طراحی تأمین اجتماعی - روش انتخاب نمونه کارها
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|24064||2004||22 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : European Economic Review, Volume 48, Issue 4, August 2004, Pages 883–904
Public social security systems may provide diversification of risks to individuals’ life-time income. Capturing that a pay-as-you-go system (paygo) may be considered as a “quasi-asset”, we study the optimal size of the paygo system as well as the optimal split between funded and unfunded pension saving by means of a theoretical portfolio choice framework. A low-yielding paygo system can benefit individuals if it contributes to hedge other risks to their lifetime resources. Numerical calculations indicate that optimal social security systems should be at least partly paygo financed in many economies. The optimal magnitude of the paygo system depends on the specified risk concept as well as the stochastic properties of stock market returns and implicit paygo-returns.
One of the most important justifications for public social security programs is presumably imperfect insurance markets. What we have in mind is non-marketability of human capital and also potential limitations in many individuals’ access to the stock market. This paper analyzes how such imperfections influence the optimal design of a social security system. Capturing that a pay-as-you-go (paygo) system may be interpreted as a “quasi-asset” (Merton, 1983; Persson, 2002), we derive the optimal size of the paygo program as well as the optimal split between funded and unfunded pension saving by means of a portfolio choice approach. The main bulk of the recent large literature on social security reforms takes as its point of departure that aging populations weaken the financial viability of social security systems, which mainly are financed on a paygo basis.1 It is well known that the implicit return of the paygo system is given by the growth of aggregate wage income, reflecting the combined effect of productivity- and labor supply growth. This implicit return is lower than the real interest rate in a dynamically efficient economy. Deterministic models therefore predict that a funded program is superior to a paygo program in steady-state, reflecting that a paygo program redistributes resources to the initial old generation from all future generations. Not surprisingly, projections of deteriorating dependency ratios have led many economies to attempt to derive a politically feasible and maybe even Pareto-optimal transition from a paygo program to a (partly) funded program.2 The conclusion that a funded program is superior to a paygo program in steady state is not always valid, however, when we take into account that returns on both paygo and funded systems are stochastic. In a stochastic framework, a lower expected rate of return on the paygo system does not necessarily imply that it is an inferior alternative. A low-yielding paygo system can benefit individuals if it contributes to hedge other risks to their lifetime resources. This reflects the basic idea that the paygo asset is not spanned by other assets due to an imperfect correlation between stock returns and the growth of aggregate wage income. Thus, the paygo system is a government created asset that allows one generation to trade in the human capital returns of the next. This paper considers three sources of risk to net individual income: (i) Wage income risk – reflecting technology shocks, (ii) fluctuations in the size of the population, which influence the aggregate labor supply, and (iii) a stochastic return on stock market investments. Employing a simple theoretical overlapping generations model, we characterize the optimal social security system under various assumptions about individuals’ participation in the stock market. The paper focuses exclusively on risk sharing issues in an overlapping generations framework with one representative individual within each generation. We disregard intragenerational redistribution and assume that labor supply is exogenous.3 It turns out that our analysis is sensitive to the definition of the relevant risk concept. Below we will consider what we define as respectively “traditional” and “Rawlsian” risk sharing. This paper adds to the fairly small literature on the design of social security systems under uncertainty. The idea that a paygo system can be considered as an asset, has recently been explicitly highlighted by Persson (2002) and Dutta et al. (2000). Persson provides a brief and verbal discussion of this idea and presents a numerical illustration based on Swedish data, which indicates that the paygo system may indeed hedge parts of the risk on a portfolio of stocks and/or bonds. Persson does not offer any formal analysis, however. Dutta et al., on the other hand, present a formalized analysis based on a portfolio choice approach. Their static mean–variance set-up abstracts from several important aspects of public social security systems, however. There is no