«BY THE CENSUS: EVIDENCE FROM SMALL AREA ETHNOGRAPHIC STUDIES by Manuel de la Puente, Ph.D. U.S. Census Bureau Center for Survey Methods Research ...»
USING ETHNOGRAPHY TO EXPLAIN WHY PEOPLE ARE
MISSED OR ERRORNEOUSLY INCLUDED BY THE CENSUS:
EVIDENCE FROM SMALL AREA ETHNOGRAPHIC STUDIES
Manuel de la Puente, Ph.D.
U.S. Census Bureau
Center for Survey Methods Research
This article demonstrates how ethnographic research provides much needed insight into the social and cultural precesses that contributed to the differential undercount of ethnic and racial minorities in the 1990 Census of Population and Housing. The article summarizes and synthesizes data and findings reported in 29 ethnographic reports sponsored by the U.S. Census Bureau and non-profit organizations throughout the country. These reports are based on ethnographic research conducted by qualified ethnographers in 29 sample areas throughout the continental U.S. and Puerto Rico. The article reports that the reasons for the differential undercount of ethnic and racial minorities are varied and complex. Empirically and in the aggregate, there is no single reason why a disproportionate number of ethic and racial minorities were not counted by the 1990 census. Rather, there are a constellation of factors that interact and contribute to the differential census undercount. These factors are listed and discussed in the article. The article concludes with recommendations to the Census Bureau for the improvement and conduct of the year 2000 census.
INTRODUCTIONIn order to better understand the reasons behind the net differential undercount and other types of census errors the Census Bureau initiated the Ethnographic Evaluation of the Behavioral Causes of Census Undercount (hereinafter referred to as the Ethnographic Evaluation) under its Research Evaluation and Experimental Programs for the 1990 Census.
The Ethnographic Evaluation took place in 29 ethnographic sample areas throughout the continental U.S. and Puerto Rico. An Alternative Enumeration (AE) was conducted in each of the 29 sample areas by experienced ethnographers working with the Census Bureau under Joint Statistical Agreements. The AE was then compared to the 1990 census count. As a result of this evaluation a total of 29 separate coverage reports (one per ethnographic sample area) have been written by the ethnographers who conducted the field research. These reports contain a wealth of qualitative information concerning causes of the differential net census undercount (and other census errors) among the respective ethnic and racial groups on a site by site basis.
This report assembles and summarizes reasons why census omissions and erroneous inclusions occurred in the 29 sample areas of the Ethnographic Evaluation. This information comes from the coverage reports, one for each sample area.1 Data analyzed in these reports were collected by experienced ethnographers using participant observation, direct observation, ethnogarphic interviews and other unobtrusive data collection methods.2 Census omission is one type of census error. Other kinds of errors that occur during the conduct of a census include misgeocoding, erroneous inclusion of persons or housing units and duplication, that is, counting persons or housing units more than once. While these mistakes often result in missed individuals, other errors have different negative impacts on the census count.
These negative or unwanted impacts can result in net undercount or net overcount. Both outcomes are undesirable because they diminish the accuracy of the census count. In the Ethnographic Evaluation we witnessed both net overcounts and net undercounts across all 29 sample areas.
Some net undercount/overcount rates come from CSMR tabulations and Brownrigg and de la Puente (1993).
For more information on the data collection methods used in the Ethnographic Evaluation see Brownrigg and de la Puente (1992 and 1993) and de la Puente (1991).
CAUSES OF CENSUS OMISSIONS AND ERRONEOUS INCLUSIONSBrownrigg and Martin (1989) in their proposed study plan for the Ethnographic
Evaluation identified five hypothesized causes of coverage error. These were:
The ethnographers used these hypothesized causes as a starting point and systematically examined these hypotheses (and other related hypotheses) using unobtrusive ethnographic methods. In general, they found that reasons why individuals were missed or erroneously enumerated by the 1990 census are varied and complex. In almost all sample areas no single cause for census omission or erroneous inclusion was noted by the ethnographers. Rather, ethnographers listed a constellation of factors that interact and contribute to the differential net census undercount/overcount. These include irregular household arrangements, irregular housing, little or no knowledge of English (and in some cases illiteracy in any language), fear of government on the part of sample area residents (leading to concealment of information), and missed or erroneously censused housing units.
It is important to keep in mind that it is often difficult to separate the relative influence of competing causes for why people are missed or erroneously included by the census. For example, irregular and complex household structure is one of the chief causes of within household omission. (Interestingly, single person households were also missed in a number of sample areas even though the housing unit was counted by the census.)3 However, in many situations other confounding influences are at play such as concealment of information by household members, residential mobility, household members who have little or no knowledge of English or cultural practices and beliefs that define "household" differently from the Census Bureau's definition of household. Ethnographers were often not able to specify which one of these causes prevailed or was most important in their sample area.
