The aging of society and the impending retirement of the

Transcription

1 Social Security Reform for the Elderly Distributional Implications of Social Security Reform for the Elderly: The Impact of Revising COLAs, the Normal Retirement Age, and the Taxation of Benefits Abstract - Several reform plans have been proposed to reduce the financial strain on Social Security. Many proposals under consideration, however, would exacerbate the inequality of income among the elderly. Using data from the Current Population Survey, this paper finds that reducing cost-of-living escalators, even when the reforms are designed to protect low-income beneficiaries, would generally worsen inequality over time and increase the number of elders with income below the poverty level. Raising the retirement age would generate similar results. However, increasing taxes on retirement benefits would improve the financial outlook for Social Security without imposing additional hardships on the poorest elderly. Richard W. Johnson The Urban Institute, Washington, D.C National Tax Journal Vol. LII, No. 3 INTRODUCTION The aging of society and the impending retirement of the post World War II birth cohorts are straining the Social Security system. Social Security actuaries estimate that outlays for Old-Age, Survivors, and Disability Insurance (OASDI) benefits will consume 6.8 percent of gross domestic product (GDP) in 2030, compared with only 3.2 percent of GDP in 1970 (U.S. Board of Trustees of the Federal Old-Age and Survivors Insurance and Disability Insurance Trust Funds, 1999). The rise in the share of social resources devoted to the elderly has led to a number of calls to reform Social Security and reduce the burden to the working population of providing financial support for retirees. Over the past few years, reform plans have included provisions that would reduce annual retirement benefits, delay the age at which retirees could begin to collect benefits, increase the taxation of benefits, reduce the annual cost-of-living escalators that are applied to Social Security income, and bring more workers into the system by covering all state and local government employees. Other plans would attempt to raise the return workers receive on their Social Security taxes by creating individual retirement accounts, thus transforming the very structure of the Social Security system. 1 1 President Clinton also proposed in his 1999 State of the Union address to invest part of the OASDI trust fund in private equities. 505

2 NATIONAL TAX JOURNAL Each of these proposed reforms to Social Security would have important distributional consequences for beneficiaries. From its earliest days, the system s redistributive properties have received much attention. Social Security was designed not only to transfer income from relatively young workers to relatively old retirees, but also to redistribute income from the rich to the poor. Social Security, created during the Great Depression, was originally intended to provide the elderly with a minimum level of retirement income, in order to prevent individuals from falling into poverty when they retired (Steuerle and Bakija, 1994). Under the current benefit structure, Social Security benefits replace a much larger fraction of earnings for those near the bottom of the income distribution than for those near the top of the distribution. Although Social Security has been successful in reducing elderly poverty rates, poverty and inequality persist among the aged. In 1997, 10.5 percent of persons age 65 and older were impoverished, according to official government statistics (U.S. Bureau of the Census, 1998). Moreover, many experts contend that the distribution of income is more unequal at older years than at younger ages (Crystal and Shea, 1990; Crystal and Waehrer, 1996; Disney, 1996; Smeeding, 1989). 2 Consequently, serious consideration should be given to the distributional implications of Social Security reform, particularly its impact on the poorest elderly. A number of recent studies have examined the progressivity of Social Security benefits and how different groups have fared under the current system. Most of these studies have estimated the rate of return that workers in different socioeconomic groups earn on their Social Security taxes. For example, Coronado, Fullerton, and Glass (1999) used longitudinal survey data from the Panel Study of Income Dynamics to construct earnings histories, lifetime tax payments, and future Social Security benefits for a sample of households. They found that the present value of benefits received net of taxes paid was higher for individuals with low lifetime income, but that the system was only mildly progressive after they accounted for income differentials in mortality. Although replacement rates are higher for low-income workers, they generally collect benefits for fewer years than highincome workers, because mortality rates decline with income. Panis and Lillard (1996) estimated rates of return for low-, medium-, and high-income individuals and found that estimates of intracohort transfers from high-income to low-income individuals fell when they accounted for differences in mortality rates by marital status and household income. Steuerle and Bakija (1994) computed lifetime Old- Age and Survivors Insurance (OASI) taxes and benefits for persons with stylized earnings histories. They found that rates of returns were substantially lower for men and women with high earnings than for those with low earnings, although returns have been dropping over time for all types of earners. Caldwell et al. (1998) employed microsimulation techniques to compute lifetime earnings histories and tax payments for a large sample of individuals. They found that persons born after World War II earned small and often negative rates of return on their Social Security taxes, and that the system treated college graduates better than nongraduates. This paper takes a somewhat different approach to investigating the distributional consequences of some Social Security reform proposals. We examined how the current level and distribution of income for a sample of elderly persons would change if some reform proposals currently under consideration had been enacted in the past. We studied the effects 2 For a dissenting view about the relative inequality of income over the life course, see Fuchs (1998). 506

