union, labour organisation gives
this behaviour institutional expression and thereby multiplies the
workers' resources of defence against the dangers they fear.
Thus in trade and service occupations, where unions have as yet
made only slight inroads, productivity has so far registered only slight
advances; in railroad transportation and mining, where unions have
long been established, large gains in productivity have occurred over
the years. 1
After this general review of the factors affecting labour productivity, it should be immediately evident that the application to duty of individual workers must play a relatively small
part in the total changes in manufacturing output per manhour over an extended period. In fact, while Dr. Fabricant's
figures indicate that manufacturing output per man-hour more
than doubled between 1919 and 1939 2 , it is difficult to believe
that the average application of several million manufacturing
workers could have undergone any such proportionate change in

those two decades.
However, in most discussions of productivity data, there is
often the implicit assumption that production per man-hour is
primarily a function of the mental and emotional attitude of the
worker toward his job.
This belief that productivity is an attribute of the individual probably
stems from two sources. First, it is a commonplace of factory experience
that employee " slow-downs " can seriously hinder production, while
high morale and general enthusiasm can appreciably increase it. These
facts, of course, apply particularly to short intervals of time rather than
to periods several years apart; they represent fluctuations about a
general trend, not the trend itself.
Second, in activities where skill and craftsmanship play an important
part, we have all observed the vast differences between experts and
beginners. A skilled painter can paint my whole house about as fast
as I can do one room—and do a better job of it, too. A skilled carpenter
1
B . M. a n d S. K. S E L E K M A N : " P r o d u c t i v i t y a n d L a b o r R e l a t i o n s " ,
in t h e Harvard Business Review, Vol. X X V I I , No. 3, May 1949, p p . 377-378.
2

S. F A B R I C A N T : op.

cit.

FACTORS AFFECTING THE PRODUCTIVITY OF LABOUR

can make my best efforts look clumsy and futile. A good assistant can
cut my work in half, while a poor one only adds to my woes. But note
that all of these pursuits are non-mechanical. A good painter with a
spray-gun can far outperform a good painter with a brush; a good
carpenter with power tools can outperform a good carpenter with hand
tools, and a good tabulating machine operator can far outdo the most
diligent longhand tabulator. The more skill and effort that can be
transferred from the worker to his tools, the less imposing does individual
artistry and attitude become. In this age of mechanisation the trend
has been increasingly away from the skill and volition of the individual
worker toward the speed, power and consistent performance of the
machine. This trend has been especially pronounced in manufacturing.1
From the preceding analysis, it should be clear that variations
in the ratio of the physical output per man-hour during different
periods can be interpreted as indicators of changes in the relative
average effort of the labour force only if—
(a) a single plant, turning out a single product, is considered;
(b) it is assumed that the product manufactured by this plant
does not undergo any changes whatever in its specifications;
(c) it is assumed that the raw materials used in the process of
production remain similarly unchanged; and
(d) it is assumed that no changes in any of the manufacturing
methods or techniques (including installation of new equipment)
take place within the plant.

1
C. B. YOUNG: "Applications and Problems of Productivity
in the Journal of the American Statistical Association, Jan. 1946.

Data",

CHAPTER III
PROBLEMS IN THE MEASUREMENT OF LABOUR
Labour is one of the two basic elements in the computation of
labour productivity. In the present chapter several points referring
to the definition and measurement of this factor will be examined.
Labour, in the most general sense, is an effort—mental or
physical—applied during a certain time. 1
Unfortunately, the quality of effort, and more particularly
of mental effort, is scarcely ever measurable:
The productivity of labour is measured by comparing the labour
input and the quantity produced during a given time. But we meet
here a very serious difficulty: how can the labour input be measured
if not by the quantity produced ? The work of the worker, or his effort,
is not merely mechanical; the choice of the point of application of his
effort is of fundamental importance. So even if it were possible to
measure physiologically the mechanical effort, how could the corresponding cerebral effort be measured ? One has therefore to admit
that labour input is not, in practice, measurable.2
The obviously measurable factor is the time during which the
effort is made. Thus, labour is ordinarily measured by the number
of man-hours spent; even when, instead of man-hours, employment
is used as a measure of labour, such a measure corresponds to
" man-years " and is therefore again a measure of time worked
(see below).
Since only the time during which the effort was contributed is
measured and not the actual effort spent, differences in the quality of effort will influence the results, and therefore effort will be one
of the factors affecting labour productivity; it may be the origin
of many false interpretations of differences in labour productivity,
which too often have been ascribed solely to differences in effort.
The preceding chapter has shown t h a t many factors other than
effort influence the ratio of output to labour time contributed.
1
In Das Kapital, Karl MARX defines labour as the use of labour power,
and the latter as the aggregate of those bodily and mental capabilities existing
in a human being.
2

G. DEURINCK: op. cit., p. 360.

PROBLEMS IN THE MEASUREMENT OF LABOUR

Had it been possible to measure effort as well as the time during
which it was contributed, such errors of interpretation would not
have been made. In fact, measures of productivity of labour
have often been computed in order to measure the effort
contributed. Conclusions on this point are correct only when
no other factor than effort has varied. This condition narrows
considerably the field within which such conclusions can be
drawn—in fact, they can only be drawn when the output of two
workers is compared who are working under identical conditions
(machines, quality of materials used, etc.). Such comparisons
cannot be extended with the same interpretation to departments,
plants or industries.
It may also be noted that in some cases technological changes
permit an actual increase in the productivity of labour (as understood by the total effort contributed for a given output) but may
not appear in productivity data as measured b y ratio of output
to labour time. Such is the case when the introduction of a laboursaving device merely lessens intensity of effort but does not decrease
time spent. For example, the fireman is required on the locomotive
whether it is hand or stoker-fired. However, any decrease in the
intensity of labour expended probably makes it possible for a
fireman to handle a longer run more effectively, and ordinarily
saving of effort and saving of time are closely related.

DIFFERENT K I N D S OF LABOUR

The amount of labour input in a given output, even when
measured in time spent, is still not completely determined. For
certain purposes, only that labour which is directly devoted to
the production of the goods under consideration will be relevant;
for other purposes, supervisory labour, management, and even
the labour entering into the manufacture of tools used or the sale
of the goods produced may have to be taken into account.
It has been suggested t h a t the following distinctions should
be considered:
Operating labour is that required directly in a particular process,
such as the operation of a brick-moulding machine. Auxiliary labour
is that required in the plant for such operations as oiling, inspecting,
adjusting and repairing the machine—in short, all plant labour that
is necessitated by the use of the machine but is not considered as engaged
in its direct operation. Embodied labour is the labour applied to the
production of the machine itself and the materials of which it is made
(pro rata over the useful life of the machine), the materials used in

METHODS OF LABOUR PRODUCTIVITY STATISTICS

machine repairs, the power, oil, grease and other materials, if any, consumed in the operation of the machine. Indirectly required labour is the
labour required beyond the manufacturing stage, for transportation
and marketing. While studies in productivity and labour displacement
ordinarily stop short of the point of allowing for all these indirect
factors, we should recognise that until they are included we have not
ascertained the real change in productivity or the real amount of labour
saved or displaced.1
An illustration of the different results that are obtained when
considering only part or, on the contrary, all of these kinds of
labour is given in table V.

TABLE V.

HYPOTHETICAL CALCULATION TO ILLUSTRATE THE
MEANING OF LABOUR SAVING

Number of
workers required *

Type of labour

Output per
man-day

Percentage
increase
in productivity

New
Old
New
Old
machine machine machine machine

Moulding machine operators
Auxiliary workers
. . . .
Total, operators
and auxiliary workers .
Total plant force . . . .
Total, plant force
plus embodied labour .
Indirectly required labour .
Total, plant force plus embodied labour and indirectly required labour

70
10

50
5

28,571

40,000

40.0

80
2,000
20

55
1,975
30

25,000
1,000

36,364
1,013

45.5
1.3

2,020
200

2,005
215

990

998

0.8

2,220

901

Source: H. JEROME: op. cil.

i For a total output of 2,000,000 bricks per day.

Embodied Labour
An important, although rather theoretical problem, is whether
" embodied labour " should be considered as forming a part of
the input of labour corresponding to a certain output.
A work that appeared recently in France puts forward the view
that estimates of labour productivity should take account of
embodied labour.
1

H. JEROME: Mechanisation in Industry, National Bureau of Economic

Research, 1934, p. 28.

PROBLEMS IN THE MEASUREMENT OF LABOUR

The French farmer has appreciably increased his productivity since
1900; it is obvious that this has been possible only because he uses
agricultural machinery and fertilisers; but this agricultural machinery
has been produced by other men; these fertilisers have been extracted
from the soil, treated, transported and delivered to the
farmer—therefore,
some men have indirectly worked for agriculture.1
The question raised seems particularly important when it is
proposed to compare the productivity of labour of two undertakings
manufacturing the same product from the same material but with
important differences in the degree of mechanisation—for instance, an
undertaking building concrete roads with the help of bulldozers and
one which has no bulldozers at its disposal: it might be contended
that if account is not taken of the labour embodied in the production of the bulldozers, the comparison lacks meaning.
However interesting this concept may be, it is nevertheless
somewhat remote from the notions commonly used in labour
productivity measures, and especially from actual possibilities of
any accurate measurement. When this idea is extended, it becomes
practically identical with a man-hour cost analysis of the production, taking into account all costs entering into the prime cost
of a product: this is so true that Mr. Fourastié, in the work referred
to, practically identifies the prime cost with the unit labour requirements as he has defined them, that is, including even the labour
" embodied " in the remuneration of capital, etc.
In fact, at the present time it would be quite impracticable to
measure with any accuracy the unit labour requirements corresponding to embodied labour. Apart from the labour embodied
in oil, grease, and other materials used to maintain the machine,
what unit labour requirements should be taken into account as
corresponding to the output of a bulldozer: the average of those
observed in various plants, the minimum, the maximum ? Should
it be the unit labour requirements observed in a highly equipped
factory, or in a much less mechanised one ?
In the present state of knowledge of labour productivity, it
appears that data covering embodied labour can only be of a very
approximate nature. Thus Mr. Fourastié suggests that the
" embodied labour " be estimated on the basis of the money
cost of the machines, divided by the hourly wage rates of the
labourer; while such computations may be useful and sometimes
necessary for economic purposes, they do not at present fall within
the scope of a direct analysis. It is only when detailed measures
1
J. FOURASTIÉ: Le grand espoir du ÄÄme siècle, Paris, Presses Universitaires de France, 1949, p. 20.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

of unit labour requirements for many industries, products, and
especially production processes have been perfected and standardised that it will be possible (with all the precautions that this will
involve) to include embodied labour in the measures and comparisons of labour productivity.
Direct and Indirect

Labour

In most studies of productivity attempts are made to consider
two main classes of labour: direct labour, generally (but not uniformly) defined as labour engaged directly in production; and
indirect labour, comprising the various workers in the plant
who are necessary for production but whose contribution
is indirect. An example of the differences between these two
kinds of labour may be found in the special enquiries into productivity conducted since 1946 by the Bureau of Labor Statistics
of the United States. In these enquiries, direct man-hours, related
directly to production, and indirect man-hours are defined in a
manner which conforms with general accounting practice in the
industries. Typical direct man-hour classifications include production or process operators, machine operators, and workers
engaged on activities such as metal working, fabrication, assembly,
sub-assembly, or finishing. Indirect man-hours generally include
supervision, maintenance, the handling of materials, shipping and
receiving, inspection, plant protection, control functions in the
plant, and other categories related to but not assignable directly
to production. General administration, office, engineering, and
sales employees are excluded.
The advantage of such distinctions is that they make it possible
to obtain more precise results:
In certain studies for individual firms we have found that elimination
of all except direct productive labour and indirect hourly labour (excluding factory supervisory and clerical labour, maintenance and repair
labour, which are more or less rigid and do not vary directly with
fluctuations in volume) results in a much more rational index or basis
for an index of per man-hour output. 1
However, standard definitions of direct
have not yet been established. These notions
use. Comparisons in time (and still more
doubtful value until specific definitions are
1

and indirect labour
are not yet in general
in space) will be of
adopted.

U N I T E D STATES NATIONAL INDUSTRIAL CONFERENCE B O A R D :

ing Labor's

Productivity,

F e b . 1946.

Measur-

PROBLEMS IN THE MEASUREMENT OF LABOUR

For example, a report concerning productivity in Sweden
states :
In the computation of hours of work, regard has been paid only
to those groups of personnel who are designated in the official statistics
as "factory workers proper", thus excluding stock and transport workers, mechanics and stokers, watchmen, etc. There is reason to assume
that, owing to the more precise instructions given to suppliers of information on the subject, the relation between these two main groups of
personnel in course of time has shifted somewhat. The tendency seems to
have been to reduce the proportion of factory workers proper, by classifying certain groups of personnel more consistently than before under other
categories, such as stock workers, etc. If so, the figures show a somewhat
greater increase in the output per working hour than corresponds with
the actual facts.1
In a recent comparison of the level of productivity in British
and United States industry, the term " direct labour " is used,
for example, for labour employed in the cement industry, and the
term " indirect labour " for labour employed in the manufacture
of cement-making machinery; in the direct labour of the cement
industry both process workers and overhead labour are included. 2
Besides the concepts of direct and indirect labour, many
similar, though slightly different, concepts have been used. Thus,
the group of " production workers " is approximately the same as
t h a t included under " direct labour ", but usually maintenance
workers are not covered by the latter term ; the term " wage
earners " corresponds to a still broader concept. The term " salaried employees " conveys a broader concept than indirect labour,
and a rather different one; it is used as a supplement to "wage
earners ". These concepts have often been used instead of those
of direct and indirect labour because the latter were not directly
available. But the variations in their scope from time to time, and
especially from country to country, render any comparisons difficult
to interpret.
One of the main reasons for the distinction between direct and
indirect labour pertains to the difficulty of assigning indirect labour
to any particular product. Of course, estimates can be made.
For instance, the indirect labour ascribed to a given product can
be estimated by multiplying the total indirect labour for the whole
1
V. KÄLLSTRÖM: "Efficiency of Industrial Production in Sweden",
in Skandinaviska
Banken (quarterly review), Vol. XXVIII, No. 1, Jan. 1947.
2
L. ROSTAS: Comparative Productivity in British and American Industry,
National Institute of Economic and Social Research, Occasional Papers,
No. XIII, Cambridge University Press, 1948, p. 2.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

undertaking by the ratio of the direct labour entering into the
product to the total direct labour.
But the unsatisfactory character of such estimates is obvious.
It therefore seems preferable to attempt to make a clear distinction
between direct and indirect labour and to compute separate figures
for each of these kinds of labour.
In such computations, the use of man-hours expended per
unit has a great advantage over output per man-hour; thus,
unit labour requirements of direct labour may be computed for
each individual product or operation in a plant; man-hours of
indirect labour expended per unit may be computed for a whole
department or for the whole plant only. Such methods, closely
related to the actualities of the production process, have been
broadly applied in the special enquiries made in the United States. 1

OUTPUT PER MAN-HOUR AND OUTPUT PER MAN

Two alternative measurements have been used in studies of
labour productivity: output per man and output per man-hour.
The former relates output to the number of persons employed,
disregarding the number of hours that these persons worked; the
latter refers to the total number of man-hours worked, disregarding
the number of persons who contributed these hours.
In many instances, output per man has been used instead of
output per man-hour, simply because this calculation was easier
from the data available; but the consideration of the essential
differences between output per man and output per man-hour
shows that these measures have a distinct meaning and illustrate
particular aspects of labour productivity; they may not, therefore,
be utilised indifferently.
At first sight, since the productivity of labour is measured
by comparing output to the corresponding input of labour, it
might appear that only that time during which labour has actually
been engaged in production should be taken into consideration.
Such a view, however, proves too narrow. It seems necessary, for
instance, to include in the time contributed to work the interruptions of work for rest; thus, if in each hour a staff of punch card
operators works 50 minutes and rests 10 minutes, it would seem
unreasonable to count only the 50 minutes as time worked, since
1

See Chapter VI.

PROBLEMS IN THE MEASUREMENT OF LABOUR

the production—and productivity—of those 50 minutes is possible
only when a rest period of 10 minutes per hour is provided.
It will be seen, then, t h a t the concept of " hours worked "
needs further definition. It can hardly be reduced to the time during
which the worker is actually expending his full effort.
If the labour contributed, measured in time, cannot be restricted to the time of full effort, the field is open for much more
enlarged concepts, which will include holidays, sickness, meal time,
etc. Finally, the widest extension of the concept of man-hours
may approach the concept of output per man in which the time
counted is the full day. However, the difference between employment and man-hours worked includes periods of time which cannot
be considered in any sense as a contribution to work—for example,
the compulsory idleness of part-time workers or the voluntaryidleness of absentees who of their own free will take a day off.
Apart from theoretical considerations, the form under which
the data are available in the undertakings gives the basis of most
of the practical concepts of man-hours worked.
In estimating output per worker, we may have in mind either the
workers employed in producing the particular product who are on the
books (i.e., colliery books or the books of cotton mills, etc.) or the
workers on the payroll or the workers actually at work. Workers on
the books may include persons who have left the industry altogether;
workers on the payroll will probably include persons who are absent
for part of the week, those who left during the week or were taken on
during the week or those who went on holidays during the week, all of
whom worked therefore only part of a week. Workers on the payroll
may or may not include those on holiday for the whole week. " Workers
actually at work " would exclude both the absentees and those on holiday
and would make allowances for those working only for part of the week.
On the other hand, in estimating " output per man-hour " there will
be differences between productive hours and hours actually paid, and
the latter, at least for time workers, include payment of time spent in
meal-breaks and other non-productive time. The widest concept is
" output per man-year ", inclusive of absentees and those on holidays,
and this is the most useful concept when estimating national incomes
or labour requirements. In measuring productivity in the technical
sense, output per worker actually at work or output per productive
man-hour (the latter being the narrowest concept) are relevant; while
measuring costs of production, " output per man-hour actually paid
for " is perhaps the appropriate concept.1
The differences, and different uses, of the two general concepts
outlined by L. Rostas as output per man-hour and output per
" man-year " have been clearly defined as follows :
Comparative Productivity in British and American Industry, op. cit., p. 25.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

While the concept " man-hour " of work is definitely circumscribed
in the sense that every hour consists of 60 minutes, the concepts " manweek " and " man-year " are ever changing because over a period of
years the standard work week may vary in the number of man-days
it contains and the man-days in turn may vary in their man-hours content. It is known, for instance, that during the last few decades the
man-hours content of a standard man-year has declined considerably.
When the standard work week consisted of six days of 10 hours each,
the man-year, allowing for 12 holidays, consisted of 3,000 man-hours;
during more recent years the standard work week in many industries
has been limited to five and a half days of eight hours each or, allowing
for holidays, to a little over 2,000 man-hours. For the purpose of measuring changing volume of output in relation to the time actually worked,
it would therefore be necessary to measure employment in terms of
man-hours of work, but from the standpoint of the number of jobs, the
3,000-hour man-year represents one full-time job for one year in the
same sense in which the 2,000-hour man-year represents one full-time
job for another year. 1
This short analysis shows that in measuring labour productivity
the choice between man-hours or employment as a measurement
of labour will depend on the object in view.
Thus, when measuring productivity with the object of determining the changing volume of ouput in relation to the time
actually worked, the productive capacity of labour, or the cost of
production in labour units, it is better to use the man-hour concept;
when measuring productivity with the object of estimating manpower requirements, employment possibilities, future national
incomes, etc., the concept of output per man is more appropriate.

HETEROGENEITY OF THE LABOUR FORCE

One limitation common to all studies of labour productivity
is t h a t the labour force is treated as a homogeneous entity. So
long as no differentiation is made between workers of different
skill, aptitude or application, man-hours or employment in no way
reflect the degree or quality of effort expended by the labour force
in the production process.
In this respect, it is well to remember that the " worker "
and the " man-hour " are statistical abstractions—that there is,
in fact, no satisfactory measurement of a unit of labour.
The " hour of work " is not a homogeneous concept, since
its influence on production as well as its " effort content " differ
1

D. W E I N T R A U B

Productivity,
Works
mittee, Mar. 1937.

and

POSNER:

Unemployment

Progress Administration,

Technical

and

Increasing

Resources

Com-

PROBLEMS IN THE MEASUREMENT OF LABOUR

widely according to the sex, age, skill and position in the factory
of the person who contributes this hour of work.
For a manual job, for instance, a man's hour of work will on
the average represent more labour than a woman's or a child's
hour, and large quantitative differences may appear in the output
of different adult males. For non-manual work, differences between
individuals are qualitative as well as quantitative and it is difficult
to find any method of measuring or assessing such qualitative
differences. The hours of work of skilled workers are not interchangeable: one hour of work expended by a loom fixer in fixing
a loom cannot be replaced by that of a shoemaker, however
skilled the latter may be in his own job. In one factory, for the
same number of persons employed or hours put in, output may
vary with the distribution of work between manual and nonmanual labour or skilled and unskilled labour. The skill of the
workers will have a direct bearing on labour productivity, and
differences in productivity figures due to this factor could be
eliminated only if the varying composition in skill, age, etc., of the
labour force could be taken into account in the measure of the
total man-hours contributed or of persons employed. If there
were an identical distribution in skill, sex, age, etc., the health
of the labour force and other factors would have to be taken into
account; for example the war-weary populations after the second
world war had, for an identical composition, a much lower ability
to work because of the continued effects of the hardships suffered
during the war.
A technological change which merely permits the substitution
of lower for higher-paid labour does not affect a productivity index;
still, even if output per man-hour remains the same, at least labour
costs have gone down and in some respects there has been progress,
which is not acknowledged by the index because it is unchanged
by modifications in the distribution of labour that do not affect
the total number of hours worked.
The nature of the available statistics usually does not permit
any distinction to be made between hours of different intensity
or hours of different kinds of workers. However, some attempts
have been made to weight the different kinds of labour. It has
often been proposed to use wage rates or earnings as a means of
taking into account the differences in effort (manual and mental)
contributed :
The quantity of employment can be sufficiently defined for our
purpose by taking an hour's employment of ordinary labour as one unit

METHODS OF LABOUR PRODUCTIVITY STATISTICS

and weighting an hour's employment of special labour in proportion
to its remuneration, i.e., an hour of special labour remunerated at double
ordinary rates will count as two units. 1
It may be that for some purposes the weighting of labour by
earnings may present a more homogeneous picture of the labour
force. But there can. be no doubt t h a t for productivity measures,
and especially for comparisons, such weighting would introduce
variations unrelated to those which are to be measured: in a
factory which always produces the same total output with the
same labour force working the same time, the weighting of manhours b y wages would give risa to variations in productivity
figures due solely to variations in the distribution of wages; for
instance, if a raise in wages was given to part (say, the unskilled
workers) of the employees. I t may be answered t h a t the earnings
should be used as fixed weights only, but then the data necessary
for such computations on a fixed weight basis are not available
and would be extremely difficult to gather.
Attempts have also been made to overcome differences in the
sex and age composition of the labour force by using a concept of
" equivalent men " or " equivalent man-hours " :
The units of measurement are in terms of man-equivalent hours,
i.e., the farm time used by average adult males in growing and harvesting,
for example, an acre of wheat or corn, or feeding and caring for a dairy
cow for a year. Because many women, children and older workers
accomplish less in an hour than an average adult male, the total of
actual hours of farm work in any given year will exceed the total of
estimated man-equivalent hours.2
Such computations can only be made on the basis of arbitrary
ratios, for instance, by using a scale where time worked b y women
is counted as only four fifths; but in any case the determination
of such scales is always very questionable.
That wide differences do exist in quality and skill is a matter of
common experience; and presumably the difficulties encountered and
the methods utilised in dealing with them would be analogous to those
involved in the computation of index numbers of physical output and
of prices. But we can do no more than mention the problem, since very
little in the way of statistical material is available. Simple aggregation of numbers is the only practicable solution. Changes in
quality and skill of labour and in effort exerted are therefore relevant
to the interpretation of these summations. 3
1

