Association for Public Policy Analysis & Management

We combine national administrative data on subsidized housing participation and adult earnings with information on HOPE VI funded public housing project demolitions to test whether demolitions (and the consequent residential moves) affect the long-term earnings of resident children. The data enable us to identify children who are between the ages of 10 and 18 at the time of a demolition from 160 HOPE VI projects and over 5,000 non-HOPE VI projects, and to observe how their subsidized housing participation, neighborhood characteristics, and earnings evolve over time. The uncertainty about how exposure to housing project demolitions affects the well-being of residents is particularly troublesome given that there is no clear theoretical prediction for the direction or magnitude of the effect. Households in demolished projects are exposed to potentially significant disruptions: adjustment costs in new neighborhoods and weakening of ties to existing social networks. Conversely, since a goal of HOPE VI was to provide opportunities to move to tenant-based voucher housing, they may experience improved housing choice, possibly enabling them to move to higher-quality neighborhoods. Further complicating any prediction about the relationship between exposure to HOPE VI and household member well-being is the considerable heterogeneity in the projects, neighborhoods, and metropolitan areas targeted by the demolition grants.

Since HOPE VI demolitions are drawn from the lowest-quality public housing projects, the core of our methodology consists of finding comparable lower-quality projects for comparison purposes. This is achieved by using data for a large number of other public housing projects to control for differences between the HOPE VI demolitions and the wider set of public housing projects. We use stratification with regression (Rosenbaum and Rubin, 1983, 1984) to classify projects into strata based on a project-level propensity score, estimate within-stratum treatment effects using individual-level ordinary least squares regressions, and aggregate up the stratum-level results to produce an estimate of the overall treatment effect. We provide primary results for eventual earnings at age 26 relative to earnings of children in comparable non-demolished projects, along with comparing effects of HOPE VI demolitions in large versus smaller Public Housing Authorities. We then shed light on the underlying mechanisms that may determine earnings outcomes at age 26. We first examine short-term housing effects (voucher, other public housing, and other housing) of the program. We then turn to potential changes in neighborhood quality. In addition to standard neighborhood measures, we exploit the richness of our longitudinal data to construct a destination census tract variable based on the age 26 earnings of all youth aged 10-18 living in any given tract as an “opportunity” measure of that tract. We also exploit additional heterogeneity in HOPE VI projects and in demographic characteristics to explore how the treatment effect differs across diverse contexts, along with other pathways to earnings outcomes.