Diane Lim Rogers, Eric Toder, Landon Jones

The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders.

Note: This report is also available in the PDF format, which many users find more convenient when printing.

---

Table of Contents

About the Authors
Future Labor Supply
Effects of Demographic Changes on National Savings
Projections of the Economic Effects of an Aging Population
Conclusions
Endnotes
References
Appendix : Simulations Based on an Alternative "Durables-Inclusive" Measure of Private Saving

Tables and Figures

Figure 1. Projected Shifts in Population
Table 1. Labor Supply Growth Rate Projections: Toder-Solanki versus Trustees
Figure 2. Labor Supply Annual Growth Rates: Toder-Solanki versus Trustees
Table 2. Measures of Labor Force Adequacy
Figure 3. Annual Labor Force Growth Rates
Table 3. NIPA Product-Side Measures of Saving, 1959–1997
Table 4. Private Income Shares and Saving Rates by Age Group, 1997
Table 5. Age-Related Government Expenditures per Capita, 1997
Table 6A. Key Demographic Factors Affecting Saving: Distribution of Children by Age of Household Head, 1997
Table 6B. Key Demographic Factors Affecting Saving: Population Shares of Key Groups Affecting Private and Public Saving
Table 7. Potential Effects of Demographic Change on Private Saving
Table 8. Projection of Age-Related Components of Government Spending Based on Demographic Changes
Table 9. Projections of Key Economic Variables in Base-Case Simulation
Table 10. Projections under Alternative (Longevity-Adjusted) Labor Supply Assumptions
Table 11. Projections Assuming Constant Private Saving Rate
Table 12. Projections Based on Constant Private Saving Rates for Adults Ages 20 to 64 and Adults Ages 65 and Over
Table 13. Projections Removing Effect of Children on Saving
Table 14. Projections Assuming Partial Ricardian Offset
Table 15. Projections under "Near Balanced Budget" Scenario
Table 16. Projections under "Balanced Budget in 2040" Scenario
Table 17. Projections Assuming Social Security and Health Transfers Grow with NNP
Table A1. Comparing NIPA-Defined to "Durables-Inclusive" Saving Measures
Table A2. Projections Based on "Durables-Inclusive" Measure of Private Saving

---

The retirement of baby boomers and the increase in the share of elderly in the population will create economic and fiscal stresses beginning in the second decade of the 21st century. These demographic developments, if not offset by changes in household behavior and government fiscal policy, will reduce the number of workers in relation to the population needing support and lower the national saving rate. The result will be slower growth in national income and consumption after 2010.

Over the next 40 years, the share of prime working-age adults will decline from about 59 percent of the population to about 56 percent (figure 1). The share of older adults (65 and over) will increase from just over 12 percent to almost 21 percent of the population. The higher costs of supporting these retirees will be offset partially by lower costs of supporting children, as the share of the population age 19 and under will drop from 29 percent to just over 23 percent.

This paper examines how demographic developments will affect the pattern of economic growth over the next 40 years. Demographic changes will alter labor supply, private and public saving rates, and the growth of national income and consumption. How these variables ultimately change will depend on behavioral adjustments by workers, savers, and governments.

The first section of this paper discusses the impact of demographic changes on future labor supply, while the second section explores the potential implications of these changes for private and public saving. The third section combines the labor supply and saving projections in a simple macroeconomic model, examining the effects of demographic changes on output, consumption, interest rates, and wages. It also examines the consequences of alternative assumptions about the following: workforce participation rates of older people, private saving behavior, and government fiscal policies.

Future Labor Supply

Projections show that the percentage of the population between ages 20 and 64 will decline after 2010 and the percentage of people over age 65 will increase dramatically. These changes reflect the short-run effect of the aging of baby boomers (1946–64 birth cohorts) and the long-run effect of reduced fertility and increased life expectancy. If labor force participation rates in each age group remain the same, the ratio of workers to retirees will decline sharply between 2010 and 2030 and continue to decline, although much more slowly, after 2030. The ratio of workers to dependents (children and older adults) will decline by less than the ratio of workers to retirees because the proportion of children in the population will also decline.

A decline in the share of workers in the population means that, if all else remains the same, output per capita and living standards will be lower than they otherwise would have been if the share of workers had remained stable. The following section briefly summarizes recent estimates of future labor supply by Toder and Solanki (1999) and compares them to projections by the Board of Trustees of the Federal Old Age and Survivors and Disability Insurance Trust Funds (Trustees) (1999). Later, these estimates will be used to project the effect of demographic changes on economic growth.

Projections of Future Labor Input

As mentioned above, the changing age composition of the population will reduce the share of workers and increase the share of dependent elderly. However, several factors may mitigate this decline. Labor force participation rates of older women are likely to increase over the next two decades as women born after World War II, with their higher lifetime participation rates, replace older women in the workforce. The increase in experience associated with an older workforce will raise average earnings and productivity per worker. General workforce education levels will also increase as less educated workers retire and are replaced by the better educated workers in later birth cohorts.