J. M. KEYNES, The General Theory of Employment, Interest and Money.
G. T. BARTON and M. R. COOPER: " Relation of Agricultural Production

to Inputs ", in the Review of Economics and Statistics (Cambridge, Massachusetts), Vol. XXX, 1948.
3

S. FABRICANT: op. cit., pp. 171-176.

PROBLEMS IN THE MEASUREMENT OF LABOUR

MEASUREMENT OF LABOUR

When it is intended to compute output per man and to construct indices of productivity of labour per head or to compare
such figures from country to country, the data which should be
brought into comparison with output are those on employment.
The problem of definitions, methods, the techniques of the
collection of data, etc., on employment have been studied at
length in the report of the International Labour Office on employment, unemployment and labour force statistics 1; the conclusions
of international debates on this point appear in Resolution I
adopted by the Sixth International Conference of Labour Statisticians in 1947.2 It does not seem necessary for present purposes,
therefore, to review these problems here.
I t has been noted above t h a t the concept of man-hours
could be understood, from a statistical point of view, in many
different ways: hours actually worked, hours paid, etc. Whatever
concept is finally adopted, the collection of the data always involves
problems, and the data commonly lack comparability even between
plants in the same industry. Data on total man-hours worked are
collected regularly and on a large scale in very few countries; in
most cases, only indirect data are available, obtained through
various methods which are often non-comparable. Even in the
United States, where important and detailed studies have been
conducted of output per man-hour in the various manufacturing
industries, the man-hour figures have often been obtained in dissimilar ways:
It should be noted at the outset that statistics for total annual
man-hours worked by wage earners or all employees, unlike those for
production and employment, are rare and usually fragmentary. The
National Research Project indexes for only three industries could be
computed directly and independently from man-hour totals. The
most common substitute method permitted by available statistics was
the multiplication of the employment indexes by relatives of weekly
hours and division of the resulting products by 100. The man-hours
indexes for 27 industries were computed in this manner. A second
method, employed principally for two industries, involved the division
of payroll relatives by an index of average hourly earnings of the wage
earners. Indexes for three other industries were derived from averages
of annual man-hours totals yielded by the two methods. Indexes for
seven others were derived by the multiplication of production and unit
1
Employment, Unemployment and Labour Force Statistics, I.L.O. Studies
and 2Reports, New Series, No. 7, Part 1.
The Sixth International Conference of Labour Statisticians, I.L.O. Studies
and Reports, New Series, No. 7, Part 4, pp. 52-60.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

labour requirements (or by the equivalent method, division of production by output per man-hour). Variants and combinations of the methods
already mentioned were employed in the construction of indexes for the
five remaining industries. Of the 13 group indexes, five were computed
for all or most years from the sums of the man-hours for the component industries, and seven were obtained independently from the
products of employment and average weekly hours. 1
It is not apparent at first sight that multiplying the index of the
average number of workers employed by the index of weekly hours
will yield an index of aggregate man-hours worked during the year,
for a question arises concerning changes in the number of weeks worked
per person per year. However, in the computation of the average number
of persons employed account is implicitly taken of change in number
of weeks per worker, and the average number of workers so calculated
is, in effect, the number of persons that would be employed if all persons
worked a full year; i.e., two persons working half a year each are counted
as one full-time worker.2
The diversity of the methods employed, and the fact t h a t
they are mostly indirect, shows that there are many difficulties
in collecting data on man-hours for an industry or a group of
industries, and productivity figures derived from such data should
be analysed with particular caution.
However, when special enquiries are made in a plant to obtain
simultaneously output and man-hours (whether hours actually
worked or hours paid), it is easier to collect the data on the restricted b a s i s 3 ; in particular, the important distinction between
direct and indirect labour is possible in this case. The reporting
of man-hour data is then confined to those undertakings with
production records which yield accurate statistics of the number
of man-hours associated with the production of specific items.
Where direct man-hour records &re not available, man-hour data
may be derived from labour costs and average hourly earnings;
but as this is computed for and within each single plant, the
information on man-hours is of much greater significance than when
the latter operation is done for an industry as a whole.

1
Production, Employment
and Productivity
tries, op. cit., P a r t I, p p . 49-50.
2

S. F A B R I C A N T : op. cit., footnote on p . 16.

See Chapter V I .

in 59 Manufacturing

Indus-

CHAPTER IV
PROBLEMS IN THE MEASUREMENT OF OUTPUT
As the productivity of labour is the ratio of output to the corresponding input of labour, any method designed to measure this
ratio will involve definitions and measurements of output. In
this chapter some of the difficulties raised by such measurements
will be reviewed; however, it is not intended to examine here all
the problems t h a t arise in measuring production per se, but only
those which have a direct bearing on the computation of productivity ratios.
As already noted in Chapter I, the problems involved in measuring the output of services are often insuperable, because of the
difficulty of measuring services in physical terms ; the only possible
measurements are in terms of value, and in many cases it may be
considered t h a t value does not actually measure output. Thus,
the following discussion will be limited mainly to output in manufacturing, mining, transport and communication, where measures
of output can be given a relatively concrete basis.

PRODUCT SPECIFICATIONS

The first difficulty lies in the unit that should be utilised for
the measurement of a given product. In many cases a number
of units are possible.
The nature of the article itself, for example an egg, may sometimes
be one thing and sometimes another. We can say it is an egg, the egg
of a domesticated hen, and even perhaps a newlaid hen's egg. But it
may be a big egg or a small egg, and its dimensions may be large or small
in relation to its weight. We have the choice, therefore, of measuring
" quantity " in terms of one egg (number), one pound of eggs (weight)
or an egg of some standard dimensions (physical volume).1
In fact, the problem of the choice of unit arises from differences
in " quality " between the different items. If these items were
1
R. WILSON : Facts and Fancies of Productivity, The Economic Society
of Australia and New Zealand, 1947, pp. 6-7.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

absolutely identical, the use of numbers, weights or physical
volumes would yield identical results. Hence differences in quality
are an important factor to be considered in measuring production,
and are especially important in productivity comparisons between
periods of time or between countries.
For instance, in Great Britain, the " saleable mined coal " of
1949 included a good deal of m&terial which before 1939 would
not have been classed as saleable; therefore, when the output of
coal in Great Britain in 1949 is compared with the output in 1937
in terms of saleable mined coal, the comparison is not a true one.
On this subject, one writer in the United States concludes t h a t
" no confusion of meaning is possible when it is said that the production of 1,000 five-cent long-filler cigars by hand requires about
33 man-hours, or that the cement industry expended about four
tenths of a man-hour to produce a barrel of cement just before
the war " 1 ; however, another writer points out that, while no
ambiguity is attached to such statements so long as no comparisons
are made over a period of time, even such commodities as five-cent
long-filler cigars and cement undergo changes in their specifications
with time:
Thus, it has been noted that a barrel of cement produced in 1936 is
chemically and physically much superior to the same cement produced
two decades previously (Philip G. Hudson: "The Construction of
Indexes of Physical Production ", in the Journal of the American Statistical Association, June, 1939, p. 249) and that cigars have been
changing in size and have been undergoing changes and improvements
in the filler contents (United States Wage and Hour Division, Research
and Statistics Section, Puerto Rico : The Leaf Tobacco Industry, Washington, April, 1941, p. 52).a
This idea is further developed in a study already quoted:
If we were to compare the products of American factories turned
out in 1937 with corresponding goods produced in 1899 the greater
number of them would be found to be more durable, more efficient,
less noisy in operation, or otherwise better. But such comparisons
might not lead to similar results if shorter intervals were considered.
To the extent that there have been fluctuations in the rate of improvement
in the quality of fabricated commodities, some of the variation in the
rate of decline in man-hours per unit of product may be accounted for,
though it is unlikely that the quality factors can explain a very large
amount of this variation.
There is some ground for the belief that the long recession following
1929 resulted in a deterioration of the average quality of some goods,
1
2

" The Meaning of Productivity Indexes ", loc. cit.
L. TEPER: Limitations of the Existing Productivity Measures and the
Need for New Studies, report prepared for the Conference on Productivity,
Washington, 1946.

PROBLEMS IN THE MEASUREMENT OF OUTPUT

not entirely counterbalanced in the revival that followed. Thus
there seems to have been a tendency, especially in consumers' semidurable goods like shoes and clothing, for factories to shift their emphasis
to the cheaper grades. Since these require much less labour in fabrication than do the more expensive qualities or styles, the change in the
character of output would be accompanied by a decline in the measure
of man-hours per unit even in the absence of other factors affecting
this ratio. 1
These remarks apply mostly to manufacturing. It should be
noted that the quality of a service is still more difficult to define
or to keep identical : the quality of a manufactured product depends
on physical facts, but the quality of a commercial transaction may
not be directly apprehended.
Modifications in quality characteristics are in fact only part of
the changes in what may be called more broadly " product specifications " :
Discussions dealing with the construction of index numbers of
production and productivity frequently refer to changes which take
place in the product make-up over a period of time as changes in " quality " if the product continues to retain the same general descriptive
name, represents the same or similar use-values to the ultimate consumer,
and is sold by the same commercial units (pounds, units, yards, etc.).
Solomon Fabricant, for example, in The Output of Manufacturing
Industries, 1899-1937 (pp. 37 et seq.), speaks of changes in the physical
characteristics of radios, such as changes in the size of the cabinet, type
of speaker, range of reception, and the ease of tuning, as quality changes.
While the quality of goods does change in time, it is but one of the
several changes which may take place. It seems much more important,
when studying production and productivity, to emphasise over-all
changes in the product make-up without limiting the concept of change
to the qualitative characteristics to the exclusion of others. Such overall changes are best described as changes in the product specifications.
Such a concept encompasses, at least theoretically, the sum total of all
possible variations in the product make-up which may take place over
a period of time. 2
The problems raised by differences in the specifications of a
product are often difficult to solve; they are obstacles to the
computation of many economic indices, such as price indices.
The solution of such problems generally rests upon specifications
of certain physical characteristics (mechanical power, capacity,
etc.), global economic use (for instance, a car can transport so
many persons such a distance at such a speed), or length of
possible use, etc.
1

S. FABRICANT, op. cit., p.

20.

Limitations of the Existing Productivity Measures and the Need for New
Studies, op. cit.
4

METHODS OP LABOUR PRODUCTIVITY STATISTICS

In a few instances, the difficulties so raised have been at least
partly overcome through a study of the component parts of a
new product.
For example, an electrical appliance manufacturer, observing that
a television set is much more complicated than any pre-war radio set,
feels that adequate productivity comparisons can be made on the basis
of established standards which break operations down into smaller
elements.
Another method employed by an instrument and controls manufacturer to measure and compare productivity, despite major changes
in product, embodies two approaches. Direct labour was related to
direct manpower standards which had been time studied. In addition,
over-all results were analysed through a cost-analysis approach. " On
this basis, by analysis between our price increases compared with our
material, wage and salary increases as related to gross profits, it seems
fairly evident that we have some increase in physical output over and
above the efficiency improvement in our direct labour ".1
DEFINITION OF A PRODUCT AND OF AN INDUSTRY

Assuming the question of product specifications has been
solved, data must be collected concerning a given product, which
must, therefore, be defined.
The definition of a product is often predetermined by the
classification used by the agency collecting the production statistics. In general, such agencies request the quantities of
" economic " products only, i.e., of products intended either for
consumption in further fabrication within or outside of a particular
industry or for sale to dealers or ultimate consumers. "Uneconomic " output, such as waste, is usually not reported, but saleable
by-products are reported. The criteria probably used in distinguishing products are physical characteristics and composition,
function or use, process of manufacture, unit and aggregate values
as compared with those of other recognised primary products of
the same industry.
It should be noted that, whatever the criterion used, each
reported product really represents a number of more or less heterogeneous items, or a " range ", within which the gradations may
be either imperceptible or marked.
The degree of detail of the classification is determined largely
by the nature of the industry's products, the traditional accounting
system, the purposes of the collecting agency, the form of the
reporting schedules, etc.
1
H . E. H A N S E N : " P r o d u c t i v i t y on the Increase ", in Survey of
Practices, J u n e 1948.

Business

PROBLEMS IN THE MEASUREMENT OF OUTPUT

Actually most of the available production statistics (censuses
of production, production figures collected for the computation of
production indices, etc.) are obtained from establishments; they
are not gathered for a given product. Most establishments produce
more than one commodity, and statistics refer to the establishment
as the unit and not to the commodity. These establishments are
grouped into industries, b u t there is usually a certain amount of
overlapping because establishments (with few exceptions) are
assigned by the statistical agencies as units to one or another
industry according to the product or group of products of chief
value, and because establishments classified in any industry are
engaged not only in the manufacture of the products normally
belorgirg to that industry, but also to a greater or less degree in
the manufacture of products normally belonging to other industries.
First, to what extent does the given industry make only products
described by the name of the industry ? That is, what percentage of
the total value of products of the industry is represented by its primary
(normal) product ? Second, of the total production of the commodities
described by the name of the industry, what percentage is made within
the industry itself ? That is, what percentage of a given group of products is produced in the industry specialising in those goods ? The two
questions are inter-related. If a given industry devotes half its energies
to its primary product, the other half will be given over to commodities
that are primary in another industry.
In 16 of the 242 industries or groups for which we have data on
degree of specialisation in 1929, the value of the primary product constituted less than 80 per cent, of the total output of the industry; in 1931
the number was 15 out of 235. More important was the extent of overlapping. In 1929, out of 224 industries, 29 produced less than 80 per
cent, of the total value of the products in which they specialised; in 1931
the proportion was 27 out of 218. Establishments classified in the cereals
industry produce some feeds as well; establishments classified in the
feeds industry make some cereal preparations also. Another type of
inter-relation, where a large industry is surrounded by satellite specialist
industries, is illustrated by the meat packing, oleomargarine, sausage,
and shortenings industries.
It should be remembered that the statistics understate considerably
the true amount of overlapping, because many of the units relate to
groups of industries. Indeed, the combination into particular groups
is probably attributable to the large degree of overlapping among these
industries. 1
D E G R E E OF INTEGRATION

The actual measure of the physical output of an undertaking
is generally independent of its degree of integration. Whether
the production process of a shoe fabric includes the punching of
1

S. FABRICANT: op. cit., pp. 335-337.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

raw leather, or whether the undertaking purchases cut-out soles,
and, in general, punched leather, its output will be measured by
the total number of shoes produced ; but the total labour required
(for a shorter task) is smaller in the second case. Hence, the
productivity, when measured in terms of the ratio of output to
labour, is higher, yet this higher ratio simply reflects a difference
in the degree of integration. Similarly, whenever industries
undertake the production of semi-manufactures previously obtained
from outside sources, the productivity ratio will decline.
This consideration is not relevant to the study of the variations
in labour productivity in a plant which does not modify its
methods of production over a period of time; but its importance
is clear when such changes do occur, or when the comparisons to be
made affect plants of different degrees of integration.
It might be thought that to overcome this difficulty the
production of a plant should be measured in terms of net output
instead of physical output. (The value of net output is usually
defined as the aggregate value of goods produced, i.e., the value of
gross output, less the value of all commodities and services purchased from other plants and consumed in the production process.)
This measurement would have the advantage of counting as
" output " only that part of the final product of the plant that
has been manufactured on the premises; comparisons would
then appear to be easy between plants of different degrees of
integration; it would even seem possible with these measurements to overcome differences in product specifications, and even
to compare the labour productivity of different industries altogether.
But the value of net output is influenced by many other factors
than the actual work done within the plant. For example, any
modification in the selling prices of the goods produced, whether
due to market fluctuations, or to changes in the wage rates, the
profits, etc., will modify the va^ue of net output, without these
changes being necessarily related to actual changes in productivity;
even more, an actual change in the physical output per man-hour
may be partly concealed, when measuring production by the
value of net output, by a corresponding change in the selling price ;
such may be the case if, for instance, in order to sell more goods,
the selling price is reduced when the prime cost is reduced, the
benefit per unit remaining the same.
Therefore, the value of net output appears to be a very unsatisfactory measure of output for the purpose of measuring absolute

PROBLEMS IN THE MEASUREMENT OF OUTPUT

productivity. However, it will be shown in Chapter V that it
may be of use as a substitute weighting factor in the computation
of labour productivity or unit labour requirement indices.
It seems that the computation of unit labour requirements may
be more helpful to solve the problems raised by the existence of
different degrees of integration.
It has already been noted in Chapter I that the computation of
man-hours expended per unit of production instead of output per
man-hour had the advantage of lending itself to addition (or
subtraction). When considering integrated plants, man-hours
expended per unit can be computed for the various phases of
the production process ; " productivity " in integrated and nonintegrated plants can then be compared for that part of the
production process which is identical or covers the same scope in
both plants.
When computing labour productivity figures for an industry
or a group of undertakings, such corrections are no longer possible.
Establishments in any particular industry are not necessarily
homogeneous with respect to the degree of integration (which may,
furthermore, vary from year to year). Thus, although all the
wage earners in a highly integrated establishment are consistently

accounted for, the quantity statistics may be defective since part
of the production—for consumption in further manufacture on
the premises—is generally not reported. The quantities of the
same goods produced for sale by less integrated establishments,
however, are always reported. Consequently, a varying percentage
of the total output of each product may be reported by each
establishment, and hence by the whole industry, from year to
year. In some instances, however, the subdivision of certain
industries will give greater representation to various important
manufacturing stages in the group production index; by grouping
other industries, the effect of transfers of establishments from
one industry to another may be somewhat mitigated.
In such computations, therefore, the influence of the classification is obvious: the more detailed the classification, the more
informative will the analysis be.
But even for a single undertaking, variations in the ratio of
output to labour may be influenced by modifications over a period
of time in the distribution of work between semi-finished and
finished goods:
If an industry should, for example, devote a larger share of its
manufacturing activity to the production of semi-finished goods at

METHODS OF LABOUR PRODUCTIVITY STATISTICS

certain times, and then complete them in subsequent periods, erratic
behaviour would be noticed in the index of manpower utilisation if the
latter is based on the output of the finished goods. Of course no data,
on an industry-wide basis, which would reflect the more precise measurement of the productive process, are known to exist. In the absence of
such information, the study of changes in manpower utilisation over a
period of time would be enhanced if each statistic were to represent a
sufficiently long production cycle to minimise the effect of production of
partly processed goods and of the completion of goods already in process.
Data covering periods of less than one year's time may be unduly
influenced by the short-term changes in the productive process and thus
be of little use in the analysis of manpower utilisation trends and fluctuations. 1

MEASURES BASED ON PRODUCTION

PROCESSES

The numerous difficulties raised by the differences from plant
to plant of the product specifications, of the degree of integration,
and in general of the differences in production lay-out has led to
the study of the advantages of measuring unit labour requirements
for a production process instead of for a product:
An electric motor produced by firm A is a different product from
an electric motor produced by firm B. The two firms may be manufacturing to meet an entirely different market, and even where the two
products are practically identical, the ratio of " bought out " to manufactured components may vary widely. Comparative assessment of
productivity on the basis of product is thus impossible, but the possibility of comparing productivity on the basis of process remains.
Product may be defined as the item actually produced. It may be
the industrial component from a machine shop, the finished drawing of
a designer or a complete assembly such as a radio or a ship. Process
may be defined as a specialised activity which may be used by companies
making various products. The process of milling, for instance, is
common to an automobile factory or a shipyard, to a toolroom or a
machine shop. The process of internal transport is common to all
factories and the process of typing to all offices. A variety of processes
is almost always necessary to produce any product, and by switching
attention from product to process a greater range of comparison
becomes possible. Many processes are common to firms making
totally different products, so that unlike factories can compare efficiencies. By the relation of the many processes necessary for the
production of a particular product to a common standard, it may
well be possible to arrive at some definite conclusion regarding the
productivity of the firm as a whole, and eventually of the industry as
a whole.
This emphasis on processes rather than product is also more in line
with the thinking of the departmental head or foreman. Indeed, one
1
Limitations of the Existing Productivity Measures and the Need for New
Studies, op. cit.

PROBLEMS IN THE MEASUREMENT OF OUTPUT

great advantage of process measurement is that standards of comparison
can be set for use, not only by top management, but by all grades of
supervisors, foremen and individual workers.1
The measurement of processes appears to offer a means of
avoiding many of the difficulties involved in measurements based
on products, particularly at the level of the plant.

DURATION OF THE PRODUCTION PROCESS

The duration of the production process has an important
bearing on the measurement of productivity. With few exceptions,
the production of a plant can only be measured in terms of
completed goods, and it is often hardly possible to estimate
the production of partly completed goods in terms of finished ones.
If the process of production is short, counting finished goods only
will not affect the measurement of labour productivity to any
great extent. If the process of production takes a long time,
however, measurements of productivity over short periods will be
of little significance.
In other words, there is some relation between the duration of
the production process and the minimum period for which productivity measures may be computed. There is a danger, for instance,
in issuing monthly figures for products involving six or 12 months'
work, that a product will be related to an amount of labour which
is not the amount contributed to the product in question. In this
case productivity is not measured, and the figures issued as measuring labour productivity will be meaningless. It is fundamental that
a certain output be related to the corresponding input of labour
expended on it, and not to any other.
It is for such reasons that the Works Progress Administration decided not to compute monthly indices of labour productivity:
The meaning of monthly indices would not be clear because of the
length of the production process: the reported output of any given
month is attributable not merely to the employment reported
for that
month but also to labour expended in preceding months.2
1
J O I N T COMMITTEE OF T H E I N S T I T U T I O N OF P R O D U C T I O N E N G I N E E R S
AND T H E I N S T I T U T E OF COST AND W O R K S A C C O U N T A N T S : Interim Report on

Measurement of Productivity, London, Dec. 1949, p p . 4-5.
2
Production, Employment and Productivity in 69 Manufacturing
tries, op. cit., P a r t I, footnote on p . 27.

Indus-

METHODS OF LABOUR PRODUCTIVITY STATISTICS

Very obvious examples of the difficulties and dangers of adopting
too short periods for the measurement of output with respect to
the time required for completion of the goods were experienced
during wartime in the United States. Emphasis in war industries
was naturally placed on deliveries of finished goods. A satisfactory
delivery rate, however, did not signify a satisfactory activity
rate, on which future deliveries depended:
'Plane deliveries have frequently been affected by accidental factors
which have little bearing on activity in assembly plants. In October
1942, 'plane deliveries fell 5 per cent, below the number for the preceding
month. The decline reflected the bunching of deliveries on 30 September, many of which under normal circumstances would have been
delivered in October, and the grouping of deliveries on 1 November,
many of which would have been included in October shipments. One
large manufacturer indicated that weather interfered with test flights
so that 59 'planes completed in October were not accepted until 1 November. The total reported for December 1942 suggests other problems
in interpretation. The figure for that month included several hundred
'planes which were accepted but went into the " pool " on the last day
of the month. The 'planes in the " pool " may not be in flying condition; they may lack some part—a propeller, wheels, instrumentation or
other essential items. 1
It is clear that when output (measured in terms of deliveries)
is compared with monthly employment or man-hours, no correct
measurement of labour productivity can be expected. But when
yearly figures, even of deliveries, are compared with yearly employment, or man-hours expended during a year, the influence of the
differences between deliveries and actual activity (which includes
the production of partly finished goods) would be practically negligible.
Hence, in most cases, monthly figures of productivity have
little significance. This is particularly true of seasonal industries:
in these industries also the year's output should be compared with
the average employment of the year or with total man-hours
expended in t h a t year.
MEASUREMENT OF HETEROGENEOUS OUTPUTS

When these problems—the problems raised by the definition
of the product or the industry producing a single product, the
different degrees of integration and the length of the production
process—are overcome, the output, and therefore the productivity
1
I. H . S I E G E L : " T h e Concept of P r o d u c t i v e A c t i v i t y " , in the Journal of
the American
Statistical
Association,
Vol. X X X I X , N o s . 225-228, 1944,
p p . 225-227.

