\\ Va^ ^)^;rll!(!n'
» • i DVi-o-lfC * * t
I
Irrcnfi Division of R^esea*^
mi i
the Journal of the American Statistical Asswetjuiim,
OOCUMENTb March, 1939, Vol. 34, pp. 49-50. '^\oN«
ROOM ^
.rJL
CONCEPTS USED IN UNEMPLOYMENT "|94Q
By John N. Webb
Works Progress Adminisiration b a
So much has been said and written about the difficulty of defining
unemployment that it may properly be asked "How have the un¬
employment surveys and censuses of recent years been able to measure
what is so hard to define?" A partial answer to this question is that a
formal definition of unemployment is not an essential element in
unemployment surveys. What is essential, however, is that the con¬
cepts which determine the final results provide a logical and compre¬
hensive substitute for a formal definition of unemployment.
Not only is there an increasing demand for accurate measurement
of unemployment, but more is expected of the measurements that are
made. A years ago a count of the number of persons out of work
was considered an important contribution; today, a count is only one
of the facts that must be determined. For example, public assistance
programs want information on unemployment by families, cases, or
households; moreover, these programs frequently require that a dis¬
tinction be made between the unemployment of primary and secondary
workers in the family group. New terms such as "employables,"
"unemployables," "reentrants," and "partly unemployed" attest the
rapid development of specific types of information that are desired in
addition to the total number of unemployed persons.
In view of the greater demands made of unemployment surveys
today, it is appropriate to examine the adequacy of the basic concepts
that have been used in surveys made during recent years. Periodic
evaluation of past procedures and constant experimentation with new
methods are essential to improved measures of unemployment.
More than 40 surveys of unemployment1 made between the years
1929 and 1937 provide the factual background for the discussion of
concepts presented here. Although these studies differ considerably as
to coverage and method, they have one basic procedure in common.
This procedure is the method by which the total population in the
civil division studied—that is, the city, county, State, or nation—is
divided into two groups, one to be included in the unemployment in¬
quiry and the other to be excluded.
The use of a common procedure may not be immediately apparent
from an examination of two surveys, one of which, for example, includes
* a paper presented in summary at the One-hundredth Annual Meeting of the American Statis¬
tical Association, Detroit, December 27, 1938.
1 A complete list of the studies examined in connection with the article may be obtained in mimeo¬
graph form from the Division of Social Research, WPA.
49
50
American Statistical Association
all persons without regard to participation in economic activity, while
the other one includes only "persons 18 years of age or over, except
women not usually employed." "The Michigan Census of Population
and Unemployment"2 is an illustration of the first type of study, and
the studies presented as "Unemployment in Buffalo"2 illustrate the
second type. Still another type of survey excludes even more of the total
population than the Buffalo studies. In 1929 the Philadelphia survey
of unemployment4 listed the number of persons in the family who were
usually employed, but enumerated the personal and economic char¬
acteristics only of those who were unemployed. This type of limitation
was carried even farther in the direction of excluding all but a particu¬
lar part of the population in the National Unemployment Census of
November 19375 which sent a schedule to every household but required
a report only from the unemployed or partly employed worker who
was able to work and wanted work or more work.
These four surveys illustrate the range of population coverage that is
found among the surveys examined. At one extreme every person in
the family or household was enumerated; at the other extreme, only
the unemployed or partly employed worker who was able to work and
wanted work is included. The method common to all of the studies
was that by which the total population was "sorted down" to the un¬
employed. When all persons are enumerated the sorting takes place
after the field enumeration; when only part of the population is cov¬
ered, the sorting takes place at the time of enumeration.
These sorting devices serve in place of a definition of unemployment
in studies using the interview method of obtaining unemployment
data. In effect, the concepts that underlie these devices become defini¬
tions of unemployment when the sequence of "sorters" in each survey
is traced. Clearly, then, the validity of the results depends upon the
soundness and efficiency of these concepts. Two considerations by
which these sorter concepts may be judged are objectivity, and appli¬
cability to the determination of unemployment at a particular point in
time.
The concepts used in separating the total population into subgroups
« Miohigan State Emergency Welfare Relief Commission, Lansing, Michigan, 1935. For reports
Eee First Series Nos. 1-10, inclusive.
* Crouton, Fred C., and Croxton, Frederick E., Speoial Bulletin No. 163, New York State Depart¬
ment of Labor; Croxton, Frederick E., and Croxton, Fred C., Special Bulletin No. 167, New York
State Department of Labor; Croxton, Frederick E., Special Bulletins Noe. 172 and 179, New York
State Department of Labor.
