"In Search of Farmworkers Missing From the 1990 Census"May/June 1993 issue of Poverty & Race
by Edward Kissam
The Decennial Census of Population has always been the paradigm of official statistics. The neat rows and columns cross-tabulating the population characteristics of the United States showily assert their scientific elegance and independence-their "definitive" nature as the picture of a nation, the mirror in which a community comes to know itself. Yet, at the same time, Census data are not used simply, or even primarily, for scientific reflection, but as the basis for planning and allocating funding among scores of federal, state, and local social programs.
To La Cooperativa Campesina de California, a statewide network of organizations serving migrant and seasonal farmworkers, it was clear that something was wrong with the Census-based statistical picture they saw.
The official 1980 Census portrait of farmworkers showed three-quarters of the nation's farmworkers to be white, when community service providers knew that most farmworkers were minorities, and showed only 5% of the farm labor force to be migrants. The 1980 statistical picture of California showed only one quarter of the "official" farmworkers to be poor-a different reality from the one they knew. Even more disturbing, in a political environment where program funding is driven by numbers, the distorted picture of the nation's farmworkers suggested that huge numbers of migrant and seasonal farmworkers had not been counted in the Census and that this gross undercount would make it harder to advocate for them.
In 1989, La Cooperativa Campesina and California Rural Legal Assistance (CRLA) joined forces to make sure the census picture which emerged from the 1990 Decennial Census would not have the same flaws. With support from the Rosenberg Foundation and PRRAC, they began an ambitious project to (a) discover the true extent of the Census undercount of farmworkers, (b) advocate the use of "best available" data, and (c) find ways to overcome the statistical travesty of the Census profile of farmworkers.
Census Bureau Response
La Cooperativa's review of the research on Census undercount showed widespread official recognition of the existence of a differential undercount of minorities and ambitious plans within the Census Bureau to correct the minority undercount through use of a statistical methodology known as "dual system estimation." In 1988, several major cities-New York, Los Angeles, Chicago and Miami-had joined with groups such as the NAACP and LULAC to sue the Census Bureau to prevent it from dismantling its efforts to measure the undercount via a Post-Enumeration Survey (PES) which would form the basis for dual system estimation of the minority undercount. The plaintiffs' case was bolstered by testimony from the former Associate Director of the Census Bureau, Dr. Barbara A. Bailar, whose affidavit detailed a politically based decision to overrule the Bureau's professional staff so as to make Census adjustment for minority undercount infeasible. A settlement was reached between the plaintiffs in The City of New York et al. v. U. S. Dept. of Commerce et al. and the Census Bureau, and the PES took place as scheduled. With PES data, it was expected that later adjustment of Census data to correct the minority undercount would be possible.
However, in 1991, the Census Bureau announced that, although there was evidence of differential undercount of minorities, it had decided not to officially adjust the 1990 Census data. The Bureau's decision was based on an argument that the statistical basis for adjustment allowed only an imperfect adjustment and that this course of action would create as many problems as it solved. There still appeared to be a political element in this decision, since the problem which would be solved was the under-representation of Blacks, Asians, and Latinos in states such as California and the problems which would be created would be diminished political power in low-minority Republican areas such as Kansas, Iowa, and other Midwestern states.
At the same time, the La Cooperativa/CRLA project team had some problems of its own with the possibility of a PES-based adjustment of Census figures. On the one hand, while the overwhelming number of migrant and seasonal farmworkers belong to an ethnic minority group, they are not simply another minority. They are, instead, a distinct class, a population which is socially and economically disadvantaged not only on the basis of race but by virtue of working in one of the most marginal occupations in the nation. Still of more concern was the fact that the Census undercount seemed to vastly exceed the undercount revealed in PES-based estimates, which tended to show systematic undercounts only in the 5-l0% range.
Estimating the Undercount
In July, 1991, La Cooperativa released a report entitled Out in the Cold. The Causes and Consequences of Missing Farmworkers in the 1990 Census, providing evidence that 1990 Census data on farmworkers (which had not yet been released) would probably omit at least 40% of the nation's farmworkers. This report was important in the context of Census politics for two reasons-the magnitude of the undercount, and recognition of the need not only for reliable estimates of overall numbers of farmworkers but for reliable demographic and socioeconomic profiles of the population (e.g., a true estimate of the numbers of farmworkers living in poverty).
