Using the December 1998 and August 2000 CPS Computer and Internet Supplements matched with subsequent CPS files, we ask which types of unemployed workers looked for work on line and whether Internet searchers became reemployed more quickly. In our data, Internet searchers have observed characteristics that are typically associated with shorter unemployment spells, and do spend less time unemployed. This unemployment differential is however eliminated and in some cases reversed when we hold observable characteristics constant. We conclude that either Internet job search is ineffective in reducing unemployment durations, or Internet job searchers are negatively selected on unobservables.

In this work I present three essays related to the rising number of female college graduates relative to men, intergenerational mobility in education, and the prevalence of overeducated workers during the great recession.

In the first chapter co-authored with Associate Professor Aashish Mehta (UC Santa Barbara), we ask whether shifting male and female employment patterns can help to explain why the US college boom between 1981 and 2005 was dominated by women. We make three contributions. First, we show that while a massive feminization of high-wage, high-skill occupations plausibly contributed to the female college boom, general, structural movements of labor (undifferentiated by gender) from industrial work into education-intensive services should have encouraged male rather than female college attendance. Previous work has suggested that both types of employment shifts would have contributed to the female college boom. Second, we show that women’s occupational upgrading was too large and ubiquitous to be explained by their growing educational advantage. This is consistent with a causal connection running from gendered employment trends to a female college boom. Third, we show that gender specializations in many occupations deepened, with college educated women gravitating towards jobs offering institutionally protected wages.

In the second chapter, I estimate the intergenerational transmission of education in the United States between 1980 and 2013. I find that intergenerational persistence in education has increased substantially among blacks in recent years while remaining stable among whites and Hispanics. I observe this trend when using data from the Panel Study of Income Dynamics as well as the National Longitudinal Surveys of Youth. I demonstrate that much of the increase in educational persistence among blacks is due to decreases in upward mobility. The increase in black educational persistence is found in both two-parent and single-parent households, and I do not find similar trends and differences when estimating intergenerational income persistence.

In the third and final chapter, I use the method introduced by Gottschalk and Hansen (2003) to analyze the rate of overeducation among workers with exactly a college degree between 2006 and 2013. To my knowledge, this is the first study to use this method to analyze trends in overeducation during the great recession in the U.S. I find that the proportion of workers with exactly a college degree working in occupations offering low college premiums increased during great recession and fell afterwards. An increase in the rate in overeducation could be due to more college-educated workers working in noncollege occupations that were noncollege in the past or because there was an increase in the number of noncollege occupations. I show that changes in the rate of overeducation are mostly due mostly to the latter. When shutting the down the flexibility for occupations to change from college to noncollege (and vice versa), the rate of overeducation increases only slightly between 2006 and 2013. Regardless, these findings run contrary to the secular decline of the rate of overeducation during the end of 20th century documented by previous research.

Vulnerable workers, workers who have recently experienced a shock that could adversely affect their labor market prospects, experience large, long-lasting earnings losses -- on average. This dissertation investigates the mechanisms behind the losses of three groups of vulnerable workers and the role of public policy in mitigating these losses. In the first essay, I identify which displaced workers, workers who lose their job as a result of a firm or plant closing, are the most vulnerable. I find that a worker's duration of joblessness depends much more on conditions within that worker's occupation than conditions within that worker's industry. This suggests a worker's vulnerability is a function of their skills and less related to the goods and services they were previously producing. In the second essay, my collaborators and I estimate the causal impacts of benefits in California's Paid Family Leave program on a second group of vulnerable workers: new mothers. We find no evidence that a higher weekly benefit amount increases leave duration or leads to adverse future labor market outcomes for mothers with earnings near the maximum benefit threshold. In the third essay, my collaborators and I find strong evidence that Disability Insurance and Paid Family Leave program take-up is substantially higher in firms with high earnings premiums. Our results suggest that changes in firm behavior have the potential to impact social insurance use and thus reduce an important dimension of inequality in America.

This dissertation consists of three essays that use and develop econometric methods to causally investigate topics in education and development economics. In the first chapter, I develop an econometric framework to correctly and efficiently draw inference in models where estimated value-added (VA) is an explanatory variable (and models where it appears as the dependent variable). Estimated VA measures have become increasingly popular metrics of worker and institutional quality, and they are now widely used in regressions by researchers seeking to establish links between worker quality and a broad range of outcomes. Although consistent standard errors are crucial to obtain correct confidence intervals and assess the validity of conclusions drawn by studies using VA measures in regressions, the literature has not yet tackled this issue. I contribute to this literature by setting up an econometric framework that allows me to show why naïve standard error estimators are inconsistent in such models, derive consistent standard error estimators, propose a more efficient estimator for models using VA measures as explanatory variables, and propose a testable condition under which naïve standard errors are consistent for models that use VA measures as dependent variables. Then, in an application using data from North Carolina public schools, I find that the increase in standard errors resulting from the required correction that I propose is larger than the impact of clustering standard errors. In the second chapter, based on joint work with Serena Canaan and Pierre Mouganie, we use VA measures to provide the first causal evidence on the impact of college advisor quality on student outcomes. To do so, we exploit a unique setting where students are randomly assigned to faculty advisors during their first year of college. We find that having a higher grade VA advisor reduces time to complete freshman year and increases four-year graduation rates by 2.5 percentage points. It also raises high-ability students' likelihood of enrolling and graduating with a STEM degree by 4 percentage points. The magnitudes of our estimated effects are comparable to those from successful financial aid programs and proactive coaching interventions. We also show that non-grade measures of advisor VA predict student success. In particular, advisors who are effective at improving students' persistence and major choice also boost other college outcomes. Our results indicate that allocating resources towards improving the quality of academic advising may play a key role in promoting college success. In the third chapter, based on joint work with Clement de Chaisemartin, we consider the case of a randomized controlled trial with individual-level treatment assignment, and we allow for individual-level and cluster-level (e.g. village-level) shocks to affect the units' potential outcomes. We show that one can draw inference on two estimands: the ATE conditional on the realizations of the cluster-level shocks, using heteroskedasticity-robust standard errors; the ATE netted out of those shocks, using cluster-robust standard errors. We show that by clustering, researchers can test if the treatment would still have had an effect, had the stochastic shocks that occurred during the experiment been different. Then, the decision to cluster or not depends on the level of external validity one would like to achieve.

In the Dutch Postcode Lottery a postal code (19 households on average) is randomly selected weekly, and prizes--consisting of cash and a new BMW--are awarded to lottery participants living in that postal code. On average, this generates a temporary, unexpected income shock equal to about eight months of income for about one third of the households in a typical winning code, while leaving the incomes of nonwinning, neighboring households unaffected. We study the responses of consumption and reported happiness of both winners and nonwinners to these shocks. Consistent with simple models of in-kind transfers, the overwhelming majority of households who won a BMW convert it into cash. With the exception of food away from home, the only 'own' effects of cash winnings we detect are on durables expenditures and car consumption; these results support a version of the permanent income hypothesis in which durable spending is used to smooth consumption. We detect social effects of neighbors' winnings on two types of consumption: cars and exterior home renovations. Six months after the fact, winning the lottery does not make households happier, nor do neighbors' winnings reduce happiness.