Printer Friendly

Undercoverage of Hispanics in household surveys.

The fact that individuals in some racial and ethnic groups have a high rate of being missed in surveys has become an important subject of discussion and debate by analysts, policymakers, social scientists, and other concerned persons. The focus of this report is on factors contributing to the undercount of the Hispanic population in the Current Population Survey (CPS).(1) The CPS measures labor force activity in a monthly sample of approximately 60,000 households selected to represent the U.S. population.(2)

Undercounts can occur when the survey enumerator misses an entire household (whole household undercoverage) or when the person reporting for the household does not list all of the occupants of the household (within household undercoverage). The major causes of whole household undercoverage have been identified as missed addresses from the decennial census and deliberately omitted households, for example, a dwelling subdivided into several unreported housing units.(3) Within household undercoverage is believed to stem from differences in definitions of household membership used by the survey and respondents in ethnic communities.(4)

Anthropological research methods were used to investigate factors associated with undercount of Hispanics in the CPS and the decennial census. The research consisted of a small-scale field study of a Salvadoran community and an analysis of data collected from eight, predominantly Hispanic, sites studied by the Bureau of the Census in its 1990 Alternative Enumeration (discussed later).

The research methods

The anthropological research methods most often used in undercoverage research are from two areas: ethnography and cognitive anthropology. Ethnography describes the culture, or way of life, of a particular society, from the point of view of members of that society. Ethnographic methods include participant observation, which requires a period of residence within the community so that the ethnographer can become immersed in the everyday realities of that society, and the ethnographic interview, a particular kind of unstructured interview that attempts to elicit the cultural patterning and context of the subject or topic under discussion. Cognitive anthropologists study the conceptual categories which members of a culture use to organize their experience of the world. A good illustration of the application of cognitive anthropological methods to survey research is Elizabeth Gerber's investigation of the concepts of "living with" and "staying at" used by inner city blacks in the District of Columbia. She reported that a man who had not been in his mother's house for 5 months would, under certain circumstances, be classified as "living with" his mother, and that his relationship to a household in which he had eaten and slept every day for the past 5 months would be termed "staying at" that household. The basis for such classification involved, among other factors, intentionality about relationships, and permanent versus temporary affiliation with household units, that were based in the everyday, commonplace experience of domestic and economic relationships within inner city black communities.(5)

The field study

In 1991, an ethnographic field study was conducted in the Washington, DC, area to investigate Hispanic respondents' definitions of household membership and residency. The largest Hispanic group within the area is Salvadoran, and this was the group chosen for study.

The field study, conducted by a team of Hispanic and non-Hispanic researchers, began with a 3-month period of participant observation. This included interviews with official and unofficial community leaders to learn about the social and economic organization of Salvadoran and non-Hispanic white households.

The research called for interviewing all household members, aged 15 years and older, in local Salvadoran and non-Hispanic white households with similar socioeconomic characteristics. Many adults in both groups worked as house cleaners, or in office maintenance or construction jobs. In addition, the households selected for the study represented the typical types of households found among Salvadorans: nuclear family (father, mother, and children), two nuclear families living together, households headed by women, and households consisting of several single men.

After 3 months of field study, respondents for the household interviews were recruited from local church congregations. Salvadoran respondents were also recruited from a free language school in the community. In addition, a community outreach worker introduced the researchers to the resident manager of an apartment complex that housed documented as well as undocumented Salvadoran immigrants.

The resident manager was an excellent source of information on household composition and residential mobility in the Salvadoran community. Through him, the researchers were introduced to households, and were vouch-safed to household members, whom it would not otherwise have been possible to interview.

Prospective respondents were informed that every member of the household 15 years of age or older would be interviewed about recent work experience, that the last names of household members were not necessary, and that information would be strictly confidential. In addition, each person interviewed would be paid $20 and the subsequent interviews would be conducted in Spanish.