According to the Trustees, the number of workers covered by Social Security will increase by 19 percent between 1997 and 2040; Toder and Solanki find a similar increase of 22 percent over the same period if 1997 labor force participation rates by age and gender remain constant (table 1). However, when Toder and Solanki adjust for greater labor force participation of postwar birth cohorts of women, increased worker experience, and higher education levels, they discover that the effective labor supply increases by about 26 percent between 1997 and 2040.¹

While total labor supply will be higher in 2040 than it was in 1997, labor supply growth is expected to decline sharply over the first four decades of this century. For example, according to Toder and Solanki, the adjusted labor supply increases by 1.04 percent per year between 2000 and 2010, followed by continually declining annual rates—0.72 percent, 0.42 percent, and 0.27 percent—in each of the next three decades (figure 2). The Trustees' projection and Toder and Solanki's unadjusted numbers exhibit the same pattern of declining growth.

However, these projections hold only if past patterns of labor force participation by age and gender continue. But, of course, behavior within age and gender groups has not always been stable and may change in unforeseen ways. For example, Toder and Solanki (1997) report that the number of workers per capita increased by about 32 percent between 1965 and 1997. About half of this growth in overall labor force participation was attributable to an increase in the share of working-age people and the other half to changes in labor force participation rates within age and gender groups. The most important source of this growth was the increase in labor force participation rates of working-age women, which outweighed the reduction in labor force participation rates of older men. However, the growth in labor force participation rates of young and middle-aged women and the decline in participation rates of older men have virtually ceased in recent years.²

Consequences for Living Standards

The slowdown in the growth of the workforce will leave fewer workers in relation to the population they must support. The term "labor supply adequacy" refers to the ratio of the quality-adjusted workforce to the total consumption needs of the population. Labor supply adequacy is one factor influencing living standards of the population; the stock of capital and technological know-how are others.

A simple measure of labor supply adequacy is the ratio of the workforce to the total population. But not all people have equal consumption needs. For example, the government spends much more per capita on the over-65 population than it does on other age groups. So what is the appropriate ratio to use in measuring labor supply adequacy?

Table 2 reports alternative measures of labor supply adequacy. The Trustees use projections of the ratio of covered workers to Social Security beneficiaries because their concern is the adequacy of Social Security financing. Between 1997 and 2040, the covered-workers-to-beneficiary ratio is expected to decline dramatically from 3.35 workers per beneficiary to 2.03 workers per beneficiary—a drop of about 40 percent. This projection is consistent with the 42 percent decline, reported by Toder and Solanki (1999), in the ratio of people between ages 20 and 64 to those who are 65 and older. By any measure, the ratio of workers to retirees will be substantially lower in 40 years than it is today.

Labor supply adequacy measures that compare workers to retirees are best suited for measuring the ability to fund Social Security but less appropriate for measuring overall effects on living standards. When a broader measure of the population that workers support is considered—one that includes workers and children—the decline in labor supply adequacy is much smaller because the nonelderly population grows more slowly than the elderly population. Compared to the 42 percent decline between 1997 and 2040 of 20- to 64-year-olds compared to those over age 65, Toder and Solanki report much smaller declines—about 11 percent for the ratio of 20- to 64-year-olds to the 20-and-over population, 5 percent in the ratio of 20- to 64-year-olds to the total population, and only 2 percent in the ratio of the quality-adjusted workforce to the total population (table 2). These quality-adjusted labor supply adequacy projections are used for the "base-case" simulation presented in the third section of this paper.

The quality-adjusted labor supply adequacy projections assume that labor force participation rates by age will remain constant. However, improvements in health and longevity of the elderly may cause workers to retire at older ages so that their expected years in retirement remain constant. This possibility is considered in a "longevity-adjusted" simulation. Under this assumption, retirement ages will increase with longevity. For example, if life expectancy of older people increases by two years between 1997 and 2040, then the labor force participation rates of older workers at age x in 2040 will be similar to the participation rates of workers at age x – 2 in 1997.

Toder and Solanki present a simulation of future labor supply under the assumption that people ages 55 to 75 maintain a constant expected number of years in retirement after 1997. This assumption raises labor supply in 2040 by 4.4 percent, compared with the assumption of constant age-specific labor force participation rates. Even under this scenario, however, labor supply adequacy declines between 2010 and 2040, as baby boomers enter retirement.

The potential decline of the workforce in relation to the population it supports is only one component of the problem created by the aging of the population. There is also a potential decline in national saving rates caused by increased transfer payments to, and lower private saving rates of, the growing elderly population. The following section turns to the effects of demographic changes on private and national saving rates.

