اصلاحات تامین اجتماعی و هموارسازی موقتی

This paper examines the welfare and distributional impact of switching from a defined benefit pension system like Social Security to a defined contribution system of personal retirement accounts (PRAs). While such a switch would do away with the transfers that low-wage workers receive through Social Security, it would have the benefit of reducing labor supply distortions. Moreover, a particular kind of PRA suggested in this paper – one that has very low (or zero) contribution rates early in life – substantially enhances life-cycle patterns of consumption and labor supply by making more resources available to credit-constrained young workers. Simulations of a life-cycle model show that in welfare terms these PRAs significantly outperform conventional PRAs because of this enhanced intertemporal smoothing.

Much of the debate over Social Security reform in the U.S. has centered on the question of whether to undertake fundamental reform that would shift from the current defined benefit policy toward a defined contribution policy of personal retirement accounts (PRAs). Such a shift would affect individuals’ welfare, and its distribution, in a variety of ways. On the one hand, a move toward PRAs would eliminate the extensive transfers inherent in the current system. These transfers serve several purposes. First, they provide insurance, within a cohort, by redistributing toward workers who experience adverse labor market shocks throughout their lives. Second, they provide insurance across cohorts, by transferring toward cohorts that experience adverse aggregate shocks. Third, the transfers redistribute toward workers with low expected lifetime earnings, which can be interpreted as insurance against being born with fewer skills. Based on these considerations alone, a shift to PRAs would adversely affect the level and distribution of individuals’ well-being. On the other hand, PRAs have advantages that could enhance welfare relative to the current system. First, the taxes and transfers inherent in Social Security distort individuals’ labor supply decisions. The benefits that an individual receives are imperfectly linked to the contributions they make, and consequently the contributions largely constitute a tax on labor. Under PRAs, that tax and its distorting effect on labor supply decisions are eliminated. A second potentially significant advantage of PRAs, which has received little or no attention, relates to the current system's distortion of life-cycle patterns of consumption and labor supply. In general, young people who anticipate higher future income would like to borrow against that future income in order to smooth their consumption over the life cycle. However, limits to borrowing impede this smoothing and result in a distorted consumption pattern that grows with income over the life cycle. As Hubbard and Judd (1987) pointed out, Social Security exacerbates the distorted life-cycle consumption profile by forcing young people to save, effectively, via their Social Security contributions. Although Hubbard and Judd do not discuss it, the life-cycle profile of labor supply is similarly distorted. Young people, who would like to work less early in life, when wages are lower and human capital acquisition is more important, cannot afford to do so because their forced contributions to Social Security make it more difficult to consume leisure. One way to reduce these distortions of life-cycle patterns, perhaps significantly, would be to allow contributions to the PRAs to be low (or zero) early in life and then to increase with age. As a result, young, borrowing-constrained workers would not be forced to save for their retirement, which would allow them to optimally increase their consumption (of both goods and leisure). Most (to my knowledge, all) of the proposed PRA policies do not consider this possibility, but the results in this paper will clearly demonstrate that they should. It is worth emphasizing, however, why a contribution schedule of this sort (low initially, then increasing with age) would be considerably less attractive in the context of a defined benefit policy like Social Security. In particular, because the imperfect benefit–tax linkage of Social Security means that contributions are largely perceived as a tax on labor, contribution rates that rose with age would discourage labor supply later in life – precisely when their effort is most valued by the market. Aside from improving life-cycle smoothing patterns, age-dependent contribution rates would potentially have another advantage. Namely, because young workers would no longer be forced to save for their retirement, it would be less costly, in utility terms, for them to accumulate a small buffer stock of liquid assets (or, similarly, to preserve some borrowing capacity) to help insure against idiosyncratic shocks to their earnings potential. This improved self-insurance under PRAs could partially, or even completely, offset the loss of insurance against idiosyncratic risk that is provided by Social Security. This paper employs a life-cycle model of consumer behavior to quantitatively assess these issues. In order to analyze the impact of different policies on the level and life-cycle pattern of labor supply, there is a labor/leisure decision in addition to the standard consumption/saving decision. There is a limit on the amount that individuals may borrow. Without such a limit, of course, life-cycle smoothing issues would become moot, as young workers would optimally smooth consumption and labor supply by borrowing heavily. Workers are born into 1 of 10 wage groups, which differ in the level of their