Summary: The number of congressional representatives a state receives, as well as district boundaries, are determined by a decennial census. A number of communities, contending that the 1980 census has missed large numbers of persons and that this undercounting will result in their citizens being shortchanged in political representations and federal fund distributions, have taken legal action to have the census count adjusted for the undercount. The Census Bureau dropped a postenumeration survey for estimating 1980 census errors at subnational levels, so that preliminary estimates would be available sooner. GAO reported on the Census Bureau's ability to develop accurate undercount estimates for subnational levels and the effect of dropping a planned post-census survey on the prospects for developing accurate undercount estimates.
Estimates of census errors at subnational levels are needed for correcting census data. Statistical analysis might help in estimating census error at subnational levels. To date, however, the agency had not set up a formal program of applied research for this purpose. The Bureau has primarily used two coverage evaluation techniques, the demographic method and matching studies. These techniques, as implemented by the Bureau, do not provide estimates at geographical levels compatible with census data user needs. The demographic method suffers from limited data, notably the absence of reliable estimates for the size of certain age groups and for the number of illegal aliens. Matching studies, both reinterview studies and reverse record checks, involve the comparison of a list of sample persons or households to the census. Matching techniques have major weaknesses. People tend to respond to a reinterview study in the same way they do to census enumeration. Persons missed by the census are likely to be missed by the reinterview study, and incomplete or invalid data prevent the Bureau from resolving whether a person has been omitted from the census. Matching is a time consuming process because it requires manual case-by-case comparison. Although other statistical methods of analysis for estimating coverage are available, the Bureau has made little use of them. In many communities, an increase in their census count will not proportionately increase their share of federal funds.