Effects of Demographic Changes on National Saving

Saving enhances society's ability to produce and consume in the future, but it requires the sacrifice of current consumption. The trend toward longer life spans and longer periods of retirement, in addition to projections suggesting growth in the share of elderly, implies that more saving will be necessary over the next several decades to maintain growth in living standards.

Private saving is defined as net national product (our measure of output) plus transfers and interest received from government debt minus taxes and private consumption. The personal and corporate components of private saving are not separately modeled; instead, it is assumed that households determine the sum. Public saving is what is left of taxes after subtracting transfers, interest paid on government debt, and government consumption. Public saving can also be thought of as government investment minus the budget deficit. National saving is net national product minus private consumption and government consumption. It also equals the sum of private and public saving. Our projections of saving depend on our assumptions about how each of the components going into this product-side definition evolve over time.

National saving equals all savings by households, businesses, and government and determines the growth of the national capital stock. Table 3 shows how the national saving rate, and its private and public components, has evolved over the past few decades. The long downward trend in the national saving rate has recently turned around because of recent improvements in government (particularly federal) fiscal positions. But the private saving rate, which has been declining since the 1980s, continues to fall. Private saving is the sum of personal and business saving. Personal saving has dropped sharply in recent years, while business saving has been relatively stable. If one believes that personal saving is a good indicator of "true" saving, this trend is alarming. Gale and Sabelhaus (1999) argue that a broader measure of personal saving, including accumulated wealth in retirement plans and consumer durables, paints a less bleak picture. Despite some theories that suggest that private and public saving are interdependent (to be discussed later), table 3 shows that the historical relationship between the two components is not obvious.

Assuming that current policies stay the same, the national saving rate will probably fall between now and 2040 for two main reasons. First, the increase in the share of elderly in the population means a relative increase in the numbers who draw down assets, thereby reducing total private saving. Second, the current transfer commitments of the government are heavily weighted toward the elderly in the form of Social Security and health (Medicare and Medicaid) transfers, implying that the ratio of transfer payments to national income will increase if per capita transfer payments for each age group grow at the same rate as income. Unless tax rates increase, or other government spending falls, the growth in transfers will significantly reduce public saving. The following sections examine how future demographic change might affect all types of saving.

Influences on Private Saving

There are many theories about what determines saving, but none is definitive. Several of the theories—especially the life-cycle model—suggest that age is an important determinant of saving behavior (box). Although past changes in demographics, on their own, cannot explain the recent trends in the saving data, future changes in demographics—in particular the ratio of elderly people to working-age people—will be dramatic. This paper considers how demographic changes might affect saving in the future if current differences in saving rates among age groups persist.

The approach taken in this paper is very similar to the one by Leibfritz et al. (1996) described in the box. The central assumption in the macroeconomic model presented here is that age-related patterns in private saving, government consumption, and transfer payments hold up over time. Various sources of household survey data are used to determine consumption as a share of annual income by age group, and the propensities to consume are preserved over time, with some adjustment for changes in the number of children per household. (Each age group's saving rate will rise if fewer children come into their group.³) Because propensities to consume vary by age, aggregate private saving will evolve as the shares of population at different ages change.

