Abstract:
|
The American Housing Survey (AHS) and American Community Survey ask respondents when their housing unit was built (year built). In both surveys, year built has a high missing rate and is not missing at random, which impacts total survey error. Yet, property and tax assessment data also contain year built. This whitepaper presents discussion and analysis on using property tax data to impute year built. We begin with a discussion of year built as a survey construct and a field in property tax data and establish the availability of year built from property tax data is sufficient for this use. We then lay the evidentiary foundation for using property tax data to replace missing year built via cold-decking. We do this by comparing aggregate distributions of year built, evaluating individual-level correspondence, and comparing hot-deck imputation for year built with the proposed cold-decking. We then evaluate an imputation method using the cumulative distribution function for respondents who do not report year built and cannot be matched to a property tax record. We conclude by discussing findings and the decision to use year built from property tax data to impute missing AHS responses.
|