The Widow(er)'s Limit Provision of Social Security

Summary

Widow benefits have been a part of the Social Security program since the 1939 Amendments to the Social Security Act (widower benefits were added later). For many years, the Social Security law called for paying a widow(er) a fraction of the deceased worker's primary insurance amount (PIA). However, the worker—while alive—may have received the full PIA as his or her retirement benefit. Over time, arguments were made that a widow(er) should be treated as generously as his or her spouse was.

The 1972 Amendments to the Social Security Act allowed for a widow(er) to receive a full PIA, subject to actuarial reductions if the widow(er) benefit was claimed before the normal retirement age (NRA) and subject to a new provision of the law commonly referred to as the widow(er)'s limit. Generally, the widow(er)'s limit specifies that if a worker received reduced retirement benefits (because the worker claimed benefits before the NRA), then the worker's widow(er) cannot receive a monthly benefit equal to the full PIA. Rather, the widow(er)'s benefit is generally limited to the amount the worker would receive if he or she was still alive. The limit provision appears to be motivated by the overall intent of the 1972 Congress to pay a benefit to a widow(er) that was comparable with what the worker received.

A number of changes to the limit provision have been discussed. The most fundamental change—abolishing the limit—would increase benefits for about 2.8 million widow(er)s and would cost about $3.1 billion a year. Most of the additional government expenditures would not go to the poor and the near poor. Another change would be more successful in aiding low-income widow(er)s: requiring that the limit amount never be set below the average PIA among all retired-worker beneficiaries. About 58 percent of the government expenditures from that option would be received by the poor and the near poor. Overall, 1.2 million widow(er)s would be helped, and the cost would be about $816 million a year.

Although the limit provision is consistent with the overall intent of the 1972 Congress, it can have effects that may have been unintended and that some policymakers might consider unusual. Persons who delay receipt of Social Security benefits usually receive higher monthly benefit amounts, but a widow(er) who faces a limit amount cannot increase his or her monthly benefit amount through delayed receipt of benefits. Thus, many persons who are widowed before the NRA face strong incentives to claim benefits early. That is somewhat unusual because the actuarial adjustments under Social Security are approximately fair, so there are no cost savings to the Social Security program from "forcing" a widow(er) to claim early as opposed to allowing him or her to delay receipt of benefits in exchange for a higher monthly amount. And many widow(er)s would be better off if they could use the Social Security program to, in effect, save (that is, delay receipt of benefits in exchange for a higher amount later).

This paper analyzes two other options that would provide widow(er)s with additional filing options under Social Security. The ARLA option would ultimately help about 229,000 widow(er)s, and the cost would be small (about $69 million a year). The SARLA option would help about 117,000 widow(er)s, and the cost would be about $41 million a year.

Robert J. Myers, a former Chief Actuary of Social Security, has offered a proposal that would provide relief from the widow(er)'s limit in cases in which the worker dies shortly after retirement. That proposal would help about 115,000 widow(er)s, and the cost would be small (about $57 million a year).

Introduction

Legislation in 1972 authorized substantial increases in the benefit amounts paid to persons receiving widow(er) benefits from Social Security. Before the legislation, a widow(er) was usually eligible for a monthly benefit amount equal to the primary insurance amount (PIA) of his or her deceased spouse multiplied by 0.825. (The Social Security Administration (SSA) computes a person's PIA based on the person's average earnings in Social Security-covered employment.) Under the 1972 act, a widow(er) could potentially receive the entire PIA rather than a fraction of it. While living, a deceased spouse may have been receiving a retirement benefit based on work in covered employment that was equal to his or her PIA, and the legislative change appears to have been guided, at least partially, by a belief that a widow(er) should not receive less than what the deceased spouse was receiving.

A deceased spouse would not have been receiving a benefit amount equal to his or her PIA if he or she filed for the retirement benefit before Social Security's normal retirement age (NRA), which is age 65 for those born before 1938. For example, if a deceased husband had filed on his 63rd birthday, he would have been receiving, before his death, a retirement benefit equal to the PIA multiplied by 0.867.¹ Following the principle that a widow(er) should receive an amount based on that of the deceased spouse, the 1972 act contained a provision that, generally, limited the widow(er)'s benefit amount to the amount that the deceased spouse would be receiving if he or she was still alive. Thus, in the example just presented, the widow of the deceased husband could not receive a monthly benefit amount greater than 0.867 of PIA. This feature of Social Security is sometimes referred to as the widow(er)'s limit. In the actual operation of the Social Security programs, it is referred to as the RIB-LIM, reflecting the fact that the retirement insurance benefit (RIB) of someone who worked in covered employment is limiting the amount of a widow(er)'s benefit.²

In the preceding example, the deceased worker filed for retirement benefits on his 63rd birthday and received a benefit less than the full PIA, namely, 0.867 of PIA. It is useful to describe how retirement benefits depend on age of filing. Consider a person born before 1938. If such an individual becomes entitled to retirement benefits in the month he or she is age 62 (the earliest age of eligibility), he or she receives a monthly benefit amount equal to 0.8 of PIA. For every month (in the period that starts with the month the person turns 62 and ends in the month before the month the person turns 65) that a retirement benefit is not claimed, the monthly benefit amount is increased by an actuarial adjustment equal to 0.00556 × PIA.³ Thus, to figure a retirement benefit amount claimed before age 65, first determine the number of months that have elapsed since the person turned 62, denoted by N. Then, the monthly benefit amount is equal to [(N × 0.00556) + 0.8] × PIA. For someone who files on his or her 63rd birthday, 12 months have elapsed since age 62 and the benefit is [(12 × 0.00556) + 0.8] × PIA, or about 0.867 × PIA. A person who files at age 65 receives the full PIA, and a person who files after age 65 receives the full PIA plus additional amounts referred to as delayed retirement credits.

