P

Total

Output

There does appear to be a positive correlation between the growth of output per worker and the ratio of investment to output. In panel Ba larger set of developed countries is considered. The relationship is now more diffuse.","

At first thought this result may be surprising. Previous sections have emphasized the point that the growth of output is related to the proportion of resources devoted to capital formation. The observed weak relationship is not, however, inconsistent with our analysis.

First, factors other than capital formation are important in determining the rate of growth of output. As noted in the Denison study, advances in knowledge, increases in education, and other

For the seven countries, the correlation coefficient is .58: for all 11 countries it is .36.

In principle, the relevant growth rate is growth of total output; however, most comparisons use output per worker. A scatter diagram of growth of total output against the investment ratio shows the same patterns as those in chart 3. The correlation coefficient falls from .74 for the seven countries to 38 for all 11 countries.

factors are as important as the increase in capital in explaining growth of output for the United States. Similar results hold for other countries." Second, in comparing countries it is necessary to consider differences in natural resource endowments, legal and political institutions, attitudes toward work and income, etc. Third, it can be shown that in the long run, the growth rate is independent of the proportion of output devoted to capital formation.

Long-run growth rates are independent of saving ratios.

The expected relationship-that growth rates depend upon the saving ratio-assumes that capital-output ratios remain constant. As capital accumulation proceeds, however, output grows but not as rapidly as capital (other things being equal), and capital-output ratios tend to increase. This rise in capital output ratios slows down the rate of growth.10

It is true that an increase in the proportion of output devoted to investment raises the growth rate initially. As capital accumulation proceeds, however, the increase in the saving ratio raises the growth path but not the growth rate.

This is depicted in chart 4. Suppose that output is growing on the path PP'. In year T, there is an increase in saving (decrease in consumption). Output begins to grow along QR but gradually approaches the path SS'. In the long run, growth rates are independent of the saving ratio.

Growth generates saving.

One of the puzzles that has fascinated economists is the secular constancy of the ratio of aggregate saving to disposable income. As shown in chart 5, this ratio has fluctuated around an average value of approximately .08 since 1900. This contradicts earlier speculations that the saving ratio would increase as income grew.

'See Edward F. Denison, Why Growth Rates Differ, The Brookings Institution, 1967.

In making cross-country comparisons, there are also differences in definitions, quality of data, periods covered, etc. It is doubtful that such differences are large enough to change our basic conclusion.

The growth rate of total output may be shown to equal the saving ratio multiplied by the ratio of output to capital (the reciprocal of the capital-output ratio). Assume, for example, that the saving ratio is 0.1. If the capital-output ratio is 3 (the output-capital ratio is .33), the growth rate is .033-i.e., 3.3 percent per year.

SOCIAL SECURITY

There have been several explanations of this constancy. One of the most ingenious has been provided by Franco Modigliani of the Massachusetts Institute of Technology." This explanation is based on the life-cycle hypothesis of consumption, to be used later in the discussion of the impact of social security on saving.

In its strictest version, the life-cycle hypothesis states that individuals save only to smooth out the stream of lifetime consumption. During their working lives, individuals save, increasing their wealth (savings deposits, securities, pension claims, etc.). After retirement, they dissave, drawing down wealth to maintain planned consumption. No estate is accumulated, nor is there saving for emergencies or other purposes. The resulting pattern of income, consumption, saving, and wealth accumulation is shown in chart 6. It is assumed that the individual enters the labor force at age 25, retires at age 65, and dies at age 80.

Suppose that population is stationary and that productivity is constant. With a stable age distribution of population, the saving of workers

"See Franco Modigliani, "The Life Cycle Hypothesis of Saving, the Demand for Wealth and the Supply of Capital," Social Research, No. 2, 1966.

BULLETIN, JULY 1975

Saving by workers will exceed dissaving by retirees. In the aggregate, saving will be positive and a constant ratio of aggregate income." The saving ratio will depend upon the rate of economic growth.

SOCIAL SECURITY, SAVING, AND
CAPITAL FORMATION

The discussion now turns to the relationship of social security to saving and capital formation. Three general questions are considered. First, what is the impact of social security on saving? Second, assuming that social security reduces saving, is it desirable to increase saving? Third, if it is desirable to increase saving what alternative policies can be considered?

What is the impact of social security on saving?

If the social security program were simply a "pay-as-you-go" tax-transfer system in which current workers did not consider anticipated benefits in making saving decisions, the analysis would be relatively simple. The reduction in disposable income resulting from the payment of payroll taxes would lead workers to reduce both consumption and saving. The increase in disposable income resulting from benefits would lead the retired to increase consumption and saving.