La Cooperativa's estimate of a minimum 40% farmworker undercount appeared three months after the Census Bureau released preliminary PES-based estimates showing that the differential undercount of Blacks and Hispanics could only be in the range of 2.5% to 4.8%. La Cooperativa's undercount estimate of 40%+ suggested that what was involved was not a statistical aberration but a gross, fundamental failure of the Census system to provide reliable data on the nation's farmworkers.
Estimating the number of farmworkers in the United States, or in a state, county, or local community, on the basis of Census data is fundamentally different from estimating numbers of minorities, because data on race and Hispanic origin are collected via the Census short-form, the "100% sample" of the U.S. population. Because farmworkers are an occupationally defined group, their numbers can only be estimated on the basis of "long-form" Census data, using a far smaller sample. In October 1992, the fast Census data on occupation became available-in the form of tabulations referred to as the EEO (Equal Employment Opportunity) tape, since the dataset shows the numbers of women and minorities in different occupational classifications.
When these data became available, La Cooperativa discovered that its earlier estimate of the farmworker undercount had, in fact, been conservative.
The EEO data indicated that the 1990 Census had counted and identified only about one out of three farmworkers in the leading farmworker states of Florida, Texas, and California-a mega-undercount in the order of 60-70%, indicating that two out of three of the nation's farmworkers had been statistically erased.
In February 1993, another Census data product became available-the Public Use Microdata Sets (PUMS), a 5% population sample allowing data analysts to examine Census data in greater detail and in different ways than would be possible simply from reviewing published tabular data. The availability of PUMS data made it finally possible for La Cooperativa to look not only at Census data on numbers of farmworkers but also to examine the demographic and socioeconomic profile of farmworkers identified in the Census to assess its reliability. A meta-analysis of California PUMS data in comparison to National Agricultural Worker Survey (NAWS) data and recent ethnographic research on farmworkers completed in mid-April 1993 both confirmed the EEO-based estimate of a 60-70% undercount of farmworkers and revealed the existence of serious sample bias in the Census profile of farmworkers.
Evidence for the Undercount
What is the evidence for a gross undercount of farmworkers? Ironically, the evidence comes in large measure from research sponsored by the Census Bureau, most importantly from research during a 1986 test census in Los Angeles which included a survey research component designed to identify the causes of and count. Based on this Causes of Undercount Survey (CUS), David Fein (then sociology graduate student) developed a regression model of Census undercount incorporating not only traditional co relates of undercount, but also a range of community "structural" factors, including prevalence of non-English speaker in a community, immigration flows, an the existence of "unusual housing units” to explain undercount.
Because the CUS analytic model addresses not only race and poverty factors related to Census undercount but also a wide range of social factors which make a population "socially invisible," provides a particularly appropriate mode for understanding farmworker undercount and, more broadly, Census undercount of an increasingly multicultural, multi-lingual populace in major immigrant states such as California and Florida. La Cooperativa, working no, in collaboration with Dr. Susan Gabbard of Aguirre International, the Project Director of the National Agricultural Worker Survey, and using Fein's m to analyze NAWS data, demonstrate that 48-52% of the nation's Latino farm workers are likely to be missed in the Census. Moreover, the CUS model predicts that the bulk of omissions stem from the fact that an entire farmworker household is omitted from the Census. This analysis is compelling because program providers' experience and ethnographic research make it clear that farm workers live under the worst housing conditions in the nation-in camps of cardboard houses in the canyons of San Diego county, in barns, toolsheds, broken-down vehicles and trailers, garages, backyards, and in single-family dwellings illegally housing fifteen to twenty migrant workers.
Not surprisingly, the CUS model also explains the extent to which a population which speaks little English (94% with limited English-speaking ability) is likely to be missed in an English-language mail survey because the decennial Census is now conducted primarily by mail. Spanish speakers can get a Spanish-language form, but only by calling an 800 number; other recent research indicates that even if farmworkers receive a Spanish-language Census long-form, most would have substantial difficulties completing it, since required levels of functional literacy are those most commonly found among persons with 9 or more years of schooling, while farmworkers' mean educational level is about 5.6 years.
