The Impact of Repealing the Retirement Earnings Test on Rates of Poverty

by Michael A. Anzick and David A. Weaver*

This article summarizes an analysis of the poverty implications of repealing the retirement earnings test (RET). Repealing the RET at the normal retirement age or older is unlikely to generate large poverty effects. Removing the test at age 62 or older, however, would possibly lead to large increases in poverty.


*The authors are with the Office of Retirement Policy and the Office of Research, Evaluation, and Statistics, respectively, within the Office of Policy, Social Security Administration.

Acknowledgments: The authors would like to thank Dan Durham, Keith Fontenot, Steve Goss, Tim Kelley, Mike Leonesio, Joyce Manchester, David Pattison, and Paul Van de Water for helpful comments.

Summary and Introduction

On April 7, 2000, President Clinton signed into law the Senior Citizens' Freedom to Work Act of 2000 (Public Law 106-182). Specifically, it:

The Commissioner of Social Security was in favor of the legislation. It repealed the RET for beneficiaries at the NRA or older but not for beneficiaries aged 62-64. He based his decision partly on the findings of the study below, which shows minimal poverty effects if the RET were repealed at the NRA through age 69 but significant poverty effects from repealing it for those below the NRA.

Using matched data from the U.S. Census Bureau's 1994 Current Population Survey and the Social Security Administration's Master Beneficiary Record, the study examines the potential changes in individual income and the number of people in poverty that could have resulted for beneficiaries if the RET had previously been repealed at either the NRA or at age 62.

Since the study was conducted prior to enactment of the law, many of the descriptions and examples used here apply to how the RET worked prior to enactment.

In general, the earlier people file for Social Security benefits the lower their monthly benefits will be. Eliminating the RET might lead some individuals to file for Social Security benefits at an earlier age than they otherwise might, which could reduce individual incomes and increase the number of people in poverty.

The study finds that even if individuals accelerated their filing for Social Security benefits as a result of eliminating the RET
at the NRA, the number of people below the poverty line would change little if at all. However, if the RET was eliminated at age 62 and individuals accelerated their filing for Social Security benefits, the number of people below the poverty line would increase.

This analysis provides a range of estimates for the impact that eliminating the RET would have on rates of poverty. Specifically, this article:

The Retirement Earnings Test

The Social Security Act of 1935 specified that beneficiaries would lose all of their benefits if they had any earnings. Over the years, however, Congress has eased the RET's restrictions (allowing beneficiaries to supplement their benefits with earnings) by increasing the amount of exempted earnings, reducing the age of exempted beneficiaries, and liberalizing the formulas for reducing benefits.

At the Normal Retirement Age

If the Senior Citizens' Freedom to Work Act of 2000 had not become law, beneficiaries aged 65 through 69 in 2000 would have had $1 in benefits withheld for every $3 they earned above $17,000. That threshold was to increase in stages to $30,000 in 2002 and increase automatically thereafter with the average wage in the U.S. economy.

Under the old law, delayed retirement credits (DRCs) provided compensation to workers at the NRA through age 69 whose benefits were withheld under the RET. The DRC provision increased the worker's retirement benefit for each month that benefits were fully withheld after the NRA.

DRCs still exist under the new law and are applied in two situations: when workers have delayed filing for benefits until after the NRA and when workers have filed for benefits but have requested that they not be paid. The DRC is 6 percent a year for workers age 65 in 2000, and it will increase 0.5 percentage points every 2 years until it reaches 8 percent a year for workers who turn 65 in 2008 and later. From 2008 on, benefits lost because of delayed retirement will generally be offset in an actuarially fair manner by the increase in benefits resulting from DRCs. See Box 1 for an illustration of how the RET and the DRC provisions interacted before the new legislation.