The computation of widow(er) benefits has a similar structure to the computation of retirement benefits. Consider a person born before 1940. A widow(er) who files for benefits on his or her 60th birthday is eligible for a benefit amount equal to 0.715 times the deceased worker's PIA (aged widow(er) benefits cannot be paid before age 60). For every month (in the period that starts with the month the person turns 60 and ends in the month before the month the person turns 65) that a widow(er) benefit is not claimed, the monthly benefit amount is increased by an actuarial adjustment equal to 0.00475 × PIA. If a person files for a widow(er) benefit on or after his or her 65th birthday, he or she receives the full PIA. All of this is subject to the widow(er)'s limit provision—that the benefit amount, generally, may not exceed what the deceased spouse would be receiving if he or she was still alive.

As an example of how benefit amounts would be calculated, consider again the man who filed for retirement benefits on his 63rd birthday and who was receiving, before his death, an amount equal to 0.867 of PIA. Assume that his wife was 59 years old when he died. If she filed on her 60th birthday, her benefit amount would be equal to 0.715 of PIA, and the widow(er)'s limit provision would be irrelevant because her amount was already below the amount her husband would be receiving if alive. If she chose to delay receipt of widow's benefits, she could earn the 0.00475 actuarial adjustment for each month she delayed receipt, so long as the widow(er)'s limit was not reached. If she delayed receipt for 32 months (she files at age 62 years and 8 months), her benefit amount would be equal to [(32 × 0.00475) + 0.715] × PIA, or 0.867 × PIA. It would then be equal to the amount her deceased husband would be receiving if alive and, because of the widow(er)'s limit, it would be at its maximum amount. ⁴

The actuarial adjustments made to widow(er)'s benefits are approximately fair, meaning that the lifetime expenditures under the Social Security programs are about the same regardless of when a benefit is filed for. Said differently, the savings to Social Security from a widow(er) benefit not being paid for a given month are balanced by the costs of paying an additional 0.00475 of PIA each month once benefits are claimed. In the last example presented, the widow(er)'s limit prevents the widow from earning actuarial adjustments after she turns 62 years and 8 months old. This feature of the law is somewhat unusual and is the focus of much of this paper. If the widow in this example is knowledgeable about the Social Security rules, she will not postpone receipt of widow's benefits past age 62 years and 8 months because doing so means that she will receive Social Security benefits for fewer months in her life with no adjustment to the monthly benefit amount (that is, her lifetime benefits will be unambiguously lower if she does not file at that age). So, the widow's limit will "force" her to file by age 62 years and 8 months. If the Social Security law was to allow her to earn fair actuarial adjustments beyond that age, there would be no additional costs to Social Security because the lifetime expenditures associated with her filing at age 62 years and 8 months are about the same as expenditures associated with her filing at some later age and receiving fair actuarial adjustments. However, allowing her additional filing options might improve her well-being. Some complications to this line of thinking are discussed later.

The widow(er)'s limit is a fairly obscure feature of the Social Security programs, but it is an important one. About 1 in 3 widow(er) beneficiaries has his or her benefits limited because of that provision. There has been legislative interest in modifying the widow(er)'s limit provision. The options for doing so range from the modest (allowing widow(er)s to receive adjustments to the capped amounts by delaying receipt of benefits) to the substantial (abolishing the widow(er)'s limit). The paper evaluates several of these options.

Abolishing the RIB-LIM

There has been much discussion about whether benefits provided to widows are adequate. ⁵ Some discussions have used as a reference point the benefit amounts the husband and wife were receiving before the husband's death. For example, at the beginning of the program, aged widows were eligible to receive 75 percent of PIA. Because the wife's benefit was equal to 50 percent of the PIA and the husband, as a worker, was eligible for 100 percent of PIA, a couple could receive 150 percent of PIA. It was reasoned that the widow could get by on 75 percent of PIA or, equivalently, one-half of the amount the couple received (U.S. Advisory Council on Social Security 1939). Note that under the original rules, if the wife died before the husband, he would still receive his worker benefit, which would exceed 75 percent of the PIA. Over time, arguments were made that the widow should be treated as generously and receive an amount equal to what the worker was receiving. In 1961, Congress took a step in that direction, and widow(er) benefits were increased to 82.5 percent of PIA. In 1972, widow(er) benefits were increased to make them comparable with benefits paid to workers.

Note that the 82.5 percent figure from the 1961 legislation still has some relevance. Although the widow(er)'s limit provision, in general, requires that a widow(er) benefit not exceed what the worker would be receiving if alive, a special provision of the Social Security law requires that the limit actually be the greater of what the worker would be receiving if alive or 82.5 percent of the worker's PIA. So, for example, if a deceased worker filed for retirement benefits at age 62, he or she would be receiving 80 percent of PIA if alive, but the worker's widow(er) would face a higher limit, namely, 82.5 percent of PIA.

Arguments For and Against Removing the Widow(er)'s Limit

The widow(er)'s limit, which generally ensures that the widow(er) does not receive more than the amount the worker would be receiving, appears to be a result of a long-running discussion about what constitutes an adequate benefit for a widow relative to the amount a couple was receiving. Should the limit be removed? One argument against its removal might be that a worker's spouse would receive better Social Security survivor protection than the worker. For example, if the husband is a worker who received actuarially reduced benefits and his wife (who, say, did not work) died before him, then the worker would be left with only the actuarially reduced benefit. If the widow(er)'s limit was removed and the husband died before the wife, then she could receive more than the actuarially reduced worker benefit. Another possible argument in favor of retaining the widow(er)'s limit is that it provides strong incentives for married workers to delay retirement. Workers who take early retirement benefits not only cause benefits to be reduced over their lifetime but also over the lifetime of the widow(er). Finally, as discussed in detail later, many widow(er)s are currently affected by the limit, and its removal would not be inexpensive. Some policymakers would object to adding significant costs to the Social Security programs at a time when the system faces a long-run actuarial deficit.

