Exploration of the Use of Empirical Bayes Procedures for Estimating Changes in Occupancy Rate and Persons per Household

The Census Bureau is carrying out research with the intent of improving the housing unit-based method for estimating population totals. One approach is to use the Decennial Census as the baseline for measures and update them annually using American Community Survey (ACS) data to estimate their change since the census year. This study looks at the possibility of using an Empirical Bayes approach to produce county estimates of change with smaller variances than direct ACS estimates. Results from national and state models are compared. Data from the 1990 and 2000 long form samples are used to represent ACS and data from the 1990 and 2000 short form are used as independent variables in the models. In the actual application, sources for the latter would be the Master Address file and Administrative Records.