PROBLEMS IN THE MEASUREMENT OF OUTPUT

of labour for a certain product, may be measured without great
difficulty. But when the productivity of labour is to be measured
for more than one product, new difficulties arise as to how the
output of a group of products is to be measured. This problem
arises not only when entirely different products are considered,
but even when different qualities of a product are studied. For
example, should the production of a shoe factory be established on the number of pairs of shoes produced, whatever their
price, or should the low-cost shoes be counted separately, i.e.,
should production be divided according to the price of the shoes,
considered as representative of their quality ?
The problems of measuring a heterogeneous output have
always received considerable attention; such measures, when used
in the computation of labour productivity, have special aspects :
So long as they concern only one specific product at a time, statements regarding productivity remain relatively clear and unambiguous.
No confusion of meaning is possible when it is said that the production
of one thousand five-cent long-filler cigars by hand requires about
33 man-hours, or that the cement industry expended about four tenths
of a man-hour to produce a barrel of cement just before the war.
Difficulties begin to arise when statements about productivity cover
more than one product. Most industries typically produce a variety
of separate products. However, for most industries there is only one
measure of labour input, a total for the industry. To measure productivity in such an industry, some composite measure of total production is obviously needed. The problem is to select a measure which is
satisfactory for the purposes of computing productivity.1
The last point should be stressed: the measurement of heterogeneous products has already been studied many times, for instance
when making production indices; the solution adopted is the use
of certain weights. But the point to be examined here is whether
the weights which are used in ordinary measurements of heterogeneous products, for the measurement of production per se, are
suitable for labour productivity computations, or whether special
weights should be used for the latter; if this should be the case,
then the meaning of " productivity " data computed with the
use of the ordinary production measures or indices should be
studied.
The questions involved in the determination of methods of
measurement for the computation of total output composed of
heterogeneous products do not in essentials differ from those of
1

" The Meaning of Pro'ductivity Indexes ", loc. cit.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

the weighting of production indices calculated for the computation
of productivity data, nor from ihose of the weighting of direct
productivity indices. They will therefore be examined in the next
chapter in direct relation with the computation of productivity
data.

CHAPTER V
THE MEASUREMENT OF LABOUR PRODUCTIVITY

For the measurement of labour productivity, as for all economic
or statistical measurement, many formulae may be developed,
according to the question to which an answer is sought*
Even with ideal data, a number of plausible productivity indices
may be constructed. Our choice is reduced to but one of these many
alternatives only when the question asked about productivity changes
is so specific as to be 2equivalent to a verbalisation of only one of the
many possible indices.
In order to analyse some of the various possible formulae and
their meaning, a few general approaches will first be examined.
The degree to which the available data lends itself to the different
measures, the results obtained and their meaning when utilising
substitute data will then be reviewed.
SOME POSSIBLE APPROACHES

Four approaches to the measurement of the variation of the
productivity of labour are examined below; others could be
envisaged. For reasons of convenience, this analysis refers to
unit labour requirements and their variations rather than to
output per man-hour; the analysis is based on comparison between
periods, but could be applied without change to comparisons
between regions of the same country.
1. A common approach is to compare, for the two periods
under consideration, the total volume of labour required to produce
the same complex of goods, under the conditions prevailing in
each period as regards unit labour requirements.
It is obvious that in this case the results will depend on the
complex chosen. What complex should be chosen ? The choice
1
Though the argument of this chapter is stated in terms of the study of
indices, it applies also to absolute measures of labour productivity. See
Appendix
I for a mathematical approach to the notions developed.
2
" The Meaning of Productivity Indexes ", loc. cit.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

is related to the question to be answered ; in other words, the choice
of a certain complex leads to results answering certain specific
questions. Among the possible complexes of goods, two are of
special interest: that representing the production composite
of the base period (to which the current situation is compared),
and that representing the production composite of the current
period (which is to be compared to certain base standards).
If computations are made on the basis of the base period production composite, the unit labour requirements index will measure
the ratio of the labour that would have been spent in the current
period to produce the base period complex of goods to the total
labour actually expended in the base period. In the following
analysis, this ratio will be termed formula I.
If the computations are based on the current period production
composite, the unit labour requirements index will indicate the
ratio of the labour actually spent to produce the current complex
of goods to the labour that would have been spent in the base
period to produce the same complex. In the following analysis,
this ratio will be termed formula II.
2. A second approach is to consider that the variations of the
unit labour requirements for a certain production composite should
be an average of individual indices of unit labour requirements.
In this case, the index of each current unit labour requirement
as compared to the base period unit labour requirement is computed for each product, undertaking or industry; the results are
then grouped in an average, which may be an arithmetic mean or
a weighted average; the weights can be chosen to represent the
relative importance of each undertaking, product or industry in
the total production composite chosen, on the basis of total labour
expended, gross or net value of production, or any other factor.
Evidently, the results obtained in the computations will depend
upon the weights used. In particular, if the weights chosen
correspond to the amount of labour expended during the base
period for the production of each product, undertaking or industry,
the results will be the same as for formula I. A more complicated
system of weights 1 would lead to the results obtained with formula II.
1
These weights would be the labour that would have been expended to
produce the current period output of each product, undertaking or industry
with the individual unit labour requirements of the base period. See
Appendix I.

THE MEASUREMENT OF LABOUR PRODUCTIVITY

3. A third approach is to compare the average unit labour
requirements needed in the current period in order to produce
the current period complex of goods with the average unit labour
requirements needed in the base period to produce the base period
complex, the averages for each period being computed as ratios
of the total labour expended to the total production. This will
involve a measure of total production for heterogeneous outputs,
and thus the use of weights for the computation of these total
productions.
The weights used to measure the current and the base period
production should be identical; if this were not the case, even
when both the individual unit labour requirements and the output
of the various goods remained unchanged, the unit labour requirements index would not be equal to 1.
This approach yields results identical with those obtained in
the first approach, with either formula I or formula II, when the
weights chosen are either the individual unit labour requirements
of the base period or those of the current period (see Appendix I).
4. Finally, a very common approach to labour productivity
or unit labour requirement measures is to relate the variations
of the total labour expended in each period to the variation in
the total output of each period; or, in other words, to compute
the ratio of an index of labour to an index of production.
Notwithstanding the apparent differences, this approach does
not differ from the preceding one. Thus, the results will again
depend upon the weights chosen; when they are the individual
unit labour requirements of one of the periods compared, the
results are the same as with formula I or formula II.
THE TWO GENERAL CONCEPTS OF LABOUR PRODUCTIVITY

The preceding analysis shows that measurements of labour
productivity may be extremely diverse. Different results will be
obtained for the same conditions, according to the approach used.
But the types of formula that may be conceived can be sorted
into two categories, corresponding to the two different basic
concepts mentioned in Chapter I.
In the first concept, labour productivity is considered as a
specific characteristic of the product, undertaking or industry for
which it is measured; in this case, averages for a group of components should remain unchanged if the individual productivity for
each component remains unchanged.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

The two first approaches envisagea correspond to this concept;
if the unit labour requirements of the current period are the same
as those of the base period, the resulting measure will indicate no
change in the average unit labour requirements for the group of
components studied.
In the second concept, labour productivity is considered in
relation to the relative importance in the group of the product,
undertaking or industry for which it is computed ; in this concept,
averages for a group of components need not remain unchanged,
even if the individual productivity for each component remains
unchanged; if some of the products, undertakings or industries
forming part of the group studied have increased or diminished
their importance within the group, although each unit labour
requirement has not changed, the index of unit labour requirements for the group will not be equal to 1.
The two last approaches examined correspond to this concept.
In these approaches, it is only when the relative importance of
each component is measured in terms of unit labour requirements
that the unit labour requirement index for the group will indicate
no change when each individual unit labour requirement has
remained the same but the relative importance of each component
has varied.
I t will be shown in the course of this chapter how computations intended to be made on the basis of the first concept (and
more particularly on. the basis of formula I or formula II) are
shifted to the second concept, owing to lack of the necessary data,
by the use of substitute data.
The theoretical but simple example given in table VI illustrates
the distinctions between the two concepts. It is assumed that the
two plants A and B are manufacturing identical products.
In spite of productivity increases in both plants, the over-all productivity declined. In one sense, this decline in productivity is real,
inasmuch as in the second year a larger proportion of the total product
was produced at a lower level of productivity than in the first year.
In another sense, however, the decline in the average is misleading,
inasmuch as it fails to indicate that productivity increased in both
plants—perhaps as a result of improvements in production techniques. 1
If the computations of unit labour requirements for the
two plants together had been based on an average weighted with
unit labour requirements—that is according to formula I or
Unemployment and Increasing Productivity, op. cit.

THE MEASUREMENT OF LABOUR PRODUCTIVITY

formula II—instead of on the ratio of total labour expended as
compared to the ratio of total production (computed with weights
equal to 1), the results would have been quite different: formula I

TABLE VI.

INFLUENCE OF WEIGHTING ON PRODUCTIVITY
MEASUREMENTS
Year

Plant

Indices
First

Second

2,000
1,000

2,100
1,000

105
100

2.00
0.50

2.10
0.48

105.0
96.0

Employment

2,000
2,000

3,100
3,000

155
150

Labour productivity . .
Unit labour requirements

1.00
1.00

1.03
0.97

103.2
97.0

4,000
3,000

5,200
4,000

130

1.33
0.75

1.30
0.77

97.7
102.6

Plant A
Production
Employment
Labour productivity . .
Unit labour requirements
Plant B

Total, both plants
Production
Employment
Labour productivity . .
Unit labour requirements

133

Source : Unemployment and Increasing Productivity, op. cit.

would yield a unit labour requirement index of 96.3; formula II,
an index of 96.4.
This example illustrates clearly the difference between the
two general concepts of labour productivity. If it is intended to
consider the latter as a characteristic specific of each product,
undertaking or industry, then the computation of the labour
productivity of a group of undertakings should be made so that the
shifts in the relative importance of each component in the total
output will not influence the results of the computation ; if, on the
other hand, it is considered that such shifts are part and parcel of
the variations in labour productivity, the measurements should be
made to allow for the influence of such variations, as in this
example.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

The Differences Between Formulae I and II
The first concept—the measurement of variations in labour
productivity considered as a specific item—has been widely used
by those who have dealt in detail with the measurement of labour
productivity in a group of industries. In particular, the Works
Progress Administration studies for the period 1919 to 1936 were
entirely based on this concept.
But even so, a choice must be made between the remaining
possible formulae, and more specially between formula I and
formula II.
Each of these formulae has specific advantages:
The use of a base-year composite has greater significance under
relatively stable conditions of production and is usually more appropriate in dealing with relatively short periods of time ; it has the advantage of permitting comparisons between any two years, since the changes
in the required volume of labour in each year are measured relatively
to the same base-year composite.
The use of a changing composite does not theoretically permit
year-to-year comparisons and can only be used to measure the change
between the base year and any other year. Such an index has analytical value particularly in periods of major shifts in the composition
of production, and is usually more suitable for studying relatively long
periods.1

Table VII, which is taken from the Works Progress Administration report, illustrates the different results obtained by using
formula I (A = indices with base-year man-hour weights) and
formula II (B = indices with changing man-hour weights).
Use of Substitute

Weights

The computation of data on the basis of formula I or II
involves two different sets of measurements. In each case, the
computation requires the measurement of—
(a) the total labour expended to produce the complex of one
of the periods with the unit labour requirements of t h a t period ;
(b) the total labour required to produce the complex of one
of the periods with the unit labour requirements of the other
period.
The first set of measurements is relatively easy to obtain, since
it merely represents the measurement of the total labour actually
expended in one of the periods; for the second set, however, it is
1
Production, Employment
tries, op. cit., p. 13.

and Productivity

in 59 Manufacturing

Indus-

THE MEASUREMENT OF LABOUR PRODUCTIVITY
TABLE VII.

COMBINED INDICES ( 1 9 2 9 =

1 0 0 ) OF PRODUCTION,

EMPLOYMENT, MAN-HOURS, AND PRODUCTIVITY FOR MANUFACTURING

INDUSTRIES, 1919-1936

Production
Year

Employment

M anil ours

Output
Per wage earner

Per man-hour

B
A

1919
1920
1921
1922
1923

98.4
100.5
78.8
91.7
100.7

99.9
100.5
75.9
91.5
100.7

63.4
67.3
54.3
70.4
81.7

69.0
72.4
58.6
73.3
83.4

64.4
67.0
68.9
76.8
81.1

70.1
72.0
74.4
79.9
82.8

63.5
67.0
71.5
76.9
81.1

69.1
72.0
77.2
80.1
82.8

1924
1925
1926
1927
1928

94.8
98.2
98.8
95.6
96.3

92.1
97.5
98.7
95.7
96.0

77.4
86.1
90.2
88.5
93.0

78.8
87.1
90.7
89.2
93.3

81.6
87.7
91.3
92.6
96.6

83.1
88.7
91.8
93.3
96.9

84.0
88.3
91.4
92.5
96.9

85.6
89.3
91.9
93.2
97.2

1929
1930
1931
1932

100.0
86.3
73.2
64.4

100.0
80.1
65.2
51.8

100.0
80.8
68.0
53.4

100.0
81.0
69.3
55.8

100.0
93.6
92.9
82.9

100.0
93.9
94.7
86.6

100.0
100.9
104.3
103.1

100.0
101.1
106.3
107.7

1933
1934
1935
1936

72.6
84.0
86.9
90.9

57.7
60.5
65.7
73.3

62.1
67.1
77.9
89.3

65.5
68.7
80.4
90.8

85.5
79.9
89.6
98.2

90.2
81.8
92.5
99.9

107.6
110.9
118.6
121.8

113.5
113.6
122.4
123.9

i For many industries data are lacking for some of the years of the period 1919-1936
The index numbers were therefore constructed by chaining links for identical industries.

necessary to have the individual unit labour requirements of either
the base or the current period, according to the formula used.
Frequently, estimates of such data are not available, even for a
few products:
In general, although fairly detailed production data can be obtained,
the available employment statistics refer to workers engaged in the
manufacture of aggregates of more or less related products. Consequently, the productivity index can in almost every instance be
constructed only as the quotient of a production index for a group of
products and the corresponding labour index.1
Indeed, formulae I and II can be modified so as to represent the
ratio of an index of total labour to an index of production, the
1
Production, Employment
tries, op. cit., P a r t I, p . 20.

and Productivity

in 59 Manufacturing

Indus

METHODS OF LABOUR PRODUCTIVITY STATISTICS

production of the current and the base periods being weighted with
the individual unit labour requirements of either the base period
or the current period. They will then be comparable with the
formulae corresponding to the third and fourth approaches, with
special weights, that is, unit labour requirements.
But the lack of data on unit labour requirements is such,
especially when it is desired to compute labour productivity figures
for long periods in the past and for an important number of industries, that production indices with such weights can seldom be
computed. Substitute weights are therefore used.
What are the possible substitute weights, and what are their
drawbacks ?
Had a choice been afforded by the available statistics, the substitute weights would have been preferred in the following order, which
probably represents the relative likelihood of their proportionality to
unit labour requirements:
(1) labour cost per unit;
(2) value added per unit;
(3) value per unit; and
(4) price.
Labour cost per unit seems to be a satisfactory substitute, since it
is equivalent to labour time per unit translated into value terms. Its
use implies that the average hourly earnings for all the workers engaged
in the manufacture of any product of an industry are the same as for
all the workers making any other product of that industry. Value
added per unit is somewhat further removed from unit labour requirements. It has, however, a fairly high labour-cost content and also
does not include raw-material costs. It is particularly appropriate for
(a) industries with establishments of varying degrees of integration
reporting their output in full (for sale, interplant transfer, and consumption on the premises), and (b) industries making products whose
raw-material costs differ considerably. With the exception of prices,
unit values are farthest removed from unit labour requirements. In
addition to labour cost, unit value includes cost of raw materials, salaries,
profits, rents, interest, depreciation charges, insurance, etc. Price
includes not only these factors but also transportation and handling
costs, discounts, etc.
Since value weights were used predominantly, an attempt was made
to test their proportionality to the preferred labour weights. Usually
the comparison could be made for only a few products classified otherwise than in the National Research Project index; frequently the comparative weights depended upon crude estimates which detract from
the conclusiveness of the test. In all, comparisons of different weighting systems or indexes embodying them were made for 11 industries.
In several instances, the value weights were found to be fairly proportional to the preferred weights; even where they were not, the indexes
based on the two sets of weights were similar. The reason may be
that any weighting system would have little influence in the index for
an industry where one product dominates, where the several products
have the same movement, where the dispersion of the weights is small,

THE MEASUREMENT OF LABOUR PRODUCTIVITY

or where the number of products is large. It appears from experiments
made in the course of the derivation of the National Research Project
measures, furthermore, that weighted indexes of the same scope but
dissimilar degrees of detail frequently have the same general movement and that unweighted indexes, which are tantamount to relatives
of but one product, frequently exhibit the same general movement as
weighted measures. These considerations lead us to believe that the
substitution of unit-value for unit-labour requirement weights results
in satisfactory production measures for most of the industries. It
should not be inferred that the problems of weighting and classification
may therefore be neglected. On the contrary, the year-to-year differ:
enees in magnitude and direction may be sufficient to preclude the use
of indexes based on less detailed classifications for the measurement
of short-term changes in production "and productivity. 1
Another analysis of possible substitute weights was made by
D. Evans:
Those who construct production indexes to be used in measuring
productivity generally substitute other weighting factors more or less
related to unit labour requirements. Unit labour costs are regarded
as a first approximation, since they will be proportional to unit labour
requirements if average hourly earnings in the production of the several
goods considered are the same. These weights are also seldom available,
so value added by manufacture per unit of product, if it is available,
may be used as a second approximation. In this case, the additional
assumption is made that wages constitute the same proportion of value
added by manufacture for each of the products. Unit value is the
next step in the line of approximations; the validity of its use depends,
in addition to all prior assumptions, on the constancy from product to
product of the ratio between value and value added. Unit value is
probably the most commonly used weighting factor since it is frequently
available for the more important products of an industry from the
census of manufactures. When unit values are lacking, recourse may
be had to wholesale prices as weights. These, of course, will be
influenced by the additional factor of distribution costs.
It thus comes about that production indexes constructed for measuring productivity changes are frequently like indexes constructed for
measuring production per se; the latter often are of the aggregative
form and also embody pecuniary weights, though by choice rather
than necessity.
The Works Progress Administration (National Research Project)
undertook, in the course of constructing its indexes, to test approximative methods wherever possible. The opportunities were few and the
tests by no means definitive. The results indicated that approximative
indexes were adequate over long periods but not so trustworthy for
year-to-year changes. In extreme instances, the unit labour requirements may tend to be inversely related to the substitute weights. In
such cases, the derived productivity index may even lie outside the
range of the productivity relatives for the individual products. No
" directly " computed productivity index would give such a result, for
1
Production, Employment
and Productivity
tries, op. cit., P a r t I, p p . 38-40.

in 59 Manufacturing

Indus-

METHODS OF LABOUR PRODUCTIVITY STATISTICS

such an index is always equivalent to an average of the productivity
relatives for the individual products. 1
An example of the differences in unit labour requirements
indices derived from production indices constructed with value and
man-hour weights was given in the Works Progress Administration
report. 2 This example, which refers to the United States cigar and
cigarette industries for the period 1919-1936, is reproduced in
table VIII below.
TABLE VIII.

INFLUENCE OF VARIOUS WEIGHTS ON INDICES
OF UNIT LABOUR REQUIREMENTS

Index of unit labour requirements (1929 = 100)
Cigars and cigarettes combined

Year
Cigars

Cigarettes
Using value
weights

Using man-hour
weights

1919
1920
1921
1922
1923

122.9
110.0
125.9
123.7
121.7

323.3
348.2
259.2
227.4
182.4

222.0
233.1
214.6
200.5
176.4

150.8
143.2
144.3
138.1
130.1

1924
1925
1926
1927
1928

116.4
116.1
109.8
112.5
109.2

164.5
145.7
123.8
129.4
129.7

158.2
144.9
128.1
127.3
117.8

123.1
120.3
111.7
114.8
112.1

1929
1930
1931
1932

100.0
101.5
86.7
95.4

100.0
95.0
92.6
89.9

100.0
94.5
80.2
81.3

100.0
100.7
87.6
94.6

1933
1934
1935
1936

88.2
79.2
66.5
63.9

81.6
98.3
88.6
73.3

70.9
64.3
52.8
47.9

87.3
81.9
69.6
65.2

In this table the production index for cigars and cigarettes with
man-hour weights was constructed so that the measure of unit labour
requirements obtained by dividing this production index into relatives
of man-hours would indicate the volume of labour required to produce
the 1929 quantities of cigars and cigarettes in each year as compared
1
2

" The Meaning of Productivity Indexes ", loc. cit., p. 107.
Production, Employment and Productivity in 59 Manufacturing
tries, op. cit., p. 7.

Indus-

THE MEASUREMENT OF LABOUR PRODUCTIVITY

with the volume of labour actually utilised in 1929. The index with
value weights is of the same form.
The index of average unit labour requirements for the two industries
combined derived from a production index with value weights is higher
in 1926 than the unit labour requirements index for either industry;
after 1929 the index of average unit labour requirements for both
industries is lower than that for either. The index of production with
value weights results in a. measure of unit labour requirements which
understates the amount of labour required after 1929 to produce the
1929 volume of cigars and cigarettes. On the other hand, the unit
labour requirements index obtained from the production index with
man-hour weights measures the percentage of the labour time utilised
in 1929 that would have been required in any year to produce the 1929
output. 1
It is clear from these various statements and from the above
example that the substitute weights are not proportional, in most
cases, to the unit labour requirement weights. The utilisation of
substitute weights corresponds in fact to a shift from the first to
the second concept of labour.productivity, as indicated above;
when such substitute weights are used, formulae I and II are
transformed into those corresponding to the third and fourth
approaches.

Influence of Changes in the Composition of the Production
It can easily be shown that the substitution of weights such
as value added per unit, price, etc., for unit labour requirement
weights {i.e., the use of formulae corresponding to the second
concept) introduces in the indices an additional factor related
directly to the change in the composition of the production (see
Appendix I).
In fact, an author already quoted has computed the separate
effects of the combined variations due merely to changes in the
unit labour requirements of each industry over a period of time
and of the changes which occurred in the composition of the
production on a somewhat similar basis; the results of these
computations were summarised as follows:
Alterations in the industrial composition of output did have substantial influence on some of the group employment-output ratios, in
several instances outweighing the effect of changes in the employmentoutput ratios of the component industries. For the tobacco products
group the findings are particularly striking: in three periods half or
more of the decline in wage earners per unit was accounted for by the
1

Idem, pp. 6-7.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

rapid rise of cigarette manufacture at the expense of the industries
producing cigars and smoking and chewing tobacco. The cigarette
industry employs relatively few workers per $10,000 of products: the
industries making cigars and other tobacco products, which require the
labour of many more workers for the same dollar amount of output,
failed to raise their levels of production from 1899 to 1937. And in
the case of another industry, transportation equipment, virtually the
entire decline of 12 per cent, in wage earners per unit from 1899 to
1909 is attributable to the increased importance of automobiles, rather
than to any reduction in the employment-output ratios of the individual
industries making up the group.
The average change in labour per unit in the component industries
was uniformly of the same sign as the change in the ratio for the group.
This was not true of those changes in a group's unit labour consumption
that were caused by shifts in the composition of group output. These
shifts sometimes acted as a counterforce, making the decline in the
group ratio less steep than it would have been if the composition of
output had remained constant. In the food group, for example, workers per unit dropped 22 per cent, from 1899 to 1937. During this
period the constituent food industries actually cut labour per unit by
as much as 47 per cent, on the average, but at the same time several
of those with large unit labour requirements increased their shares in
the group's output, causing a rise of 26 per cent, in workers per unit
of output for the group as a whole. The 22 per cent, decline, then, is
the net result (rounded off) of a drop of 47 per cent, and an increase
of 26 per cent. A similar situation is found in the transportation
equipment group for the same period. During the 38 years the average
decline in employment per unit for the group was 63 per cent., a
reflection not only of the average decline in the employment-output
ratios of the various transportation equipment industries, but also of
the development of the motor car at the expense of other means of
transportation. The decline of 63 per cent, in workers per unit of
transport equipment, though large, ranks the group only in third place
in this respect; it follows tobacco and petroleum products, in which
the corresponding changes in employment per unit of product between
1899 and 1937 were 81 and 74 per cent, respectively. If the effect of
the shift in the relative importance of motor cars were excluded, however,
the average decline in employment per unit in the transport equipment
group would be 94 per cent., much larger than the corresponding cut
for any other group. 1
The results obtained (summarised in table IX) are computed
with a formula t h a t may be considered as representative of the
second concept of labour productivity; this table, when compared
with table V I I 2 which is based on the first concept, will show the
differences between the results, even for the large group of manufacturing industries.
The differences between these results and those published by
the Works Progress Administration are particularly striking for
the years 1919 to 1921, 1923 to 1925, and 1927 (for output per
1

S. FABRICANT: op. cit., pp. 40-41.

See p. 61.