Box 1.
The effect of the retirement earnings test (RET) on beneficiaries at the normal retirement age (NRA) or older if the Senior Citizens' Freedom to Work Act had not become law

In 2000, a worker files for Social Security retirement benefits at age 65 (the NRA) and receives the full benefit of $1,000 per month. If that worker had delayed filing for benefits for another year, the benefit would have been increased by 6 percent, to $1,060 per month.

That increase accounts for the fact that the worker had not received benefits for one full year for which he or she was eligible. The delayed retirement credit (DRC) would amount to 6 percent of the full annual benefit in 2001 and later.

If the worker filed for benefits at age 65 and continued to work, he or she might have been affected by the RET, depending on the amount of earnings. For instance:

Example 1:
If earnings never exceeded the RET earnings threshold, the RET would have had no effect on the amount of the benefit. The worker would have continued to receive the full monthly benefit of $1,000 for the rest of his or her life.
Example 2:
If earnings exceeded the threshold to the extent that all of his or her benefits at age 65 would have been withheld, the worker would have received a DRC of 6 percent. The credit would have increased the benefit to $1,060 per month, accounting for the fact that he or she would have received no benefits at age 65 because of the RET.

These examples represent the extreme cases. If the worker had received partial benefits, then the DRC would have adjusted the benefit accordingly. Also, a worker who had earnings after age 65 could have had his or her full benefit increased. Annual cost-of-living adjustments would also raise the benefit amount.

At Age 62

In 2000, beneficiaries between age 62 and the NRA have $1 in benefits withheld for every $2 earned above $10,080. (That amount is adjusted annually to reflect the growth in the average wage in the U.S. economy.) Their benefits are also actuarially reduced for each month that they receive benefits before the NRA. For example, persons born in 1938 who start to collect retired-worker benefits at age 62 receive 79.2 percent of what they would have received at age 65 and 2 months (their NRA). If benefits are withheld before the NRA because of the RET, the actuarial reduction is adjusted at the NRA to exclude those months, so that benefits are not permanently reduced (Box 2).

Box 2.
The effect of the retirement earnings test (RET) on beneficiaries aged 62 through the normal retirement age (NRA)

In 2000, a worker files for Social Security retirement benefits at age 62 rather than waiting until age 65 and 2 months (his or her NRA). If the worker had waited until the NRA to file, the full benefit amount would have been $1,000 per month. Choosing to take early retirement benefits at age 62 reduced the worker's benefits by 20.8 percent, to $792 per month.

That reduction accounts for the fact that the worker will receive benefits for 38 additional months. However, over his or her lifetime, he or she is expected to receive the same total amount of benefits (based on actuarial projections of life expectancy and adjustments for interest).

If the worker continued to work after filing for benefits at age 62, he or she might be affected by the RET, depending on the amount of earnings. For instance:

Example 1:
If earnings never exceed the RET earnings threshold, the RET will have no effect on the amount of the benefits. The worker will receive 79.2 percent of the full monthly benefit amount ($792) for the rest of his or her life.
Example 2:
If earnings exceed the threshold, to the extent that the worker's benefits are partially or fully withheld in every month before reaching the NRA, benefits will be recomputed at age 65 and 2 months as 100 percent of his or her full benefit amount of $1,000 per month, accounting for the fact that he or she never received full benefits earlier because of the RET.

These examples represent extreme cases. If the worker receives partial benefits before he or she reaches age 65 and 2 months, then the adjustment to his or her benefit at age 65 and 2 months will reflect that benefit payment in an actuarially fair manner. Also, a worker who has earnings after age 62 could have his or her full benefit increased. Annual cost-of-living adjustments would also raise the benefit amount.

Methodology

To analyze the effects of eliminating the RET on poverty, data from the March 1994 Current Population Survey--a nationally representative survey--are matched with the Social Security Administration's Master Beneficiary Record. The data indicate how much each person received in Social Security and other income in 1993. (See the appendix for a detailed description of the data and methodology.)

