Reprinted for private circulation from
The American Journal of Sociology, Vol. XLII, No. 4, January, 1937
. PRINTED IN THE U.S.A.
Am
ANOTHER CENSUS OF UNEMPLOYMENT?
HOWARD B. MYERS AND JOHN N. WEBB
ABSTRACT
In view of the numerous demands for an unemployment census, an analyzing of the
possibilities and limitations of the census method is called for. As indicated by the 1930
experience, it is doubtful whether a census can provide a satisfactory answer even to the
relatively simple question of how many persons are out of work on a given date. Proper
measurement of unemployment duration and accurate determination of the occupation
and industry to which a worker is properly assigned require more detailed information
than it is practicable to secure through the census method. The method is of little
assistance in answering the important and complex questions of what proportion of the
unemployed population is actually employable, and what work the employables are
best fitted to do. The inability of the census method to provide the foregoing reasonable
minimum suggests that attention should be focused on the development of current un¬
employment statistics through a national system of unemployment insurance and em¬
ployment exchanges, rather than on further efforts to report a highly dynamic phenome¬
non through the device of another unemployment census.
The recognized need for more information concerning employ¬
ment and unemployment has led to numerous demands for an un¬
employment census. To a number of commentators in the daily
press the failure to count the jobless during the last three years is
inexplicable. This point of view is clearly stated in the following
excerpt:
Why the Administration steadfastly refused during these last three years to
take a census of the jobless will rank, in our opinion, as one of the minor mys¬
teries of history. Politics aside, it seems utterly irrational that throughout the
greatest period of unemployment that this or any country has known no de¬
pendable figures on the subject have been collected.1
A census of unemployment seemed certain to be taken in 1934, in
1935, and again in 1936, but each time failed to materialize. There is
no doubt but that demands for a census will continue to be heard,
particularly when the next Congress is confronted with the problem
of unemployment relief appropriations. It seems pertinent, there¬
fore, to inquire whether a census would assist materially in dealing
with relief and unemployment problems, as suggested by its pro¬
ponents.
In dealing with this question it may be helpful to indicate briefly
1 Editorial, "How Many Unemployed?" Detroit News, August 1, 1936.
S2i
522 THE AMERICAN JOURNAL OF SOCIOLOGY
some of the more essential facts needed and then to consider whether
a census is likely to supply these facts. An intelligent attack upon
the unemployment problem would seem to require current informa¬
tion on at least the following points: (i) How many are unemployed?
(2) How long have they been unemployed? (3) From what occupa¬
tions and industries do they come? (4) What proportion is employ¬
able? (5) What work are the employable persons fitted to do? These
merely include some of the simpler items on which we need infor¬
mation. More complex matters, such as the proximate causes of loss
of employment, the personal and social effects of unemployment, the
effectiveness of various methods for reducing unemployment or miti¬
gating its consequences, and the possibilities of vocational training
and rehabilitation, also need to be considered. The five simpler
items listed above, however, when supplemented by data on per¬
sonal and family characteristics (age, sex, marital status, family
composition, etc.) would state the problem numerically, furnish data
which are essential to the most effective planning of unemployment
relief, and provide some of the facts which are basic to any con¬
sideration of a better adjustment of the available labor supply to
the organization of business and industry.
How far can a census go in answering these questions?—It is ob¬
vious that the census technique cannot deal effectively with the
more complex matters which have been mentioned. Its fact-gath¬
ering possibilities are sharply restricted even on the simpler items,
owing largely to limitations inherent in the census method. The
census technique is necessarily limited to simple, objective inquiries;
the size of the undertaking requires that a considerable period of
time elapse before the data become available; and the cost of col¬
lection and tabulation of data on such a vast scale makes the publi¬
cation of census information impracticable except at infrequent in¬
tervals.
To illustrate the difficulties, it is instructive to examine the results
of the Unemployment Census taken in conjunction with the decen¬
nial Census of 1930. It might be expected that answers to the first
three of the questions listed above would be available directly from
the tabulations of the data scheduled, and these answers, in turn,
might throw at least some light on the last two questions. In fact,
ANOTHER CENSUS OF UNEMPLOYMENT?
