Quasi-Facts

The point of the following discussions of what we have called quasi-facts

is not to suggest that such measurements are of doubtful value. It is rather
to indicate that much of what we need to learn about our society forces us
to deal with phenomena that are subjective in part, if not in whole; and that
the solution of simply ignoring the subjective realm is not a feasible one.

The Subjectivity of Ethnicity

Tom W. Smith

Basic background variables are commonly seen as concrete and objective

factors. In fact they often have a large subjective component. Ethnicity is a
prime example. First there is the difficult problem of defining what ethnici-
ty is. It is most frequently seen as some form of a cultural heritage or
identification that is defined by some combination of nationality, language,
religion, and race (Isajiw, 1974). We will not even try to disentangle how
one plucks "ethnicity" out of these and related factors. Are Jews from Po-
land Jews, Poles, Polish Jews, or Jewish Poles? If we look at Catholic and
Protestant Germans, do we see two ethnic groups or one ethnic group
broken down by religion? Are West Indian blacks in the United States
blacks, West Indians, some combination, or Africans? Are Creoles and
Amerindians of Mexico both Mexican, both Hispanic, or separate? Are
people from Sicily Sicilians or Italians? Certainly we could promulgate a
complex set of standards to objectively resolve these and a long string of
related ambiguitie,, but while ~ e r h a p sobjective in the sense of being sus-
ceptible to consistent repetitive application by different enumerators, such
standards would be based on arbitrary classifications that undermine a full
sense of objectiveness.
This chapter concentrates on the narrower problem of nationality.' Most
social scientists give more weight or emphasis to this factor in their concep-
tualization and/or classifications, and people usually express their ethnicity
in what might be considered nationality groups. If we count nationality
 The Subjectivity of Ethnicity
groups within multinational states as nationalities (for example, Serbs and -If response to Q. 2 was "no":
Croatians), then almost all respondents respond to even vague terms such 2A. Which country was your father born in?
as "origin," "ancestry," or "descent" in terms of nationality. (The major 2B. Which country was your mother born in?
exceptions are blacks, who will either respond in terms of their race or with -If response to Q. 2 was "yes" or "don't know":
a general reference to Africa.) 2C. Do you remember which country your family came from originally on
Attempts to determine nationality use three basic approaches: (1) the your father's side?
natal, (2) the behavioral, and (3) the subjective.2 The natal approach identi- 2D. Do you remember which country your family came from originally on
fies a respondent's nationality by determining a person's place of birth, the your mother's side?
places of birth of his or her parents, grandparents, and so forth. The behav-
ioral approach determines a respondent's nationality according to some This begins to shift from being an objective measure to a subjective or
practice, &liation, or membership such as language spoken or voluntary self-identification measure. Inquiring about "which country your family
group membership. The subjective approach simply asks respondents what came from" does not measure the national origins of the respondent's
nationality they consider themselves to be or where their ancestors came ancestors as a whole, but rather elicits a single origin from among potential-
from. ly several different ancestral lines. A strict (although obviously impractical)
The natal approach to ethnicity is typified by the traditional item used by natal approach would inquire about the place of birth of all ancestors until
the Bureau of the Census. I t asks the places of birth of the respondent and all lines were traced back to a country of origin.
his or her mother and father: This shift from the objectivity of place of birth to the subjectivity of self-
identification proceeds one step further in a standard Michigan Election
13a. WHERE WAS THIS PERSON BORN? If born in hospital, give State or Study question which asks: "In addition to being an American, what do you
country where mother lived. If born outside U.S., see instruction sheet: dis- consider your main ethnic or nationality group?' This question emphasizes
tinguish Northern Ireland from Ireland (Eire). one's "main" background but does not make clear how this concept is to b e
operationalized. Furthermore, the question moves away from the place-of-
T h i s State birth definition by referring to "ethnic or nationality group" rather than to
OR country of origin in a geopolitical sense. Similar in kind is the Current
(Name of State or foreign country; or Puerto Rico, Guam, etc.) Population Survey (CPS) item asking: "What is 's origin or