THEORIES OF SAVING One of the most prominent theories of saving is the life-cycle model (LCM), which predicts that people will save in order to translate their fluctuating levels of income into smooth paths of consumption. Because earnings tend to rise first and then fall over a lifetime, smooth consumption implies that households borrow when young, save when middle-aged, and spend savings, or "dissave," when old. The pure version of the LCM assumes that people consume all their wealth by death (no bequests) and that people have unlimited access to capital markets at a single interest rate paid by borrowers or received by savers. Given these assumptions, the pure LCM implies pronounced differences in annual saving rates by age, with consumption fluctuating with changes in permanent income but not transitory income. Many economists, including Bosworth (1996), are unimpressed with the life-cycle theory, because it fails to explain recent trends in saving. However, empirical evidence is mixed; cross-country data appear to be more supportive of the LCM, while the U.S. historical time series or household survey data are less so. Recent declines in private saving have occurred even as the baby boomers are entering their high-saving years. The household survey data suggest that the decline in saving has been widespread and not related to differences across age groups or changes in the age distribution (see Bosworth, Burtless, and Sabelhaus 1991). Finally, when productivity growth slowed after 1973, saving decreased, instead of increasing as the LCM would predict. (If consumption is a function of expected future income, then an expected slowdown in future income should decrease consumption and increase saving.) Research has exposed other inconsistencies between the LCM and empirical facts. Given the pattern of age-earnings and age-consumption profiles, the pure LCM cannot explain the size of the aggregate capital stock; hence, intergenerational transfers and bequest motives ("dynastic models") may be important in determining saving and asset accumulation. Also, consumption appears more sensitive to fluctuations in current or transitory income than theory would suggest. Therefore, rules of thumb, mental accounts, and other theories involving less-than-lifetime planning horizons might better explain saving behavior. More recent explanations of the downturn in saving have considered the role of uncertainty and the need for precautionary saving (less of such saving is needed with recently expanded credit markets) and the expansion of public transfers to the elderly (everyone saves less on their own in anticipation of greater government support in retirement). Economists have taken a variety of approaches to embedding demographic effects into macroeconomic models. In any data set, one can find particular empirical relationships between saving and consumption and the age distribution of the population, but it is unclear if such relationships will be maintained in the future. Bosworth (1996) and the Congressional Budget Office (CBO) (1997) both model saving as a fixed share of income so that the saving rate does not depend on the age distribution. This approach is reasonable if one does not believe that age-related saving rates in current data will remain stable over time. The Federal Reserve Board’s (1996) macroeconomic model of the United States specifies an aggregate consumption function that represents a combination of life-cycle and liquidity-constrained behavior based on different propensities to consume out of different types of income and wealth (see Brayton and Tinsley 1996). Because income and wealth composition vary by age, the consumption function implicitly allows changes in the age distribution to affect the aggregate saving rate. Auerbach and Kotlikoff (1987) use an explicit overlapping generations model that tracks the paths of consumption that maximize lifetime utility subject to lifetime budget constraints and perfect capital markets. In the Auerbach and Kotlikoff model, changes in the age distribution affect the aggregate saving rate by changing the relative numbers of people at different (saving and dissaving) points in the life cycle. An OECD analysis by Leibfritz et al. (1996) starts with age-dependent saving rates, keeping track of trends in the numbers of people in different age groups and assuming behavior by age group is constant. While the central case considered generally reflects a life-cycle pattern of saving by age, they also adjust the age-saving profiles to consider patterns possibly more consistent with other theories of saving (such as bequest motives). This approach of modeling age-based saving rates differs from an explicit utility-based overlapping generations framework, however, because it provides no accounting for lifetime utility and lifetime budget constraints. It also is an empirically based, reduced-form type of approach that does not apply the pure LCM.

Table 4 shows the saving rates by adult age group that are derived from household survey data for 1997, the starting point in the simulation. "Private income" in the table is a measure that includes both personal and corporate income. It is equal to disposable income, as shown in the National Income and Product Accounts (NIPA), plus corporate retained earnings. Private saving is computed as the difference between private income and private consumption. In making this computation, labor income, capital income, taxes, transfer payments, and consumption are calibrated to NIPA data. The distribution of income, transfer payments, and consumption by age is based on other data sources. Labor income by age is from Census data, while the distribution of capital income is from the distribution of wealth in the Survey of Consumer Finances.⁴ The age distribution of transfers comes from the various program agencies.⁵ Household consumption by age of household head comes from the Consumer Expenditure Survey. Taxes by age are assigned by applying national effective tax rates on capital, labor, and consumption to each age group's tax bases.

The data suggest strong implications for private saving associated with changes in the relative numbers of both young and old people. Calibrating aggregate private saving to the 6.55 percent rate observed in 1997, we obtain a wide distribution of saving rates across the adult age categories. Middle-aged adults are the largest savers, saving nearly a quarter of private income. In contrast, the youngest adults have the largest negative saving rate (over a quarter of their private income) and also produce more total dissaving than the elderly (negative 5.21 percent toward the aggregate private saving rate versus negative 0.07 or negative 0.19 percent, as seen in the fourth column of the table), because they have a much larger share of private income. This effect at the young end of the age distribution has received little attention from other researchers; for example, Leibfritz et al. focus on saving rates of the elderly relative to the rest of the population.

This high rate of dissaving by young adults may reflect inconsistencies among the several sources of data combined to derive saving rates by age. To test the sensitivity of results to the assumed distribution of saving rates by age, simulations with alternative age patterns of saving were run. The high measured rates of dissaving among young adults may in part reflect the fact that NIPA treats purchases of durable goods as current consumption. In the appendix, we examine the consequences of using a measure of private saving that treats net increases in the stock of consumer durables as saving.

Previous estimates of the effects of dependency ratios on saving rates have found a more negative effect of the elderly dependency ratio than the child dependency ratio; as mentioned, those analyses did not look at differences in consumption by age among working-age adults.⁶ To examine how much this matters, a more aggregate distribution of private saving rates across adults is considered, one that differentiates only between adults between ages 20 and 64 and adults 65 and older. The distribution of saving rates corresponding to this view is also shown in table 4. The 64-and-under population saves about 8 percent of private income, while the 65-and-over population dissaves at a very slow rate of 0.08 percent of private income, so that the aggregate private saving rate is maintained at the 1997 level of 6.55 percent. In the model simulations presented here, this and other alternative patterns of saving rates by age are examined. In addition, the model explores the saving rate's sensitivity to assumptions about how the number of children per adult affects consumption.

