میزان مصرف، بازنشستگی و تامین اجتماعی: بررسی بهره وری از اصلاحاتی که فرصتهای شغلی دیگر را تشویق می کند

This paper proposes and analyzes a Social Security reform in which individuals no longer face the OASI payroll tax after, say, age 54 or a career of 34 years, and their subsequent earnings have no bearing on their benefits. We first estimate parameters of a life-cycle model. Our specification includes non-separable preferences and possible disability. It predicts a consumption–expenditure change at retirement. We use the magnitude of the expenditure change, together with households' retirement-age decisions, to identify key structural parameters. The estimated magnitude of the change in consumption–expenditure depends importantly on the treatment of consumption by adult children of the household. Simulations indicate that the reform could increase retirement ages one year or more, equivalent variations could average more than $4000 per household, and income tax revenues per household could increase by more than $14,000.

As the U.S. population ages and the moment approaches when Social Security benefit outlays will exceed payroll tax receipts, discussions of Social Security reform naturally focus on the system's solvency. Renewed concern about the Federal deficit has drawn still more attention to Social Security's assets and liabilities. This paper argues that issues of efficiency deserve greater attention as well. The current Social Security rules may generate or exacerbate labor–supply distortions; these distortions may contribute substantially to the system's social cost; and demographic trends may augment their importance. This paper proposes and evaluates a simple Social Security reform aimed at alleviating distortions to private retirement decisions that the current system may create. The proposed reform would establish a long vesting period (say, 34–40 years of contributions). After vesting, a worker would no longer face the old-age and survivors insurance (OASI) payroll tax, his benefits schedule as a function of retirement age would be fixed, and he would not face any “earnings test.” In fact, we would maintain the existing benefit formula, but base it only on earnings prior to the vesting age. Individuals who continue to work after vesting would thus receive a 10.6% payroll tax reduction. To maintain revenue neutrality within the system, there would be a small increase in the payroll tax during the vesting period.1 Following the tradition of Auerbach and Kotlikoff (1987) and others, we evaluate this reform in the context of a certainty equivalent life-cycle model. In contrast with that tradition, we estimate the parameters of the model using microeconomic data on earnings, consumption, and retirement. We employ what we think is a novel estimation strategy to recover key structural parameters. The strategy uses both panel data from the Health and Retirement Study (HRS) and pseudo panel consumption expenditure data from the Consumer Expenditure Survey (CEX). Simulations of the estimated model indicate that the proposed reform could raise retirement ages by more than a year, on average; equivalent variations from the reform could average $4000 per household (2005 dollars, present value age 50) or more; and, society's additional income tax revenues could average more than $14,000 per household. The logic of the proposed reform echoes a literature on age-dependent taxation that points to efficiency gains from using age to target lower tax rates at households with higher elasticities of labor supply.2 Intuitively, the reform aims to eliminate the substitution effects of Social Security taxes late in life, when labor supply is especially elastic, while leaving other potential distortions of the system unchanged. To see better how efficiency gains can arise, it helps to know that a standard assumption of our model implies that the income and substitution effects of Social Security taxes offset one another on average. Social Security benefits also generate an income effect, which leads to earlier retirement, and a substitution effect, which leads to later retirement. In the case of benefits, the income effect tends to dominate; the substitution effect is slight because the present value of benefits is quite insensitive to marginal earnings for households with long work histories. On balance, therefore, the existing Social Security system tends to promote earlier retirement. Our proposed reform eliminates the payroll tax late in careers — but before most households' optimal retirement age — canceling, for many, the tax's adverse substitution effect on work incentives. Although the positive substitution effect from the present system's benefit formula will be eliminated at the same time, its magnitude is smaller. Income effects from both taxes and benefits remain unchanged. On net, we hope to reduce work disincentives from the current Social Security system, taking advantage of the relatively high elasticity of labor supply at the age of retirement to attain significant efficiency improvements. To quantify the effects of our reform, this paper develops a life-cycle model in which households choose their retirement age as well as their lifetime consumption/saving profile, jobs require full-time work, and retirement is permanent. The benefit to a household of later retirement is greater lifetime earnings; the cost is forgone leisure — and, more generally, lost time at home. A household derives a flow of services from its consumption expenditure and time at home. The service flow, in turn, yields utility through a conventional, concave utility function. Although our baseline model ignores health considerations, we present a second formulation with an insurable chance of disability. The model is simple. It abstracts, among other factors, from uninsured income risk, uncertain longevity, and liquidity constraints. A benefit of simplicity is that the model offers analytic insights. One of these insights is the prediction of a discontinuous change in expenditure at a household's retirement, a change attributable to the abrupt increase in leisure and the intratemporal complementarity of expenditure and leisure.3 A number of empirical studies have described a drop in household consumption expenditure at the time of retirement (Banks et al. (1998), Bernheim et al. (2001), Hurd and Rohwedder, 2003 and Hurd and Rohwedder, 2005, Haider and Stephens (forthcoming), Aguiar and Hurst (2005), Blau (2006), and others). Our analysis shows how to use the magnitude of the drop, which this paper measures from CEX data, as well as age of retirement, measured from the HRS, to identify the model's key parameters in a simple way.4 We simulate our estimated model and find potentially substantial behavioral and welfare consequences from reform. We find, for example, that stopping the Social Security OASI payroll tax after a vesting period of 34 years of contributions could lead households to postpone their retirement by a year and a half or more, on average. We calculate that consumers, on average, would pay as much as $4000 (2005 dollars, in present value at age 50) to participate in the post-reform system. When we account for the social gain from income taxes on longer careers, the total social benefit could increase to more than $18–20,000 per household. Certain assumptions of our model — such as jobs requiring full-time work, the permanence of retirement, the absence or insurability of many forms of risk, and a lack of liquidity constraints — likely amplify the behavioral consequences and efficiency gains from reform.5 However, we believe that the estimated magnitudes of the gains in our model indicate that this paper's reform is worth further consideration. This paper joins a large literature aimed at evaluating the effects of Social Security on labor supply. See Feldstein and Liebman (2002) for a review. By applying an explicit life-cycle model, we differ from much of this literature, which seeks reduced form estimates. Implementing a structural model allows us to evaluate the life-cycle effects on retirement and consumption of counterfactual reforms. By estimating the parameters of a fully-specified model, our paper also joins a smaller literature that provides structural estimates of life-cycle models of retirement (see, for example, Gustman and Steinmeier (1986), Rust and Phelan (1997), Bound et al., 2005 and French, 2005, and van der Klaauw and Wolpin (2005)). Our work is distinguished from this literature by its emphasis on a particular reform and by its use of both earnings and consumption data. Our estimation differs from many recent structural models of retirement in its certainty equivalent approach. Policy simulations, however, often employ such a framework, and we believe that it provides a rich yet tractable formulation — permitting analytic as well as numerical insights. The organization of this paper is as follows. Section 2 describes our basic model and its formulation with stochastic disability. Section 3 discusses our pseudo-panel data on consumption expenditure, our HRS data on lifetime earnings and retirement ages, and our parameter estimates. Section 4 discusses how the model's parameters are identified, and details our estimation strategy. Section 5 qualitatively and quantitatively analyzes the Social Security reform outlined above. Section 6 concludes.