Arguments in favor of removing the widow(er)'s limit might focus on adequacy issues. As noted above, one of the problems in removing the limit is that it would have the seemingly unusual effect of providing better Social Security survivor protection to the spouse than to the worker. However, the spouse may need a higher benefit from Social Security because the spouse is less likely to have access to other types of income, such as pension income or earnings. In 1998, the poverty rate among widows aged 65 or older was 16.8 percent, which was higher than the poverty rate for the overall U.S. population (12.7 percent) and for the overall aged population (10.5 percent) (U.S. Congress 2000). Some policymakers might believe that removing the RIB-LIM is warranted because doing so would provide a more adequate benefit structure for widows. ⁶

Other arguments can be made for removing the limit. One is that the limit-affected widow(er) had no formal say in a decision (early retirement of the worker) that dramatically affected the widow(er)'s economic well-being. Of course, workers may consult with their spouses about the early retirement decision, but that will not always occur. Workers may not know about or understand this complex provision of the law, or, in some cases, may not take into account the concerns of their spouses. Also, surviving divorced spouses are subject to the limit just as widow(er)s are. Certainly, in many cases, workers would not consider the well-being of ex-spouses when making decisions about early retirement.

Distributional Effects and the Costs of Removing the Limit

If the goal of removing the limit is to provide increased benefits to persons with low or moderate income, then one would want to know the percentage of additional government expenditures that such persons would actually receive. If the goal of removing the limit is something else, a secondary concern is still likely to be whether much of the public's money is transferred to those whose income is already adequate. The distributional results will help inform policymakers as to whether the primary or secondary goals are being met. The cost estimates presented here will help policymakers assess whether such a change is feasible in light of the solvency problems facing Social Security.

The distributional results are based on the March 1994 Current Population Survey (CPS) exactly matched to benefit records from the Social Security Administration (SSA). About 81 percent of adult respondents to the CPS provided enough information to match SSA records. It is impossible to identify widow(er) beneficiaries in the CPS without the matched benefit records, and any sample of widow(er) beneficiaries used in the distributional analysis excludes respondents for whom records were not matched. The CPS sample weights have been adjusted to account for the missing respondents. ⁷ To protect the confidentiality of CPS respondents, use of this matched file is restricted. It may only be used for research and by those who receive authorization from the U.S. Census Bureau.

To provide some perspective, I first present some basic information on the economic well-being of widow(er) beneficiaries affected by the limit, all widow(er) beneficiaries, all Social Security beneficiaries, and the U.S. population as a whole (see Table 1). ⁸ Economic well-being is measured by a person's welfare ratio, which is the ratio of family income to the appropriate federal poverty threshold. For 1993 calendar year income, 25 percent of the total U.S. population had a welfare ratio below 1.5, 50 percent below 2.78, and 75 percent below 4.57. I use those cutoffs and the poverty cutoff to describe the economic well-being of the various groups.

About 1 in 5 limit-affected widow(er) beneficiaries is poor (that is, the welfare ratio is less than 1). That incidence of poverty is similar to that for all widow(er) beneficiaries but is higher than that for both the overall Social Security population (12.6 percent of whom are poor) and the overall U.S. population (15.1 percent of whom are poor). That basic pattern holds when other welfare-ratio cutoffs are used to compare groups. Note that 81 percent of limit-affected widow(er)s have welfare ratios below the national median. These figures suggest that changes to the widow(er)'s limit will mainly affect persons with low to moderate income.

For each limit-affected widow(er) in the CPS sample, I calculated the additional government income the widow(er) would have received in calendar year 1993 had the limit provision not existed in that year. As an example, consider a woman who was widowed after she reached the normal retirement age, who received widow benefits in each month of 1993, and who was married to a man who would be receiving 0.867 × PIA if he was alive. In 1993, Social Security would have paid the widow (12 × 0.867 × PIA). If the RIB-LIM provisions had not been in effect in 1993, she would have received the full PIA each month (because she was widowed after the normal retirement age), and her additional government income would be (12 × PIA) − (12 × 0.867 × PIA).⁹

Tabulations reveal that 14 percent of the additional government expenditures that would have occurred in 1993 (had there been no limit) would have been received by persons with income below the poverty threshold, 40 percent by those with income below 150 percent of the poverty threshold, 77 percent by those with income below 278 percent of the poverty threshold, and 91 percent by those with income below 457 percent of the poverty threshold. There would have been 88,000 fewer poor persons in 1993 had the widow(er)'s limit not been in effect. ¹⁰

If the policy goal is to channel money to elderly persons with very low income, then removing the widow(er)'s limit will be an inefficient way to do so, because the overwhelming percentage of additional government expenditures will be received by the nonpoor. If, instead, the goal is to target increased expenditures toward those with low or moderate income, the option is more successful. Also, some policymakers may object to the limit on fairness grounds (for example, they may believe that the widow(er) should not suffer a reduction because of something the worker did). In that case, policymakers can be assured that a secondary effect of the policy change will not be a transfer of the public's money to affluent persons, because a large majority of the additional expenditures will be received by those whose income is below the median income of the U.S. population as a whole.

To assess the costs associated with removing the RIB-LIM, tabulations were made from several historical 1-percent samples of the benefit records maintained by SSA (those samples are not related to the CPS in any way). The 1-percent samples are pulled every December and contain information on benefits paid for the month of December; those benefits contain the cost-of-living adjustment (COLA) for the year and are actually received by beneficiaries in January. For each limit-affected widow(er) in a December sample, the additional amount he or she would receive from Social Security in the absence of the limit was calculated. Total cost for the year following a December pull was calculated by summing the individual amounts over the sample, multiplying by 1,200 (to reflect the 1 in 100 sampling and to put the costs in annual terms), adjusting for COLAs through December 1998 (to put costs in 1999 dollars), and applying an upward adjustment to reflect the costs associated with newly entitled widow(er) beneficiaries. ¹¹

Had the RIB-LIM provision been repealed for benefits payable for December 1998 and later, Social Security would have paid out an additional $3.1 billion in benefits in calendar year 1999. All widow(er) benefit payments for 1999 total an estimated $60.1 billion. Removing the RIB-LIM would therefore have increased widow(er) benefits by about 5 percent. Table 2 provides cost estimates and the ratio of costs to all widow(er) benefit payments.