descent?" The CPS establishes no criteria and uses the somewhat less spe-
14. WHAT COUNTRY WAS HIS FATHER BORN IN?
cific terms "origin" and "descent." Likewise, the 1980 Census asked: "What
U n i t e d States
is this person's ancestry? I F UNCERTAIN ABOUT H O W T O REPORT
OR
(Name of State or foreign country; or Puerto Rico, Guam, etc.) ANCESTRY: SEE INSTRUCTION GUIDE. (For example, Afro-Amer.,
English, French, German, Honduran, Hungarian, Irish, Italian, Jamaican,
15. WHAT COUNTRY WAS HIS MOTHER BORN IN? Korean, Lebanese, Mexican, Nigerian, Polish, Ukrainian, Venezuelan,
U n i t e d States etc.)" The question uses the very vague term "ancestry" (although from its
OR examples it makes clear that it is referring to nationality) and offers no
(Name of State or foreign country; or Puerto Rico, Guam, etc.) criteria for determining ancestry. The instructions make the nationality ref-
erence explicit, stating, "Ancestry (or origin or descent) may b e viewed as
A variant of this approach, used by the Michigan Election Studies, asks the nationality group, the lineage, or the country in which a person or
parallel information about the respondent and his or her parents, but then person's parents or ancestors were born before their arrival in the United
inquires about the general ancestral origins of respondents who are third (or States." The instructions also set a general standard for handling multiple
later) generation: nationalities: "Persons who are of more than one origin and who cannot
identify with a single group should print their multiple ancestry (for exam-
1. Where were you born? (IF UNITED STATES) Which state? ple, German-Irish)." The instructions also make explicit the subjective com-
2. Were both your parents born in this country? ponent in the question by telling the informant to "Print the ancestry group
 The Subjectivity of Ethnicity
with which the person identifies." In sum, by failing to collect detailed error aggravated by the complexity of categorizations and situations. The
information on nativity and asking people to select an ancestry on the basis approaches do vary, however, in how well they deal with each of these
of identification, the 1980 Census has established a largely subjective mea- problems.
sure of ethnicity. A basic drawback of the two relatively objective approaches (natal and
Even more explicitly subjective is the General Social Survey (GSS) (Da- behavioral) is that they lead to a much higher rate of nonidentification than
vis, Smith, and Stephenson, 1980) question that asks: "From what countries the subjective approach. Responses to the 1970 Census nativity question,
or part of the world did your ancestors come? I F MORE THAN ONE for example, showed that only 4.7 percent of the population were immi-
COUNTRY IS NAMED: Which of these countries do you feel closer to?' grants, and 11.8 percent were native-born of foreign-born or mixed parent-
Not only is no criterion specified by which to choose between origins in the age. That leaves without any indication of ethnicity the 83.5 percent who
initial question, but in the followup question, people giving multiple origins were native-born of native-born parents. In theory one could extend the
are told to use a subjective standard-feeling closer to--rather than to nativity question back until each ancestral line was traced to foreign shores,
choose a particular lineage or their most frequent origin. but this is impossible since the number of ancestors increases geometrically
Clearly the line between the natal and subjective approaches is often a across generations, while knowledge declines in a similarly precipitous fash-
fine one. In general one passes from the natal approach when one moves ion. Approximately 56 percent of the adult population report that all four of
from asking information about the place of birth of specific persons (the their grandparents were native-born. For this large segment of the popula-
respondent and his or her ancestors) to nonspecific information about one's tion, the place of birth of at least eight great-grandparents (and many more
background, descent, ethnicity, nationality, or origin. ancestors if all eight were not foreign-born) would have to be known in
More clearly separated from the natal or subjective approaches is the order to have complete information. This is well beyond the knowledge of
behavioral approach, which classifies a person according to some practice or most people.4 Without complete information, identification could be made
affiliation such as language spoken or membership in certain voluntary as- only by assuming that the missing data agreed with the available data or via
sociations. An example of the language approach comes from the 1970 Cen- some other imputation procedure.
