SSI Recipients in Households and Families with Multiple Recipients: Prevalence and Poverty Outcomes

Summary

Supplemental Security Income (SSI) is designed to provide an income floor for qualifying individuals and couples. Analysis of data from the Survey of Income and Program Participation (SIPP) matched to Social Security Administration records indicates that one in five SSI recipients lives with at least one other SSI recipient who is not a spouse. This estimate is much higher than previously thought.

Program rules implicitly assume that substantial economies of scale in consumption arise from married couple recipients living in the same household. Therefore, couples are subject to a federal income guarantee equal to 150 percent of the federal benefit rate (FBR) for individuals. However, no economies of scale are assumed to arise from nonmarried-couple recipients sharing a household—each recipient (some of whom may be a member of a cohabiting couple) is guaranteed the full individual FBR. In effect, the federal income guarantee for nonmarried multiple recipients (noncouple multirecipients, or NCMs) can be expressed as the product of the individual FBR and the number of recipients in the household.

In contrast to SSI rules, the official poverty measure and popular experimental alternatives are neutral with respect to marital status and assume that economies of scale result from family or household size and other factors. However, those assumptions are hotly debated, and it is not immediately obvious what yardstick is appropriate to use in measuring the effects of SSI program rules on poverty among SSI recipients.

Beyond describing NCM families and households, this article focuses on the effects that current program rules have on poverty among SSI recipients. It assesses the sensitivity of those distributional outcomes to the unit of observation (SSI unit versus family versus household), the time frame of income measurement (monthly versus annual), and the scale used to measure poverty (the federal poverty threshold versus a three-parameter experimental scale). The apparently lower prevalence of poverty among NCMs compared with married couples is not sensitive to such measurement issues and is essentially driven by assumptions imbedded in the SSI benefit formula regarding the relationship between living arrangements and economies of scale.

The outcomes for individuals who do not live with another SSI recipient are somewhat more ambiguous, but in general poverty is more prevalent in this group than among married couples or NCMs. The magnitude of the differences is sensitive to assumptions regarding economies of scale. The high prevalence of poverty among individual SSI recipients is concentrated among recipients living alone; poverty among those who live with nonrecipients tends to be substantially reduced by the income of the nonrecipients. Distributional outcomes are further affected by age group (elderly versus disabled adult versus disabled child) and can vary substantially with the number of people living in the household or family and with differences in the earned and unearned income of those who have no direct link to the SSI program.

Introduction

This article provides new estimates of the prevalence of households with two or more nonmarried recipients of Supplemental Security Income (SSI)—estimates based on more accurate data and on analysis of the poverty status of families and households with multiple SSI recipients. Three types of households are examined:

• One-recipient households,
• Households with two married SSI recipients and no other recipients, and
• Households with multiple recipients other than married couple recipients.

The last category is referred to as noncouple multirecipient (NCM) households. The recipient or recipients in each category may or may not live with nonrecipients. SSI recipients are also studied by age group: children (age 17 or younger), working age (18 to 64), and elderly (65 or older). Because of differences in the basis of eligibility, deeming rules, and patterns of income sources and resources of these age groups, separate analyses of each can provide important insights concerning the effects of NCM households on poverty among SSI recipients.

The analysis is based on the 1996 panel of the Survey of Income and Program Participation (SIPP) matched to the Social Security Administration's Supplemental Security Record (SSR) at the individual level.¹ The SSR contains highly accurate information on the receipt and amount of SSI benefits, including both federal benefits and federally administered state supplements, but it lacks comprehensive information on household relationships. In contrast, SIPP self-reports of the receipt and amount of SSI benefits are somewhat less accurate, but the survey is designed to provide highly accurate information on other sources of income, as well as on household and family composition. In particular, the SIPP data are used to give a complete picture of household structure, including some information on cohabitation.

Unless otherwise noted, data are taken for 3 reference months: March 1996, January 1998, and December 1998. In the analyses of microdata, SSI receipt and SSI benefit refer to the sum of the federal SSI benefit and any federally administered state benefits.² Annual poverty measures include income in the reference month and the following 11 months. All of the estimates are weighted to reflect the noninstitutionalized population of the United States.