age–wage profile. This allows us to analyze the distributional impact of different policies. Within each wage group, average wages for the group increase with age, reflecting productivity-driven real wage growth and age/experience effects. Individuals within a group experience shocks that cause their wages to deviate from the group average. This idiosyncratic risk is necessary to compare the level of insurance provided by different policies. Four different policies are analyzed using this life-cycle model. First, we consider a policy that resembles the current Social Security program (SS). Given the distortions that result from the distributional aspect of SS, special attention is given to capturing the most important details of both the tax and benefit aspects of SS. Thus, in addition to the state variable that tracks liquid assets, there is a state-variable that tracks indexed average earnings in a way very similar to the average indexed monthly earnings (AIME) measure recorded by the Social Security Administration. We also model the cap on earnings that are subject to Social Security taxes and that figure into the average earnings calculation. As with Social Security, benefits are a concave, piecewise-linear function of this measure of earnings. Second, we consider a conventional PRA policy in which individuals, in addition to voluntarily accumulating liquid assets, must contribute a constant fraction of their labor earnings to their own illiquid retirement account. We call this policy PRA-FLAT because the contribution schedule, as a function of age, is flat. For the third policy, PRA-VAR, contribution rates vary with age. More specifically, contribution rates between ages 21 and 30 are zero and then increase linearly until retirement. By considering both PRA policies, we can examine the benefits of low contribution rates for young individuals – both in terms of the life-cycle patterns of consumption and labor supply, and in terms of self-insurance against idiosyncratic risk. The last policy that we consider is a laissez-faire (LF) policy, with no mandatory retirement savings. This policy is useful for comparisons with other policies, as it indicates the savings patterns that individuals would choose of their own volition. Of course, LF should outperform both PRA-FLAT and PRA-VAR since the involuntary nature of PRAs constrains individuals’ decisions – retirement savings are mandatory and illiquid – but has no benefits (at least in the context of the model). One question of interest is to what extent does the schedule of contributions under PRA-VAR actually constitute a binding constraint, in that the life-cycle path of accumulated retirement savings is actually different from the path of savings accumulated (voluntarily) under LF. If the involuntary contributions of PRA-VAR do not constitute a significant constraint, then there is little cost to making retirement saving mandatory.1 In fact, simulations of the four policy regimes show that LF and PRA-VAR do perform very similarly; however, they both differ significantly when compared with SS and PRA-FLAT. In particular, because young individuals are not forced to save for their retirement under PRA-VAR and LF, the life-cycle patterns of consumption and labor supply associated with those policies differ considerably from the patterns under SS and PRA-FLAT. For example, consumption by young people is nearly 10% higher than that under SS. Labor supply increases over the life cycle by an additional 15% under PRA-VAR and LF, compared with SS. This steeper path for labor supply is optimal given the growth in wages over the life cycle. Labor supply differs, across policies, not only in patterns but also in levels. Hours worked under each of the three alternative policies is greater than under SS, reflecting the distortionary effect of SS's taxes and benefits. The differences in hours worked are greatest for LF and PRA-VAR, with increases relative to SS of almost 6% for some wage groups. Differences in labor earnings are even greater, since much of the additional labor supply occurs later in life, when wages are higher. The impact of the four policies on welfare, as measured by compensating variations, differs considerably across wage groups. The lowest seven deciles of the wage distribution prefer SS to PRA-FLAT, while only the lowest five deciles (the fourth and fifth by very little) prefer SS to PRA-VAR or to LF. Some of the preference for SS stems from the fact that it offers better insurance against idiosyncratic risk, though the magnitude of this effect is small and is relatively uniform across wage groups. By far, the redistributive nature of SS is the factor most responsible for the fact that the lower end of the wage distribution prefers SS. When the impact of the transfers in neutralized, by equalizing the averages of simulated consumption and labor supply under the three alternative policies with the SS averages, all wage groups strongly prefer PRA-VAR and LF to SS and PRA-FLAT. This result stems from the superior life-cycle patterns of consumption and labor supply. It is worth noting that when we consider a lower intertemporal elasticity of substitution – which for the baseline parameterization is toward the upper end of the typical range of values considered – these results are even more strongly reinforced. Interestingly, the welfare benefits of allowing lower contribution rates early in life are robust to relaxation of the borrowing limit. The compensating variations are affected very little even when the borrowing limit from the baseline parameterization is doubled. Though easing borrowing limits does facilitate life-cycle smoothing, it does