The first Salvadoran respondent recruited said that there were three persons in her household--herself, a son, and a daughter. On the night of the interview, another woman was observed sitting in the family's living room. When asked if this other woman also lived in the house, the daughter replied, "She is just staying here." When asked how long the woman had been "staying" in the house, the daughter answered, "Two years." The daughter explained that the woman was "staying" rather than "living" in the household because the woman's daughter was expected to arrive from Panama in the next few months and the woman would then live with her daughter. The woman in question readily agreed to be interviewed along with the other household members.

On subsequent visits to other Salvadoran households, the existence of additional household members would come to light well after the interviews had begun. In late night interviews especially, one or two adults, usually men, would arrive home about 10 p.m. The respondents did not attempt to hide the presence of their boarders. Rather, the boarders were usually peripheral to the life of the household, and their presence in the household was not spontaneously reported. The practice of not reporting boarders was not necessarily an attempt to protect undocumented individuals. In one household, the two male boarders were documented, and one worked in a county government-sponsored job.

Table 1 shows the distribution of men and women 15 years of age and older who were reported, and those not reported but observed, in the 17 Salvadoran and 15 non-Hispanic white households surveyed. There were no boarders and no unreported persons in the non-Hispanic white households. In the Salvadoran households, 11 percent of the men and 3 percent of the women observed to be living in the households were not reported by the respondents. The Salvadoran households consisted of almost equal numbers of men and women even though 6 of the 17 households interviewed were headed by women. The fact that six of the eight boarders in the households were men accounts for the almost equal male-female distribution. life of the Salvadoran immigrant household. This is best illustrated by the following excerpt from a respondent's account of the changing composition of her household:

I moved to this house with my daughter, my mother, brother, brother's wife, and their daughter, early in 1979. After 6 months, my father came for a visit from El Salvador and returned to El Salvador with my daughter. My stepmother, 2 half-sisters, and a nephew came from El Salvador to stay in my house in 1981. Two months later, an uncle and a brother-in-law came from El Salvador to stay with us. In October of 1981, my 2 half-sisters, nephew, step-mother, and brother-in-law moved to Chicago.

We asked if any of the relatives who were described as "visiting" from El Salvador for 1 or 2 months did any paid work while in the United States. Respondents replied that it was necessary for these visiting relatives to work to cover their expenses during their visit. The CPS does not ask about the activities of persons who are visiting in the household.

Alternative Enumeration research

In 1990, the Bureau of the Census conducted an alternative enumeration to evaluate behavioral factors contributing to census undercount.(6) The project consisted of ethnographic surveys of 29 ethnic minority communities throughout the continental United States and Puerto Rico that were expected to have high census omission rates. The researched communities, which were predominately black, Hispanic, or Native American, or composed of recent immigrants from Asia, were characterized by a high percentage of foreign-born persons, non-English speaking households, relatively low household incomes, and individuals who worked in blue-collar occupations.


An alternative enumeration, conducted in each research site by experienced ethnographers in the 3 months following census day, was compared with the 1990 census count. A comprehensive set of match and resolution procedures allowed the ethnographers to determine the enumeration status of every individual and housing unit located in each of the sites on the day of the 1990 census.

Alternative Enumeration data provided the basis for a second phase of the research on the pattern of underreported boarders found in the field study of Salvadoran households. Analysis of Alternative Enumeration data from Hispanic communities would further corroborate or refute the field study finding in a larger sample, even though census data collection methodology differs from that used in the CPS. (The decennial census relies, to a large extent, on self-enumeration; in contrast, CPS data are collected by interviewers.)

