Content area
Abstract
Census undercount is arguably the most difficult public policy issue ever to confront the U.S. Census Bureau. Serious disagreements exist over how best to reduce disproportionately high rates of omission of the poor and of racial and ethnic minorities in the census.
In Chapter 1 of this dissertation, I argue that the search for a solution to the census undercount problem should be informed by theoretically-grounded research on the causes of undercount. In Chapters 2 through 5, I commence the task with an investigation of the social factors causing high minority omission rates.
A population's social, cultural and economic attributes exert basic constraints on the effectiveness of census operations. In Chapter 2, I identify two "fundamental" causes (poverty and ethnicity) and five "proximate" social factors (complex living arrangements, respondent inability to report accurately, housing unit characteristics, perceived costs and benefits of reporting) which could explain high minority omission rates.
In Chapter 3, I test whether the disproportionate omission of racial and ethnic minorities in the 1980 census was due to high minority poverty rates. Data are from the 1980 Post-Enumeration Program, which matched Current Population Survey records to the 1980 census to identify census omissions. I find that income and ten other socioeconomic variables together explain 60% of the average difference between the proportion of whites missed in the census and the proportions missed in six racial and ethnic minority groups. I speculate that a substantial portion of the residual 40% is due to unmeasured group-specific factors.
In Chapter 4, I assess the social sources of omission in 1986 Los Angeles test census, also using post-enumeration survey data. I identify a number of non-economic factors associated with ethnicity which appear to have constrained enumerability.
The number and complexity of social causes of census omission suggest that it is unlikely that race-ethnicity differences in omission rates can be greatly reduced through census operational innovations. I elaborate in Chapter 5 and note some implications of current social trends for census coverage in 1990. In Chapter 6, I offer suggestions for research on the upcoming 1990 census.
