Essays in Labor Economics
This dissertation contains three independent chapters on topics in labor economics. In the first chapter, I examine the sources of decline in the job-finding rate over the spell of unemployment. In particular, I distinguish between dynamic selection and the adverse effect of longer unemployment durations. In the second chapter, Carter Bryson and I explore the role of broad sectoral shifts in labor demand in explaining the divergence of employment outcomes of Black and White men during the second half of the 20th century. Finally, in the third chapter, I propose a nonparametric estimator for the discrete-time version of the Mixed Proportional Hazard (MPH). In the first chapter, "Duration Dependence and Heterogeneity: Learning from Early Notice of Layoff,'' I disentangle the sources of decline in the observed job-finding rate over the spell of unemployment. It is possible that an individual worker's likelihood of exiting unemployment declines with the duration of unemployment (e.g., due to employers discriminating against long-term unemployed). However, workers also differ in their employability, their desperation to find a job, or perhaps their ability to look for jobs. Such heterogeneity across workers, which is often unobserved, implies that the observed job-finding rate declines even when an individual worker's exit probability does not. As the more employable workers exit early, the still unemployed workers are increasingly the less employable ones. A long literature tries to disentangle the role of structural duration dependence from heterogeneity, but it has proved to be quite challenging. I develop and implement a novel approach to answer this question by leveraging variation in the length of notice an individual receives from their previous employer before being laid off. The key idea behind my approach is that workers with longer notice start searching earlier and are more likely to exit unemployment early. If workers are heterogeneous, then the composition of long-notice workers should be worse at later durations as the more employable workers from this group have already left. This observation enables me to pin down the extent of heterogeneity and estimate structural duration dependence. My estimates imply that there is substantial heterogeneity in the likelihood of exiting unemployment across workers. I find that roughly 40% of the decline in the observed job-finding rate over the first five months is due to dynamic selection. Further, an individual's exit probability increases up to unemployment insurance (UI) exhaustion and remains constant after that. This is in contrast to the observed exit rate, which continues to decline even after benefit exhaustion. Recently, researchers have tried to explain this decline in the observed exit rate after UI exhaustion. For instance, it can be rationalized with storable offers (Boone and van Ours, 2012) or reference-dependent utility (DellaVigna et. al., 2017). My estimates suggest that most of the decline in the observed exit rate after UI exhaustion is due to changes in the composition of workers. In the second chapter which is joint work with Carter Bryson, "Understanding the Racial Employment Gap: The Role of Sectoral Shifts,'' we quantify the extent to which sectoral reallocation can explain the divergence in employment outcomes of Black and White men during the last three decades of the 20th century. Using a shift-share strategy, we document that local employment-to-population ratios for Black men are relatively more responsive to local labor demand shocks. We also document substantial population responses for both groups of workers. Finally, we provide a framework incorporating frictional unemployment and imperfect mobility across locations to aggregate these local responses. We find that sectoral shifts can explain roughly half of the observed exacerbation in the employment-to-population ratio differential between Black and White workers from 1970 to 2000. Furthermore, our findings indicate that the increase in the differential due to sectoral shifts results from the greater responsiveness of Black workers to local labor demand shifts rather than a higher concentration of these shifts in areas or sectors with a higher share of Black workers. In the final chapter, "The Discrete-Time Mixed Proportional Hazard Model'', I propose a nonparametric estimator for the discrete-time MPH model. Hazard models of event durations are widely employed in economics to analyze unemployment spells, retirement decisions, and an array of other topics. As the findings from the first chapter highlight, ignoring unobserved heterogeneity while analyzing duration data can lead to inaccurate inferences. The MPH model explicitly accounts for such heterogeneity but estimating this model can be challenging. I set up a discrete-time MPH model and propose an estimator for it that is based on the Generalized Method of Moments and is easy to implement.