Population undercount and social science research

The undercount problem in the decennial census has important implication for social science research based directly or indirectly on census data. Because undercount rates (or coverage rates) vary by age, race, residence, and other factors typically studied in social research, important conceptual difficulties arise in using census results to corroborate sampling frames or to validate survey results. Differential undercount, particularly for analyses based on small areas, could produce substantial variability in prevalence rates in cases where the denominators for those rates are derived from the census. Several examples where the undercount problem arises in social science research, including survey research, are considered. The adjustment problem-whether to adjust, how to adjust, and how much to adjust-is also considered from the point of view of social sicence research.