Dynamic Models of Labor Supply and Retirement

This dissertation contains three separate essays on the dynamic models of labor supply and retirement. The first essay documents "sharp retirement"--retirement accompanied by a discontinuous decline in labor supply--across three data sets, which previous literature found difficult to explain. I propose and estimate a life-cycle labor supply model with habit persistence wherein sharp retirement can be explained by workers quitting "cold turkey." In much the same way that one might quit smoking, workers with accumulated "working habit" exit the labor force with a pronounced, discontinuous decline in labor supply. The working habit model is consistent with the data, where workers reduce yearly labor supply by scaling back more in hours worked per week (over 50% reduction) than in weeks worked per year (20% reduction). The fixed costs approach, which has been the standard model used to understand sharp retirement, cannot explain these trends. After estimating the model, counterfactuals show that reducing Social Security benefits by 20% causes individuals work an additional 8.6 months. Individuals choosing sharp retirement respond mostly on the extensive margin by delaying retirement eight months, while individuals choosing smooth retirement respond mostly on the intensive margin by increasing yearly labor supply and delaying retirement only one month. The second essay develops and estimates a Ben-Porath human capital model in which individuals make decisions on consumption, human capital investment, labor supply, and retirement. The model allows both an endogenous wage process (which is typically assumed exogenous in the retirement literature) and an endogenous retirement decision (which is typically assumed exogenous in the human capital literature). This integration is important to obtain unbiased estimates, which are critical for most counterfactual analysis. For instance, when evaluating the effect of increasing the Social Security Normal Retirement Age (NRA) on workers' labor supply and retirement decisions, not only does one have to consider how the policy change affects the retirement decision directly, one also needs to consider how it affects the wage process and therefore affects retirement indirectly. We estimate the model using the Method of Simulated Moments to match the life-cycle profiles of wages and hours from the PSID data. Counterfactuals of delaying NRA and removing Social Security earnings test are conducted. We find significant increases in one individual's human capital investment at old ages, which leads to over 20% increase in the wage profile near retirement. Finally, the third essay tests for asymmetric employer learning in the labor market using a three-period model with a match component of wages. When a worker makes her quit/stay decision in a labor market with three periods, she must consider the signaling effect of her decision in subsequent periods. This breaks down some implications derived from two-period models, which are mostly used in the empirical literature. The unconditional quit rate is not necessarily negatively connected with ability in this three-period asymmetric learning model. I suggest two alternative hypothesis tests for asymmetric employer learning in the model. The first test scrutinizes the negative relationship between conditional quit rates and abilities. The second test examines the evolution of weighted average within-group ability variation. Under this model, the variation should decrease over one worker's career history due to sorting on ability. I use the NLSY79 Work-History data and find evidence of asymmetric employer learning from these tests.

Copyright - Copyright ProQuest, UMI Dissertations Publishing 2012 Last updated - 2013-04-18 First page - n/a

normal retirement age

69244