Risk Aversion & Retirement Decisions: Using Policy Variation to Identify and Estimate a Structural Model of Retirement

Dayanand Manoli, UCLA
Kathleen Mullen, RAND
Mathis Wagner, University of Chicago

Abstract

How do individuals' retirement decisions respond to changes in retirement benefits? Are the results from structural and reduced-form methods for estimating labor supply responses to changes in retirement benefits consistent with one another? We develop a standard dynamic model of retirement decisions and show that the model implies a relationship between the coefficient of relative risk aversion and labor supply responses to retirement benefits. This relationship indicates that can be estimated via two approaches: (1) a reduced-form approach that relies on identification of labor supply elasticities and (2) a structural approach that explicitly models the micro-foundations of the economic environment. We use administrative data from the Austrian Social Security Database and exploit variation from five pension reforms in Austria between 1984 and 2003 to implement both approaches. Using Cox proportional hazards specifications, we estimate reduced-form income and price elasticities of roughly 0.43 and -2.90 respectively. Next, we implement the widely-used structural approach of estimating based on matching retirement patterns across ages. We show that, even though they are unified under the same theoretical model, the two approaches yield different estimates of risk aversion because they are based on different sources of identifying variation. We propose a solution to reconcile these differences while exploiting the policy variation for identification based on estimation via the method of Indirect Inference.