636 – Real-World Approaches to the Knotty Problems of Outliers, Faulty Values, and Covariates in Complex Sampling Designs
Self-assessed Housing Values in the American Community Survey: An Exploratory Evaluation Using Linked Real Estate Records
W. Kingkade
U.S. Census Bureau
This investigation uses a linked dataset of occupied housing records from the 2009 ACS and information on the same housing units in a commercial database of national scope obtained from Core Logic Inc. The database, derived largely from county-level administrative records, contains data on the characteristics of the property, including measures of property value. With the linked dataset, the characteristics of the residents can be related to the registered characteristics of the property. The analysis examines the relationship between the self-reported value obtained from the ACS householder and measures of value derived from the administrative records by methods including Weighted Least Squares Regression with suitable controls. The analysis is limited to single family owned homes, for which category ACS respondents' estimates of value are most straightforward. The match rate is 80 percent of available ACS records for such homes. The statistical analysis of these data relies on the R Survey package (Lumley, 2012) for analysis of complex survey samples. The package accommodates the replicate weights used for variance estimation from the ACS.