The estimates clearly indicate that the real cost of removing the limit has been rising over time. The constant-dollar cost rose about 43 percent over the 1994–1999 period. In 1999, 2.8 million widow(er)s would have received more in Social Security if the RIB-LIM provisions had been repealed. The number of widow(er)s who would receive more in Social Security if the RIB-LIM was abolished has also risen over time (by about 19 percent over the 1994–1999 period). The increased costs and numbers affected are probably related to trends in early receipt of worker benefits. Real costs may be tapering off, however. The Office of the Chief Actuary (OCACT) has estimated the costs of repealing the RIB-LIM over the 2000–2004 period (Chaplain 1999a). Over that period, real costs rise by only 9.4 percent. ¹²

Raising the Widow(er)'s Limit

One way to think about how the widow(er)'s limit works is to realize that it is the lesser of two amounts that is actually paid to a widow(er). The first amount—the initial benefit amount—is what the widow(er) would receive if the limit provision did not exist. Basically, the initial benefit amount is equal to the PIA if the widow(er) is entitled after the normal retirement age; otherwise, the initial benefit amount reflects actuarial reductions because the widow(er) claimed a benefit early. The second amount—the RIB-LIM amount—is the higher of either the amount the worker would be receiving if alive or 82.5 percent of PIA. Under current law, the widow(er) is paid the lesser of the initial benefit amount and the RIB-LIM amount. One can express these relationships mathematically as follows.

1. RIB-LIM amount = MAX (82.5 percent of PIA, amount worker would receive if alive)
2. Actual Payment to widow(er) = MIN (RIB-LIM amount, initial benefit amount)

Note from equation 1 that the 82.5 percent of PIA acts as a "floor" to the RIB-LIM amount. If a deceased worker (born before 1938) filed for reduced retirement benefits at age 62, he or she would be receiving 80 percent of PIA if he or she was still alive. However, the RIB-LIM amount is not set that low because it cannot fall below 82.5 percent of PIA. Although the 82.5 percent of PIA acts as a floor to the RIB-LIM amount, it does not mean that the benefit the widow(er) actually receives can never be below 82.5 percent of the PIA. If the widow(er) files for benefits before the NRA, his or her initial benefit amount might be below 82.5 percent of PIA, and, following equation 2, it is the initial benefit amount that is paid.

One possible change to the widow(er)'s limit would be to augment the current floor to the RIB-LIM amount. Specifically, consider a change that would set the floor at the maximum of either 82.5 percent of the worker's PIA or the average PIA among all retired workers (AVGPIA). ¹³ Under such an option, equation 1 would read as follows:

1. RIB-LIM amount = MAX (82.5 percent of PIA, AVGPIA, amount worker would receive if alive)

The option could, in a straightforward way, be implemented under the current general structure of benefits: if AVGPIA exceeds 82.5 percent of the deceased worker's PIA, use AVGPIA instead of 82.5 percent of PIA when calculating widow(er) benefits. ¹⁴

This option would provide relief from the widow(er)'s limit but in a manner that helped low-income widow(er)s. Consider a limit-affected widow(er) (that is, one whose current-law benefit is equal to the RIB-LIM amount). If that widow(er)'s current-law benefit amount exceeded AVGPIA, he or she would not receive increased benefits because AVGPIA would not be the maximum of the three amounts in the revised equation 1. So, relief would be provided to widow(er)s who, under current law, have relatively low benefit amounts. Such widow(er)s would probably have low overall economic status. Indeed, tabulations from the March 1994 CPS reveal that 35 percent of additional government expenditures under this option would be received by the poor, 58 percent by the poor and near poor (below 150 percent of poverty), and 85 percent by persons with income below the national median.

This Social Security option may be as well targeted as some proposed changes to means-tested programs, such as SSI. Tabulations from the March 1994 CPS indicate that a proposal to increase the general income exclusion of SSI from $20 to $80 would result in 41 percent of additional government expenditures reaching the poor and 65 percent reaching the poor and near poor. ¹⁵ In other words, this Social Security option targets low-income persons almost as well as an SSI proposal. According to tabulations from the CPS, the option would lift about 59,000 persons out of poverty.

Based on Social Security administrative records, the estimated annual cost of this option would be about $816 million (see Table 2). Around 1.2 million widow(er)s would receive higher benefits.

Finally, note that this proposal could be viewed as being consistent with current law. Although the widow(er)'s limit generally requires that a widow(er) benefit not exceed that of a deceased spouse, the law does make an exception by providing a floor to the RIB-LIM amount. This option augments the current-law floor. ¹⁶

Adjustments to RIB-LIM Amounts (ARLAs)

Another option would be to allow some widow(er)s to receive adjustments to RIB-LIM amounts (ARLAs). For reasons that will become clear, this option only affects persons widowed before the normal retirement age. ARLA is described in detail in the appendix, but the basics of the option can be illustrated using two examples.

Example 1: The deceased worker is a man who was born before 1938 and who filed for reduced retirement benefits at exactly age 63. He dies on his wife's 62nd birthday. She was born before 1940.

In example 1, the worker, if alive, would be receiving 0.867 of PIA because he became entitled when he was 63 years old. That is the RIB-LIM amount. The widow's initial benefit amount is based on the worker's PIA and on when the widow claims benefits. The widow's limit requires that the smaller of the two amounts be paid. If the widow claims a benefit immediately (on her 62nd birthday), the initial benefit amount is 0.829 of PIA. That amount would already be below the RIB-LIM amount, and the widow's limit would not be in effect.

Basically, there are only two ways the widow in example 1 could be affected by the widow's limit. First, she may be affected if she postpones receipt of benefits. Each month that she waits to claim benefits, the initial benefit amount rises by 0.00475 of PIA. If she postpones receipt long enough (in this case, past age 62 and 8 months), the initial benefit amount will overtake the RIB-LIM amount, and the widow's limit will go into effect. Second, if she works, the widow's limit can be triggered. Persons who file for benefits before the NRA can have them suspended because of Social Security's earnings test. Benefits are suspended if earnings exceed specified thresholds (generally, for 2000, $1 in benefits is suspended for every $2 earned above $10,080). In general, benefits are suspended, not lost. When the beneficiary reaches the NRA, his or her monthly benefit amount will be increased to reflect months that benefits were suspended. Specifically, for widow(er) benefits, for each month benefits are suspended because of the earnings test, the widow(er) benefit is increased (at the NRA) by 0.00475 of PIA. ¹⁷ But the widow(er)'s limit could prevent those increases from being paid because the amount of the widow(er) benefit cannot be above the RIB-LIM amount.