4 Dewhurst, J. Frederic and Tupper, Ernest A. . . . Social and Economic Character of Unemploy¬
ment in Philadelphia . . . University of Pennsylvania, Wharton Sohool of Finance and Commerce.
Industrial research department. June 1930.
8 Census of Partial Employment, Unemployment and Occupations: 1937, Washington, D. C., 1938,
Vols. I to HI.
Concepts Used in Unemployment Surveys
51
provide the first step in the determination of unemployment in field
surveys. The sorter concepts used in the principal surveys of recent
years are the familiar ones of "gainful worker," "ability to work,"
"wanting (willingness) to work," "age," and "seeking work." Varying
combinations of these concepts were used in the several surveys. For
instance, to be included in the unemployment returns of the survey
reported as "The Incidence of Work Shortage in New Haven, 1931"6
a person had to be a "gainful worker," over 14 years of age, "able to
work," and "wanting work."
The relation of the sorting devices to the final determination of un¬
employment may be seen from the following illustration. In the Decen¬
nial Census of 1930 the sorter sequence that eventually determined un¬
employment began on the population schedule (Form 15-3b) in the
form of an entry in column 25, of the "gainful occupation pursued
. . . ; or 'none' (that is, no gainful occupation)" for every person enu¬
merated. By means of this entry the total population was divided
into two broad groups, gainful workers, and persons who were not
gainful workers. There was no definite time factor involved; that is,
the entry did not depend upon activity at any particular point of time
or for any specified duration.
The next step in the sequence was also on the population schedule in
the form of an entry (column 28) of "yes" or "no" to the question
"Employment; whether actually at work yesterday (or the last regular
working day)." At this point, a time factor came suddenly into opera¬
tion, so suddenly in fact, that it required quite a bit of explanation to
make it practicable in a variety of special situations encountered dur¬
ing the enumeration.
Having sorted the total population down into gainful workers and
other persons, and then having sorted the gainful workers into those
working and those not working, the unemployment inquiry was under¬
taken on a separate schedule (Form 15-95). Entries were to be made
on the unemployment schedule only for those gainful workers on the
population schedule who were not at work, that is, who answered "no"
in column 28 of that schedule.
Since all quantitative surveys of unemployment use some sequence
of sorters in determining the volume of unemployment, and since these
sorter sequences serve in place of a definition of unemployment, it is
appropriate to examine them in terms of objectivity and applicability
to the determination of unemployment at a particular point in time.
Among the surveys examined, the gainful worker concept easily
1 Hogg, Margaret H., The Incidence of Work Shortage, Russell Sage Foundation, New York, 1932.
52
American Statistical Association-
ranks first in order of importance. Any person who has done gainful
work satisfies the primary qualification for inclusion among gainful
workers, though he may be excluded from the unemployment count
on other grounds. The fact of gainful employment can, in most in¬
stances, be readily established; as a result there can be little question
about the objectivity of this concept.
Despite this advantage, the gainful worker concept has some very
obvious limitations, particularly during a period of prolonged depres¬
sion. The one most frequently noted is its automatic exclusion of "new
workers," i.e., inexperienced first job seekers, all of whom should be
counted as unemployed in the labor market sense, and most of whom
would be counted as unemployed in the needs sense.7 This particular
limitation of the gainful worker concept may be stated more broadly
as follows: the total number of gainful workers available for employ¬
ment at any one time is not the same as the total available labor
supply at that time.
The size and membership of the total available labor supply varies
with economic conditions. A household may have only one worker
when the head of the household is working full-time, but may have two
or more when the head loses his job, and the wife or children suddenly
become active job seekers. Unless they find work, these temporary
entrants into the labor market cannot be regarded as gainful workers.
In measuring unemployment, however, it is the total labor supply
—entrants, reentrants, temporary as well as permanent participants
—that must be examined; and this examination must refer to a specific
point in time. Thus, while the gainful worker concept meets the re¬
quirement of objectivity, it is open to question on the requirement of
applicability to the determination of unemployment at a particular
point in time.
A growing realization of the limitations of the gainful worker con¬
cept as a base for measuring unemployment has led to placing greater
reliance on the concepts of "able to work," and "willingness (wanting)
to work." One or both of these concepts have frequently been used in
connection with the gainful worker concept, so that the shift is largely
one of emphasis. Ability to work and willingness to work are widely
held to be logical and unambiguous tests that can readily be applied
to the measurement of unemployment. The difficulty with these con¬
cepts lies in their lack of objectivity—just the reverse of the case with
the gainful worker concept.