THE MEASUREMENT OF LABOUR PRODUCTIVITY

TABLE IX.
PRODUCTIVITY OF LABOUR IN MANUFACTURING INDUSTRIES
Output
Year

1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939

Per wage
earner

Per manhour

61
65
69
80
79
80
86
90
93
96
100
98
98
86
90
85
96
104
100
94
109

57
63
69
76
75
79
84
88
89
96
100
102
105
102
108
111
120
120
117
120
131

Source: S. FABRICANT, op. cit.

man-hour) and 1919, 1920, 1922, 1930, 1931, 1934 and 1936 (for
output per man).
These illustrations should be sufficient to show that the two
general concepts of labour productivity are radically different,
though neither is essentially better than the other. It is therefore
important to choose between the two, and to state clearly which
has been used in any computations that may be published.
The theoretical advantage of an index computed on a fixed weight
base lies in the fact that it measures but a single element of change
to the exclusion of others. Yet the use of such a formula does not
prevent the infusion of many factors, each of which changes in time,
and several of which may be deemed to possess peculiar secular patterns
of their own. So long as such influences are not eliminated, and the
index reflects more than one element of the dynamic changes which
take place in the economy, one can properly raise a question as to
whether an index number, which would incorporate in its make-up the
effect of all of the dynamic changes which affect the economy (i.e.,
include those which are responsible for the growth in the production
of some goods and the decline in others) would be of greater significance
for economic analysis. Such an index would reflect changes in the
relative importance of the different products in the course of time, and

METHODS OF LABOUR PRODUCTIVITY STATISTICS

would permit the inclusion of new products as they make their appearance. An index number so constructed then could be described in
terms of its significance as the index of manpower utilisation (measured
in units of time input) needed at different times to produce a unit of
physical output in accordance with specifications, efficiency and nature
of economic organisation at each of the periods under consideration.1
Thus, labour productivity data obtained by the division of
production figures (or indices) measured per se (that is, weighted
by net value, price, etc.) by employment or labour figures (or
indices) corresponds to the second (and larger) concept of productivity of labour, in which all factors of variation have their
influence, including those due to changes in the composition of
production. In particular, when the total output of the nation
is compared to total employment or total man-hours expended, it
is this, last concept t h a t is implicitly or explicitly considered, since
gross national product, for instance, is a production complex
weighted by value added.
COMPARISON OF PRODUCTION INDICES WITH LABOUR INDICES

Once it has been clearly shown to what concept the ratio of
production indices to labour indices corresponds, some of the main
drawbacks of this method of computing data concerning the
productivity of labour may be examined.
Limitations of available data or oí the possibilities of collecting
data are such that production can seldom be measured in all
branches of activity; this is also true of the measurement of labour.
In addition, the scope of the possible production and labour
indices often differs.
Labour productivity indices are thus for the most part ratios
of production indices computed for a limited number of industries
(considered as measuring the variation of production in all the
industries concerned) to labour indices computed for another
limited group of industries (considered as measuring the variations
of labour expended in all the industries concerned).
If the limitations of data are too great (that is, if the number
of industries for which data cannot be collected is too large), it
might be preferable not to try to combine the individual indices
and not to issue a " general " index:
A proper question can be raised regarding the wisdom of combining
data for a number of selected industries into one index of productivity,
1
Limitations of the Existing Productivity Measures and the Need for New
Studies, op. cit.

THE MEASUREMENT OF LABOUR PRODUCTIVITY

if the makers of such a combined index do not believe that the combination has any independent significance. When this is done . . ., it
invites misuse by those who, not having been cautioned, are prone to
interpret such a combined index as one representative of industry as
a whole.1
But production figures are sometimes not even available;
output is often measured by figures relating to a basic material
t h a t is found to vary, more or less, as the product does. In such
cases—bread and bakery products and newspaper and periodical
publishing are examples—the index of production should be
regarded only as a rough approximation.
In the absence of a directly constructed index of physical
production, a production index may be derived by deflating a value
series, such as value of shipments, by an appropriate price series.
Such indirect procedures, while subject to several important
sources of error, are nevertheless capable of indicating gross changes
in productivity. Detection of changes of the order of 2 or 3 per
cent., such as occur annually in many industries, is rarely possible
by using such procedures, but changes of the dimensions of those
t h a t occurred between 1930 and 1940 can be detected.
Sometimes the method of computation of the production indices
makes it impossible to use them in the computation of productivity
data:
A striking feature of the Federal Reserve indexes since the 1940
revision is the inclusion of adjusted man-hours to represent current
activity in industries having a long production cycle, making numerous
or heterogeneous goods, or manufacturing products which differ from
their peacetime output. After the December 1941 revision, man-hours
series accounted for 38 per cent., and after the October 1943 revision
41 per cent., of the base period aggregate for private manufacturing.
Such series account for even greater percentages of aggregates for more
recent periods
and of aggregates which include Government manufacturing.2
Since the implicit assumption in the use of these data as production indices is that labour productivity remains unchanged,
they cannot be used in the computation of indices of labour
productivity.
Even when production figures or indices are not computed in
such a way as to make it impossible to use them, a number of
divergencies between production indices and employment or man1
Limitations of the Existing Productivity Measures and the Need for New
Studies, op. cit.
2
" The Concept of Productive Activity ", loc. cit., pp. 227-228.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

hours indices computed for the purpose of measuring output,
employment or man-hours render comparability between these
two sets of data very doubtful.
Differences in the Classification of Production and Labour Data
Among the difficulties encountered in comparing production
figures with labour figures, the allocation of undertakings to
different groups in production and labour statistics should be
noted.
It often happens that in one country the classification utilised
for production statistics differs from the one used in regrouping
labour figures. For instance, if in production statistics the classification is based on the criteria of the goods produced, and on the
other hand the classification used in the collection of labour statistics is based on the main raw material used, the data on production and on labour will not be comparable in scope; thus, in such
cases, all undertakings making toys will be grouped in the toy
industry, and all undertakings making musical instruments in
musical instruments industry, while in the labour classification the
undertakings making toys or musical instruments out of wood will
be separated from those producing similar goods out of metal.
It is true that such difficulties will not be encountered in the
data obtained from censuses of manufactures, where production
and labour figures are collected simultaneously. Proper comparability, however, is difficult to achieve even in this case, because
of the form of the available production statistics and the fact that
in the censuses the establishment is used as the basis for industry
definition. The employment reported for an industry usually
includes workers engaged in the manufacture of products not
normally belonging to that industry; on the other hand, production statistics often include the output of primary products of one
industry contributed by workers in other industries.
In some instances it was possible to increase the degree of correspondence between the production and employment statistics by redefining a census industry so that it refers not to establishments but to
the total output of the products normally belonging to it. In the
instances where no adjustment could be made, it is very likely that a
negligible net distortion results from the inclusion in the production
index of primary output contributed by other industries and the
inclusion in the employment index of the wage earners engaged in the
manufacture
of products not normally belonging to the industry in
question.1
1

Production, Employment
tries, op. cit., p. 25.

and Productivity

in 59 Manufacturing

Indus-

THE MEASUREMENT OF LABOUR PRODUCTIVITY

In other words, in the classification of production and labour
data taken from the censuses of manufactures, and still more in
data collected separately (by different agencies, for instance) there
are basic differences of allocation. When an undertaking produces
entirely different products, the output of each may be allocated
to each corresponding industry: if an automobile factory also produces wheelbarrows, these two outputs will be allocated, most
commonly, to different groups in the industrial classification; b u t
this separation is not feasible for labour data, and employment or
man-hours expended in an establishment are practically always, in
labour statistics, allocated to one industry, in general that of the
main product of the establishment.
Differences in the Size of the Undertakings on which Production
and Labour Statistics are Based
When data on production and labour are collected separately,
there are often serious differences in the scope of the figures. The
size of the undertakings from which data are collected may not
be the same; thus, production indices may be based on the reports
of establishments employing, say, more than 20 persons, and labour
indices on the reports of establishments employing more than
10 persons. If, as will usually be the case, the productivity of
labour of establishments employing between 10 and 20 persons
varies differently from that of larger establishments, the productivity indices will be biased. Such considerations are still more important when the computation of absolute productivity figures (output
per man-hour or unit labour requirements) is contemplated. The
lack of comparability of production and employment data is even
greater if the latter are obtained through labour force sample surveys or complete counts based on social security data, for instance,
since the data include a number of persons (self-employed workers,
persons employed in very small shops, etc.) for whom production
figures may not in practice be collected. Obviously, the biases
are reduced when data concerning both production and labour
are obtained from censuses of manufactures; but the scope of the
figures might vary in the course of time and introduce other biases :
The scope of the census canvas was reduced slightly in 1921 by
the exclusion of establishments whose annual output was valued at less
than $5,000; the minimum in 1919 and earlier years was $500. Although
the comparability between the production and labour indexes is maintained, a slight upward bias may be imparted to the productivity
measures between 1919 and 1921, since the small establishments excluded
in the latter year were probably less efficient than the average. That

METHODS OF LABOUR PRODUCTIVITY

STATISTICS

the bias is negligible, however, is indicated by the fact that in 1919
establishments with products valued at $5,000 or more employed
99.5 per cent, of the reported number of wage earners in all manufacturing industries
and accounted for 99.8 per cent, of the total value
of products.1
Combination of Errors
However carefully they are computed, production and labour
indices do not represent the " true " trend of variations in production or labour; they are only approximations to the phenomenon they are meant to measure. The most important
drawback in the computation of productivity indices by the
division of production indices by labour indices is the fact
that in such a division these approximations or errors are
compounded. They may cancel each other out but they may
also be added to each other; most likely the effects will vary,
and the bias of the productivity index will be erratic and may
finally be greater than the bias of either the production or the
labour indices.
A more detailed analysis of the production and labour indices
would reveal other points of non-comparability. Therefore, it
appears that great caution should be taken when computing, and
especially when interpreting, labour productivity data obtained
by the comparison of production and labour figures prepared
primarily for the measure of production per se, or of employment
or man-hours.
Many of the deficiencies of such measurements disappear when
use is made of production and employment (or man-hour) indices
specially computed for the purpose of obtaining labour productivity
data. The Works Progress Administration presented such computations in the past, and the United States Bureau of Labor
Statistics uses this method for a number of industries. However,
these special computations are far from removing all points of
non-comparability; in addition, the data obtained are of a
general character, and yield no information concerning the factors
affecting labour productivity.
In view of these difficulties, it may seem advisable to turn to
special methods of measuring labour productivity, that is, special
enquiries into the productivity of labour: this is particularly to be
recommended if detailed international comparisons of labour
productivity are to be made, since, as will be shown in Chapter VII,
1
Production, Employment and Productivity in 59 Manufacturing
tries, op. cit., Part I, pp. 27-28.

Indus-

THE MEASUREMENT OF LABOUR PRODUCTIVITY

the differences in the various national practices add to the difficulties encountered in national comparisons.
Special enquiries into labour productivity will be studied in
Chapter VI. A brief reference will first be made, however, to the
measurement of labour productivity in non-manufacturing industries and for a nation as whole.
MEASURES OF LABOUR PRODUCTIVITY IN NON-MANUFACTURING
INDUSTRIES

Scattered attempts have been made to measure labour productivity in non-manufacturing industries. Here, however, the
special difficulties in the measurement of production mentioned
in Chapter I are met with, in addition to the difficulties encountered
in manufacturing industries. Some of the solutions adopted will
be briefly reviewed here.
In agriculture the United States Bureau of Agricultural Economics has developed two measurements of agricultural output
which are used to derive separate measurements of productivity:
farm output and gross farm production. Both these measurements
have been developed by the use of annual production indices for
the individual crops included.
Gross farm production measures the total product of United
States farm land and farm labour resources in each calendar year.
The index records changes in the volume of production in both
farm " producers' goods " and the output of products for human
use. Total gross production thus includes all crop production,
pasture grazed by all livestock, and the production of livestock
and livestock products. Livestock and livestock production
include only the " product added " by livestock in the conversion
of feed and pasture into livestock and livestock products for human
use and into farm-produced animal power. Farm-grown livestock
feed and pasture are included in total crop production and are thus
counted but once in arriving at total gross farm production.
The index of farm output measures total gross farm production
minus farm-produced power. Horse and mule feed and the
products of farm labour and other resources used in raising and
maintaining farm horses and mules are thus excluded.
For the telephone industry, the production index in the United
States is based on the weighted aggregate number of originating
local and toll calls. The weights represent average revenue per
local and per toll call in 1939.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

The production index for the telegraph industry is based on
series for: (1) domestic messages (land-line companies); (2) foreign
messages (land-line companies); (3) domestic messages (oceancable companies); and (4) foreign messages (ocean-cable companies).
Each series is weighted by the corresponding unit revenue in 1939.
The employment series for these two industries is based on data
adjusted to represent annual averages by the use of employment
figures :
The " production " indexes for the telephone and telegraph industries are obviously not measures of physical units produced, but rather
measures of service performed for the public expressed in terms of the
number of telephone calls made or the number of messages sent. The
man-hours factor for each industry is an average measure of labourtime spent by all wage earners who have in any way contributed to
the volume of service provided the public. The productivity measures,
therefore, express the relationship between the volume of public service
and labour-time expended by all wage earners, whether directly or
indirectly engaged in providing that service.1
In the construction industry, some attempts to measure productivity have been made in particular branches. Thus, in
concrete road building, productivity was measured by the average
man-hours required per cubic yard of dirt moved, and by the
average man-hours required per cubic yard of concrete road
constructed. 2
Because so little is known of changes in productivity in the construction industry it is only natural that attempts to use indirect procedures, at least potentially capable of detecting large changes, have
been frequent. In the case of the construction industry, such methods
are not merely subject to large errors; they are impossible. A minimum
requirement for such indirect methods would be the existence of three
independently derived statistical series, the dollar volume of construction, construction costs or prices and man-hours expended. Until
recently only one of these three, dollar volume of construction, has
been available. Estimates of employment in the construction industry
were derived from the dollar volume series by means of a series of key
assumptions, one of which involved productivity assumptions. Indirect
methods based on such data would merely reproduce the productivity
assumptions used in deriving the employment estimates. More recently,
estimates of construction employment have been derived from direct
employer reports, so that this particular problem no longer exists. The
series on construction costs or prices now available are still not independent, however. They are obtained, not on the basis of direct field
reports, but synthetically from data on building materials prices,
construction wage rates, and assumptions about productivity. Any
1

UNITED

STATES

DEPARTMENT

LABOR,

BUREAU

LABOR

STA-

TISTICS: Major Sources of Productivity Information,
W a s h i n g t o n , J u n e 1949,
p . 44.
2
Monthly Labor Review, Nov. 1932, p p . 1027-1029.

THE MEASUREMENT OF LABOUR PRODUCTIVITY

estimates derived from these data will merely reproduce the productivity
assumptions. Until further data on prices, physical volume or unit
labour requirements, are obtained by direct field enumeration, even
the roughest estimates of changes in construction productivity will
be impossible.1
Thus, the main difficulty encountered in most non-manufacturing industries lies in the acquisition of a correct indicator
of output, that is, a measure of what has actually been " produced ". The efforts made in this direction have not so far been
very successful, especially in the measurement of the output of
distribution or Government services. Detailed studies of this
subject would be well worth while.
LABOUR PRODUCTIVITY OF A NATION

•Numerous studies have been undertaken to measure and to
compare between various periods the " labour productivity of a
5 nation ".
Such studies are generally based on gross national product per
worker or man-hour, or on real national income per head.
The preparation of such data is particularly difficult, and its
interpretation involves many problems which lie outside the

scope of this report.
In particular, the reconciliation of the results of such studies
with those obtained through the direct measurement of labour
productivity implies a number of assumptions and estimates based
on various data. It should be recalled t h a t such measures all
refer to the second type of concept of productivity.
Perhaps the best way of looking at productivity is to regard it as
an indicator of the way the country is utilising its existing resources
—its natural wealth, capital equipment, and manpower. Increased
productivity means better utilisation of the country's productive
forces ; it implies achieving a higher volume of production from a given
volume of resources. The ability to sustain, let alone increase, the
present standard of living depends almost entirely on increased productivity. This is the only way of providing exports in exchange for
imports of foodstuffs and raw materials, of supplying industry with the
necessary capital equipment and of securing for the people a flow of
consumption goods as high if not higher than before the war.2
Gross national product designed to measure the total national
output is an involved and complex concept, in which the pro1
J. CORNFIELD: Productivity in the Construction Industry,
Productivity Panel Conference, Washington, Jan. 1948.
2
" Output a Head ", loc cit.

Fourth

METHODS OF LABOUR PRODUCTIVITY STATISTICS

duction of an industry or branch of the economy is represented
by the total value of its output, minus purchases from other
sectors:
In order to develop a measure approaching physical output for the
economy, the Department of Commerce of the United States is currently
engaged in computing gross national product in terms of 1939 prices,
rather than in terms of current-year dollars. In general, the computation of gross national product in " real " terms is accomplished by
deflating gross national product in current dollars for each minor
expenditure segment by appropriate price series. Among the difficulties
inherent in this approach are the lack of comparable price series for
some industries, the conceptual difficulties of attempting to measure
real output for such groups as the armed services, retail trade, etc.,
and problems involving the purchase of materials by one segment from
another. 1
The observation of the variations in gross national product per
worker or man-hour is of economic importance; however, these
variations should not be interpreted as due solely to variations in
productive efficiency:
Gross national product per worker or per man-hour is not a measure
of production efficiency, primarily because the measure is affected by
changing patterns of production from year to year. For example,
value added by manufacture per worker (which is numerically close to
gross value added per worker) ranged in 1939 from less than $1,000 in
various garment industries to more than $6,500 in the petroleum
refining and chemicals industries. Thus, a shift in the production
composite can substantially change the value of gross national product
per worker (both in terms of current dollars or 1939 dollars) for the
economy without any change in physical output per worker having
occurred in any industry. 2
In addition, the practical difficulties encountered in the measures
of gross national product or real national income are numerous.
I t may be noted briefly t h a t :
(a) the estimates for long periods are very inaccurate;
(b) the deflation of the price element is hindered by the
absence of accurate deflators;
(c) special difficulties are met in estimating the value of the
net product in service industries.
From the point of view of the data concerning labour to be
used in such measures, a number of problems of concept and of
1
2

Major Sources of Productivity Information, op. cit.
Recent Productivity Trends and their Implications, op. cit.

THE MEASUREMENT OF LABOUR PRODUCTIVITY

measurement also arise: gross national product or real national
income may be compared with total man-hours worked, total
employed persons, total labour force or total population, according to the purpose of the calculation. For instance, when the
increase of productivity of labour in a plant is due to a technical
improvement which results in the dismissal of a number of workers, it might be considered that the nation's productivity has
actually increased only if these dismissed workers have been reemployed ; otherwise their " productivity " is nil, and when combined with the increased productivity of the plant may show no
improvement in the nation's productivity. This would be an
argument in favour of considering the unemployed in the computation of a nation's productivity.1
In conclusion, it appears that the measure of the productivity
of a nation is based on highly intricate concepts of total production
(real national income, gross national product) and on general
concepts of labour (total labour force, total employed) which all
involve estimates and approximations in measures, and leave room
for a number of diverse interpretations; the results of such computations are subject to wide margins of variability. Such broad
calculations present great difficulties to any standardisation in the
near future.

1
For further discussion of this point, see Unemployment and Increasing
Productivity, op. cit.

CHAPTER VI
SPECIAL ENQUIRIES
PAST EXPERIENCE

Labour productivity measures based on data collected for the
measurement of production and employment per se present many
drawbacks, as was seen in Chapter V. Even when, as in some
industries, the available data can be corrected in order to furnish
comparable estimates of production and employment, the measures
of labour productivity so obtained are of a very general character,
throwing no light on many important points, such as, for example,
factors affecting labour productivity.
The United States Bureau of Labor Statistics, in its attempt
to compute labour productivity data for each of the manufacturing industries, with a view to obtaining eventually an index of
labour productivity for manufacturing as a whole, has encountered
various difficulties in the use of data collected through the census
of manufactures and the current employment surveys ; in order to
achieve its aims it has resorted to special enquiries.
Such enquiries had already been undertaken in the United
States as far back as 1898, when Carroll D. Wright, Commissioner
of Labor of the United States, included in his 13th annual report
a detailed study of labour productivity in all industries.1
Throughout the detailed tables of this report, the names of the
operations necessary in the production of a unit were shown, and
for each operation the following indications were given:
(1)
(2)
(3)
(4)
(5)
(6)
(7)
1

the
the
the
the
the
the
the

machine, implement or tool used;
power used;
number of persons necessary on one machine;
number and sex of employees occupied;
name of the occupation of each employee;
age of each employee occupied;
time expended by the employee;

Hand and Machine Labor, op. cit.

SPECIAL ENQUIRIES

(8) the rate of pay of the employee, per hour, day or week,
as may be the case;
(9) the labour cost.
This study was a comparison of productivity or unit labour
requirements according to two different methods of production,
the " hand " method (in which simple tools such as the hammer,
the chisel, the saw, etc., might be used) and the machine method.
It provided for a very detailed analysis, permitting interpretations which no over-all measure would yield.
Subsequently, and up to 1940, detailed studies for various
industries were conducted by the United States Bureau of Labor
Statistics, and the results were presented in a number of special
bulletins giving statistics and analyses of changes in output per
man-hour. These reports, developed from detailed data concerning industry, contained a wide variety of information concerning
industrial techniques and processes. Among the topics included
in a typical report were a history of the industry, a description
of production methods for each major product, employment and
wage figures, an analysis of technological change and its relation
to labour displacement, output per man-hour and labour cost for
machine and hand production methods. Figures were presented
showing output per man-hour (or labour time per unit of output)
for a number of products, classified according to process, geographic region, type of machine employed, etc.
The conclusions obtained through such detailed enquiries
proved to be remarkably interesting. As an illustration, extracts
of conclusions reached in the study of the cigar manufacturing
industry (conducted in 1935-1936) * are given in tables X and XI.
Unfortunately, while extremely useful and instructive, these
studies took a great deal of time, and only a limited number of
them were conducted. The reports, issued after long delays, were
of historical interest rather than of direct application; they did
not always contain sufficient information to permit interpretation
of the figures, and, above all, they were sporadic and covered
only a few industries.
After the second world war, the general interest in labour
productivity strongly increased, and since available statistics on
production and employment often proved to be insufficient, the
1
D. EVANS: Mechanisation and Productivity of Labor in the Cigar Manufacturing Industry, United States Bureau of Labor Statistics, Bulletin No. 660.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

United States Bureau of Labor Statistics undertook a new project
—" direct productivity reports ".
TABLE X.

APPROXIMATE AMOUNTS OF LABOUR REQUIRED TO

MAKE 1 , 0 0 0 FIVE-CENT CIGARS BY VARIOUS MANUFACTURING
METHODS (UNITED STATES,

1935-1936)

Number of man-hours
Long-filler

Short-filler

Operation
Handmade

Four-operator cigar
machine

Machinebunched
hand-rolled

Two-operator cigar
machine

1.41
1.64
25.19
1.96

1.41
1.64
9.55
1.96

1.41
1.64
19.64
1.96

1.41
1.64
5.58
1.96

0.36
0.66
2.16

0.36
0.16
0.88

0.36
0.66
2.16

0.36
0.16
0.88

Total, all above
operations . . .

