امنیت اجتماعی و روند جمعیتی: نظریه و شواهد از تجربیات بین المللی

The worldwide problem with pay-as-you-go, defined-benefits social security systems is not just financial. Through a dynamic, overlapping-generations model where forming a family and bearing and educating children are choice variables, we show that social security taxes and benefits generate incentives to reduce both family formation and fertility, and that these effects cannot be fully neutralized by counteracting inter-temporal or intergenerational transfers within families. We implement the model using calibrated simulations as well as panel data from 57 countries over 32 years. We find that PAYG tax measures account for a non-trivial part of the downward trends in family formation and fertility worldwide, especially in OECD countries.

Social security has become a subject of intense policy concern because of its financial vulnerability. We focus on a related, but no less important issue: social security’s impact on the real economy, particularly through demographic trends. Data from 57 countries show that average annual marriage net of divorce rate per 1000 people age 15 and over fell from 9.72 in 1960 to 6.40 in 1990, and that average total fertility rate (TFR) fell from 3.82 in 1965 to 2.07 in 1989. In the US, the proportion of married persons in the population age 15 and over fell from 67.9% in 1950 to 56.5% in 2000, and TFR fell from 3.09 in 1950 to 2.06 in 2000. These dramatic trends stem from secular trends in basic determinants of family formation and fertility worldwide. Our theoretical and empirical analyses indicate, however, that the defined-benefits, pay-as-you-go (PAYG) social security systems operating in most countries have independently contributed to these trends. More specifically, controlling for a number of theoretically relevant variables that affect the incentive to form families and bear children in a panel of 57 countries over the period 1960– 1992, we find that upward shifts in the ratio of social security’s pension benefits to GDP, which approximates the PAYG system’s equilibrium tax rate (PEN), have depressed the net marriage rate in the adult population (the rate of marriage net of divorce) and the total fertility rate. These effects are especially large in OECD countries, and they are generally not observed in countries where social security is a provident fund, rather than a defined-benefits, PAYG system. The insights we offer to rationalize these effects are based on a model of endogenous growth where human capital is the engine of growth, family choices include fertility and investment in children’s education (quantity and “quality” of children) by altruistic parents, and family formation itself is a choice variable. Theoretically, we focus on the way the scale of the PAYG system, as indicated by the level of taxes and defined-benefits, affects these decisions. The basic source of these effects is an externality inherent in the PAYG system. Old-age benefits are “defined”: they are fixed largely independently of one’s own contributions, and regardless of one’s children’s contributions, or whether one has any children. Individuals thus have little incentive to take such contributions into account in making fertility and related decisions, and the incentive to form a family is affected by the implicit subsidy that defined benefits provide to single (childless) households. The possible effects of PAYG systems on fertility, typically measured by the total fertility rate (TFR), were analyzed in several previous studies using different model specifications.1 All implicitly reach a common conclusion, however, which is a central proposition in Ehrlich and Lui (EL) (1998): exogenous increases in PAYG taxes must exert opposite effects on at least one of the parental choice variables.We go beyond all previous studies by considering the role of family formation, since families are more likely to bear children than single households.2 Consideration of family formation adds to our understanding of the way social security taxes affects TFR, since changes in the latter may come about in principle from changes in the percentage of families among all households, as well as from changes in the within-families fertility rate (FFR). This formulation also uncovers different welfare implications of social security for single and married households.Second, we develop our inferences using an endogenous-growth framework, which accounts for the impact of social security taxes at any given stage of development, as well as for transitional dynamic movements in family formation and fertility following an economy’s takeoff to a perpetual-growth regime. This enables us to apply the model empirically to a panel of countries that are in different stages of economic development. Our basic predictions are that exogenous increases in higher social security taxes generally depress on the margin both family formation and the number of children per family, and that the fall in TFR as a result of higher taxes therefore increases the financial vulnerability of the PAYG system. We derive these predictions as closed-form solutions of a benchmark model in which social security is the only source of old-age consumption. The results remain robust, however, even when we allow for inter-temporal or intergenerational transfers within families—i.e., we find that the effects of higher PAYG taxes on social security are not neutralized by offsetting movements in savings or old-age support from children to old parents. We assess the quantitative impact of exogenous increases in proportional payroll taxes via both a calibrated simulation analysis using US data, and a regression analysis in which we implement our model using international panel data. The results we derive via these two estimation methods corroborate one another. Although theoretically we treat social security taxes as an exogenous policy variable, empirically we allow for their possible endogeneity as well. We find that higher tax rates may have generated significant reductions in our key demographic variables. For example, our calibrated simulations using US data suggest that social security taxes may account for 26.8% of the fall in TFR, and for 27.8% of the fall in the proportion of married households in the US between 1950 and 2000.