so for all policies. That is, it does so without reducing the relative advantage that PRA-VAR and LF have in that regard. For clarification, it is worthwhile to point out at least two issues that the model of this paper, in the interest of tractability, is not designed to address. First, there is only one (riskless) financial asset in the model, so the fact that personal retirement accounts have the potential benefit of facilitating the investment of retirement savings in assets with higher returns, such as equities, is not addressed. Second, the model is not general equilibrium in nature: wages and the rate of return are taken as given. A general equilibrium formulation would require that we take seriously the fact that SS is an unfunded pension system, whereas PRAs are inherently funded. As it is, to put the various regimes on an equal footing for the sake of comparison, we assume (counterfactually) that SS is fully funded. In this way, the effective return on retirement contributions is equal across policy regimes,2 and attention can be focused on the issues of primary interest here. Moreover, for general equilibrium results to have meaning, we would have to address the issue of how to finance the transition from unfunded to funded, i.e. how to pay for the currently existing liabilities of SS. While these issues are no doubt important, to treat them seriously would complicate the analysis by confounding these general equilibrium considerations with the issues of greatest interest here. Accordingly, the paper focuses more narrowly on the implications that the various policy regimes have for individuals’ consumption and labor supply decisions, and for their welfare, in terms of the transfers, insurance against idiosyncratic risk, and ability to achieve optimal life-cycle patterns of consumption and labor supply. The paper is organized as follows. The next section briefly discusses the related literature. Section 3 describes the features of the model that are common to all policy regimes and then explains the decision problems of individuals under the four policy regimes. Calibration of the model's parameters is discussed in Section 4. Section 5 examines how the consumption and labor supply decisions are affected by the different policies. Then, in Section 6, the welfare and distributional impact of the policies is explored, with a focus on the relative importance of transfers, insurance against idiosyncratic risk, and life-cycle smoothing for the welfare differences. Section 7 examines how robust the results are to different assumptions about borrowing constraints and about risk aversion. Finally, Section 8 offers concluding comments.

This paper has utilized simulations of a life-cycle model with endogenous labor supply, uncertainty in earnings potential and in lifespan, and heterogeneity in expected wage profiles to analyze the impact that different policy regimes for retirement saving would have on labor supply and on the welfare of different wage groups. The redistributive nature of Social Security means that there is an imperfect link between taxes paid and benefits received. This has the effect of discouraging labor supply. The results show that a switch from Social Security to PRAs could increase labor supply by 4%. The distributional analysis has examined the extent to which the welfare impact of the various policy regimes can be attributed to transfers, to insurance against idiosyncratic risk, and to life-cycle patterns of consumption and labor supply. The bottom seven deciles of the wage distribution fare better under Social Security than under personal retirement accounts with constant contribution rates. This is almost entirely attributable to transfers received by these individuals. Insurance against idiosyncratic risk is only marginally better under Social Security than it is under personal retirement accounts. PRAs with contribution rates that are low early in life fare considerably better than PRAs with constant contribution rates because they allow more efficient life-cycle patterns of consumption and labor supply. The number of deciles that prefer Social Security to these PRAs is five, instead of seven, and, furthermore, they prefer it by considerably smaller compensating variations. The results of this paper, which show the importance of life-cycle smoothing considerations, raise more general questions about social insurance programs. The fact that the welfare gains that low-wage workers achieve under Social Security are overwhelmingly attributable to the transfers that they receive, and not to any significant improvement in insurance, suggests that perhaps the transfers would be best achieved outside the scope of Social Security. More specifically, transfers for low-wage individuals would have their greatest impact early in life, when individuals are most in need of resources. Because transfers made earlier in life would enhance, rather than diminish, life-cycle smoothing, the size of the transfers needed to achieve the welfare gains achieved by the transfers under Social Security would be smaller. Furthermore, to the extent that agents can actually affect their lifetime earnings potential through human capital investments, shifting the timing of transfers to earlier in life would also potentially facilitate greater investment in human capital by young, borrowing-constrained workers. A clear difficulty with this kind of policy (transfers early in life) would be the matter of how to identify at an early stage whether a worker has low lifetime earnings potential. This difficulty suggests that greater subsidies of higher education, for example, may be preferable to direct transfers.