The goal of this work is to simultaneously estimate the value of staying in a job, the value of leaving either to retire or accept a new job and the exit behavior of older workers. Unlike previous studies, which have used proxies for the value of a job, the objective of this study is to measure job values my numerical calculation of the implicit dynamic problem solved by workers. The resulting dynamic programming algorithm is embedded within an iterative maximum likelihood routine in order to solve for parameters of wage functions and retirement values. Job changing and retirement probabilities must be estimated jointly with job values as job switching behavior causes a truncation problem in the estimation of job values. This truncation is caused by workers prefering high value activities to low value ones. Hence values of observed activities are necessarily greater than alternative ones. The value of retirement and alternative jobs in turn determine transition probabilities revealing the necessity of joint estimation. Once the model is estimated it can be used in many other applications related to the work environment of older workers. These include estimation of the effects on retirement behavior of changes in the Social Security system, Medicare, and other government programs important to older workers and analysis of employer behavior towards older workers.
