Benefit Adequacy Among Elderly Social Security Retired-Worker Beneficiaries and the SSI Federal Benefit Rate

by Kalman Rupp, Alexander Strand, Paul Davies, and Jim Sears
Social Security Bulletin, Vol. 67 No. 3, 2007

The federal benefit rate (FBR) of the Supplemental Security Income program provides an inflation-indexed income guarantee for aged and disabled people with low assets. Some consider the FBR as an attractive measure of Social Security benefit adequacy. Others propose the FBR as an administratively simple, well-targeted minimum Social Security benefit. However, these claims have not been empirically tested. Using microdata from the Survey of Income and Program Participation, this article finds that the FBR is an imprecise measure of benefit adequacy; it incorrectly identifies as economically vulnerable many who are not poor, and disregards some who are poor. The reason for this is that the FBR-level benefit threshold of adequacy considers the Social Security benefit in isolation and ignores the family consumption unit. The FBR would provide an administratively simple but poorly targeted foundation for a minimum Social Security benefit. The empirical estimates quantify the substantial tradeoffs between administrative simplicity and target effectiveness.


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

Acknowledgments: We would like to thank Jasmin Kakar for expert programming support and Anne DeCesaro for outstanding research assistance. Many of our colleagues provided useful comments on various drafts. We particularly appreciate Ed DeMarco's sage and insightful feedback.

The findings and conclusions presented in the Bulletin are those of the authors and do not necessarily represent the views of the Social Security Administration.

Summary

Both target effectiveness and administrative simplicity are desirable properties in the design of minimum benefit packages for public retirement programs. The federal benefit rate (FBR) of the Supplemental Security Income (SSI) program has been proposed by some analysts as a potentially attractive basis of establishing a new minimum benefit for Social Security on both of these grounds. This type of proposal is related to a broader array of minimum benefit proposals that would establish a Social Security benefit floor based on the poverty rate.

In contrast to Social Security, the SSI program is means tested, including both an income and asset screen and also a categorical eligibility screen (the requirement to qualify as aged or disabled). The SSI FBR provides an inflation-adjusted, guaranteed income floor for aged and disabled people with low assets.

The FBR has been perceived by proponents as a minimal measure of Social Security benefit adequacy because it represents a subpoverty income level for a family of one or two depending on marital status. For this same reason it has been seen as a target-effective tool of designing a minimum Social Security benefit. An FBR-based minimum benefit has also been viewed as administratively simple to implement; the benefit can be calculated from Social Security administrative records using a completely automated electronic process. Therefore—in contrast to the SSI program itself—an FBR-based minimum benefit would incur virtually no ongoing administrative costs, would not require a separate application for a means-tested program, and would avoid the perception of welfare stigma.

While these ideas have been discussed in the literature and among policymakers in the United States over the years, and similar proposals have been considered or implemented in several foreign countries, there have been no previous analyses measuring the size of the potentially affected beneficiary population. Nor has there been any systematic assessment of the FBR as a measure of benefit adequacy or the tradeoffs between potential target effectiveness and administrative simplicity.

Based on a series of simulations, we assess the FBR as a potential foundation for minimum Social Security benefits and we examine the tradeoffs between administrative simplicity and target effectiveness using microdata from the 1996 panel of the Survey of Income and Program Participation (SIPP). Our empirical analysis is limited to Social Security retired-worker beneficiaries aged 65 or older. We start with the assessment of the FBR as a measure of benefit adequacy. We are particularly concerned about two types of error: (1) incorrectly identifying some Social Security beneficiaries as "economically vulnerable," and (2) incorrectly identifying others as "not economically vulnerable." Operationally we measure economic vulnerability by two alternative standards. One of our measures considers beneficiaries with family income below the official poverty threshold as vulnerable. Our second measure is more restrictive; it uses a family income threshold equal to 75 percent of the official poverty threshold.

We find that a substantial minority of retired workers have Social Security benefits below the FBR. The results also show that the FBR-based measure of Social Security benefit adequacy is very imprecise in terms of identifying economically vulnerable people. We estimate that the vast majority of beneficiaries with Social Security benefits below the FBR are not economically vulnerable. Conversely, an FBR-level Social Security benefit threshold fails to identify some beneficiaries who are economically vulnerable. Thus an FBR-level minimum benefit would be poorly targeted in terms of both types of errors we are concerned about. An FBR-level minimum benefit would provide minimum Social Security benefits to many people who are clearly not poor. Conversely, an FBR-level minimum benefit would not provide any income relief to some who are poor. The administrative simplicity behind these screening errors also results in additional program cost that may be perceived as substantial. We estimate that an FBR-level minimum benefit would increase aggregate program cost for retired workers aged 65 or older by roughly 2 percent.

There are two fundamental reasons for these findings. First, the concept of an FBR-level minimum benefit looks at the individual or married couple in artificial isolation; however, the family is the main consumption unit in our society. The income of an unmarried partner or family members other than a married spouse is ignored. Second, individuals and couples may also have income from sources other than Social Security or SSI, which is also ignored by a simple FBR-based minimum benefit concept.

The substantial empirical magnitude of measurement error arising from these conceptual simplifications naturally leads to the assessment of the tradeoff between target effectiveness and administrative simplicity. To facilitate this analysis, we simulate the potential effect of alternative screening methods designed to increase target effectiveness; while reducing program cost, such alternatives also may increase administrative complexity. For example, considering the combined Social Security benefit of a married couple (rather than looking at the husband and wife in isolation) might substantially increase target effectiveness with a relatively small increase in administrative complexity. Adding a family income screen might increase administrative complexity to a greater degree, but also would increase target effectiveness dramatically. The results also suggest that at some point adding new screens—such as a comprehensive asset test—may drastically increase administrative complexity with diminishing returns in terms of increased target effectiveness and reduced program cost.

Whether a broad-based minimum benefit concept that is not tied to previous work experience is perceived by policymakers as desirable or not may depend on several factors not addressed in this article. However, to the extent that this type of minimum benefit design is regarded as potentially desirable, the tradeoffs between administrative simplicity and target effectiveness need to be considered.

Introduction

The Supplemental Security Income (SSI) program's monthly income guarantee—the federal benefit rate (FBR)—has entered policy discussions of the adequacy of benefits for Social Security beneficiaries in two ways. First, it has been described as one possible standard to judge the adequacy of the benefits provided by the Old-Age and Survivors Insurance (OASI) program. Second, the FBR is the basis of some Social Security minimum benefit proposals.

Thompson (2004) describes the federal SSI guarantee for a single individual as one of several adequacy benchmarks. The various benchmarks Thompson discusses—such as the poverty line for a single individual or the minimum wage—differ in their generosity and rationale. In contrast to the poverty line, the FBR may be seen as an appealing standard of adequacy because it represents an existing income guarantee for the elderly, as opposed to a measurement tool. In addition, Social Security benefit amounts can be directly observed in administrative records, while establishing family poverty status requires survey interview or other data. However, while the poverty measure considers the family as the unit of measurement and accounts for all sources of income, using the FBR as a measure of Social Security benefit adequacy limits the analysis to Social Security benefits and, thus, ignores all other sources of income. Further, when using the FBR, the focus of the analysis becomes the Social Security benefits of the individual and his or her possible spouse, and it moves away from the income of the family, which is the principal consumption unit in our society.

The minimum benefit is not a new concept in Social Security policy. A broadly applicable minimum benefit was established by the 1939 Amendments to the Social Security Act. Subsequently it has been criticized as insufficiently targeted and was eventually eliminated by the 1981 amendments. A more targeted, "special" minimum benefit was established by the 1972 amendments, but it affected only a small and diminishing group of beneficiaries (Olsen and Hoffmeyer 2001/2002).1 In fact, Feinstein (2000) estimates that it will be impossible for anyone who becomes entitled to Social Security benefits in 2013 or later to receive the special minimum. Major Social Security reform proposals such as Kolbe-Stenholm, H.R. 1793 (1999), Graham, S. 1878 (2003), and the minimum benefit provisions of Models 2 and 3 of the President's Commission to Strengthen Social Security (2001) also target low earners with long-term attachment to the labor force.2 The application of the SSI FBR as a potential tool in establishing a Social Security minimum benefit is relatively new to policy discussions.

The proposal to establish a Social Security minimum benefit at the FBR level (Herd 2005) is related to a broader array of less-targeted minimum benefit proposals that would establish a Social Security benefit floor based on the poverty rate or some multiple thereof, with little or no conditioning on prior earnings history (McGarry 2000; Wasow 2004; Smeeding 1999; Smeeding and Weaver 2002).3 The "Resident Minimum" proposal (Herd 2005) is universal and guarantees a flat benefit set at the federal SSI level for all elderly residents of the United States. The minimum benefit scenario analyzed by McGarry (2000) is also universal, but sets the minimum at the poverty line. The "Senior Income Guarantee" proposal (Smeeding and Weaver 2002) provides a minimum benefit guarantee of 75 percent of the poverty line and would provide benefits to all Social Security beneficiaries at or above the normal retirement age. Wasow (2004) proposes a "New Minimum Social Security Benefit" that would provide a Social Security benefit guarantee at the poverty line for households of retirees who receive at least 75 percent of their income from Social Security.4 According to a recent review (OECD 2007), minimum pensions play some role in almost half (14 of 30) of the "first tier" of public pension systems in Organisation for Economic Co-operation and Development (OECD) countries. The appropriate roles of more universalistic minimum benefits versus means-tested pension system components are widely discussed among experts in the developed world. In contrast to the United States, some OECD countries have substantial present or past experience with universalistic minimum benefit components in their public pension systems.

