High item nonresponse rates on the American Community Survey (ACS) are a major concern for the accuracy of housing estimates used in policymaking and distribution of federal funds. Commercial data and administrative records can provide valuable, low-cost information to adjust survey estimates for nonresponse. This research will study how CoreLogic commercial housing data, aggregated from a variety of data sources nationwide, can be used to improve ACS estimates for such topics as property values, real estate taxes, year structure built and number of rooms. The estimation approach does not treat the CoreLogic data as a “gold standard.” Instead, I propose using CoreLogic indirectly in model-based imputation for ACS nonrespondents. The research will develop estimates that can be used to study biases in current ACS estimates. Finally, the research will investigate the sensitivity of findings to different assumptions about nonresponse patterns by comparing point identified estimates to the partially identified interval estimates discussed in Manski (2009).
|Effective start/end date||7/1/15 → 6/30/16|
- Bureau of the Census (YA1323-15-SE-0097)