sus, which inquires: "What language, other than English, was spoken in The behavioral approach suffers even more seriously from nonidentifica-
this person's home when he was a child?"3 The affiliation approach is com- tion. The 1970 Census found that only 12 percent of the population did not
monly used when a list sample is employed to select respondents. Under report English as their native language, and the General Social Surveys find
this method, membership in the association that the list represents that only 3 percent of adults report membership in a "nationality group."
becomes the definition of nationality. This might include congregations, Again, in theory one could extend this approach across generations, but the
mutual benefit societies, or other groups (for examples, see Vrga, 1971; results would probably be even less fruitful than in the case of nativity.
Masuda, Matsumoto, and Meridith, 1970; Barton, 1975). Also included in Only the subjective approach succeeds in classifying a substantial majori-
this appioach are lists based on such documents as baptismal and marriage ty of people. The GSS question elicited some ethnic identification from 86
registers (which frequently include persons who are not actually members percent of the population (during the period 1972-80), while in the 1976
of such congregations). Another hypothetical example would be a survey Michigan Election Study, 89 percent of the white population mentioned an
that asked a series of ethnic-orientation questions (such as foods eaten, mu- ethnicity. In sum, nonidentification is a general problem, and there re-
sic preferred, and so forth) and then assigned ethnicity according to the mains a significant share of the population (10 to 15 percent) with no mean-
responses. ingful ethnic identification. The subjective questions do, however, mini-
Each of these approaches has particular strengths and weaknesses. We mize the problem by asking for a simple, summary identification, while the
can broadly evaluate them by considering how each handles three major more objective approaches, by necessitating more voluminous and exact
p-oblems in measuring ethnicity: (1) nonidentification, (2) multiple identifi- information than typical respondents possess, are able to come up with
cation, and (3) misidentification. Ideally, an ethnic measure would maxi- complete nativity data for only a minority of the population.
mize the number of identifications, simplify the handling of multiple na- The performance of the strictly natal approach can be enhanced consider-
tionalities, and minimize erroneous identifications. No approach (nor any ably, however, if it is modified to ask about the general ancestral origins of
combination) can avoid or solve these problems completely, since noniden- people of the third (or perhaps fourth) generation, as in the Michigan Elec-
tification and multiple identification are intrinsic to the subject, and mis- tion Study example cited above. In the 1972 Michigan Election Study, for
identification results from the general and basic problem of measurement instance, 62 percent of the third-generation white Americans were able to
 QUASI-FACTS The Subjectivity of Ethnicity
give both maternal and paternal ancestral origins, and 18 percent more with multiple ethnicities (for example, one Irish, one Irish-English, one Polish,
were able to specify the origin of one lineage. This modification of the natal and one Lithuanian). In the 1980 GSS, of those giving nationalities for both
approach greatly reduces the problem of nonidentification, but only by parents, 43.2 percent had parents with a single, common nationality; 8.7 per-
abandoning a strictly natal approach for a hybrid of the natal and subjective. cent had different mixtures of nationalities but one shared nationality; and 48.1
Nonidentification can also b e minimized by the use of a combination of percent had different nationalities for their parents. Under these common cir-
approaches. While a simple subjective question identifies a high proportion cumstances the assigning of a primary nationality would be impossible or arbi-
of all possible identifiers (see Smith [1980] on the reason for nonidentifica- trary.
tion), a natal approach or hybrid approach can identify the national origins
The behavioral approach does not typically have much of a problem with
of an additional segment of the population. In the 1972-74 Michigan Elec-
overidentification (since it suffers so severely from nonidentification).,, but if
tion Studies, for example, the additional information from the hybrid ap-
it was extended across generations, it would have the same problem as the
proach reduced the portion of whites unidentified on the subjective ques-
nativity question does. The subjective alternative either minimizes the
tion from 28 percent to 12 percent.