The model uses a straightforward method of capturing the effect of demographic changes on saving, but it leaves out important potential effects. Consumption and saving by age are simply fixed shares of private income. Thus, both wealth, which could be defined in various ways, and changes in the rate of return to capital are assumed to have no effect on saving. In addition, the assumption that consumption propensities by age are fixed makes the model subject to the "Lucas critique," a term applied to the use of equations that might not be stable as policy changes. In this particular context, consumption and saving propensities may depend on pension and transfer policies, which could change as the age distribution changes. Finally, there may be systematic differences among cohorts in saving propensities.⁷

Influences on Public Saving

The nation saves through its government sector when governments collect more than they spend. According to NIPA data, public saving equals taxes minus government consumption (public goods and services), transfer payments, and interest on the debt. This measure of public saving is equal to government investment minus the deficit. By this measure, public saving fell from the 1960s to the mid-1990s but has risen in recent years, largely reflecting changes in the federal budget position. Government investment has been relatively stable. While increases in health-related costs continue to put upward pressure on government transfers, and accumulating debt has necessitated greater payments of interest, higher taxes and decreases in government consumption in the 1990s have created overall surpluses.

Future public saving will be profoundly affected by the aging of the population because the major government transfer programs—Social Security and the health programs (Medicare and Medicaid)—disproportionately benefit the elderly. Part of government consumption, expenditures on public education, goes disproportionately to the young, who will become a smaller share of the population. This will reduce government spending in the future but only partially offset the upward pressures from the transfer system. Public education spending was about half of Social Security and health transfers in 1997, and the increase in the elderly population will be much more pronounced than the decrease in the number of children.

Public saving is projected by assuming that the age-related patterns in transfers and education spending found in the initial-year data largely hold up in the future. These patterns are shown in table 5. For each of these programs, we assume that per capita expenditures by age group will grow with national output per capita. This makes total government spending grow relative to output, because of the concentration of transfer payments among the more numerous elderly.

Regarding Social Security, the projections presented in this paper of changes in aggregate benefits are based solely on the evolution of the age distribution. CBO's projections, in contrast, account for changes in program structure, adjusted for differences between CBO's and the Social Security Administration's (SSA) economic assumptions. But using both the age-related pattern in the data and SSA's aggregate projections is somewhat inconsistent because some of the movement in the aggregates reflects program changes that may affect the age distribution of benefits. In this paper, the demographic changes drive what happens to aggregate Social Security transfers, assuming the program's benefits-by-age structure stays the same. Despite the slightly different methodology, the projections presented here for the growth of Social Security turn out to be quite similar to those of CBO.

The projections presented in this paper of health transfers differ more substantially from CBO's, mainly because it is assumed that per-beneficiary health costs by age grow with productivity (output per capita), but not any faster. Recent research suggests productivity growth has offsetting effects on per-beneficiary costs, but the net effect is to raise costs. Prolonged longevity and improved health may allow health costs for those age 85 and older to fall, but technological advances may continue to put upward pressure on per capita costs, holding health status constant. Cutler and Sheiner (1999) believe that the technological cost factor will outweigh the improved health factor. Lee and Skinner (1999) also conclude that technology will increase the overall share of GDP spent on health care, but at a substantially lower rate than government projections. The projections presented here of the growth in aggregate health spending simply reflect demographic effects, but not the effect of sector-specific increases in medical costs. CBO's memo on its long-term model shows that its assumption about additional medical costs is critical in driving its result that health spending more than doubles as a share of national income over the 1997–2040 period. When CBO's assumptions about increases in health costs are applied to the model presented in this paper, very similar projections (not shown in this paper) are produced.⁸

Putting It Together: Demographic Influences on National Saving

Given that national saving is the sum of private and public saving, the effects of demographic change on national saving are driven by the age-related patterns of private saving rates and the age-related patterns in government transfers. Each age group's reliance on public versus private sources of funds is held constant. For example, the elderly will continue to do the bulk of their consuming through the public sector. The key age distributions affecting the two components of saving are summarized in tables 6A and 6B. Table 6A shows the 1997 distribution of children among the adult age categories. This distribution is held constant over time, but the total number of children, relative to the total number of adults, changes according to the Trustees' midrange population projections. Because children are disproportionately found in households headed by young adults (between the ages of 20 and 44), any change in the relative number of children will disproportionately affect the consumption and saving of those young adult groups. Table 6B shows why these young adults—who are "high spenders" according to the 1997 saving rates shown previously (in table 4)—are an important part of the private saving story. There will be relatively fewer of them in the future and they will have fewer children. In contrast, the share of high-saving middle-aged people will be slightly higher in 40 years, while the share of elderly dissavers will increase dramatically. A focus on the oldest groups alone would suggest a dramatic fall in private saving. The drop in the share of the younger group, with its high consumption propensity, and the increase in the 45-to-64 age group, with its high saving propensity, push aggregate private saving upward.