Both our calibrated simulations for the US and our regression results corroborate our theoretical predictions and, when comparable, produce similar quantitative results. First, both analyses confirm our theoretical predictions that higher equilibrium tax rates, θ (PEN), depress our measures of family formation and fertility, p and n, thus TFR (2pn). Other predictions of our model concerning the effects of survival probabilities and measures of the opportunity costs of mating and child-bearing are also confirmed. Moreover, both analyses show that the absolute elasticities of p (or NETMARRY) and n with respect to our θ measure increase at higher levels of θ , and at more advanced development levels. In the simulations this is shown in parts A and B of Table 1, while in the regressions this is documented in part (3) of Table 3, comparing the OECD countries, which have relatively much higher PEN-levels, with non-OECD countries.Second, we are able to decompose the impact of PEN on TFR into its components. Our calibrated simulations identify directly the elasticities of p and n with respect to θ , which are found to be of approximate equal weight (see Table 1.A). Specifically, the 9.41% reduction in TFR from 2.3471 in 1950 to 2.1363 in 2000 in the US as a result of the rise of θ from 0.33 to 4.15 can be decomposed to a 4.79% reduction in p (50.9%) and a 4.62% reduction in n (49.1%). Third, the simulation and regression analyses also produce quantitative elasticity estimates of similar order of magnitude regarding the social security tax rate effect on our proxies of TFR, which are comparably defined in both analyses. The elasticity of TFR (= 2pn) with respect to θ in Table 1 at θ = 0.036 is found to be −0.090. The elasticity of our empirically measured TFR with respect to PEN is imputed to be −0.106 in the US, using the regression results of model 2 of Table 2 where NETMARRY is excluded as a regressor.15 Furthermore, the explanatory power ascribed to the effect of the increase in PEN on TFR over this period is also similar across the simulation and regression analyses. Our calibrated simulations in Section 1.4 ascribe 26.8% of the actual change in TFR to the rise in PEN from 0.33% in 1950 to 4.15% in 2000. Our regression results, as imputed for the US based on the modified OLS regression model 2 of Table 2 (with NETMARRY excluded), ascribe 31.8% of the actual change in TFR to the rise in PEN over the same period. This regression-based estimate has a standard error of 0.032.16 Ironically, the financial vulnerability of the defined-benefits, PAYG system, assessed by the magnitude of the tax rate required to support a given pension per recipient at a given per capita income level (see Eq. (8)), has increased over time partly as a result of the tax level. A key index of the system’s financial viability is the “worker support ratio” (WSR)—the number of contributing adult workers per social security pension beneficiary—which is theoretically given in our analysis by WSR = (π1/π2)ptnt . In the US, this index fell from 5.1 in 1960 to 3.3 in 1991. Over this period, the ratio of survival probabilities to adulthood relative to old age, (π1/π2) fell from 1.4747 to 1.2739, accounting for 33.6% of the fall in WSR. Based on Table 1 projections for the US, however, ptnt has further fallen from 1.1180 to 1.0539 as a result of the upward shift in PEN from 0.0232 in 1960 to 0.0468 in 1991. The rise in the social security tax rate therefore accounts for 13.6% of the fall in WSR in the US over this relatively short period Our analysis indicates that the structure of the defined-benefits, PAYG social security system has played a non-trivial role in contributing to the decrease in family formation and TFR worldwide, but especially in OECD countries. This, in turn, has also contributed to the decline in the potential domestic labor force in OECD countries. Even though our focus in this paper has been just on demographic variables, our expanded model in Section 1.3 offers potentially important implications for labor force participation, savings, and economic growth as well. These topics deserve, however, a separate treatment, which we plan to pursue in future work (see also Ehrlich and Kim, 2005).