523
however, the 1930 Census provided a satisfactory answer to none of
these questions. From the time the first Census returns were pub¬
lished they were challenged as an understatement of the number of
unemployed workers. This uncertainty of coverage, of course,
brought into question all of the data derived from the unemploy¬
ment schedule. Moreover, by the time the data were available they
were of little value in answering the very practical questions of how
many persons were out of work at the time and what should be done
about unemployment.
Thus the results of the Unemployment Census of 1930 were
criticized on two important points: the validity of the count and
the time required to make the data available. There are sound rea¬
sons for believing that the results of another unemployment census
will be open to the same criticisms despite the experience gained in
1930.
The fault lies not with the census method itself, nor with the
persons responsible for its direction, but with the application of the
method to the measurement of phenomena for which it is ill adapted.
The census method has been most useful when applied to the meas¬
urement of population characteristics which change slowly from year
to year. A report at ten-year intervals on age, sex, color, nativity,
and similar slowly changing characteristics of the population has in
the past largely met the need for this type of information, although
even here there has been a growing demand for reports at shorter
intervals of time.
In contrast, no well-informed person would argue that vital sta¬
tistics, particularly mortality statistics, at ten- or even at five-year
intervals would be adequate to chart the changes in the human
resources of the nation even though births and deaths can be pre¬
dicted with much greater accuracy than can employment and its
corollary, unemployment. The development of current reports of
births and deaths illustrates the need for a method of reporting the
dynamic aspects of constantly changing phenomena. This need is
obviously even greater in the case of unemployment, in view of its
extremely rapid fluctuations and its changing characteristics at dif¬
ferent periods within a single year. It would seem, therefore, that
the census method of securing a cross-section view of unemploy-
524
THE AMERICAN JOURNAL OF SOCIOLOGY
ment at widely spaced intervals of time must be judged inadequate
because of its inability to measure intervening change.
Another equally important aspect wherein the census methods
fail to provide adequate information on unemployment has been
touched upon in the preceding discussion. The census method when
applied to the national scene must be restricted to the measurement
of a limited number of simple and objective facts. The method re¬
quires the use of a large number of enumerators during a brief
period of time; these persons are, in the main, untrained in the
technique of gathering economic and social data. For this reason
census schedules must be simple of structure and must require a
minimum of judgment on the part of enumerators if the results are
to be worth the cost of tabulation.2
Unfortunately, unemployment is a complex phenomenon. It can¬
not be reported in the simple and objective terms applicable to age,
race, and sex. The results of the 1930 Census of Unemployment
furnish ample evidence of the difficulties which arise when an at¬
tempt is made to report a complex situation through the use of a
method that depends for its success upon simplicity of enumeration.
Since the 1930 Census failed to provide a satisfactory answer to the
least difficult of the questions it set out to answer—How many
workers were unemployed on the Census date?—it is hardly fair to
expect that it, or another national census, would provide adequate
information on the more difficult questions of duration of unem¬
ployment, employability, and the occupations in which unemployed
workers are best fitted to engage.
Apparently the complexity of reporting unemployment, and the
consequent weaknesses of the census method as applied to this
problem, are not fully recognized by most proponents of another
unemployment census. It may be well, therefore, to consider briefly
some of the simpler questions proposed above as a minimum require¬
ment of adequate statistics on unemployment.
How many are unemployed?—Employment and unemployment
are complementary states into which the total labor supply is di¬
vided at the time of observation. Neither can be defined for purposes
of measurement without at the same time defining, positively or
2 See Leon E. Truesdell, "Methods Involved in the Federal Census of Population,"
in Methods in Social Science, ed. Stuart A. Rice (Chicago, 1931).