The demographics affecting public saving are more straightforward. Although the decreasing share of children will reduce the growth in public education spending, the driving force affecting government outlays in the future will be the increase in the "high public spending" (65 and older) share of the population. Social Security transfers are skewed toward people ages 65 and older, with very little going to the remainder of the population. Thus, Social Security transfers as a share of national output are expected to increase substantially over time. Health transfers are enjoyed by more people throughout the age distribution but also favor older adults. Because the growth of the oldest elderly will be even more significant than the growth of the elderly as a whole, health transfers should also increase dramatically as a result of demographic change alone.⁹

Tables 7 and 8 quantify these effects. Table 7 shows that the age-related private saving rates in table 4 imply that changes in the age distribution of adults will make the private saving rate higher in the future than it is today. The evolution of the adult age distribution alone implies that the private saving rate would first rise significantly—from 6.55 percent of private income in 1997 to 8.71 percent of private income in 2010—and then fall. By 2040, the saving rate would remain nearly a percentage point above the 1997 level. This rather rosy view of the implications of demographic change for private saving is due to the fact that the biggest private spenders in the economy, the young adults, are shrinking in relative numbers between 1997 and 2040. Combined with the projection that the share of high-saving middle-aged people will rise at first, these trends cause the private saving rate to increase substantially in the next 20 years. In the subsequent 20 years, although the share of young adult spenders continues to fall, the share of middle-aged savers also drops slightly while the elderly share rises sharply. Thus, the overall private saving rate falls from its peak but remains higher than in the initial year, 1997.

The effect of changes in the number of children relative to adults reinforces the optimistic story about private saving, as table 7 shows. The marginal effect of this "kid factor" builds over time as their share falls. Although changes in the mix of adults is a more significant influence over most of the 1997-2040 period, the effect of there being fewer children relative to adults becomes equally important by 2040. Alone, it boosts private saving by nearly a percentage point over the 1997 rate.

The prediction of the effect of demographic change on private saving depends on distinguishing among prime-age adults in terms of their saving rates. If, instead, the only differentiation is between people who are 20 to 64 and people 65 and over, a very different story about how the aging adult population affects aggregate private saving emerges, as shown in the last row of table 7. Although saving rates initially rise slightly (because of the slight growth in the younger share of the population), they eventually fall below the 1997 level as the share of older adults rises to over 20 percent of the population. Still, the private saving rate in 2040 is only slightly lower than in 1997.

Table 8 examines what demographic change may imply for the evolution of age-related components of government spending and hence public dissaving. Allowing transfers per capita by age category to rise with national output (net national product, or NNP), and then accounting for the changes in the age distribution of the population, the table shows the implied changes in their aggregates over time as a share of NNP. Transfer programs are skewed toward older adults, so transfers as a share of NNP rise as the share of the population over 65 increases. Public education spending is skewed toward the young, so the aggregate as a share of NNP falls as the relative numbers of young people decrease. These trends will put substantial upward pressure on public spending because the drop in education spending is outweighed by transfer spending and the growth in the share of older adults. Demographic trends raise spending of all levels of government by about 6 percent of NNP between 1997 and 2040.

Possible Interdependencies between Private and Public Saving

Changes in private saving could be unrelated to changes in public saving, as the above calculations suggest, but private and public saving may affect one another. The theory of Ricardian equivalence asserts that changes in the intergenerational distribution of taxes and transfers will be perfectly offset by changes in private transfers, so that changes in public saving would ultimately have no effect on national saving. However, the extreme case of full offset, or "strict Ricardian equivalence," has been rejected by most empirical studies (Leibfritz et al. 1996). Thus, a simple rule to look at the effects of a partial Ricardian offset is CBO's approach of letting private saving increase by half the decrease in public saving. Most of the simulations reported below assume that private saving is independent of public saving; but one variant assumes a partial Ricardian offset, according to CBO. With such an offset, there are many different ways that aggregate private saving can increase. For example, it can increase proportionately across age categories according to each category's initial saving or it can increase in proportion to the distribution of tax liabilities. The first approach is taken in the following analysis.

Projections of the Economic Effects of an Aging Population

This section uses a simple macroeconomic growth model to show how demographic changes may affect total output, wages, interest rates, saving rates, and consumption through 2040. The model illustrates how the economy might evolve under varying assumptions about labor supply growth, the effects of age composition of the population on private saving rates, and government fiscal policies.

The Macroeconomic Model

The macroeconomic model is a simple neoclassical growth model.¹⁰ Output depends on effective labor supply, the amount of capital (capital stock), and the state of technological knowledge. Labor supply and saving are exogenously determined, but alternative assumptions about how labor supply and private saving rates evolve over time are considered. The model in its current form has no foreign sector, so that the growth of the capital stock depends only on net national saving. The parameters of the model are calibrated to data from the 1997 National Income and Product Accounts (NIPA).

