UC Berkeley

The purpose of this dissertation is to examine the effects of environmental conditions on graduation and dropout rates in the United States. Neighborhood effects studies are not uncommon, however they typically fail to account for serious selection issues that potentially bias results. The contribution of this dissertation is to offer an alternative approach, natural experiments, to deal with such methodological problems. The dissertation consists of one theoretical chapter and two analytical chapters. In chapter one, I establish the econometric difficulties of estimating neighborhood effects and outline how a natural experiment can obviate these problems if applied correctly. This chapter sets up the rest of my dissertation, which includes two empirical chapters using major urban riots and county mass layoffs as natural experiments.

In chapter one, I describe the econometric difficulties of estimating neighborhood effects within the context of youth educational well-being. I outline the three levels of selection bias, individual, family and school, that researchers normally confront when attempting to examine the effects of community conditions on education outcomes. While most studies attempt to minimize individual and family level bias, they often neglect to account for school level context. Doing so will lead to biased results because model parameters will fail to reflect the lack of independence between units due to clustering within schools and the deep selection issues caused by the interwoven relationship between neighborhoods and schools. I then outline the conditions under which a natural experiment can solve these problems. This chapter sets the stage for the two empirical articles that make up the rest of my dissertation.

In the first analytical chapter, I examine the effects of large urban riot shocks to city wide conditions on aggregate dropout rates. Using decennial census data, I examine the 1960s civil rights riots on city level school non enrollment rates. I find that the 1960s urban riots had negative short- and long-term consequences on school enrollment rates. In order to test the robustness of these results to a more contemporary setting, I estimate the effect of the 1992 Los Angeles riots on city and census tract level dropout rates. I find evidence of a short-term effect, but no long-term impact.

In the second empirical chapter, I estimate the effects of diminished county conditions due to mass layoff occurrences in the U.S from 1996 to 2010 on county level graduation and dropout rates. Although I find no evidence of a negative single year effect on schooling persistence rates, I uncover lagged and cumulative effects. Specifically, I find that a large two-year lagged shock has a negative effect on current graduation rates. I also find that minor and large four-year average shocks have a significant effect on graduation and dropout rates. These results indicate that point-in-time measurements may not capture the temporal effects of neighborhood disadvantage, especially for processes like schooling persistence that may not be sensitive to minor, short-term shocks.