Sick Leave and Retirement

Just under twenty percent of the US workforce is employed by federal, state, or local governments. Most of these employees receive a pension upon retirement. As both the size and age of the retired population has increased, these pensions have become increasingly burdensome, particularly for state budgets. As a result, legislators across the US are rethinking and redesigning fringe benefit schemes for government employees.

The paper presented in this Mannheim Applied Seminar uses individual-level panel data on sick leave and retirement decisions to study optimal fringe benefit design and time preferences among the largest subset of state and local employees in the US – public school teachers. The sick-leave scheme faced by these teachers is very common in the US, but unique among other developed nations. Teachers earn sick-leave credits every year that they work and unused credits accumulate over time. When teachers fall ill and cannot work, they “spend” these credits to avoid wage penalties. Importantly, unused credits carry substantial financial value upon retirement. As such, in this paper the authors design and estimate a dynamic structural model that features an agent (i.e., a teacher) who makes a sequence of sick-leave decisions over the course of her career while forming expectations over future uncertain events (i.e., illness shocks). Her career then ends with a retirement decision. The key tension in the model arises from the fact that teachers like current time off, but also like future earnings, which must be sacrificed in order to take a sick day.

They use data on teachers from the Scott County School System in central Kentucky to estimate the unknown parameters of the model. Among the parameters estimated is a discount factor, which measures preferences for current versus future benefits. Discount factors are not typically separately identified from preference parameters in dynamic structural models; however, a unique feature of the pay schedule in Scott County enables identification. The first significant contribution of this work is the estimation of this discount factor using observational data.

Once the unknown parameters of the model are estimated, the model is used to simulate sick leave and retirement behavior under counterfactual fringe benefits designs. For example the authors’ model allows them to test how teacher sick leave and retirement behavior responds to (i) an increase/decrease the amount paid to teachers per accumulated sick day upon retirement, (ii) an increase/decrease in the number of sick days allowed per year, and (iii) an increase/decrease the cap on accumulated sick days. This counterfactual analysis provides the second major contribution of this work, as the existing literature is only able to evaluate the impact of historical policy changes that have actually occurred.