The model has only one production sector, but it has seven representative households—one for each of seven groups based on the age of the head of the household (20 to 34, 35 to 44, 45 to 54, 55 to 64, 65 to 74, 75 to 84, and 85 and over). Children who are 19 or younger are assigned to the seven household groups in proportion to their current distribution among households by age of household head, as reported in the Current Population Survey (CPS). Children do not have their own income or consumption, but their presence affects the saving rates of households.

Output is measured as NNP and is determined by a Cobb-Douglas production function, with weights on labor and capital that reflect the 1997 proportions of national income going to labor income and capital income in the NIPA, respectively.¹¹ Increases in total factor productivity are captured by a parameter that scales up the entire production function and grows over time. This represents increases in output that would result from technological improvements even if the quantities of labor and capital did not change.

Both the private sector and government sector employ capital and labor in production, but one production function covers both sectors; in other words, both sectors are assumed to follow the same technology in producing output. Given this production function, the growth of national output depends on the growth of total factor productivity, labor, and the capital stock. The demographic changes over the next 40 years will significantly affect the growth of labor and capital; they may also affect growth in total factor productivity, but no position is taken on the direction or magnitude of any such effect.¹²

According to CBO assumptions (1997), total factor productivity increases by 1 percent per year. The projections by Toder and Solanki (1999) of the quality-adjusted workforce, as summarized in the first section of this paper, determine how labor supply changes over time. The model can incorporate different assumptions about the growth of both total factor productivity and labor supply.¹³

The capital stock in any year is equal to the previous year's capital stock plus net saving. Net saving, in turn, is equal to the difference between NNP and consumption, where consumption is the sum of private and government consumption. The key factor determining the evolution of the capital stock, therefore, is the percentage of NNP devoted to private and public consumption.

Private Consumption

Private consumption as a share of NNP depends on the ratio of private consumption to private income and the ratio of private income to NNP. Private income is NNP minus all taxes (including indirect business taxes) plus transfer payments and interest on government debt. Private income differs from the NIPA concept of disposable income because we do not distinguish between corporations and individuals; corporate retained earnings are included in both private income and private saving.

As discussed in the second section of this paper, alternative consumption functions are used in the model presented here. Consumption is divided into nonhealth consumption and health consumption. The base-case simulation maintains a fixed ratio of each type of consumption per adult equivalent to income per adult in each age group.

Two equations are used: one for private nonhealth consumption and another for privately financed health consumption (out-of-pocket expenses plus insurance premiums paid by households and employers). The equations represent both components of consumption as proportional to private income (less government health transfers) multiplied by the ratio of adult equivalents to adults.¹⁴ Including the ratio of adult equivalents to adults in the equations makes the ratio of consumption to income fall as the child dependency ratio increases. This means that child care costs displace saving instead of other consumption. An alternative simulation removes this ratio from the equations, which makes consumption of child care costs displace other consumption instead of affecting the saving rate.

Taxes and Transfers

The ratio of private income to NNP depends on the evolution of taxes, transfer payments, and government interest on the debt. Taxes are classified by category, and rates are set equal to the ratio of 1997 liability to income for each category. The four tax categories are (1) labor income taxes, (2) capital income taxes, (3) consumption taxes, and (4) taxes on nonhealth transfers (including Social Security benefits). In setting the initial tax rates from NIPA data on federal, and state and local, tax liabilities, the following allocations were made: payroll tax revenues to labor income taxes; corporate income tax, property tax, and estate tax revenues to capital income taxes; and sales and excise tax revenues and customs duties to consumption taxes. Individual income tax liabilities are divided among labor income taxes, capital income taxes, and taxes on nonhealth transfers, based on rough estimates of the share of federal tax payments attributable to each source of income.¹⁵ In the base-case projection, taxes remain roughly constant as a share of NNP over time, although they are affected to some degree by changes in the ratios of different private income sources and consumption to NNP. Transfer payments per capita for each age group are assumed to rise in proportion to the growth of NNP. As described in the previous section, projected transfer payments rise relative to NNP over time due to the aging of the population. Changes in interest payments relative to NNP reflect variations in public-sector debt, due to past government deficits or surpluses.

In the base-case scenario, private income rises relative to NNP due to both rising transfer payments and the effects of these transfer payments on government interest payments via the accumulation of public-sector debt. This makes the ratios of both private consumption and private saving to NNP rise over time in relation to the ratios of consumption and saving to private income.

Public Consumption

Most public-sector consumption is assumed to grow in proportion to NNP; the model does not allocate the benefits of defense, highways, and most other public goods by age group. But it does incorporate an age distribution for public educational expenditures, which are concentrated among the youngest age groups. The result is that projected public educational expenditures as a share of NNP decline along with the share of young people in the population.

While public-sector consumption declines in relation to NNP, public saving also drops. This reflects the increase in transfer payments that reduces public income, defined as taxes minus transfers and interest payments.

Wage Rate, Rate of Return, and Income Distribution across Age Groups

The model computes the annual wage rate per quality-adjusted labor unit and the rate of return per unit of capital. Combining NNP with the constant factor shares from the Cobb-Douglas production function gives total labor and capital income in any year. The wage rate is then equal to total labor income divided by labor supply; the rate of return on capital equals total capital income divided by the stock of capital.

The distribution of income among age groups depends on a variety of characteristics, including factor prices (the wage rate and rate of return) and the shares of labor income, capital income, interest on the public debt, and transfer payments received by each age group. The distribution of labor income by age group in 1997 comes from CPS data on earnings by age group and evolves according to the projections of labor supply by age group discussed in the first section of this paper. The shares of transfer payments depend on the 1997 amounts of relative transfer payments per capita by age group and population shares.

The distribution of income from capital among age groups is complicated. For 1997, capital income is allocated among age groups in proportion to wealth data reported in the Survey of Consumer Finances (SCF). The same distributions are assumed for both wealth in the form of capital assets and private wealth in the form of government bonds. Asset accumulation over time is not tracked by age group, partly because it is unclear how bequests should be distributed by age. Instead, private wealth is increased by the amount of private saving and distributed among age groups based on the benchmark distribution of per capita wealth by age. This implicitly assumes that intergenerational transfers operate to preserve the current distribution of private wealth by age.

Base-Case Scenario

Table 9 reports the base-case simulation of the path of selected macroeconomic variables between 1997 and 2040. Tables 10 through 17 report alternative simulations based on varying assumptions about future labor supply growth, private saving behavior, and government fiscal policy.

In the base-case simulation, private consumption is determined using the two equations—one that determines private nonhealth consumption and another that determines privately financed health consumption—mentioned above. Labor supply grows according to the projections of Toder and Solanki (1999) described in the first section of this paper. Government consumption (except for education) remains a fixed percentage of NNP. Age-related components of government spending (education outlays and transfer payments) increase to keep the ratio of per capita spending to NNP fixed in all age groups. The concentration of transfer payments among the more numerous elderly causes the ratio of transfer payments to NNP to rise by almost 7 percent of NNP between 1997 and 2040. Average tax rates on labor income, capital income, consumption, and nonhealth transfer payments are kept fixed at their 1997 level.

The growth of NNP per capita decreases after 2010, due mainly to a decline in labor supply per capita. The growth of capital stock per capita also declines after 2010, but the capital-to-output ratio continues to increase, and rates of return continue to decline through 2030. The private saving rate rises through 2010, then declines slightly but remains higher than its 1997 level throughout the period; the national saving rate plummets after 2010 as increased transfer payments drive government deficits up. (Tax revenues as a percentage of NNP increase slightly, reflecting the increase in taxable transfer payments and consumption as a percentage of NNP.) Government debt as a share of NNP quintuples between 2020 and 2040. Even though government is absorbing more of private saving, interest rates in 2010 are still below their 1997 level. This continued adequacy of capital occurs, in spite of the decline in saving, because the decline in workforce growth reduces the demand for capital in production.

Growth rates of per capita private income and consumption decline between 2010 and 2020 but stabilize afterward. Private income and consumption grow faster than NNP because they are maintained by increased government transfer payments. In the long run, therefore, the base-case simulation implies an unsustainable path for the economy. But there is little slowdown in the growth of living standards through 2040.

Alternative Labor-Supply Behavior

Table 10 shows the effects of assuming faster labor supply growth. In this simulation, we assume that labor force participation rates of older workers increase to maintain a constant length of retirement as life expectancy increases after 1997. By 2040, labor supply per capita is 4.4 percent higher than in the base-case simulation.

The increase in labor supply makes NNP and consumption 4 percent higher per capita in 2040 than in the base-case simulation, but it is not sufficient to keep the growth in NNP from declining after 2010. Because transfer payments rise with NNP, the slightly faster growth rate does not ease the government fiscal position. The national saving rate is just slightly higher than in the base case, because households save some of their increased factor income.

Alternative Saving Behavior

Tables 11 through 14 show projections based on alternative assumptions about private saving behavior. Tables 11 and 12 alter the influence of changes in the relative numbers of adults at different points in the life cycle; table 11 assumes there are no age-related differences in private saving rates, while table 12 differentiates only between two groups of adults (basically retirement-age versus others) in terms of private saving rates. Tables 13 and 14 consider alternative specifications of the private consumption function. Table 13 removes the effect of changes in the ratio of children to adults, while table 14 presents a partial Ricardian offset in which private saving reacts to changes in public saving.