Background and Motivation

The SSI program provides a nationwide cash assistance safety net for elderly and disabled individuals and couples with limited financial means. Most SSI beneficiaries are also eligible for Medicaid. The federal income guarantee (the federal benefit rate, or FBR) is not designed to provide couples or individuals living alone with an income equal to the federal poverty level, although it may raise income above that threshold, sometimes substantially, when combined with other sources of family or household income. Such sources might include supplemental state SSI benefits, earned income, and non-SSI unearned income.

For an individual or couple to receive federal benefits, three conditions have to be met: the person or couple must be aged 65 or older, or be disabled or blind at any age; have countable resources at or below a legislated threshold; and have countable income below the FBR. The first two screens affect eligibility for federal cash benefits but not the amount; the third affects both eligibility and amount. Specifically, the federal cash benefit equals the FBR minus countable income.

Current SSI rules recognize only spousal and parent-child relationships when assessing eligibility and benefits. Spousal relationships are essentially limited to legally married couples, although SSI rules also treat a man and woman who live in the same household as married if they hold themselves out as husband and wife to the community in which they live.³ Other types of relationships between SSI recipients and household members do not affect benefit payments unless the recipient lives in the household of another person and receives in-kind support and maintenance from that person. Thus, SSI units are classified as either individuals or married couples, with no explicit recognition of NCMs, although recipients may well live in a family or household with additional persons and additional recipients who are not part of a parent-child or husband-wife relationship.⁴

SSI rules guarantee married couples 150 percent of the individual FBR, while guaranteeing each recipient in an NCM household the full individual FBR. In addition, NCM households that include children benefit from the assumptions regarding economies of scale (which guarantees children the full individual FBR), as well as from the fact that the guardians of childhood beneficiaries may have substantial earnings without disqualifying the child, while earnings among adult beneficiaries are inherently limited because of qualifying disabilities or old age. Regarding the treatment of widows and other single recipients living alone, the FBR for such individuals is 67 percent of the FBR for married couples, whereas the federal poverty threshold for one-person families is around 80 percent of the threshold for two-person families.

Research Questions and Methodology

The following sections describe the research questions and methodology used to determine the prevalence of NCM households and to identify the resulting distribution of SSI benefits and their effects on poverty among recipients.

Prevalence of Multirecipient Households

Previous estimates of the prevalence of households with two or more nonmarried SSI recipients were based on administrative records and an internal Social Security Administration survey conducted by the Office of Quality Assurance (Kochhar and Scott 1997; Social Security Administration 1999). Relying on these data to establish household relationships tends to underestimate multirecipiency, for a number of reasons: SSI rules do not explicitly recognize all household relationships and therefore all relevant data are not collected, SSI applicants may not accurately report marital status and parent-child relationships, and updating records to reflect changing household patterns is burdensome and expensive.

One innovative and inexpensive approach to supplementing the administrative data uses address-matching techniques to link the Supplemental Security Records of household members.⁵ Another approach is the Office of Quality Assurance's survey of SSI recipients, who are identified from administrative records.⁶ Both methods lack an independently derived, reliable, and comprehensive household roster to guide the collection and analysis of information. Because these methodologies are more likely to result in undercounting rather than overcounting, the net effect is believed to be an underestimation of the number of people in NCM households.

The structure of SIPP helps solve some of the problems encountered in previous studies. The establishment and updating of a household roster is a cornerstone of the survey, and household relationships (relative to a reference person) are established independent of and prior to recording receipt of SSI benefits.⁷ However, SIPP's self-reports of SSI recipiency status contain nontrivial measurement errors, whereas administrative records are highly accurate at measuring monthly SSI recipiency at the individual level (Huynh, Rupp, and Sears 2002).

This study is based on an exact match of SIPP's household information to administrative information on SSI receipt. The analysis thus capitalizes on the strengths of both SIPP and administrative data in a manner that would not be feasible without the matched data set. In addition, the study technique is low-cost because it does not require conducting a special-purpose SSI survey of households.

Multirecipiency and Distributional Outcomes

The SSI program rule guaranteeing married couples only 150 percent of the FBR for individuals reflects the policy assessment that economies of scale in consumption arise from married couples sharing household expenses, compared with individuals living alone. For eligible individuals other than couples regarded to be married by SSI rules, economies-of-scale considerations do not typically affect benefits.⁸ While one can question the reasonableness of SSI rules taken by themselves, the question of balancing fairness and cost-effective targeting needs to be assessed in the broader contexts of the family and the household. This article presents such an assessment, based on more accurate and comprehensive data than have been available for previous analyses.