problem of multiple identification by asking for a simple summary national-
In sum, nonidentification is an intrinsic problem. Even given the best
ity, as the Michigan Election Study question does, or by asking people to
possible combination of approaches, 10 to 15 percent of Americans have no
choose between nationalities on some subjective ground. The GSS, for ex-
ethnic identification and no information on their national origins. The non-
ample, asks people giving two or more nationalities: "Which one of these
response problem can b e minimized by combining a subjective question
countries do you feel closer to?" The 1980 Census instructions ask for the
with a natal or hybrid natal-subjective item. As a single item the subjective
approach is most useful; strictly natal and behavioral approaches identify a ancestry a person "identifies" with. These kinds of followups can greatly
simplify the problem of ethnic identification. On the 1972-80 General So-
smaller proportion of people.
cial Surveys, for example, 68.5 percent of those naming more than one
The second major problem in ethnic identification is just the opposite of
nationality were able to name a primary ethnicity.
nonidentification (that is, overidentification or multiple-identification). On
It can be fairly argued that complexity and detail are desirable attributes
the 1972-80 General Social Surveys, 35 percent of respondents named two
of an ethnic measure and that they should not be compressed away in
or more countries when asked about their national origins. In addition, the
analysis. Condensation, however, is often a practical necessity and may
1980 GSS found that among those who named one ethnicity, 24 percent
even be more meaningful than detailed information on the birthplaces of
mentioned two or more nationalities for their parents. For these people the
several generations of one's ancestors. Given that some simplification may
task is to try to sort out a main ethnicity or to otherwise handle the multiple
identification. The natal approach generates a large and rich array of iden- be useful, the question becomes whether a genetic approach, such as
nativity, or a subjective choice is more useful. Compare, for example, the
tifications, but contains no device to distill the data or select a main identifi-
approaches used in the Census and the GSS. In the 1970 Census, father's
cation. Several solutions are available for handling multiple ethnicities on
country of origin is used to determine a person's nationality when the par-
nativity questions:
ents have different nationalities. O n the 1980 GSS we were able to compare
1. Each combination could be treated as a separate group. Thus English and i the nationalities of parents with the summary nationality of the respondent.
German as well as English-German would be groups. This would soon lead to The standard GSS ethnicity question asked: "From what countries or part
;i
so many combinations with minuscule numbers that most would have to be of the world did your ancestors come? IF MORE THAN ONE COUNTRY
collapsed into a residual of other combinations. 1i IS NAMED: which of these countries do you feel closer to?" As part of a
supplement to the 1980 survey, respondents were asked the country of
2. A simple trichotomy of all-English, some English, and no English could be
established, but this would obviously hamper interethnic comparisons. birth of their parents. For parents born in the United States respondents
3. A set of rules to choose a primary ethnicity could be devised. If, for example, were asked: "What countries or parts of the world did your (mother'slfa-
a person reports one Danish and three Swedish grandparents, then it might be ther's) ancestors come from?" Up to two responses were coded for each
reasonable to code Swedish as the primary nationality. Unfortunately the dif- parent.5 W e compared the summary nationality data from the standard GSS
ference is often less clear-such as choosing between parents with different question with the parental nationality data. W e looked at instances in which
nationalities (for example, one Irish and one Swiss) or between grandparents (1) different ethnicities were reported for the parents, (2) two different eth-
 The Subjectivity of Ethnicity

geopolitical entities, but uses nationality groups as units instead. Likewise,

nicities were reported for the same parent, and (3) interparental and intra-
accidents of place of birth will typically not confound the data, since the
parental combinations of differences occurred. We found that among the
respondent will screen out such false evidence and report national identity
cases where one or more different nationalities were reported for the par-
ents, 20.4 percent had no ethnic identifications, 5.9 percent chose an iden- instead. On the other hand, since the self-identification approach elicits an
ethnicity from many more people than does the natal or behavioral ap-
tification different from that of either parent,6 14.5 percent selected an
proach, it includes many people with weak and nominal identification. As a
ethnicity shared by their parents, 26.4 percent chose their mother's ethnici-
result, it probably receives a higher proportion of labile responses than do
ty, and 35.8 percent chose their father's ethnicity. Among those who chose
the natal or behavioral approaches.
