Intertemporal Substitution in Retirement Decisions: Evidence from Policy Discontinuities

by Dayanand Manoli, University of California, Los Angeles

This project proposes to use a combination of administrative data and policy discontinuities to examine intertemporal substitution in retirement decisions. The policy discontinuities arise because of features in the Austrian pension system. Specifically, if an individual accumulates at least ten years of tenure by retirement, the Austrian pension system mandates that the employer must make a lump-sum payment to the individual at the time of his/her retirement equal to one-third of the last year’s salary. If the individual accumulates less than ten years of tenure by retirement, no payment is made. If the individual accumualtes at least 15, 20, or 25 years of tenure by retirement, the payment amounts are stepwise increased up to one full year of salary. This rule therefore discontinuously increases wage rates just prior to the thresholds. For example, the wage rate for the tenth year of tenure is discontinuously higher than other years since individuals will receive their standard annual wage rate plus the payment when they complete the tenth year. Futhermore, the wage increases can be fully anticipated at earlier years of tenure. The combination of administrative data and changes in anticipated financial incentives presents a unique setup to gain insights in the determinants of retirement entry that are relevant for several countries, including the United States.

Our research strategy is based on examining how individuals’ retirement decisions respond to the anticipated changes in financial incentives at the tenure thresholds. We first present graphical evidence documenting individuals’ labor supply responses to the policy discontinuities. We examine behavior before and after these tenure thresholds to determine if individuals delay their retirements in response to the anticipated wage variation. The graphical evidence indicates excess retirements just after the thresholds and reduced retirements just prior to the thresholds. We then use a standard labor supply model to develop a semiparametric estimator for the elasticity of intertemporal substitution in labor supply based on the graphical evidence. Once the elasticity is recovered, we aim to use the estimated elasticity to calibrate a model of retirement decisions. The calibration allows us to simulate a variety of hypothetical pension reforms that are currently being considered in the United States and many other countries. Furthermore, we are able to use policy simulations based on the calibrated model to evaluate the cost effectiveness of providing incentives for individuals to delay their retirements using lump-sum retirement benefits as opposed to annuitized benefits.

Back to 2010 Sandell Grant Program Recipients