In 2005, the monthly SSI FBR was $579 for individuals and $869 for couples.5 The effective level was slightly higher for Social Security beneficiaries ($599 and $889, respectively) because the first $20 of Social Security or other income is exempted from the SSI payment calculation. The poverty threshold for a one-person family with a householder aged 65 or older with no children was $9,367 per year in 2005. The corresponding figure for a two-person family with an elderly householder was $11,805. These thresholds are higher than the annualized effective SSI FBR of $7,188 for an individual (a monthly benefit of $579 plus $20 multiplied by 12) and $10,668 for a couple in the same year (a monthly benefit of $869 plus $20 times 12). The effective FBR amounts to roughly between 77 percent and 90 percent of the applicable poverty threshold for one- and two-person elderly families. Both the FBR and the official poverty threshold are indexed to inflation. The FBR increases with the same automatic cost-of-living adjustment (COLA) that is applied to Social Security benefits each January.6

The FBR may be a potentially attractive tool for designing a minimum benefit because of its promise to avoid some perceived drawbacks of alternative approaches. In contrast to minimum benefit provisions that are conditional on substantial work experience, an FBR-level minimum OASI benefit guarantee could be applied to all elderly OASI beneficiaries.7 Compared with minimum benefit approaches that are similar to the existing SSI program, the OASI minimum benefit would be an administratively simple way of reaching the targeted OASI beneficiaries without the imposition of a resource test. Yet, a minimum benefit based on the FBR may not be as target efficient as minimum benefits based on other approaches. Further, it may be less cost effective. This article presents evidence relevant to the tradeoffs between administrative simplicity, target efficiency, and program cost.

The analysis here provides empirical data necessary to assess (1) the usefulness of the SSI FBR as a measure of Social Security benefit adequacy, and (2) minimum benefit proposals that focus on the provision of FBR-level minimum Social Security benefits. Administrative simplicity is part of the appeal of this approach; the information necessary to measure benefit adequacy and to administer the proposed minimum benefit would be available from administrative records. This simplicity, however, may result in error in classifying beneficiaries by economic vulnerability. We are particularly concerned about two types of classification error: (1) incorrectly screening in those who are not economically vulnerable, and (2) incorrectly screening out those who are economically vulnerable.8

Although there have been discussions on these issues in the literature and among policymakers, no reliable data have been published about the proportion of elderly retired-worker beneficiaries with benefits below the FBR, and no estimates are available to assess the target efficiency of FBR-related minimum benefit proposals. Without such information it is difficult to assess complex tradeoffs involving administrative simplicity, distributional outcomes, and program cost. This study intends to fill this information gap, but does not attempt to judge the policy merits of specific reform proposals.

The rest of the article is organized as follows. First we briefly describe the data and methodology for the empirical analysis, and then provide information on the prevalence of Social Security benefits below the effective FBR among elderly retired-worker beneficiaries. What follows is an analysis of SSI participation among elderly retired-worker beneficiaries with Social Security benefits below the effective FBR. Next we determine the quality of the FBR as a yardstick in assessing the adequacy of benefits using family income relative to the poverty threshold as the measure of economic well-being. In the section that follows, we assess the tradeoffs between administrative simplicity and effective targeting, and finally we conclude by discussing areas for potential future research.

Data and Methodology

The source of data for this study is the 1996 panel of the Survey of Income and Program Participation (SIPP) matched to Social Security administrative records. The sample universe here is limited to Social Security retired-worker beneficiaries aged 65 or older in the United States' noninstitutional population in November 1996. The institutional segment of the elderly population (for example, those in nursing homes) are not included in our empirical estimates. Beneficiaries are defined on the basis of Social Security participation (current-pay status) as reflected in records matched to the SIPP from the Social Security Administration's (SSA's) Master Beneficiary Record (MBR). In this article, "retired-worker beneficiary" is defined as a fully insured Social Security beneficiary who receives benefits as a result of his or her own earnings record. Former disabled workers who automatically converted to OASI at the full retirement age are included in this definition of retired-worker beneficiary. Only retired workers are counted as reference persons in our individual-level analysis file; other OASI beneficiaries (such as dependents and survivors) are excluded from the sample frame.9

Our study methodology is based on the Office of Retirement and Disability Policy's Financial Eligibility Model (FEM). The FEM is a static simulation model focusing on SSI financial eligibility, participation, and the assessment of various SSI policy options. The key elements of the FEM are described in Davies and others (2002). The basic structure of the FEM is similar to the SSI model that has been developed by McGarry (1996, 2000), except that the FEM utilizes administrative records matched to the survey data and contains a more detailed algorithm to establish SSI financial eligibility. This study extends the application of the FEM to the measurement of Social Security benefit adequacy and the assessment of OASI minimum benefit proposals.

We briefly describe some key elements of the FEM below as we applied them to the subject of this study. A key element of the FEM is a financial eligibility calculator that estimates potential SSI income and resource eligibility for any sample member regardless of actual program participation.10 The eligibility calculator is based on detailed SSI income and asset eligibility rules applied to survey data on income and assets reported in the SIPP. For those deemed financially eligible for SSI, the FEM calculates expected (hypothetical) federal SSI payments based on the applicable FBR (individual or couple unit) and countable income from the SIPP.11

In this study we establish potential financial eligibility for "FBR-level" minimum Social Security benefits with some appropriate modifications. Since up to $20 of Social Security income can be excluded from countable income, we define an "effective" FBR measure, derived simply by adding $20 to the applicable SSI FBR.12

We define a retired-worker "unit" as a retired worker without a spouse present (individual unit) or a retired worker with a spouse present (couple unit). If both spouses are aged 65 or older, this is identical to the SSI unit concept. If there is a nonelderly spouse, the SSI determination of whether to apply the individual or couple FBR is more complicated. A sensitivity analysis indicated that the inclusion or exclusion of retired workers with a nonelderly spouse makes very little difference in the estimates. Thus, we include the spouses of all Social Security retired-worker beneficiaries aged 65 or older, if any, without regard to the age of the spouse.

For each individual in the sample, we calculate both an "effective individual FBR" and an "effective unit FBR." The effective individual FBR concept applies to each individual in the sample regardless of the presence of a spouse. The effective unit FBR concept is equal to the individual SSI FBR plus $20 for sample members without a spouse present and the couple SSI FBR plus $20 for those with a spouse present. By comparing the monthly retired-worker benefit recorded in the MBR to one of these "effective FBR" thresholds, one can establish whether a sample member has Social Security benefits below or above the FBR.

These measures in conjunction with other data on beneficiary characteristics reported in the SIPP are then used to assess benefit adequacy and tradeoffs between administrative complexity, distributional outcomes, and potential program cost. In comparing various outcomes of interest, we focus on patterns and magnitudes of substantive importance. However, we also provide standard error estimates to facilitate the calculation of confidence intervals or to perform simple tests of differences in means that may be of interest to some readers.13 We do not model behavioral responses to alternative policy options—a simplification that is probably more reasonable for the benefit-claiming behavior of retired-worker beneficiaries beyond the full retirement age than would be the case for some other beneficiary groups such as disabled workers or early retirees.14

Prevalence of Social Security Benefits Below the Effective FBR Among Elderly Social Security Retired-Worker Beneficiaries

In order to provide an empirical estimate, one needs to deal with an ambiguity. As noted earlier, the SSI program distinguishes between two kinds of units—"individuals" and "couples." In order to account for economies of scale in consumption, the individual FBR is set at about two-thirds of the couple FBR. Are we to apply the individual FBR to the OASI benefits of the retired worker regardless of the presence or absence of a spouse, or should we apply the couple FBR to the combined benefits of the retired worker and spouse for married couples? The answer to this question has substantial effects on the estimates (see Table 1). When the individual FBR is applied to the individual benefit amount of the retired worker, we find that approximately one-fourth (23 percent) of retired workers have benefits below the FBR.15 However, when the unit concept is used, the proportion drops to 15 percent.

Table 1. Percentage of Social Security retired-worker beneficiaries aged 65 or older with Social Security benefits below the effective SSI federal benefit rate
Measure Single Married All
OASI benefit below effective individual FBR a 19.3
(0.7)
25.2
(0.7)
22.6
(0.5)
OASI benefit below effective unit FBR a 19.3
(0.7)
12.1
(0.5)
15.2
(0.4)
N 2,966 3,700 6,666
SOURCES: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTES: Standard error estimates (in parentheses) reflect the assumption of simple random sampling. See U.S. Census Bureau (2001) for the adjustments that are needed to account for the SIPP sample design effect.
SSI = Supplemental Security Income; OASI = Old-Age and Survivors Insurance; FBR = federal benefit rate; N = the unweighted count of the number of observations for the denominator of the estimated percentages; SIPP = Survey of Income and Program Participation.
a. The effective FBR (for individual or unit) equals the applicable FBR plus $20 to account for the exclusion of up to $20 from any source, including Social Security, in the benefit calculation.

The difference, of course, is attributable to married couples. Using the individual FBR, we see that about a quarter (25 percent) of married elderly retired-worker Social Security beneficiaries appear to have Social Security benefits below the FBR, while the consideration of the husband's and wife's combined Social Security benefits against the couple FBR cuts this estimate by more than half, to 12 percent. The relative position of the two groups is reversed as well. Using the "individual" concept would make the Social Security benefits of married retired workers look relatively inadequate. In contrast, when the unit concept is used, the results are consistent with the generally accepted notion of greater economic vulnerability of the single elderly person.

Although policy discussions of applying the SSI FBR to the measurement of the adequacy of Social Security benefits are often unclear about the proposed use of the individual or couple FBR, the empirical differences are substantial. In the next two sections we will use the unit concept because it appears to provide a more reasonable measure of "adequacy." Later we revisit the relationship between the two measures of benefit adequacy as potential screening variables in establishing a Social Security minimum benefit.

SSI Participation Among Elderly Social Security Retired-Worker Beneficiaries with Social Security Benefits Below the Effective Unit FBR

What is the extent of SSI participation among elderly retired-worker beneficiaries with Social Security benefits below the effective unit SSI FBR? Overall, only about 20 percent of elderly retired-worker beneficiaries with Social Security benefits below the effective unit FBR participate in the SSI program.16

What are the reasons for SSI nonparticipation? The main reason for nonparticipation is the lack of SSI financial eligibility. As Chart 1 shows, we estimate that only about 30 percent of elderly retired-worker beneficiaries with Social Security benefits below the unit FBR are financially eligible for SSI. This amounts to 4.7 percent of all Social Security retired-worker beneficiaries aged 65 or older.17 All elderly persons are categorically eligible for SSI, but applicants also need to meet an income and asset test. Almost half of all beneficiaries with below-FBR Social Security benefits (46 percent) fail to meet both the income and asset screens. An additional 17 percent meet the asset test but have incomes that are too high to qualify for SSI, while a smaller group of 7 percent meet the income screen but have countable assets above the asset threshold. Another way to look at these numbers is to observe that the majority of Social Security beneficiaries with below-FBR Social Security benefits (63 percent) have countable income from sources other than Social Security benefits that would disqualify them from receiving SSI payments regardless of the asset screen.

Chart 1.
Percentage distribution of Social Security retired-worker beneficiaries aged 65 or older with Social Security benefits below the effective SSI federal benefit rate for individuals or couples, by SSI income and asset eligibility status
Pie chart described in previous paragraph.
SOURCE: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTE: SSI = Supplemental Security Income.

Because SSI is a voluntary program, not all elderly who might be financially eligible for SSI actually participate. In addition to financial eligibility, the person (or couple) also has to apply—provide SSA with the necessary personal financial information—and be determined eligible by SSA. We estimate that about 63 percent of financially eligible retired-worker beneficiaries participate in SSI. Thus over one-third do not participate in SSI, forming about 10 percent of all beneficiaries with Social Security benefits below the FBR.

What are the key characteristics associated with SSI nonparticipation among financially eligible retired workers aged 65 or older with below-FBR Social Security benefits? A description of the demographic characteristics of the two principal subgroups: participant and nonparticipant eligibles is presented in Table 2.

Table 2. Percent with selected characteristics among participant and nonparticipant SSI eligibles a among Social Security retired-worker beneficiaries aged 65 or older
Characteristic SSI participation status
Participant Nonparticipant
Percent Estimated
standard
error
(percent)
Percent Estimated
standard
error
(percent)
Women 65 3 57 4
Married 20 3 31 4
Resides in metropolitan statistical area 67 3 63 4
Black 35 3 28 4
Hispanic 18 2 10 3
Former DI beneficiary 22 3 16 3
Self-reported poor health 29 3 25 4
Self-reported fair or poor health 66 3 55 4
High school graduate 21 3 38 4
Aged 80 or older 20 3 23 4
N 247 137
SOURCES: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTES: Standard error estimates reflect the assumption of simple random sampling. See U.S. Census Bureau (2001) for the adjustments that are needed to account for the SIPP sample design effect.
SSI = Supplemental Security Income; DI = Disability Insurance; N = the unweighted count of the number of observations for the denominator of the estimated percentage; SIPP = Survey of Income and Program Participation.
a. SSI eligibility has been estimated using the Office of Retirement and Disability Policy's Financial Eligibility Model (FEM) based on the SIPP. Participants who were estimated to be financially ineligible by the FEM are excluded from this table.

Several of the estimated differences between participants and nonparticipants are fairly minor. Other differences are noteworthy—even though not all of them would meet stringent statistical significance requirements because of the small sample size and the SIPP design effect. Participants are more likely to be Hispanic and women than nonparticipants, and they are also less likely to be married. High school graduates are substantially overrepresented among nonparticipants. This may reflect perceived stigma or other factors associated with high school graduation status. Former Disability Insurance (DI) beneficiary status is positively related to SSI participation. These beneficiaries usually have extensive past involvement with Social Security and may have previously received SSI on the basis of being categorically disabled. Access to Medicaid has additional value for participants and may contribute to the explanation of the pattern of relatively high rate of participation among those with poor and fair self-reported health and former DI beneficiary status.

Financial incentives should also be considered here because there is considerable evidence showing that expected SSI payments are associated with the decision to participate in the SSI program. Consistent with past research, participants are eligible for a higher SSI payment than nonparticipant eligibles would be if they applied (Table 3). We estimate that the expected SSI monthly payment18 of nonparticipants is only 68 percent of that of participants. This difference is counterbalanced by the higher average Social Security benefit of nonparticipants. Note, however, that SSI nonparticipation still results in a nontrivial average amount of foregone income among nonparticipants. The model-predicted foregone SSI payment amounts to about 23 percent of the retired worker's Social Security benefit.19 The net result is that the combined Social Security and SSI benefit of nonparticipants ($489) is slightly lower than the corresponding value for participants ($519). However, if we assume SSI application and award among nonparticipants, the combined value of Social Security and SSI benefits for the "unit" would be 12 percent higher for nonparticipants (see last row of Table 3).

Table 3. Actual and predicted Social Security benefit and SSI payment among Social Security retired-worker beneficiaries aged 65 or older who are estimated to be eligible to receive SSI payments, by SSI participation status
Benefit type SSI participation status Nonparticipant
minus
participant
difference
(1996 dollars)
Nonparticipant
average
as a
percentage of
participant average
Participant Nonparticipant
Average
monthly
amount
(1996
dollars)
Estimated
standard
error
(percent)
Average
monthly
amount
(1996
dollars)
Estimated
standard
error
(percent)
Social Security benefit of retired worker 334 6 393 9 59 118
Social Security benefit of "unit" a 371 8 477 21 106 128
Model predicted SSI payment of retired worker b 134 6 91 6 -43 68
Observed SSI payment of retired worker 134 6 0 0 -134 0
Observed SSI payment of "unit" a 148 7 13 5 -135 9
Social Security plus SSI of "unit" a 519 6 489 22 -29 94
Model predicted Social Security plus SSI payment of "unit" a, c 519 6 580 21 61 112
N 247 137 -- --
SOURCES: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTES: Standard error estimates reflect the assumption of simple random sampling. See U.S. Census Bureau (2001) for the adjustments that are needed to account for the SIPP sample design effect.
SSI = Supplemental Security Income; SIPP = Survey of Income and Program Participation; N = the unweighted count of the number of observations for the denominator of the estimated percentages; -- = not applicable; FEM = Financial Eligibility Model.
a. For individuals with spouse present includes benefit of retired worker and of spouse. For others it includes benefit of retired worker only.
b. This row represents hypothetical benefits calculated from SIPP data by the FEM model. For participants it is expected to be close to the observed SSI payment. For nonparticipants it is a hypothetical amount predicting the SSI payment the retired worker would be entitled to receive conditional on application and award. In order to distinguish these hypothetical amounts from the observed amounts for other variables we use italics for this row.
c. The average monthly amounts are calculated by summing the observed Social Security benefit of the retired worker "unit," the model-predicted SSI payment of the retired worker and the observed SSI payment of the spouse (if any). A simplifying assumption is that the model-predicted SSI payment would equal the observed SSI payment for the spouse. Since the average of this estimate is small, the sensitivity of the overall estimates to this assumption should be minor.

In conclusion, only about one in five retired-worker beneficiaries with Social Security benefits below the FBR participates in the SSI program. The main reason for nonparticipation is the failure to pass the SSI financial eligibility screens; 70 percent of the total is estimated to be ineligible for SSI.20 We estimate that about 10 percent may be financially eligible, but do not participate.

Economic Well-being Among Elderly Retired-Worker Beneficiaries with Benefits Below the Effective Unit FBR

The results of the previous section imply that the vast majority of elderly retired-worker beneficiaries with Social Security benefits below the SSI FBR are not economically vulnerable if the yardstick of economic vulnerability used is the SSI means test, literally applied. Nevertheless some of those beneficiaries might be classified as economically vulnerable if a broader measure of economic vulnerability, such as poverty status is applied. While the poverty line is not a foolproof "gold standard," and in fact has been subject to methodological criticism,21 the poverty rate is still widely used as a social indicator and is useful for the assessment of broad patterns of economic vulnerability. Whereas the SSI means test has been developed to administer a cash-assistance program and was not designed to serve as a general measure of economic well-being, the poverty line has been explicitly developed and is used for purposes of measuring economic well-being.

Applying poverty status as an indicator of economic vulnerability might result in a different pattern of economic well-being than indicated by SSI financial eligibility status for several fundamental reasons. First, the SSI eligibility rules use the "unit" concept that distinguishes only between "individual" and "couple" status. However, people live in a family, which is widely recognized as the appropriate consumption unit. Thus the presence and income of other family members, as well as other factors—such as economies of scale assumptions—affect comparisons between the two measures. Second, in some sense the SSI income eligibility measure is stricter than the poverty threshold because it ensures only a subpoverty level of income (Koenig and Rupp 2004). Third, the SSI income test is also less strict in some aspects because of exclusions from "countable" income. The test disregards up to $20 of income from any source, up to $65 of any additional earnings from work, and 50 percent of the remainder of earnings. This results in the SSI income test being less strict in certain situations, which is not as important in the context of the present study because earned income is relatively infrequent among the elderly. Fourth, SSI financial eligibility is affected by both an income and an asset test, while poverty status is strictly an income measure. The inclusion of an asset screen makes SSI financial eligibility a stricter measure than it would be if based on the SSI income test alone.

The distribution of all retired workers with Social Security benefits below the SSI unit FBR by family income as a percentage of the poverty threshold is shown in Table 4. The categories therein roughly correspond to various measures of policy relevance. The 75 percent threshold indicates a strict measure of economic vulnerability, providing a useful measure in light of the SSI program's target of guaranteeing income for individual and couple units at a level that is below the poverty threshold; 101–125 percent of the poverty line is often used to identify the "near poor." Various programs—other than SSI—use income eligibility thresholds above 125 percent of the poverty threshold, typically not surpassing 200 percent of the poverty threshold. While the definition of "201 percent or more" as the top family income category is somewhat arbitrary, people with incomes above twice the poverty line form a category that may be considered to represent beneficiaries that are not meant to be targeted by cash-assistance programs that focus on the neediest.

Table 4. Percentage distribution of family income relative to the poverty threshold among Social Security retired-worker beneficiaries aged 65 or older with Social Security benefits below the unit FBR, by SSI financial eligibility status
Family income as a percentage
of poverty threshold
All Social Security
retired-worker
beneficiaries
with benefits below
the effective unit FBR a
Subgroup
SSI eligible Not SSI eligible
Percentage
distribution
Estimated
standard
error
(percent)
Percentage
distribution
Estimated
standard
error
(percent)
Percentage
distribution
Estimated
standard
error
(percent)
75 or below 11 1 27 2 4 1
76–100 20 1 44 3 9 1
101–125 8 1 6 1 8 1
126–200 16 1 12 2 18 1
201 or above 46 2 11 2 61 2
Total percent 100 0 100 0 100 0
N 1,089 370 719
SOURCES: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTES: Standard error estimates reflect the assumption of simple random sampling. See U.S. Census Bureau (2001) for the adjustments that are needed to account for the SIPP sample design effect.
FBR = federal benefit rate; SSI = Supplemental Security Income; N = the unweighted count of the number of observations for the denominator of the estimated percentages; SIPP = Survey of Income and Program Participation.
a. For retired-worker beneficiaries without a spouse present, the individual SSI FBR is used. For retired-worker beneficiaries with a spouse present, the couple SSI FBR is used.

The table displays wide disparities. Almost half (an estimated 46 percent) have family income above twice the poverty threshold, while about 30 percent are poor. Only 11 percent fall into the subpoverty category of 75 percent or less of the poverty threshold, and the proportion "near poor" is relatively small.

Subgroup differences are also informative. In particular, there is substantial heterogeneity by SSI eligibility status. Table 4 compares the distribution for the 70 percent who are not SSI eligible with the distribution of the 30 percent who are SSI eligible. Clearly, the majority of the group that is not eligible for SSI is relatively well off, and only about 13 percent are poor. In contrast, the rate of poverty is 71 percent for the SSI-eligible group. Thus it appears that employing the SSI financial eligibility screen is helpful in identifying those who are economically vulnerable.

How do our subgroups identified by the four principal reasons for SSI nonparticipation compare in economic well-being? We are particularly interested in two aspects: (1) the proportion that is clearly economically vulnerable and (2) the proportion that is clearly not economically vulnerable. Chart 2 compares the five subgroups using a subpoverty threshold (75 percent of poverty line) and the poverty threshold (100 percent of poverty line). Only about 13 percent of the income-ineligible group is poor. In contrast, the proportion poor is around 70 percent for the two subgroups of eligibles and for the group that is ineligible as a result of the SSI asset test alone. There is a notable difference between eligible participants and the other two groups on the stricter 75 percent threshold measure. All but 15 percent of eligible participants have family income higher than the 75 percent subpoverty threshold.22 In contrast, a larger percentage of nonparticipating eligibles and the group ineligible because of the asset test alone have family income at or below the 75 percent subpoverty threshold.

Chart 2.
Percent of elderly retired-worker beneficiaries with Social Security benefits below the unit FBR in each eligibility and participation status category with family income at or below 75 percent and 100 percent of the poverty threshold
Bar chart linked to data in table format.
SOURCE: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTE: FBR = federal benefit rate; SSI = Supplemental Security Income.

What about the proportion of elderly retired workers that appears clearly not economically vulnerable? Chart 3 shows the proportion in each SSI eligibility/ participation category with family income above 200 percent of the poverty line. Not surprisingly the proportion is highest (76 percent) for those ineligible because of both the asset and income screens. Consistent with Table 4, a relatively large portion of those who are income ineligible have family income over 200 percent of the poverty line. The three groups that were characterized by high rates of poverty—eligible participants, eligible nonparticipants, and ineligible nonparticipants as a result of the asset test alone—have relatively low proportions with family income above 200 percent of the poverty line. Interestingly, the "eligible nonparticipant" group that had the highest proportion with family income below 75 percent of the poverty threshold also has the highest proportion above 200 percent of poverty among the three groups mentioned above. Thus family income well above the poverty line may contribute to SSI nonparticipation among eligibles.

Chart 3.
Percent of elderly retired-worker beneficiaries with Social Security benefits below the unit FBR in each SSI eligibility and participation status category with family income above 200 percent of the poverty threshold
Bar chart linked to data in table format.
SOURCE: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTE: FBR = federal benefit rate; SSI = Supplemental Security Income.

The implications of the above findings are less definitive for the subgroup that is income eligible, but is asset ineligible according to current SSI standards. The poverty line measures only income. Thus it is possible that some of those with countable incomes below the FBR but countable assets above the SSI asset threshold might have very large assets and therefore would not be economically vulnerable in a broader sense.23

One way to approach this problem is to perform a sensitivity analysis. One of our sensitivity analyses excludes people whose assets are high enough to label them not economically vulnerable. This allows for the assessment of the economic vulnerability of the remaining group that is asset ineligible, but income eligible using SSI standards. The selection of the asset threshold for this sensitivity analysis is somewhat arbitrary. We present results using the median value of countable assets ($10,000) as the cutoff point.24

Another approach to sensitivity analysis that avoids the use of an arbitrary cutoff point is to transform assets to an income debit and treat this debit as countable income. Rupp and others (2003) and Davies and others (2004) consider this approach in investigating SSI reform options that focus on modifying the asset test.25 This approach results in a modified income screen that compares the sum of countable income under the status quo program and the annuitized value of countable assets to the FBR to establish a simulated SSI eligibility indicator.

Both approaches reduce the size of the "asset-ineligible" target group by half. The remaining half is deemed economically vulnerable for purposes of this sensitivity analysis. The exclusion of those with assets above the median would result in a poverty rate of 68 percent for the remaining subgroup that is deemed economically vulnerable using this technique. Likewise, the exclusion of those who would lose income eligibility as a result of the addition of annuitized assets would result in a poverty rate of 59 percent for a similar subgroup deemed economically vulnerable. Both results are qualitatively consistent with the overall finding of a relatively high poverty rate (70 percent) for the asset ineligible group (Chart 2). The results of the sensitivity tests with respect to the proportion below 75 percent of the poverty threshold also are comparable with the estimate presented in Chart 2 for the asset-ineligible group.

The qualitative conclusion from our sensitivity analysis is that reclassifying people with "high" assets as not economically vulnerable would reduce the size of the "asset ineligible" group judged to be economically vulnerable, but the remainder of the group would contain a relatively high proportion of economically vulnerable persons. Thus there are complex tradeoffs related to asset testing that arise from the conflict between the potential for substantial screening out error under a strict asset-test regime and a potentially salient screening in error in the absence of asset testing with clear implications for administrative complexity.

Another perspective is provided by comparing poverty-related outcomes for the baseline with a hypothetical unit FBR-level minimum benefit. Table 5 provides this comparison for Social Security retired-worker beneficiaries aged 65 or older with benefits below the effective unit FBR. The first panel provides information on poverty outcomes, and the second panel shows subpoverty outcomes using 75 percent of the poverty threshold as the operational measure. In both panels, the first column gives the relevant outcome under the status quo; the second column gives it for the hypothetical minimum benefit, and the third column gives the difference (status quo less hypothetical) in percentage points.26 The difference is a measure of the magnitude of reduction attributable to the hypothetical unit FBR-minimum benefit. Overall, the data show substantial reduction in the proportion below the 75 percent subpoverty threshold (7.1 percentage points, representing a drop of over half of the baseline rate), and a more modest, 2.5 percentage-point reduction in the rate of poverty. This pattern is not surprising because the simulated minimum uses the FBR threshold for individual or couple units, which is below the poverty threshold for couple units and roughly equals 75 percent of the poverty threshold for individual units. The subgroup patterns are not surprising in that the percentage-point reductions are largest for the subgroups that are the most disadvantaged under the status quo by the given outcome measure, although this conclusion does not hold for the poverty outcome in relative terms. Importantly, despite the larger percentage-point reductions, the two subgroups most disadvantaged under the status quo (SSI eligible and ineligible because of assets alone) are clearly the most disadvantaged under the simulated minimum benefit as well.

Table 5. Comparison of poverty outcomes under status quo baseline and hypothetical unit FBR-level minimum Social Security benefit
Subgroup of Social Security
retired-worker beneficiaries
N Percent with family income below
official poverty threshold
Percent with family income below
75 percent of official poverty threshold
Status quo Hypothetical Difference Status quo Hypothetical Difference
All Social Security beneficiaries aged 65 or older with benefits below effective unit FBR 1,089 30.6
(1.4)
28.1
(1.4)
2.5
(0.5)
11.1
(1.0)
4.0
(0.6)
7.1
(0.8)
Of which:
SSI eligible 370 70.7
(2.4)
67.7
(2.4)
3.1
(0.9)
27.1
(2.3)
10.5
(1.6)
16.6
(1.9)
Not SSI eligible because of—
Assets only 73 70.1
(5.4)
65.6
(5.6)
4.5
(2.4)
42.1
(5.8)
10.5
(3.6)
31.5
(5.5)
Income only 199 12.8
(2.4)
11.1
(2.2)
1.7
(0.9)
0
(0)
0
(0)
0
(0)
Both assets and income 447 5.6
(1.1)
3.4
(0.9)
2.2
(0.7)
0.5
(0.3)
0.5
(0.3)
0
(0)
SOURCES: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTES: Standard error estimates (in parentheses) reflect the assumption of simple random sampling. See U.S. Census Bureau (2001) for the adjustments that are needed to account for the SIPP sample design effect.
FBR = federal benefit rate; N = the unweighted count of the number of observations for the denominator of the estimated percentages; SSI = Supplemental Security Income; SIPP = Survey of Income and Program Participation.

In conclusion, retired workers with Social Security benefits below the SSI FBR form a fairly heterogeneous group in terms of economic vulnerability. Almost half of them have family incomes above 200 percent of the poverty threshold. This proportion is particularly high among those who are income-ineligible for SSI, reflecting the importance of income sources other than the retired worker's (and spouse's) Social Security benefit. While SSI participants are often poor, SSI participation is associated with a low proportion of persons in extreme poverty. Two subgroups of retired-worker beneficiaries that stand out with relatively high prevalence of extreme poverty are nonparticipating SSI eligibles and those who are ineligible for SSI as a result of the asset test alone. As noted above, some people in the latter subgroup may appear economically vulnerable on the poverty measure but would not be treated as such by some other measure that would consider both asset levels and current income in defining economic vulnerability in some fashion.

Administrative Simplicity and Effective Targeting: What are the Tradeoffs?

As noted in the Introduction, the idea of providing a Social Security minimum benefit at the SSI FBR level has been suggested by some (for example, Herd 2005) as a method to reach the most economically vulnerable in a manner that is administratively simple and that avoids welfare stigma. Policymakers may consider the tradeoffs between these potential advantages and other relevant factors such as program cost and target efficiency.

Given that only a minority of Social Security retired-worker beneficiaries with benefits below the effective FBR are eligible for SSI and given that SSI participation among eligibles is less than universal, the effect of a minimum benefit at the effective SSI benefit level would be more than merely substituting OASI for SSI benefits on a dollar-for-dollar basis. On the contrary, the net change would be a 25 percent increase in combined OASI and SSI benefits for affected individuals. We estimate the change in total program cost to be around 2 percent of aggregate OASI benefits to all retired workers aged 65 or older, with nontrivial implications for Trust Fund balances.27 Note that these estimates assume no behavioral effects on OASI participation, an assumption that may be more or less valid depending on the specific way an FBR-based minimum benefit might be implemented.28

Although the effective SSI payment standard is below the poverty level, the additional expenditures would not necessarily go to recipients in poverty for two reasons. First, the additional income sources of people with OASI benefits below the effective SSI payment standard may move them out of poverty. Second, people with very low Social Security retired-worker benefits may live in families that are not in poverty because of the income of other family members. We estimate that only 18 percent of the additional hypothetical spending would accrue to poor retired-worker beneficiaries. This figure is low compared with that of the SSI program, which uses income and resource testing to target around 78 percent of program spending to people in poverty.29 The 18 percent figure is also low compared with all but one of the Social Security reform options targeting economically vulnerable elderly beneficiaries analyzed by Anzick and Weaver (2001).30

Given the relatively low-target efficiency of the proposal to raise Social Security benefits to the FBR level, a closer look at the tradeoffs between administrative simplicity and effective targeting is warranted. We are particularly concerned about two types of classification error: (1) incorrectly screening in nonpoor beneficiaries, and (2) incorrectly screening out the "severely impoverished," which we operationalize by classifying beneficiaries with family income below 75 percent of the poverty line as severely impoverished. The choice of using both a poverty and subpoverty level threshold in the analysis is warranted by the fact that SSI was designed to provide subpoverty level income. As previously noted, the FBR is set at 90 percent of the poverty threshold for two-person couple families and 77 percent for one-person families. Thus income above 100 percent of the poverty threshold is clearly above what can be considered as "SSI level," and income below 75 percent of the poverty threshold is clearly below "SSI level." Income between 75 percent and 100 percent of the poverty threshold may be considered as representing a "gray area."

Using our measures of classification error we assess the potential tradeoff between administrative simplicity and effective targeting. We start out with a measure identifying individual retired-worker beneficiaries with Social Security benefits below the effective FBR for individuals as the target population. This is the simplest operational measure in that it requires only the comparison of the individual's Social Security benefit with a constant dollar value regardless of the presence or absence of a spouse, family structure, income, or assets. The individual's Social Security benefit is easily identifiable using Social Security administrative records on a monthly basis. Next we replace this measure with one that uses the "unit" concept of the FBR for individuals or couples.31

We continue our analysis by incrementally adding an income and an asset screen to the effective FBR for the retired-worker unit to explore whether there is a tradeoff between the increased administrative complexity introduced by these additional screens and the accuracy of targeting. We use the SSI income and asset screens for this illustration, but note that there might be some other (perhaps simpler) ways of defining an income and an asset screen for purposes of establishing a minimum Social Security benefit that have somewhat different properties in terms of administrative complexity and targeting error (for example, see Rupp and others 2003; Smeeding 1999).

Table 6 presents the screening properties of four alternative screening scenarios using 100 percent of the poverty threshold as the classification variable. The screening variable categories provide a mutually exclusive and exhaustive classification of all retired-worker beneficiaries aged 65 or older by the combination of poverty status (poor versus nonpoor) and screening status (screened in versus screened out) using the four different screening criteria identified by the rows of the table. Table 7 presents similar statistics using the 75 percent of the poverty threshold measure. While all of the statistics presented in these two tables are interesting and relevant, as noted before, the two most important statistics here are the "percent nonpoor screened in" (Table 6) and the "percent below 75 percent of the poverty threshold screened out" (Table 7). Chart 4 highlights these two key measures that can be seen as error rates in some sense.

Table 6. Percentage distribution of all Social Security retired-worker beneficiaries aged 65 or older, by poverty and screening status; percent of all nonpoor beneficiaries screened in; percent of all poor beneficiaries screened out
Screening variable a Total
percent
Poor Nonpoor Percent
of all
nonpoor
screened
in
Percent
of all
poor
screened
out
Screened
in
Screened
out
Screened
in
Screened
out
OASI benefit below individual FBR 6,666 100.0 4.5 2.8 18.1 74.6 19.6 38.0
OASI benefit below unit FBR 6,666 100.0 4.6 2.6 10.5 82.2 11.3 36.0
OASI benefit below unit FBR plus
SSI income eligible 6,666 100.0 3.9 3.3 1.6 91.1 1.8 46.0
SSI income plus resource eligible 6,666 100.0 3.2 4.0 1.3 91.4 1.4 55.7
N -- -- -- -- -- 6,107 b 559 c
SOURCE: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTES: OASI = Old-Age and Survivors Insurance; FBR = federal benefit rate; SSI = Supplemental Security Income; -- = not applicable.
a. N refers to the unweighted count of Social Security retired workers aged 65 or older.
b. N refers to the unweighted count of nonpoor Social Security retired workers aged 65 or older.
c. N refers to the unweighted count of poor Social Security retired workers aged 65 or older.
Table 7. Percentage distribution of all Social Security retired-worker beneficiaries aged 65 or older, by income below and above 75 percent of the poverty threshold and screening status; percent of all with family income above (below) 75 percent of the poverty threshold screened in (out)
Screening variable a Total percent Income below 75 percent of poverty threshold Income above 75 percent of poverty threshold Percent of all with family income above 75 percent of the poverty threshold screened in Percent of all with family income below 75 percent of the poverty threshold screened out
Screened
in
Screened
out
Screened
in
Screened
out
OASI benefit below individual FBR 6,666 100.0 1.7 0.2 21.0 77.2 21.4 9.6
OASI benefit below unit FBR 6,666 100.0 1.7 0.1 13.5 84.7 13.7 7.9
OASI benefit below unit FBR plus
SSI income eligible 6,666 100.0 1.7 0.2 3.9 94.3 4.0 9.7
SSI income plus resource eligible 6,666 100.0 1.2 0.6 3.3 94.9 3.4 32.8
N -- -- -- -- -- 6,532 b 134 c
SOURCE: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTES: OASI = Old-Age and Survivors Insurance; FBR = federal benefit rate; SSI = Supplemental Security Income; -- = not applicable.
a. N refers to the unweighted count of Social Security retired workers aged 65 or older.
b. N refers to the unweighted count of Social Security retired workers aged 65 or older with family income above 75 percent of the poverty line.
c. N refers to the unweighted count of Social Security retired workers aged 65 or older with family income below 75 percent of the poverty line.
Chart 4.
Distributional effects of four alternative screening scenarios
Bar chart linked to data in table format.
SOURCE: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTE: FBR = federal benefit rate.

First we compare the percentage with an OASI benefit below the individual and unit FBR screens. The differences in terms of administrative complexity are relatively minor here. The unit FBR screen performs unambiguously better on both screening indicators. Compared with the individual FBR screen, the unit FBR screen reduces the percent nonpoor (incorrectly) screened in from 20 percent to 11 percent, and it reduces the percent below 75 percent of the poverty threshold (incorrectly) screened out from 10 percent to 8 percent. These findings support our decision to focus on the properties of the unit FBR measure in earlier sections of this article.32

How does this improved performance of the unit (as compared with the individual) FBR screening variable translate into a reduction in the proportion incorrectly screened in among all who are screened in? A comparison of the first two bars of Chart 5 answers this question. By switching to the unit based screen, the percentage of nonpoor who are screened in is reduced by only about 10 percentage points—from 80 percent to 69 percent. These high percentages of screening-in error are explained by the dominance of the nonpoor in the overall sample of Social Security retired-worker beneficiaries—about 93 percent of all Social Security retired-worker beneficiaries are nonpoor (statistics not shown in the tables). Thus it is not surprising that target efficiency is relatively low even when the unit concept is used as we have seen above.

Chart 5.
Nonpoor as a percent of all those screened in under four different screening scenarios
Bar chart linked to data in table format.
SOURCE: Authors' calculations based on November 1996 data from the Survey of Income and Program Participation.
NOTE: FBR = federal benefit rate.

Given the high percentage of nonpoor incorrectly screened in using the FBR-level benefit screen, one may reasonably ask whether imposing additional screens that may increase administrative complexity (as well as intrusiveness) have potential benefits in terms of improved target efficiency. We address this issue by first adding the SSI income screen and then adding the SSI asset screen incrementally to the unit FBR-level OASI benefit screen. Either of these screens would introduce some means testing, which is arguably not desirable because the OASI program has always been an earned benefit program. However, adding one or both of these screens incrementally has the advantage of avoiding explicit means testing for the top 85 percent of elderly OASI retired-worker beneficiaries—those receiving OASI benefits that already exceed the effective SSI FBR (see authors' calculation from Table 1).33

The next addition—the SSI income screen—reduces the proportion of the nonpoor who are incorrectly screened in from 11 percent to 2 percent (Chart 4). As a result, the proportion nonpoor among those who are screened in drops from almost 70 percent to almost 30 percent (Chart 5). This should be weighed against increased administrative complexity, administrative costs and intrusiveness, as well as against a modest increase in those persons below 75 percent of the poverty threshold who are (incorrectly) screened out from 8 percent to 10 percent (Table 7).34

In contrast, the incremental addition of an SSI-style asset test would reduce the percent nonpoor who are screened in only slightly—from an estimated 1.8 percent to an estimated 1.4 percent (Chart 4)—but would dramatically increase the percent below 75 percent of the poverty threshold who are screened out from 10 percent to 33 percent (Chart 4). As noted earlier, our measure of economic vulnerability is solely income based, and therefore screening out error may be overstated from a broader perspective that considers very high assets to be a legitimate reason for screening out regardless of very low income.35 Overall, while the incremental addition of the SSI asset test would reduce program cost somewhat—as the introduction of any additional screen is expected to do—this is to be balanced against increased administrative complexity, increased administrative costs, possibly increased screening out error, potential additional welfare stigma, and other negative factors. Among these other factors we acknowledge behavioral effects widely discussed in the literature. Perhaps the most problematic is the "spend-down" effects of the asset test: marginally disqualified people face strong incentives to reduce assets to a level that is below the applicable threshold.36 In addition, certain asset classes (housing, automobile) are favored through exclusions, while defined contribution pensions are not favored.37

In summary, the potential advantages of the proposal to raise the minimum Social Security benefit for retired workers to the level of the SSI federal income guarantee are to be balanced against potential disadvantages. The disadvantages include relatively large program cost and relatively low target efficiency. Modifying the proposed approach by introducing some additional income screening could result in reduced program cost and increased target efficiency, but at the expense of increased administrative complexity and the possibility of an increase in perceived welfare stigma. As we have seen, however, if income screening is to be implemented incrementally (in addition to a "prescreening" based on OASI administrative records), about 85 percent of elderly Social Security retired-worker beneficiaries would not be subjected to this additional, explicit, test. Although our results are less definitive concerning asset testing, they suggest that the incremental addition of an asset test (in addition to an income test) might substantially increase screening out error without obvious gains in program cost or target efficiency. These results concerning the asset test are less definitive than the findings concerning the effects of prescreening based on administrative records or income screening; some who appear to show high-economic vulnerability based on current income may not be regarded as such once spend down (or potential spend down) from assets is explicitly considered. More work is needed on studying alternative approaches to asset testing and on examining the relationship between income and asset testing.

Concluding Comments

In this article we focused on the SSI FBR as a potential basis for designing a minimum Social Security benefit and limited our attention to elderly Social Security retired-worker beneficiaries. Future research may consider a broader range of minimum benefit proposals, as well as additional target groups. We briefly discuss these potential extensions.

Analysts have proposed the poverty standard as a potential basis for evaluating the adequacy of Social Security benefits and as a basis for determining a minimum Social Security benefit.

Although the poverty threshold is somewhat more generous than the FBR, the tradeoffs related to administrative implementation appear very similar. If one were to use the poverty threshold for a one- or two-person family—depending on the presence of a spouse—as a minimum benefit threshold, the policy implementation would be similarly simple as with an FBR-level minimum benefit. The policy would, however, increase program cost even more than an FBR-level minimum benefit and would be even less target efficient. The tradeoffs between administrative complexity, program cost, target efficiency, and potential welfare stigma should also be very similar. Thus, the conclusions of this article seem applicable to a broader array of approaches that focus on the poverty measure.

Future research may also explore the SSI FBR as a measure of benefit adequacy and as a potential tool for establishing a minimum Social Security benefit for groups other than retired workers aged 65 or older. The most important of these other groups are elderly widow(er) beneficiaries, many of whom are economically vulnerable and a relatively high proportion of whom are SSI recipients (Rupp and others 2003). Other groups of Social Security beneficiaries, such as retired persons who retired before reaching the full retirement age and have not reached it by the survey reference month, raise additional policy issues not addressed in this article. Workers are eligible for early retirement beginning at age 62, and 8.7 percent of all retired-worker beneficiaries are aged 62 to 64.38 This age group is not automatically eligible for SSI (for persons aged 64 or less, a disability screen also has to be met), and early retirees are subject to an actuarial reduction of their Social Security benefit. Thus, whether and how to implement an FBR-related minimum benefit for these beneficiaries raises important additional issues. Another important group, disabled-worker beneficiaries, differs from retired-worker beneficiaries in many relevant aspects. Of particular relevance in the context of this study is that earned income is more important in this working-aged group than among retired workers as a result of the presence of nondisabled spouses and other family members. The practical effect is that some disabled-worker beneficiaries have family income well above the poverty threshold. However, others, particularly those who are living alone, may have little or no income from sources other than Social Security, SSI, and the Food Stamp program.39 There are other issues related to smaller groups of beneficiaries, such as the workers' compensation offset, that would need to be carefully considered in terms of tradeoffs related to administrative simplicity.

Aggregate program cost could be substantially higher if these additional groups are considered as well. Each group would also raise somewhat distinct issues about potential behavioral effects, a subject we did not address here. Nevertheless, many of the qualitative findings in this study are expected to hold for each of these additional groups of Social Security beneficiaries.

Another direction for future research would be to examine Social Security minimum benefits in the context of the transition to a solvent Social Security system. The broad-based minimum benefit proposals we focused on in this analysis were treated in the context of the current, status quo, safety net for the elderly. In contrast, Social Security solvency proposals often include minimum benefits targeted toward individuals with long work histories but with low levels of earnings and thus low Social Security benefits. A recent study by Favreault and others (2006) discusses both types of proposals, but provides longer-term estimates only for a set of reform scenarios with highly targeted minimum benefits. A logical follow-up study would assess the more universal minimum benefit proposals in the context of the long-term solvency of the Social Security system. Other things equal, movement toward a solvent Social Security system would be expected to increase the subset of beneficiaries that may qualify for the minimum benefit. However, the prevalence of poverty among the elderly is expected to decrease in the long run for reasons other than the reforms themselves, and this reduction may be substantial (Favreault and others 2006). Learning about the net effect of these opposing factors would be helpful for assessing the pros and cons of alternative minimum benefit proposals.

The relationship between Social Security and the SSI program may also be a subject for fruitful additional research. A related avenue for future research in evaluating the economic well-being of the elderly would be to further explore the role of assets owned by the elderly. This study used the standard poverty measure, which is based on income. The findings reported herein suggest that an asset screen could screen out many economically vulnerable people, yet the standard poverty measure fails to consider the characteristics and use of those assets in providing economic support.

Notes

1. An important context here is the establishment of the SSI program (Public Law 92-603, enacted October 30, 1972) that offers a guaranteed income floor for all elderly Americans who meet an asset test. The first SSI payments were made in January 1974.

2. See also FitzPatrick and others (2003) and Diamond and Orszag (2004). Sandell and others (1999) simulate the estimated effects of minimum benefit provisions similarly structured as the subsequent Kolbe-Stenholm plan. Favreault and others (2006) provide a comprehensive review of recent minimum benefit proposals, most of them conditioning the benefit guarantee on years of covered earnings.

3. The original Social Security minimum benefit was similar to these more recent proposals in that it established a broadly applicable benefit floor, but differed in that it was not tied to a measure of benefit adequacy such as the official poverty threshold or the SSI FBR. We note, however, that the original minimum benefit amounted only to 56 percent of the individual FBR in 1980 (authors' calculations based on Kollmann (2000) and the Annual Statistical Supplement to the Social Security Bulletin, 2002, Table 2.A27 and Table 2. B1). The concerns leading to the "freezing" of the original minimum benefit by the 1972 amendments were largely influenced by the perceived windfalls that would have otherwise occurred under the 1972 act as a result of large anticipated increases in the minimum benefit relative to the poverty line.

4. Favreault and others (2006) discuss these proposals in more detail.

5. These rates apply to individuals and couples living in their own household. The FBR for individuals and couples living in the household of another is lower. SSI rules also establish a separate (much lower) FBR for persons living in Medicaid institutions. In the Social Security minimum benefit simulations, we use the FBR for individuals and couples living in their own household. Note that Social Security administrative records do not contain information on living arrangements unless the beneficiary is a concurrent recipient of SSI.

6. The 2006 FBR is 4.1 percent higher than the 2005 figures cited in the text. The 2006 FBR is $603 for eligible individuals and $904 for eligible couples. The corresponding 2007 values are $623 and $934, respectively. This represents an additional 3.3 percent COLA increase. In 2008, the FBR is $637 for individuals and $956 for couples.

7. Insured status for OASI benefits generally requires 40 quarters of Social Security–covered employment, which is roughly equivalent to 10 years of employment.

8. Salkever and others (2006) formalize judgments about the relative importance of these two sources of error in a cost-benefit framework.

9. Note, however, that spouses are considered in measuring individual or couple unit status and in measuring Social Security benefits and SSI payments. Also, income-based measures consider the income of other family members.

10. SSI policy generally refers to "resources" rather than "assets" as is common in the analytic literature. Assets generally only involve an ownership test, but there is both an ownership and availability test for resources as defined by SSI program rules. Thus, while all resources are assets, not all assets are resources. In this article we use SSI rules for identifying countable resources, but often use the broader term of "assets" throughout the study to clarify the analytic distinction between "income" and "assets" as these terms are defined by economists.

11. In our simulations of FBR-level minimum Social Security benefits, we consider only the SSI federal cash benefit guarantee for individuals and couples living in households. We ignore SSI rules that reduce SSI payments because of the receipt of in-kind support and maintenance. Note that the SSI program also includes optional state cash benefits, and SSI recipiency status enters into the determination of eligibility for various in-kind benefits, such as Medicaid, food stamps, and housing assistance. Although important in their own right, none of these features of the SSI program are relevant to measuring an FBR-level minimum Social Security benefit.

12. The qualitative results are fairly robust to the use of the traditional SSI FBR or the "effective" FBR measure.

13. In tables focusing on estimated means for various population segments, we provide standard error estimates that assume simple random sampling (SRS). Because the SIPP has a complex sample design, these estimates tend to underestimate the true standard errors. See U.S. Census Bureau (2001) for the adjustments that are needed to account for the SIPP sample design effect.

14. Note, however, that we had to make some assumptions about participation rates under alternative minimum benefit scenarios. We used the simplifying assumption of 100 percent participation, which is reasonable as long as receiving the minimum does not require any action other than the standard application for Social Security benefits. For some minimum benefit scenarios involving income or asset testing, this may be an upper bound depending on how the application process is operationalized.

15. This estimate reflects the U.S. noninstitutional population in November 1996 as measured by the SIPP. We do not have comparable record data accounting for both the institutional and noninstitutional population for November 1996. We were able to derive the corresponding statistics from the 100 percent Master Beneficiary Record for December 2003; this estimate is somewhat lower, 19.4 percent. The differences may reflect a combination of true differences in the proportion and SIPP measurement error. Because the SIPP excludes the institutional population and has sampling error, some difference is expected even if there were no true differences in the November 1996 and December 2003 proportions.

16. This amounts to roughly 3 percent of all Social Security retired-worker beneficiaries aged 65 or older in the U.S. noninstitutional population.

17. The standard error estimate assuming SRS is 0.3 percent. See U.S. Census Bureau (2001) for the appropriate adjustment for the SIPP design effect.

18. By expected payment we mean hypothetical benefits that would be paid conditional on application and award. These amounts are calculated from the FEM on the basis of SIPP data for participating and nonparticipating eligibles using an identical algorithm. Note that the model-predicted hypothetical average benefit is virtually identical to the actual average for participants, suggesting that the model produces fairly accurate estimates.

19. This is calculated by dividing the model-predicted average SSI benefit of $91 for nonparticipants by the average Social Security benefit ($393) of the retired worker and taking percentages (authors' calculations, 1996 dollars).

20. We note that some of those who are estimated not to be financially eligible, in fact, do participate in SSI. About 1.8 percent of all beneficiaries with Social Security benefits below the SSI FBR are estimated to participate in SSI, although they are classified as financially ineligible by the FEM. This amounts to about 8 percent of all SSI beneficiaries in the sample. The discrepancy may be attributed to a combination of measurement error (reporting error in SIPP or error in the measurement of financial eligibility in the FEM) and possible financial ineligibility among SSI participants. The overall results are highly robust to the possible misclassification of some participants as financially ineligible. We also note that some participants and nonparticipants we classify as financially eligible may in fact be ineligible because of the same measurement problems, but we do not have additional data to gauge the potential magnitude of this problem.

21. See Citro and Michael (1995) for an overview of poverty measurement issues. U.S. Census Bureau (2005) provides poverty estimates using alternative measures of poverty. Koenig and Rupp (2004) analyze the robustness of using the official poverty measure by comparing it with a three-parameter experimental scale in estimating poverty outcomes for SSI recipients and discuss the economies of scale assumptions of SSI program design and alternative poverty measures. Rupp and others (2003) use three alternative poverty measures as tools for examining the effects of SSI reform options on elderly women. Zagorsky (2004) develops alternative measures of poverty that considers both income and wealth. Hurd and Rohwedder (2005) compare income- and consumption-based poverty measures and address the implications of problems with survey measures of asset income for poverty measurement. Koenig and others (2004) simulate the effects of converting imputed asset income to countable income in calculating SSI financial eligibility of the elderly on the distribution of income relative to the poverty threshold.

22. For both single and married couple units, the FBR is above the subpoverty threshold. Thus, it may sound counterintuitive for SSI participants to have family incomes below 75 percent of the poverty line, but there are several legitimate reasons. Perhaps most importantly, SSI recognizes only "individual" and "couple" units, while the poverty line is family based. Thus if there is an additional person in the family who is not part of the SSI unit (such as the sibling of an elderly SSI beneficiary) with zero income, family income may drop below 75 percent of the poverty line. Of course, SIPP measurement error may also result in family income measured to appear lower than 75 percent of the poverty line.

23. Accounting for assets in assessing the economic vulnerability of the elderly is an issue with wider implications for policy evaluation of the relationship between aging, widowhood, and economic vulnerability among the elderly. The conventional wisdom—derived from studies using income-based measures of economic vulnerability—is that the older subgroups of the elderly (for example, those aged 80 or older) are much more economically vulnerable than their younger peers. However, using consumption-based measures, Hurd and Rohwedder (2005) suggest that these discrepancies might be substantially smaller once the effects of the life-cycle patterns of asset accumulation and spend down on current consumption are accounted for. Zagorsky (2004) finds that the elderly are among the population subgroups whose poverty status is relatively sensitive to the consideration of wealth in addition to income.

24. We considered different methods to establish a cutoff point. The $10,000 value seems reasonable from three different perspectives. First, it roughly corresponds to the inflation-indexed value (to account for changes in prices) of the 1974 SSI asset thresholds for individual and couple units that have not changed at all from 1974 to date in nominal terms. Second, even if one takes a generous view of the income-producing capacity of $10,000 countable assets (a conservative assumption in this context), the imputed monthly income stream would be relatively low. For example, with a 12 percent annual nominal rate of return the imputed monthly asset income would be roughly $100. This is reasonably low in light of the average of $374 SSI-countable income for this subgroup. Adding the $100 imputed asset income results in an average of $474, a value that is below the effective FBR for both individuals and couples in 1996 (all numbers are in nominal 1996 dollars.) Finally, $10,000 is the median, a statistic with a clear intuitive meaning of representing "the middle."

25. See also Zagorsky (2004) for the sensitivity of the amortization of asset stocks to income flows to assumptions about the interest rate and time horizon.

26. Note that the percentages reflect the subuniverse of retired-worker beneficiaries aged 65 or older with benefits below the effective unit FBR. Because only 15.2 percent of the universe of retired workers aged 65 or older have benefits below the effective unit FBR (Table 1), the percentages in Table 5 translate into much smaller percentages relative to this broader universe.

27. These estimates assume instantaneous change in Social Security benefit and SSI payment amounts for retired-worker beneficiaries and their spouses (if any) as a result of the simulated policy change. In this analysis we used November 1996 data from the SIPP matched to Social Security administrative records. The analysis was limited to retired workers aged 65 or older. For sample members, we calculated the status quo monthly Social Security benefits received (primary and secondary benefits combined), the individual's SSI payment received, and the same quantities for a spouse beneficiary as applicable. For SSI recipients with an eligible spouse, half of the SSI couple benefit was allocated to the sample member and half to the spouse.

For the simulation scenario, we conducted separate analyses for (a) retired workers without a spouse beneficiary, and (b) retired workers with a spouse beneficiary. For retired workers without a spouse beneficiary we identified those sample members whose monthly Social Security benefit were less than the individual FBR + $20. For 1996 this amounted to $490 per month. We assumed that the Federal SSI payment for the individual was to be eliminated under the simulation scenario, and the individual's total Social Security benefit was raised to $490 per month in 1996. We calculated the net trust fund cost for the individual as $490 less the individual's combined Social Security benefit and federal SSI payment under the status quo. We created an aggregate amount of net trust fund cost by multiplying the per retired-worker cost with the weighted total of the retired workers. For retired workers with a spouse beneficiary, we used a similar procedure applied to the presumed "couple unit" and used the couple FBR in the calculations. We allocated 50 percent of the net trust fund cost (net benefit increase) for the couple to the sample retired worker. Finally, we summed the aggregate net trust fund cost estimate for the above two groups of retired workers.

Next we calculated the aggregate net trust fund cost estimate for retired workers as a percentage of the status quo and aggregate Social Security and federal SSI costs for the affected retired workers, as a percentage of the status quo aggregate Social Security benefit amount for all retired workers and as a percentage of status quo aggregate federal SSI payments for all retired workers. The denominators for these percentages were derived on the basis of the November 1996 SIPP sample to assure internal consistency.

28. Interactions with Social Security's early retirement program seem particularly relevant here. An FBR-based minimum Social Security benefit without changes related to the early retirement program may produce strong incentives for early retirement among low-income individuals. One possible way to deal with the issue would be to implement a minimum benefit that preserves an actuarial reduction for those who choose early retirement. The detailed discussion of interactions with early retirement is beyond the scope of this study.

29. For more information, see Davies and others (2004).

30. One of the options analyzed by Anzick and Weaver (2001) has an estimated target efficiency of 14 percent. However, the target efficiency of the other four options ranges from 28 percent to 35 percent.

31. In the previous sections, we focused on this second concept because it is relatively close to the "individual" measure in terms of administrative complexity, however it is based on a more reasonable assumption about economies of scale and is closer to the SSI approach. Nevertheless, we note that a number of operational issues would arise if a couple status measure would become an integral part of establishing the minimum benefit eligibility status of retired-worker beneficiaries.

32. There are other relevant differences between these two measures that we acknowledge, but do not focus on here. Most importantly the "unit-based" measure is not neutral with respect to marital status, but arguably more reasonable in terms of the underlying consumption economies of scale assumption.

33. We realize that the unit FBR approach may require some changes in the way marital status is represented in administrative records or in the use of administrative records currently available with the understanding that this may introduce some measurement error.

34. Although screens for targeted assistance programs are designed to exclude those who are not intended to benefit from the policy, they may also have the unintended consequence of excluding some who are targeted. In our case, the SSI income screen is based on a unit concept considering an individual and the spouse if present, but not other members of the family. If for example, there is an additional elderly family member with zero income, the unit income may not pass the SSI income screen but family income may be below 75 percent of the applicable poverty threshold.

35. Smeeding and Weaver (2002) consider whether their Senior Income Guarantee (SIG) proposal should include an asset test. They note that the Canadian Guaranteed Income Supplement program does not include an asset test, but argue that such an approach is probably not politically viable in the United States. Thus the SIG proposal includes an asset test more generous than the SSI asset test, with thresholds of $20,000 in liquid assets for an individual and $30,000 for a couple, indexed for inflation for future years.

36. There is considerable literature on the negative effects of means-tested programs on savings. Neumark and Powers (1998) provide evidence suggesting that SSI reduces savings among households with heads who are near elderly and who are likely participants in the program.

37. Defined contribution (DC) assets are countable. Defined benefit (DB) pensions are considered only in the income test. For more information, see Parent (2006).

38. This statistic represents retired-worker beneficiaries in current-pay status. A much higher proportion of the stock of beneficiaries in current-pay status receives reduced benefits (72 percent) as a result of early retirement. The major reason is that about half of new retired-worker awardees are aged 62 or older, and some additional retired-worker beneficiaries are awarded benefits before reaching the normal retirement age. The vast majority of this inflow of early retirees stay in the program beyond the normal retirement age with reduced benefits. In 2003 about 69 percent of new retired-worker awardees were aged 62–64. An additional 9 percent were converted from the DI program with full benefits, and 22 percent were new awardees aged 65 or older (authors' calculations based on the Annual Statistical Supplement to the Social Security Bulletin, 2004, Table 5.B5 and Table 6.B5).

39. Technically, food stamps are regarded as in-kind benefits and are not included in traditional income measures. However, there is a wide consensus among policy analysts that food stamp benefits are highly liquid and therefore better regarded as cash-like benefits. Policy analyses sometimes use income and poverty measures that treat food stamp benefits as cash income. The National Research Council's panel on poverty measurement recommended the inclusion of food stamps (and other "near-money" in-kind benefits) in their proposed measure of family resources (Citro and Michael 1995, 66). The qualitative results of our study appear invariant to the inclusion or exclusion of food stamp benefits in measuring family income and poverty status.

References

Anzick, Michael, and David A. Weaver. 2001. Reducing poverty among elderly women. ORES Working Paper, No. 87, Social Security Administration, Office of Policy, Office of Research, Evaluation, and Statistics, Washington, DC (January), 1–25.

Citro, Constance F., and Robert T. Michael, eds. 1995. Measuring poverty: A new approach. Washington, DC: National Academy Press.

Davies, Paul S., Minh Huynh, Chad Newcomb, Paul K. O'Leary, Kalman Rupp, and Jim Sears. 2002. Modeling SSI financial eligibility and simulating the effect of policy options. Social Security Bulletin 64(2): 16–45.

Davies, Paul S., Kalman Rupp, and Alexander Strand. 2004. The potential of the Supplemental Security Income program to reduce poverty among the elderly. Journal of Aging and Social Policy 16(1): 21–42.

Diamond, Peter A., and Peter R. Orszag. 2004. Saving Social Security: A balanced approach. Washington, DC: Brookings Institute Press.

Favreault, Melissa M., Gordon B.T. Mermin, and C. Eugene Steuerle. 2006. Minimum benefits in Social Security. Project report for AARP Public Policy Institute, #2006–17. Washington, DC: AARP, 1–66.

Feinstein, Craig A. 2000. Projected demise of the special minimum PIA. Actuarial Note No. 143, Office of the Actuary. Baltimore, MD: Social Security Administration (October), 1–8.

FitzPatrick, Christina Smith, Catherine Hill, and Leslie Muller. 2003. Increasing Social Security benefits for women and men with long careers and low earnings. Paper presented at IWPR's Seventh International Policy Research Conference: Women Working to Make a Difference, Washington, DC: National Women's Law Center, 1–11.

Graham, Lindsey. 2003. Social Security solvency and modernization act of 2003, S. 1878, 108th Congress. http://www.govtrack.us/congress/bill.xpd?bill=s108-1878.

Herd, Pamela. 2005. Ensuring a minimum: Social Security reform and women. The Gerontologist 45(1): 12–25.

Hurd, Michael, and Susann Rohwedder. 2005. Consumption and economic well-being at older ages: Income- and consumption-based poverty measures in the HRS. Paper prepared for the 7th Annual Conference of the Retirement Research Consortium: Towards a Secure Retirement System, Washington, DC (August 11–12).

Koenig, Melissa, Mark Nadel, Kalman Rupp, Jim Sears, Alexander Strand, Steve Wamhoff, and Michael Wiseman. 2003. Options and tradeoffs in strengthening Supplemental Security Income for the elderly. Paper presented at the Annual Meeting of the Association for Public Policy Analysis and Management, Washington, DC (November).

Koenig, Melissa, and Kalman Rupp. 2004. SSI recipients in households and families with multiple recipients: Prevalence and poverty outcomes. Social Security Bulletin 65(2): 14–27.

Kolbe, Jim, and Charlie Stenholm. 1999. Twenty-first century retirement act of 1999, H.R. 1793, 106th Congress. http://www.govtrack.us/congress/bill.xpd?bill=h106-1793.

Kollmann, Geoffrey. 2000. Social Security: Summary of major changes in the cash benefit program. Washington, DC: Congressional Research Service (May 18). http://usinfo.state.gov/usa/infousa/society/socwelf/ss2000.htm.

McGarry, Kathleen. 1996. Factors determining participation of the elderly in Supplemental Security Income. Journal of Human Resources 31(2): 331–358.

———. 2000. Guaranteed income: SSI and the well-being of the elderly poor. NBER Working Paper No. 7574, Cambridge, MA: National Bureau of Economic Research (March), 1–37.

Neumark, David, and Elizabeth T. Powers. 1998. The effect of means-tested income support for the elderly on pre-retirement saving: Evidence from the SSI program in the U.S. Journal of Public Economics 68(2): 181–206.

Olsen, Kelly A., and Don Hoffmeyer. 2001/2002. Social Security's minimum benefit. Social Security Bulletin 64(2): 1–15.

OECD. 2007. Pensions at a glance: Public policies across OECD countries. Paris, France: Organisation for Economic Co-operation and Development.

Parent, Rene. 2006. Defined contribution pension plans and the Supplemental Security Income Program, OP Policy Brief No. 2006-01, Washington, DC: Social Security Administration, Office of Policy.

President's Commission to Strengthen Social Security. 2001. Strengthening Social Security and creating financial wealth for all Americans: Report of the president's commission. http://www.ssa.gov/history/reports/pcsss/reports.html.

Rupp, Kalman, Alexander Strand, and Paul Davies. 2003. Poverty among elderly women: Assessing SSI options to strengthen Social Security reform. Journal of Gerontology: Social Sciences 58B(6): S359–S368.

Salkever, David S., Stephen Johnston, Mustafa C. Karakus, Nicholas Ialongo and Eric Slade. 2006. Using target efficiency to select program participants and risk-factor models: An application to child mental health interventions for preventing future crime. NBER Working Paper No. 12377, Cambridge, MA: National Bureau of Economic Research (July), 1–37.

Sandell, Steven H., Howard M. Iams, and Daniel Fanaras. 1999. The distributional effects of changing the averaging period and minimum benefit provisions. Social Security Bulletin 62(2): 4–3.

Smeeding, Timothy M. 1999. Social Security reform: Improving benefit adequacy and economic security for women. CRR Working Paper No. 1999–16, Chestnut Hill, MA: Center for Retirement Research, 1–25.

Smeeding, Timothy M., and R. Kent Weaver. 2003. The senior income guarantee (SIG): A new proposal to reduce poverty among the elderly. CRR Working Paper No. 2001–12, Chestnut Hill, MA: Center for Retirement Research at Boston College (December).

Thompson, Lawrence H. 2004. Social Security reform and benefit adequacy. Urban Institute Brief No. 17, Washington, DC: Urban Institute (March), 1–7.

U.S. Census Bureau. 2001. Source and accuracy statement for the 1996 public-use files from the Survey of Income and Program Participation. Washington, DC: U.S. Census Bureau. http://www.sipp.census.gov/sipp/sourceac/s&a96_040501.pdf.

U.S. Census Bureau. 2005. The effects of government taxes and transfers on income and poverty: 2004. Washington, DC: U.S. Census Bureau. http://www.census.gov/hhes/www/poverty/effect2004/effect2004.html.

Wasow, Bernard. 2004. A new minimum benefit for Social Security. The Century Foundation Idea Brief, New York, N.Y.: The Century Foundation (April). http://www.socsec.org/publications.asp?pubid=463.

Zagorsky, Jay L. 2004. Measuring poverty using both income and wealth. Journal of Income Distribution 13(3–4): Fall 2004–Winter 2005.