only between their parents, 58 percent selected the paternal line while 42
Thus, subjective identifications are likely to have lower test-retest relia-
percent chose the maternal. This evidence indicates that the census had
bility than the more objective measures. In addition, reliability would also
some basis for favoring paternal lineage over maternal, but in more than 40
tend to be lower since a person with multiple nationalities might switch his
percent of the cases it results in the assignment of an ethnic identity with
or her subjective identification depending on personal factors and social
which a respondent does not actually identify.
pressures (National Research Council, 1978). Also, the subjective approach
Depending on the research purpose, a person may wish to use various
can lead to certain questionable or inappropriate classifications (similar to
methods for handling multiple identities. In a study of Italian assimilation,
the birthplace of children of diplomats) such as the black Irishman who calls
for example, it might be desirable to include initially all people with any
himself Spanish, or a person adopting a spouse's nationality.* In brief, by
Italian lineage whether or not they identify themselves as Italian. Similarly,
relying on the nationlstate of birth, the natal approach will probably create
studies of father-son mobility have focused on paternal origins (Featherman
more erroneous identifications than most behavioral approaches (for exam-
and Hauser, 1978:523-528). An investigation of the pattern of ethnic iden-
ple, language) or the subjective approach, either of which would reduce
tification might compare various nativity combinations with subjective iden-
mistakes due to geopolitical peculiarities and circumstances of birth. The
tification. If the researcher is looking for a main identification, then subjec-
subjective approach, however, may have lower test-retest reliability be-
tive identification is probably more reasonable than is reliance on some
cause it encompasses more weak identifiers.
type of genetic weighting or arbitrary assignment.
The previous discussion shows that a strictly objective approach to the
Finally, there is the problem of misidentification. Some misidentification
measurement of ethnicity and nationality is difficult because (1) many
is inevitable, given the vagaries of memory, misunderstandings, errors of
people Iack sufficient information to supply complete data on ancestral na-
transference, and so forth. In the natal approach the special type of misi-
tional origins, (2) multiple nationalities create difficult problems for objec-
dentification that occurs involves disparities between place of birth and
tive methods, (3) objective ways of handling multiple nationalities probably
nationality. A strict geopolitical reporting will lose all "stateless" nationali-
produce classifications that are less personally and sociologically meaningful
ties, such as Serbian, Kurdish, Walloon, French Canadian, Armenian, or
than those of a subjective approach, and (4) emphasizing the country or
Lithuanian. Such reporting will also be influenced by the shift of bound-
place of birth distorts classifications because of multinational states,
aries and the creation and destruction of states. Polish nativity prior to
changed boundaries, national minorities, and other "accidents" of birth and
World War I is only discernible in the U.S. censuses by subtracting from
geography. In sum, a solely objective approach to ethnicity or nationality
the Russians, Germans, and Austro-Hungarians those with Polish as a
would produce less information, take more effort, and result in less relevant
mother tongue. Likewise, accidents of place of birth, such as birthplaces of
data than an approach that incorporates a subjective element.
children of military personnel or of those in the diplomatic corps, will con-
The limitations of the strictly natal approach are tacitly recognized by the
found the situation. Behavioral identifications such as mother tongue do not
fact that the standard ethnicity questions used by the government and aca-
create as many problems as does geopolitical reporting, although it is im-
demia (see page 119) all either implicitly or explicitly incorporate a subjec-
possible to distinguish between nationalities speaking the same language
tive element (Lowry, 1980). The preferred approach will depend on the
(for example, Irish, English, and Scottish; or German and Austrian). On the
precise research objective, but generally a combination of all three methods
other hand, identification by mother tongue can separate out such groups as
would be desirable. Nativity questions provide important information on
French Canadians, the Flemings and the Walloons, or the Serbians and the
immigrant generations and heterogeneous lineage. A behavioral question
Croatians. The subjective approach can avoid the stateless nationality and
such as that of language can clarify various ambiguous identities and pro-
boundary problem as long as it does not rigidly structure itself in terms of
 The Subjectivity of Ethnicity
vide evidence of the strength of the identification. A subjective approach (four) of their great-grandparents. Among respondents with at least one native-born parent,
paternal andlor maternal origin was unknown for 24 percent. Among the 76 percent with both
will minimize nonidentification, handle multiple identifications in a simple a known maternal and paternal origin, few could probably successfully track all ancestral lines
and relevant fashion, and reduce some types of misidentifications. When as a strict natal approach would require (see also Davis and Smith, 1980).
used together, each method can both buttress data obtained by the others
and add valuable additional information that is missed by them. In addition, 5. Only 20 percent of those with one nationality gave a second nationality, so the restriction
by using all three methods along with other items on strength of identity, to code only two nationalities per parent probably lost little information.
importance of identity, and behavioral consequences, we would be able to 6. We looked at these odd cases in which a respondent reported an ethnicity different from
study the meaning, sources, and consequences of ethnicity. 1 that of either parent. Around a quarter were instances in which people expressed an ethnicity
In summary, the measurement of ethnicity and the narrower element of in different ways, such as Mexico versus Spain or England versus Canada. The other combina-
nationality involves a strong subjective aspect. This is apparent in most tions are not readily explainable. They probably result from such factors as mixing references
governmental and scholarly approaches. We have further found that if one to natural and adopted parents, contradictions between place of birth and nationality, and
coding or processing errors.
is forced to rely on a quick and simple approach, a well-crafted subjective
question is the best single indicator for most purposes. Even the preferred 7. In 1973 and 1974, subsa~nplesof the GSS were reinterviewed about 1 month later; 82
method--combining the behavioral, natal, and subjective approaches--de- percent either selected an ethnicity or chose no identity both times. Among the consistent
pends in large part on a subjective element. 1 identifiers, 89 percent selected the same nationality. This gives 74 percent as consistently
defining their ethnicity (or lack of same). This percentage tends to be lower than that for other
! demographics: 97 percent were consistent on region of residence at age 16, 92 percent on
religious preference, 85 percent on father's education, 70 percent on community type at age
I! 16, and 69 percent on. number of siblings. The consistency was higher when collapsed ethnic
groups were used. For example, using English versus non-English, 79 percent were consis-
I tent.
Notes I
! 8. We assume that distortion from conscious attempts at racial/ethnic passing would be
similar across approaches. A person wishing to be identified with a Inore prestigious nationality
1. Schemes that use multiple variables to construct ethnic categories still collect the con- would presumably alter natal as well as subjective responses.
stituent parts as separate variables, so one is dealing with what usually are clearly distin-
guished components. i

2. In addition to these, there are several other methods of identifying national origins, such
as by surname or by physical characteristics. Neither these nor other techniques are generally
reliable or co~nznonlyemployed.
I
References
For example, the Current Population Survey (CPS) found that "among all persons with a
Abramson, J. J. (1975) The religioethnic factor and the American experience: another look at
Spanish surname in the United States in March 1971, only about two-thirds reported that they
the three-generations hypothesis. Ethnicity 2:163-177.
were of Spanish origin . . . among all persons in the United States who reported they were of
Spanish origin, about two-thirds had a Spanish surname and one-third did not" (Bureau of the American Council of Learned Societies (1932) Report of Committee on Linguistic and Na-
Census, 1975:2). On the difficulty of using surnalnes for classification in general, see Anlerican tional Stocks in the Population of the United States. In Annual Report of the American Histori-
Council of Learned Societies (1932). cal Association for the Year 1931. Washington, D.C.: Government Printing OEce.
Barton, J.J. (1975) Peasants and Strangers: Italians, Rumanians, and Slovaks in an Ameri-
3. For problems with this item, see Bureau of the Census (1974).
can City, 1890-1950. Cambridge, Mass.: Harvard University Press.
