انگیزه های مالی و بازنشستگی: شواهدی از کارگران خدمات مدنی فدرال

We examine the retirement behavior of federal civil service workers employed by the Department of Defense. These workers provide an interesting population for studying retirement because they face relatively simple financial incentives, high quality administrative data are available, and they are not covered by the Social Security system. We find that these workers respond to their financial incentives in a similar manner to what others have found when analyzing much different retirement systems. We also find no evidence of “excess retirement” at key ages of the Social Security system, which does not support the existence of societal-wide norms regarding retirement.

Numerous studies have examined the impact of financial incentives on retirement behavior.1 Understanding such behavior is fundamental to assessing the solvency of public and private retirement programs and evaluating the potential impacts of any proposed changes. In this paper, we examine the impact of financial incentives on retirement behavior for a unique set of workers, the federal civil service workers in the Department of Defense. There are several benefits to examining these workers. First, they face a simple retirement scheme because we study workers who are covered by the Civil Service Retirement System (CSRS). The CSRS is a typical defined benefit plan in which pension benefits are a function of years of service and salary, and workers covered by the CSRS are not covered by the Social Security system. Increasing evidence suggests that individuals often know very little about their retirement incentives (e.g., Gustman and Steinmeier, 2001). The simplicity of the CSRS suggests that individuals are much more likely to know their incentives, and thus, these data provide a “best case” in assessing individual retirement responses. The second benefit to studying CSRS workers is that they provide important evidence regarding whether there is “excess retirement” at ages 62 or 65. Several previous studies have concluded that there exists excess retirement because, after the incentives in Social Security and Medicare are accounted for, more people retire at ages 62 and 65 than models suggest (e.g., Lumsdaine et al., 1996).2 This excess retirement is attributed to the existence of social norms influencing retirement behavior. Our data provide an interesting test of whether there is excess retirement at ages 62 and 65 because the workers in our sample do not face special incentives to retire at these ages: none of the pension rules relevant to our sample involve ages 62 and 65, and upon retirement, all employees have access to the same health insurance they had while working.3 The third benefit is that we have high-quality administrative data for hundreds of thousands of CSRS workers. These data include a complete description of pension rules and the corresponding information necessary to compute pension values for each worker in the sample. Such information minimizes the potential for measurement error in computing financial incentives. Our results suggest that civil service workers delay their retirement probability by 4% for each additional US$10,000 of the expected pension wealth they would gain for working another year. This estimate is very similar to those found in previous studies using much different data and analyzing much different retirement schemes. We also find no evidence that the civil service workers in our sample retire disproportionately at ages 62 or 65, which undermines the conclusion that there exists societal-wide norms influencing the retirement decision.

It is of immense interest to understand the impact of financial incentives on retirement. Numerous papers have focused on many aspects of these incentives. We contribute to this literature by examining the retirement behavior of federal civil service workers in the Department of Defense. These workers provide an interesting case study because they do not participate in the Social Security system, they are only covered by a standard defined benefit pension plan, and high quality administrative data are available. Relying on an option value framework, our main result suggests that federal civil service workers respond to their retirement incentives in a manner that is quite similar to the response other studies have found. This similarity is despite the fact that we study a retirement system that is much different than those that are typically studied. Such a finding provides important confirmation that these reduced form estimates could be useful for policy simulations. Second, we find less evidence of “excess retirement” at the key retirement ages of the civil service workers (55 and 60 in our sample). This finding, that the reduced form estimates predict retirement better in our data than in previous studies, could be due to less measurement error in calculating the pension incentives or due to the workers' better understanding of the pension rules. Finally, we find no evidence of excess retirement at age 62 or 65. Several studies have concluded that the excess retirement is due to social norms because they rule out a role for Social Security and pension incentives by relying on their model, and Medicare eligibility by focusing on individuals with retiree health insurance. Our method for ruling out Social Security and pension incentives is much more transparent: civil service workers do not have any incentives to retire at age 62 and 65 and they possess relatively generous retiree health insurance benefits. The lack of a spike at ages 62 and 65 suggests that there is at least no society-wide “social norm” associated with the key Social Security retirement ages.