ANOTHER CENSUS OF UNEMPLOYMENT? 525
negatively, its complement. Thus definition becomes the axis upon
which the problems of measurement turn. A workable definition of
employment must include the activities of the factory employee
with regular hours and pay, the playwright who writes much and
sells little, the youth who works regularly on his father's farm with¬
out wages, the farmer or business man who works all year at a loss,
the lawyer with an office and no clients, the absentee corporation
president who draws a salary but does no work, and others for whom
no single test of time, income, or specific activity is adequate to
determine the fact of employment. Unemployment is the absence of
such activity on the part of persons relying upon the results of their
efforts for full or partial support. In addition, a satisfactory defi¬
nition of unemployment must differentiate between those in and
those out of the labor market. It would seem that the factors which
must be taken into account in a full and satisfactory definition of
employment and unemployment are too complex to admit the
simple statements appropriate to a national census.
At first thought this statement seems to be contradicted by the
section on employment in the instructions to enumerators3 for the
1930 Census. There (p. 38) employment was defined with apparent
simplicity as "whether actually at work yesterday." In special cases
"yesterday" was to be interpreted as the last regular working day
preceding the enumerator's visit. This seemingly simple and prac¬
tical definition was actually not so simple as it appears. It left un¬
settled a number of problems related to the question of when a
person is "at work"; in addition the fact of employment was to be
determined only for persons defined elsewhere in the instructions as
"gainful workers." In effect this shifted the burden of definition
from employment, which would have been difficult to define for
census purposes, to gainful workers, which was an established con¬
cept in the Census Bureau procedure.4
3 Instructions to Enumerators, Population and Agriculture: Fifteenth Census of the
United States (rev. ed.; Bureau of the Census, Washington, D.C., 1930).
< However, the Bureau of the Census does not appear to be satisfied with this defi¬
nition. In the article cited earlier, Dr. Truesdell stated: "The gainful employment
classification is dependent in considerable measure on another artificial or 'manu¬
factured' definition. The borderline along the edge of gainful employment is rather
indefinite, and even the definition given to enumerators is not as specific as one might
wish." It should be noted that this article is dated October 30, 1928.
526
THE AMERICAN JOURNAL OF SOCIOLOGY
This proved to be an unhappy solution of the difficulty presented
by the need for a comprehensive definition of employment. It was
not that the definition of a gainful worker was inaccurate, but that
the definition was made to serve a purpose wherein it was inade¬
quate. This may be seen from the description of a gainful worker.5
249 [col. s, Unemployment Schedule], Does this person usually work at a
gainful occupation?—Write "Yes" if the person is usually employed at any occu¬
pation yielding an income of any amount. This applies to part-time workers
even though they work only a few hours a day or only a day or two a week
250. Persons will be found who have been long unemployed because of change
in industry, the introduction of machines, or the decline of production in certain
lines. If able and willing to do work of any kind, these persons should be re¬
turned as usually working at a gainful occupation, without regard to the length
of the period of idleness, provided they still expect to find employment and re¬
sume work.
The inclusion of the qualification "usually" requires either a judg¬
ment on the part of the enumerator or an opinion from the person
enumerated; omit this qualification throughout and the difficulties
inherent in the definition are apparent.
It is doubtful if this description was adequate to the task of
distinguishing within the total population of 1930 those persons to
whom the simple question, "Were you at work yesterday?" was
applied for the purpose of establishing the primary distinction be¬
tween the employed and the unemployed. It seems likely that this
description permitted the inclusion of persons who were not prop¬
erly a part of the total labor force simply because they had a usual
occupation when work was available. It is clear that it excluded
others—recent additions to the labor force—simply because they
had never previously been employed. Certainly this indirect method
of establishing the basis for determining employment and unem¬
ployment left many troubling questions to be answered when the
jobless were grouped into Classes A to G according to their employ¬
ment status and reasons for idleness.
It seems evident that an attempt to count the number of em¬
ployed and unemployed workers without first having defined what
meaning was to be attached to these terms was to invite criticism
s Instructions to Enumerators, op. cit., p. 43.