explicit modeling of how the paygo system transfers resources between generations, and they do not capture how individuals’ own saving or portfolio decisions are influenced by the public decisions on the design of the social security program. We also note that Dutta et al. do not consider different risk sharing concepts (their analysis seems consistent with what we have defined as traditional risk sharing). Moreover, we argue below that mean–variance preferences in combination with a proper dynamic overlapping generations setting will imply that the derived optimal social security program is time-inconsistent. The general insight that a mixed paygo and funded system may be optimal due to diversification dates back to Merton (1983). Merton characterizes optimal tax-transfer programs within a theoretical general equilibrium model. While our approach is to consider the three stochastic variables (wages, stock market returns and population growth rates) as exogenous to the model, Merton explicitly models the interdependencies between these variables. In particular, the capital market returns are determined by the stochastic variables productivity, factor-shares and population growth. At the outset, Merton's approach seems beneficial from a theoretical point of view. We will still follow Hassler and Lindbeck (1997) and argue that this is likely to put overly strong restrictions on the relationship between these variables. Below we will present some empirical observations which indicate that the correlation between stock market returns and respectively economic growth and aggregate wage income is lower than suggested by most general equilibrium models in many countries.4 We also note that Merton does not address the distinction between various risk concepts. A recent line of research with links to Merton (1983) studies the risk sharing effects of social security in general equilibrium models, see for example Storesletten et al. (1999), Barbie et al. (2000), Bohn (2001) and Krueger and Kubler (2002). On the one hand these models do not allow any derivation of explicit expressions for optimal individual portfolio shares or optimal investments in paygo and funded social security. On the other hand, they provide a richer set-up that may include relevant additional features not captured by the model framework of this paper. In particular, a general equilibrium framework captures that an introduction of a paygo system has adverse effects on the real capital stock, which in turn contributes to lower wages and higher interest rates. According to numerical simulations performed by the calibrated models of Storesletten et al. and Krueger and Kubler, the welfare gains caused by increased risk sharing may be dominated by the crowding out effect of paygo social security on capital. In Section 6 below we discuss how the conclusions and policy messages of this paper are sensitive to the apparent trade-off between risk sharing and capital accumulation. This paper is also related to studies of how a paygo program may contribute to intergenerational sharing of income-risk; see for example Gordon and Varian (1988), Enders and Lapan 1982 and Enders and Lapan 1993, Thøgersen (1998) and Wagener (2003). With an exception of Wagener's contribution, these papers assume only one source of risk and they do not consider a split between funded and paygo programs. We will argue below that the main insight from these papers, namely that a paygo program leads to increased intergenerational income-risk sharing, hinges on specific stochastic properties of the income (or output) path over time. Wagener does capture both stochastic wage growth and stochastic interest rates. Comparing paygo systems with respectively fixed contribution rates and fixed replacement rates along similar lines as Thøgersen (1998), his focus is quite different from this paper, however. The next section presents our theoretical model framework and defines the risk concepts. 3 and 4 study the optimal design of social security systems in the cases of respectively traditional and Rawlsian risk sharing. We derive the optimal size of the paygo program as well as the optimal split between funded and unfunded pension saving. Intuitively, a lower correlation between stock market returns and the growth of aggregate wage income increases the size of the optimal paygo program. We demonstrate that the case of traditional risk sharing implies a larger paygo program than in the case of Rawlsian risk sharing under reasonable assumptions about the coefficient of relative risk aversion. Section 5 provides some numerical illustrations based on data for Sweden, Norway, the US and the UK. Our calculations suggest a role for paygo-systems in the three latter countries – but not in Sweden. Taking limited stock market participation into account, a mixed paygo/funded system is optimal for Norway, the US and the UK, while a fully funded system is optimal in Sweden. Finally, we discuss transition issues and offer some concluding remarks in Section 6.
