Comparisons and extensions of reduced form and structural approaches to retirement decision of Missouri public school teachers
There have been growing calls to reform teachers' pension systems. Designing of pension reforms requires modeling retirement decisions first. My thesis compares different models of teachers' retirement decisions. To further predict behavioral responses to changes in pension rules, I estimate a structural model of retirement decisions (the option value model by Stock and Wise (1990) by maximizing likelihood (ML). In chapter three, I focus on a number of technical issues in ML estimation. I show that the commonly used frequency simulator for likelihood evaluation is computationally costly when the data set has a large number of teachers and long panels. In addition, the maximum obtained by a hill-climbing algorithm may not be global when the objective function is ill-shaped. Therefore, I propose to evaluate the option value model's likelihood by using the GHK simulator, in place of the frequency simulator, and obtain robust estimates via simulated annealing instead of the hill-climbing algorithm. The results suggest that the GHK simulator is much more efficient than the frequency simulator and that the aggregate behavior that is predicted on the basis of the estimates of the GHK simulator is reasonably robust. The last chapter compares the t of the probit model with the option value model and discusses the pros and cons of the two competing models for teachers' retirement decisions.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License.