In the United States, a collection of methods have historically been used to develop estimates of the population for comparison with decennial census counts. These estimates are developed from various types of demographic data in order to build a historical accounting of population change. The term Demographic Analysis (DA) has been used to refer to this approach for assessing the quality of the census. For the 2010 Census, the U.S. Census Bureau is again using the DA estimates along with results from operational indicators and a post-enumeration survey to assess the quality of the decennial census.

In past decades, the Census Bureau has produced the DA estimates at the national level for single-year birth cohorts by sex and two broad race categories, Black and non-Black. Reflecting the contributions of many demographers, two distinct methodologies have evolved over time for separately estimating the population under 65 years of age and the population 65 years and over. Administrative records on births, deaths, and estimates of international migration are used to estimate the population under 65. Estimates of the population 65 and over are developed from data on Medicare enrollment and estimates of the number of those not enrolled. The DA estimates have been limited to the national level and the Black and non-Black race categories because of the limitations of the available historical data.

This paper describes the development of the 2010 DA population estimates. In order to show the sensitivity of the DA estimates to the use of different but plausible input values, alternative estimates for each component of the DA methodology were developed. Specific combinations of alternative inputs were then used to produce five series of DA estimates by single year of age, sex, and the DA race categories for use in the demographic analysis of the 2010 Census counts.  The specific combinations of alternative estimates were selected to communicate the uncertainty around the DA estimates and include a low series, three middle series, and a high series. All series were considered to be plausible. These estimates were produced without knowledge of the 2010 Census results and will provide an essentially independent basis for assessing the accuracy of the 2010 Census counts.