نتیجه گیری انگلیسی
During the recent years a large part of the literature on social security systems has dealt with comparisons between funded and paygo program as well as the design of potential transitions from paygo financing to funded systems. Adopting a portfolio choice approach, this paper has provided a different perspective on the design of social security systems. Interpreting the paygo system as a “quasi-asset” along the lines of Merton (1983) and Persson (2000), the analysis has focused on the optimal size of the paygo system and the optimal split between the paygo part of pension savings and the funded part. The optimal exposure to the stock market can – from a representative individual's point of view – be handled individually if access to the stock market is perfect, or by the government if this access is imperfect. The general insight from our analysis is that a low-yielding paygo system can benefit the representative individuals if the correlation between the implicit return on the paygo program and the stock market returns is not too high. We have derived analytical formulas for the optimal size of the paygo system and the optimal magnitude of the funded pension saving in the stock market. The optimal size of the paygo program depends on the risk concept. It turns out that the optimal paygo system is smaller under “Rawlsian risk sharing” (which captures all risks to the net lifetime income of the representative individual at birth) than under “traditional risk sharing” (which capture only risks which are realized in the representative individual's second period of life). This reflects that the paygo system increases the exposure to wage income risk when wage income shocks are not transitory. We provide numerical illustrations for Sweden, Norway, the US and the UK. Adopting the traditional risk sharing concept, our calculations indicate that the paygo-asset should play a role in pension savings in the three latter countries but not in Sweden. The Rawlsian risk sharing concept reduces the scope for a paygo program in the sense that it ceases to be an interesting alternative in the US (as in Sweden) and the optimal magnitude is reduced in the remaining two countries. These conclusions are sensitive to the stochastic properties of respectively stock market and paygo returns – and the specified degree of risk aversion. With limited stock market participation, our calculations suggest that a mixed paygo/funded system is optimal (except in Sweden where the optimal system should be fully funded). Our analysis has not captured the transition to the new pension system. While an explicit modeling of such a transition combined with the portfolio set-up in the present paper is an ambitious task that calls for additional research, we will still highlight some insights. We first note that the transition issue is not problematic if the economy initially is characterized by either (i) no pension system at all, (ii) a fully funded system or (iii) a paygo system that is smaller than the paygo part of the new system. In these cases the transition generation (i.e. the old generation at the time of enactment of the program) obviously gains because it receives an unexpected paygo transfer from the next generation. Thus, the introduction of the new system will be Pareto-improving. The transition issue is crucial if the initial pension system has a larger paygo component than the new system, however. In such a case comparisons of the steady-state welfare properties of alternative pension programs cannot be the sole basis for policy recommendations because the welfare loss of the transition generation (i.e. the double taxation of the initial young generation in this case) is not captured. Recalling that the actual social security reform agenda in several OECD countries involves attempts to introduce funded parts in systems that have been entirely paygo financed at the outset, we will still claim that our analysis in an indirect sense has bearings on the transition issue. To a large extent this reform agenda is motivated by the basic insight from deterministic models that the return on funded programs is higher than the implicit return on paygo programs. In such a context deviations from a fully funded new program can be regarded as the consequence of restrictions caused by the transition issue – or equivalently lack of public financial resources to compensate the transition generations. Our analysis highlights that a maintenance of parts of the paygo system should not necessarily be regarded as an unhappy consequence of weak public finances – but as a part of a potentially optimal pension system. Loosely speaking, the insight that a paygo scheme improves risk diversification implies that the transition costs identified by deterministic analyses are mitigated. Finally, we will return to a shortcoming of our analysis that was briefly mentioned in the introduction. Our partial equilibrium approach does not capture the effects of social security on real capital accumulation and factor prices. This is fine as long as we imagine a small open economy, but obviously harder to accept in the case of a closed (or large) economy. As pointed out by Storesletten et al. (1999) and Krueger and Kubler (2002), we might suspect that the welfare gains due to the risk sharing effects of paygo social security might be offset by the adverse effects on capital accumulation. This trade-off is obviously crucial if we consider a closed economy that initially has no (or a small) paygo system. As argued above the most relevant case is an economy with a large paygo system at the outset, however. Then the magnitude of the paygo system that yields optimal risk diversification is hardly much larger than the initial system – and it might well be smaller. Consequently, the “new” system will imply small adverse effects on the real capital stock or, alternatively, even an increase in the capital stock.