Measuring Neighborhood Change Using Postal and Housing Choice Voucher Data: Results from a Pilot Analysis of Four Metropolitan Areas in Washington, D.C. and Ohio

Alena Stern
Urban Institute

This article presents the results of a pilot effort to model neighborhood change in near real-time by supplementing time-lagged demographic data from the American Community Survey (ACS) with realtime U.S. Postal Service (USPS) and Housing Choice Voucher (HCV) data. The author first defines and measures three key types of neighborhood change—gentrification, decline, and inclusive growth—in the selected metropolitan statistical areas (MSAs). She then uses machine learning methods to create a model that identifies neighborhood change at the census tract level. The model identifies neighborhood change with 76 percent accuracy and 76 percent precision; that precision exceeds models trained with just ACS data or just USPS and HCV data. The model is strong at predicting neighborhood decline and less accurate at identifying gentrifying neighborhoods. These results suggest a promising application of the USPS and HCV data to model neighborhood change.