Nine of the 29 sites were predominantly Hispanic areas. The rates of within household, whole household, and total omissions among Hispanics in the nine predominantly Hispanic sites included in the Alternative Enumeration Study are summarized by Manuel de la Puente.(7) (See table 2.) The total omission rates ranged from 5 percent for New Orleans, Louisiana, where community outreach organizations had vigorously promoted Hispanic participation in the census, to 66 percent for Long Island, New York, where landlords had illegally subdivided large, single family homes into as many as 14 dwelling units.

de la Puente analyzed the demographic characteristics of individuals who were correctly enumerated or missed by the census in the nine predominantly Hispanic sites. Along with higher omission rates for men and for young adults, there was a sharp increase in the omission rates for individuals as the relationship to the household head becomes more remote: 17.8 percent for nuclear family relatives; 35.6 percent for non-nuclear family relatives; and 44.8 percent for those who were not related to the head of the household.(8)

An analysis of the effect of household relationship status on both number and type of omissions in the 1990 census was conducted for eight of the nine predominantly Hispanic sites in the Alternative Enumeration data base. (The data from Santurce, Puerto Rico, were not included in the analysis to eliminate possible factors not relevant for surveys of mainland U.S. populations.) The analysis shows that for total omissions, boarders and housemates/roommates have the highest proportion (approximately 50 percent) of missed individuals. (See chart 1.)

Chart 2 divides the number of persons missed in the 1990 census (shown in chart 1) into "whole household" omissions, where the entire household was missed, and "within household" omissions, where individuals who resided in the households were missed. The pattern that emerges in chart 2 expands on de la Puente's finding that census omissions increased as the relationship to the household head moves from close relative, to distant relative, to nonrelative. We observed that another phenomenon accompanied this increase in omission rates: there was a shift from a predominance of "whole household" missed for close family members to "within household" missed for other relatives and nonrelatives of the household head.


Table 3 compares omission rates by "within household" and "whole household" between relatives and nonrelatives in the eight survey sites. The difference in distribution between the two omission rates is significant at p <.001.

According to the findings of the field study of the Salvadorans, the household members who were not reported on the household rosters were boarders who were peripheral to the social life of the household core. Pursuing the line of reasoning that an individual's likelihood of being missed as a "within household" omission increases as ties to the household head become more remote, we reasoned that boarders--whose ties to the household head are weaker than those of housemates or roommates--would have the highest rate of "within household" omissions in the nonrelative categories.

Table 4 supports this reasoning by comparing omission rates between boarders and housemates/roommates, for "within household" and "whole household" omissions.

In the eight survey sites, the majority (91 percent) of the boarders missed by the census were "within household" omissions; by contrast, fewer than half (42 percent) of the housemates/roommates missed by the census were "within household" omissions. The difference between boarders and housemates/roommates in percentage of "within household" omissions in significant at p < .001. Thus, it appears that it is the boarder who contributes most to the greater tendency of non-relatives to be missed as "within household" omissions.
Table 3. Within household and
whole household
misses for relatives
and nonrelatives,
aggregate of eight
sites, 1990
 Within Whole
 Household Total household household
relationship misses misses
 Total 390 189 201
Relatives 305 124 181
Nonrelatives 85 65 20
NOTE: A chi square test, measuring the
difference between relatives and nonrelatives
in the distribution of within household misses
and whole household misses, resulted in
[X.sup.2] = 34.139, df = 1, p <.001.
Table 4. Within household and
whole household
misses for boarders
and housemates/roommates,
aggregate of eight
sites, 1990
 Household Within Whole
relationship total household household
 Total 81 61 20
Boarders 55 50 5
 roommates 26 11 15
NOTE: A chi square test, measuring the
difference between boarders and housemates/roommates,
in the distribution of within household
misses and whole household misses,
resulted in [X.sup.2] = 22.426, df = 1, p <.001.
 A third category of nonrelatives, termed
"Other nonrelatives" was excluded from this
analysis because of the small number (4) in this


Evidence from a field study of Salvadoran and non-Hispanic white households suggests the two groups manifest cultural differences in household composition, concepts of household membership, and household reporting patterns. The field study found that Salvadoran households tended not to report boarders as household members. This findings was tested and confirmed in a larger sample of eight Hispanic communities. Analysis of this larger data base also revealed that a complex set of relationships exists between household relationships status and the rate and type of census omissions.