The article first analyzes the federal income guarantee in the context of SSI program rules, without regard to family or household context. Then it considers SSI in the context of other sources of household income—after all, the FBR was designed to provide a guaranteed income, not a guaranteed level of SSI benefits. This is an important distinction, because eligibility for SSI payments depends on other sources of income presumed to be available to the individual or couple and because the SSI payment formula reduces the federal benefit by these other sources, with an adjustment for earned and unearned income. To assess distributional outcomes and, particularly, how well SSI functions as a tool for preventing poverty, one must consider not only the other sources of income but also the unit of observation, the time frame, and the economies-of-scale assumptions embedded in the poverty measure.

The assessment of outcomes depends to a critical extent on the assumptions about economies of scale that underlie the federal poverty thresholds (Citro and Michael 1995). This is particularly important when evaluating the level of SSI benefits for single recipients living alone, married couple recipients, and NCMs. Because the federal poverty measure's assumptions about economies of scale have been the subject of controversy recently, this analysis uses a different yardstick, in addition to the federal poverty threshold, to measure the effects of SSI on poverty—namely, the three-parameter experimental equivalence scale, which is widely considered to contain more realistic assumptions regarding economies of scale (Short 2001; Working Group on Revising the Poverty Measure 2000). The three-parameter scale makes it possible to assess the robustness of conclusions concerning SSI's effects on poverty in light of the economies-of-scale assumptions embedded in the federal yardstick. Finally, this article describes the overall implications of the SSI program's assumptions by age group and household size as well as the number of recipients in the household.

Prevalence of Multirecipiency

A substantially higher proportion of SSI recipients live in NCM households than was previously believed, according to estimates based on the matched data set and on the Office of Quality Assurance study (October 1996 to September 1997) (Social Security Administration 1999).⁹ Overall, the matched data estimate that approximately one in five SSI recipients lives in an NCM household (Chart 1).¹⁰ This proportion is almost twice as large as that identified by the quality assurance study. In fact, the matched data show larger proportions in each age group living in an NCM household, with the relative differences increasing somewhat with age.

Both data sources find that children are the most likely and elderly persons are the least likely to live in NCM households. Analysis of the matched data set shows that the same is true of SSI recipients who live in NCM families (Table 1). Because SIPP considers individuals who are in the same family to be in the same household, but not vice versa, it is not surprising that the percentage of SSI recipients in an NCM family is slightly smaller than the percentage in an NCM household.¹¹ The main difference between NCM families and NCM households is that some NCM households are made up of unrelated recipients (such as unmarried partners, friends, or tenants) living under the same roof. The age distribution of recipients in NCM households and families differs from the overall age distribution of SSI recipients (Table 2). Although children make up only 13 percent of the recipient population, around one-quarter of recipients in NCM households and families are children. The majority of SSI recipients in NCM households and families are working-age adults. Fewer than one recipient in five is elderly.

Identifying the composition of NCM households and families by recipients' age is important because different policy issues and rules apply to various age groups. For example, elderly persons who pass the income and asset screens qualify for assistance regardless of whether they are disabled, whereas children and working-age applicants must pass the disability screen. The treatment of earned income is more important for working-age applicants (whose benefits are affected by a spouse's earnings) and for children (whose benefits are affected by a parent's earnings), whereas unearned income—specifically, Social Security—is an important source of income for the elderly. Policy changes that focus on one age group affect other recipients in the same household, regardless of age.

The composition of NCM households and families is striking (Table 3). Forty percent of recipients in an NCM household live with at least one child recipient. Almost 85 percent are in households with working-age disabled recipients, and 32 percent live in households with elderly recipients. However, only 5 percent of SSI recipients live in NCM households composed just of elderly recipients. When NCM families are considered, the proportion of multirecipients increases in categories that include child recipients and decreases in other categories.

The matched data set shows that only a small fraction of SSI recipients—about 3.7 percent—cohabit with other SSI recipients. Cohabitation is more common among younger people (Bumpass and Sweet 1989), and 87 percent of cohabiting couples are disabled working-age recipients rather than elderly recipients. Although this measure of cohabitation reveals that fewer than 1 percent of all SSI recipients are cohabiting, the percentage is higher among working-age couple recipients: 16 percent of this group are cohabiting, and the remaining 84 percent are married.