Note that the widow in example 1 can avoid the widow's limit. All she has to do is file for benefits immediately (or at least before the initial benefit amount becomes too high) and keep her earnings below the earnings test threshold (so as not to have benefits suspended). The first part of the ARLA option is to abolish the widow(er)'s limit for widow(er)s who could avoid the limit by filing early and keeping earnings low. To understand the rationale for the option, note that under current law, the widow in example 1 maximizes her lifetime Social Security widow benefits by filing no later than age 62 and 8 months and by keeping earnings below the threshold ($10,080 in 2000). If she files at age 62 and 8 months, the initial benefit amount and the RIB-LIM amount are both 0.867 of PIA, and that is the amount she will be paid. If she waits another month to file, her initial benefit amount is 0.872 of PIA, the RIB-LIM amount is 0.867 of PIA, and she will be paid 0.867 of PIA. In other words, waiting past age 62 and 8 months to claim her benefit does not increase her monthly benefit amount; it only results in her receiving benefits for fewer months over her life. A similar result would occur if the widow's earnings caused benefits to be suspended; the widow's limit would prevent the increase in the monthly benefit amount at the NRA, with the result being that the widow would simply receive benefits for fewer months over her life.

One potential problem with current law is that the widow(er)'s limit does not necessarily produce program savings, but it does encourage a widow such as the one in example 1 to file early and to keep earnings low. Suppose that widow files for benefits at age 62 and 8 months and keeps earnings below the threshold because delaying receipt of benefits or having higher earnings would lower her lifetime Social Security widow benefits. So, she would receive 0.867 of PIA for the rest of her life. Under the ARLA option, the widow(er)'s limit would be abolished, and she might choose to wait past age 62 and 8 months to file for benefits or to earn above the earnings test threshold. For example, she might choose to file at age 65 (her NRA) and receive 100 percent of PIA. That would improve the widow's well-being (because she prefers this option) but would be cost neutral from Social Security's perspective. Because the actuarial adjustments to widow(er) benefits are about fair, the cost of paying 0.867 of PIA starting at age 62 and 8 months is approximately the same as paying 100 percent of PIA starting at age 65 (see the appendix).

Example 2: The deceased worker is a man who was born before 1938 and who filed for reduced retirement benefits at exactly age 63. He dies on his wife's 64th birthday. She was born before 1940.

In example 2, note that even if the widow files immediately and keeps her earnings low, she will still be affected by the widow(er)'s limit. If she files immediately (at age 64), the initial benefit amount will be 0.943 of PIA, the RIB-LIM amount is 0.867 of PIA, and her monthly benefit amount will be 0.867 of PIA. The second part of the ARLA option does not abolish the widow(er)'s limit for such persons but does allow them to "earn" adjustments to their widow(er) benefits if they do not claim them immediately or if they have them suspended because of the earnings test.

Under the ARLA option, the widow's benefit would be increased by 0.00475 of PIA for each month before the NRA that she waits to claim widow benefits (or for each month her widow benefits are suspended because of the earnings test). So, if this widow waited 12 months after her 64th birthday (until her NRA), her benefit would increase by 12 × 0.00475 × PIA, or 0.057 of PIA. Her total benefit would equal the RIB-LIM amount (0.867 of PIA) plus the 0.057 of PIA she "earned" by waiting 12 months, for a total of 0.924 of PIA.

Again, the rationale for this part of the ARLA option is rooted in the fact that, under current law, this widow maximizes lifetime Social Security widow benefits by filing immediately (and keeping earnings below the threshold). Under this option, she could wait 12 months (until her NRA) and claim a benefit equal to 0.924 of PIA. She might prefer that approach, and it would not add to the cost of Social Security because the lifetime cost associated with paying 0.867 of PIA starting at age 64 is about the same as the cost of paying 0.924 of PIA starting at age 65 (her NRA). Note that ARLA does not abolish the widow(er)'s limit in this case. If the widow(er)'s limit was abolished, the widow in example 2 could file as soon as the worker died and receive a higher monthly amount than current law provides, which would unambiguously increase program cost. It would also be contrary to the ARLA option's cost-neutral structure.

Almost all beneficiaries under the normal retirement age receive higher monthly benefit amounts if they postpone receipt of benefits or earn above the earnings test threshold. One group that does not are persons who face the widow(er)'s limit. ARLA would change that.

One potential strength of ARLA is that, generally, it provides additional options to widow(er)s without imposing additional costs on the Social Security program. Also, it encourages behavior that policymakers may consider positive. Specifically, it encourages widow(er)s to save and to work. Under the ARLA option, widow(er)s may choose to forgo Social Security for some months in exchange for a higher monthly benefit amount at a later date (that is, save) and will be reimbursed for any benefits lost to the earnings test. A widow(er) may find the option to "save" using ARLA especially valuable because the higher Social Security benefits are paid for life and are adjusted for inflation. In short, this option allows persons widowed at an early age to "build" a higher benefit for their later years through deferred receipt of benefits and through work.

Although the ARLA option has elements of cost neutrality, it would not be completely cost neutral. One reason is that beneficiaries make filing mistakes. Consider again the widow in example 2 and assume that her labor market experience, health, or other factors are such that she cannot earn more than the threshold amount of the earnings test. Under current law, if she files for benefits immediately (at age 64), her monthly benefit amount is 0.867 of PIA. If she waits 12 months to file (at age 65), her monthly benefit amount is still 0.867 of PIA. She should not wait the 12 months to file because her monthly benefit amount does not increase and she misses out on 12 payments. However, some widow(er)s may not understand the widow(er)'s limit provision and do wait to file for benefits. The Social Security program saves money because of these filing mistakes, but under ARLA that would no longer be true. It is doubtful that policymakers consider filing mistakes by widow(er)s a good outcome even if the programs currently save money.

