An Evaluation of Housing Unit and Block Cluster Eects on Small Area Census Coverage Variability in the 2006 Census Test

Models are developed for the correlation effect of block clusters and housing units on the capture/recapture model of Census coverage measurement using data from 2006 Census test. The purpose of understanding this possible correlation is two-fold: 1) as noted by Malec and Maples (2005) a model for within small area variability is needed because design based estimates may be imprecise. Also, using 2000 census coverage information, Keller (2008) has shown evidence of both variable census capture rates and correct enumeration rates between block cluster and 2) the heterogeneity effects of between correlation due to block clusters and housing units on estimates of coverage have not been thoroughly evaluated. Random effects and the choice of a random effects model will be evaluated as a tool for measuring variability and correlation at these small levels.