Just what does "ability to work" mean? In large part it means the
7 It should be noted that in reoent years most unemployment surveys make special provisions for
including new workers.
• Concepts Used in Unemployment Subveys
53
absence of physical handicaps to employment; but how is a determina¬
tion of this kind to be made in field surveys by the average enumerator
except .by taking the respondent's word in all but the most obvious
cases of disability. Ability to work is a relative matter. When, for
instance, does age become a handicap as real as the loss of an arm or a
leg? Ask the unemployed miner of 55 years if he is able to work and the
answer most likely will be "yes;" ask him why he lost his job and he is
very likely to answer "too old." Ability to work, therefore, is not a fact
that can be determined objectively in many important instances.
Ability to work is frequently paired with "wanting work" or with
"willing to work" in such a way that an affirmative response to both is
required to qualify the individual as eligible for the unemployment
inquiry. In application, the "able" and "willing" concepts are subject
to the same objection, i.e., they lack objectivity.
"Wanting work" or "willing to work" however the concept may be
expressed, is again a relative matter that can be determined accurately
only in the presence of an actual job offer. A large number of Mexicans
in San Antonio, Texas are "willing to work" for the low wage of 4 to 5
cents an hour in the pecan shelling industry; a surprising number of
youths—new entrants to the labor market—have been found willing
to work for nothing in order to gain experience for lack of which they
have been unsuccessful in their job seeking. On the other side, most
highly skilled workers—machinists, carpenters, bricklayers, etc.—are
willing to work only for a wage that contrasts sharply with the pre¬
vailing rate among unskilled workers. Moreover, "willing to work" is
not only relative as between an individual and a particular job, it is
also relative as between the individual and economic conditions. Thus,
even if willingness to work could be tested by an actual job offer, the
test would have to be highly variable.
Another objection to this concept is that it cannot be expressed in
neutral terms for field inquiries. The prolonged discussion in the public
press about whether the unemployed—particularly those on relief—
would work if jobs were offered has so colored the words in which
"wanting" and "willing" to work can be expressed that the question is
"loaded" when asked.8 A good illustration of this coloration is found
in the oft repeated statement: "Anyone who really wants a job can
find one."
The three concepts discussed up to this point have, in one combina¬
tion or another, determined what part of the total population was eli¬
gible for an inquiry about unemployment. This is true despite the fact
8 For an interesting discussion of the bias resulting from the use of "colored" terms in survey
questions, see Spingarn, Jerome H., "These Public Opinion Polls," Harpers Magazine, December 1938.
54
Ameeican Statistical Association-
that age limits have been used in most of the important surveys of
recent years. Age, however, is less a concept of unemployment than
a mechanical sorting device for excluding the very young and the very
old who are considered to be outside the customary age range for
economic activity.
There remains only one other concept—"seeking work"—that has
been used to any extent in surveys of unemployment. Its use has been
largely that of a secondary "sorter" applied to those who had previously
passed the test of "usually works," "able to work," or "wanted work."9
It is difficult to see why the "seeking work" concept has not been
more extensively used in unemployment surveys; not as a secondary
sorter, but jointly with "working" as the best means of identifying the
total labor supply. Like "working," "seeking work" is an activity that
can be reported in terms of what the individual is doing at the time
of inquiry. It does not depend upon a judgment by the respondent;
and it can readily be phrased in neutral terms. Moreover, it meets the
test of applicability in the determination of unemployment at a par¬
ticular point in time more closely than any of the other concepts.
The best demonstration of an individual's ability and willingness to
work is an active search for work, a fact that has been recognized in
instructions to enumerators where it is sometimes suggested as a means
of judging cases that offer difficulty of classification in terms of "abil¬
ity" and "willingness" to work. In addition, an active search for work
demonstrates availability for suitable employment as required by the
eligibility rules of unemployment compensation systems. Unemploy¬
ment determined by the seeking work concept is what the United
States Employment Service reports by an inventory of its active files.10
Finally, unemployment measured by means of the "seeking work"
concept distinguishes the unabsorbed portion of the active labor supply
in the sense used by the economist in analyzing the supply of, and the
demand for, workers. Thus, to objectivity and applicability there may
be added the further advantage of uniformity of meaning with the
theoretical approach to the problem of unemployment.
The advantages claimed for "seeking work" over the concepts used
in most surveys made in recent years can be more fully appraised from
an actual outline of an unemployment survey11 which uses this concept
8 "Actively seeking work" was one of the concepts used in the sample survey taken to check the
results of the voluntary registration census of November 1937, which did not use this conoept. See
Dedrick, Calvert L., and Hansen, Morris H., The Enumerative Check Census, VoL IV, of the final report
of the National Unemployment Census, Washington, D. C., 1938.