3 Social Security Reform for the Elderly of reductions in cost-of-living adjustments (COLAs), increases in the normal retirement age, and changes in the taxation of benefits. Because we examined income for a sample of elderly persons, we did not have to make assumptions about rates of mortality or their relationship to income. As other studies have done, we did assume that individuals did not alter their behavior in response to changes in Social Security. We found that reductions to the COLA formula, even when designed to protect low-income beneficiaries, often had regressive effects, and over time would increase the number of elderly persons with income below the poverty level. Raising the normal retirement age to 70 would have similar effects. However, increasing taxes on retirement benefits would improve the financial outlook for Social Security without imposing additional hardships on the poorest elderly. PROPOSED SOCIAL SECURITY REFORMS Reductions in COLAs 507 To maintain the real value of retirement benefits after workers retire from the labor force, Social Security payments are tied to changes in the Consumer Price Index (CPI). Under current law, once retirees begin receiving benefits, payments increase each year by the percentage change in the CPI. COLAs have become increasingly important as life expectancy rises and retirees spend more years collecting benefits. The typical woman who reaches age 62 in 2000, for example, can expect to live another 22 years (Bell, Wade, and Goss, 1992). If the inflation rate were 3 percent per year, the real value of Social Security benefits would decline by 48 percent during her expected 22 years in retirement if benefits were not indexed and Congress did not intervene to maintain the real value of her retirement income. Because other sources of retirement income, such as private pensions and savings, generally do not automatically adjust in response to changes in the cost of living, Social Security provides one of the few means of inflation protection available to the elderly. Some recent studies suggest, however, that the CPI overstates the true rate of inflation. The CPI estimates changes in the price level by computing the cost to consumers of purchasing a fixed basket of goods and services at different points in time. However, this basket becomes less representative over time as consumer spending patterns change in response to price changes and new choices. For example, as the prices of particular items in the basket rise, consumers may replace these products with less expensive substitutes. The ability to substitute for relatively expensive goods means that the true cost increase experienced by consumers is generally less than the change in the CPI. Similarly, consumers may respond to higher prices by finding less expensive outlets where they can shop. Improvements in the quality of products, which can enhance durability and reduce the cost of repairs, are not well measured, further biasing upward estimates of the true cost of living. In addition, new products are added to the market basket with long lags, if at all, so that the CPI basket may not always accurately reflect the types of products consumers purchase. The Advisory Commission to Study the Consumer Price Index (1996), known as the Boskin Commission, concluded in 1996 that these factors lead the CPI to overstate the true increase in the cost of living by 1.1 percentage points per year. Partly in response to the Commission s recommendations, the Bureau of Labor Statistics has instituted a number of changes in the past few years in the way it computes the CPI. The U.S. Congressional Budget Office (1999) estimates that these changes will reduce the annual growth in the CPI by about 0.7 percentage points per year.

4 NATIONAL TAX JOURNAL Because the CPI has been overstating changes in the cost of living, a number of Social Security reform plans incorporate lower COLAs, either by assuming CPI reform or by proposing changes in the COLA formula. 3 Senator Daniel Patrick Moynihan (D-NY), for example, introduced a bill (S. 21) in January 1999 that would reduce COLAs to one percentage point below the annual percentage change in the CPI, and Representative Mark Sanford (R-SC) introduced a bill (H.R. 250) that same month to reduce COLAs to onehalf percentage point below the annual percentage change in the CPI. Federal Reserve Board Chairman Alan Greenspan (1999) recently advocated that reductions to COLAs be considered as part of Social Security reform. The Advisory Council on Social Security (1997) recommended that changes be made to the CPI to remove bias from the cost of living estimates, although they did not support changes to the COLA formula. Aaron (1998) and Reischauer (1998) of The Brookings Institution have also put forward a Social Security reform plan that assumes that the CPI would be corrected. Although reduced COLAs would ease the social burden of retirement benefits, they would also have significant effects on the level of income received by the elderly. 4 Moreover, reductions in COLAs would not affect all elderly equally. Reducing Social Security COLAs may exacerbate inequality at older ages, because poor elderly rely upon Social Security benefits for almost all of their income, whereas higher-income elderly have other sources of income. Figure 1 reports the percentage of income received by the elderly from different sources, by quintile of total income. For the elderly in the bottom two income quintiles, Social Security retirement benefits accounted for 80 percent of their income, compared to only 19 percent for elderly in the top income quintile. Thus, reducing Social Security COLAs would affect almost all of the income received by the poor elderly but only a relatively small portion of the income received by the affluent. In addition, the proposed reductions would have larger effects for the oldest old than for relatively young beneficiaries. Changes to Social Security s COLA formulas would have no effect on the initial benefits received by retirees, because they are based on past average indexed monthly earnings (AIME), which are adjusted for changes over time in the median wage received by covered workers. Once workers begin collecting benefits, however, small annual reductions in payments that have only small effects on income in the early years can compound into large declines after many years. For example, a 1 percent annual reduction in benefits compounded for 23 years translates into a 26 percent reduction for 85 year olds who began collecting benefits at age 62, whereas 65 year olds who were collecting benefits for three years would experience only a 3 percent reduction in benefits compared to what they would have otherwise received. Because across-the-board reductions in Social Security COLAs would disproportionately affect the poor, some policymakers have proposed modified COLA cuts designed to maintain the full value of benefits for the low-income elderly. For example, Representative Sanford (R-SC) has introduced another bill in Congress in 1999 (H.R. 249) that would set the maximum annual increase in the COLA equal to the increment in benefits received by those at the 30th percentile of 3 The CPI does not necessarily overstate inflation for the elderly, however. Elderly persons devote a much larger portion of their income to health care than younger persons, and health care costs have been rising more rapidly than the overall increase in the CPI. 4 The Advisory Council on Social Security estimated that a one percent reduction in COLAs would lower the long-term deficit in the OASDI trust funds by more than one-half. 508