33.38

15.96

27.83

11.99

Leaf preparation . . .
Stripping
Making
Packing
Cellophaning and
Miscellaneous labour

i Includes all labour not previously assigned departmentally (watchmen, elevator
operators, cleaners, ofllce help, shipping clerks, electricians, nurses, maids, locker attendants, etc.).

TABLE X I .

OVER-ALL MAN-HOUR PRODUCTIVITY IN PLANTS

MANUFACTURING FIVE-CENT CIGARS BY DIFFERENT METHODS

All plant labour
Method of manufacture

Four-operator machine-made, long-filler. . .
Machine-bunched, hand-rolled, short-filler .
Two-operator machine-made, short-filler . .

i Computed from table X.

Productivity,
cigars per
man-hour

Productivity
relative to
hand-made
long-filler
cigars

30.0
62.7
35.9
83.4

1.00
2.09
1.20
2.78

SPECIAL ENQUIRIES

DIRECT PRODUCTIVITY REPORTS IN THE UNITED STATES

The direct productivity reports are mainly designed to provide,
for a specific industry, detailed statistics and supplementary information regarding the levels and trends of labour expended per
unit of production; they are particularly useful in those industries
for which such statistics could not previously be prepared, and
where detailed data, not obtainable from published sources, are
deemed to be essential.
Preparation of the Survey
The first step is a preliminary study of the industry. Using
all available information (census of manufactures, industrial directories, telephone directories, technical books, publicity catalogues,
bibliographies, etc.), a definition of the industry in terms of manufactured products may be established. This definition is based
on practical considerations, and therefore does not always correspond to that used in the census of manufactures. As an example,
for the electrical appliances industry, on the advice of the trade
unions and the manufacturers, the survey is extended to include
refrigerators, washing machines, vacuum cleaners, etc., products
which are not covered by the definition of the industry used in
the census of manufactures.
The employers' associations and the trade unions are approached
in order to secure the co-operation of the industry. (The surveys
are conducted on an entirely voluntary basis.) These contacts
indicate whether the survey is technically possible; for instance,
the undertakings must keep adequate records; there must be a
sufficient proportion of the products manufactured which remain
comparable from one year to another; and, of course, the manufacturers must also be willing to co-operate. With the information collected on these points, and paying due attention to the
estimated cost of the survey, the Bureau of Labor Statistics is in
a position to make a decision as to the advisability of starting a
survey.
In July 1950, reports had already been issued or were about
to be issued for the following industries: construction machinery;
fertilisers; footwear; industrial equipment; leather; machine tools;
men's dress shirts; soap and glycerine; cane sugar refining; electrical machinery; household electrical appliances; home radio
receivers; metal forming equipment; mining machinery; railroad
freight cars; and luggage.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

The initial reports generally cover the period 1939 to 1947 or
1948, and are supplemented by reports furnishing yearly data for
1948 and 1949.
The actual preparation of a survey begins with the choice of
the products on which the survey will actually be based. This is
done after consultation with management officials, employers'
associations and other persons familiar with the industry. It is
neither necessary nor, most of the time, even possible to include
all products manufactured by the industry in the survey. Thus,
for the machine tool industry, by taking 16 products more t h a n
80 per cent, of the total value of production of t h a t industry is
accounted for.
Products chosen as a basis of reporting are those with specifications typifying the industry's output with respect to type, size,
complexity of design, method of fabrication and volume of production. Only relatively standardised items regularly produced
are selected, since man-hour data pertaining to goods made to
order generally represent a small segment of production and are
not comparable from year to year.
For example, the report on the radio industry stated:
Information on man-hours required per unit of production was
requested for 24 different types of radio receivers. These radios,
classified in six general categories, were selected to represent the output
of the industry in terms of value and units produced . . . Eight
kinds of table model receivers, for example, were chosen to represent
the range of table model sets produced by the industry . . . The
selection of the 24 types of receivers was made with the assistance of
management officials in the industry and the Radio Manufacturers'
Association. Detailed specifications were drawn up for each of these
items with respect to product characteristics such as type of power
supply, number of tubes, and tuning bands. These specifications were
sufficiently broad that the products of several manufacturers could be
combined for the purpose of making a product index, yet narrow enough
that man-hours would be reasonably comparable for the products of
different establishments reported within each specification.1
Thus, the products included in the survey are not all those
manufactured by the industry b u t only those t h a t are considered
representative, and t h a t remain sufficiently comparable over a
period of time. For instance, in the above-mentioned survey
" the coverage excludes some products (such as radiotherapy
machines, industrial electronic equipment, and short-wave receivers
for amateur, police, and military use) because of their specialised
1

UNITED

STATES

DEPARTMENT

TICS: Home Radio Receivers

OP L A B O R , B U R E A U

(1939 to 1947),

OF L A B O R

May 1950, p . 4.

STATIS-

SPECIAL ENQUIRIES

nature and the lack of comparability for purposes of trend
analyses ".
The list of establishments manufacturing the specified products
retained is then drawn up. When the industry is very concentrated,
an attempt is made to secure data from all the establishments
concerned. In most cases, however, only a sample is used. When
the industry comprises a large number of establishments, the sample
will contain, in general, all the large establishments and a selection
of the smaller ones, the latter selection being made on the basis
of scientific sampling. In practice, the coverage of the actual
sample of establishments reporting differs from the initial list
because there is always a certain proportion of manufacturers
unable or unwilling to co-operate. When possible, establishments
retained in the sample but not co-operating are replaced by establishments similar in size, method of production, etc. However,
as the survey becomes established, the number of refusals decrease
and more manufacturers find it possible and useful to maintain
records which are suitable for reporting.
Collection of Data
A questionnaire is then prepared, intended for the manufacturer to fill in. The questionnaires vary in content from one
survey to another, but they are drafted according to a general
pattern. They are sometimes given a trial period with a few firms
who are willing to co-operate, and then modified according to
the suggestions considered useful. In addition to identification of
the company and the plant, they usually ask for:
(a) specifications of products to be reported;
(b) output for each year or for representative production
periods ;
(c) total direct man-hours used in the manufacture of the
output ;
(d) total indirect man-hours used in the manufacture of the
output ;
(e) details concerning the methods used to compute the data
furnished.
Additional questions concerning plant capacity, prices, distribution of production according to production methods,
mechanisation, etc., may occasionally be included.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

These questionnaires are supplemented with detailed instructions to the manufacturer concerning the methods to be used in
order to produce the data, in particular with respect to direct and
indirect man-hours.
Thus, the instructions on the questionnaire used in preparing
the report on the production of home radio receivers indicate that
" the classification of direct man-hours should include all factory
man-hours involved in actual production of the models reported
and included in the company's accounting procedure as direct
labour cost " and that " the classification of indirect man-hours
should include all man-hours in general factory and departmental
overhead costs applied in the company's accounting procedure to
the output of the model being reported. Exclude man-hours
in general administration, office, engineering, and sales." Time
devoted to sales, product development research and administration are generally excluded in order to eliminate areas of activities
where considerable variation prevails among establishments in the
same industry. Hours paid for but not worked (such as paid
vacations, sick leave, call-in time, etc.) are also generally excluded.
Direct man-hours are to be obtained on the basis of one of the
three following methods:
(1) Computations from records such as time tickets (this is the
method preferred).
(2) Conversion of direct labour cost data (available in dollars)
to an estimate of direct man-hours by dividing the cost data by
average hourly earning figures computed for all groups of direct
workers employed in the production of the item.
The establishments generally have available data on average
hourly earnings, correctly weighted. In some questionnaires,
information on the correct method of computation of these average
hourly earnings is added to the instructions.
(3) Estimates from standard cost accounting. Many firms
establish yearly a standard budget, based on previous records,
according to which they expect a certain number of working hours
(the standard) to be expended per unit of production during the
coming year. (The computations may also be on the basis of a
standard cost of labour per unit of product.) In addition, the firm
also computes periodically the number of hours actually spent
(or the actual labour cost) in a production period for the production
of a group of products, or for a department.

SPECIAL ENQUIRIES

It is then possible to estimate the number of direct man-hours
actually spent in the production of a given product by adjusting
the standard hours for this product by the ratio of direct manhours actually expended (or the actual labour cost) for the production of a related group of products or of a department to the
standard direct man-hours (or the corresponding standard labour
cost) set out for that group of products or that department.
Indirect man-hours may be estimated by the reporting establishment on the basis of any one of the methods described for
obtaining direct man-hours; if, as is usually the case, data on
indirect man-hours are available only for a group of products or
for a department, estimates of the indirect man-hours expended
on the product under consideration can be obtained by applying
to the aggregate indirect man-hours the ratio of direct man-hours
expended on the product to the direct man-hours expended on
the aggregate production. Alternatively, estimates may be based
on labour cost ratios.
Field Agents
An important feature of these direct enquiries into labour
productivity is that the questionnaires are not sent by mail, but
are presented to the firm by a field agent.
The role of the field agent is essential. He visits the manufacturers at the beginning in order to secure their co-operation,
and for this purpose he generally addresses himself to the president
of the company. In a large measure the success or failure of a
survey will depend on the ability of the field agent to convince
the management of the usefulness of the enquiry. He will base his
arguments, for example, on the necessity for the Government to
know what is happening to productivity in general; or on the
advantages for a co-operating undertaking to be able to compare
its performance with the average performance of the other undertakings in the industry.
Having secured co-operation, the field agent will present the
questionnaire to the firm, and explain in detail the requirements
and the methods of obtaining the information asked for; he will
assist in obtaining the data from the records if requested by the
manufacturer.
He will also fill in a personal preliminary report according to
a schedule drawn up by the central office. This report contains
additional information on a great number of points intended to.
help in the interpretation of the data and the determination of its

METHODS OF LABOUR PRODUCTIVITY STATISTICS

reliability; it includes plant identification, description of the product, and information concerning the types of records kept by
the plant, the posts classified as direct labour and indirect labour,
the coverage of man-hour data in relation to the various kinds of
hours paid for but not worked, the degree of integration of the
plant (purchased parts, contract work), etc.
The questionnaire (of which three copies are established initially
for each product manufactured by the establishment, one to be
kept by the manufacturer, and the others to be sent directly to
the central office) and the field agent's preliminary report are
analysed in detail by the central office. One copy of the questionnaire is then returned with the comments of the central office
to the field agent, who contacts tlie firm again in order to resolve
any inconsistencies that may have appeared in the data collected
and submits to the central office a final report, somewhat more
detailed than the preliminary one, on the organisation, set-up,
methods of production, employment conditions, changes in the
design of the product, etc., of the establishment.1
Analysis and Publication
After the major discrepancies have been so resolved, and the
field agent's final reports have been received by the analyst in the
central office, the latter is in a position to proceed to a final critical
evaluation of the data collected. Among the problems he will
have to deal with is that of the comparability of the products
from year to year. No product remains absolutely constant from
one year to another. Very often the changes may be disregarded,
since they are merely in design, and do not affect the unit
labour requirements to any significant extent. When, however,
important changes have occurred, either through the accumulated
effect of small changes over a period of years or, more usually,
because of a major change at a definite time, corrections are made,
for instance by using overlap periods and splicing values by the
link relative method. If such corrections are not possible or are
too debatable, the data will have to i e abandoned.
Similar problems arise because of variations in plant integration; it may be that a plant previously manufacturing parts now
1
The field agents, who are highly trained officials, generally possessing
an economic and technical background, have other duties relating to productivity statistics. They will answer any particular request for information,
and of course will keep in touch with establishments they have surveyed,
checking for instance that annual reports for periodical surveys are sent on
time.

SPECIAL ENQUIRIES

buys them from an outside establishment. Again, corrections are
made, by link relative procedures, or, when data are available, by
recomputing the past figures or the new ones on a comparable
basis: for instance, by deducting from the unit labour requirements previously computed the labour required to manufacture
the parts now bought from outside establishments.
It is in such cases that the information contained in the field
agent's reports will be essential. In addition, difficult cases will
give rise to discussions with the field agent and, if necessary,
with managers. The analyst will also visit a few plants of the
industry on which he is writing a report in order to be in a better
position to understand the main problems involved and to interpret the data correctly.
Unit man-hour requirements are then computed for each
product and each plant.
With few exceptions (e.g., cane sugar refining) no absolute
data on man-hours per unit of production are published. The
data are normally transformed into indices, to show the trends
over a period of years. These indices are computed as follows:
(1) A product index is calculated for each product. This is
the ratio of a weighted average of the unit labour requirements
observed in the various establishments, for a given year, to the
similarly weighted average for the base year. The weights used
are the quantities of the product produced by each establishment
in the base year (or sometimes in another year).
(2) A product group index is then calculated for each group
of products. This is a weighted average of all the product indices
of the items in the group, the weight for each product index being
the total hours of work expended in the total production of each
product in the base period. These weights have generally to be
estimated from various sources, since in general they are not
available from the reports, the reporting establishments representing only a part of the total number of establishments; these
estimates are frequently obtained by multiplying the total production of a particular product by the average unit labour requirements for the product as reported by the sample.
(3) An industry index is finally computed. This is a weighted
average of all product group indices, the weight for each product
group index being the total number of hours of work expended
in the base period in the total production of the group of products

METHODS OF LABOUR PRODUCTIVITY STATISTICS

represented by a product group index; these weights are also
estimated.
For the home radio receivers industry, for example, a product
index will be computed for an " A.G.-D.G. power supply, four
tubes plus one rectifier tube or selenium disc, one amplitude
modulation band "*'; a product group index will be calculated for
all table model radios, and the industry index will refer to all
home radio receivers.
Thus, it is clear that the computations are such that the indices
are not affected by shifts of production from one product to another
or one establishment to another.
Any variation in the indexes must be interpreted in terms of factors
affecting individual plants. The indexes would not be affected, for
example, by a tendency for the industry's output to concentrate in the
most efficient plants, although such a development would reduce the
average number of man-hours required by the industry as a whole to
turn out a given machine. The question whether or not production
tends to concentrate in the plants with
lowest man-hour requirements
is not within the scope of this report.1
Of course, the above description corresponds to ideal conditions.
In practice, it often happens that data are lacking for an establishment for one or more years (including the base year) or that
estimates of total man-hours expended in the production of a
given product are available only for a year other than the base
year, etc. In such cases, various statistical techniques, and in
particular link relative methods have to be used, inevitably
introducing some deviations in the results.
The computations are not confined to over-all indices. Indices
of year-to-year trends are presented by type of factory labour for
the industry as a whole (as represented by the composite of the
reported items) and for individual products, processes or departments. Within the limits of available statistics, these studies
provide data on average man-hoizr trends for groups of companies
classified on the basis of geographic location, size of plant, method
of production, utilisation of plant capacity, extent of technological
change, type of wage payment plan, diversification of output and
similar factors pertinent to the analysis of productivity. While
no data are shown that might allow identification of individual
plant figures, the trends of the indices of individual firms are sometimes given as an indication of their wide differences in behaviour.
1
UNITED STATES BUREAU OF LABOR STATISTICS: Monthly Labor Review,
Vol. LXV, 1947, p. 188.

SPECIAL ENQUIRIES

Supplementing these statistics, the surveys include extensive
information relating to the dynamic influences affecting the levels
of man-hours expended per unit. The trends are explained in
terms of such factors as increases and decreases in the volume of
production, different levels of capacity utilisation, changes in
quality of product, product design modifications, abnormal shortages of raw materials, changing skills and efforts of management
and labour, etc.
CONCLUSIONS

The direct enquiries into labour productivity, as at present
conducted by the United States Bureau of Labor Statistics, do
not seem as yet to have spread to any other country. However,
in the United Kingdom, under the impulse given by the AngloAmerican Council on Productivity, some steps have been taken
in this direction. For instance, the interim Report on the Measurement of Productivity issued by the Joint Committee of the Institution of Production Engineers and the Institute of Cost and Works
Accountants in December 1949 indicates the broad outline of the
practical action which is envisaged. The main feature of the
study is the suggestion to measure labour productivity for production processes rather than for products. This point has already
been developed in Chapter IV.1
It should be stressed that most of the difficulties involved in
the comparability between production and labour figures or indices
are avoided in direct enquiries.
The chief advantage of these enquiries is that, in contrast to
productivity indices based upon ordinary published data, they
provide detailed information and facilitate realistic analysis of the
factors affecting productivity trends. This approach is particularly
useful in those industries (notably durable goods) which produce
a heterogeneous variety of final products, and for which over-all
indices do not always have much meaning.
In addition, the detailed data collected permit comparisons
between plants, and therefore this method of enquiry has a
definite advantage in international comparisons.
Moreover, in addition to computations made on the basis of
fixed weights, as conducted by the Bureau of Labor Statistics,
computations could easily be made with changing production
1

See p. 50.

METHODS OF LABOUR PRODUCTIVITY

STATISTICS

composites, or with weights taking into account the change in the
composition of production. Therefore, these special enquiries
enable the statistician to answer questions concerning the variation
of labour productivity according to both concepts as defined in
Chapter I.
The main drawback to such enquiries is their cost: consultations with employers, workers and technicians, field operations
and the interpretation of data by qualified officials involve considerable expense.
It is also true that, owing to insufficiencies in the data collected,
many estimates and corrections are necessary. But the need for
such statistical treatment is not confined to special enquiries;
estimates and corrections are similarly used for obtaining the
basic data concerning production and employment which enter
into the computation of over-all productivity indices.

CHAPTER VII
INTERNATIONAL COMPARISONS OF LABOUR
PRODUCTIVITY
The various countries of the world have been paying increased
attention to the development of labour productivity data, because
of the importance of such data in the elaboration of production
and labour productivity policies. Studies have therefore been
undertaken of the variations of the ratio of output to employment
or man-hours, and the comparisons of such ratios or their reciprocals,
man-hours expended per unit of production, between plants of
different size, location, production processes, etc., producing the
same product.
But differences in labour productivity within an industry or
between different industries in the same country have often
appeared to be smaller than over-all differences with other
countries. Therefore, the study of such over-all differences has
been—especially since the second world war—of particular concern
to countries which find difficulties in exporting as much as they
must import, or to countries envisaging a new development in
industrialisation. Thus, international comparisons have been
attempted, either on a somewhat rough basis or through extensive
and detailed computations, as in the basic work of L. Rostas on
comparative productivity in British and United States industry 1,
which includes an analysis of the meaning, methods and limitations
of international comparisons of labour productivity.
In general, international comparisons of labour productivity
will be based on the physical output (per head or per man-hour) of
a given product, or a group of similar products (shoes of different
quality, for example) ; but in some cases, use has been made of the
value of net output (per head or per man-hour) of an industry or
a group of industries; this latter method involves, besides other
difficulties, the conversion of such net values into the same
monetary unit.
1

Comparative Productivity in British and American Industry, op. cit.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

The data are often obtained from censuses of manufactures;
another method is to collect data through special enquiries in each
industry, as was done in the Piatt Report in the United Kingdom.1
Whatever the method adopted, numerous difficulties will be
encountered in such comparisons. Many of the problems are
identical with or similar to those arising in national computations ;
others, such as the problem of exchange rates to be used in the
conversion of values of net output, are special to international
comparisons.
In the present chapter, therefore, the particular impact on
international comparisons of the problems examined in the previous
chapters is briefly reviewed and the special problems arising in
international comparisons are analysed in more detail.

PROBLEMS IN THE MEASUREMENT OP LABOUR

The various difficulties outlined in Chapter III in the measurement of labour (whether through employment or man-hours) also
arise in international comparisons, but in a more acute form.
(a) In national statistics, the definition of direct and indirect
labour varies according to the administration collecting the data;
the divergencies of concepts on this point between the various
countries are much greater than in any one country.
No comparability can be expected between the various national
existing concepts of direct and indirect labour, production workers,
wage earners or salaried employees. For instance, while in one
country the concept of " production workers " will exclude supervisory personnel, maintenance workers, etc., in other countries
these groups will be included.
In some countries wage earners are distinguished from salaried
employees merely on the basis of their system of remuneration
(pay periods of a month or of less than a month).
In others, the distinction is based on the kind of work performed :
wage earners include production process operators, maintenance
workers, etc. ; salaried employees include shipping and accounting
clerks, salesmen, etc. In both cases the allocation of supervisory
personnel varies.
To an unspecified extent the census employment figures include
persons employed not in the manufacturing process proper, but in
1
MINISTRY OF PRODUCTION: Report of the Cotton Textile Mission to the
united States of America, March-April 1944.

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

repair and construction work, transport or generating electricity. This
factor affects the margin of error in comparisons only in so far as the
proportion of such people to manufacturing employees differs in the
two countries—a question on which hardly any factual information is
available.1
Therefore, little comparability as between countries can be
expected between figures obtained for " output per wage earner "
or " man-hours of direct labour expended per unit of production "
computed on such a basis.
Even the definition of total employment lacks comparability
between the various countries. The Sixth International Conference
of Labour Statisticians has made recommendations concerning the
definition of total civilian labour force, employed persons, etc., in
order to remedy these discrepancies. International definitions of
wage earners and salaried employees, production workers, direct
and indirect labour, however, remain to be established.
(b) Differences in the content of the concept of man-hours
are equally wide; many conflicting interpretations are given to the
various notions noted in Chapter III of man-hours paid, man-hours
actually worked, etc.
Here again, standard definitions are lacking and therefore
comparison is difficult. Moreover—
when making productivity comparisons, whether in time or internationally, one further point has to be considered. Shorter hours of
work may be due to unemployment and the spreading of work over a
larger number of workers ; or they may mean a genuinely shorter working week, in the sense that the workers take part of their real earnings
in increased leisure. This differentiation is especially important owing
to the prima facie causal relationship between the number of hours
worked and output per man-hour. A number of studies made by
industrial psychologists have indicated that unduly long hours of work
reduce output; at the same time it is held that the relatively higher
output per man-hour in the United States is made possible to some
extent by shorter working hours, or in other words that the mass production methods employed in the United States could not be operated
permanently by operatives working long hours. If this argument is
true, and it requires careful investigation, it would mean that the
United States and the United Kingdom per man comparisons are
perhaps more realistic than per man-hour comparisons, and further
that it would not be possible in Britain both to have higher output
as well as maintaining longer working hours, since higher output per
man-hour could be achieved only by a shorter working week.2
(c) Another important point is the much greater differences
in the composition of the labour force in sex, age, skill, etc., be1
2

Comparative Productivity
Idem, p p . 26-27.

in British

and American

Industry,

op. cit., p. 20.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

tween various countries than between two periods of comparison
in one country or between labour employed in undertakings in the
same industry in one country. For instance, in 1938-1939—
41 per cent, of the United States population was in the labour
force, against 47 per cent, in the United Kingdom. The reasons for
this are the higher proportion of persons under 16 in the United States
(27 per cent, against 23 per cent.) and the lower proportion occupied
in any class. The difference in the proportion occupied is most marked
in the groups aged 16 to 19 and to a lesser extent 20 to 29. The
differences can be attributed almost exclusively to more extensive
education in the United States. Applying the United Kingdom percentages to the United States population, the United States labour
force is deficient, according to British standards, by 6.5 million workers
aged 16-64 (2.5 million men and 4.0 million women). Of these, 3.5 million are aged 16-19, 2.2 million aged 20-29 (almost all near 20) and
only 0.8 million above the age of university graduation. The composition of the two labour forces is also different, the proportion of men
to women being higher in the United States, and the proportion of
juveniles to adults lower.1
Such differences, which correspond to differences in the effort
t h a t labour can contribute, may have a definite influence on
productivity data, and therefore on comparisons. In international
comparisons, as in the case of national statistics, attempts have
been made to overcome them b y weighting the different types
of labour:
It would lead to an important refinement of our methods if, instead
of aggregating all operatives without distinction, a weighted figure,
expressed in " men units ", could be calculated. The method usually
available for weighting would be relative wage rates or hourly earnings
which would express broadly the differences between potential efforts.
For obvious reasons this is not always the case, especially in a country
like Britain with fairly strong delimitations of different kinds of jobs.
Unfortunately, data are not fully available for making such calculations. 2
L. Rostas used the ratio of 6/10 for women (i.e., counting
each woman worker as 0.6 " equivalent man "). Using this ratio,
the results of the productivity comparisons between the United
Kingdom and the United States would be about 10 per cent, less
favourable to the United States; in view of the fact that the
United States labour productivity in the selected manufacturing
industries for 1935-1939 was estimated as more than 2.5 times
t h a t of the United Kingdom, the adjustment would not have
1
" T h e P r o d u c t i v i t y of L a b o u r : I t s Concept and M e a s u r e m e n t " , loc.
cit., p p . 211-212.
2
Comparative Productivity in British and American Industry, op. cit., p . 23.