These findings are roughly comparable with others concerning the prevalence of cohabitation, although the measure misses cohabiting couples if neither partner is listed as a household head in the SIPP.¹² Therefore, the estimates of cohabiting SSI beneficiaries given here can be considered a lower bound. Using this direct measure, Baughman, Dickert-Conlin, and Houser (2002) found that 3.9 percent of SIPP respondents were cohabiting in March 1996, compared with 5.1 percent of respondents when the researchers used the Adjusted Persons of the Opposite Sex Sharing Living Quarters measure.¹³ The latter measure is more inclusive, however, and risks identifying persons who are not actually in a cohabiting relationship.

Poverty and Multirecipiency in SSI

The next four sections

• analyze the relationship between the federal income guarantee and recipiency status;
• assess the role of SSI in the context of other sources of household income;
• analyze the key issues in poverty measurement affecting the perceived role of SSI in alleviating poverty; and
• provide an overall assessment of the relationship between poverty and multirecipiency status, considering the findings from those analyses.

For the remainder of this article, recipients' multi-recipiency status is determined by household rather than family membership. There are several reasons for this. First, the question of the differences between married and nonmarried recipients cannot be analyzed solely in the family context. Second, policy discussions as well as programmatic rules tend to use the household as the unit of observation. Third, any analyses of family issues can be treated as a more detailed look at household relationships, whereas the reverse is not possible. And finally, household- and family-based determinations of recipiency status are closely related to each other, as demonstrated in the preceding analysis and do not affect the key empirical findings presented below.

Nevertheless, the importance of the family needs to be stressed when moving from data analysis to policy implications. The family is a fairly straightforward consumption unit, whereas the household is not. Unrelated people in a household may or may not share expenses, and when they do, the extent of sharing varies substantially across households. Thus, considerations regarding economies of scale that are straightforward in the family setting may be problematic in the household context. In addition, program administration is inherently more difficult in the household context. In part, the difficulty stems from presuming that unrelated household members share any relationships or responsibilities other than those arising directly from shelter. Program administration challenges are further magnified by the more transient nature of household membership.

The Federal Income Guarantee and Recipiency Status

The first step in analyzing the effects of the federal income guarantee on poverty is to examine recipiency units. A recipiency unit, as defined here, is related to the "SSI unit" concept but is somewhat broader. A recipiency unit may consist of an individual recipient living without another recipient present, married couple recipients living together without any other recipient present, or recipients living with one or more other recipients who are not a spouse (that is, NCMs). The recipiency unit is different from the SSI unit in important respects: the SSI unit classifies all recipients as either individuals or couples, and it does not explicitly recognize NCMs. Recipiency units recognize NCM households, but they follow SSI program logic in that they include only SSI recipients and ignore anybody else who may be living with recipients. Artificial as they are, recipiency units reflect conventional programmatic thinking and are consistent with previously published studies that were based on similar definitions of NCM status.

The federal poverty threshold varies with the number of persons in the recipiency unit and assumes lower consumption needs for elderly one- and two-person units; the FBR, in contrast, does not vary with age. Because the federal poverty measure is based on annual income, whereas SSI payment eligibility is determined on a monthly basis, one-twelfth of the federal poverty threshold is used to create a monthly poverty threshold. The applicable FBR for the recipiency unit is then compared with the monthly poverty threshold.

When the federal income guarantee is expressed as a percentage of the monthly poverty threshold, the result confirms the widely held view that federal SSI assistance provides only subpoverty income for couples and for individuals living without another recipient present. For NCMs, in contrast, SSI apparently guarantees a monthly income above the poverty threshold. The monthly SSI federal income guarantee for NCMs relative to the federal poverty threshold markedly increases as the number of nonmarried recipients increases (Table 4). Because the poverty threshold differs by age and household size, SSI seems to provide somewhat greater protection against poverty for the elderly than for the nonelderly. The data also suggest that the monthly SSI federal income guarantee for individual recipients, such as widows, is relatively low not only in comparison with NCM recipients but also in comparison with that for married couples.

SSI and Other Sources of Household Income

Recipients of federal SSI payments may benefit from other sources of income, some of which reduce the federal payment—namely, countable earned and unearned income. Up to $20 in income from any source may be excluded. An additional $65 of earned income and 50 percent of any remaining earned income may also be excluded. Thus, an SSI recipient who lives alone and is eligible for Social Security benefits is allowed to realize only $20 over the FBR (a mere 4 percent increase over the 1998 individual FBR). Higher Social Security benefits would reduce the federal SSI payment dollar for dollar and might result in ineligibility for SSI payments altogether. On the other hand, a recipient could earn twice the amount of the FBR and still qualify for a federal payment, albeit a small one because of the effective 50 percent benefit reduction.¹⁴ Work among elderly and disabled recipients is relatively rare, so recipients' earnings seldom increase income substantially.

Perhaps more important are sources of income that do not result in the reduction of federal SSI payments. One of these is the state SSI supplement, which varies widely but is substantial in some states. In addition, unless deeming or in-kind support and maintenance rules apply, the income of other members of the recipient's family or household is not counted against the applicable FBR. For example, an adult recipient may live in a household with one or more other adults who have substantial earned or unearned income and still qualify for SSI. Such situations may arise when individual or couple SSI recipients live with nonrecipient adults or with unrelated SSI recipients.

The bulk of SSI recipients' household income is derived from other sources (Chart 2). When only non-SSI income is considered, recipients who live with no other recipients have the highest average household income, followed by those in NCM households, then by married couple recipients. That picture changes substantially when SSI income is added. SSI increases household income most for those in NCM households and least for individual recipients who live with no other recipients. As a result, recipients in NCM households have the highest income, followed by recipients in one-recipient and married couple households.

The averages presented in Chart 2 mask potentially important variations in the importance of non-SSI income in each of the three groups. The key variables are the presence or absence of state supplements, disregarded income of the recipient or recipients, the presence of household members not included in the recipiency unit, and the absence or presence (and amount) of earned and unearned income of nonrecipient household members.

Shifting the focus of analysis from the recipiency unit to the household changes perceived income. In some cases, it improves perceived income, as when a recipient lives (or recipients live) with individuals in the household who are outside the recipiency unit and who have substantial income. In other cases a recipient (or recipients) lives with persons who have little or no income of their own, resulting in a worsening perceived income. Total household income cannot, by definition, decrease with larger units of observation, but per capita income can, substantially undermining the income guarantee.

The share of average annual household income from various sources depends on the type of household (Table 5). In all three types—one recipient, married couple recipients, or noncouple multirecipients—average SSI income is less than 40 percent of total household income; in one-recipient households, it is less than 20 percent. Earnings are a substantial source of income in all three categories, but especially in one-recipient households.

SSI income reduces substantially the prevalence of poverty among recipients, regardless of household type (Table 6). Relying only on non-SSI sources, well over half of recipients in each group would have incomes below the federal poverty level, ranging from around 70 percent for married couple and NCM households to 58 percent for one-recipient households. Adding SSI income improves the situation considerably for all three types of households.

Recipients in one-recipient households appear worst off. After adding SSI income, almost half of them are still in poverty, compared with only about 20 percent of NCMs. About 40 percent of married couple recipients remain poor after receiving SSI payments. This pattern is qualitatively consistent with the naive view of subgroup differences shown in Table 4, but the appropriateness of the assumptions regarding economies of scale in the federal poverty measure has yet to be addressed.

The picture becomes clouded when considering recipients at the high end of the poverty distribution. Although recipients in one-recipient households are the most likely to be in poverty, they are also the most likely to have a total household income of twice the poverty threshold or higher—with or without SSI payments (Table 6). Without SSI payments, they are almost twice as likely as recipients in the other two types of households to have incomes above 200 percent of the poverty threshold. For recipients in married couple and NCM households, the marginal effect of SSI income is a more substantial factor in producing household incomes above 200 percent of the poverty threshold; nevertheless, they are consistently less likely than recipients in one-recipient households to have incomes at the higher end of the poverty distribution.

The Sensitivity of Poverty Outcomes to the Yardstick Used

Assessments of SSI's impact on poverty may be affected by the unit of observation, time frame, and economies-of-scale assumptions embedded in the poverty measure. This section addresses each one briefly.

Chart 3 shows how the monthly poverty status of adult SSI recipients in various households differs depending on the unit of observation used to measure income—the SSI unit (which recognizes individual and married couple recipients only, regardless of who else lives in the household), the family unit, or the household (the latter two units as defined by the SIPP).¹⁵ The comparisons reveal several important factors affecting poverty status:

• While federal payments guarantee only a subpoverty income for recipients in an SSI unit, not all adult recipients in SSI units live in poverty. About half of married couples escape poverty when other sources of income received by the couple are considered. However, the vast majority of adults in one-recipient and NCM households are poor. This reflects the fact that few recipients in either category receive sufficient state supplements or disregarded unearned or earned income to make up for the shortfall between the individual FBR and the federal poverty threshold for one person.
• The picture changes dramatically when the unit of observation is the family or household. The rate of poverty drops most precipitously for NCM households, mainly because SSI does not apply any assumptions regarding economies of scale to recipients in these households, whereas the poverty measures (federal or otherwise) do.
• The perceived poverty status of recipients in a family or household is also affected by the non-SSI income of other family or household members. Such income may increase or decrease the prevalence of poverty.
• For individual and married couple recipients, the prevalence of poverty is generally reduced by the "natural" safety net of grouping individuals into families and households. In some cases, however, the addition of nonrecipient members to the family or household may increase measured poverty. For example, if a nonrecipient with no income (such as a child or a disabled adult without earnings who does not meet the Social Security Administration's stringent disability screen) lives with the individual or couple recipient, that added person would not affect income but would increase the applicable poverty threshold.
• Similarly, the addition of family or household members without SSI income to NCM households may increase income or leave it unaffected. However, the implications for poverty status are bound to be asymmetrical. Because SSI does not apply economies of scale to NCMs, the federal income guarantee for NCMs is above the federal poverty threshold (as seen in Table 4). In some cases, the addition of family or household members without SSI and with little, if any, other sources of income may result in poverty. The data show that although the federal income guarantee applicable to NCMs exceeds the poverty threshold, over 17 percent of NCMs have household income below that threshold.
• For each group, the percentage of recipients living in poverty is smaller, on average, when based on the household unit than when based on the family unit. Moving from the family unit of observation to the household may increase the number of persons included in the unit, thereby increasing the applicable poverty threshold. However, the additional persons may also contribute income to the household.

These comparisons produce two main conclusions. First, when family or household income from any source is measured against the federal poverty threshold, NCMs tend to be better off, on average, than recipients in one-recipient or married couple categories. Second, in contemplating policy implications, the substantial heterogeneity of non-SSI sources of income behind these averages needs to be considered. In particular, while somewhere between 45 percent and 60 percent of one-recipient and couple recipient beneficiaries live in nonpoor families or households, a nontrivial minority of those in NCM households live in poor families or households.

The remaining analysis of poverty status is based on the household unit only. Although poverty is lower in households than in families, the relationships among the three groups (one recipient, married couple recipient, and NCM families or households) are preserved. In addition, because recipients are categorized on the basis of the number of SSI recipients in the household, household income is compared with the applicable household poverty threshold.

Whether annual or monthly measures of poverty are used, the overall patterns among the three groups are similar, with 47 percent to 49 percent of one-recipient households, 40 percent to 43 percent of married couple recipients, and 21 percent of NCM households living in poverty based on the annual and monthly measurements, respectively. Poverty status can vary with annual and monthly measurements for a number of reasons, including program dynamics of SSI and Old-Age, Survivors, and Disability Insurance; changes in household composition due to births or deaths; changes in marital status or household membership; and changes in non-SSI sources of income of household members. The changes may cause differences in either direction: people who are poor on the basis of the annual measure may not be poor on the basis of a given month's observation. It is also possible, however, for people to be classified as poor for several months yet not be poor on the annual measure because they had high income in the other months of the year. Nevertheless, no major systematic differences were observed.

Finally, the federal poverty threshold and the three-parameter experimental scale are used to compare the percentage of poor recipients in one- and two-person households (Chart 4).¹⁶ This comparison gets to the heart of SSI's impact on poverty because the two measures differ substantially in the assumptions they apply to individuals living alone and in larger units. Essentially, the three-parameter scale's assumptions regarding economies of scale for one- and two-person units are much closer to SSI's assumptions concerning individual versus couple beneficiaries. Thus it is possible to analyze the extent to which the economies-of-scale assumptions embedded in the federal poverty threshold affect the percentage of recipients living in poverty. Moreover, both the federal measure and the three-parameter scale are blind with respect to marital status. The relative status of recipients in the different types of households varies only slightly with the scale used to measure poverty (Chart 4). In particular, both scales show that two recipients living in a two-person NCM household are less likely to be poor than individual recipients living alone or with a nonrecipient and married recipients in a two-person household. Also consistent is the finding that overall, poverty is less prevalent among individual recipients in a two-person household than among married couple recipients.

Both measures also show that individual recipients who live alone are much poorer than any other group of recipients. However, the federal poverty measure shows a difference of 36 percentage points between individual recipients living alone and married couple recipients, whereas the three-parameter scale indicates a difference of only 24 percentage points. Although the economies of scale assumed in the federal poverty measure may overstate this important difference, the disparity remains regardless of which measure is used: the relatively high incidence of poverty among SSI recipients living alone also persists regardless of the scale used. The reason is largely attributable to factors other than SSI policy per se, and the implications for SSI policy are not immediately obvious.

Overall Poverty Outcomes

Using either the federal or the three-parameter poverty yardstick, analysis shows that, overall, non-SSI sources of income do not alleviate poverty for the vast majority of recipients (Table 7). Interestingly, when SSI benefits are excluded from household income, one-recipient households fare best, while NCM households fare only slightly better than married couple households. When SSI payments are included, one-recipient households fare worse and NCM households emerge as the least disadvantaged group. Importantly, the difference in prevalence of poverty between recipients in one-recipient households and married couple households disappears when the three-parameter scale is used instead of the federal poverty threshold.

Differences Among Age Groups. The prevalence of poverty differs substantially by age group (Table 8). Patterns among the elderly are basically consistent with the naive view presented in Table 4; that is, elderly NCM households are less likely to be poor than elderly married couple recipients, and individual recipients living without another recipient in the household are the most likely to be poor, regardless of the scale used to measure poverty. This finding suggests that the higher risk of poverty among individual recipients compared with couples does not result exclusively from the economies-of-scale assumptions embedded in SSI program design.

The situation among working-age recipients is reversed, with married couples having a higher poverty rate than one-recipient or NCM households. Many of the married couples in this group are made up of two severely disabled spouses, a situation that is likely to limit the availability of income from other sources. In contrast, some individual recipients may live in households with an adult who is not an SSI recipient, a situation likely to lift many of them out of poverty.

For children, there is no difference in the poverty rate of recipients living with no other recipient versus recipients in NCM households. This finding reflects the way disability interacts with other factors. In particular, disabled youths in one-recipient households may live with one or two working-age parents who are not SSI recipients, whereas youths in NCM households may live either with a disabled adult or an elderly caregiver, as Table 3 suggests, or with another disabled child. Either of these conditions reduces an NCM household's earnings potential. Reduced earnings, in turn, counterbalance the more generous SSI benefits that accrue to youths in NCM households.

Comparisons across age groups also support this interpretation. Child recipients living without another recipient in the household appear less likely to be poor than working-age or elderly recipients in such households, irrespective of the measure used. This is not surprising, considering the likelihood that an adult non-SSI recipient lives in the child's household (even though the parents' ability to work is probably constrained by the demands of caring for a disabled child). However, child recipients in NCM households, which may also include adult recipients, are more likely to be poor because the adult recipients in those households are severely disabled or elderly and less able to work. Child recipients in NCM households are also more likely than adult recipients in such households to be poor.

Differences by Household Size. Finally, poverty varies with household size and the number of recipients in the household (Table 9). The patterns largely reflect the factors analyzed above, with NCM households less likely to be poor than other household types (see Chart 4). The sharp contrast between individual recipients living alone or with a nonrecipient persists as household size increases, although the rate of poverty rises markedly in very large households (those with eight or more persons). Among NCM households, those with two nonmarried persons have the lowest poverty rates; the apparent advantage of NCM status is clearly diluted by larger household size. These findings vary little with the poverty scale used, and they highlight the importance of the interaction between household size and the number of recipients in the household. This key relationship in determining whether recipients live in poverty lies largely outside SSI program design.