These data are used to determine the amount of the Social Security benefit the person would have received in 1993 had there never been an RET at 62 or older or at the NRA or older. Essentially, the approach calculates the effects of changes in filing behavior on poverty. The analysis assumes that many people would have claimed benefits earlier had the RET never existed and that many would therefore have had lower Social Security income (because, for example, benefits claimed before the NRA are permanently reduced). Once the difference in Social Security income is estimated, one can determine whether that difference would change the person's poverty status. One can also determine how many more people would be in poverty if the RET had never been in effect.

Historical Approach

The historical approach examines the Social Security population at a point in time in the recent past and explores how Social Security income and poverty status would be different assuming that the RET had never existed. It does not take a cohort of people approaching their retirement years and forecast the poverty effects from repealing the RET at 62 or at the NRA.

Using a historical approach has limitations and advantages. Such an approach does not fully reflect the recent increases in women's labor force participation, real (inflation-adjusted) increases in Social Security benefits, or increases in the number of beneficiaries retiring earlier. Estimates for 1993 may therefore be somewhat larger than those for future years. Advantages include being able to know definitively at what age people in the sample claimed benefits and to determine how their Social Security income would have been different if they had claimed benefits earlier. Another advantage of using a historical approach is that it measures effects on people of all ages within the Social Security population. That is important because as beneficiaries age, they exhaust other income sources, and the importance of Social Security to total income rises (Chart 1).

Chart 1. Social Security benefits as a percentage of total income rise with age across income groups.

The historical approach does not measure changes in income (non-Social Security) that could have occurred if the RET had never existed. For example, people might choose to work and earn more in the absence of an earnings test. However, economic research has found that the RET has only modest effects on labor supply (Leonesio 1993), possibly because workers consider a number of factors when making decisions about work and retirement. Those factors include the availability and size of private pensions, health status, job characteristics, personal preferences, and so forth.

Eliminating the RET at 62 would probably cause some people to file for permanently reduced benefits, but what they would do with those reduced benefits is not known. They may save or invest a portion of them and have higher asset income later in life. Although this analysis does not take such effects into account, economic research suggests that individuals at or near the poverty level are not likely to save the additional income. Bosworth, Burtless, and Sabelhaus (1991) report negative saving rates for people in the lowest two income quintiles.

Estimating the Poverty Effects of Eliminating the RET

Eliminating the RET is likely to encourage some beneficiaries to apply earlier for benefits and, as a result, receive a lower monthly benefit in the long run. Depending on the amount of their income from other sources, that reduction in monthly benefits may reduce their total income below the poverty threshold ($7,990 a year for an aged individual and $10,070 for an aged couple in 1999).

At the Normal Retirement Age

Eliminating the RET at the NRA (currently age 65) may encourage some workers who now plan to retire later to file for benefits at the NRA, but it would have only a small effect on the incidence of poverty for several reasons. Few insured workers (only about 9 percent) currently delay receipt of Social Security benefits beyond age 65, and those who do so typically have income well above the poverty level. Furthermore, accelerating the receipt of a worker's benefit to age 65 would generally not lower benefits enough to cause the beneficiary (or the beneficiary's survivor) to become poor; benefits for a surviving spouse (widow or widower) would generally not be reduced below the deceased worker's full benefit.

Since it is unclear to what extent respondents in the 1994 CPS would have filed (that is, established entitlement to a benefit) at different ages had there never been an RET at the NRA, the poverty effects are estimated using four separate sets of assumptions about their behavior. The estimates assume that among respondents who filed for benefits after age 65, the percentage who would have opted to file at age 65 had there never been an RET at the NRA is zero (under best-case assumptions), 20 percent or 50 percent (under two intermediate assumptions), or 100 percent (under worst-case assumptions).

The resulting poverty effects would be minimal in all cases (Table 1). The estimates of the number of people who would be moved into poverty range from negligible in the best case to 2,000 in the worst case. The poverty rate for persons aged 62 or older would remain at 12 percent.

