زمانبندی اقدام بازنشستگی - شواهد جدید از زنان کارگر در سوئیس

We investigate the responsiveness of individual retirement decisions to changes in financial incentives. A reform increased women's normal retirement age (NRA) in two steps from age 62 to age 63 first and then to age 64. At the same time retirement at the previous NRA became possible at a benefit discount. Since the reform affected specific birth cohorts we can identify causal effects. We find strong and robust behavioral effects of changes in financial retirement incentives. A permanent reduction of retirement benefits by 3.4% induces a decline in the age-specific annual retirement probability by over 50%. The response to changes in financial retirement benefits varies with educational background: those with low education respond most strongly to an increase in the price of leisure.

Since an understanding of the magnitude of workers' responsiveness to institutional reforms is crucial for policy design, it is important to provide reliable empirical estimates. A large literature attempts to quantify the effect of retirement incentives, and the problems involved in identifying their causal effects are widely discussed (see, e.g., Lumsdaine and Mitchell, 1999, Coile and Gruber, 2007 and Chan and Stevens, 2004). Much of the literature identifies behavioral responses to financial incentives based on the cross-sectional comparison of individuals with different benefit claims and focuses on the appropriate representation of dynamic incentives (for cross-national comparative studies see, e.g., Gruber and Wise, 2004 and Duval, 2004). This approach mostly neglects the possibility of unobserved heterogeneity in tastes for retirement which might affect both incentives and responses. In their study of retirement expectations Chan and Stevens (2004) find that such heterogeneity strongly affects the estimates of responses to retirement incentives. Some studies rely on natural experiments to obtain estimates of the effect of financial incentives that are not biased by unobserved heterogeneity: Krueger and Pischke (1992) show that workers affected by reduced retirement benefits in the U.S. in 1977 did not respond as strongly as would have been expected based on prior findings. Mastrobuoni (2009) investigates whether the 1983 reform of the U.S. normal retirement age affected retirement behavior. He finds that every 2-months-increase in the NRA (normal retirement age) at actuarially fair benefit reductions for early retirement increases the mean age of benefit claiming by one month. Similar to these studies we take advantage of a reform in the retirement system to identify the effect of financial incentives on retirement behavior. The 1991 reform of the Swiss mandatory retirement insurance introduced two separate institutional modifications. On the one hand the normal retirement age for females was raised in two steps from 62 to 64. On the other hand the possibility of early retirement was introduced at the expense of a benefit discount. As these measures reflect policy options available in about every social security system, it is both interesting and important to study their effects. Also, since the retirement reform is tied to the year of birth as a fixed individual characteristic, the experiment is not subject to endogenous sorting into treatment. This study contributes to the literature in various ways: first, it identifies the labor supply response to retirement incentives by comparing the behavior of birth cohorts which differ only with respect to the financial incentives of a policy regime. In contrast to studies which rely on the cross-sectional identification of incentive effects, we can take advantage of an exogenous institutional reform. We know its precise timing and can therefore avoid measurement error. In addition, we avoid the problem that individuals may not be informed about their retirement incentives (Asch et al., 2005): the reform we look at here was subject to intense public debate due to a national public referendum (Bütler, 2002). Second, we evaluate the heterogeneity of the behavioral response to the policy reform, i.e. labor force exit, across various levels of individual human capital. Song and Manchester (2007) find that there are large differences in the response to changes in the social security earnings test along the income distribution. Similarly, Mastrobuoni (2009) finds stronger responses to retirement incentives among men with less formal education. Third, we investigate whether the behavioral response to the institutional change happens instantaneously or whether the adjustment process takes time. If retirement age is strongly affected by social norms the response to policy reforms might be dampened and protracted. Social norms play a key role in the debate on excess retirement at age 62 and 65 in the U.S. (cf. Lumsdaine et al., 1996, Coile and Gruber, 2007 and Duflo and Saez, 2003), where they are discussed as a potential explanation. Finally, while many studies in this literature focus on males we take advantage of a retirement reform specifically for female workers. For historical reasons, benefit eligibility rules for retiring females are often more lenient than those for retiring males. Therefore, adjustments specifically for the female labor force are a relevant policy issue in many countries. In addition, both health and financial restrictions may cause different responses to given policy changes for men and women. We find clear behavioral adjustments in response to changes in retirement incentives. Labor supply elasticities differ across population groups with heterogeneous educational backgrounds. The estimation results are robust to controls for endogenous panel attrition. The evidence suggests that the adjustment of retirement behavior to changed institutional circumstances intensifies over time

This study identifies the effect of financial incentives on retirement behavior taking advantage of the natural experiment of an exogenous institutional reform in Switzerland. This source of identification helps to avoid the substantial biases of up to 50% that e.g. Chan and Stevens (2004) found when they added controls for individual unobserved heterogeneity: the marginal effect of pension incentives on subjective retirement expectations robustly dropped by at least half after the authors added individual fixed effects to their model. The reform of the Swiss retirement insurance increased the normal retirement age for females born after 1940 in two steps from 62 to 64 years. After the reform, female retirees at age 62 incurred a benefit reduction of initially 3.4 and later 6.8%. The modification of the normal retirement age is a potent policy instrument as it affects both, the length of the contribution period as well as the duration of benefit payments. We apply a difference-in-differences type procedure and confirm the robustness of our results with respect to alternative model specifications and estimators. We observe a strong response to the shift of the normal retirement age in connection with benefit reductions: a reduction in benefits by 3.4% caused a decline in the retirement probability at age 62 from 46 to 22%, i.e. half of those who would have left the labor force at age 62 prior to the introduction of benefit discounts now remain in the labor force. The probability of retirement at age 63 drops from 40 to 30%, i.e. by 25%, after benefit discounts of 3.4% were implemented. Behavioral adjustments intensify significantly over time. This finding is in line with social norms affecting behavioral choices of Swiss workers, or with a time-lag in recalculating expected benefits. Females with low education show the strongest response to changes in retirement incentives and appear to be most reluctant to incur a decline in benefit payments. Overall, retirement behavior responds strongly to changes in incentives. However, the effect for Switzerland may indicate a lower bound on the effect that is possible in other countries, because the Swiss reform affects only the first pillar of the retirement insurance system leaving the other pillars unchanged. We expect stronger effects if a reform comprehensively addresses all funding sources for retirement. On the other hand our analysis is limited to the extensive margin of labor force participation. De Grip et al. (2012) show that reducing benefits and delaying retirement may adversely affect mental health of older workers. Such effects, which are beyond our analysis, may limit the effectiveness of retirement benefit reforms. Our findings confirm prior studies (e.g. Asch et al., 2005) and suggest that financial retirement incentives can substantially affect the retirement plans of the generations to come. This may contribute to solve the funding problems of retirement insurance funds.