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

brought any appreciable change in the conclusions. But such
adjustments might be relatively of much greater importance when
comparing countries where the difference in labour productivity
is less marked than between the United States and the United
Kingdom (between Germany and the United Kingdom, for
instance).
T. Barna used weights to take into account not only the sex but
also the age composition of the labour force.1 The following
weights were assumed: persons under 16, one fifth; 16 to 19,
one half; 20 to 64, men, one; 20 to 64, women, two thirds; 65 and
over, men, two thirds; 65 and over, women, one half.
It is clear that such scales are arbitrary; if they are based on
wages, for instance, it may be noted that in many countries post-war
wages for women have increased considerably more than those of
men, while the ratio of output has presumably remained unchanged.
Physical effort could serve as a basis of such scales only in purely
manual jobs, and would not be relevant in most of the skilled ones.
Similar remarks could be made for any of the weights that have
been proposed for the conversion of the various kinds of labour
into a homogeneous unit.

PROBLEMS IN THE MEASUREMENT OF PRODUCTION

As in the measurement of employment or man-hours, the
difficulties encountered when measuring output in a country arise
in a more acute form in international comparisons.
(a) The differences in product specifications are greater than
when national comparisons are undertaken:
In our comparisons we compare the direct labour requirements
needed to produce the unit quantity of output, whether a ton of cement
or a pound of yarn, and therefore we neglect the quality factor. In
some cases, it is possible to account for some of the measurable aspects
of the quality factor, such as the differences in count of yarn or horsepower of cars, but it is never possible to account for all the quality
differences due to variety, size, shape, durability and style. While it
is important to bear in mind the importance of the quality factor in
productivity comparisons, it would be mistaken
to regard low physical
output necessarily as a mark of high quality. a
. . . But even beyond such quality differences, which can be easily
observed and appreciated by the layman, there are further differences
in " technical qualities " of apparently similar products. These " technical differences " may take the form of differences in chemical, physical
1
2

" The Productivity of Labour: Its Concept and Measurement ", loc. cit.
Comparative Productivity in British and American Industry, op. cit., p. 2.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

or similar properties of the product or they may consist of differences
in the tolerances of manufacturing specifications. These differences
may be often invisible to the consumer, and yet each manufacturer
may attach particular value to the specific technical qualities of his
product, and a higher technical quality may involve the use of more
labour per unit. 1
The influence of changes in product specifications is particularly
important when dealing with international comparisons: changes
in product specifications in the course of time may often be slow
and therefore have little influence on year-to-year comparisons in
a given country; when comparing outputs from one country to
another at a given time, differences in the specifications of products
bearing the same name might correspond to important differences
in labour required for the production of what appears at first sight
to be an identical " unit ".
Therefore, any computation of productivity figures would
need, to be of use, to be accompanied by a detailed description of
the product specifications.
(b) The difficulties arising out of the definition of the product
or the industry have been noted as one of the main problems of
productivity computations in national statistics. These difficulties
are increased when international comparisons are required.
Individual industries, as classified by the censuses of manufactures, each produce a group of products and by-products which
are not identical in the different countries as regards the relative
importance of individual types within the group. As a result,
comparisons of physical output per head can be made only for a
certain number of industries for which quantitative data are available and where the output can be reduced to a sufficient degree
of homogeneity. A second complication arises from the difficulty
of covering precisely the same range of productive processes in
the two countries compared. Therefore, in manufacturing industry,
comparisons are often confined to industries which have a rather
simple product structure, that is, which produce only one or a few
products, and to industries which produce approximately homogeneous products. Experience shows that even two firms in
the same industry of the same country seldom produce identical
products, and this will be even more true of whole industries in
different countries, since they will be differently organised, and
market requirements, climate, taste, etc., will all be different.
1

Idem,

p p . 13-14.

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

Full details relating to these various difficulties have been
noted by L. Rostas :
To start with, it is possible only to compare industries the product of which is broadly homogeneous. That is, they produce
mainly one product, such as cement, or different sub-groups of the
same product, such as men's shoes, women's shoes and so forth. It
is not possible to compare industries with a heterogeneous product
structure, such as chemicals or plate and jewellery. In many cases the
difficulties are purely statistical, where we have an insufficient breakdown of the industrial classification.
Even in cases where there is one product in the industry, if
this product consists of a number of sub-products, such as different
types of shoes or different types of motor cars, the difficulty arises in
converting these sub-products into one homogeneous product. In this
respect international comparisons present perhaps a smaller difficulty
than, say, inter-regional comparisons, as there is more likelihood that
the product structure of two countries will be broadly similar. The
proportion of men's and women's shoes in the total number of shoes
produced in the United Kingdom and in the United States will perhaps
not differ very much, whereas the product structure of, say, England
and Wales would differ.
A similar problem arises when instead of many sub-products there
is one main product and a number of by-products, all specific to
the industry under investigation. Thus, in the case of the coke
industry, coke is obviously the main product, but a number of important
gases, etc., which are by-products of the trade, have to be considered
too. 1
The output of the industries compared will consist not only of
the output of the main product or products in which we are interested
but of some other by-products or ancillary products, and therefore
part of their labour force will be used for producing these latter products.
A corollary of the [latter] difficulty is that part of the main
product in which we are interested is produced as the by-product or
ancillary product of some other industry. Thus the motor car industry
may produce such ancillary products as aircraft engines and parts,
wireless apparatus, electrical machinery, tractors, etc. At the same
time, some part of the motor cars may be produced in other industries
than the motor car industry. In both cases mentioned the proportions
of the labour force and of the product involved may differ in different
countries.
Industries as defined by the censuses may cover a different part
of the productive process. Cement-making may include or exclude
quarrying of lime and chalk. Again, in all the textile industries,
textile finishing may be treated as part of the cotton or woollen industry
or as a separate industry. This will depend on the actual technical
integration in the industries concerned and secondly on the census
treatment of such subordinated industries. The usual procedure is to
classify each plant into one industry on the basis of its main activity.
But if the census regards the subordinate processes as separate industries and data relating to the performance of the plant covering the
whole process can be divided up, this is usually done.
1

Idem,

p p . 12-13.

METHODS OF LABOUR PRODUCTIVITY STATISTICS

Industries compared may perform a different number of processes, even if they cover roughly the same scope. The steel industry
is the most obvious example. Two steel industries compared may
differ because one is producing all its raw steel, while the other is
importing part of the bars, etc. Or they may differ because one is
taking rerolling a stage further than the other.
Industries compared may differ in scope when part of the production process consists of making parts and accessories, and the
other part consists of assembly. This problem is especially important
in practically all engineering industries. The motor car industry is a
good example. To make a comparison between the motor car industries of two countries one can attempt one of two things: to compare
the assembly work solely, which is a difficult proceeding considering
that in practice assembly and the making of parts and accessories are
not at all easily distinguishable, and this would be reflected in the
statistical material available; or one tries to compare the whole process
of the making of the motor car, including the making of parts and
accessories. The choice of processes and activities to be included is
sometimes arbitrary—making radiators is regarded as part of the
motor car industry, while making tyres is not, but in so far as an
identical range of activity is covered in the countries investigated the
arbitrary factor will not have much effect.
Ancillary manufacturing activities may be performed either within
the industry investigated or outside the industry. For example,
making packing material or printing, etc. In so far as these activities
are not treated as separate industries by the censuses, differences will
arise.
Repair and maintenance of installations may be included in the
industry (i.e., in the labour force) or they may be treated separately.
Non-manufacturing activities may be performed outside the
industry. The two main examples are transport and generating electricity. 1
(c) Differences in integration are generally sharper between
the various countries than between the various plants of an industry
within a country:
If we compare cotton weaving in the United States with cotton
weaving in the non-integrated British mills, the basic figures—quantity
and value of piece goods produced, the total value of gross output,
number of persons—on the surface appear to be comparable. The
total value of gross output is largely accounted for by the value of
piece goods produced. In reality, however, there is a vast difference,
and a large, in fact the major, part of the employees in the United
States has been engaged in making yarns, without this additional
activity being reflected in any of the figures used for comparison. In
this particular case the intermediate product was transferred within
the same plant or within the different plants of the same firms and
did not appear in the final value of gross output. If a great proportion
of the intermediate product were sold within the industry, the data
relating to the final product of the industry would be the same, but
1

Idem,

p p . 11-13.

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

owing to statistical duplication there would have been a substantial
difference between the value of total gross output and the value of the
final product or products. 1
Another type of difficulty arises when the two industries compared produce the same product, but also produce for sale intermediate products, and the proportion of intermediate products
for sale varies in relation to the total product. For example,
the British cotton industry—so defined as to include spinning
and weaving—has two main products : yarn and cloth ; yarn is also
an intermediate product for its other main product, cloth, whereas
the United States cotton textile industry has, broadly speaking,
only one main product, cloth, and the bulk of the yarn produced
is used in the same plant or is transferred within the industry.
A similar type of problem arises in engineering, where parts
and accessories are mostly assembled into the finished product,
such as motor cars or wireless sets, but the same intermediate
products are also sold to the final consumer for replacement.
Thus, a comparison of the labour productivity of steel industries
in the United States and the United Kingdom based on the output
of ingots (as a measure of output of the steel industries) would
not be correct, since it would make no allowance for the considerable quantities of imported ingots and semi-finished steel which
are worked up in the United Kingdom.
Finally, in some industries the measure of output of heterogeneous finished goods is often estimated b y using as an indicator
the quantity of raw materials consumed. This may lead, in international comparisons, to wrong conclusions.

COMPARABILITY OF PRODUCTION AND LABOUR DATA

(a) In a given country, the proportion of a certain product
which is produced outside the industry, and, within the industry,
the proportion of the main product to by-products, may remain
fairly constant over not too considerable periods of time; but
such proportions are seldom the same in every country.
Therefore, adjustments are necessary in order to allow for the
number of persons employed in the production of by-products
and ancillary products not taken into consideration in the comparison, as well as for those persons producing part of the main
1

Idem, p. 20.

100

METHODS OF LABOUR PRODUCTIVITY STATISTICS

product but classified in other industries. Such adjustments are
made, for instance, on the following basis:
The total number of persons employed as given in the censuses
relates to the total value of the gross output of the trade. As the
total value of those products which are included in the comparison is
also available, the number of persons employed can be corrected for
differences. The correction factor is given by the ratio of the value
of included products to the total value of gross output . . . The
unstated assumption underlying this method is that the value of gross
output per head is the same in respect of the main product produced
in the industry and outside the industry . . .
If the total value of the product specified both by quantity and
value does not account for the major part of the total value of gross
output, the method of adjustment described above is unsatisfactory,
in the sense that the number of persons associated with the particular
output cannot be reliably estimated. Another case when this method
is not applicable is when it can be assumed that the total value of
gross output contains a large amount of duplication. Duplication in
the value of gross output is a clear warning that great care should be
taken in the comparison of the processes covered.1
(b) The influence of variations over a period of time in the
size of the establishments included in the censuses of production
or in the computation of production or labour indices has been
noted in Chapter V. When proceeding to international comparisons, the same type of difficulty is encountered, because the
scope of the various national censuses is not identical.
For the British census for 1935 and for the 1937 Import Duties
Act enquiries, " detailed returns were not obtained from persons
employing not more than 10 persons as a yearly average " ; the excluded
small firms gave only the number of persons employed; while, for the
United States censuses, data have been collected only from establishments reporting products to the value of $5,000 or more. As the
value of product per person engaged amounted to $6,270 in 1937, the
scope of the United States census is wider than the British census
(except for repairs and " custom " industries). In 1935, in the factory
trade covered by the British census, 536,600 persons were estimated
to have been employed by small firms (with not more than 10 employees)
as compared with 5.2 million persons employed by firms with 10 or
more employees. In individual industries, of course, this ratio is even
higher. For the United States, the only indication of the excluded
persons is given in the 1939 Census of Service Establishments, where
persons employed in such categories as automobile repairs and services,
other repair services, and especially custom industries, are shown;
categories which are broadly comparable to, although more extensive
than, the excluded British small firms. But the number of persons
employed in these services was relatively small in comparison with the
9.1 million persons employed by the census industries.
1

Idem,

p p . 19-20.

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

101

The effect of these differences in the scope of the censuses on relative
productivity depends on the relative efficiency of different-sized establishments and plants. Available evidence suggests that small plants
and small firms have a lower output per head than bigger plants, and
thus their exclusion from the British census makes the comparison1
more favourable for Great Britain than it would be by their inclusion.
The importance of such considerations becomes obvious when
comparisons are attempted between countries of completely
different industrial organisation, such as the United Kingdom and
China, for instance.
In a study of industrial production and employment in pre-war
China 2 intended to present data comparable to that shown b y
L. Rostas 3 , some of the definitions and limitations in scope were
stressed. Factories in China were defined as those firms employing
more than 30 employees; " h a n d i c r a f t " covered all other firms,
as well as family subsidiary works and independent craftsmen.
The scope of manufactures in the other three countries is much
wider (in the United Kingdom, all firms employing more than
10 persons; in Germany, firms employing five or more employees;
in t h e United States, firms producing an output worth $5,000
or more).
But the authors of the comparison noted that in terms of output
produced, Chinese firms with 30 employees rarely attain $5,000
worth of output, or the output of a United Kingdom shop with
10 employees. The ratio of factory employment (as defined here)
to total employment compared with that in the other three countries as follows: China 0.25 per cent., Germany (1936) 9.12 per cent.,
the United States (1935) 6.7 per cent., the United Kingdom (1935)
11.3 per cent.
SPECIAL PROBLEMS OF INTERNATIONAL COMPARISONS

(a) In a given country, the proportion of the output of a
given industry to the total does not generally vary rapidly over
a period of time. Therefore, in year-to-year changes, t h e influence
of this factor may often be of secondary importance.
In international comparisons, on the contrary, the different
structure of the output of each country, combined with sometimes
1
2

Idem, pp. 24-25.
PAO-SAN OU and FOH-SHEN WANG: " Industrial Production and Employ-

ment
in Pre-War China ", in the Economic Journal, Vol. LVI, 1946.
3
L. ROSTAS: "Industrial Production, Productivity and Distribution
in Great Britain, Germany and the United States ", Economic Journal,
Vol. LIII, Apr. 1943.

102

METHODS OF LABOUR PRODUCTIVITY STATISTICS

considerable differences of productivity in the various branches of
economic activity, may have a predominant effect on the results
of the computations.
Here, as in time-to-time comparisons within a country, two
measures (or sets of measures) can be made. One type consists of
comparisons based on identical production composites; the productivity of each individual industry is weighted with the same
weights in each country (these weights may be those of one of the
countries compared, or of an intermediary or standard production
complex). The second type consists of comparisons based on the
production composite of each country as it stands; the productivity
of each individual industry is weighted by the importance of t h a t
industry in the country.
In the latter case, in addition to the differences due to different
productivity rates in each industry, the importance of each industry
in each country, and especially the relative importance of high
productivity to low productivity industries will intervene.
The meaning of these two sets of measures has been examined
in detail in Chapter V. It would appear useful, for international
comparisons, to compute over-all productivity data according to
both methods, as was done by L. Rostas in his study of comparative
productivity in the British and United States manufacturing
industry: misinterpretations of the results would then be avoided.
(b) Among the many factors affecting labour productivity, it
has been pointed out in Chapter II t h a t the rate of capacity used
was one of the most important. Therefore, especially when comparing productivity data compiled from censuses of production,
special attention should be paid to the consideration of the rate of
capacity used, which is influenced, for instance, by general economic
conditions.
This important point is emphasised by the fact that the years
for which censuses of production are taken are seldom the same in
the various countries: for this reason a comparison is relevant
only if the years chosen in the countries to be compared are reasonably comparable as to the rate of unemployment and the use of
equipment in relation to capacity.
Considerations of this order had a definite influence on the
choice of the data to be compared in a study made by A. W. Flux.
The years to which the American data apply have not, so far, been
the same as the years selected for the British enquiries, and this circumstance may be an important obstacle in the way of carrying out
the desired comparisons. Thus, the first census of production in the

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

103

United Kingdom dealt with the year 1907, and the United States data
for 1909, the nearest available data, may reflect industrial conditions
which had been affected by the business crisis of 1907 sufficiently to
invalidate, in part, many of the deductions that might otherwise be
made from the recorded data. The second of the British series of
censuses related to 1912; the United States data for comparison, being
those for the year 1914, may have failed to represent normal conditions,
in view of the war demand on the resources of a great neutral manufacturing country. After the war the re-establishment of the British
census was not effected until 1924, to which year the third of the series
relates. For comparison we have the results of the United States
Census of Manufactures for 1925. The fourth of the British censuses
dealt with the year 1930, and the available material for comparison
relates, in the case of the United States, to 1929. It may, perhaps,
appear that a six-year interval would be better for comparative purposes, and that either the American data for 1923 should be used for
the earlier comparison or those for 1931 for the later. It is, however,
desirable to put as long an interval as may be feasible between the
disturbed industrial conditions resulting from the feverish war activity
and the dates selected for such inter-country comparisons, and it
appeared, therefore, preferable to set the British data for 1924 against
the United States data for 1925, rather than those for 1923.1
(c) Although the use of net outputs to measure the production
of an industry is subject to such drawbacks as have been indicated
in Chapter IV, international comparisons have sometimes been
attempted on this basis, particularly because such measures seem
to overcome the difficulties due to differences in product specification, degree of integration, etc. But when comparisons of values
of net output, per head or per man-hour, are made between countries, a very, important additional problem is encountered : that
of the conversion of the data for each country into the same monetary unit.
The choice of the rate of conversion is of fundamental influence,
but involves great difficulties. It would be necessary to make the
conversion at the purchasing parity rate for the products investigated.
For this purpose it is necessary firstly to list the main products
of the industries to be compared, and, secondly, to ascertain the prices
of these products in both countries. It would be even more appropriate
to ascertain the purchasing parity rate for both the finished products
and for the materials which were used in their manufacture, and thus
to calculate an exchange rate which is more corresponding to the
concept of net output. Prices, of course, must relate to identical
products as well as to identical conditions of sale, for example, " ex
works " or " delivered at site ", and so on. The ratio of such prices
gives the exchange rate in terms of these products.
1
A. W. FLUX: " Industrial Productivity in Great Britain and the United
States ", in the Quarterly Journal of Economics, Vol. XLVIII, Nov. 1933,
pp. 1-2.

104

METHODS OF LABOUR PRODUCTIVITY STATISTICS

The ascertainment of comparable prices is not easy, if possible at
all. Two supplementary methods can be applied, neither of them
being very satisfactory.
(1) Instead of the purchasing parity rate the official exchange
rate merely can be applied, which expresses price differences in terms
of all commodities and not only of those under investigation.
(2) The exchange rate in terms of average factory values (as
ascertained from the census) can be calculated: this would be identical
with the purchasing parity rate only if the average product of the
countries compared were identical (which is very seldom indeed), but
the differences between this rate and the general exchange rate usually
gives some indication of the direction in which the purchasing parity
rate would deviate from the general exchange rate. 1
An example of the considerable differences that might be
involved in such computations can be shown on the basis of a
comparison made between the United Kingdom and Germany for
1935-1937. 2
In this study, the rate of Í7.08 Reichsmark to the pound
(used by the German Institute of Business Research) was adopted
as the appropriate rate of exchange for measuring relative purchasing power in 1935-1937.
Under this assumption, the German output of manufacturing
industries in 1936 was 127 for 100 in the United Kingdom in 1935.
Had the official rate been used (12.2 Reichsmark to the pound),
the German to the United Kingdom ratio would have been 40 per
cent, higher, t h a t is, 177. It is unlikely that this rate actually
represented the purchasing power parity; but this example should
show the considerable differences t h a t might be encountered if care
is not taken to use the correct rates. The same German Institute
arrived at a purchasing power rate of $4.68 to the pound, instead
of the $4.92 used by L. Rostas; the difference between these
two estimates is 5 per cent.—a not inconsiderable extra error to
introduce into calculations of labour productivity.
When net outputs per head are compared, one may thus
wonder whether the differences observed express a difference in
the prices ruling in the two countries or the difference in the productive capacity of the two organisations:
The figures showing the nominal increases in net output per head
are difficult of interpretation, as they reflect in the main the changes
in the value of money over the period examined. As already noted,
the fact that the year 1907, for a large part of its duration, was a time
of relatively high prices and flourishing trade, may well account for a
1
2

Comparative Productivity in British and American Industry, op. cit., p. 10.
" Industrial Production, Productivity and Distribution in Great Britain,
Germany and the United States ", loc. cit.

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

105

smaller proportionate advance in prices from 1907 to 1924 than from
1909 to 1925, without calling for explanation based on different degrees
of industrial progress in the two countries.1
The results obtained when substituting factors of conversion
such as official exchange rates, average factory values, etc., for
the purchasing parity rates, may thus vary considerably. Moreover,
the use of a general factor of conversion may be entirely erroneous,
since the assumption that the ratio of prices in the individual
industries is the same as the ratio of general price levels is seldom
true. It might be proposed to convert the value of output of each
industry at the purchasing parity rate of that industry; if the
amount of research work involved in such computations is considered, it will be realised t h a t there is little to be gained by using
this method rather than comparisons in terms of physical output
per head.

SPECIAL ENQUIRIES

Since numerous difficulties are involved in the international
comparison of production and labour data taken from censuses of
manufactures, or from data collected for the measurement of
production and labour per se, it is useful to examine whether
special enquiries into labour productivity would not, as in the
collection of national data, yield better results.
The advantages of such enquiries lie in the fact that data
concerning production and labour are collected simultaneously;
these enquiries enable the collection of detailed data of unit labour
requirements per operation; they furnish information per plant,
therefore permitting of comparisons according to differences in
size, location, production process, etc.; and they are easily supplemented by detailed information concerning technological differences in the processes followed.
When comparing labour productivity in an industry in
different countries, it must be borne in mind that the production
processes (machines used, nature and quality of raw materials, etc.)
vary widely from one country to another, and while an over-all
result is certainly of great interest to show the differences in
numbers of man-hours or men required for a certain production,
interpretations of such differences cannot be made if details are
not available on the internal structure of the industry.
1

A. W. FLUX, loc. cit., p.

30.

106

METHODS OF LABOUR PRODUCTIVITY STATISTICS

Thus, the fact that " the American weaving equipment and
organisation are based on the automatic loom, which constitutes
about 95 per cent, of the American loomage, and the British on
the non-automatic loom, representing also about 95 per cent, of
the total looming in Britain 1 ", lias to be known if correct interpretations of the lower British over-all productivity of labour in
the cotton industry are to be made.
One of the main advantages of special enquiries is that they
allow separate comparisons of each phase of the productive process and thus indicate at what stage and in which part of the
process substantial differences may be observed. The report of
the Cotton Textile Mission to the United States, for example,
indicated that although output per man-hour is much higher
at every single stage of cotton textile production in the United
States than it is in the United Kingdom, it is particularly high
in the stage of winding and beaming.
The main drawbacks of such a method lie in:
(1) its need for preciseness, which often confines it to clearly
measurable outputs;
(2) its cost, since it is undertaken separately from current
collection of data concerning production or labour.
The main problem in the making of special enquiries lies in
the correct selection of plants in each country, so that they are
representative of the industry as a whole. This is far from being
an insuperable problem, especially with the development of sampling methods.
Once data have been collected in each country for a representative sample of plants of the industry (representative in size,
location, production processes, etc.), several computations may be
made, all of which are useful and necessary for the interpretation
of the actual situation.
Comparisons can be made:
(i) between the productivity of two " identical " plants of the
same size, operating the same processes, etc.;
(ii) between the unit labour requirements expended to produce
identical products from identical raw materials through entirely
different processes;
1
Report of the Cotton Textile Mission to the United Slates of America, MarchApril 1944, op* cit.

