Using Successive Censuses to Reconstruct the African-American Population, 1930-1990

The Census Bureau's program to estimate the completeness of decennial census counts for age, sex, and race groups relies principally upon what it terms "demographic analysis." The essence of this approach is to introduce extraneous information on the number of births, deaths, and migrations, derived from non-census sources, to estimate the true size of each birth cohort at the time of a census (Robinson et al., 1993; Himes and Clogg, 1992). Comparison of this alternative estimate to the census count provides an estimate of the degree of under - or over-enumeration in the census, often termed the census undercount. Acceptance of the estimated undercount implies that the census itself is irrelevant to estimating the true size of the population; whatever deficiencies it contained would be accurately and completely revealed by comparison to the estimate based on demographic analysis.