Distributional implications of the social security spouse benefit.

I. Introduction

The social security retirement program is often characterized as serving dual objectives of individual equity and social adequacy. These are conflicting objectives. Under a pure equity standard, retirement benefits would be distributionally neutral within a given age cohort of retirees. Social adequacy, on the other hand, relates benefits to the relative "need" of qualified beneficiaries and, correspondingly, is explicitly redistributive within an age cohort.

If one accepts the legitimacy of introducing considerations of need, then some adjustment based on household size is appropriate. Thus the spouse benefit has traditionally been justified as an important "adequacy" component of the retirement benefit formula. However, the method of computing the spouse benefit has been the subject of considerable controversy. Most of the critical scrutiny has focused on perceived horizontal inequity in the formula |2; 4~. This paper, in contrast, examines the spouse benefit from the perspective of vertical equity. A desire for progressivity clearly motivates intragenerational redistribution in the social security retirement program. The basic retirement benefit formula provides a higher implicit return on payroll tax payments to workers with lower career earnings than to workers with higher earnings. The issue to be examined in this paper is whether the spouse benefit in its current form complements or weakens a progressive relationship between household benefits and household earnings.

II. The Social Security Retirement Benefit Formula

An individual's retired worker benefit from the social security program, his or her Primary Insurance Amount (PIA), is based on earnings in covered employment prior to retirement. All of the individual's earnings which were subject to payroll taxation after 1950 or the individual's 21st birthday, whichever was latest, are indexed based on average wage growth. Five years of lowest earnings are excluded and Average Indexed Monthly Earnings (AIME) are computed for the remaining years. The individual's monthly PIA is then computed by applying a formula to the AIME. In 1991, an individual retiring at age 65 qualified for a retirement benefit equal to 90% of the first $370 of AIME plus 32% of the next $1860, plus 15% of any AIME in excess of $2230. The dollar brackets in the formula are adjusted annually using the Consumer Price Index. If the individual is married to a spouse of retirement age, that spouse receives either his or her own retired worker benefit based on own earnings in covered employment or a spouse benefit equal to one half of the PIA of the retired worker. Thus a couple is guaranteed a minimum benefit equal to 150 percent of the PIA of the highest earner.

If one regards the appropriate unit of analysis as the household rather than the individual, the spouse benefit in its current form can be criticized on horizontal equity grounds. Because of the either-or nature of the spouse benefit, it is possible for two couples with identical household earnings and payroll tax liabilities during their working lives to qualify for different levels of combined retirement benefits based on how covered earnings were divided between the spouses in each household. The formula clearly favors couples in which one spouse is responsible for the greatest share of earnings as opposed to couples in which earnings are more equally divided.

In addition to the problem of horizontal equity, the spouse benefit in its present form has the potential to alter the degree of progressivity in the basic retirement benefit formula. Because it is a supplemental benefit, it would increase overall progressivity if its recipients were to be disproportionately concentrated among households with relatively low lifetime earnings. If, on the other hand, upper income households were the primary beneficiaries, overall progressivity could be lessened and potentially eliminated. A third possibility would be for the benefit to be randomly distributed among retired households. In this last case, no serious vertical equity concerns would be present although the issue of horizontal equity would remain. These issues are the focus of this paper.

III. Methodology

Our approach to evaluating the vertical equity implications of the spouse benefit is to compare the distribution of benefits under the current formula with two alternatives, both of which eliminate the spouse benefit. The first alternative allows individuals to qualify for benefits solely on the basis of their own earnings histories. Household benefits derived from a straightforward summation of individual PIAs does not eliminate the problem of horizontal equity. Because of the progressive benefit formula, total benefits for a couple remain dependent on the distribution of earnings between the two spouses. The second alternative eliminates this problem by dividing the combined earnings of the couple equally between them for purposes of computing each spouse's individual benefit. Couples with the same combined earnings would then qualify for the same total benefits regardless of the distribution of those earnings. Under this earnings sharing approach, a progressive relationship between total earnings and total benefits would be guaranteed by the progressive benefit formula.

