This dissertation estimates the impact of two geographically targeted economic development programs on the employment of people living in the targeted areas. This question is difficult to address for a number of reasons. Unlike in most program evaluation problems, the process that determines the outcome of interest (resident employment) happens at a lower level of aggregation than the process that determines selection for treatment. Therefore, standard program evaluation techniques have to be modified to address this issue. The programs I study, the enterprise zone programs of California and Florida, were designated at a very detailed level of geography, making it hard to measure the location and the characteristics of the zones.

I develop a methodology to address the unusual selection process of these programs. The first step of the methodology is to create a neighborhood-level measure of the component of residents' employment probabilities explained by the neighborhood that is conditional on the characteristics of area residents. To do this, I estimate the component of employment probability correlated with residential neighborhoods, which I call tracts' conditional employment probabilities. The next step is to estimate the effect of enterprise zones on resident employment by comparing the conditional employment probabilities of neighborhoods containing enterprise zones with those of comparable areas. I do this with tract-level propensity score matching. I also carefully measure the location and attributes of enterprise zones.

I find that a substantial portion of the variation across neighborhoods in employment rates can be explained by controlling for the attributes of residents. This indicates that it is important to control for resident characteristics when making cross-neighborhood comparisons. Using propensity score matching, I find a large pool of non-zone tracts that are observationally similar to tracts containing enterprise zones. I use these non-zone tracts to create an estimate of what the conditional employment probabilities in zone tracts would have been in the absence of the programs. Even though I focus on two very targeted and generous enterprise zone programs, I find no evidence that the programs impacted the employment of zone residents.