Another reason the ARLA option would entail costs is that some widow(er)s will want to have relatively high earnings even if they do not maximize lifetime Social Security widow(er) benefits. Consider again the widow in example 2 and assume that she has the ability to earn more than the threshold amount of the earnings test. Her lifetime Social Security benefits are maximized by filing at age 64 and keeping earnings low. Suppose, despite that, she works and has benefits suspended until age 65 (her NRA). Current law does not reimburse her for the suspended benefits, but the ARLA option would. Although the option will increase program costs, the current structure may discourage work among some widow(er)s (because suspended benefits are not restored). Policymakers will have to weigh that trade-off. Finally, the ARLA option may increase costs because of dual-entitlement provisions and because of some marriage rules that exist in the current law (see the appendix).

Tabulations from a sample of SSA's benefit records can provide information on the number of persons who might be affected by this option and on the costs associated with it. Among beneficiaries receiving widow(er) benefits for December 1998, there are 895,000 persons who were widowed before the normal retirement age and who were married to workers who received reduced benefits. Had the option always been a part of Social Security, those 895,000 persons could have "earned" benefits that exceed the maximum amounts available under current law. So, if the ARLA option was implemented, a "snapshot" of widow(er) beneficiaries in the future might indicate that around 895,000 widow(er) beneficiaries were given additional choices under Social Security (assuming the current situation is a reasonable guide to the future).

It is useful to consider what the costs and the number affected would be if current widow(er) beneficiaries were allowed to take advantage of the ARLA provisions. Current beneficiaries cannot generally change their behavior in response to such a change (they have already selected an age of filing and made decisions about earnings). However, some current beneficiaries filed or worked after their widow(er) benefits had reached their maximums, and they would have their benefits recomputed if the ARLA provisions were made available to them. Table 3 presents estimates of costs and number affected for various years (the figures for a given year measure the effects assuming that those provisions were made available to beneficiaries starting in the prior December). If the ARLA provisions applied to all widow(er) beneficiaries starting in December 1998, then 112,000 widow(er)s in 1999 would have had their benefits recomputed, and program costs would have been $69 million higher. However, a "snapshot" of widow(er) beneficiaries in the future would reveal a larger number of widow(er)s with higher benefits because some widow(er)s, first eligible after the option was implemented, would choose to file later than they would under current-law provisions. (This does not mean that real costs would be substantially higher in the future—persons who choose to file later receive higher benefits only because they receive them for fewer months.)

As noted previously, 895,000 widow(er) beneficiaries would have had additional options had the ARLA provisions always been a part of the Social Security system. Of those beneficiaries, it is known that 112,000 would have had higher benefits had ARLA always been in effect. At minimum, an estimated 15 percent of the remaining 783,000 widow(er)s would have chosen to receive benefits later than they did (or chosen to work more), and an additional 117,000 would have had higher benefits in 1999. ¹⁸ So, in the future, a "snapshot" might reveal that 229,000 (or so) widow(er)s had higher benefits because of the ARLA provisions.

ARLA would affect widow(er) beneficiaries who were widowed before the NRA and who were married to workers who claimed early retirement benefits. Among such widow(er) beneficiaries in the 1994 CPS, 14.6 percent were poor and 34.5 percent were poor or near poor. About 74 percent had income below the median income for the United States as a whole.

A Simplified ARLA (SARLA) Option

A much simpler version of the ARLA option would be to abolish the widow(er)'s limit for those who are widowed before age 62. Note that such persons can avoid the widow(er)'s limit by filing before age 62 and keeping earnings low. That is because the RIB-LIM amount can never be below 0.825 of PIA, and the initial benefit amount of a widow(er) who files before age 62 is always less than 0.825 of PIA. ¹⁹ In other words, under current law, a person who is widowed before age 62 is affected by the widow(er)'s limit only if he or she postpones receipt of benefits or earns above the earnings test threshold. SARLA would remove the limit for many widow(er)s, allowing them to build higher monthly benefit amounts for their later years through saving (that is, postponing receipt of benefits in exchange for a higher monthly amount) or through work. In many cases, the SARLA option would be cost neutral. Current law encourages early receipt of benefits (and low earnings); allowing someone to delay receipt of benefits in exchange for a higher monthly amount would not be costly because the actuarial adjustments to widow(er) benefits are about fair.

The advantage of the SARLA option relative to the ARLA option is its simplicity. Only two pieces of readily available information (date of death of the worker and date of birth of the widow(er)) are needed to implement the change. In fact, SARLA would make the program easier to administer because SSA would not have to be concerned about enforcing the RIB-LIM provisions for some beneficiaries.

A disadvantage of SARLA is that widow(er)s in similar situations will be treated differently. For example, a person widowed at age 62 would not have the RIB-LIM abolished, but someone widowed at age 61 and 11 months would (under ARLA, those similar widow(er)s would receive similar relief from the limit). Note, however, that current law treats similar widow(er)s differently and that the SARLA option would relieve some of these inequities. Consider two women who are widowed before age 62. Assume that both husbands intended to file for benefits as early as possible (that is, on their 62nd birthday). One of the husbands dies a month before he turns 62 and the other a month after he turns 62. The two women are nearly identical, but the first does not face the widow(er)'s limit whereas the second does. The SARLA option would cause Social Security to treat those similar widows the same (that is, neither would face the widow(er)'s limit). The point of this example is that given the complexity of the Social Security programs, legislative proposals almost always create some inequities and relieve others. Policymakers will have to assess whether SARLA improves or worsens overall equity and decide whether equity issues are more or less important than allowing many widow(er)s additional options under Social Security.

