پس انداز بافر در حساب های بازنشستگی

We use a dynamic programming model to explore the possibility and extent of precautionary saving in tax-sheltered accounts such as the 401(k). The main policy experiment examines the behavior of saving for different levels of unemployment insurance (UI), which is a perfect substitute for precautionary saving against job loss. Our results indicate that increasing the generosity of UI crowds out 401(k) contributions made by younger workers, who save primarily for precautionary reasons. At the aggregate level, we find that 401(k)s increase national saving and that the magnitude of the effect depends on the generosity of UI.

In 2004 President Bush outlined a proposal to reform Social Security that would allow workers to invest a fraction of payroll taxes in private accounts. If enacted, the proposal will further the trend toward tax-sheltered accounts that began in the late 1970s with the inauguration of the IRA and 401(k) programs. Critics of private accounts point to the poor performance of 401(k)s during the recent decline of the U.S. stock market in 2001 to argue that private accounts are too risky. But if the stock market decline highlighted the riskiness of 401(k)s, the recession that followed illustrated a potential advantage: flexibility. Unlike traditional pensions or Social Security, the 401(k) rules allow participants to make pre-retirement withdrawals when they leave their jobs. The withdrawals are penalized at 10%, but they can still provide a buffer against income loss during unemployment. While the riskiness of 401(k)s has received a lot of attention from the media, the potential for 401(k)s to provide self-insurance against income fluctuations has been largely ignored. We examine the conditions under which a 401(k) plan can serve as a precautionary saving vehicle and explore the implications this has for national saving. The economy in the paper extends the life-cycle frameworks of Engen et al. (1994) and Laibson et al. (1998) by modeling job loss and unemployment insurance (UI). The inclusion of UI allows us to measure the extent of precautionary saving in the 401(k) by observing how contributions and saving respond to changes in the UI replacement rate. Since UI is a perfect substitute for buffer saving against unemployment shocks, its effect on saving both inside and outside the 401(k) provides information about the motive for saving. In particular, if an account is used as a precautionary vehicle against employment risk, higher rates of UI should reduce saving in that account during periods when the precautionary motive is dominant. Results from the model simulations show that increasing the generosity of UI crowds out 401(k) contributions made by younger workers, who save primarily for precautionary reasons. Using the benchmark parameters, we find that raising the UI rate from 10% to 70% reduces the ratio of contributions to income by an average of 86% for the first 10 years of working-life. In contrast, saving in the unsheltered account falls only slightly in response to higher rates of UI. The possibility of early withdrawals from the 401(k) has fueled concern about leakage from retirement saving. Munnell and Sundén (2004) report that individuals who took a lump-sum distribution from a 401(k) plan cashed out a median amount of $6,000 (in 2001 dollars). Poterba et al. (2001) estimate the loss in retirement savings associated with these cashouts and find it to be about 5%. These studies measure the losses due to early withdrawals, but they do not account for the increase in contributions attributable to the possibility of self-insurance in the 401(k). Instead of interpreting early withdrawals as a consequence of bad decision making, we find that a rational saver will not only withdraw from the 401(k) during unemployment, she will actually plan to do so. The introduction of 401(k) plans affects national saving in the model economy through two channels. First, the flexibility of the plans allows agents to use the 401(k) to self-insure against unemployment shocks, and consequently, there is a substitution of saving from the unsheltered account to the 401(k). Second, the substitution is complemented by higher saving because the 401(k) is a more attractive vehicle for life-cycle saving. Simulations suggest that 401(k)s have a positive effect on overall saving. Relative to a model economy without 401(k)s plans, we find that introducing 401(k)s increases the steady-state national saving rate by between 1 and 3 percentage points, depending on the assumptions about UI. Consistent with previous simulations of the UI system (e.g., Engen and Gruber, 2001), we find that increasing the generosity of UI decreases national saving for both economies, but that the effect is largest in the economy without 401(k)s. Whether 401(k)s actually increase overall saving remains a subject of some controversy in the empirical literature.1 The remainder of the paper proceeds as follows. Section 2 discusses institutional features of the 401(k) that are relevant for the model. Section 3 presents the model. Section 4 discusses the data, the choice of parameters, and the solution technique. Section 5 calculates the extent of self-insurance in the 401(k) and the net effect on national saving. Section 6 concludes.

Our simulation exercises suggest that saving is done in the 401(k) for two reasons: to accumulate resources for retirement and to build a buffer stock against potential income loss due to unemployment. While the first reason is obvious, the second is novel and has implications for the effect of employment risk on saving. Two results in the paper support the idea that some precautionary saving is done inside the 401(k) plan. First, savings in the plan are regularly tapped in the event of unemployment, and the size of the withdrawals is large. In an economy with a 50% match and a 50% replacement rate of UI, the average withdrawal from the 401(k) during unemployment is approximately $8,000. Second, increasing the generosity of UI, which is a perfect substitute for precautionary saving against employment risk, is associated with a large reduction in 401(k) saving for the young. Saving in the unsheltered account falls as well, but by much less than in the 401(k). In contrast to the popular perception that early withdrawals from the 401(k) represent poor planning on the part of the individual, our interpretation is exactly the opposite. A fully rational saver will not only withdraw from her 401(k) plan during unemployment, but more significantly, she will actually plan to do so. This does not mean that for a given level of savings in both accounts the individual would prefer to withdraw from the 401(k). Rather, the opportunity cost of saving in the outside plan, in terms of forgone matching and tax benefits, is too high. The possibility of self-insurance in the 401(k) also appears to have important effects at the aggregate level. When we compare a model economy with a 401(k) plan to one without, we find that introducing the 401(k) increases the national saving rate by between 1 and 3 percentage points in the steady state. The magnitude of the impact on saving depends on the generosity of UI and the extent of employer matching. Saving in both economies falls with higher replacement rates of UI, but the reduction is greatest in the economy without a 401(k) plan. The analysis in the paper could be extended and improved in several ways. First, the analysis should ideally be carried out in general equilibrium. The assumption of partial equilibrium is inaccurate to the extent that responses of saving to changes in UI and 401(k) incentives feed back into the capital stock, returns, and wages.18 The partial equilibrium assumption is not critical to the paper's main conclusion about the possibility of self-insurance in the 401(k) plan, but it does make a difference for the results pertaining to aggregate saving. Second, the simulation results pertaining to self-insurance in the 401(k) need to be tested empirically. One approach to this would be to measure the degree of precautionary saving in the 401(k) using state variation in unemployment insurance using data from the Survey of Income Program Participation. Empirical analysis is especially important in light of recent research suggesting that 401(k) participants behave in ways quite different from the optimizing agent in the model.19 If the empirical results corroborate the simulation findings, the model could have important policy implications for the current debate about privatizing Social Security in the U.S. Although this debate has been framed broadly in terms of risk and return, one could argue that this is really part of a more general tension between paternalism and flexibility. The current system is paternalistic, in that it provides a reasonably secure flow of retirement income, but it is also completely illiquid for people in their working years. Our results suggest that there might be gains associated with making the retirement saving system more flexible. Of course, if saving behavior is not rational, greater flexibility could create leakages from retirement savings that are inefficient in their own right. Understanding the trade-offs inherent in the design of saving programs will be a goal of future research.