Developing PIK Mobility Scores for the Enhancement of the Person Characteristic Frame

This CES technical note provides an overview of the development of PIK Mobility Scores for the enhancement of the Census Bureau’s Person Characteristic Frame (PCF). PIK Mobility Scores are the predicted probability that a PIK observed in a given MAFID in year y will be found at the same MAFID in year y+1. These scores are developed using the enterprise Demographic Frame (Demo Frame) and use its Person-Place model to place PIKs in MAFIDs on a given reference date. To model PIK mobility, we supplement the Demo Frame by appending data from the United States Postal Service on change of address filings, the Internal Revenue Service on income, Black Knight on home ownership, the Master Address File on place characteristics, the American Community Survey on tract characteristics, the Planning Database on tract response rates, the Federal Emergency Management Agency on disaster declarations, and the Bureau of Economic Analysis on cost of living. Then, we use logistic regressions to model move probabilities and compare these scores to observed mobility outcomes. PIK Mobility Scores accurately predict whether a PIK is a mover or non-mover for more than 80% of PIKs. About 95% of PIKs predicted to be non-movers actually do not move. About half of PIKs predicted to be movers actually move. Thus, these scores are best for use cases that rely on identifying PIKs that remain in the same MAFID year-over-year, such as In-Office Enumeration. We conclude by discussing opportunities to improve future versions of these scores and related projects.