Table 1.
Estimated poverty effects of eliminating the retirement earnings test
Filing behavior assumption Net increase in poverty among those aged 62 or older
Number Rate
(percent)
Elimination at normal retirement age
 
0 a a
20 percent 500 a
50 percent 1,000 a
100 percent 2,000 a
 
Elimination at age 62
 
0 a a
20 percent 140,400 0.4
50 percent 351,100 0.9
100 percent 702,200 1.9
 
Memorandum:
Before change 4,407,200 12.0
a. The effect would be negligible.

At Age 62

By contrast, eliminating the RET at age 62 could raise the number of beneficiaries in poverty, primarily because more workers would file earlier and therefore receive benefits that would be permanently reduced below the full-benefit level. Evidence suggests that the effects of increased work efforts would be unlikely to offset those reductions.

Filing for benefits before the NRA is advantageous for workers in the short run, but it can be disadvantageous later on--particularly for their surviving spouses. For workers born in 1938, the reduction in benefits for filing at age 62 is 20.8 percent. That amount is scheduled to rise, reaching 30 percent for workers born in 1960 or later. The reduction is intended to be actuarially fair so that beneficiaries, on average, will receive the same total lifetime benefits as they would have received if they had filed for benefits at the NRA. Note, however, that much of the reduction will pass through to surviving spouses (because of Social Security's widow(er)'s limit provision) and could make their benefits inadequate.1

Estimates of the poverty effects of eliminating the earnings test at age 62 are derived using the same four assumptions about filing behavior--in this case, the assumed percentage of respondents who did not file at age 62 but would have if there had never been an earnings test at that age.2 The resulting poverty effects range from negligible (none file earlier) to substantial. The number of poor persons aged 62 or older was 4,407,200 in 1993, but had the RET at age 62 never existed, the figure could have been higher by 702,200. Among persons aged 62 or older, the poverty rate could have risen 1.9 percentage points--from 12.0 percent to 13.9 percent in 1993 (Table 1).

The effects of eliminating the RET at 62 would have been different for various demographic groups. The effects would be most severe among women, widow(er)s, beneficiaries receiving worker-only benefits, and those aged 70-79. The effects are summarized below and in Table 2, by sex, marital status, type of benefit received, and age.

Table 2.
Estimated poverty effects of eliminating the retirement earnings test at age 62 among persons aged 62 or older, by selected demographic characteristics
Demographic characteristic Net increase in the number of poor persons using alternative filing assumptions Before change Poverty rate after change using alternative filing assumptions
0% 20% 50% 100% 0% 20% 50% 100%
Total a 140,400 351,100 702,200 12.0 12.0 12.4 12.9 13.9
 
Sex
Men a 40,400 100,900 201,800 8.2 8.2 8.5 8.8 9.5
Women a 100,100 250,200 500,400 14.8 14.8 15.2 16.0 17.1
 
Marital statusb
Married a 47,000 117,600 235,200 5.8 5.8 6.1 6.4 6.9
Widowed a 77,300 193,300 386,700 19.6 19.6 20.3 21.4 23.2
Divorced/separated a 10,200 25,500 51,000 23.7 23.7 24.1 24.7 25.6
Never married a 5,900 14,600 29,300 22.5 22.5 22.8 23.4 24.3
 
Benefit typec
Worker-only a 76,600 191,400 382,900 9.1 9.1 9.4 10.0 11.0
Spoused a 13,100 32,800 65,500 8.1 8.1 8.3 8.7 9.4
Surviving spousee a 47,600 119,000 237,900 19.2 19.2 20.0 21.1 22.9
 
Age
62-69 a 12,300 30,900 61,700 10.0 10.0 10.1 10.2 10.4
70-79 a 69,700 174,200 348,400 11.4 11.4 11.9 12.6 13.8
80-89 a 50,900 127,300 254,600 17.5 17.5 18.3 19.7 21.9
90 or older a 7,500 18,700 37,500 19.1 19.1 19.9 21.0 22.8
a. The effect would be negligible.
b. Respondents to the Current Population Survey indicated their marital status at the time of the survey. They may or may not be receiving a benefit based on their marital status.
c. Some persons affected by this proposal are not beneficiaries (they share a household with a beneficiary), so rows do not add up to the total number of persons affected.
d. Includes spouse beneficiaries who are divorced or dually entitled.
e. Includes surviving-spouse beneficiaries who are divorced or dually entitled.