As the U.S. population ages and public programs strain to support larger populations of retirees, policymakers look for ways to encourage longer careers. This paper proposes and evaluates a simple reform of the Social Security system that a standard theory predicts will reduce the distortions of the system and incentivize many to extend their worklives. When we estimate a life-cycle model and simulate that reform, we predict large increases in average career lengths and economically substantial increases in welfare result from this straightforward change in policy. Certain assumptions of our model — such as jobs requiring full-time work, the permanence of retirement, and the absence or insurability of many forms of risk — likely amplify the behavioral consequences and efficiency gains from reform we have studied. Overall, however, we think the simulated responses of Table 5 are so large as to justify more attention to Social Security reform of this type. More generally, if distortions from tax policy are this of this magnitude, reform of the type analyzed here might sensibly be extended to other forms of taxation beyond the payroll taxes that fund Social Security. Indeed, our increases in labor supply would be even larger if the entire payroll tax (rather than just its OASI component) were discontinued at our vesting age. This paper's life-cycle framework is simple. This stripped-down nature has the advantage of allowing us to estimate parameters rather than relying on calibration. And yet the model is sufficiently flexible that we can incorporate into our simulations the behavioral responses from individual households in the HRS sample, and thus preserve a realistic degree of heterogeneity of earning profile shapes and family composition. Nevertheless, it would be valuable to add further sophistication to the model. Looking ahead, it would be especially useful to enrich the model to better account for female labor supply. The growing labor force participation of married women is one of the most dramatic changes in U.S. society over the past 40 years, and central to policies aimed at encouraging longer careers. While our formulation incorporates both male and female labor supply, one of the elaborations that we are most anxious to pursue is to model female labor force participation decisions in much greater detail.