INTERNATIONAL COMPARISONS OF LABOUR PRODUCTIVITY

107

(iii) within each country, between the performance of plants
of various sizes, locations, production processes, etc.;
(iv) between countries, of the " average " productivity, considering the actual distribution in size, location, etc., of the plants;
international comparisons on this latter basis will indicate differences due to the compound effects of differences in production
processes, size, etc., and efficiency.
Finally, with this method, computations corrected for any of
the factors affecting labour productivity are possible, whereby
information is gained on the influence of that factor.
It must also be recalled that, as noted for national computations in Chapter VI, once basic data are collected through special
enquiries, they may always be combined with any weights that
are desired, either to produce comparisons of productivity as a
characteristic of each undertaking or to study differences in
productivity, including differences in the composition of production.
Special—or sample—enquiries thus appear to have, from the
international point of view, the considerable advantages of eliminating most of the drawbacks due to the lack of comparability of
production figures with labour figures or their non-comparability
in time, and, in addition, permit of detailed comparisons and
analysis of the influence of the various factors affecting labour
productivity. They require the collaboration of technicians,
because of the necessity of complete technical knowledge of the
processes of the industry studied; but they appear essential if
complete information is to be gathered on international labour
productivity differences.

CHAPTER Vili
INTERPRETATION OF LABOUR PRODUCTIVITY DATA
The previous chapters have been mainly concerned with the
difficulties encountered in the measurement of labour, output and
labour productivity, and of the insufficiencies of the data actually
used to measure and compare labour productivity for products,
undertakings, industries and countries.
An attempt will be made in the present chapter to bring out
the meanings of the various data currently issued on labour
productivity, in the light of the solutions generally adopted to
the main problems raised by the measurement of labour or production.
THE CONCEPTS

The interpretation of measures referring to the labour productivity of a single independent worker, or of a gang of workers,
is simple ; difficulties arise when the unit under consideration is
an undertaking and become complicated when a group of undertakings forming an industry is considered; the interpretation of
labour productivity data for a group of industries, or for a branch
of economic activity, is extremely intricate ; finally, the computation of the labour productivity of a nation as a whole may involve
the use of such complex concepts as the national income, and are
extremely difficult to interpret rightly.
The data to be collected depend—or should depend—on the
purpose of the investigation. Reciprocally, interpretations of
the results obtained will depend on the data actually used; the
use of data of a certain kind allows correlative conclusions to be
drawn, but no other conclusions. Thus, the concepts used, explicitly
or implicitly, must be determined before any interpretation may
be undertaken, and particularly the distinction must be made
between the two main concepts discussed in Chapters I and V.
For instance, when an index of unit labour requirements is
computed in the United States for the cigar and cigarette industries by dividing a production index with value weights by a

•

INTERPRETATION OF LABOUR PRODUCTIVITY DATA

109

man-hour index, the index understates the amount of labour
required after 1929 to produce the 1929 volume of cigars and
cigarettes 1 , but provides an indication of the total labour actually
required to produce, say, the 1936 production composite of cigars and
cigarettes, as compared with the total labour actually required to
produce the 1929 production composite. Information obtained on
the basis of this index with value weights would not be furnished
by an index weighted with unit labour requirements, and vice
versa; both indices are useful indicators, however, of different
aspects of the same facts.
MEASUREMENT

The way in which results are influenced by the main features
of the various measures currently used in labour productivity
computations are reviewed below.
Labour Productivity and Labour Data
(1) Actual effort, particularly non-manual effort, is scarcely
ever measured. Labour is commonly measured by the time spent
at work; variation in the quality of effort contributed is one of
the factors reflected in changes in labour productivity. Variations
in labour productivity measure effort alone only when all other
factors affecting labour productivity are maintained unchanged,
but this is very rarely the case.
(2) The meaning of the results varies according to the kinds
of labour taken into consideration; within an establishment, the
productivity of direct labour will vary differently from that of
indirect labour, and the productivity of the total labour expended
in the establishment may thus have a trend different from either.
Whenever possible, it will be useful to collect data for direct and
indirect labour separately, since this will throw light on the
influence of each of these kinds of labour. Standard international
definitions of direct and indirect labour are not at present available, so that general conclusions on comparisons of the productivity
of direct or indirect labour separately from country to country are
not at present possible.
The problem of embodied labour, or labour spent in the building
of the machines used in a given process, is of great importance.
No accurate statistics can at present be set up to measure such
1

See Chapter V, p. 65.

110

METHODS OF LABOUR PRODUCTIVITY STATISTICS

labour, and therefore the measures of labour productivity
currently issued do not take it into account. It must thus be kept
in mind, when interpreting labour productivity data, that the
data obtained correspond in general to the labour spent on the
premises of the establishments only, that is, they correspond to
certain given conditions of production and equipment, and that
comparisons between establishments might be considerably
influenced by the differences in equipment.
(3) In the computation of man-hours contributed, most
measures of labour productivity make no allowance, because of
the lack of suitable weights, for differences in the output of persons
of different sex, age or skill. As a consequence, differences in
sex, age or skill in the labour force will affect comparisons of
labour productivity. Obviously, data concerning the composition
of the labour force are essential for evaluating the influence of
differences in the composition of the labour force on the differences
observed in productivity.
(4) The difference between " output per man " and " output
per man-hour " need not be recalled in full detail. Each kind of
data serves different purposes : " output per man " data should
be compiled with a view to estimating employment possibilities,
future national income, etc., in other words, when interest is
focused on the number of jobs; "output per man-hour" data
should be used when the productive capacity of labour, the cost
of production in labour time, etc., are under analysis.
Labour Productivity and Production Data
(1) In the interpretation cf productivity data, particular
attention should be paid to variations in the course of time and
to the non-comparability from undertaking to undertaking of
product specifications ; many of the apparent variations in labour
productivity, especially over short periods, and many of the
apparent differences in labour productivity between one undertaking and another may be due, at least partly, to differences in
the quality, shape, size, etc., of products designated by the same
name. Measures of labour productivity for a production process,
rather than for a product, may ensure more comparability,
and thus help to avoid misinterpretations.
(2) Problems of interpretation of labour productivity data
similar to those raised by the consideration of " embodied labour "
are raised by differences in the degree of integration of various

INTERPRETATION OF LABOUR PRODUCTIVITY DATA

111

establishments. Such differences may explain, at least partly,
differences in unit labour requirements for two establishments
producing the same goods:
If output per man-hour goes up because a plant now buys its
electricity instead of making its own, it is very likely that some social
and private gain has resulted, but it will not be measured at all accurately by the decline in unit labour requirements in that plant. 1
Such will be the case whenever an establishment expands to
take over the production of semi-finished goods previously
obtained from other sources or restricts its scope and specialises
in part of the production it was previously carrying out. As
output is still measured by the weight, or the number of units, etc.,
of the same finished product, variations in computed labour
productivity might appear which are caused solely by changes in
the production process: they correspond to displacements in the
total labour producing an article, but not necessarily to a decrease
or increase in the unit labour requirements for the article considered.
The use of computations based on the value of net output is
of little help here, especially for absolute productivity data,
because of the great number of factors unrelated to productivity

by which such computations are affected. However, simultaneous
information on the variations of gross and net output per unit
will be of great assistance in the interpretation of the data.
An example of the difference in meaning of data based on gross
output from that of data based on net output, and of the conclusions that might be drawn when in possession of both sets of data,
is shown by comparing the two major indices issued by the Bureau
of Agricultural Economics of the United States for measuring
labour productivity in agriculture. 2
The indexes of gross farm production per worker do not measure,
specifically, changes in the effectiveness with which farm labour is
used. The indexes do reflect the increases in production per worker
which are due to the substitution of city-made tractors, trucks and
automobiles for farm-produced horses and mules, the use of better
seeds, more fertiliser, soil conservation practices, etc. All in all, the
indexes of gross farm production per worker can be said to measure
only changes in the quantity of production per farm worker that emerge
from agriculture. The indexes of farm output per worker reflect changes
in the volume of product per worker resulting from the diversion of
feed for workstock to feed to produce a greater volume of livestock
and livestock products for human consumption. In other words,
1
2

Measuring Labor Productivity, op. cit.
For the definition of these indices, see p. 73.

112

METHODS OF LABOUR PRODUCTIVITY STATISTICS

changes in gross production more nearly represent changes in the
product for which farm workers were responsible.1
To overcome the difficulties arising from different degrees of
integration, unit labour requirements figures may be computed
for the various processes within an establishment; comparison
between plants of different degrees of integration are then possible,
taking into account for the comparisons only those unit labour
requirements corresponding to the same phases of the production
process.
Labour Productivity Formulae
The concepts and formulae on which the computations of
labour productivity or unit labour requirements indices are based
are intimately linked. It has been shown2 that they can be grouped
in two main types. It is essential, in order to interpret the measures
correctly, to know clearly what formula and what weights, implicit
or explicit, have been used, or at least to what main type they
belong.
In particular, the ratios of production indices to employment
or man-hour indices—indices computed to measure production or
labour variations, and not especially in view of productivity
measures—are influenced by changes in the composition of output ; if
the influence of the latter is to be avoided, special measures
must be used, in which individual outputs will be weighted by

unit labour requirements. Whenever possible, it would of course
be preferable to secure data according to various types of formula,
so as to be in a position to throw light on all aspects of the variations of the facts observed, including changes in the composition
of production, which are as important as any to the interpretation
of labour productivity changes.
THE INTERPRETATION OF OVER-ALL MEASURES

Measures of labour productivity as computed in the past have
clearly shown one important fact: labour productivity levels and
trends often show considerable differences from one undertaking
to another, one industry to another, or one branch of economic
activity to another:
Realisation of the striking differences in productivity trends of the
various industries is extremely important. These diverse trends among
1
2

Major Sources of Productivity Information,
I n Chapter V a n d A p p e n d i x I .

op. cit.

INTERPRETATION OF LABOUR PRODUCTIVITY DATA

113

industries characterise all periods of time. The index for manufacturing
as a whole, then, is an average of the indexes for the component industries and represents the net effect of many widely dissimilar separate
trends. As a consequence, no industry can safely assume that its
productivity is increasing on the basis of an increase in the index for
all manufacturing.1
Variations in unit labour requirements in the United States
from the beginning of the century to the outbreak of the second
world war afford an illustration of this fact:
The changes in wage earners per unit between 1899 and 1937 range
over a wide scale, from a net rise for forest products of 10 per cent.
to a drastic cut of 80 per cent, in the group of industries making tobacco
products. Transportation equipment, petroleum and coal products,
and printing and publishing were all characterised by extreme declines
in the employment-output ratio: in 1937 these industrial groups
employed only one fourth as many wage earners per unit of product
as they had 38 years earlier. In the paper products and chemical
groups, the declines, though smaller, exceeded 50 per cent.; in all
the remaining groups the cut in workers per unit was less than the
50 per cent, average for manufacturing as a whole.2
This quotation refers to variations observed for industrial
groups; the divergencies between the variations of individual
industries are of course greater. The 1937 index of wage earners
employed per unit of product (1899=100) varies from 12 (automobiles, including bodies and parts) to 225 (locomotives).
Thus, averages for manufacturing conceal wide variations from
industry to industry. These wide differences between various
industries in the variations in labour productivity or unit labour
requirements over a period of time have also been noted by
J. Fourastié, who considers them as the fundamental factor
influencing modern economy, and points out that, while in most
manufacturing industries the decreases in unit labour requirements
over time are considerable—
very large sectors of economic activity have remained practically
unaffected by technical progress. For example, the men's hairdresser
does not attend more customers in 1948 than in 1900; entire occupations have not modified their methods of work from 1900 to 1930.3
D. Weintraub has also remarked that an analysis of productivity increases in basic industries and services shows t h a t
1

Productivity, Employment and Living Standards, op. cit.

S. FABRICANT: op. cit., pp. 31-32.

J. FOURASTIÉ: op. cit.,

27.

114

METHODS OF LABOUB PRODUCTIVITY STATISTICS

productivity of services has practically not increased. He points
out, however, that—
even if similar relative increases in productivity had occurred in both
fields, a levelling off of the index of labour required per unit of output
would have taken place if, during the same period, service activities
had accounted for a rapidly growing proportion of the total o u t p u t . . . .
The divergency of the trends of employment in the several industrial
groups discussed and the variations in the extent of the productivity
changes make it clear that the over-all productivity ratio derived
from the data on total national income and total employment cannot
be interpreted as measures of the extent of technological advance in
individual industries any more than the per capita monetary national
income figures can be used as measures of the incomes of individual
groups in society. . . . Over-all productivity ratios reflect a variety
of factors in addition to the mechanical improvements usually characterised as " technological ".x
It appears that the explanation of the apparent divergencies
between various labour productivity data is to be found at least
partly in these wide differences from plant to plant, from industry
to industry, or from one branch of economic activity to another.
As the most easily measurable labour productivity is that of
manufacturing, because it is possible to define production more
easily in this branch of economic activity than in others, and
because data concerning production and employment (current
statistics and censuses of production) are more generally available,
most comparisons made up to the present between various
countries have referred to manufacturing. Such comparisons are
often in apparent contradiction with other data, and especially
with comparisons of real national income per head. Therefore,
some authors have attempted a " reconciliation " between various
international comparisons of labour productivity and real national
income per head.
The following table illustrates the kind of computations made.
(The data refer to the period 1935-1939.) A similar table was also
established by T. Barna. 2
To give its full weight to this table, it should be mentioned
t h a t J. R. N. Stone estimated the per capita national income of the
United Kingdom in 1938 as amounting to £97 and that of the
United States to £103 (the United States national income being
converted into pounds at the official rate of exchange). These
figures correspond to an index of 106 for the United States as
1
2

Unemployment

and Increasing

cit.,

B A R N A : op.

216.

Productivity,

op.

cit.

INTERPRETATION OF LABOUR PRODUCTIVITY DATA
TABLE XII.

115

RECONCILIATION OF PRODUCTIVITY COMPARISONS WITH

REAL INCOME COMPARISONS, UNITED KINGDOM AND UNITED STATES

Output
Basis of comparison

Estimated output per man-hour in manufacturing
. .
Estimated output per worker in manufacturing . . . .
Estimated output per worker in all branches of the
national economy
Allowing for distribution of working population coupled
with lower relative output per worker in the different
industries of the same country
Allowing for difference in the ratio of working population
to total population
Allowing for higher United States unemployment in the
year of comparison
Allowing for slightly higher British income from foreign
investments

U.K.

U.S.A.

100
100

273-292
212-224

100

163-183

100

150

100

128

100

124

100

118

Source: Comparative Productivity in British, and Amsrican Industry, op. cit., p. 92.

compared with 100 for the United Kingdom. 1 Similarly, the level
o f consumption of final goods and services of the United Kingdom
was estimated at between 10 and 20 per cent, below that of the
United States. 2
NECESSITY FOR DETAILED

INFORMATION

The complexity of the interpretation of labour productivity
data has been reviewed and stressed. At all levels—the plant,
the industry, the branch of economic activity and the national
level—the meaning of absolute figures, indices or comparisons is
difficult to interpret, and the variations in the figures are influenced
by a considerable number of factors. In addition, the meaning of
the data is also dependent on the formulae used and in general on
the methods of collection and tabulation of the data:
One may well ask, at this point, what significance do figures have
which show the relationship, over a period of time, between output
and employment ? Clearly they are not directly related to productive
efficiency, and one would do well to avoid the sin of commission in
using them as such. They do offer, however, a crude indicator of
1
J . R. N . S T O N E : " The Measurement of National Income a n d E x p e n diture ", in t h e Economic Journal, Sept. 1947, p . 297.
2
The Impact of the War on Civilian Consumption in the United
Kingdom,
the United States and Canada (H.M. Stationery Office, London, 1945),
Appendix XI.

116

METHODS OF LABOUR PRODUCTIVITY STATISTICS

industrial demand for manpower. We learn from such data, for example,
that a particular industry needs more or less man-hours to complete
a unit of product. Of course, it may not be exactly the same product
as that turned out in the past. The industry might have changed its
method of operation. But whatever the causes, the figures are apt
to indicate the change in the demand for labour. Of course, we do not
know whether workers of the same or different skill, aptitude or ability
are in demand now as heretofore. Nevertheless, the global figure gives
us some idea of industry's need for manpower.1
It is clear that this complex effect of all factors affecting labour
productivity should not be considered identical with change in
" efficiency ", or with any other single aspect of manufacturing
development.
If the reduction in labour requirements per unit of product, whether
calculated in numbers employed or in hours worked, can be held to
reflect a multitude of changes in the processes and means of production,
it cannot be regarded also as a measure of change in the efficiency,
amount or character of any one factor of production. Those who consider the figures cited to be indexes of labour efficiency, of quantity
of capital invested, or of improvement in capital equipment, are in
effect ascribing to one or another factor the result of changes in all
factors. For the same reason, one cannot reasonably focus attention
exclusively on any one motivating force behind the far-reaching changes
m e a s u r e d b y t h e d e c l i n e in l a b o u r p e r u n i t .

If w e s t r e s s m a n a g e m e n t

as their initiator we may underestimate the other factors that must
have stimulated management: for example, trade union efforts to
raise wage rates, and encouragement of standardisation of products
and regulation of hours and conditions of labour by governmental
agencies.2
It is thus clear t h a t over-all measures of labour productivity or
of unit labour requirements, while useful in a first stage, are
insufficient to throw light on the thousand and one questions which
arise in their interpretation.
For this reason, as L. Teper points out—
Research should be undertaken on the plant level of the different
industries in order to find out, in greater detail, what actually takes
place there over a period of time. How are changes in productive
efficiency brought about ? What is the influence of machinery, management, individual effort, flow of work, demand for goods, specifications
and variety of product, nature of raw materials, and so on ? . . .
We need also to get more information on the influences outside the
plant on its productive efficiency. To what extent do changes in the
demand for goods induce variations in the production methods ? What
is the effect of competitors' activities on changes within the plant ?
How rapid is the process of change as between the different plants ?
1

" This Thing called P r o d u c t i v i t y ", loc. cit.

S. F A B R I C A N T : op.

cit.,

pp.

26-27.

INTERPRETATION OF LABOUR PRODUCTIVITY DATA

117

What explains the survival of some of the outmoded methods of production ? What is the relation between capacity output and productive
efficiency ? How is productive efficiency affected by changes in the
nature of integration between the plant and other businesses ? What
is the effect of developments in the field of communications and transportation on the plant ?
The present interest in the question of productivity does not stem
so much from the controversy about the meaning of a particular
statistic, but from a universal desire to find the clue to a better standard
of living. If it is held that the answer lies in greater productive efficiency,
there is no question but that we must supplement our econometric
studies by investigations of causes, effects and by-products of such
changes. It is to this area that the attention of investigators of productive efficiency must turn. 1
Thus a correct interpretation of labour productivity data is
possible only when the data collected are accompanied with as
much information as possible on all the influential factors. In
this respect special enquiries into labour productivity are superior
to any other method: they permit the presentation of detailed
data according to many classifications, which can throw light on
the various factors affecting productivity; they may be combined
according to various formulae, and this method also allows for the
collection of non-numerical data which are extremely useful in
the evaluation of the figures.

Limitations
of the Existing
New Studies, op. cit.

Productivity

Measures

and

the Need

for

CHAPTER IX
GENERAL

CONCLUSIONS1

The general importance of labour productivity has been stressed
in the introduction to this report; productivity is a corner-stone
of the economy of the future, and since the end of the second
world war it has been in the foreground of the preoccupations of
most countries. In order to throw light on a number of economic
problems, such as the relation of technical progress to unemployment and employment, the possibilities of realising a high production, or the differences in real net national incomes per head
between countries, statistics should be developed to measure as
accurately as possible differences and changes in labour productivity
in individual processes, separate industries, and, as far as possible,
countries as a whole.
Throughout this report the numerous difficulties of defining and
measuring labour productivity and of interpreting the data have
been emphasised.
DEFINITIONS

It appears extremely difficult to give a detailed definition
of labour productivity covering all concepts to which the term is
applied; it seems preferable to retain the general notion of the
ratio of output to the corresponding input of labour and, whenever
figures are issued, to state clearly the methods of computation,
formulae and weights used. The computations will generally refer
to either (a) output per man, or its reciprocal, the number of
persons employed per unit of output; or (b) output per man-hour,
or its reciprocal, man-hours expended per unit of output. In the
latter case, the measures refer to the ratio of output to labour
actually contributed; in the former case, the measures concern the
ratio of output to the labour force. Both measures are useful;
1
These conclusions were presented to the Seventh International Conference of Labour Statisticians in the form of a draft resolution. The Conference
suggested that, in the course of the revision of the report before printing,
this resolution should be replaced by a chapter summarising the conclusions
of the report.

GENERAL CONCLUSIONS

119

they characterise different aspects of the relation of production
to labour.
Whatever the computations contemplated, it is indispensable to
make provision for obtaining data concerning output and labour in
comparable terms; this is particularly important when considering
computations made on the basis of production and labour figures
which are not collected at the same enquiry. Of course, whenever
possible, measures per man and per man-hour should be prepared
simultaneously, since the juxtaposition of these two series will
yield considerable information.
FORMULAE

Whenever labour productivity is measured for more than a
single plant, that is, for a group of undertakings or a group of
industries, problems concerning the formulae of combination will
arise. Two general concepts have been noted; it is important to
state clearly the concept on which the measurements are based.
The first concept is one in which the results are not affected by
changes over time or differences between countries or regions in
the composition of the outputs of the undertakings or industries
on which the measures are based ; this concept implies that labour
productivity is a characteristic of the plant in which it is measured.
The second concept is one in which the results may be affected by
changes over time or differences between countries or regions in the
composition of the outputs of the undertakings or industries on
which the measures are based; the use of this second concept is
implicit in computations where average labour productivity for a
certain production composite is compared with average labour
productivity for another production composite. It should be noted
that, when the formulae corresponding to the first concept are
modified so as to represent the comparison of a labour index to a
production index, this production index should be weighted with
unit labour requirements ; if this is not done, and if use is made of
substitute weights, the results correspond to those obtained under
the second concept, that is, the variations in the productivity
indices are influenced by variations in the production composites.
MAN-HOURS PER UNIT OF OUTPUT

Though the computation of man-hours per unit of output
appears at first sight to be merely the reciprocal of output per

120

METHODS OF LABOUR PRODUCTIVITY STATISTICS

man-hour, the use of man-hours per unit of output offers considerable advantages over the use of figures of output per man-hour.
The first advantage is that man-hours per unit of output are
expressed in the same unit, an hcur of work, whatever the output
considered, while output per man-hour is always expressed in
different units—shoes, machines, bulldozers, number of sheets
of glass, etc. The second and more important advantage is that
man-hours per unit of output have an additive character: it is
possible to compute the number of hours of work necessary to
perform the different phases of a production process and, therefore,
to compare from country to country, region to region or plant to
plant the labour productivity of undertakings of different degrees
of integration; for this purpose, it is merely necessary to isolate
in the measures those man-hours corresponding to the same
phases of a production process. This is not possible with measures
of output per man-hour.