For reasons mentioned in the discussion of ARLA, the SARLA option will have some costs associated with it. Table 3 presents estimates of costs and number affected if SARLA was made available to current beneficiaries. In 1999, 53,000 widow(er)s would have had their benefits recomputed, and program costs would have been higher by $41 million. Had SARLA always been part of Social Security law, 479,000 persons currently receiving widow(er) benefits would have had additional options under the program. Of those beneficiaries, it is known that 53,000 would have had higher benefits had SARLA always been in effect. Assuming that 15 percent of the remaining 426,000 widow(er)s would have chosen to receive benefits later than they did (or decided to earn more), an additional 64,000 widow(er)s would have had higher benefits in 1999. So, at some point in the future, 117,000 (or so) widow(er)s might have higher benefits because of the SARLA option.

The economic status of widow(er)s who would be affected by SARLA can be approximately assessed by examining the economic status of widow(er) beneficiaries who were widowed before age 62 and who were married to workers who claimed early retirement benefits. Among such widow(er) beneficiaries in the 1994 CPS, 14.2 percent were poor and 37.1 percent were poor or near poor. About 74 percent had income below the median income for the United States as a whole.

A Proposal by Robert J. Myers

A former Chief Actuary of SSA, Robert J. Myers, has offered a proposal that would make modest adjustments to the RIB-LIM provisions. His proposal would apply in RIB-LIM cases in which workers died before the normal retirement age. Specifically, his proposal would exclude months of death (before the NRA) when determining the RIB-LIM amount. Consider an example in which a man born before 1938 filed for benefits at age 63, received benefits for 1 month, and then died. In determining the RIB-LIM amount, SSA would apply 24 months of actuarial reductions because the worker filed for benefits 24 months before the NRA (the RIB-LIM amount would be 0.867 of PIA). However, the worker received benefits for only 1 month. Under Myers's proposal, only 1 month's worth of actuarial reductions would be applied in determining the RIB-LIM amount (the RIB-LIM amount would be 0.994 of PIA, as opposed to 0.867 of PIA).

In general, Myers's proposal is designed to provide relief from the widow(er)'s limit in cases where the worker dies shortly after retirement. One rationale for the proposal is that the current structure imposes difficult choices on couples, particularly those in which the worker has health problems and may have difficulty delaying retirement. A worker with a health problem may claim benefits immediately, only to die shortly thereafter. He or she will have received retirement benefits for only a short period of time but will leave the widow(er) with a sharply reduced benefit for the rest of the widow(er)'s life. Another rationale is that the proposal would lead to a more general treatment of months in which no payment was made. A deceased worker may have had one or more nonpayment months (before the NRA) because of the earnings test. The worker, if alive, would be receiving a benefit (at the NRA and later) that reflected upward adjustments that compensated for the benefits suspended because of the earnings test. And current law would take that into account when computing the RIB-LIM amount (that is, the amount the worker would be receiving if alive). Under Myers's proposal, nonpayment months (before the NRA) because of death would also be accounted for.

Estimates of costs and number affected if Myers's proposal was made available to current beneficiaries are shown in Table 3. In 1999, 96,000 widow(er)s would have had their benefits recomputed, and program costs would have been higher by $57 million. Also, had the proposal always been part of the law, some widow(er)s would have had additional options under the program (just as they would have under ARLA and SARLA). I estimate that an additional 19,000 widow(er)s would have postponed receipt of benefits (or had higher earnings) had Myers's proposal always been part of the law. So, a "snapshot" in the future might reveal 115,000 widow(er)s with higher benefits under the proposal.

Myers's proposal would affect widow(er) beneficiaries in cases in which workers received reduced retirement benefits and died before the NRA. Among such widow(er)s, 13.6 percent were poor and 29.7 percent were poor or near poor. About 71 percent had income below the median income for the United States as a whole.

Conclusion

If Congress was to abolish the RIB-LIM, it would be the most substantial change to widow(er) benefits in three decades. About 2.8 million widow(er)s would receive increased benefits, and the program would pay out an additional $3.1 billion a year. Most of the additional expenditures would not be received by the poor and the near poor. A proposal to increase the widow(er)'s limit, using the average PIA among retired workers, is more successful at targeting increased expenditures to low-income widow(er)s: a majority of expenditures reach the poor and the near poor. Overall, about 1.2 million widow(er)s would be helped by the option, and costs would be about $816 million a year.

Social Security faces a long-run actuarial deficit and, in recent years, Congress has taken a modest approach to improving benefits for widow(er)s. For example, the 1983 amendments allowed some persons who were widowed before age 60 to receive a more favorable computation of the PIA. About 500,000 widow(er)s have higher benefits because of that change, and annual program costs are higher by about $200 million (Chaplain 1999b). The RIB-LIM provisions can be changed in a modest way. The ARLA and SARLA options are inexpensive, and either one would give hundreds of thousands of widow(er)s additional options under Social Security. Myers's proposal is also inexpensive and would address some inequities in the current law.

Appendix: The Option to Adjust RIB-LIM Amounts (ARLA)

In the following description, it will be useful to keep in mind that widow(er)s have to "earn" ARLAs by forgoing widow(er) benefits for 1 or more months before the normal retirement age (NRA), much in the same way actuarial adjustments are earned. The ARLA option can be described by the following steps.