18 Survey of Employment Service Information, February 1938, United States Employment Service,
Department of Labor, p. 1.
11 The author wishes to acknowledge the assistance of his associates, particularly Katherine G.
Parker and Gordon M. Murray, in the planning of this survey.
• Concepts Used in Unemployment Surveys
55
to determine unemployment. The remainder of this article is given
over to a description of such a survey.
The survey to be described was formulated by the Division of Social
Research, WPA, for use in an exploratory study of unemployment in
selected urban areas, including some in which the long-time unem¬
ployed offer a particularly difficult problem of relief administration.
In order to minimize the risk of error in departing from established
procedures in unemployment surveys, a few basic rules were set up
for guidance in planning the field schedule. The rules were:
Every person in the unit of enumeration (the household) must be accounted
for either in terms of participation in the labor market or in terms of the
reason for nonparticipation. In other words, there must be no "pre-sorts"
of the total population that is to be examined for unemployment.
The objective tests of "working," and "not working but actively seeking
work" must be used in a way that distinguishes sharply the active part of the
labor supply at a particular point of time.
At the same time the inactive part of the labor supply must be distinguished
with sufficient clarity to permit combination with the active part for some
analyses, and for comparison with other studies.
The field schedule must not be unduly complicated or make more than
ordinary demands upon the ability of the enumerator.
The number and arrangement of schedule questions that conform to
these rules is by no means fixed. The number and arrangement outlined
below is simply that which seemed best adapted to the needs of a par¬
ticular inquiry. For convenience in discussion, the questions are ar¬
ranged by sections as they appeared on the test schedule. The unit of
enumeration was the household, and the time interval, one week.
The first section "A," is concerned with the personal characteristics
of each member of the household.
A
Personal characteristics
Line
No.
Name
Relationship
to head
Age
last
birth¬
day
Sex
Color
and
na¬
tivity
Mari¬
tal
status
Last
school
grade
com¬
pleted
(1)
(2)
(3)
(4)
(5)
(6)
(7)
This section requires no comment other than that the questions are
the standard ones found in most population studies, and that this
section lists the total population for subsequent examination as to
participation or nonparticipation in economic activity.
The way in which "working" and "seeking work" can be used to dis¬
tinguish the active part of the labor supply, and at the same time dis-
56 American Statistical Association-
tinguish the inactive part from the remainder of the population, is
shown in section B.
B. Aotivity during oensus week of through
Working (include persons assigned to WPA, NYA,
and other Emergenoy Work Programs or CCC)
Not working
but active¬
ly seeking
work. (Give
date this
period of
seeking
work began)
Mo. Yr.
Not working
actively seeki
Reason for
not actively
seeking work
and not
ng work
Date
reason
became
effec¬
tive
Mo. Yr.
Occupation
Industry
Class
of
worker
or
pro¬
gram
Hours
worked
Date
job
began
Full
time
Part
time
Mo. Yr.
(8)
(91
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(On the actual schedules, columns 8, 9, and 15 will be made much wider.)
The entries in section B distribute all members of the household into
three groups depending on their activity during the census week. These
groups are: persons actually at work during the census week who re¬
ceive pay, profit, goods or services for their efforts, or who assist in the
production of marketable goods without pay (i.e., unpaid family
workers); persons not working but actively seeking work during the
week; and persons who are not actively in the labor market during the
week.
The first group—persons working—includes the employed in the
sense of persons engaged in productive activity at some time during
the week. The inclusion in this group of persons employed on emer¬
gency work programs such as the WPA, and persons engaged in unpaid
family work, not only simplifies the definition of employment by
making it depend on activity within a week, but also takes account of
the fact that many emergency program and unpaid family workers
consider themselves employed in the ordinary meaning of the term.
The entry of type of worker or program12 for each occupational and
industrial entry identifies these workers and permits their segregation
for purposes of analysis.
This method of handling entries of employment relieves the enumer¬
ator of the task of deciding what types of work constitute "employ¬
ment" before making entries for persons working. The fact that emer¬
gency program and unpaid family workers are tabulated as separate
groups in the working population allows the user of the returns to
11 i.e., wage earner, employer, own account, unpaid family worker not seeking work for pay,
unpaid family worker seeking work for pay, apprentices without pay and persons working for experi¬
ence only, WPA, NYA, other emergency work programs, and CCC.
• Concepts Used in Unemployment Surveys
57
decide whether, for his purposes, they were to be considered as em¬
ployed, unemployed, or as a special group.