ANOTHER CENSUS OF UNEMPLOYMENT? 527
of the results. Moreover, the group to be included in an unemploy¬
ment count should have been limited to those actually in the labor
market at the time of the census, without reference to experience,
rather than to those who were "usually employed." Considering the
experience of the Census Bureau officials and their demonstrated
success in obtaining other types of population data, the criticisms
of the 1930 unemployment returns must logically be attributed to
the shortcomings of the census method.
How long have they been unemployed?—Next in importance to the
number of unemployed persons is the duration of their unemploy¬
ment. Without this information the social and economic effects of
unemployment cannot properly be appraised. Short periods of un¬
employment may have very little significance: e.g., unemployment
resulting from transfer from one job to another, vacations without
pay, periods of inventory, the breakdown of machinery. As the du¬
ration of unemployment increases beyond these short periods, the
effect varies with the earning capacity of the worker. Unemploy¬
ment for a period of several months frequently forces workers at
the lower wage levels to apply for public assistance, whereas the
same period of idleness is likely to be tided over by means of reserves
accumulated by workers at the higher wage levels. In addition to
the importance of duration data in the planning and administration
of public relief or unemployment benefits, there is its importance in
determining the effects of prolonged idleness on skill and thereby
upon employability.
Granting the importance of accurate information on duration
of unemployment, there arises the question of how it shall be meas¬
ured. From the last job of any kind, provided that job lasted more
than one day, as in the Census of 1930? From the last job at the
usual occupation, no matter how long held? From the last regular
job at any occupation? In short, how shall a job be defined as to
nature and length for purposes of measuring duration of unemploy¬
ment?
The difficulty of obtaining satisfactory answers to these questions
can be demonstrated from the employment record of a worker inter-
528 THE AMERICAN JOURNAL OF SOCIOLOGY
viewed in one of the studies made by the Division of Social Research,
Works Progress Administration:
Occupation Industry From To
Machinist Laundry ma- January, 1930 March, 1931
chinery (beginning of
record)
Unemployed Lived on savings April, 1931 May, 1933
Laborer Fruit farm June, 1933 August, 1933
Unemployed On relief September, 1933 May, 1934
Machine operator.... Basket factory June, 1934 August, 1934
Unemployed On relief September, 1934 January, 1935
(end of record)
In this case, which is by no means an extreme one, how shall the
duration of unemployment be measured, and from what occupation
and industry was the worker unemployed? He was employed as a
machinist when the record began, and this occupation was indicated
in the proper space as his usual occupation. If duration of unem¬
ployment is measured from the last job at his usual occupation, this
worker had been unemployed for three years and nine months when
the record was made. If duration of unemployment is measured
from the last job of any kind, this worker had been unemployed for
four months when the record was made.
Moreover, if unemployment is measured from the last employ¬
ment at usual occupation, the worker is properly returned as a
skilled worker and is so listed in the inventory of available labor
supply which is an important by-product of unemployment surveys.
If, on the contrary, unemployment is measured from his last job,
this worker is counted among the semiskilled workers.
It may be instructive to take another illustration, again from
actual records, of situations that present considerable difficulty:
Occupation Industry From To
Student College January, 1930 June, 1931
(beginning of
record)
Unemployed Dependent July, 1931 August, 1932
(but seeking work)
Student College September, 1932 August, 1933
Teacher Public school September, 1933 March, 1934
Laborer C.C.C. camp April, 1934 September, 1934
Unemployed On relief October, 1934 January, 1935
(end of record)
ANOTHER CENSUS OF UNEMPLOYMENT? 529
This is a case of a worker entering the labor market during the
depression. Not only is it difficult to determine the duration of
unemployment, but it is also difficult to determine what occupation
and industry is to be charged.
Additional examples of cases wherein duration of unemployment,
usual occupation, and industry present difficult problems of defini¬
tion could be multiplied at will from the unemployment relief
records. A few of the more common are: women entering the labor
market after a long period of absence because of the insufficient
earnings of the head of the household; seasonal, migratory, casual,
and part-time workers; gainful workers leaving the labor market
voluntarily or otherwise; newcomers in the labor market who have
never had gainful employment or who have been employed only on
work-relief jobs.