In addition to de la Puente's finding of a sharp increase in omission rates for "other relatives" and nonrelatives of the household head, we observed a striking shift in proportions of "whole household" omissions and "within household" omissions as the kin relationship to the household head diminished. Persons missed by "whole household" omissions tended, by and large, to be relatives of the head of the household. In contrast, nonrelatives of the household head tended to be missed as "within household" omissions. A more detailed analysis of the type of omission for the household relationship statuses included under "nonrelative" revealed that it is the boarders alone who account for the predominance of "within household" omissions.

The pattern of respondents not reporting their boarders was found to be widespread across ethnically diverse Hispanic communities with demographic characteristics predictive of high census omission. A cultural explanation for the omission of the boarders was also identified.

These findings have important implications for improving Hispanic coverage in surveys. Boarders are being missed primarily in Hispanic households which are responding households. Hispanic respondents are not reporting those individuals who they perceive as nonhousehold members. Recent recommendations to improve Hispanic coverage in the census and other surveys call for greater community outreach efforts and for providing incentives to individuals for fuller reporting.(9) Our research suggests that consideration must also be given to linguistic bridge-building between definitions of household membership used in surveys and by Hispanic respondents.


(1) Gary Shapiro and Paul Bettin, "Coverage In Household Surveys," Paper presented to the Census Advisory Committee of the American Statistical Association and the Census Advisory Committee on Population Statistics at the Joint Advisory Committee Meeting, October 1992.

In 1986, the undercoverage rate was 20 percent for Hispanics, compared with 6 percent for non-Hispanic whites. The authors note that CPS undercoverage is in addition to decennial census undercoverage, and adding census undercoverage yields an undercoverage rate in the CPS for Hispanics of "probably worse than 31 percent for Hispanics males."

(2) The CPS is conducted by the Bureau of the Census for the Bureau of Labor Statistics.

(3) Shapiro and Bettin, "Household Surveys."

(4) Peter Hainer, Catherine Hines, Elizabeth Martin, and Gary Shapiro, "Research on Improving Coverage in Household Surveys," Proceedings of the Census Bureau Fourth Annual Research Conference (U.S. Bureau of the Census, 1988), pp. 513-39.

(5) Elizabeth R. Gerber, "Calculating Residence: A Cognitive Approach To Household Membership Judgments Among Low Income Blacks," Paper presented to the American Association of Public Opinion Research Meeting, 1990.

(6) Leslie A. Brownrigg and Elizabeth A. Martin, "Proposed Study Plan for Ethnographic Evaluation of the Behavioral Causes of Undercount," Paper presented to the Census Advisory Committee of the American Statistical Association and the Census Advisory Committee on Population Statistics at the Joint Advisory Committee Meeting, 1989, Alexandria, VA.

(7) Manuel de la Puente, "An Analysis of the Underenumeration of Hispanics: Evidence From Hispanic Concentrated Small Area Ethnographic Studies," Proceedings of the Bureau of the Census 1992 Annual Research Conference (U.S. Bureau of the Census, 1992), pp. 45-69.

(8) de la Puente, "An analysis of underenumeration."

(9) Shapiro and Bettin, "Household Surveys."
COPYRIGHT 1993 U.S. Bureau of Labor Statistics
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Current Population Survey
Author:McKay, Ruth B.
Publication:Monthly Labor Review
Date:Sep 1, 1993
Previous Article:Measuring education in the Current Population Survey.
Next Article:Developments in aerospace.

Related Articles
Job losses among Hispanics in the recent recession.
Acculturation Erodes The Diet Quality Of U. S. Hispanics.
Michigan Changing: Diversity in Hispanic Populations in Southwest Michigan. (Geography).
Retirement expenditures for Whites, Blacks, and persons of Hispanic origin: the relationship between income and expenditure level reveals differences...
Antibiotic use in Hispanic households, New York City.
In 2002, 15 percent of U.S. households (16 million) experienced one of more violent or property crimes as measured by the National Crime...
Hispanic suicide rate lower than national rate.
Direct mail, gracias.
Hispanic restaurant consumers and acculturation trends.
Study: 54% of all Hispanics in Miami-Dade are now Cuban.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters