Analysis of Social Security Proposals Intended to Help Women: Preliminary Results

by Sharmila Choudhury, Michael V. Leonesio, Kelvin R. Utendorf, Linda Del Bene, and Robert V. Gesumaria
ORES Working Paper No. 88 (released January 2001)

One aspect of the current debate about changing the Social Security program concerns how new rules might affect elderly women, many of whom have low income. This paper examines three possible changes: (1) a reduction in spousal benefits combined with a change in the computation of the survivor benefit, (2) a redefined minimum benefit, and (3) a 5 percent increase in benefits for persons aged 80 or older. The paper assesses the cost, distributional consequences, and antipoverty impact of each option.


The authors are with the Division of Economic Research, Office of Research, Evaluation, and Statistics, Office of Policy, Social Security Administration

Acknowledgments: The author would like to thank Joyce Manchester, David Pattison, Gene Steuerle, David Weaver, and especially Ben Bridges, for helpful comments and advice.

Working papers in this series are preliminary materials circulated for review and comment. The findings and conclusions expressed in them are the authors' and do not necessarily represent the views of the Social Security Administration.

Introduction

One aspect of the current debate about changing the Social Security program concerns how new rules might affect elderly women, many of whom have low income. This paper examines three possible changes: (1) a reduction in spousal benefits combined with a change in the computation of the survivor benefit, (2) a redefined minimum benefit, and (3) a 5 percent increase in benefits for persons aged 80 or older. The paper assesses the cost, distributional consequences, and antipoverty impact of each option.

Study Data and Methodology

Our first task was to construct a database that would support the analysis of a variety of proposals that affect the financial well-being of women aged 62 or older. An attractive candidate is the Survey of Income and Program Participation (SIPP), which offers a number of advantages for the analysis of income changes associated with tax and transfer policies. Among the many strengths of the SIPP are accuracy in classifying and reporting income sources (that is, SIPP records monthly income for 65 separate categories of cash income and 7 in-kind transfer programs); monthly detail for examining interactions among labor force participation, transfer programs, and living arrangements; and good measures of assets, liabilities, and taxes. The sample sizes for both the 1992 and 1993 SIPP panels consist of approximately 62,000 respondents. The overlapping design feature of the panels has meant that for any calendar year, two SIPP panels are active, allowing the investigator the option of doubling the sample size by pooling observations from consecutive panels.1 Accordingly, the core of our database is the SIPP information for calendar year 1993 provided by respondents from both the 1992 and 1993 panels.

Because the central interest is to examine the distributional consequences of changes in the Social Security program, we merged information on benefits and earnings from the Social Security Administration's (SSA's) administrative records with the public-use survey data for the SIPP respondents. For persons aged 62 or older, match rates for record data on earnings (SSA's Summary Earnings Records, or SER) were 90 percent and for benefits (Master Beneficiary Record, or MBR), 79 percent.2 Further deletions from the file, made mostly for reasons of miscellaneous data inconsistencies, reduced the final sample to 12,296 unweighted cases.3 We then reweighted the sample to conform to independent Census Bureau estimates of 1993 population totals. In the end, the database used in this study consists of the rich set of demographic and socioeconomic characteristics gathered from the SIPP interviews, SSA's record of benefits paid in 1993, plus the lifetime earnings records of respondents as maintained by the agency and used in the computation of benefits. Basic demographic statistics for the beneficiary subsample of the extract, categorized by beneficiary type, are displayed in Table 1.

Table 1. Descriptive characteristics of the sample beneficiary population aged 62 or older
Characteristic Women Men
Retired worker
Average age (years) 72.6 72.1
Percent married 49.5 77.4
Percent black 9.8 6.7
Percent Hispanic 5.1 4.2
Average total family income (dollars) 26,840 31,948
Disabled worker
Average age (years) 63.0 63.1
Percent married 69.2 69.8
Percent black 17.3 7.8
Percent Hispanic 0.0 7.6
Average total family income (dollars) 23,761 28,064
Spouse
Average age (years) 70.3 74.4
Percent married 95.6 92.1
Percent black 3.8 7.7
Percent Hispanic 4.2 9.5
Average total family income (dollars) 32,874 36,595
Survivor
Average age (years) 74.5 74.7
Percent married 3.6 8.1
Percent black 9.0 17.8
Percent Hispanic 3.2 6.2
Average total family income (dollars) 19,668 16,934
SOURCE: 1992 and 1993 SIPP, CPS look-alike file matched to SSA earnings and benefit records. Results weighted using SIPP sample weights adjusted to represent the U.S. noninstitutionalized population, December 1993. Dual beneficiaries are classified here by their secondary benefit (either spouse or survivor). The spouse category includes divorced spouses.

With the construction of the database completed, our basic approach is to analyze "current year" effects of proposed changes. That is, we begin with the actual distribution of income and benefits in 1993 documented in the data set, recompute benefit amounts using revised program rules, and calculate the changes in the distribution of benefits and income.

That type of analysis, although reasonably straightforward, has some noteworthy limitations. Most problematic is that many proposed changes in Social Security would take effect years after legislation is enacted and would often be gradually phased in. Rather than simply apply a different set of Social Security program rules to the population in the past, it would be better to compare the distribution of benefits and income in the future under both current and alternative policies. A credible analysis requires a simulation model that projects the relevant set of economic variables that ultimately determine the impact of any changes.

We present a more limited analysis that examines how the financial circumstances of the retirement-age population would have looked had different program rules existed in 1993. No behavioral changes are modeled even though some options plausibly entail incentives to alter the timing of benefit entitlement, late-life labor supply, and other economic choices. The analysis also assumes no offsetting effects from the Supplemental Security Income (SSI) or other transfer programs that contribute to income of the elderly. We are in the process of expanding the model to encompass effects of Social Security changes on SSI.4 Furthermore, our results are dependent on the economic histories and circumstances of the population in 1993. That limitation could be important with regard to program changes that affect women in light of the secular changes in lifetime earnings patterns and marital histories that have been well-documented. So, for example, if 30 years from now women born after 1970 are much less likely to have been in marriages that last 10 years, and if they exhibit stronger lifetime labor force attachment than earlier generations of women, then the basis for their Social Security benefits under current rules would be markedly different than that for retirees in 1993.

Those caveats aside, we believe that our approach is instructive and has its own merits. Our procedures are simple, reasonably straightforward to interpret, and can be easily adapted to examine a wide range of program changes. Although more complex microsimulations have some advantages, they also introduce problems associated with forecasting errors. Because our methodology does not project the effects of policy changes, it is likely to provide better indications of the consequences of changes that are fully implemented fairly quickly than of those that are slowly phased in over many years.

Analysis of Policy Options

To provide a context for judging the effects of alternative Social Security policies, we begin by presenting data for the actual distribution of Social Security benefits among the population aged 62 or older in 1993. Table 2 displays summary statistics by sex and age for December 1993, as represented in the exactly matched SIPP extract.

Table 2. U.S. population aged 62 or older, by sex, age group, and beneficiary status as of December 1993
Sex and age group Total Beneficiary status Not
receiving
benefits
Primary Secondary
Number Average
benefit
amount
(dollars)
Number Average
benefit
amount
(dollars)
Men 62 or older 15,605,785 13,259,751 782 152,812 480 2,193,221
62–64 2,795,283 1,599,291 743 7,738 503 1,188,253
65–69 4,395,405 3,812,905 779 33,765 444 548,735
70–74 3,653,514 3,425,866 758 45,175 524 182,473
75–79 2,522,279 2,350,909 850 28,150 341 143,220
80 or older 2,239,304 2,070,780 781 37,984 557 130,540
Women 62 or older 21,210,431 6,600,160 573 12,269,514 565 2,340,757
62–64 3,209,615 840,620 506 1,433,912 474 935,083
65–69 5,339,922 1,836,766 546 3,022,383 513 480,773
70–74 4,789,932 1,414,713 544 3,042,603 575 332,616
75–79 3,583,171 1,168,907 625 2,195,796 623 218,468
80 or older 4,287,791 1,339,153 636 2,574,821 614 373,817
SOURCE: 1992 and 1993 SIPP, CPS look-alike file matched to SSA earnings and benefit records. Results weighted using SIPP sample weights adjusted to represent the U.S. noninstitutionalized population, December 1993. Primary beneficiaries are defined as retired workers and disabled workers. Secondary beneficiaries are all others. Those who are entitled to both primary and secondary benefits are classified here as secondary beneficiaries.

Beneficiaries are categorized as either primary or secondary. Primary beneficiaries receive benefits solely on the basis of their own earnings in covered employment, whereas secondary beneficiaries receive benefits on the basis of their relationship to a primary beneficiary. In this analysis, primary beneficiaries are essentially retired workers who receive benefits under the Old-Age and Survivors program and disabled workers who are Disability Insurance program beneficiaries.

The important secondary beneficiary groups in our sample are aged spouses, divorced spouses, and aged survivors. Many older women, and far fewer men, are dually entitled beneficiaries. That is, the amount of a secondary benefit for which they qualify exceeds the amount of a primary benefit for which they also qualify on the basis of their own work record. In such cases, the primary benefit is paid in full, and the secondary benefit is reduced by the amount of the primary benefit. So, in effect, the dually entitled beneficiary receives the secondary benefit amount. Whether to classify dually entitled beneficiaries as primary or secondary beneficiaries is a practical question with different answers depending on the purpose to be served. Here we classify them as secondary beneficiaries for reasons suggested by Weaver (1997). Changes in Social Security law that would increase primary benefits often have little effect on the benefits received by dually entitled persons, because the amount of their secondary benefit is reduced by the amount that their primary benefit rises. On the other hand, policies that increase secondary benefits would affect benefit amounts paid to dually entitled beneficiaries.

Option 1. Increase Survivor Benefits and Reduce Spouse Benefits.

The first policy change that we examine reduces the dependent-spouse benefit from one-half to one-third of the retired-worker's benefit and increases the survivor benefit to three-quarters of the couple's combined benefit under the new law. Like the similar policy changes outlined in the 1994–96 Advisory Council Report (1997), this option specifies that the benefit of a surviving spouse would be determined by the highest of (1) his or her own worker benefit, (2) the deceased spouse's worker benefit, and (3) 75 percent of the couple's combined benefit under the new law.5 The couple's benefit is the sum of the deceased spouse's worker benefit and the higher of the living spouse's worker benefit or spousal benefit. A key goal of the option is to increase benefits paid to survivors of two-earner couples.

Three main categories of beneficiaries are potentially affected by the option. Table 3 summarizes how benefits would change under this option. First, spousal beneficiaries would face benefit reductions of two types. Many spouses would see their benefit amount fall by one-third, but some dually entitled spouses would see their benefits fall by less than one-third because their own monthly benefit amount (MBA) exceeds one-third the value of the MBA of their husband or wife. The latter group would switch from being dually entitled spousal beneficiaries to retired-worker-only. The option would not affect spouses who were currently married and claimed retired-worker benefits.

Table 3. Percentage change in benefits under Option 1 from current law, by beneficiary type
Ratio of wife's MBA
to husband's MBA (R)
Beneficiary status under current law
Spouse* Survivor
(widow)
Retired worker
(widow)
R<0.33 -33.3 0 ***
0.33<=R<0.5 -33.3 to near 0** 0 to near +12.5 ***
0.5=R<1 *** 12. to near +50 ***
1<=R<3 *** *** +50 to near 0**
3<=R *** *** 0
* Includes married spouses and divorced spouses.
** Individuals switch to a different category under Option 1.
*** Not applicable—these designations do not yield the highest entitled benefit.

Second, some survivor beneficiaries would receive increased benefits, but the benefit amount for survivors whose MBAs were one-third or less than the MBA of their deceased spouse would not change. In cases in which a dually entitled survivor beneficiary had an MBA that was greater than one-third but did not exceed the MBA of the deceased spouse, the survivor benefit would increase by as much as one-half, the upper limit occurring when the MBAs were equal.

Third, some nonmarried retired-worker-only beneficiaries could convert to dually entitled survivors. For them, the increased survivor benefit would then exceed the previously dominant retired-worker benefit to which a widow or widower had been entitled. That would occur when the retired-worker's MBA was one to three times the MBA of the deceased spouse. The proportional increase in benefits could range from near zero (as the ratio of the MBAs approached 3) to one-half (when MBAs were equal). Survivors of two-earner couples with similar earnings histories might fare particularly well under such an option.

Table 4 summarizes the distribution of women's benefits in 1993, categorized by current marital status, beneficiary type, and age (five age groups: 62–64, 65–69, 70–74, 75–79, and 80 or older).6 Women are categorized by marital status (married, nonmarried) as reported in the survey data and by beneficiary type as determined from SSA's benefit records. We show married widow beneficiaries as a separate subgroup because widows who remarry after the age of 60 can claim survivor benefits on a deceased husband's earnings record. Similarly, divorced spouses who are currently married (having remarried after age 60) can also claim benefits on earnings records of ex-husbands; however, there were too few cases to permit us to show that subgroup separately in the table. The table shows, for each benefit category, the aggregate annual benefits paid to the group, the average monthly benefit amount, and the group poverty rate. Also shown are total annual benefits and average benefits by marital status, age group, and type of highest benefit.7 In the presentation of results in Tables 4 through 6, however, we hold beneficiary type constant. For example, if a dually entitled beneficiary switches to retired-worker-only under the proposal, we compute the higher benefit but report the new amount under the original or current-law beneficiary designation. By doing so, we are better able to track changes in the financial circumstances of each beneficiary type as we move from the baseline to full implementation of the option. Table 4 presents statistics for current law and Option 1.

Table 4. Effects of Option 1 on women's benefits, by marital status, beneficiary type, age group, and income quintile, 1993
Benefit category Age group Income quintile
Current law Option 1 Current law Option 1
62–64 65–69 70–74 75–79 80+ All 62–64 65–69 70–74 75–79 80+ All 1st 2nd 3rd 4th 5th All 1st 2nd 3rd 4th 5th All
Total benefits paid (billions of dollars)
Nonmarried
Retired worker 0.96 5.18 4.49 5.39 8.41 24.43 0.96 5.18 4.49 5.39 8.41 24.43 6.09 6.06 4.30 4.39 3.58 24.43 6.09 6.06 4.30 4.39 3.58 24.43
Widow 3.93 10.08 13.44 12.01 16.45 55.90 4.03 10.55 14.32 13.10 17.66 59.65 17.80 15.06 9.50 7.75 5.78 55.90 18.96 16.12 10.17 8.28 6.12 59.65
Divorced spouse 0.16 0.29 0.21 0.16 a 0.83 0.12 0.21 0.15 0.12 a 0.60 0.36 0.15 0.16 0.10 0.06 0.84 0.26 0.11 0.12 0.07 0.05 0.61
Married
Retired worker 2.92 6.86 4.74 3.37 1.81 19.71 2.92 6.86 4.74 3.37 1.81 19.71 1.59 2.94 4.48 5.50 5.20 19.71 1.59 2.94 4.48 5.50 5.20 19.71
Widow 0.10 0.42 0.45 0.66 0.41 2.04 0.10 0.44 0.49 0.74 0.44 2.21 0.19 0.35 0.38 0.46 0.65 2.04 0.21 0.38 0.40 0.53 0.70 2.21
Spouse 3.45 7.58 6.78 3.55 1.99 23.35 2.43 5.38 4.79 2.48 1.37 16.45 1.50 4.56 6.01 5.19 6.09 23.35 1.06 3.24 4.24 3.66 4.25 16.45
Total benefits 11.52 30.40 30.12 25.14 29.06 126.25 10.56 28.61 28.99 25.20 29.70 123.06 27.55 29.12 24.82 23.40 21.37 126.26 28.18 28.86 23.71 22.43 19.91 123.07
Average monthly benefits (dollars)
Nonmarried
Retired worker 537 596 586 656 670 628 537 596 586 656 670 628 472 652 696 712 827 628 472 652 696 712 827 628
Widow 662 703 719 735 664 698 679 736 766 802 713 745 593 730 764 793 803 698 632 781 817 846 850 745
Divorced spouse 330 364 403 396 a 371 249 267 284 291 a 271 331 337 470 403 570 373 241 251 341 272 459 272
Married
Retired worker 482 513 510 581 514 518 482 513 510 581 514 518 387 471 523 531 592 518 387 471 523 531 592 518
Widow 673 596 688 755 617 670 673 630 744 851 668 729 617 573 616 694 781 670 651 627 649 796 837 729
Spouse 361 381 413 411 384 391 254 270 292 287 265 276 271 365 400 402 444 391 192 260 283 283 310 276
Average monthly benefit 480 526 565 624 623 569 440 495 544 626 636 554 510 586 575 583 611 569 522 580 549 559 569 554
Poverty rate (percent)
Nonmarried
Retired worker 0.24 0.15 0.16 0.20 0.16 0.17 0.24 0.15 0.16 0.20 0.16 0.17 -- -- -- -- -- -- -- -- -- -- -- --
Widow 0.13 0.15 0.14 0.15 0.18 0.16 0.13 0.13 0.13 0.12 0.14 0.13 -- -- -- -- -- -- -- -- -- -- -- --
Divorced spouse 0.20 0.27 0.35 0.45 a 0.30 0.27 0.38 0.42 0.53 a 0.40 -- -- -- -- -- -- -- -- -- -- -- --
Married
Retired worker 0.01 0.03 0.03 0.04 0.06 0.03 0.01 0.03 0.03 0.04 0.06 0.03 -- -- -- -- -- -- -- -- -- -- -- --
Widow 0 0 0.10 0 0.11 0.04 0 0 0.10 0 0.11 0.04 -- -- -- -- -- -- -- -- -- -- -- --
Spouse 0.02 0.02 0.04 0.01 0.05 0.03 0.03 0.03 0.06 0.03 0.07 0.04 -- -- -- -- -- -- -- -- -- -- -- --
NOTE: -- = not applicable.
a. Sample size too small.

Under current law, of the six benefit categories, the largest share of women's benefits is paid to nonmarried widow beneficiaries. That is true across all age groups, with the share paid to widows mostly rising as age increases. The pattern is not surprising given that women generally live longer than men and may switch from other benefit categories to the survivor category when their husband dies. Average monthly benefits received by women as survivor beneficiaries are generally higher than the average amounts received by other types of female beneficiaries. Although many women have work histories that qualify them to receive retired-worker benefits on the basis of their own earnings, those lifetime earnings are typically lower than those of their husband. As a consequence, when widowed they claim a survivor benefit based on the husband's earnings. Because so many women claim survivor benefits under current law, one might expect that proposals to change survivor benefits would have a potentially large effect on the distribution of women's Social Security benefits. As shown in Table 4, the share (in dollar amounts) of spousal benefits paid decreases as women grow older and their benefits are converted to survivor benefits, at least at ages 65 and older. Proposals that would change the spousal benefit also have the potential to significantly affect the distribution of benefits for women under age 75 but would be expected to have a much smaller impact on the distribution of benefits for women aged 75 or older.

Two noteworthy points can be made regarding poverty rates 8 under current law. First, as shown in Table 4, there is a striking difference between the poverty rates of nonmarried versus married women that holds across all age groups. For all benefit categories, married women are much less likely to reside in poor families than are nonmarried women.9 That difference in poverty incidence has long been documented and has been the impetus behind proposals that would shift benefits from married to nonmarried women. Second, the poverty rate of female divorced-spouse beneficiaries is markedly higher than for all other beneficiary subgroups among both married and nonmarried persons except for nonmarried retired workers aged 62–64. Even though their average monthly benefits are roughly similar, divorced-spouse beneficiaries typically have smaller amounts of other financial resources than do married spouses. Those beneficiaries number about 200,000 and represent a little over 1 percent of the total female beneficiary population. The higher divorce rates of the past several decades will continue to make that group a growing proportion of future beneficiaries. Changes that differentially affect married versus nonmarried female beneficiaries or that have an impact on divorced-spouse beneficiaries have the potential to substantially alter the distribution of benefits among women.

Consistent with expectations, the simulation of Option 1 shows improved financial circumstances of women receiving survivor benefits, with decreased benefits for those who originally received spousal benefits. Total annual benefits paid to widow beneficiaries, both married and nonmarried, increase by 7 percent under the option. Total benefits for spouses decline by 30 percent; 28 percent for divorced spouses. Overall, Option 1 decreases the total annual benefits paid to women by about $3.2 billion.10 Mean monthly benefits increase under the option for widow beneficiaries and decrease for those receiving benefits as a married or divorced spouse.

The induced changes in family poverty rates shown in Table 4 are generally in accordance with expectations. For every age group except those aged 62–64, the poverty rate for nonmarried widow beneficiaries decreases as a result of Option 1.11 Poverty rates fall by about 20 percent for unmarried widows aged 75 or older, but rates for married widow beneficiaries of all ages are unchanged. Married retired-worker beneficiaries should not be much affected by this option because its provisions do not have a direct impact on them. There are, however, a few instances of married female retired workers with husbands who are spouse beneficiaries, whose potential reduction in benefits would cause family income to fall. The infrequent occurrence of such cases, however, leaves the poverty rates of female married retired workers unaffected.

The poverty rate for women who receive spousal benefits is higher under Option 1. For the 62–64 and 65–69 age groups, the average benefit reduction for married spouses is usually small relative to total family income, and therefore the option causes a small increase in the poverty rate for those younger beneficiaries. By contrast, older spouse beneficiaries experience a larger increase in poverty rates. The reason is that a decline in their spousal benefit constitutes a proportionately larger reduction in total family income because their older spouses often have lower incomes.

The greatest impact on poverty rates occurs for divorced-spouse beneficiaries. For all ages, poverty rates increase by roughly 20 percent or more. Under the option, almost half of divorced women aged 70 or older are officially poor. Discussions of policies that affect spousal benefits vis-à-vis survivor benefits seldom explicitly address the treatment of divorced spouses. Clearly, proposals similar to Option 1 that do not entail additional protections targeted toward older divorced women could have adverse consequences for them.

The lower panel in Table 4 shows how changes in Social Security benefit rules affect women aged 62 or older in different parts of the income distribution. The analysis conveys information about an option's effectiveness in improving the economic well-being of women in lower-income families. The quintile groups under current law were derived by rank-ordering the ratio of total family income to the poverty line for the entire population of men and women aged 62 or older. As one would expect, women are disproportionately represented in the lower end of the income distribution in 1993. Table 4, which focuses exclusively on women, therefore has unequal numbers of observations in the five quintiles. The table shows how benefits change for individuals in each of the current-law quintiles. Therefore, when comparing the proposal with current law, it is important to keep in mind that the table varies the type and amount of benefit that individuals receive but not their income quintile.12

The current-law results in Table 4 show that, generally, the aggregate benefits paid to women in the nonmarried benefit categories decline from the first through the fifth quintiles and follow the reverse pattern for women in the married benefit categories. There are many more nonmarried women at the lower end of the income distribution relative to married women. The current-law results also show, within benefit categories, a positive relationship between total family income and average benefit amount.

Option 1 redistributes benefits from both divorced- and married-spouse beneficiaries to widow beneficiaries. Table 4 shows a large decline in mean benefits for both spouse benefit categories and a large increase in mean benefits for individuals in the widow benefit category. How effective is Option 1 in targeting those widows at the bottom end of the income distribution? Nonmarried widow beneficiaries in the first quintile receive a mean monthly benefit increase equal to $39 (6.6 percent). But widow beneficiaries in all other quintiles receive roughly similar larger increases in mean monthly benefits in percentage terms. Married widow beneficiaries exhibit more variation in their new benefits, ranging from a 5.5 percent increase in the first quintile to a 15 percent increase for those in the fourth quintile. Spouse beneficiaries—both current and divorced—incur substantial losses under Option 1. The smaller average dollar losses occur in the first quintile, with married spouses losing $79 per month and divorced spouses losing $90 per month. The loss across all the quintiles is quite high in percentage terms. On average, married-spouse beneficiaries see a drop of 30 percent in their benefits across quintiles. Divorced-spouse beneficiaries see a slightly lower drop of about 27 percent except in the two highest quintiles. In sum, the results show that Option 1 improves the economic circumstances of widow beneficiaries but does not necessarily favor those in the lower quintiles of the income distribution.

Option 2. Redefine a Minimum Benefit.

The second option that we examine would institute a new minimum benefit.13 A minimum MBA equal to 60 percent of the poverty threshold would be guaranteed to a retired worker with 80 quarters of coverage. That percent would increase by 2 percentage points for every additional year of coverage up to a maximum of 40 years (160 quarters of coverage), at which point the income guarantee would equal the poverty line. The 1993 poverty threshold for a single person aged 65 or older was $6,930.

The option targets additional benefits toward workers who have substantial labor force experience but in low-wage employment. The top panel of Table 5 14 shows that benefits increase for most benefit categories for all age groups, resulting in increased program costs of about $1.3 billion.15 The table also shows that the percentage increases in aggregate benefits paid are relatively small across all age groups, ranging from a low of about 0.4 percent for the group aged 80 or older to a high of about 1.5 percent for those aged 65–69. The magnitude of those increases is dictated by the criteria for receiving a higher MBA under Option 2: a worker has to have at least 80 quarters of coverage and must have a current-law benefit that is smaller than the poverty threshold. Many workers qualify for a minimum benefit that is less than the current-law benefit, which leaves their position unchanged.

Table 5. Effects of Option 2 on women's benefits, by marital status, beneficiary type, age group, and income quintile, 1993
Benefit category Age group Income quintitle
Current law Option 2 Current law Option 2
62–64 65–69 70–74 75–79 80+ All 62–64 65–69 70–74 75–79 80+ All 1st 2nd 3rd 4th 5th All 1st 2nd 3rd 4th 5th All
Total benefits paid (billions of dollars)
Nonmarried
Retired worker 0.96 5.18 4.49 5.39 8.41 24.43 0.98 5.29 4.63 5.47 8.41 24.77 6.09 6.06 4.30 4.39 3.58 24.43 6.29 6.12 4.34 4.42 3.60 24.77
Widow 3.93 10.08 13.44 12.01 16.45 55.90 3.94 10.15 13.50 12.09 16.55 56.23 17.80 15.06 9.50 7.75 5.78 55.90 18.00 15.12 9.53 7.77 5.81 56.23
Divorced spouse 0.16 0.29 0.21 0.16 a 0.83 0.17 0.30 0.21 0.17 a 0.85 0.36 0.15 0.16 0.10 0.06 0.84 0.38 0.16 0.16 0.10 0.06 0.87
Married
Retired worker 2.92 6.86 4.74 3.37 1.81 19.71 3.01 7.08 4.90 3.43 1.81 20.23 1.59 2.94 4.48 5.50 5.20 19.71 1.68 3.04 4.62 5.62 5.28 20.23
Widow 0.10 0.42 0.45 0.66 0.41 2.04 0.10 0.42 0.45 0.66 0.41 2.04 0.19 0.35 0.38 0.46 0.65 2.04 0.19 0.35 0.38 0.46 0.65 2.04
Spouse 3.45 7.58 6.78 3.55 1.99 23.35 3.46 7.61 6.80 3.55 1.99 23.41 1.50 4.56 6.01 5.19 6.09 23.35 1.53 4.58 6.01 5.19 6.10 23.41
Total benefits 11.52 30.40 30.12 25.14 29.06 126.25 11.66 30.84 30.50 25.37 29.17 127.53 27.55 29.12 24.82 23.40 21.37 126.26 28.08 29.36 25.03 23.58 21.50 127.55
Average monthly benefits (dollars)
Nonmarried
Retired worker 537 596 586 656 670 628 549 609 603 666 670 637 472 652 696 712 827 628 487 658 702 717 832 637
Widow 662 703 719 735 664 698 664 708 722 740 668 702 593 730 764 793 803 698 600 733 766 795 806 702
Divorced spouse 330 364 403 396 a 371 346 371 412 414 a 383 331 337 470 403 570 373 344 360 473 403 570 384
Married
Retired worker 482 513 510 581 514 518 496 530 526 591 514 531 387 471 523 531 592 518 407 486 539 543 601 531
Widow 673 596 688 755 617 670 673 596 688 755 617 670 617 573 616 694 781 670 617 573 616 694 781 670
Spouse 361 381 413 411 384 391 362 383 414 412 384 392 271 365 400 402 444 391 276 367 400 402 444 392
Average monthly benefit 480 526 565 624 623 569 486 534 572 630 625 575 510 586 575 583 611 569 520 591 580 587 615 575
Poverty rate (percent)
Nonmarried
Retired worker 0.24 0.15 0.16 0.20 0.16 0.17 0.19 0.12 0.16 0.19 0.16 0.16 -- -- -- -- -- -- -- -- -- -- -- --
Widow 0.13 0.15 0.14 0.15 0.18 0.16 0.13 0.14 0.14 0.14 0.17 0.15 -- -- -- -- -- -- -- -- -- -- -- --
Divorced spouse 0.20 0.27 0.35 0.45 a 0.30 0.20 0.27 0.35 0.45 a 0.30 -- -- -- -- -- -- -- -- -- -- -- --
Married
Retired worker 0.01 0.03 0.03 0.04 0.06 0.03 0.01 0.02 0.03 0.04 0.06 0.03 -- -- -- -- -- -- -- -- -- -- -- --
Widow 0 0 0.10 0 0.11 0.04 0 0 0.10 0 0.11 0.04 -- -- -- -- -- -- -- -- -- -- -- --
Spouse 0.02 0.02 0.04 0.01 0.05 0.03 0.02 0.02 0.04 0.01 0.05 0.03 -- -- -- -- -- -- -- -- -- -- -- --
NOTE: -- = not applicable.
a. Sample size too small.

The benefit changes here are modest compared with those resulting from Option 1. As with Option 1, dually entitled individuals might switch from one benefit category to another, depending on which type of benefit generates the highest value.16 Again, the table shows individuals assigned to their benefit category as determined by current law, although we report the highest benefit to which they are entitled. Average monthly benefits do not increase much for most benefit categories. Correspondingly, the effects of Option 2 on family poverty rates are also small. Among nonmarried women, retired-worker beneficiaries show the largest drop in poverty rates, especially those aged 62–64. Widow beneficiaries in all but two of the age groups see a small reduction in their poverty rates. Among married women, who have lower poverty rates in general, we find only one group—the 65–69 retired workers—with a drop in poverty rates. Weak work histories of divorced spouses largely keep them from receiving this version of a minimum benefit.

As expected, total benefits paid and mean benefit amounts under Option 2 generally increase from the current-law amounts across all quintiles.17 Women in the first quintile tend to receive the largest relative benefit increases across all of the benefit categories. Option 2 is therefore effective in providing higher relative benefit increases to those at the lowest end of the family income distribution. The largest percentage increases in mean benefits, in general, are received by married and nonmarried retired-worker beneficiaries across all quintiles.

Option 3. Increase the Primary Insurance Amount at Age 80 by 5 Percent.

The third option increases the primary insurance amount (PIA) at age 80 by 5 percent.18 The rationale is that an individual's real Social Security benefit is essentially fixed at retirement. In an economy with real wages that increase over time, successive retiring cohorts tend to have higher average real benefit levels than their predecessors. Thus, the relative benefit levels of older retirees compared with those of new retirees decline over time. For older retirees, benefits become notably small relative to levels received by young retirees. It is well-documented that poverty rates of the Social Security beneficiary population tend to increase with age. The Social Security Administration (2000) reports that among women beneficiaries, the poverty rate for those aged 85 or older is substantially higher than the rate for those between ages 65 and 74.19 Increasing benefits at age 80 by 5 percent is a crude but simple method for reducing that difference.

Although a 5 percent increase in benefits for that age group boosts both their total annual and average monthly benefits, the impact on poverty rates across benefit categories varies (see Table 6). Only nonmarried retired-worker and nonmarried widow beneficiaries show slight declines in their poverty rates. That relatively small improvement in poverty rates entails an increase in program costs of about $1.4 billion.20 In addition, the option fails to accurately target low-income beneficiaries—less than 30 percent of the additional benefits generated under Option 3 accrue to those in the bottom quintile of the income distribution.

Table 6. Effects of Option 3 on women's benefits, by marital status, beneficiary type, age group, and income quintile, 1993
Benefit category Age group Income quintile
Current law Option 3 Current law Option 3
62–64 65–69 70–74 75–79 80+ All 62–64 65–69 70–74 75–79 80+ All 1st 2nd 3rd 4th 5th All 1st 2nd 3rd 4th 5th All
Total benefits paid (billions of dollars)
Nonmarried
Retired worker 0.96 5.18 4.49 5.39 8.41 24.43 0.96 5.18 4.49 5.39 8.83 24.85 6.09 6.06 4.30 4.39 3.58 24.43 6.19 6.19 4.38 4.46 3.64 24.85
Widow 3.93 10.08 13.44 12.01 16.45 55.90 3.93 10.08 13.44 12.01 17.27 56.72 17.80 15.06 9.50 7.75 5.78 55.90 18.10 15.27 9.62 7.87 5.87 56.72
Divorced spouse 0.16 0.29 0.21 0.16 a 0.83 0.16 0.29 0.21 0.16 a 0.83 0.36 0.15 0.16 0.10 0.06 0.84 0.37 0.15 0.16 0.10 0.06 0.84
Married
Retired worker 2.92 6.86 4.74 3.37 1.81 19.71 2.92 6.86 4.74 3.37 1.90 19.80 1.59 2.94 4.48 5.50 5.20 19.71 1.60 2.97 4.49 5.52 5.22 19.80
Widow 0.10 0.42 0.45 0.66 0.41 2.04 0.10 0.42 0.45 0.66 0.43 2.06 0.19 0.35 0.38 0.46 0.65 2.04 0.20 0.36 0.38 0.47 0.65 2.06
Spouse 3.45 7.58 6.78 3.55 1.99 23.35 3.45 7.58 6.78 3.55 2.09 23.45 1.50 4.56 6.01 5.19 6.09 23.35 1.51 4.59 6.03 5.21 6.11 23.45
Total benefits 11.52 30.40 30.12 25.14 29.06 126.25 11.52 30.40 30.12 25.14 30.51 127.70 27.55 29.12 24.82 23.40 21.37 126.26 27.97 29.52 25.06 23.62 21.55 127.72
Average monthly benefits (dollars)
Nonmarried
Retired worker 537 596 586 656 670 628 537 596 586 656 704 639 472 652 696 712 827 628 479 666 708 722 841 639
Widow 662 703 719 735 664 698 662 703 719 735 697 708 593 730 764 793 803 698 603 740 773 804 814 708
Divorced spouse 330 364 403 396 a 371 330 364 403 396 a 371 331 337 470 403 570 373 332 337 470 403 570 373
Married
Retired worker 482 513 510 581 514 518 482 513 510 581 540 520 387 471 523 531 592 518 390 475 525 533 594 520
Widow 673 596 688 755 617 670 673 596 688 755 647 676 617 573 616 694 781 670 630 582 623 699 785 676
Spouse 361 381 413 411 384 391 361 381 413 411 403 393 271 365 400 402 444 391 273 368 401 403 445 393
Average monthly benefit 480 526 565 624 623 569 480 526 565 624 654 575 510 586 575 583 611 569 518 594 581 588 616 575
Poverty rate (percent)
Nonmarried
Retired worker 0.24 0.15 0.16 0.20 0.16 0.17 0.24 0.15 0.16 0.20 0.13 0.16 -- -- -- -- -- -- -- -- -- -- -- --
Widow 0.13 0.15 0.14 0.15 0.18 0.16 0.13 0.15 0.14 0.15 0.14 0.14 -- -- -- -- -- -- -- -- -- -- -- --
Divorced spouse 0.20 0.27 0.35 0.45 a 0.30 0.20 0.27 0.35 0.45 a 0.30 -- -- -- -- -- -- -- -- -- -- -- --
Married
Retired worker 0.01 0.03 0.03 0.04 0.06 0.03 0.01 0.03 0.03 0.04 0.06 0.03 -- -- -- -- -- -- -- -- -- -- -- --
Widow 0 0 0.10 0 0.11 0.04 0 0 0.10 0 0.11 0.04 -- -- -- -- -- -- -- -- -- -- -- --
Spouse 0.02 0.02 0.04 0.01 0.05 0.03 0.02 0.02 0.04 0.01 0.05 0.03 -- -- -- -- -- -- -- -- -- -- -- --
NOTE: -- = not applicable.
a. Sample size too small.

Conclusion

Analyses of public policy proposals often reveal that changes are likely to be somewhat less effective than advocates might anticipate. For instance, the 5 percent increase in the PIA at age 80 is shown to be a poorly targeted antipoverty measure for elderly women generally and for women in the lowest two quintiles of the income distribution in particular, whose monthly benefits would increase very little. Also problematic can be the "law of unintended consequences,*quot; as occurs with Option 1, in which a policy designed to help one disadvantaged group (elderly widow beneficiaries) ends up harming another disadvantaged group (divorced-spouse beneficiaries). Distributional analyses of the type presented here can help identify such weaknesses in policy design.

Earlier we alluded to the problem of evaluating policy proposals using simulations that assume the changes were fully implemented during a recent year. That approach can lead to particular difficulties in the case of women's Social Security benefits because those benefits depend on complex interactions among women's lifetime earnings patterns, marital histories, and the correlation between the earnings histories of married women and those of their husbands. Those factors are briefly discussed in turn.

First, it is well-established that women's wages (relative to men's) have been steadily increasing during the past several decades. The trend is thought to have a number of causes, including greater investments in education by young women, rising wage premiums for skills, occupational entry patterns for women that more closely approximate those of men, and a higher degree of lifetime labor force attachment than had previously been the case. Those developments in the labor force activity of women suggest that in the future, a higher proportion of women reaching retirement age will be eligible for their own Social Security retired-worker benefits than was the case in 1993. As a corollary, the fraction of women entitled as spouse-only beneficiaries will decline.

Second, marriage patterns have also been changing. The increased incidence of divorce has been widely discussed and, on its own, would lead to a smaller proportion of women near retirement age qualifying for secondary Social Security benefits. Perhaps just as important is the lower propensity of women to enter marriage at all. In general, secondary benefits paid in old age require the beneficiary to have been married to an insured worker for a minimum of 10 years. Marriage and divorce trends both suggest that women reaching old age in the future will be less likely to qualify for secondary Social Security benefits of any type.

Third, for a woman who has a 10-year marriage, a key determinant of her eventual treatment by Social Security is the relationship of her earnings history to that of her husband. Although aggregate statistics on earnings might indicate that women's average earnings are increasing relative to those of men, they do not identify the trend in the average wife/husband earnings ratio within couples, nor do they inform us about changes in the distribution of that ratio. Nonetheless, the relationship between a woman's lifetime earnings and those of her husband determines the type of Social Security benefit that she will eventually receive. Research on that issue remains limited. The reader should therefore keep all of these economic and social trends in mind when making inferences about the likely future effects of policies using the results of simulations conducted for 1993.

One lesson that emerges from this study is the importance of distinguishing between current marital status and the type of Social Security beneficiary. Although the terms are similar (for example, divorced woman versus divorced-spouse beneficiary), they sometimes describe decidedly different population subgroups. Many Social Security policy analyses are conducted using survey reports of marital status, which can lead to very different conclusions about a policy's impact than if one uses the SSA's benefit distinctions.21 To demonstrate the point, consider our initial division of the sample into groups based on the type of benefit received, as reported in the MBR file. When that categorization of the sample is used, divorced-spouse beneficiaries receive lower benefits under Option 1 because of the spouse benefit reduction.22 But, if we divide the sample into groups based on current marital status as reported in the SIPP, we find that those individuals who report themselves to be divorced or separated women actually incur much smaller (in percentage terms) decreases in mean monthly benefits. In fact, as Table 7 shows, divorced or separated women in the group aged 80 or older actually show an increase in mean monthly benefits.

Table 7. Monthly benefits of divorced or separated persons (marital status as reported in the SIPP)
Age Current law Option 1 Percentage difference
62–64 505 493 -2.37
65–69 554 546 -1.36
70–74 577 573 -0.71
75–79 570 565 -0.87
80 or older 652 659 1.22
All (62 or older) 565 560 -0.93
 

Why the apparent discrepancy? Most divorced or separated women are not divorced-spouse beneficiaries; they receive other types of benefits. Closer examination of their beneficiary status shows that the share of widow beneficiaries is larger than that of spouse beneficiaries, by about 10 percent. Nearly 65 percent claim benefits on their own work histories. Thus, a reduction in spouse benefits for divorced-spouse beneficiaries does not adversely affect the survey-reported divorced spouses taken as a group.

We close with a final problem to consider in interpreting and evaluating the effects of alternative policy changes that are not equal-cost options. Well-targeted, higher cost policies should, to no one's surprise, have larger effects on attaining a policy objective. For instance, the results in Tables 4–6 imply that implementing Option 1 in 1993 would lower the cost of women's Social Security benefits by $3.2 billion (remember, men are not represented in those tables); Options 2 and 3 add $1.3 billion and $1.4 billion, respectively, to Social Security's expenditures. In other words, two of the three program changes exacerbate the long-run solvency problem. Further analyses that address that point could rerun the simulations, constraining program expenditures to be consistent with program revenues, by adjusting all benefits proportionately during the reference year to be consistent with current law payroll tax rates ( Social Security Advisory Board 1999, p. 35).

Notes

1. The original 32-month, 8-wave design of SIPP was recently abandoned by the U.S. Census Bureau. Beginning with the 1996 SIPP, about 37,000 households are now interviewed 12 times over a 4-year period, with a new panel selected every 4 years. The 4-year panels will be supplemented by 3-year panels that interview about 11,000 households. The 3-year panels will commence in all years in which a longer, larger 4-year panel does not start. See Weinberg (1999) for further details.

2. These match rates are generally consistent with previous studies that have linked SIPP data with SSA's administrative files. Earnings record data are available for respondents who have worked in Social Security-covered employment, thus excluding those without work experience and workers who spent entire careers in noncovered jobs. Benefit records are established only when an application for Social Security benefits is made. Consequently, many individuals at the younger end of our age range have no benefit record. In addition, reporting and verification of Social Security numbers is imperfect, resulting in some mismatches and nonmatches between survey and administrative files.

3. The study's initial age restriction reduced the sample size to 13,894 individuals. After linking survey data to administrative records, we compared year of birth, sex, and race variables to identify obvious mismatches. That reduced the sample by 202 cases. We also dropped 1,396 survey records that were not linked to either SER or MBR data.

4. Under current SSI rules, many proposals that would increase Social Security benefits would result in dollar-for-dollar decreases in SSI payments, leaving a low-income beneficiary's income unchanged. The implicit assumption here is that a Social Security change intended to boost the income of the poor would be accompanied by accommodating SSI program changes.

5. Some advocates suggest that this new survivor benefit should be subject to a cap, thereby limiting the extent to which new benefits would accrue to high-income survivors. For example, the National Council of Women's Organizations has suggested that the benefit be capped at the amount of the maximum earner's retired-worker benefit. The analysis presented here has no cap.

6. The article's remaining tables are devoted strictly to the financial circumstances of female beneficiaries. Most proposed changes to the Social Security program are gender-neutral in their language. We restrict our examination of program changes to women, even though it is clear that a lesser number of men would also be affected. In evaluating the financial status of married women, we take into account any changes in her husband's Social Security benefits that would result from the stipulated program changes.

7. Our results show that approximately 1.2 million spouse beneficiaries switch to the retired-worker category. Due to data limitations, we do not report any cases of widows who claim retired-worker benefits switching to survivor benefits. There are 1.9 million widows who claim retired-worker benefits in our sample. We do not have access to Social Security record data for their deceased spouses, which precludes a calculation of the couples' benefit under alternative scenarios. Therefore, we are unable to determine the number of switchers. Keep in mind that only those widows whose ratio of MBA to that of their husbands is between 1 and 3 will switch to the survivor beneficiary category; all others remain unaffected by this plan.

8. The poverty rates in this paper are calculated using total family income and the poverty cutoff points provided in the SIPP data files. Total family income is changed by the amount of benefit increases or decreases accruing to the family due to the implementation of the plan.

9. "Families" include SIPP respondents who do not live with any other family members.

10. Estimated annual program costs increase by about $180 million for men because a small number of widowed men also receive higher benefits under these changes. The increase in costs is also due to the fact that far fewer men than women claim spouse benefits, the beneficiary group that sees the largest drop in benefits under this Plan.

11. Under current law, Social Security income is at least 1.5 times the benefit paid to the higher lifetime earner. If the higher earner dies first, the benefit paid to the survivor falls by at least 33 percent of the couple's total benefit amount. Yet, the official U.S. poverty guidelines for families with a head of household aged 65 or older show that the poverty line for a single-person family is 21 percent lower than the threshold for a married couple. Thus, the combination of Census Bureau procedures for measuring poverty and current Social Security law implies that widowhood tends to lower the survivor's income-to-poverty ratio, other things—in particular, income amounts—being equal. This relationship is particularly relevant for lower-income aged families, where Social Security benefits usually represent a high proportion of the family's income. The change in the computation of survivor benefits described here dampens the tendency to lower the survivor's income-to-poverty ratio for widows of 2-earner couples.

12. In fact, Social Security program changes often cause many women to change their relative positions in the income distribution. We experimented with an alternative table design in which, for each proposal, individuals were reclassified according to their post-plan location in the income distribution. So, for example, a woman from the lowest quintile might receive a higher benefit, moving her into the next higher quintile, which might serve to lower the mean benefit in each of these two quintiles. These types of difficulties in interpreting the results led us to adopt the current design that classifies women according to their current law quintile.

13. This type of proposal has been discussed by the National Council of Women's Organizations and by Representatives Kolbe and Stenholm, among others.

14. Note that the current-law information in Table 5 is identical to that in Tables 4 and 6. It is repeated for ease of comparison to the new numbers of the plan(s) under discussion.

15. When men are included in the calculation, program costs increase to $1.6 billion.

16. Approximately 29,000 spouses and 21,000 widows would switch to the retired-worker category in order to receive higher benefits.

17. Again, individuals are kept in the same quintile and the same benefit category as they were under current law, though the highest possible benefit they were eligible for is reported in the table.

18. The primary insurance amount is the monthly amount payable to a retired worker who begins to receive benefits at the normal retirement age.

19. The Social Security Administration (2000) reports that in 1998 the poverty rate for women beneficiaries aged 65–74 was 8.8 percent; the rate for women beneficiaries aged 85 or older was 14.6 percent. Poverty rates were markedly higher for minorities and nonmarried persons.

20. Another $1 billion is attributable to increased men's benefits.

21. We note that internal agency analyses are sometimes based strictly on administrative record data, where marital status is often unavailable.

22. The mean decrease in monthly benefits going to divorced-spouse beneficiaries using this categorization is 27.1 percent.

References

Advisory Council. 1997. Report of the 1994–1996 Advisory Council on Social Security: Findings and Recommendations. Washington, D.C.: Social Security Administration.

Social Security Administration. 2000. Income of the Population 55 or Older, 1998. Office of Research, Evaluation, and Statistics, Washington, D.C. March.

Social Security Advisory Board. 1999. 1999 Technical Panel on Assumptions and Methods: Report to the Social Security Advisory Board. Washington, D.C. November.

Weaver, David A. 1997. "The Economic Well-being of Social Security Beneficiaries, with an Emphasis on Divorced Beneficiaries," Social Security Bulletin, Vol. 60, No. 4, pp. 3–17.

Weinberg, Daniel H. 1999. "The Survey of Income and Program Participation—Recent History and Future Developments." Washington, D.C.: U.S. Census Bureau. January.