COMPUTATIONS BASED ON NET OUTPUT

Because of the difficulty of combining the physical output of
entirely different goods, many comparisons of labour productivity,
and especially international comparisons, have been based on
the value in monetary terms of net output per head or per manhour. The danger of such computations lies in the fact that many

factors other than the one they are intended to measure intervene
in the computations. Though they have, within limits, the
advantage of discarding the difficulties due to the varying degrees
of integration, they nevertheless introduce in the computations
variations due to differences in wages, profits, selling prices, etc.
But, above all, they cannot be used in international comparisons
without the help of exchange rates. The determination of such
exchange rates involves an arbitrary and subjective element.
There is no basis for the assumption that official exchange rates
may be considered as the correct ones to be used, and there is
no sound basis for the determination of any other " exact "
exchange rate. It therefore appears preferable to confine international comparisons of labour productivity to measures based on
physical output per head or per man-hour. Unfortunately, this
may practically restrict such international comparisons to agriculture, manufacturing, mining and transport.

GENERAL CONCLUSIONS

DIVISION OF PRODUCTION INDICES BY EMPLOYMENT

121
INDICES

Indices of labour productivity have often been prepared in
the past by dividing current production indices by current employment or man-hour indices; these measures were made because
the two sets of indices were available and because such a division
seems to afford an approximate indication of the variation of
the factor studied; but they should be used with extreme caution.
Production and employment or man-hour indices are known to
have biases, which are additive in such operations. Changes in
labour productivity, with certain rare exceptions, are extremely
small in the short run, whether production and employment vary
greatly or little, and therefore small errors which may be negligible
for most purposes in production indices or in employment indices
are no longer negligible when considering indices of labour productivity obtained by dividing one by the other. The main
difficulty lies in the lack of comparability in most cases of the
two indices, since production indices are generally computed by
agencies entirely independent of those preparing employment
indices. Data are not collected simultaneously, or even from
the same undertakings. The scope of the indices is different,
as well as the concepts on which they are based, and it could be
held that the variations observed in the indices of labour productivity computed by these means are due much more to differences
in the methods of computation of the indices than to variations
in labour productivity itself.
However, when the production and employment (or man-hour)
indices are based on data collected simultaneously in each undertaking, and are prepared by the same agency or by closely coordinated agencies, or when separately collected data have been
submitted to careful correction, the comparability is improved.
These remarks do not apply to important variations of labour
productivity such as may be observed over a long period of 10 or
20 years, or in periods when special conditions prevail, as in
wartime, during which labour productivity may show sharp
increases or decreases.
P E R I O D S OF COMPUTATION

In this connection it should be noted that, whatever the method
of computation, labour productivity figures should not be based
on such short periods as would lead to erroneous conclusions

122

METHODS OF LABOUR PRODUCTIVITY STATISTICS

regarding the significance of apparent changes and trends. In
many cases the length of the production process is considerably
longer than a month, and sometimes longer than a year (as in
boat building, house building, etc.). While, with the use of
approximations and estimates, it may be possible in such cases
to estimate monthly production figures, such figures lack real
significance in labour productivity computations. It would appear
also that in most cases where seasonal variations in the production
process have to be considered labour productivity figures should
not refer to periods of less than a year. At any rate, changes in
labour productivity are so slow that computations for periods of
less than a year seldom yield useful information.

SPECIAL ENQUIRIES

It has been stressed throughout the present report that over-all
measures, while useful, are entirely insufficient to explain the
influence of the various factors affecting labour productivity.
It must be realised that the ultimate objective is to find out not
whether labour productivity has increased or decreased but why
such changes have occurred. Therefore, whenever possible, analyses
or special studies should be undertaken to explain and measure
the influence of the many and various factors affecting labour
productivity. For these purposes the method of special enquiries
into labour productivity should be recommended for serious
consideration. In the collection of national and international data
special enquiries have many advantages : they permit simultaneous
collection of data on production and labour of the same scope;
they allow the presentation of data according to the different
phases of the production process ; they provide analyses illustrating
the influence of various factors such as size, location of plant,
mechanisation, etc., and they also permit the collection of information concerning the techniques followed and any other factors
that may influence labour productivity. On the basis of data so
obtained from individual plants, computations can be made
according to whatever formula may be considered desirable.
Productivity for a group of undertakings, a group of industries
or for two or more countries can be computed on a fixed weight
base or on a changing weight base with the same data, with very
little additional work. In short, the method of special enquiries
into labour productivity seems to be the best method of collecting

GENERAL CONCLUSIONS

123

such data for the analysis of differences and variations from plant
to plant, industry to industry and country to country.
PUBLICATION

Finally, the publication of labour productivity data should not
be confined to the issue of bare figures, but should always be
accompanied by detailed descriptions of the methods of collecting
the data, the methods of computation used (including formulae
when necessary) and, as far as possible, by an interpretation of
the data in the light of the methods followed.

APPENDICES

APPENDIX I
ANALYSIS OF LABOUR PRODUCTIVITY FORMULAE
In Chapter V of the present report several approaches to labour
productivity or unit labour requirements indices have been indicated,
and the relations between these approaches have been reviewed. However, the exposition of the various ideas is facilitated when use is
made of mathematical notations. Furthermore, some statements require
a more formal demonstration. Such is the purpose of the present
appendix.
For a single product the number of man-hours expended per unit
of output, or unit labour requirement, may be expressed by:

where q is the output of the product, m the man-hours utilised in the
production of q, and I the number of man-hours per unit of output, or
the unit labour requirement. The productivity of labour for any
product, which is the reciprocal of the unit labour requirement, may be
expressed as follows:
g
P= J = m
where p is the labour productivity.
For reasons of convenience the following analysis is based on the
consideration of unit labour requirements; it refers to comparisons between the base period (noted o) and the current period (noted i), but
the argument would apply equally well to comparisons between two
regions o and i.
The following notations are used :
l0 represents unit labour requirements for one of the products at the
period o;
Zi, the unit labour requirements at the period ¿;
qx, a certain quantity of that product;
q0, the quantity of that product produced at the period o;
qi, the quantity oí the same product produced at the period i;
m0, the man-hours utilised in the production of q0 and
nu, the man-hours utilised in the production of qi.

125

APPENDICES

According to these definitions, for each product:
m
.
mi
lo = —0 , ti = — , and
Ço

rtio — l0qo, m = h gi .

First approach. In the first approach to the setting up of unit
labour requirements indices, as indicated in Chapter V, the total volume
of labour required to produce a given complex of goods in the period
i is compared to the total volume of labour required to produce the
same complex in the period o; the formula of the unit labour requirements index is then x :
"Zqxlo
The unit labour requirements index thus depends on the complex
considered. If the complex chosen is that formed by the output of
each product during the period o (qx = q0), then:
Xqxli _ S ^ l j __ Sgok
Zqxlo ~ Zqolo ~ Sffio ~

°Jo
SOT 0

(

This formula I can be expressed as the ratio of the total labour that
would have been spent to produce the base period complex of goods,
with the current unit labour requirements, to the total labour actually
spent in the base period.
If the complex chosen is that formed by the output of each product
during the period i (qx — qi), then:

This formula II can be expressed as the ratio of the total labour actually
spent to produce the current complex of goods to the total labour that
would have been spent to produce the same complex had the unit
labour requirements been those of the base period.
Formulae I and II differ in one major respect. When data are
obtained for several periods, if formula I has been used, comparisons
1
The correct writing of this formula, and similarly for the following ones,
would be:

3=1

1, 2, 3, . k, . n corresponding to the various goods forming the complex.
However, to avoid overloading the presentation, the formula will be written
in the simplified way as indicated.

126

METHODS OF LABOUR PRODUCTIVITY STATISTICS

are possible between any two years. Thus, if the period / is to be
compared to the period i the index should be:
S go h
Zqoh'

This is equal to the division of the index for the period / as compared
to period o by the index of the period i as compared to the same
period o:
S go lj
Sgp/o

S q0 lj

2 q0 li

'Zqok
SçcZo

If, however, formula II is used, the latter computation would not
furnish the desired result, since the ratio
Sgj/i

is not equal to
S q¡ lj
S qj lg
-

S gift
2 qi l0

Second approach.

In the second approach a weighted average of

each individual unit labour requirement index 7- is computed, the
lo

weights being a.

The formula will then be:
to

Sa "
The interpretation of this formula will depend on the weights a chosen.
When a = l0,

"ÊTo"'

that is, the ratio of the total unit labour requirements of the period i
for all components to the total unit labour requirements of the period 0.
When a — 1,
Xa

T0_

1 sfe
n

lo '

that is, the arithmetic mean of the individual unit labour requirements
indices.

127

APPENDICES

In this approach, the formula is reduced to formula I if the weights a
correspond to the labour expended for the output of each component
at the base period:
T.

S a 7-

S m0 7=

Em«

- (I)
{
''

The same approach leads to formula II if the weights adopted correspond
to the labour that would have been expended to produce the current
eriod output of each product with the unit labour requirements of the
ase period:

S a 7to

Sg-iío X 7-

„

ío

Zqrfo

^qi H _

.jj.

~ S^i/o ~

[ll)

Third approach. In the third approach, the average unit labour
requirements in the period ¿ are compared to the average unit labour
requirements in the period 0, the averages for each period being computed
as the ratio of the total labour expended to the total output. The
formula is then:
S mi
Stvji

Sm 0
Swo 0

where w represents the weights used in the computation of the total
outputs at the periods ¿ and 0. It must be noted that the weights
used to measure these two outputs should be the same; otherwise, if
weights w were used to measure the current output, and different
weights w' to measure the base period output, while, when both the
output of each good and the unit labour requirements remained
unchanged the index should be equal to 1, this would not be true, since
the formula would then become :
S nu
Zwqj
Sm0
S M/ q0

S m0
Swgp
S m0
S w' q0

_ Su/q 0
Ewq0

If the weights chosen to measure the outputs are w = k, or w = l0,
the resulting formula is:
S mi
SZi?¡

S » i Zi
SZi?i

Sm 0
SZi^o

S<7o¿o

SZiffo

Xqolo

128

METHODS OF LABOUR PRODUCTIVITY

STATISTICS

or
S¿Q<¡Tj
S Wo

SZo<7o

Zqih
2>l0qi
2<¡r0Zo
SZ 0 ?o

^q%lo

= (II).

Fourth approach. In the fourth approach, commonly used, the
variations between the two periods considered in the total labour
expended (or labour index) are compared with the variations in the
total production between the two periods (or production index):

Sm 0
~Zwqa

This formula is, in fact, the same as the previous one, and therefore is
also reduced to formula I or formula II when the weights chosen are
w = k or w = lo respectively.
According to this analysis the two formulae designated as I and II
can be expressed in quite different terms, depending on the approach
used.
Formula I corresponds to any one of the following expressions :
(a) the ratio of the total labour required to produce the production
composite of the base period with the unit labour requirements of the
present period to the total labour actually expended in the base period
to produce the same composite;
(b) the weighted average of the individual unit labour requirements
indices, the weights corresponding to the total labour expended for the
output of each product at the base period;
(c) the ratio of the average unit labour requirements of the current
period to those of the base period, the weights used to measure the
output of the two periods corresponding to the unit labour requirements of the current period;
(d) the ratio of the index of total labour expended in each period to
the index of total output, the weights used to measure the latter corresponding to the unit labour requirements of the current period.
Similarly, formula II corresponds to any one of the following
expressions :
(a) the ratio of the total labour actually expended to produce the
current complex of goods to the total labour required to produce the
same complex with the unit labour requirements of the base period ;
(b) the weighted average of the individual unit labour requirements
indices, the weights corresponding to the labour which would be necessary to produce the current output of each product with the base period
unit labour requirements;
(c) the ratio of the average unit labour requirements of the current period to those of the base period, the weights used to measure

129

APPENDICES

the output of the two periods corresponding to the unit labour requirements of the base period ;
(d) the ratio of the index of total labour expended in eac h period
to the index of total output, the weights used to measure the latter
corresponding to the unit labour requirements of the base period.
In the first two approaches, whatever the production composites
considered or the weights used, the unit labour requirements index is
equal to k when for each product k — kl0 (k being a constant), that
is, when unit labour requirements have varied in the same proportion for each product, undertaking or industry:
If it = klo for each component,
S QxH

Zi ([xklo

S Çx lo

k 2JÇX 'o

= *,
S qx ¿o

S Çx lo

and
- h
h_
Sa

klo

h
Sa

_ « Sa
"~Sä"

On the contrary, in the last two approaches, unit labour requirements
indices might differ from k even when k — kl0 for each component,
since in this case :
Sffít

Swt/i =
Sffip ~
Sffji

S/re¿
S£¿Z¡
s?¿z¡
IiMp _ Zgolo
Sç-oio _
'Lwqi ~
liWqj ~ 'Lwqj
'Zwqo
Ew>ç0

kZqilo

Sy 0 ¿ 0
"Lwqj
Swço

_ ,

2l0qi
Zloqo

Xwq0
Swffj'

This is equal to k only when
S ¿agi _ S(V<¡Ti
S lo q0

S w q0 '

which is true when for each component either w is proportional to k
( = klo), or when
wqi _ wq0
Hwqi

Tiwqo'

Thus, the first two approaches correspond to the first concept of a
labour productivity (or unit labour requirements) index, in which the
index remains unchanged when each individual productivity (or unit
labour requirements) index has remained unchanged; the last two
approaches correspond to the second concept, in which the labour
productivity (or unit labour requirements) index might vary even when
each individual labour productivity (or unit labour requirements) index
remains unchanged.
It may be shown that the replacement of the data necessary to
compute formula I or formula II by some substitute data corresponds
to a shift from the first concept to the second one. Formula I can

130

METHODS OF LABOUR PRODUCTIVITY STATISTICS

be written so as to be very similar to the mathematical expressions
of the third and fourth approaches:
S q0 li
Zçoio ~~

S q0 lo
Eçik
2<7o/i

ïfflo

Síigi
2 h q0

imilarly formula I I can be written:

S qi k
"Lqilo

Xqdi
S q0 lo
S qi l0
Xq0l0

Xmi
Sfilo
S/o^t
S lo q0

When so written, formulae I and II correspond to the ratio of a labour
index to an index of output, provided that the weights used in the
latter are the unit labour requirements of the current or the base period.
If substitute weights are used because of the lack of available data, the
results obtained do not correspond any longer to formula I or II and
in fact involve a shift from the first concept to the second one. It can
easily be shown that the formulae where weights w are not proportionate either to k or to l0 correspond to the introduction in the unit labour
requirements index of an additional factor directly related to the change
in the relative importance of each product, undertaking or industry in
the complex ; the following identities can easily be verified :
Sffii
Sfflo
*Zwqi
~Lwq0
SWi
Sfflo

Zwq,
~Lwq0

Sgok
Zq0lo

'Zqili , S(vg 0
SZ0<¡r0
S$t/o

Zwqo
S lo q0

Xkqi
2/o</i

Vk I wqi
| w \Lwq%

^ | lo ( wqj
\jv X'Zwqi

wq0 \
Zwqo)

wq0
I,wq0

The differences between the formulae corresponding to the third and
fourth approaches and formulae I and II are nil when:
(a) w = kk, or w = klo (k being a constant), for each component;
(b)

——ii- = —-3^- for each component.

It is on the basis of formulae very similar to those above that
S. Fabricant computed the relative influence of " actual " changes in
labour productivity and of changes due to the different composition of
outputs. 1 Computations for the theoretical but simple examples
proposed by D. Weintraub a will give a practical indication of the
differences in results that may be observed:
1
2

See pp. 65 et seq.
See p. 58.

131

APPENDICES

First

Approach.

(I) = Cá»jí = 0.963 ;
£§0Z0
'
Second

(II) = - 1 ^ = 0.964.
Sgi¿0

Approach.
La

If'a = 1,

-^r1

= - S r = 0.961; if a = Z0, - ^ r - 2 = v T = 0- 9 6 3 -

TAW a«<Z Fourth Approaches.
The weights for each product as used by D. Weintraub are logically
equal to 1, since the plants are manufacturing an identical product:

nu
(where w = 1) = 1.026.

s wqi
Zwqo

Finally, according to t h e relations shown above,

S g-,o

0.062.

In this case the differences between the various formulae corresponding to the first concept are negligible; on the contrary, differences
between those formulae representing different concepts are considerable
(6 per cent.), so much so t h a t they outweigh the actual changes in the
unit labour requirements t h a t take place in each plant (decreases of 4
and 3 per cent, respectively).

APPENDIX II
RESOLUTION ADOPTED BY THE
SEVENTH INTERNATIONAL CONFERENCE OF LABOUR
STATISTICIANS (GENEVA, OCTOBER 1949)
The Seventh International Conference of Labour Statisticians,
Having been convened at Geneva by the Governing Body of the
International Labour Office, and having met from 26 September to
8 October 1949, and
Recognising the importance of labour productivity in relation to
social and economic welfare, and therefore the importance of developing
accurate methods for measuring the productivity of labour for use in
national as well as in international comparisons and analyses, and
Noting the especial need for international agreement on the elements
of the data to be collected as well as on the methods by which labour
productivity is to be measured,
Believing that the report prepared by the International Labour
Office provides an extremely valuable analysis of the problems encountered in the development and interpretation of statistical measures of
labour productivity,
Requests the Governing Body of the International Labour Office:
1. To direct the International Labour Office—
(a) to communicate this report, together with an account of the discussions on the subject at the present Conference, to Governments
as soon as possible, with a request for their comments;
(b) to continue its studies of the statistics on labour productivity and
in particular to collect and publish data on labour productivity in
the different countries;
(c) to prepare recommendations on standard international definitions
of wage earners, salaried employees, production workers, and on
direct and indirect labour for use in labour productivity statistics;
(d) to undertake studies of the best methods of measuring labour
productivity in non-manufacturing industries;
(e) to examine the possibility of obtaining comparable data in the
different countries on labour productivity in important basic
industries by means of special enquiries, intended to provide
breakdowns according to the different phases of the production
process, analyses illustrating the influence of various factors, such
as size, location of the plant, etc. ; and the collection of information
on the techniques followed and on other factors that may influence
labour productivity;
(f) to collaborate with the United Nations and with other specialised
agencies in their work on the development of production and
productivity statistics.

APPENDICES

133

2. To place the subject of statistics of labour productivity on the
agenda of the next session of the International Conference of Labour
Statisticians.
3. To give consideration to the desirability of convening, in advance
of that Conference, a small meeting of experts on productivity statistics
to assist the Office in formulating the proposals to be placed before
the Conference.

APPENDIX III
SELECTED REFERENCES
BARN A, T.: " T h e Productivity of Labour: Its Concept and
Measurement ", in the Bulletin of the Oxford University Institute of
Statistics, Vol. VIII, No. 7, July 1946.
BARTON, G. T., and COOPER, M. R.: " Relation of Agricultural
Production to Inputs ", in the Review of Economics and Statistics
(Cambridge, Massachusetts), Vol. XXX, 1948.
BERGER, H., and LANDSBERG, H. H.: American Agriculture, 1899-

1939, A Study of Output, Employment and Productivity, National Bureau
of Economic Research, New York, 1942.
BRITISH MINISTRY OF PRODUCTION: Report of the Cotton Textile
Mission to the U.S.A., March-April 1944, London, 1944.
CLAGUE, E.: Productivity, Employment and Living
Standards,
Conference on Productivity, Milwaukee, 4 June 1949.
COMMISSARIAT GÉNÉRAL DU PLAN: Programme français pour Vaccroissement de la productivité, Paris, Feb. 1949.
CONGRESS OF INDUSTRIAL ORGANIZATIONS: An Evaluation of Bureau
of Labor Statistical Data on Productivity, 5 Jan. 1946.
DEURINCK, G.: "Mesure de la productivité", in Organisation
Scientifique (Brussels), Dec. 1948.
EVANS, D., and SIEGEL, I.: " The Meaning of Productivity Indexes ",
in the Journal of the American Statistical Association, Vol. XXXVII,
Mar. 1942.
EVANS, D.: " Recent Productivity Trends and their Implications ",
in the Journal of the American Statistical Association, Vol. XLII,
June 1947.
FABRICANT, S.: Employment in Manufacturing Industries, 1899-1939,
National Bureau of Economic Research, New York, 1942.
FLUX, A. W.: " Industrial Productivity in Great Britain and the
United States ", in the Quarterly Journal of Economics (Cambridge,
Massachusetts), Vol. XLVIII, Nov. 1933.
me
FOURASTIÉ, J. : Le grand espoir du XX
siècle, Presses Universitaires de France, Paris, 1949.
HANSEN, H. E.: "Productivity on the Increase", in Survey of
Business Practices, National Industrial Conference Board, New York,
June 1948.
JEWKES, J. : " Is Britain's Industry Inefficient ? ", in The Manchester
School, Vol. XIV, No. 1, Jan. 1946.
JEROME, H.: Mechanisation in Industry, National Bureau of Economic Research, 1934.

135

APPENDICES

KÄLLSTRÖM, V. : " Efficiency of Industrial Production in Sweden ",
in Skandinaviska Banken (quarterly review), Vol. XXVIII, No. 1,
Jan. 1947.
KURLAT, A.: Notas para una encuesta continental sobre productividad
obrera, Consejo interamericano de comercio y producción, Montevideo, 1946.
MARKUS, B. L. : " The Stakhanov Movement and the Increased
Productivity of Labour in the U.S.S.R.", in the International Labour
Review, Vol. XXXIV, July-Dec. 1936.
Ou, PAO-SAN, and WANG, FOH-SHEN: " Industrial Production and
Employment in Pre-War China ", in the Economic Journal, Vol. LVI,
1946.
ROSTAS, L. : Comparative Productivity in British and American
Industry, Occasional Papers, No. XIII, National Institute of Economic
and Social Research, Cambridge University Press, 1948.
— " Productivity of Labour in British, American and German
Agriculture ", in London and Cambridge Economic Service, Vol. XXIV,
Bulletin No. 3.
SELEKMAN, B. M. and S. K. : " Productivity and Labour Relations ",
in Harvard Business Review, Vol. XXVII, No. 3, May 1949.
SIEGEL, I. H.: " T h e Concept of Productive Activity", in t h e
Journal of the American Statistical Association, Vol. XXXIX, Nos. 225228, 1944.
TEPER, L.: "This Thing called Productivity", in The American
Federationist, Nov. 1948.
— Limitations of the Existing Productivity Measures and the
Need for New Studies, Conference on Productivity, Washington, 1946.
UNITED STATES BUREAU OF LABOR STATISTICS:

Productivity and

Unit Labor Cost in Selected Manufacturing Industries, 1919-1940,
mimeographed, Feb. 1942.
— Productivity and Unit Labor Cost in Selected Manufacturing
Industries, 1939-1945, mimeographed, May 1946.
— Major Sources of Productivity Information, Washington, June
1949.
— Selected References on Productivity, Washington, Oct. 1946.
— Summary of Proceedings of the Conference on Productivity,
October 28-29, 1946, Bulletin No. 913.
UNITED

STATES

NATIONAL

INDUSTRIAL

CONFERENCE

BOARD:

Measuring Labor's Productivity, Feb. 1946.
UNITED STATES WORKS PROGRESS ADMINISTRATION, NATIONAL
RESEARCH PROJECT: Production, Employment and Productivity in
59 Manufacturing Industries, 1919-1936 (MAGDOFF, H.; SIEGEL, I. H.;

and DAVIS, M. B.), May 1929, 3 parts.
WEINTRAUB, D.: " Some Measures of Changing Labor Productivity
and their Use in Economic Analysis ", in the Journal of the American
Statistical Association, Vol. XXXIII, Mar. 1938.
— and POSNER, H. L.: Unemployment and Increasing Productivity,
Works Progress Administration, Technical Resources Committee, Mar.
1937.

136

METHODS OF LABOUR PRODUCTIVITY

STATISTICS

WILSON, R.: Facts and Fancies on Productivity, The Economic
Society of Australia and New Zealand, 1947.
WRIGHT, C. D. : Hand and Machine Labor, United States Department
of Labor, 13th Annual Report, 1898.
WUBNIG, A. : " The Measurement of the Technological Factors in
Labor Productivity ", in the Journal of the American Statistical
Association, Vol. XXXIV, June 1939.

See also various articles in the Monthly Labor Review, United
States Bureau of Labor Statistics, Washington.