Evidence of undercount stems as well from the skewed demographic and socioeconomic profile of farmworkers in Census data. This evidence, comparing the English-speaking ability of Census-profiled farmworkers against recent high-quality ethnographic and survey data, suggests that the Census data have omitted 70-80% of the Spanish-speaking farmworkers in California, a finding closely in line with the CUS-based analysis. Farmworkers profiled in Census data are much more likely to speak English better than the actual population, hold more stable jobs, have higher hourly earnings, live in better housing, and to migrate less than the actual farmworker population.
La Cooperativa also compared the Census count of California farmworkers with estimates of farmworker population based on several different administrative data sources, including analysis of unemployment insurance records (which, in California, show how many workers worked in any given industry) and crop production data. These comparisons clearly show that the numbers of farmworkers identified in the Census could not possibly have harvested all of California's crops or generated the number of hours of employment recorded from agricultural workers.
In summary, La Cooperativa, by developing a meta-analysis of farmworker census undercount, using both different analytic strategies and different (and independently generated) datasets, has definitively demonstrated that no less than half and, more probably, two-thirds of the nation's farmworkers are invisible in decennial Census data. About half of the nation's Latino farmworkers (i.e., about 35-40% of all farmworkers in the nation, since 75-80% are Latino) are omitted from the Census altogether. Another 20% are missed in the Census day "snapshot" of the population on April 1 or misidentified because they are making ends meet by working temporarily in another kind of job which "blocks" their being identified as farmworkers. The cumulative undercount of farmworkers, including those missed and those who are misidentified, thus is in the 60-70% range.
Impact of the Undercount
What does this mean for farmworker well-being? Plenty. A range of federal and state programs specially designed to help farmworkers-health, children's education, housing, job training, food stamps-are designed and funded based on what is officially reported about this population. Some $640 million in federal funds now goes for these programs. More rational planning for larger appropriations depends on knowing true numbers and true characteristics.
What does this mean for Census data politics? It likely means that the dual-enumeration methodology for assessing minority undercount provides an inadequate measure of undercount for many populations. If so, even minority advocates have underestimated the true extent of the Census data problem.
Analyses of Census undercount, and social inequality, based only on race tend to understate the situation of the poorest and most marginal sub-populations among ethnic minorities. For example, California Census data identify only 20 individuals out of 20,000110,000 Mixtec farmworkers in California, a population sub-group known to live in some of the worst housing accommodations in the state, known to hold the least stable jobs, and isolated by language from access to effective social services. Similarly, Kanjobal-speaking farmworkers of Mayan origin are likely to be obliterated from the "official" picture of South Florida's farm labor force, making it difficult to convince reluctant local, state, and federal officials of their need for services.
Research on the causes of Census undercount of farmworkers has important implications for many different immigrant groups throughout the nation. For the Census Bureau, a far-reaching conclusion may be that radically new approaches are needed to accurately capture the ethnic and linguistic diversity of the United States in the 21st Century. Promising strategies include replacing large-scale "standardized" surveys with targeted ethno-surveys involving members of minority communities themselves in developing the sampling frame and research questions of investigations, as well as in conducting field research. Whatever social science strategies are adopted for policy-oriented research and program planning, a key element will be responsiveness to the cultural pluralism of the United States.
What advocacy use now will be made of this research? CRLA and the California Human Development Corp. believe this is an appropriate matter for litigation. Administrative advocacy with the U.S. Department of Commerce also is being pursued: specifically, a demand to do a mid-Decennial Census to collect more accurate data. And legislative remedies are being sought as well, via the House Subcommittee of the Census and individual legislators. There is a crying need to remedy the damage done to this vulnerable group by shamefully biased official statistics gathering.
|Poverty & Race Research Action Council | 740 15th St. NW, Suite 300, Washington, DC 20005|
©Copyright 1992-2018 Poverty & Race Research Action Council