In the second group—persons not working but actively seeking work
—are included those workers who are unemployed in the strictest sense
of the word, namely, persons without jobs of any kind who were ac¬
tively seeking employment. For purposes of comparison with other
studies it is a simple matter to add to this group the emergency pro¬
gram workers and persons whose reason (column 15) for not actively
seeking work during the census week is that their separation from
employment is of a temporary nature.
Persons not actively seeking work during the census week consist
principally of individuals who are outside the labor market—children,
students, housewives, retired workers, invalids, the physically in¬
capacitated, etc. But it will also include and identify workers who do
not fit readily into either the employed or the unemployed groups dur¬
ing the week of inquiry; this includes workers on strike, seasonal work¬
ers who are idle but are not seeking work, and workers who, being at¬
tached to industries (e.g., coal mining) which have stopped operations
for an indefinite period of time, are not seeking work because they
know none can be had.
These "in-between cases," which are neither employed nor unem¬
ployed in the ordinary sense, are troublesome far beyond their numeri¬
cal importance. For example, there is the perennial problem of what
entry should be made for the worker on layoff in a one-industry town
where that industry has closed for an indefinite period of time. To
report workers in this situation as employed because they would have
a job if and when the industry reopened is to bring into question the
meaning of the term "employed." The fact in this case and in many
others is that the worker and the market for his services have become
separated; and it seems logical to show this separation, particularly
since it would disclose "stranded" workers within the total population.
A further advantage of recording the reason for not actively seeking
work is that it rounds out a full report on the total population being
studied. The entire group is classified according to one concept—labor
market activity during a specific week. The total available labor supply
is identified, and reasons are given to explain why some part of the
population is not included in the available supply at the time of in¬
quiry.
It would be possible to stop here and still obtain a good picture of
employment, unemployment, and the distribution of the total popu¬
lation according to what they are doing at a particular point in time.
It would not, however, be possible to obtain from so brief an inquiry,
58
American Statistical Association-
much needed information on duration of unemployment from the
last nonrelief job, the duration of the last job, the "usual occupation,"18
and the length of time since the usual occupation was followed. A third
section, therefore, has been added which although not essential to the
determination of employment and unemployment, materially increases
the amount of information obtained by the survey.
C. Activity prior to census week of
Last full time private job of 2 weeks or longer
Usual occupation
Occu¬
pation
In¬
dustry
Class
of
worker
Date
Began
job:
Ended
Occu¬
pation
In¬
dustry
Last job of 2
or longer
Date job:
eeks
Class
of
worker
or
program
Began
Ended
Mo. Yr.
Mo. Yr.
Mo. Yr.
Mo. Yr.
(17)
(18)
(19)
(20)
(21)
(22)
(23)
(24)
(25)
(26)
(On the actual sohedules, columns 17, 18, 22, and 23 will be made much wider.)
In addition to extending the amount and the scope of the informa¬
tion obtained on employment and unemployment, section C achieves
a needed simplicity by bringing together in one place information on
the worker's past experience without in any way altering or complicat¬
ing the determination of what he was doing at the time the inquiry
was made. In a sense, section C is a supplement to section B, and allows
the inquiry outlined in this paper to be used either in a "short form,"
i.e., sections A and B, or in a "long form," consisting of sections A,
B and C.
Viewed as a supplementary inquiry, section C offers a compromise
between the minimum of information required of unemployment sur¬
veys, and the much desired work-history record that is not feasible
except on a very small scale.
In presenting this outline of a survey of unemployment, the main
purpose has been to emphasize the objectivity of the inquiry from the
point of view of field enumeration. The objectivity of this procedure
arises out of the simple fact that it reports what the worker is doing
within a particular unit of time. If a person is not working, he either is
18 There is much variation in definition of what constitutes "usual occupation." The intent of this
question is, in general, to find out what trade or profession the worker customarily follows. While it
may have been easy at one time to obtain a clear-cut answer to this question, it is certainly no simple
matter after 8 years of widespread unemployment. In the plan outlined above, the usual occupation
has been defined as: "The occupation that the worker, in the light of his work experience and training,
regards as his usual occupation."
• Concepts Used in Unemployment Surveys
59
or is not actively seeking work. To the question, then, "Are you ac¬
tively seeking work?" the person need merely answer on the basis of
his actual performance.
It is urged, then, that activity as a test for determining unemploy¬
ment is not only simple of definition and of application to enumeration,
but that it would also clarify the meaning of returns by relating them
to the way in which the labor market does operate. And finally, its
use would provide a basis for standardization of reports that is much
needed at the present time.