1. Calculate the proportion of the primary insurance amount (PIA) that the worker would be paid if he or she was still alive. For this proposal, count the number of months in the period that begins with the month the worker turns 62 and that ends with the month before the month the worker first becomes entitled to retirement insurance benefits (RIB). Add to that the number of months (before the NRA and on or after the worker first becomes entitled to RIB) in which the worker had benefits fully or partially suspended because of the earnings test of Social Security. If the worker was born before 1938, multiply the total number of months by 0.00556 and add to 0.8 to get the proportion of PIA the worker would be paid. If the worker was born in or after 1938, use the figures that apply under the 1983 amendments. An additional requirement is that this proportion can never be less than 0.825.
2. Calculate the proportion of PIA that the widow(er) would receive if he or she was entitled to widow(er) benefits "as soon as possible." For this option, if the worker died in or before the month the widow(er) reaches age 60, the proportion would be equal to 0.715. Otherwise, determine the number of months in the period that begins with the month the widow(er) reaches age 60 and that ends with the month before the month the worker dies. If the widow(er) was born before 1940, multiply the number by 0.00475 and add to 0.715 to get the proportion of PIA. If the widow(er) was born in 1940 or later, perform a similar calculation but use the appropriate actuarial adjustment that applies to the widow(er) under the 1983 amendments.
3. If the amount in (2) is less than that in (1), calculate the widow(er) benefit as if the current-law RIB-LIM provisions were not applicable.
4. If the amount in (2) is greater than that in (1), award an ARLA for every month that is (a) after (or the same as) the month of the worker's death and before the month the widow(er) reaches the normal retirement age and (b) a month for which entitlement as a widow(er) has not been established or a month for which entitlement has been established but widow(er) benefits have been partially or fully suspended because of the earnings test. Each ARLA would be equal to the actuarial adjustment that applies to the widow(er) (0.00475 for those born before 1940) multiplied by the PIA. The widow(er)'s benefit would be equal to his or her current-law amount plus the ARLAs. ARLAs based on suspensions because of the earnings test would not be added to current-law amounts until the widow(er) reaches the NRA. Otherwise, ARLAs would be added upon entitlement.

Multiplying the proportion in step (1) by the PIA yields the RIB-LIM amount, or the maximum amount the widow(er) can receive. Multiplying the proportion in step (2) by the PIA yields the amount the widow(er) can receive if he or she becomes entitled at age 60 or in the month of the worker's death (if after the widow(er)'s age 60), ignoring the RIB-LIM feature. If that amount is below the RIB-LIM amount, the ARLA option simply abolishes the RIB-LIM and allows a widow(er) to earn actuarial adjustments without ever "bumping" into the RIB-LIM amount. If the amount based on step (2) is already above the RIB-LIM amount, a widow(er) will receive adjustments to benefits if he or she does not file immediately. Although I have referred to RIB-LIM "amounts" in this discussion, the ARLA option is based on the proportions outlined in the four steps. ²⁰

Fairness of Actuarial Adjustments

Are the actuarial adjustments for reduced widow(er) benefits, which would underlie the ARLA option, approximately cost neutral? Consider a widow born before 1940 who files on her 62nd birthday (for this discussion, assume that her deceased husband did not receive reduced benefits and the RIB-LIM provisions are therefore not relevant). Assume, for simplicity, that there is no inflation, so that the PIA is constant over the widow's lifetime (no cost-of-living adjustments are made). Social Security will pay her [(24 × 0.00475) + 0.715] × PIA, or 0.829 × PIA, for the rest of her life. The present discounted costs to the program of this benefit stream—measured at the time of the widow's 62nd birthday using a rate of interest equal to 3 percent and using survival probabilities for all women—is 154.92 × PIA.²¹ If the widow was to claim her benefit on her 63rd birthday instead, she would receive a benefit equal to [(36 × 0.00475) + 0.715] × PIA, or 0.886 × PIA, from age 63 until the end of her life. The present discounted costs to the program of this benefit stream, measured at the time of the widow's 62nd birthday, would be 154.94 × PIA. The costs associated with age 64 and age 65 filing would be 154.04 × PIA and 152.28 × PIA. All these amounts are close, which indicates that the 0.00475 adjustment is approximately cost neutral. For widow(er)s born in 1940 or later, the actuarial adjustment is very gradually getting smaller. So, over the near term, the adjustment should still be approximately fair. For widow(er)s born in 1962 or later, the adjustment will be 0.00339. That may still be about fair because life expectancy should be higher for later birth cohorts.

Costs Due to Dual Entitlement

Widow(er)s who have worked in covered employment may be eligible for both a retirement benefit and a widow(er) benefit. Some widow(er)s will claim a retirement benefit first and then, sometime later, will claim a widow(er) benefit, thus becoming dually entitled (about 42 percent of widow(er) beneficiaries are dually entitled). The widow(er) will receive the retirement benefit in full plus the difference between the widow(er) benefit and the retirement benefit. Consider an "average" dual-entitlement case. The husband's PIA is equal to $943.30 and the wife's PIA based on her own work is $588.20. Those are the average PIA amounts for men and women for December 1997 (U.S. Social Security Administration 1998, p. 210). Assume that both claimed worker benefits on their 62nd birthday and the woman was widowed on her 63rd birthday. The RIB-LIM amount is 0.825 × 943.3, or 778.2. (The husband would be receiving 0.8 × 943.3 if he was alive, but, as noted earlier, the RIB-LIM amount can never be below 0.825 × PIA.) Under current law, the widow should claim the benefit as soon as possible (on her 63rd birthday) because it is at its maximum. When she does, her retirement benefit will be equal to (0.8 × 588.2), or 470.6, and her widow benefit will be (778.2 − 470.6), or 307.6.

Now, if the ARLA option was implemented and she decided to delay receipt of widow benefits, she would forgo only a partial widow benefit (307.6) but would receive an ARLA based on the full widow benefit (the husband's PIA). That would be costly to Social Security relative to current law. The present discounted cost of paying the current-law worker and widow benefits to this widow, measured at her 63rd birthday, is $141,664. If the ARLA option was implemented and she chose to wait until age 65 to claim her widow benefit, the cost would be $151,472, which is about a 7 percent increase. So, for dually entitled widow(er)s who take advantage of ARLA, there will be a modest increase in program costs.

Costs Due to Marriage Rules

A person who is widowed before age 60 and who remarries before that age may generate costs under the ARLA option. Such a person cannot collect a widow(er) benefit on the prior spouse's work record until the remarriage ends (through divorce or through the death of the most recent spouse). Consider a woman who is widowed at age 55 when her first husband dies. Suppose she remarries before age 60 and is widowed by her second husband at age 65. Only at age 65 can she file for widow benefits on her first husband's record. If her first husband filed for a reduced benefit, she could not receive the full PIA at age 65 because of the RIB-LIM feature of current law. Under ARLA, she could receive the full PIA because her first husband was deceased from the time she was age 60 until she was age 65, and she did not establish entitlement on his work record. Thus, in that case, the ARLA option would generate costs relative to current law.