Another case in point is irregular housing. This cause of census error can interact with other causes such as concealment of information on the part of sample area residents. As I shall discuss and illustrate below, in at least two sample areas, landlords or property managers of buildings with illegally converted housing units kept this information from the Census Bureau.
This resulted in missed housing units. However, irregular housing can also be a cause of For example see Hamid (1992).
erroneous census enumerations. In fact, this occurred in a number of sample areas and resulted in net census overcounts.
Irregular and Complex Household Arrangements One key reason why individuals are missed or are erroneously included by the census in households (within household omission or within household error) is irregular and complex household structure.4 In irregular and complex households, members cannot be easily related to person 1 on the census form and may not be listed on the census roster. Alternatively, household members that should be included may be excluded resulting in an enumeration error. Still another possibility is the inclusion of all household members but because the household item on the census form requires that household members be listed in relation to person 1, the complexity of unconventional households is not fully captured on the census form. Although, technically speaking, this is not an erroneous enumeration it, does distort the actual household structure and the interrelationships among its members.
The above mentioned situations arise when one or more of the following conditions apply.
Household members do not understand census rules of residence. This can happen when household members have little or no knowledge of English or they are semi literate or illiterate.
Another reason is the disjunction between how "household" is defined by the Census Bureau and how "household" is defined by respondents.
In general, irregular and complex households have one or more of the following features:
(a) unrelated individuals, (b) mobile or ambiguous household members, (c) households formed for the sole purpose of sharing the rent and/or other living expenses or (d) households that contain two or more "nuclear" families.
Irregular and complex households as a contributing cause of census omission or other census error were noted, to a greater or lesser extent, in virtually all coverage reports. However, the reasons why some households were complex, with erratic or irregular membership, were varied.
Complex households were common in sample areas populated with recent immigrants5, Other reasons for within household omission include concealment of information, residential mobility and a misunderstanding of the Census Bureau's definition of household and Census Bureau residence rules. These reasons for within household omission will be discussed in later sections.
Sample areas with recent Hispanic immigrants include: Houston, TX (Rodriguez and Hagan 1991), rural Santa Barbara County, CA (Garcia 1992), Santurce, PR (Duany 1992), San Francisco, CA (Romero 1992), San Diego, CA (Velasco 1992), rural Marion County, OR especially Hispanic immigrants. Martin Dale Montoya, the ethnographer who conducted the field work in rural Marion County, OR coined the term "Ad Hoc Household" to describe the complex households that he encountered among Mexican migrant workers. Montoya defines Ad Hoc
Household as a household arrangement that generates and maintains:
"....relationships which can only be described as loosely tied, ephemeral, and alienated (no responsibility to household) because each slot in the household is allocated by money and not necessarily kinship. Housemembers come and go as they please with little concern for the housing unit itself, individual household members or groups." (Montoya 1992: 7).
Montoya argues that Ad Hoc Households are formed as a response to poverty. Household members come together not out of familial sense but rather as a practical response to poverty and
a lack of affordable housing. Ad Hoc Households are very difficult to enumerate. Montoya notes:
" In the Ad Hoc household, if all members are not present, the likelihood of obtaining the data pertaining to persons outside, asleep, at work, or temporarily absent is virtually impossible. It is as if those persons do not exist. However, even when the number of housemates is determined or provided, the personal data for those other persons is still unattainable. This is because Ad Hoc households protect their identity. This means that coverage of the Ad Hoc household will be determined, to a great extent, by coincidence (who is actually present during the visit) and/or the perseverance of the enumerator." (Montoya 1992: 7).
This type of household was also reported by other ethnographers in sample areas populated by predominately Hispanic immigrants. For example, in the San Diego sample area,
populated mainly by Mexican immigrants, Alfredo Velasco notes:
Similar observations regarding unrelated males sharing the same housing unit were made in other sample areas with sizeable numbers of recent immigrants. One such sample area was located in the south Bronx in New York City. Boanerges Dominguez conducted his research in (Montoya 1992), Long Island, NY (Mahler 1993) and Bronx, NY (Dominguez 1993). Sample areas with recent Haitian immigrants are: Fort Lauderdale, FL (Wingerd 1992) and Miami, FL (Stepick 1992). And sample areas with Asian immigrants are: Chicago, IL (Straus 1991), Koreatown, CA (Kim 1991), Chinatown, NY (Sung 1991), Long Beach, CA (Bunte 1992), North Beach, CA (Shaw and Guthrie 1992), South St. Louis, MO (Rynearson 1992) and Queens, NY (Kang 1992).
this sample area. The population in this sample area consisted of immigrants from Mexico and the Dominican Republic. Dominguez also noted that Mexican immigrants tended to reside in overcrowded apartments with beds lining the living room walls. He found that the interaction among the unrelated household members was minimal. Since the men worked different shifts it was difficult to catch all of them in the apartment at one time. According to Dominguez, getting one household member to divulge information about other members of the household was close to impossible (Dominguez 1992: 9).