5 Social Security Reform for the Elderly Figure 1. Sources of Income for the Elderly,

6 the income distribution. For beneficiaries with Social Security income below the 30th percentile, the COLA would remain equal to the full percentage change in the CPI. However, because some elderly who receive virtually all of their income in the form of Social Security benefits may have limited income but receive Social Security benefits above the 30th percentile, it is not clear how these types of progressive COLA cuts would affect the distribution of income in later life. Increases in the Normal Retirement Age 510 NATIONAL TAX JOURNAL Proposals to increase the retirement age are generally based on the belief that most workers are now able to work until older ages than they could in the past because of improvements in health and declines in the physical demands of work. In 1940, when the Social Security system was just beginning, men who reached age 65 lived an additional 12.7 years on average. By contrast, Social Security actuaries now estimate that males born in 1985 can expect to live another 18 years after they reach age 65 (in 2050) (Bell, Wade, and Goss, 1992). Increases in longevity do not necessarily imply that persons will be able to work until older ages, because they may spend the additional years in poor health and disability. However, recent evidence suggests that health has also been improving over the past 20 years. For example, Manton, Corder, and Stallard (1997) found that disability rates among persons ages 65 and older declined during the 1980s and early 1990s. Workers may also be increasingly able to delay their withdrawal from the labor force because the physical demands of jobs continue to decline over time. For example, the proportion of workers in jobs that require them frequently to lift or carry objects weighing more than 25 pounds fell from 20.3 percent in 1950 to 7.5 percent in 1996 (Johnson, Spiro, and Steuerle, 1999). As technological change increases the prevalence of relatively sedentary jobs in the economy, fewer older workers will be forced to retire from the labor force because they are physically unable to perform their employment duties. The combination of improvements in health and reductions in the physical demands of work has led a number of policymakers to conclude that the state no longer needs to offer full retirement benefits at age 65. Under legislation passed in 1983, the Social Security normal retirement age is already scheduled to increase gradually, beginning in the year Workers who reach age 62 in 2022 would not be eligible for full retirement benefits, or 100 percent of their primary insurance amount (PIA), until they reached age 67. Workers would continue to be eligible for reduced benefits at age 62. However, the penalty for the receipt of early benefits at age 62 would increase from 20 percent of PIA for workers who begin collecting benefits in 1999 to 30 percent for workers who begin collecting in A number of Social Security reform proposals currently under consideration would increase the retirement age even further. For example, the reform plan put forth by the National Commission on Retirement Policy (1999) would increase the normal retirement age by two months per year, until it reached age 70 at the end of Henceforth, the retirement age would rise with increases in life expectancy, so that the number of expected years in retirement would remain constant over time. Under the Commission s plan, the early retirement age would also gradually increase, until it reached age 65 in Two of the three different reform plans proposed by the Advisory Council on Social Security (1997) the Individual Accounts plan and the Personal Savings Account plan would also raise the normal retirement age to 70, although the increase would be more gradual, so that the normal retirement age would not hit age 70 until the year Plans pro-

7 Social Security Reform for the Elderly posed by Aaron (1998) and Reischauer (1998) and by Representative Sanford (H.R. 250) would also raise the retirement age to 70. Changes in the Taxation of Benefits For most of its 60-year history, all Social Security benefits were exempt from federal income taxation. It was not until the 1983 amendments went into effect that any portion of Social Security retirement benefits were subject to the federal personal income tax, and even then no more than 50 percent of benefits were taxable for income tax filers whose income exceeded certain thresholds. Many observers argue that the taxation of Social Security benefits can be justified on the basis of horizontal equity, which requires that persons receiving the same level of income should pay the same level of taxes. If benefits were not taxed, an elderly person with Social Security income would pay lower federal income taxes than a younger person who received the same level of income but who received it in the form of wages. Moreover, an elderly person who receives most of her income in the form of Social Security benefits would pay lower taxes than one who receives most of her income in the form of private pensions, even when both individuals receive the same amount of income, if Social Security benefits were not taxed. Except for previously taxed employee contributions, private pension income is taxable when it is received. Consequently, although federal income taxes were paid on OASDI contributions when the beneficiary (or spouse) was working, horizontal equity considerations suggest that the amount of Social Security benefits in excess of employee contributions should also be taxed when they are received. Social Security Administration actuaries estimate that past contributions account for only about 15 percent of the benefits received by current beneficiaries (Pattison and Harrington, 1993). Under current law enacted in 1993, up to 85 percent of Social Security benefits are taxable but only for beneficiaries with relatively high income. If modified adjusted gross income (MAGI), defined as the sum of adjusted gross income (AGI), nontaxable interest income, and one-half of Social Security benefits, exceeds certain thresholds, then a portion of Social Security benefits are taxed. For every dollar that MAGI exceeds $25,000 for single filers and $32,000 for joint filers, 50 cents of OASDI benefits are taxable. For every dollar that MAGI exceeds a second threshold, set at $34,000 for single filers and $44,000 for joint filers, 85 cents of OASDI benefits are taxed. A maximum of 85 percent of benefits are taxable. The income thresholds are not indexed, so that over time the proportion of benefits subject to the federal income tax has been increasing. Many Social Security reform proposals would increase the taxation of OASDI benefits. Senator Moynihan introduced a bill into Congress (S. 21) that would include all Social Security benefits in AGI, thus eliminating the income threshold and subjecting 100 percent of benefits to the income tax. Other plans, such as the Maintenance of Benefits plan and the Individual Accounts plan proposed by the Social Security Advisory Council and the proposal advanced by Aaron (1998) and Reischauer (1998), would eliminate the income thresholds and include 85 percent of Social Security benefits in AGI. 5 One 5 The current rule taxing up to 85 percent of Social Security benefits is designed to approximate the tax treatment of contributory private pensions. For recipients of private pension income, pension benefits in excess of past employee contributions are taxable. For the highest paid workers, past contributions account for only about 15 percent of pension income, so that 85 percent of pension income is generally taxable. For many current Social Security beneficiaries, however, past contributions account for less than 15 percent of Social Security benefits. As a result, proposals that would tax 85 percent of Social Security benefits would treat Social Security income more generously than income from contributory private pensions for most taxpayers. 511