The effect on aggregate saving would depend upon the level of benefits (assumed equal to taxes) and the difference between the marginal propensities to consume" of workers and beneficiaries. For example, in 1973, old-age and survivors insurance (OASI) benefits were approximately $46 billion. Assume that workers' propensity to consume is .9 while that of beneficiaries is 1.0. Then the reduction of saving would be $4.6 billion, equal to the reduction in saving by workers.

A similar result is obtained if productivity is grow ing at a constant rate.

13 Studies of the effect of social security on saving generally restrict consideration to retirement benefits. The analysis here is also limited to retirement benefits. "The marginal propensity to consume measures the proportion of additional income that is spent for consumption.

10

Workers anticipate future benefits in their saving decisions.

The problem is not that simple. Workers undoubtedly do anticipate future benefits to some extent in making their spending-saving decisions. In the life-cycle model described earlier, if perfect foresight and quid pro quo are assumed, personal saving would be reduced during worklife and dissaving correspondingly reduced during retirement. That is, the worker would view social security contributions as compulsory public saving and therefore a substitute for private saving. With current financing, payroll tax revenues are used to finance benefits and are not accumulated in a fund. Therefore, aggregate private saving would be reduced.

What is the empirical evidence?

Some empirical evidence suggests that social security increases saving.

As noted above, there has been no discernible downward trend in the percentage of income saved since the introduction of the social security program. Moreover, indirect evidence provided by two studies of the relationship between private pension coverage and saving behavior suggests that social security might have increased personal saving.

A 1965 study by Phillip Cagan of Columbia University analyzed the saving behavior for 1958-59 of more than 15,000 Consumer Reports subscribers. Cagan found that individuals covered by pension plans saved more than those not covered." He attributed his results to a "recognition effect": participation in a pension plan calls attention to retirement needs and leads individuals to increase provision for retirement.

Recall, however, that for a stationary economy the life-cycle model would imply zero aggregate saving even without a social security system. The reduction in saving implied by the life-cycle hypothesis assumes that the economy is growing.

"Phillip Cagan, The Effect of Pension Plans on Aggregate Saving, National Bureau of Economic Research, 1965,

"Cagan's results have been criticized by Alicia Munnell. See Alicia Munnell, The Effect of Social Security on Personal Saving, Ballinger, 1974 (study based on Ms. Munnell's dissertation completed at Harvard Uni versity and financed by a Social Security Administration grant). A summary of the study appears in the Social Security Bulletin, November 1974, pages 29-30.

SOCIAL SECURITY

A second 1965 study, by George Katona of the University of Michigan's Survey Research Center, was based on personal interviews with approximately 2,000 families in 1962-63.18 Like Cagan, Katona found that pension plans increase personal saving. Katona's explanation was in terms of "goal feasibility" and "level of aspirations"that pension plans made retirement goals feasible and that workers both raised their retirement income goals and intensified their saving effort.

The explanations offered by these studies imply that the individual's preference for future vs. present income is changed by participation in a pension plan. That is, as a result of participating in a pension plan, workers reduce consumption (and correspondingly increase saving) during their working years in order to finance increased consumption during retirement. These conclusions have generally been extended to social security.

Recent studies suggest that the social security program reduces saving.

Martin Feldstein of Harvard University has offered an alternative explanation of the empirical evidence consistent with the predictions of the life-cycle model." Feldstein hypothesizes that the social security program has lowered the age of retirement. With a shortened period of earnings and longer retirement period, the worker would have to increase his saving rate. This would offset the reduction in saving resulting from the substitution of anticipated retirement benefits for personal saving. The net effect on personal saving would depend upon the relative strength of these offsetting forces.

An example may clarify his hypothesis. Assume that a worker enters the labor force at age 25, earns $10,000 per year, plans to retire at age 70, and expects to die at age 80. Suppose that he allocates his lifetime resources to provide a retirement income of $5,000 per year and plans to leave no estate. If the interest rate is 5 he would percent, have to save $242 per year-2.42 percent of his earnings during his working life.

Suppose now that he is promised a retirement

George Katona, Private Pensions and Individual Saving, University of Michigan, Survey Research Center, 1963.

Martin Feldstein, "Social Security, Induced Retirement, and Aggregate Capital Accumulation," Journal of Political Economy, September-October 1974.