Appendix: Data and Methods

The data set analyzed in this article is the 1996 Survey of Income and Program Participation (SIPP) panel matched to the Social Security Administration's Supplemental Security Record (SSR), which provides complete monthly benefit information on all persons receiving Supplemental Security Income (SSI). Whereas SIPP includes self-reports of SSI benefits, the SSR data reflect actual payments and thus avoid reporting error. The weight provided by SIPP for each month is used, but for the annual poverty analysis, which is based on information from respondents who were present for 12 interview months, the SIPP weight is adjusted upward. The unweighted counts and weighted population of SSI recipients used in the analysis are presented in Table A-1.

The SSR includes all SSI beneficiaries, but SIPP reflects only the noninstitutionalized population. Although institutionalized persons live in a group setting, different SSI rules apply to them, and they are not regarded as members of multirecipient households. Thus the estimated percentage of recipients who live in a multirecipient household, based on the SIPP noninstitutionalized population, is slightly higher than if it were based on all SSI recipients. SIPP respondents who live in noninstitutional group quarters, such as rooming and boarding houses, college dormitories, convents, and monasteries, are also excluded. SIPP classifies such persons as sharing a household, but that classification is not suitable for this analysis. Only 59 of the identified recipients lived in group quarters, less than 1 percent of the unweighted sample. Those individuals were included in a previous analysis, and the results were not qualitatively different.

Unfortunately, not all individuals interviewed in SIPP can be matched to administrative data, because not all of them report a valid Social Security number (SSN). Individuals who do not have an SSN match are identified by the Summary Earnings Record (SER), which provides the universe of SSNs for matching with SIPP data. In a small number of cases, the same SSN was assigned to more than one SIPP respondent; these individuals are treated as nonmatches. In March 1996, 82.5 percent (unweighted) of individuals had an SSN match; in January 1998, 83.1 percent did; and in December 1998, 82.9 percent did. The match rate varies with age group, however. For example, in March 1996, 84.8 percent of individuals aged 65 or older had an SSN match, 83.7 percent of individuals aged 18 to 64 did, and 79.1 percent of individuals aged 17 or younger had a match.

The flag for nonmatches is summed across households to identify who in the household reported a valid SSN. SIPP defines a household as a group of persons who occupy a housing unit. Two variables are used to identify persons in the same household: the sample unit ID (SSUID) and the current address ID (SHHADID). The sample unit ID and the family ID number (RFID) are used to identify persons in the same family. Persons are considered by SIPP to be in the same family if they are related by blood, marriage, or adoption. Cohabiting couples, even if they share a child, are not classified as being in the same family. The current address ID and the family ID are monthly variables. One advantage of SIPP is its ability to pick up monthly household and family changes that might not be caught by a yearly measure.

If everyone in a household has an SSN match, the SSR is used to determine receipt status and benefit amount for each person. If one or more persons in the household do not have an SSN match, the self-reported SSI benefits in SIPP are used. By this method, SSR information is used for 71.5 percent of individuals from March 1996, for 70.1 percent from January 1998, and for 69.1 percent from December 1998.

When SIPP information is used for an individual, the benefit amount is the self-reported sum of federal SSI and federally administered state supplementation (see Huynh, Rupp, and Sears 2002). The variable used from the SSR corresponds with what a respondent would report in SIPP, because it reflects the actual payment to the recipient rather than eligibility status. When SSR payment information is used, the benefit amount is the sum of the federal SSI payment and any federally administered state SSI payment. Therefore, individuals who receive only a federally administered state SSI payment are classified as recipients.

An individual's record in SIPP includes the person number used to identify his or her spouse. Information on a spouse's benefits is matched (where applicable) to the individual's observation in order to flag married couples in which both members receive benefits in a given month. The count of cohabiting couples is a rough proxy because an individual's SIPP record does not include the person number of a cohabiting partner, if applicable. However, SIPP does provide a relationship variable that enables respondents to indicate their relationship to the household reference person. Respondents who were coded as the household reference person or as the unmarried partner of the reference person were flagged. This method misses cohabiting couples in which neither partner is the household reference person and can therefore be regarded as a lower-bound estimate of cohabiting couples.