8 NATIONAL TAX JOURNAL attraction of the increased taxation of benefits for Social Security reform is that it is a relatively progressive way to reduce net retirement benefits. Eliminating the income thresholds or increasing the proportion of benefits subject to the tax would not reduce the after-tax income of poor elderly whose income falls below the standard deduction and personal exemption, because they do not pay any federal income taxes. Even among those who pay taxes, the effective benefit cut that would result from increasing the proportion of benefits that are taxable would be greatest for high-income elderly because of the progressivity of the marginal income tax rates. Eliminating income thresholds, however, would reduce after-tax benefits primarily for middle-income beneficiaries, who would become subject to the tax on benefits only if the thresholds were lowered. At the same time that increases in taxation are being proposed in the name of Social Security reform, other bills introduced in Congress in the first few months of 1999 would reduce the taxation of Social Security benefits in the name of tax relief. These bills would repeal the 1993 legislation and set the maximum amount of benefits that are taxable at 50 percent (H.R. 48, H.R. 107, H.R. 688, S. 137, and S. 286), eliminate nontaxable interest income from MAGI (H.R. 291), or eliminate the taxation of Social Security benefits altogether (H.R. 761, S. 488). METHODS To examine the distributional effects of a number of Social Security reform proposals, we analyzed data from the March 1998 Current Population Survey (CPS). The CPS is an ongoing monthly survey of about 50,000 households, which includes information each March about the type and amount of income received in the previous year. Our unit of analysis was the aged unit, defined here as unmarried persons aged 65 or older and married couples in which at least one spouse was aged 65 or older. Our sample included 11,879 aged units. Because of age topcoding in the CPS, we set age equal to 90 for respondents who were age 90 or older. When classifying units by age when both spouses were aged 65 or older, the age of the husband was used. We measured income as the per-capita income of the aged unit, consisting of all money income received by the unit, before any deductions such as taxes, union dues, or Medicare premiums. Nonmoney transfers, such as food stamps and health benefits, and any income from other persons who may have lived with members of the aged unit, such as their adult children, were excluded. 6 On the one hand, by omitting in-kind benefits received by the poor elderly, such as food stamps and Medicaid benefits, comparisons of money income between the rich and the poor tend to overstate inequities in economic well-being. On the other hand, the failure of simple comparisons of money income to account for differences in health care costs tends to understate inequities in well-being, because the portion of income spent on medical expenses is almost four times as high for the elderly in the bottom income quintile as for those in the top income quintile, despite the availability of Medicaid (Crystal et al., 1998). Although our definition of the aged unit is consistent with that used by the Social Security Administration, a different approach is taken by the Bureau of the Census, which defines units by the age of the head of the household. Under the Census definition, elderly persons living with younger householders, such as their adult children, would be excluded from the 6 Although there is evidence that the CPS underreports some types of income, particularly interest and dividend income (U.S. Bureau of the Census, 1993), we have not attempted to correct for income underreporting here. 512

