It is essential to be able to accurately represent the target population in order to produce unbiased estimates. In both the Decennial Census and the Bureau’s demographic surveys, there is evidence of significant coverage problems for some segments of the population. Under-coverage occurs when a person eligible for the Census (or survey) does not have a chance to be represented or included in the Census, either intentionally or unintentionally. Historically, there are coverage problems with low income and inner-city communities, with household surveys being worse than the Decennial Census. This is of major concern with the American Community Survey (ACS) data since the Bureau’s long range plan is to possibly replace long form Census data with ACS data and provide local area long form data on an ongoing basis.

In an effort to identify any possible coverage problems associated with the ACS, a research project was initiated to compare coverage in the 1996 ACS to the 1990 Decennial Census by looking at the distribution of household size by various demographic characteristics and mode of response. From those results, we hope to identify possible causes, such as forms designs and field and processing procedures, that might contribute to the under-coverage and suggest further research and testing.

Other research projects will address within household coverage, look at residence rules, suggest alternative rostering and questionnaire design, assess the impact of nonresponse on coverage, evaluate the completeness of data for persons from large households, and experiment with methodologies to improve whole household coverage in frames for sampling.