BULLETIN, JULY 1975

benefit of $2,500 per year. With no change in preferences, he would reduce his saving by half, saving $121, or 1.21 percent of his income, per year. Suppose, however, that he is required or induced to retire earlier, say at age 65. Then, to maintain the same retirement income, he would have to save $215 per year, or 2.15 percent of his earnings. Thus, the increase in saving resulting from a shorter work-life tends to offset the reduction in saving resulting from substitution of the anticipated benefit for private saving.

20

The model used by Feldstein to estimate the effect of social security on saving is an aggregate consumption function based on the life-cycle hypothesis. Consumer expenditures are assumed to depend upon permanent income and wealth. Wealth is divided into two components-household assets and estimated social security wealth. Social security wealth is constructed by assuming the growth path of benefits and discounting to the present, taking survival probabilities into account. In 1971, Feldstein's estimate of social security wealth was $2,029 billion, 60 percent of other household assets.

The Feldstein study estimates the relationship using aggregate U.S. data for the period 1929-71. The results may be approximated by the following equation:

The example abstracts from changes in the preferred life-cycle distribution of consumption and assumes that the worker still will seek to maintain retirement income of 50 percent of earnings.

Permanent income is an estimate of expected average lifetime income.

The equation actually estimated was C = 228 + 530 YD + 120 YD., +356 RE+014 HW+.021 SSW where YD is lagged disposable income and RE is corporate retained earnings. The latter is a proxy for the permanent component of capital gains. The method of estimation used was ordinary least squares. The regression coefficients pass standard statistical significance tests.

11

formation, and output. He estimates that aggregate personal saving has been halved by social security and that total private saving has been reduced by 38 percent. In the long run, the private capital stock would also decrease by 38 percent, reflecting the decline in the rate of private saving.

This reduction in private capital formation implies a substantial reduction in GNP. Feldstein estimates that GNP would be 15 percent higher in the absence of social security.

Feldstein's results must be validated.

Feldstein's work has raised important issues; however, his results must be viewed with caution. A few points deserve particular attention.

First, the values and statistical reliability of key regression coefficients of his estimated relationship are sensitive to both the specification of the model and the period of estimation. One must be cautious in basing estimates of the reduction in saving on regression coefficients whose value and reliability is suspect.

Second, the social security wealth variable may be a proxy for changes that have occurred during the period since the introduction of social security. Unemployment compensation, private pensions, health insurance, and other income security programs have been greatly expanded. There has been an increase in the ratio of the dependent population-young and old-to the working population. Expenditures for education have increased substantially. (Such expenditures are included in consumer expenditures; in the broad sense, however, they represent capital formation.) It is likely that these changes have influenced saving behavior.

Third, the construction of the key variable, social security wealth, depends upon a particular set of assumptions about how workers perceive the present value of future benefits. Alternative assumptions would yield different wealth series and may well lead to different conclusions.

Professor Feldstein has raised an important issue. His estimates of the impact of social security on saving may be correct, although the authors suspect that they are high. Further research is necessary to test the validity of his conclusions and to refine the estimates of the magnitude of the effect on saving.

12

Is there too little saving?

Suppose that we accept the hypothesis that social security reduces saving. Is this necessarily bad?

The rate of growth depends on society's choice. As noted earlier, a higher rate of capital formation is achieved at the sacrifice of present consumption. Thus, the desirability of growth turns on the desirability of sacrificing present consumption for the benefit of future generations.

Arguments can be made for such sacrifice. Since our consumption derives from the capital endowment of past generations, there is equity in making similar sacrifices on behalf of future generations. Some of us will live to enjoy this increased future consumption. On the other hand, since it is likely that succeeding generations will be richer than we are, is it necessary to make additional sacrifices on their behalf? The difficult question is how do we weigh the claims of the future against the claims of the present.

There is a related problem. If we decide that current consumption should be reduced to increase capital formation, the question arises: whose consumption should be reduced? In the context of the social security system, for example, the alternatives might be to reduce the consumption of workers or the consumption of beneficiaries.

The existence of such alternatives may give rise to conflicting goals. If workers are unwilling to sacrifice present consumption for the benefit of the future, then the burden must fall on the retired. If we are committed to supporting the aged, this solution is unsatisfactory. We may prefer to accept a lower growth path in order to maintain transfers to the aged. Thus, even if we accept the conclusion that capital formation and potential output have been reduced by social security, we may agree that this growth path is preferred.

Is there a shortage of capital?

Concern about the impact of social security on saving is associated with the belief that there is a shortage of capital. Several recent studies have projected a chronic shortage of investment funds over the next decade and concluded that a large increase in saving is required.

SOCIAL SECURITY