9 Social Security Reform for the Elderly analysis. In fact, there are about 13 percent fewer elderly households in the country based on the Census definition than aged units as defined here (Grad, 1998). By focusing on the aged unit, we are not necessarily measuring the average standard of living that the elderly would experience under different reform options, because many aged units are actually components of larger households that include adult children. Because adult children usually share resources with their coresident parents, the standard of living experienced by elderly persons who reside with their children will generally exceed their own income. However, our focus on aged units does enable us to examine changes in the financial capacity of the elderly to support themselves and indicates whether they would need financial support from their families to be selfsufficient. We examined the effects of reducing Social Security COLAs, increasing the normal retirement age, and revising the tax treatment of Social Security benefits by assuming that each policy change had been implemented at a specified point in the past but everything else remained the same. We then compared the simulated distribution of income after the policy change was introduced with the actual distribution observed in For each reform proposal, we estimated the mean level of per-capita Social Security retirement income and the mean level of total per-capita income for the aged unit. We compared the distribution of income for aged units after the policy change was introduced with the actual distribution of 1997 income by computing Gini coefficients and by estimating the mean level of income within each decile of the distribution. We also computed the percentage of persons with per-capita aged unit income below the poverty level before and after the simulated reductions in the COLA and increases in the retirement age. 7 Throughout our analysis, we assumed that these policy changes did not generate any behavioral responses, such as increased saving or labor supply. By ruling out these types of responses, which would tend to offset declines in retirement income, our estimates can be viewed as upper bounds on the effects of proposed policy changes on elderly income. We considered the effects of several different reforms to the COLA formula. First, we assumed that the COLA formula used to adjust Social Security retirement benefits had been revised in the past to the percentage change in the CPI minus one percentage point. Then, following some legislative proposals, we set the COLA formula equal to the full annual percentage change in the CPI, as under current law, but capped the annual dollar adjustment at the amount equal to that received by persons with Social Security retirement income at the 30th percentile. However, because some commentators have noted that some elderly persons with Social Security benefits near the 30th percentile have low incomes and might be seriously harmed by reductions in their COLAs (Moon and Mulvey, 1996), we also simulated income when the COLA was capped at higher levels, equal to the dollar adjustment received by persons at the 50th and 60th percentiles of the distribution of Social Security retirement benefits. We also examined the effects of basing the COLA caps on total income, instead of Social Security benefits, because some persons with relatively high retirement benefits may have low income if all of their income comes from Social Security. For this final scenario, we set the COLA cap equal to the dollar adjustment received by elderly persons at the 50th percentile of the percapita income distribution for aged units. 7 Our calculations of the percentage of persons with income below the poverty level differs from the official poverty rates published by the Census Bureau, which defines elderly households by the age of the householder and does not use the concept of the aged unit. 513

10 NATIONAL TAX JOURNAL For each COLA reform scenario, we assumed that the CPI itself was not altered and COLAs for other government programs did not change. In our simulations, reductions to Social Security benefits from the change in COLAs began the year the revision was assumed to be implemented or at age 62 the age at which Social Security benefits are calculated and tied to changes in the CPI whichever came later. We considered COLA revisions that had been implemented 1 year prior, 10 years prior, and 30 years prior to 1997, but we based our distributional analysis on the simulated incomes that would prevail had the new COLAs been in effect for 30 years. After 30 years, all the Social Security retirement benefits ever received by the members of our sample would have been subject to the new COLA formulas. For the change in the retirement age, we considered the effect on elderly income of having increased the normal retirement age to 67 and to 70 at some point in the past. Because we assumed that retirement behavior does not respond to changes in Social Security policy, delays in the retirement age would reduce benefits but not affect their timing. To compute the reduction associated with the increase in the retirement age, we assumed that all respondents in our sample began collecting Social Security benefits at age 62, the age at which most current beneficiaries did begin collecting. 8 Thus, under current rules, we assumed that the level of benefits observed in our sample in 1997 was based on 80 percent of respondents Primary Insurance Amount (PIA) or the spouses PIA if they are collecting spousal benefits because benefits are permanently reduced by 5/9 of 1 percent below full PIA for each month that they were first received before age 65. When simulating an increase in the retirement age to 67, we followed the rules enacted with the 1983 amendments, which will in fact raise the normal retirement age to 67. Benefits for those who begin collecting at age 62 will be based on 70 percent of PIA once the normal retirement age reaches 67. Under the new rules, benefits first received between ages 64 and 66 will be reduced by 5/9 of 1 percent below full PIA for each month that they were received before 67, and benefits first received at age 62 or 63 will be reduced by 20 percent plus an additional 5/12 of 1 percent for each month they were received before age 64. Thus, when the normal retirement age reaches 67, benefits for those who first begin collecting at age 62 will be only 70 percent of PIA. To simulate an increase in the retirement age to 70, we assumed that the same rules would apply. In particular, we assumed that benefits for persons who first began to collect between ages 64 and 69 would be reduced by 5/9 of 1 percent below full PIA for each month that they were first received before age 70. For persons retiring at age 62, then, we assumed that benefits would equal 50 percent of PIA when the normal retirement age reaches 70. Relative to benefits actually received in 1997, our simulations assumed that Social Security benefits would be 10 percent lower when the normal retirement age was 67 and 30 percent lower when the normal retirement age was 70. We also assumed that the increases were implemented gradually, at the rate of two months every year. At this rate, the 5-year rise in the normal retirement age would take 25 years, and every member of our sample would not have been subject to the age 70 retirement structure until the transition had been underway for 53 years. In our analyses, we focused on the distribution of income that would prevail had the increase in retirement ages begun 55 years ago, so that the higher age was in effect when every member of our sample reached Information on the age at which respondents actually began to collect Social Security benefits is not available in the CPS. In 1997, 55 percent of new beneficiaries were age 62, while only 15 percent were age 65 (U.S. Social Security Administration, 1998). 514