Because of these and other difficulties it seems evident that
duration of unemployment cannot be measured accurately from a
single entry of last employment. At least two entries are required:
the last job at the usual occupation and the last job of any kind.
The inclusion of a time qualification on the length of these last
employments is so important that a third entry is probably essential
in order to distinguish the last regular employment from makeshift
or stop-gap jobs.
In a sense, an increase in the number of job entries needed for
measurement of duration of unemployment is a compromise between
the inadequate single entry and a complete job record for a period
of several years preceding the inquiry. It is only from job histories
that the real significance of employment and unemployment can
be obtained for use in planning how to relieve and reduce the effects
of enforced idleness. However, job histories are virtually impossible
in large-scale census enumerations, and even addition of one or two
questions concerning last jobs, as suggested above, would create
difficulties in securing uniform returns.
What occupations and industries?—Adequate unemployment sta¬
tistics should provide an occupational inventory of the surplus labor
supply, together with an indication of the industries responsible for
this surplus. Correct occupational information on the unemployed
is particularly important because of the commitment in this country
53°
THE AMERICAN JOURNAL OF SOCIOLOGY
to work relief in place of direct relief. If work projects are to be
most useful to both the worker and to the community, the work
provided must take account of the available skills. This is particu¬
larly important in the long-range planning of public projects which
will serve to cushion the downward swing in employment and keep
the surplus labor force occupied productively until needed again in
private industry.
Without accurate occupational data work projects are likely to
be initiated which force a downward shift in occupational skills.
For instance, too many road construction projects in an area may
readily force skilled manual and white-collar workers to accept as¬
signments which are better suited to the unskilled. This is counter
to the real objective of public work projects, which aim to preserve
or develop skills through use of socially desirable activities.
A worker's occupation is less subject to change than is his em¬
ployment status (i.e., employed or unemployed). Largely for this
reason the census method can be used to good effect in providing
an inventory of available skills, provided care is exercised in deter¬
mining the occupations reported. It should be obvious from the
two examples of relief cases used above that an occupational inven¬
tory based upon an arbitrarily determined single entry, particularly
if that entry were the last job held during a depression period,
would result in an inaccurate account of occupational skills for use
on work projects or for reporting to private industry the nature of
the surplus labor supply. An accurate occupational inventory of
available workers cannot be taken unless some provision is made for
distinguishing stop-gap and made work from regular employment
or, what amounts to much the same thing but is more difficult,
unless the definition of usual occupation is better than any here¬
tofore available.
The importance of correct industrial returns, as distinct from but
accompanying the occupational entries, lies in the information they
provide on the nature of unemployment. Since many occupations
are common to more than one industry, it is frequently difficult to
determine whether a surplus of one occupational group represents a
short- or long-term change in the industrial structure. Moreover,
when duration of unemployment is classified by industry there is
ANOTHER CENSUS OF UNEMPLOYMENT? 531
provided valuable information both on the incidence of unem¬
ployment and on the probability that these industries will reabsorb
the existing labor surplus.
What proportion of the unemployed is employable?—Most studies of
unemployment do not attempt to report the employability of those
out of work, although this should be one of the most important
inferences to be drawn from these studies. It is generally assumed
that the definition of a gainful worker as a person who is usually
employed at an occupation yielding an income is sufficient indication
of employability. This assumption is incorrect in many instances.
The fact that a person was usually employed prior to the loss of his
last job is no basis for assuming that he can or will return to gainful
work as employment opportunities increase. In many cases the loss
of the job that initiated the period of unemployment under observa¬
tion represented, in fact, the act of discarding of a worker because
he was no longer productively employable according to the stand¬
ards of that particular industry. In many, if not most, instances the
worker does not know why he was separated from employment.
Certainly this is true of those who are discharged or "laid off" be¬
cause they were marginal or obsolete labor in the judgment of the
employer. This fact indicates that the employability of the unem¬
ployed cannot be determined with any confidence from the worker's
own opinion of his chances of returning to his former job.