Using a sample of households from the Social Security Retirement History Survey (RHS) |3~, we compute the combined benefits received by both spouses under each of the three alternative benefit systems.(1) Households are then sorted into quintiles based on average real monthly earnings for the period between 1951 and the year the youngest spouse attained age 62.(2) Using benefits under the current system as a base, the percentage change in average benefits in each quintile is computed for each of the two alternative benefit formulas. If the percentage change is the same in all quintiles, the spouse benefit is neutral with respect to the average progressivity of retirement benefits. Differential percentages are evidence of nonneutrality.(3)

The method used to index earnings for the purpose of determining AIMEs is not neutral with respect to when earnings are received. Earnings are adjusted using a wage index rather than a price index. Because money wages tend to increase more rapidly than money prices, wage indexing means that a younger worker with the same real covered earnings prior to retirement as an older worker will nonetheless have a higher AIME. Because of this we did not use AIMEs to rank households in quintiles, but rather computed Average Real Monthly Earnings (ARME) using the GNP deflator. We also included all covered earnings after 1950 in our calculations. Thus households were ranked on the basis of actual real earnings.

It is, of course, fairly common for spouses to be different ages. In order to compute combined benefits for a couple after both had reached the assumed retirement age of 65, we computed the PIA of the oldest spouse at age 65 and then adjusted that benefit on an annual basis using the CPI until the year that the youngest spouse reached age 65 and qualified for a retirement benefit. This is the actual adjustment that would be made under the Social Security benefit formula. At that point in time the benefits of the two spouses were combined to generate the household retirement benefit. This combined benefit was then computed for all households in 1977 dollars using the GNP deflator. This adjustment was necessary because of differing ages across households.

A sensitivity of real benefits to the age of the beneficiary presents an obvious problem in trying to isolate the independent effect of the spouse benefit when the ages of beneficiaries in the sample population vary. However, a comparison of the results using our sample with results for a sample in which the age of the head was the same for all households supports an assumption that wage indexing did not make a significant difference. Its effect appears to become significant over a much longer term than the time period involved in our sample.

A final point can be made about the relationship between ARME and permanent income. The availability of earnings histories over many years allows the ARME to be a much better proxy for permanent income than a measure of annual earnings during a given year. However, because it includes only market compensation, the ARME underestimates full household income by the value of household production and leisure. This suggests the possibility that any ranking of households on the basis of full income might be significantly different from a ranking based on money income. Two households with identical ARMEs could have very different levels of real income, for example, if both spouses were employed full time in one household, while one spouse devoted full time to household production in the other. However, we have not tried to adjust our household rankings for the following reasons. First, imputing a shadow value for household production and leisure would be extremely difficult. It is not done in the vast majority of distributional studies and ours is no different in that regard. More importantly, we do not believe that a failure to rank households on the basis of full income is serious for the problem at hand. The reason is that, when full income is considered, the spouse benefit is a regressive element within a money income cohort. A household in which one spouse is a full-time homemaker surely has greater full income TABULAR DATA OMITTED than another household with identical money income in which both spouses are employed full-time in the labor force. Yet the former household would receive the supplemental spouse benefit. This reasoning can be extended to support our original proposition that if recipients of the spouse benefit were to be disproportionately concentrated among households with relatively high lifetime earnings, it would be a regressive feature of the benefit structure.

IV. Empirical Evaluation of the Spouse Benefit

The results of our paper are reported in Table I. The households are arrayed in quintiles based on ARMEs. As the table indicates, the top 20 percent of households, Quintile 5, accounted for over 34 percent of total ARME in the sample. The lowest quintile, in contrast, only accounted for about 5 percent of total ARME.

Columns 3-5 in the table report comparative results of the three alternative approaches to determining household benefits. Column 3 reports a comparison of the current system with its separate spouse benefit with the alternative of simply summing the retired worker benefits of the individual spouses. The spouse benefit consistently provides higher benefits in all quintiles than would be provided if only retired worker benefits were granted. However, with the exception of the top quintile, the percentage change in benefits is not significantly different across quintiles. This implies that the spouse benefit is fairly neutral with respect to progressivity of the Social Security benefit for the first 4 quintiles but increases the progressivity of the benefit formula because of the relatively small amount received by the highest income quintile.

