Skip to main content

Measuring Neighborhood Quality With Survey Data: A Bayesian Approach


The goal of Cityscape is to bring high-quality original research on housing and community development issues to scholars, government officials, and practitioners. Cityscape is open to all relevant disciplines, including architecture, consumer research, demography, economics, engineering, ethnography, finance, geography, law, planning, political science, public policy, regional science, sociology, statistics, and urban studies.

Cityscape is published three times a year by the Office of Policy Development and Research (PD&R) of the U.S. Department of Housing and Urban Development.


Volume 12 Number 3

Measuring Neighborhood Quality With Survey Data: A Bayesian Approach

Brent D. Mast

The contents of this article are the views of the author and do not necessarily reflect the views or policies of the U.S. Department of Housing and Urban Development or the U.S. government.


Although neighborhood quality is important for shaping public policy, it is also difficult to quantify. This study measured subjective neighborhood quality using data from two sources: (1) the 2002 American Housing Survey (AHS) and (2) the U.S. Department of Housing and Urban Development's (HUD's) Customer Satisfaction Survey (CSS) of Section 8 Housing Choice Voucher Program (HCVP) households. Survey responses were analyzed regarding neighborhood quality, home quality, and crime perceptions. Tract-level Bayesian estimates were computed using AHS metropolitan-level data and CSS census tract data.

The new Bayesian estimates have fewer outliers than the original CSS data, and the use of prior information allows for estimation for tracts with lower sample sizes than would be practical to estimate using only CSS data.

I compared the CSS and Bayesian estimates with other measures of neighborhood quality, such as poverty rates, median income, and indicators for tracts receiving low-income housing tax credits. The CSS and Bayesian indicators are highly correlated, and both the CSS and Bayesian estimates correlate well with the auxiliary variables used in this study. For tracts with large differences between the CSS and Bayesian estimates, correlations are much stronger for the Bayesian estimates.

Previous Article   |   Next Article