The most frequent device for determining employability in un¬
employment is that of ability and willingness to work. Ability is
vaguely defined as freedom from physical or mental handicaps,
including age—"too young or too old"—which would render work
impossible. Willingness to work is variously defined as actively seek¬
ing employment, expressed desire for a job, registration with an
employment agency, etc. In so far as ability and expressed willing¬
ness to work actually report employability, it does so from the point
of view of the worker rather than from that of the employer. An
unskilled worker of fifty years may be without a physical flaw, sound
of mind, and actively looking for work, and still be unemployable
in the judgment of the employment office that released him, because
of the competition of younger men.
One of the arguments advanced in support of a work-relief pro-
532 THE AMERICAN JOURNAL OF SOCIOLOGY
gram is that it preserves skills and work habits; i.e., it helps to
retain employability against a return of the demand for these work¬
ers in private industry. It seems extremely important that there be
some practical and effective measure of employability in terms of
ability to return to private employment. Obviously there is no single
objective measure of employability. However, there are objective
and measurable factors which provide a basis for fairly sound infer¬
ences concerning the general question of employability.
Chief among these factors are age, physical condition, duration of
unemployment, usual occupation and industry, and continuity and
duration of work experience. In combination, these data promise to
provide a sound basis for estimating the employability of the unem¬
ployed. For instance, a study6 of urban workers receiving unemploy¬
ment relief shows that of persons leaving the relief rolls because of
private employment those having a short duration of unemployment
from the usual occupation go first; and that there is a marked inverse
relationship between age and duration of unemployment among
unskilled and semiskilled workers. Moreover, it is known that the
"hard core" of unemployment, representing the least employable
part of the idle population, has an overrepresentation of workers
from specific occupational and industrial groups. Far too little has
been done thus far in examining the relation between the character¬
istics of the unemployed and their work histories for the purpose of
determining the likelihood of return to private employment.
What work are the employables fitted to do?—The preceding discus¬
sion suggests the need for determining the usual occupations at
which unemployed workers are employable. In addition, it is im¬
portant to know whether the occupation followed in the past is the
one which the worker is best fitted to follow in the future. The
utility of this information is obvious, both for the planning and
operation of a public-work program and for the replacement of
jobless workers in private industry. The difficulties of securing the
information are also obvious; the question is one which requires
6 Palmer and Wood, Urban Workers on Relief, a forthcoming monograph of the Divi¬
sion of Social Research, Works Progress Administration. Unpublished data in the files
of the Division of Social Research show a consistent relationship between relief turnover
and duration of unemployment.
ANOTHER CENSUS OF UNEMPLOYMENT? 533
careful and detailed analysis of such factors as previous work his¬
tory, age, physical condition, and the attitude and inclinations of
the worker. At the same time, the type of vocational analysis sug¬
gested is essential to the maintenance of a properly functioning
labor market, and contribution to such analysis is clearly one of the
most important end results to be desired from studies in the unem¬
ployment field.
Summary and conclusions.—The problem of collecting the infor¬
mation needed to deal effectively with unemployment is by no
means as simple as it appears to most persons pressing for an unem¬
ployment census. The census method of gathering data on a national
scale is most useful when applied to phenomena which can be defined
in simple, objective terms, and which change slowly with time. It
is not well adapted to dealing with the complex and rapidly chang¬
ing situations presented by unemployment. Because of limitations
inherent in the method, the census technique fails, in large part, to
provide a reasonable minimum of reliable and timely information
on the increasingly important problem of unemployment.
Recognition that the census method has serious limitations when
applied to the collection of essential data on unemployment should
emphasize the need for a national system of current unemployment
statistics. Eventually such a system must be developed in connec¬
tion with the administration of unemployment insurance and some
form of national employment exchanges. For the present it seems
that more would be gained from a concerted effort to further this
eventuality than from an attempt to overcome the difficulties of the
census method.
Division of Social Research
Works Progress Administration