Column 4 compares benefits received under earnings sharing with those provided by summing individual retired worker benefits. Because of the progressive benefit formula, earning sharing should provide an increase or no change in benefits depending on whether or not the two spouses were in different marginal benefit brackets in benefits in all quintiles. However, if the spouses are of different ages, the switch to earnings sharing could involve the elimination of different years of lowest earnings.(4) This presumably explains the slight decline in benefits in the top quintile. As was the case with our previous comparison, the switch to earnings sharing slightly increases the relative share of benefits going to the bottom quintile and significantly decreases the share claimed by the top quintile. There is very little change in relative shares across the middle three quintiles. Comparing the top and bottom quintiles, however, a switch to earnings sharing from the simple summation of the husband's and wife's retirement benefits would add to overall progressivity.

The final column compares household benefits under the current system with earning sharing. There is a consistent decline in benefits across all quintiles when the earning sharing formula replaces the current system. As in our previous comparisons, the percentage changes in the bottom four quintiles are roughly equal. Benefits in the top quintile fall by a much smaller percentage. Our interpretation of this result is that the top quintile of households would benefit from a shift to earning sharing at the expense of the lower four quintiles if the reform were instituted in such a way as to hold total benefit outlays constant. Our own calculations allowed total outlays to fall and the percentage decline in benefits in the top quintile was considerably smaller than the decline in the bottom four quintiles. This implies an increase in the share of total benefits received by the top quintile. If all benefits were to be increased equiproportionately in order to maintain constant total outlays, households in the top quintile would gain benefits at the expense of the bottom four.

V. Conclusions

Our results indicate that the spouse benefit makes a small contribution to progressivity. It does not appear to be a significant factor in determining total benefits for households in the top quintile, undoubtedly because this quintile is dominated by couples in which both spouses have significant earnings histories. The spouse benefit is an important component of benefits in the bottom four quintiles, but does not appear to have a strongly disproportionate effect across those quintiles.

Ironically, the issue of vertical equity would be a factor in either of the proposed reforms because both would tend to increase the share of total benefits going to upper income households. In this context, reform of the spouse benefit would appear to go against the grain of much of the current general discussion of Social Security reform which is focused on reducing benefits to upper income households, primarily via changes in the taxable portion of benefits.

1. Households in the RHS did not actually retire under the current benefit structure which took effect in 1977. However, if the within household earnings structure has not changed significantly, the hypothetical distributional implications drawn from our sample should shed some light on the current implications of reforming the treatment of spouses.

2. An individual's retirement benefit at age 65 is based on earnings prior to age 62. An individual who retires before reaching age 65 receives a reduced benefit.

3. Total benefit outlays will differ under the alternative formulas. If an alternative to the current system were actually adopted, some adjustment would probably be made in the basic benefit formula to hold total outlays constant. We do not make this adjustment in our calculations. However, we assume that any such actual adjustment would not alter the relative benefits implicit in the new formula. Our approach will still isolate any change in shares resulting solely from the new benefit structure.

4. Benefits based on an earning sharing formula are not computed until both spouses have retired. When the oldest spouse retires, his or her PIA is computed in the standard way. When the youngest spouse retires, annual covered earnings for both are combined for each year after 1950 and divided by two to create individual earning records. The earnings are then indexed and individual AIME's are computed after elimination of the five years of lowest earnings. Each spouse then has an identical AIME and, correspondingly, an identical PIA. Total household benefits is the sum of PIAs. This method is based on the Modified Earnings Sharing Plan as contained in Earnings Sharing Options for the Social Security System |1~.

References

1. Earnings Sharing Options for the Social Security System. Congress of the United States, Congressional Budget Office, January 1986.

2. Flowers, Marilyn R. Women and Social Security: An Institutional Dilemma, Washington D.C.: American Enterprise Institute for Public Policy Research, 1977.

3. Longitudinal Retirement History Survey. Washington D.C.: Social Security Administration, 1969-1979.

4. "Report of the Subcommittee on the Treatment of Men and Women under Social Security," Reports of the (1975) Quadrennial Advisory Council on Social Security. Washington D.C., 1975.