11 Social Security Reform for the Elderly If COLA formulas or retirement benefits were revised, the erosion in income would have led some persons to become eligible for Supplemental Security Income (SSI), offsetting at least part of the loss in Social Security benefits. Elderly persons were eligible for SSI in 1997 if their countable monthly income was below $484 for unmarried individuals or below $726 for couples, and if they had limited assets. Countable monthly income excludes the first $20 of income, $65 in earnings, and one-half of any earnings above $65. However, only about one-half of the elderly who are currently eligible for SSI actually receive benefits, possibly because of the stigma associated with receiving need-based transfer payments (McGarry, 1996). We assumed that all persons who became eligible for SSI received benefits, except for those who were already eligible for benefits in 1997 before the simulated change in Social Security benefits but did not receive SSI. By assuming that all those newly eligible for SSI receive benefits, our results may overstate the extent to which SSI offsets the fall in Social Security for the very poor, and thus may understate some of the distributional effects of reducing Social Security benefits. Finally, we examined four different reforms that would alter the taxation of Social Security income. We considered the effects of fully taxing all Social Security retirement income; subjecting 85 percent of Social Security benefits to taxation for persons of all income levels; repealing the 1993 tax increase, so that at most 50 percent of benefits were included in the tax base for couples with MAGI above $32,000 and for singles with MAGI above $25,000; and eliminating the taxation on Social Security income altogether. For comparison purposes, we also computed after-tax income for the elderly under the current system. A number of assumptions were incorporated into our simulations of tax liabilities. We assumed that itemized deductions equaled 20.5 percent of AGI, the 515 mean level of deductions taken by taxpayers ages 65 and older in 1993, according to Statistics of Income data. We then assumed that tax filers took the larger of the standard deduction or the itemized deductions. We also assumed that all pension and interest income was fully taxable, that all married couples filed jointly, and that no one in our sample claimed any dependents. RESULTS Simulated Impact of Social Security Reform on the Level of Elderly Income Table 1 reports the mean level of Social Security retirement income received by the elderly under the different COLA reform scenarios. Although COLA reform would have only limited impact on the level of Social Security benefits received by the elderly in the years immediately after the changes were introduced, 10 years and especially 30 years later the effects on benefits would be pronounced. As reported in Table 1, after the reforms had been in place for 30 years, so that all of the OASI benefits received by the elderly in our sample had been subject to reduced COLA formulas, the mean level of benefits would have been 12 percent smaller than its actual value in 1997 if the COLA equaled the annual percentage change in the CPI minus one percentage point. The average reduction in benefits would have been even greater if the COLA had instead been capped at the dollar amount received by beneficiaries at the 30th percentile of the benefit distribution. The reduction would have been smaller, however, if the COLA had been capped at higher levels. In each of the scenarios, the decline in mean benefits after 30 years was greater for the oldest old than for the relatively young elderly, because the reduced COLAs, which generally begin at age 62, were compounded for only a few years for the relatively

12 NATIONAL TAX JOURNAL TABLE 1 CHANGES IN PER-CAPITA SOCIAL SECURITY RETIREMENT INCOME AFTER COLA REFORMS Actual 1997 Mean Income COLA = change in CPI minus one percentage point 7,478 6,986 8,155 8,063 Income if COLA Reduction Began 1 Year Ago Mean Amount 7,405 6,918 8,075 7,984 Percent Change 1.0% 1.0% 1.0% 1.0% Income if COLA Reduction Began 10 Years Ago Mean Amount 6,886 6,513 7,404 7,317 Percent Change 7.9% 6.8% 9.2% 9.3% Income if COLA Reduction Began 30 Years Ago Mean Amount 6,608 6,474 6,917 6,327 Percent Change 11.6% 7.3% 15.2% 21.5% COLA capped at $ amount received at 30th percentile of benefits distribution COLA capped at $ amount received at 50th percentile of benefits distribution COLA capped at $ amount received at 60th percentile of benefits distribution COLA capped at $ amount received at 50th percentile of income distribution 7,478 6,986 8,155 8,063 7,478 6,986 8,155 8,063 7,478 6,986 8,155 8,063 7,478 6,986 8,155 8,063 7,419 6,932 8,088 7,997 7,441 6,953 8,113 8,023 7,451 6,962 8,124 8,034 7,450 6,960 8,123 8, % 0.8% 0.8% 0.8% 0.5% 0.5% 0.5% 0.5% 0.4% 0.3% 0.4% 0.4% 0.4% 0.4% 0.4% 0.4% 6,755 6,440 7,191 7,119 7,017 6,640 7,534 7,472 7,134 6,728 7,687 7,630 7,132 6,729 7,681 7, % 7.8% 11.8% 11.7% 6.2% 5.0% 7.6% 7.3% 4.6% 3.7% 5.7% 5.4% 4.6% 3.7% 5.8% 5.3% 6,277 6,386 6,363 5,324 6,678 6,601 6,951 6,177 6,859 6,698 7,219 6,568 6,901 6,705 7,265 6, % 8.6% 22.0% 34.0% 10.7% 5.5% 14.8% 23.4% 8.3% 4.1% 11.5% 18.5% 7.7% 4.0% 10.9% 15.7% young elderly. The one percentage point reduction in the COLA formula, for example, would reduce mean benefits for those ages 85 and older by 22 percent, compared with only 7 percent for those between the ages of 65 and 74. The predicted impact of COLA reform on the per-capita income of aged units was relatively small on average, except for those ages 85 and over. As reported in Table 2, lowering COLAs to one percentage point below the percentage change in the CPI would reduce total per-capita income by 4.5 percent after 30 years, and capping COLAs would reduce total income somewhere between 3 percent and 6 percent, depending on how low the cap was set. For those ages 85 and older, however, income would fall by 10.5 percent after 30 years if the COLAs were reduced by one percentage point, and by 17 percent after 30 years if annual adjustments were capped at the dollar amount received by those at the 30th percentile of the benefit distribution. Table 3 reports the mean level of Social Security retirement income and total percapita income of aged units if the normal retirement age were increased. Because the higher ages would be phased in gradually, the effects on income would be small in the first several years after the transition began. For example, after 20 years, mean Social Security benefits would be six percent lower than the actual 1997 level if the retirement age were raised to 67 and nine percent lower if the 516

13 Social Security Reform for the Elderly TABLE 2 CHANGES IN TOTAL PER-CAPITA INCOME OF AGED UNITS AFTER COLA REFORMS Actual 1997 Mean Income COLA = change in CPI minus one percentage point 18,848 20,699 16,531 15,842 Income if COLA Reduction Began 1 Year Ago Mean Amount 18,776 20,632 16,452 15,764 Percent Change 0.4% 0.3% 0.5% 0.5% Income if COLA Reduction Began 10 Years Ago Mean Amount 18,265 20,232 15,792 15,114 Percent Change 3.1% 2.3% 4.5% 4.6% Income if COLA Reduction Began 30 Years Ago Mean Amount 17,999 20,194 15,323 14,184 Percent Change 4.5% 2.4% 7.3% 10.5% COLA capped at $ amount received at 30th percentile of benefits distribution COLA capped at $ amount received at 50th percentile of benefits distribution COLA capped at $ amount received at 60th percentile of benefits distribution COLA capped at $ amount received at 50th percentile of income distribution 18,848 20,699 16,531 15,842 18,848 20,699 16,531 15,842 18,848 20,699 16,531 15,842 18,848 20,699 16,531 15,842 18,788 20,644 16,464 15,777 18,811 20,665 16,489 15,803 18,821 20,674 16,500 15,814 18,819 20,673 16,498 15, % 0.3% 0.4% 0.4% 0.2% 0.2% 0.3% 0.2% 0.1% 0.1% 0.2% 0.2% 0.2% 0.1% 0.2% 0.2% 18,125 20,153 15,567 14,899 18,387 20,352 15,910 15,252 18,503 20,441 16,063 15,410 18,502 20,442 16,057 15, % 2.6% 5.8% 6.0% 2.4% 1.7% 3.8% 3.7% 1.8% 1.2% 2.8% 2.7% 1.8% 1.2% 2.9% 2.7% 17,654 20,099 14,745 13,152 18,048 20,314 15,328 13,961 18,229 20,410 15,595 14,349 18,270 20,418 15,641 14, % 2.9% 10.8% 17.0% 4.2% 1.9% 7.3% 11.9% 3.3% 1.4% 5.7% 9.4% 3.1% 1.4% 5.4% 8.0% retirement age were increased to 70. The impact would be more severe for those in their late sixties and early seventies than for the oldest elderly, who would generally be unaffected by the higher retirement ages because they would have retired under the previous system. Age differentials in the impact on benefit income would be eliminated 55 years after the transition began, because all of the elderly would have reached age 62 after the higher retirement ages were in place. For all age groups, Social Security benefits would have fallen by 10 percent compared to actual 1997 levels if the retirement age were 67 and by 30 percent if the retirement age were 70. However, increases in the retirement age would lead to somewhat larger reductions in total per-capita income after 55 years for the oldest old than for the relatively young old, because Social Security benefits represented a larger share of total income for the oldest old. Table 4 reports after-tax income and average tax rates for the elderly under different tax reform plans. Under the tax code in effect in 1997, we estimated that the elderly paid 7.8 percent of their income in taxes. If all Social Security benefits were included in the tax base, annual federal income taxes would increase on average by $601, to 11.0 percent of income. Subjecting 85 percent of all Social Security benefits to the federal income tax, regardless of total income, would raise mean taxes by $433 over the mean taxes paid by the elderly under current law. Re- 517

14 NATIONAL TAX JOURNAL TABLE 3 CHANGES IN PER-CAPITA SOCIAL SECURITY RETIREMENT INCOME AND TOTAL PER-CAPITA INCOME AFTER INCREASING THE NORMAL RETIREMENT AGE Social Security retirement income Actual 1997 Mean Income Income if Transition Began 20 Years Ago Mean Amount Percent Change Income if Transition Began 55 Years Ago Mean Amount Percent Change Normal retirement age = 67 7,478 6,986 8,155 8,063 7,021 6,345 7,885 8, % 9.2% 3.3% 0.1% 6,730 6,288 7,340 7, Normal retirement age = 70 7,478 6,986 8,155 8,063 6,842 6,067 7,826 8, % 13.2% 4.0% 0.1% 5,235 4,890 5,709 5, Total per-capita income Normal retirement age = 67 18,848 20,699 16,531 15,842 18,398 20,068 16,265 15, % 3.0% 1.6% 18,113 20,011 15,730 15, % 3.3% 4.8% 5.0% Normal retirement age = 70 18,848 20,699 16,531 15,842 18,224 19,798 16,207 15, % 4.4% 2.0% 0.1% 16,703 18,683 14,200 13,571 Note: Computations assume that all individuals begin collecting Social Security benefits at age % 9.7% 14.1% 14.3% TABLE 4 MEAN AFTER-TAX TOTAL PER-CAPITA INCOME AND AVERAGE TAX RATES FOR THE ELDERLY, UNDER VARIOUS TAXATION REFORM PROPOSALS All Elderly Persons Ages Ages Ages 85 and Older Mean before-tax total per-capita income 18,848 20,699 16,531 15,842 Current tax law Mean after-tax income Average tax rate 17, % 18, % 15, % 14, % 100% of Social Security benefits taxable Mean after-tax income Average tax rate 16, % 18, % 14, % 14, % of all Social Security benefits taxable Mean after-tax income Average tax rate 16, % 18, % 15, % 14, % Maximum of 50% of Social Security benefits taxable Mean after-tax income Average tax rate 17, % 18, % 15, % 14, % Eliminate taxation of Social Security benefits Mean after-tax income Average tax rate 17, % 19, % 15, % 14, % 518

15 Social Security Reform for the Elderly pealing the 1993 tax increase by taxing no more than 50 percent of benefits, as a number of bills introduced in Congress in 1999 proposed, would have a small effect on taxes, reducing the mean tax liability by only $85. Abolishing taxes altogether on Social Security benefits would reduce the average tax rate for the elderly to 6.7 percent of income and would have saved the elderly $214 on average in Simulated Impact of Social Security Reform on the Distribution of Elderly Income The overall impact of Social Security reform plans on beneficiary income may mask important variation in individual effects, especially because the impact is likely to vary by the income level of the individual. Table 5 reports the predicted mean level of total per-capita income of aged units, by income decile, assuming the COLA reforms we investigated had been implemented 30 years earlier. The table indicates that the financial pain associated with the one-percent reduction in COLAs would fall disproportionately on the elderly near the bottom of the income distribution. For those in the bottom income decile, mean total per-capita income would fall by 8.3 percent, even with the availability of SSI which can offset part of the erosion in Social Security income. For elderly persons in the third decile, who are not eligible for SSI benefits, mean income would fall by 10.2 percent after the one percentage point reduction in the COLA had been in effect for 30 years. By contrast, total per-capita income for those in the top decile would drop by only 1.5 percent, because Social Security benefits represent only a small share of their income. Throughout the income distribution, the drop in income associated with the one percentage point COLA reduction would be larger for those ages 85 and older. For example, in 519 results not reported in the table, we found that among the oldest old, after 30 years, mean income would fall by 14.5 percent in the bottom decile and by 18.6 percent in the fourth decile. The modified COLA reductions that cap adjustments in order to protect the income of those with limited Social Security benefits would not protect all of the elderly with limited income from reductions in their COLAs. For example, if the COLA formula were revised so that the annual adjustment could not exceed the dollar amount received by those at the 30th percentile of the benefit distribution, mean income would fall by 2.3 percent for those in the bottom income decile and by 5.8 percent for those in the second income decile. Mean income would fall for these groups because some elderly with limited total income receive Social Security benefits above the 30th percentile. The COLA caps would protect income for all of those near the bottom of the income distribution only when the caps are tied to total income, which may be difficult administratively to achieve. All of the COLA cap schemes we examined impose little financial pain on those near the top of the income distribution. The distributional effects of delaying the normal retirement age were similar to the effects of reducing COLAs. As reported in Table 6, the percentage change in mean total per-capita income associated with increases in the retirement age was larger for persons near the bottom of the income distribution than for those near the top of the distribution. If the normal retirement age were raised to 67, mean total per-capita income would fall after 55 years by 7.1 percent for those in the bottom income decile, by 8.2 percent for those in the third decile, and by only 1.4 percent for those in the top decile. As with COLA reductions, the skewed impact arose because Social Security represents a small share of income for high-income elderly.

16 TABLE 5 CHANGES IN THE MEAN TOTAL PER-CAPITA INCOME OF AGED UNITS 30 YEARS AFTER REVISIONS TO COLA FORMULA WERE FIRST IMPLEMENTED, BY INCOME DECILE Bottom Second Third Fourth Fifth Sixth Seventh Eighth Ninth Top Actual 1997 income 3,293 6,600 8,231 9,985 11,907 14,121 17,174 21,944 29,333 66,501 COLA = CPI minus one percentile point Mean income Percent change 3, % 6, % 7, % 8, % 10, % 13, % 16, % 21, % 28, % 65, % 520 COLA capped at $ amount received at 30th percentile of benefits distribution Mean income Percent change COLA capped at $ amount received at 50th percentile of benefits distribution Mean income Percent change 3, % 3, % 6, % 6, % 7, % 7, % 8, % 9, % 10, % 10, % 12, % 13, % 15, % 16, % 20, % 20, % 27, % 28, % 64, % 65, % COLA capped at $ amount received at 60th percentile of benefits distribution Mean income Percent change COLA capped at $ amount received at 50th percentile of income distribution Mean income Percent change 3, % 3,293 6, % 6,600 7, % 8,231 9, % 9,985 11, % 11, % 13, % 13, % 16, % 16, % 21, % 20, % 28, % 28, % 65, % 65, % NATIONAL TAX JOURNAL