4. Both because of mortality and associational patterns (that is, with each prior generation Bureau of the Census (1974) Census of Population and Housing: 1970 Evaluation and Re-
the ancestor is less likely to be alive during the respondent's lifetime, and, if alive, is likely to search Program, PHC(E)-9: Accuracy of Data for Selected Population Characteristics as Mea-
have less contact with the respondent), knowledge about ancestors quickly diminishes with sured by Reintemiews. Washington. D.C.: Government Printing Office.
each intervening generation. A National Opinion Research Center mobility study (Davis and Bureau of the Census (1975) Comparison of Spanish Surname and Persow of Spanish Ori-
Smith, 1980) found, for example, that while 98 percent of respondents knew their father's gin in the United States. By Edward W. Fernandez. Technical Paper No. 38. Washington,
occupation, only 76 percent knew the occupation of their paternal grandfather. Knowledge D.C.: Government Printing OEce.
about paternal great-grandfathers could be expected to decline as sharply. Schneider and
Cottrell (197565-66) found that while 55 percent of their white, middle-class Chicago sample Davis, J.A., and Smith, T. W. (1980) Looking backward: a national sample survey of ances-
could give the first or last names of all four grandparents, only 14 percent could identify half tors and precessors, 1980-1850. Historical Methods 13:145-162.
 Davis, J.A., Smith, T.W., and Stephenson, C.B. (1980) General Social Surveys: 1972-1980:
Cuinulative Codebook. Chicago: National Opinion Research Center.
Featherman, D.L., and Hauser, R.M. (1978) Opportunity and Change. New York: Academ-
ic Press.
Isajiw, W.W. (1974) Definitions of ethnicity. Ethnicity 1:111-124.
Lowry, I.S. (1980) The science and politics of ethnic enumeration. Paper presented to the
American Association for the Advancement of Science, San Francisco, January.
Masuda, M., Matsumoto, G.H., and Meredith, G.M. (1970) Ethnic identity in three gener-
ations of Japanese-Americans. Journal of Social Psychology 81:199-207.
National Research Council (1978) Counting the People in 1980: An Appraisal of Census
Plans. Washington D.C.: National Academy of Sciences.
Plax, M. (1972) On studying ethnicity. Public Opinion Quarterly 36:99-100. Measuring Employment and
Schneider, D.M., and Cottrell, C.B. (1975) The American Kin Unitierse: A Genealogical
Study. Chicago: University of Chicago Press.
Smith, T.W. (1980) Ethnic measurement identification. Etlznicity 7:78-95. Unemployment
Vrga, D.J. (1971) Serbians in America: the effects of status discrepancy on ethno-religious
factualism. In 0. Feinstein, ed., Ethnic Groups in the City. Lexington, Mass.: Heath.
Barbara A. Bailar and Naorni D . Rothwell

Acknowledgments
This chapter draws on material used in Smith (1980). I would like to thank Theresa
DeMaio, Otis Dudley Duncan, Baruch Fischhoff, Lester R. Frankel, William Kruskal, Stanley
Lieberson, and Robert Parke for their cotnments.

Whether a person is employed, unemployed, or out of the labor force may

seem at first a factual item, and one that can easily be verified. There are
people, however, for whom labor force status is an attitude that cannot be
verified from records. Whether or not a person will be reported in statisti-
cal tabulations as employed, unemployed, or out of the labor force may
depend on various factors such as the weather or the interviewer's skill;
who responds for the person; the respondent's interest, mood, or percep-
tions about the purposes and uses of the interview; the time of day; who
else is present; other things that happened that day; and the questions
asked. The effects on labor force classification of some of these variables will
be illustrated here after some facts about the Current Population Survey
are described.
The Bureau of the Census conducts the Current Population Survey (CPS)
monthly to estimate employment and unemployment. The survey, begun
in 1940, has undergone a number of conceptual changes, which are de-
scribed in the chapter on historical development of the 1979 report of the
National Commission on Employment and Unemployment Statistics (1979: