Doctor of PhilosophyDepartment of EconomicsDong LiThe first essay tests whether the returns to education are different between entrepreneurs and regular employees. If the signaling model of education is correct, entrepreneurs should receive lower returns from education (relative to employees) because they have no need to signal their productivity to an employer. However, this result should only hold if the researcher is able to control for selection into self-employment and the endogeneity of ed- ucation. This is illustrated using a stylized model of signaling. The relationship between self-employment and the returns to education is tested using data from the 1996 Survey of Income and Program Participation. This rich panel dataset makes it possible to control for many business-specific characteristics, like business equity, that have been previously unaccounted for in the literature. Ordinary least squares regressions find the correlation between education and earnings to be weaker for entrepreneurs. To control for selection, I utilize a Heckman selection model using spousal health insurance and housing equity as instruments. It shows that selection biases downward the correlation between education and income for entrepreneurs. Finally, a fixed effects model is employed to control for any time invariant unobserved heterogeneity. This approach indicates that education is as valu- able, if not more valuable, to entrepreneurs as it is to employees. This does not support the signaling hypothesis. The finding is robust to different measures of entrepreneurial earnings. The second essay explores whether unemployed workers make successful transitions into self-employment. It is well established that unemployed workers are more likely to transition into self-employment than individuals coming from paid employment. A growing body of literature suggests that these formerly unemployed entrants tend to exit self-employment earlier than typical entrants. It is tempting to attribute this result to differences in ability between the two groups. However, using an adapted version of Frank (1988)’s Intertemporal Model of Industrial Exit, I show that this is not the case. In this model, entrants to self- employment receive noisy information about their true entrepreneurial ability from their earnings in the market. I show that low ability entrants to entrepreneurship should be no more likely to exit self-employment than high ability entrants to self-employment. This is because although low ability entrants will earn less as entrepreneurs, their outside wage in paid employment will also be proportionately lower. Survival in self-employment, therefore, is a function of how initial expectations match reality. This leads me to suggest that the high exit rates out of self-employment for the formerly unemployed may be because this group systematically overestimates their entrepreneurial ability at entry. This hypothesis is justified by evidence from the psychology literature that low ability individuals tend to overestimate their performance. Duration analysis on data from the 1996 and 2001 panels of the Survey of Income and Program Participation confirms that the formerly unemployed are more likely to exit self-employment. I also find preliminary evidence consistent with the hypothesis that the unemployed overestimate their likelihood of success in self-employment. These findings should give policymakers pause before incentivicing the unemployed to enter self-employment.