Sex. More than 500,000 women could be moved into poverty, accounting for 71 percent of the total moved into poverty. Their poverty rate could increase from 14.8 percent to 17.1 percent.

Marital Status. Widow(er)s could account for 55 percent of the total moved into poverty (up to 387,000). The poverty rate for widow(er)s could increase from 19.6 percent to 23.2 percent. Married couples could account for 33 percent (235,000) of the total moved into poverty, and their poverty rate could increase from 5.8 percent to 6.9 percent.

Benefit Type. Beneficiaries receiving worker-only benefits could account for 55 percent of the total moved into poverty. As many as 238,000 surviving-spouse beneficiaries could be moved into poverty (34 percent of the total), and their poverty rate could rise from 19.2 percent to 22.9 percent.

Age. About half of the total moved into poverty could be aged 70-79. The poverty rate for that group could rise from 11.4 percent to 13.8 percent. Beneficiaries aged 80-89 could account for 36 percent of the total moved into poverty. Their poverty rate could increase from 17.5 percent to 21.9 percent.

Notes

1. The widow(er)'s limit provision operates as a ceiling on survivor benefits, ensuring that a survivor benefit does not exceed the greater of either the benefit the deceased worker would be receiving if alive or 82.5 percent of the deceased worker's primary insurance amount.

2. In 1998, 48 percent of insured workers opted for benefits at either 62 and 0 months or 62 and 1 month; 79 percent of insured workers opted for benefits before age 65.

References

Bosworth, Barry; Gary Burtless; and John Sabelhaus. 1991. "The Decline in Saving: Evidence from Household Surveys." Brookings Papers on Economic Activity, No. 1, pp. 183-256.

Leonesio, Michael V. 1993. "Social Security and Older Workers." Social Security Bulletin, Vol. 56, No. 2, pp. 47-57.

Appendix on Data and Methods

Data

The data used for the poverty analysis are from two sources: the March 1994 Current Population Survey (CPS) and the Master Beneficiary Record (MBR). The 1994 CPS is a nationally representative, cross-sectional survey conducted by the U.S. Census Bureau. Among other things, it contains the information on family income and family composition needed to determine whether a respondent is above or below the appropriate federal poverty threshold. The MBR is an administrative-record database maintained by the Social Security Administration. It contains information on Social Security benefits, such as the amount and type of benefit and the date of entitlement.

The CPS and the MBR can be linked for the large majority of respondents in the CPS--Social Security numbers are available for about 81 percent of CPS respondents aged 15 or older. An exact-match file is used for the poverty analysis (sample weights were divided by 0.81 to adjust for nonmatches). To protect the confidentiality of respondents, use of this exact-match file is highly restricted. It may be used only for research purposes and only by persons who receive authorization from the Census Bureau.

An important limitation of the CPS data is that respondents tend to underreport income. A comparison of CPS-based estimates with independently derived estimates indicates that overall income is underreported by about 11 percent in the CPS (U.S. Census Bureau 1992).

How underreporting influences the results of this study is unclear. A respondent who underreports income may inaccurately be classified as poor initially. In the simulations, such a person will never move from above poverty to below poverty. Had income been measured accurately, such a move could have occurred. Therefore, too few persons may be "pushed" into poverty in the simulations. In other cases, though, under-reporting could cause an opposite effect (some persons being artificially "pushed" into poverty).

Simulating Changes in Social Security Income

Our goal is to estimate how much higher or lower Social Security income would have been for CPS respondents in calendar year 1993 had there never been a retirement earnings test (RET) at 62 (the early retirement age) or older and, separately, had there never been an RET at 65--the normal retirement age (NRA)--or older. The results are an approximation of the long-run effects of eliminating the RET at those ages. However, our approach is historical. It examines the Social Security population at a point in time in the recent past. The historical approach has limitations in that recent trends in women's labor force participation and in early retirement are not fully reflected. It does have advantages, though. We have detailed information regarding actual Social Security benefits. If we forecast the effects of eliminating the RET for persons not yet retired, the accuracy of the forecasts on a number of the key variables would be uncertain.

Our focus is on how filing decisions would have differed had the earnings tests never been in effect. As an example, consider a 70-year-old man in the CPS who receives retired-worker benefits that he filed for 5 years earlier (at age 65). He may have had substantial earnings from ages 62 to 65 and may not have bothered to file for Social Security before age 65 because an earnings test would have prevented benefits from being paid. If an earnings test had never been in effect, he could have received benefits as early as age 62 years and 1 month and might have filed for them.1 These age 62 and 1 month benefits would be permanently reduced benefits. Therefore, had an earnings test never been in effect, we might have observed such a 70-year-old in the CPS with permanently reduced benefits rather than full (age 65) benefits. Even if the 70-year-old man had only limited earnings before age 65, he may have thought an earnings test would have prevented the payment of benefits and not filed until age 65. In the absence of an earnings test, he might have claimed permanently reduced benefits.

Eliminating the RET at Age 62 or Older. Most of the discussion in this section focuses on changes in Social Security income under a set of assumptions about filing behavior we refer to as "worst-case" assumptions. The assumptions are labeled such because they produce an estimate of the largest increase in poverty. Next, we discuss the poverty impacts under "best-case" assumptions and two intermediate sets of assumptions.

Worst-case assumptions. The worst case assumes that had the RET never been in effect, all persons would have filed for benefits as early as possible. First, consider those who were receiving benefits at the time of the CPS.2 A hypothetical benefit amount is computed that reflects the assumption of early filing. Next, the ratio (R) of the hypothetical benefit to the actual benefit the person received at the time of the CPS is calculated. The MBR file provides the amount of Social Security the person received in calendar year 1993 (SSANNUAL). The change in the person's Social Security income (CHGSS) due to the assumption of early filing is estimated to be (R - 1) × (SSANNUAL). For example, consider again the 70-year-old retired worker who filed for benefits at age 65. Because he filed at the NRA, his actual monthly benefit amount is equal to his full primary insurance amount (PIA). Suppose his PIA is $1,000. SSANNUAL would likely be 12 times that amount (or $12,000). Had he filed as early as possible (age 62 and 1 month), his monthly benefit amount would be about 80.6 percent of his PIA (or $806). Therefore, R is 0.806 and CHGSS is equal to - $2,328 (that is, -$2,328 = (0.806 - 1) × $12,000). Thus, if this man had filed as early as possible, his annual Social Security benefits would be 19.4 percent lower ($2,328 = 0.194 × $12,000) than benefits claimed at the NRA.

An important group in the simulation are those who receive widow(er) benefits from Social Security at the time of the CPS. There are two assumptions involved in calculating their hypothetical benefits. First, we assume that those persons would have filed for widow(er) benefits as soon as possible, and second, we assume that their deceased spouses would have claimed retired-worker benefits as soon as possible in the absence of the RET.

The second assumption is unique to widow(er) beneficiaries and is needed because of the widow(er)'s limit feature of the Social Security law. For example, consider an 80-year-old woman in the CPS whose husband died 5 years earlier (when the woman was 75 years old). If her deceased husband first received retired-worker benefits at age 65, then the actual widow's benefit would be equal to the husband's PIA. Since the woman was already past the NRA when her husband died (she was 75), she cannot file for reduced widow benefits. Consequently, the first assumption does not affect the calculation of the widow's hypothetical benefit. However, had her husband filed for his retired-worker benefit as early as possible (age 62 and 1 month), the widow's limit in the law would reduce her benefit to 82.5 percent of her husband's PIA. Because we assume the husband would have filed early in the absence of the RET, the widow's hypothetical benefit is set at 82.5 percent of his PIA. In the simulation, her annual Social Security income would drop by 17.5 percent.

In calculating a hypothetical benefit, only one benefit is considered. For dually entitled beneficiaries, the hypothetical benefit is based on the auxiliary or survivor benefit. For example, the hypothetical benefit of a person who is dually entitled to a retired-worker benefit and a widow benefit is calculated as if the person receives only a widow benefit and assuming the earliest possible filing date.3

Note that a person who is receiving only one benefit at the time of the CPS might be eligible for other benefits. Those other benefits are not reflected in the simulations. For example, a woman who is receiving only a widow benefit at the time of the CPS might be eligible for some other benefit, such as an unclaimed retired-worker benefit or survivor benefit based on another marriage. In the simulations, only her current widow benefit is considered.

For many CPS beneficiaries, we simulate no change in Social Security income. Many retired-worker and spouse beneficiaries actually filed for benefits as early as possible. Assuming that the RET never existed would not change that, so those respondents do not have simulated changes in Social Security. Also, disabled-worker benefits are not reduced for age. Persons who receive those benefits or who received them until they were automatically converted to retired-worker benefits at age 65 do not have simulated changes in Social Security income.

For persons receiving benefits at the time of the CPS, the simulated change in Social Security income is always zero or negative. For persons who were not receiving benefits at the time of the CPS, we sometimes simulate increases in Social Security. Consider, for example, a man who turns 63 in 1993 and who files for unreduced retired-worker benefits in 1996. Had the RET never existed, we assume that he would have filed for benefits at age 62 and 1 month and received reduced benefits in calendar year 1993. The amount he would have received is estimated and used as the simulated change in Social Security. Similar procedures were used for spouse and widow(er) beneficiaries. Note, however, that the matched MBR records do not contain information after October 1996 (the date the records were actually pulled from the MBR). Thus, some persons who would have received benefits in 1993 under the early-filing assumption could not be identified.

Best-case assumptions. The best case assumes that eliminating the RET will have no effect on filing decisions. Under that assumption, it is asserted that there would be no poverty effects from eliminating the RET. That, of course, is not exactly right. Some persons file for benefits and then have them suspended because of the earnings test. Even if those persons did not change their filing behavior, there would be changes in their Social Security income and potential poverty effects. However, the large majority of persons whose benefits are suspended because of the earnings test are aged 65 to 69. Those who still enjoy high earnings at such relatively late ages are unlikely to be drawn from the lower-income part of the population. The estimates therefore assume that the poverty effects would, for practical purposes, be zero.4

Intermediate assumptions. We also estimate poverty effects for two sets of intermediate assumptions. One set assumes that 20 percent of those who postpone receipt of Social Security would have filed as early as possible had the RET never existed. Under the other set, that figure is 50 percent. Rather than specifically simulate the poverty effects, a linear interpolation is taken between the "worst-case" and "best-case" poverty effects. For example, the net change in the number of poor persons (aged 62 or older) is +702,200 under the worst-case assumptions and is zero under the best-case assumptions. The 20 percent estimate is 140,400 (140,400 = 0.2 × 702,200). The 50 percent estimate is 351,100 (351,100 = 0.5 × 702,200).

Eliminating the RET at the Normal Retirement Age or Older. Had the RET never existed at the NRA or older, the simulations assume that decisions about whether to file before the NRA would not have changed. For example, consider someone who filed for benefits at the NRA. Had an earnings test been in place before age 65 but not after, that person probably would not have changed his or her filing decision. Thus, in simulating changes in Social Security, we focus only on CPS respondents who filed after the NRA.

The worst-case assumptions are that persons who filed after the NRA would have filed at that age in the absence of an earnings test (and thus would not have acquired delayed retirement credits). The calculations are analogous to those described previously. Again, the best-case assumptions are that there would be no poverty effects, and the intermediate assumptions are based on linear interpolations between worst-case and best-case results.

Changes in Overall Income

Generally, we do not simulate changes in income other than Social Security, and that approach may impose an upward bias on the resulting poverty estimates. For example, consider the 70-year-old man in the CPS who filed for full benefits at age 65. Under worst-case assumptions, he would be receiving permanently reduced benefits at age 70 had the RET at age 62 or older never existed, but what he would have done with the early benefits is unknown. He may have saved or invested a portion of them and had higher asset income at age 70. The simulations do not account for such second-round effects. However, since persons from the lower part of the income distribution may have a low propensity to save, those second-round effects may not affect the poverty results very much.5 Eliminating the earnings tests could increase labor supply and earnings, but the simulations also do not account for those effects. A review of economic studies indicates, however, that eliminating the earnings tests would have modest effects on labor supply (Leonesio 1993) perhaps because many factors (private pensions, health, job characteristics, personal preferences, and so forth) influence work and retirement decisions.

The simulations account for some interactions with the Supplemental Security Income (SSI) program. If a person reports receiving SSI in the CPS and his or her Social Security income is simulated to decrease (increase), then we simulate an increase (decrease) in SSI. That approach ignores some SSI interactions. For example, if a person's Social Security decreases, he or she may become eligible for SSI. That type of interaction is not simulated.

Once we have determined the changes in the income of individual respondents, we calculate the changes in the income of families. We add or subtract the change in family income from the family income reported in the CPS and determine whether the new family income is above or below the federal poverty threshold appropriate for the family. If it is below, all persons in the family are classified as poor. From the CPS, we know which persons are initially poor. We then determine the net change in the number of poor persons. That approach is consistent with the official measurement of poverty in the United States, which compares family income with a family threshold. A person who has no change in his or her individual income can be pushed into poverty if a family member has a change in income. Thus, persons of any age and persons who are not eligible for Social Security can have a change in their poverty status.

Notes

1. The Social Security law does not allow most persons to receive retirement benefits for the month they turn 62. Also, "filing" for a benefit at a particular age refers to establishing entitlement to the benefit at that age.

2. Receiving benefits at the time of the CPS refers to receiving benefits for December 1993.

3. A person entitled to a worker benefit and to a larger auxiliary or survivor benefit is subject to Social Security's dual entitlement provisions. The effect of those provisions is that the total benefit paid is approximately or exactly equal to the amount that would be paid if the person were entitled to only the auxiliary or survivor benefit.

4. The results of Pattison and others (1990) support the idea that persons aged 65 to 69 whose benefits are suspended because of the earnings test are not from the lower part of the income distribution.

5. Bosworth, Burtless, and Sabelhaus (1991) report negative saving rates for the lowest two income quintiles.

References

Bosworth, Barry; Gary Burtless;and John Sabelhaus. 1991. "The Decline in Saving: Evidence from Household Surveys." Brookings Papers on Economic Activity, No. 1, pp. 183-256.

Leonesio, Michael V. 1993. "Social Security and Older Workers." Social Security Bulletin, Vol. 56, No. 2, pp. 47-57.

Pattison, David; Benjamin Bridges, Jr.; Michael V. Leonesio; and Bernard Wixon. 1990. Simulating Aggregate and Distributional Effects of Various Plans for Modifying the Retirement Earnings Test. ORS Working Paper No. 46, Office of Research and Statistics, Social Security Administration (July).

U.S. Census Bureau. 1992. "Measuring the Effect of Benefits and Taxes on Income and Poverty: 1979 to 1991." Current Population Reports (Series P-60, No. 182-RD). Washington, D.C.: U.S. Government Printing Office.