The TANF/SSI Connection

Summary

Interactions and overlap of social assistance programs across clients interest policymakers because such interactions affect both the clients' well-being and the programs' efficiency. This article investigates the connections between Supplemental Security Income (SSI) and Temporary Assistance for Needy Families (TANF) and TANF's predecessor, the Aid to Families with Dependent Children (AFDC) program. Connections between receipt of TANF and SSI are widely discussed in both disability policy and poverty research literatures because many families receiving TANF report disabilities.

For both states and the individuals involved, it is generally financially advantageous for adults and children with disabilities to transfer from TANF to SSI. States gain because the federal government pays for the SSI benefit, and states can then use the TANF savings for other purposes. The families gain because the SSI benefits they acquire are greater than the TANF benefits they lose. The payoff to states from transferring welfare recipients to SSI was substantially increased when Congress replaced AFDC with TANF in 1996. States retained less than half of any savings achieved through such transfers under AFDC, but they retain all of the savings under TANF. Also, the work participation requirements under TANF have obligated states to address the work support needs of adults with disabilities who remain in TANF, and states can avoid these costs if adults have disabilities that satisfy SSI eligibility requirements. The incentive for TANF recipients to apply for SSI has increased over time as inflation has caused real TANF benefits to fall relative to payments received by SSI recipients.

Trends in the financial incentives for transfer to SSI have not been studied in detail, and reliable general data on the extent of the interaction between TANF and SSI are scarce. In addition, some estimates of the prevalence of TANF receipt among SSI awardees are flawed because they fail to include adults receiving benefits in TANF-related Separate State Programs (SSPs). SSPs are assistance programs that are administered by TANF agencies but are paid for wholly from state funds. When the programs are conducted in a manner consistent with federal regulations, the money states spend on SSPs counts toward federal maintenance-of-effort (MOE) requirements, under which states must sustain a certain level of contribution to the costs of TANF and approved related activities. SSPs are used for a variety of purposes, including support of families who are in the process of applying for SSI. Until very recently, families receiving cash benefits through SSPs were not subject to TANF's work participation requirements.

This article contributes to analysis of the interaction between TANF and SSI by evaluating the financial consequences of TANF-to-SSI transfer and developing new estimates of both the prevalence of receipt of SSI benefits among families receiving cash assistance from TANF and the proportion of new SSI awards that go to adults and children residing in families receiving TANF or TANF-related benefits in SSPs.

Using data from the Urban Institute's Welfare Rules Database, we find that by 2003 an SSI award for a child in a three-person family dependent on TANF increased family income by 103.5 percent on average across states; an award to the adult in such a family increased income by 115.4 percent. The gain from both child and adult transfers increased by about 6 percent between 1996 (the eve of the welfare reform that produced TANF) and 2003.

Using data from the Department of Health and Human Services' TANF/SSP Recipient Family Characteristics Survey, we estimate that 16 percent of families receiving TANF/SSP support in federal fiscal year 2003 included an adult or child SSI recipient. This proportion has increased slightly since fiscal year 2000.

The Social Security Administration's current procedures for tabulating characteristics of new SSI awardees do not recognize SSP receipt as TANF. We use differences in reported TANF-to-SSI flows between states with and without Separate State Programs to estimate the understatement of the prevalence of TANF-related SSI awards in states with SSPs. The results indicate that the absolute number of awards to AFDC (and subsequently) TANF/SSP recipients has declined by 42 percent for children and 25 percent for adults since the early 1990s. This result is a product of the decline in welfare caseloads. However, the monthly incidence of such awards has gone up—from less than 1 per 1,000 child recipients in calendar years 1991–1993 to 1.3 per 1,000 in 2001–2003 and, for adult recipients, from 1.6 per 1,000 in 1991–1993 to 4 per 1,000 in 2001–2003.

From these results we conclude that a significant proportion of each year's SSI awards to disabled nonelderly people go to TANF/SSP recipients, and many families that receive TANF/SSP support include adults, children, or both who receive SSI. Given the Social Security Administration's efforts to improve eligibility assessment for applicants, to ensure timely access to SSI benefits for those who qualify, and to improve prospects for eventual employment of the disabled, there is definitely a basis for working with TANF authorities both nationally and locally on service coordination and on smoothing the process of SSI eligibility assessment.

The Deficit Reduction Act of 2005 reauthorized TANF through fiscal year 2010, but with some rules changes that are important in light of the analysis presented in this article. The new law substantially increases effective federal requirements for work participation by adult TANF recipients and mandates that adults in Separate State Programs be included in participation requirements beginning in fiscal year 2007. Thus SSPs will no longer provide a means for exempting from work requirements families that are in the process of applying for SSI, and the increased emphasis on work participation could result in more SSI applications from adult TANF recipients.

Introduction

Like the Aid to Families with Dependent Children (AFDC) program it replaced in 1996, the Temporary Assistance for Needy Families (TANF) program supports many poor families in which one or more members has a disability (Nadel, Wamhoff, and Wiseman 2003/2004). In many cases these disabilities are sufficiently severe to make the adult or child a candidate for Supplemental Security Income (SSI). Eligibility for TANF benefits is determined by income, not disability, and can be established relatively quickly, so it is common for poor families with children and disability problems to apply for TANF first and then seek to qualify for SSI. Many analysts claim that the intensity with which families pursue SSI is a function of both the potential financial gain to the family and the fiscal consequences of such moves for state governments (Kubik 2003; Schmidt and Sevak 2004; Ziliak 2004).

Differences between TANF and SSI benefits are substantial, and the transition from AFDC to TANF may have increased incentives for recipients who were potentially eligible for SSI to apply for benefits and for states to encourage such efforts. Between December 1996 and December 2003, the nonelderly SSI caseload increased by 8.6 percent (SSA 2004b, 21). Although it is reasonable to ask how much of this growth can be attributed to welfare reform, we do not try to answer this question in full here. Rather, our intention is to throw new light on just how substantial the connection is between TANF and SSI. We argue that the growing enrollment of TANF families in what are termed Separate State Programs has led to an undercount of the number of SSI awards that go to TANF recipients and, in consequence, an underappreciation of the magnitude of the interaction between the two programs. We present estimates that correct for this problem.

We begin with an overview of the two programs, the basis for overlap, and the incentives for TANF recipients to seek SSI awards. We then look at the interaction from two perspectives, considering first the prevalence of SSI receipt among families currently receiving TANF and then the extent to which new SSI awardees come from TANF cases. TANF was reauthorized through 2010 by the Deficit Reduction Act of 2005, which was signed into law by President Bush in February 2006. We suggest that certain provisions of the new legislation are likely to increase states' interest in promoting the transfer of TANF recipients to SSI. We conclude with a review of the implications of our results for possible collaboration between the Social Security Administration (responsible for SSI) and the Department of Health and Human Services (responsible for TANF).

Background

Supplemental Security Income is the nation's safety net for adults and children with major disabilities as well as for poor elderly persons. SSI provides a monthly cash payment to persons or their caretakers based on uniform national eligibility standards for disability and need. All persons who meet those standards are enrolled. The program is funded from general federal revenues and is administered by the Social Security Administration. Some states supplement the SSI benefit, but supplements account for less than 14 percent of benefits nationwide (SSA 2004a, 3).

Temporary Assistance for Needy Families is a federal/state program that provides assistance to needy families with children. Details, including standards of need and benefit payments, are generally left up to the states, subject to certain federal requirements and restrictions. TANF law mandates some conditions for eligibility, sets requirements for participation in work and work-related activities for recipients, places a time limit on use of federal funds for benefits to some families, specifies certain "maintenance-of-effort" levels of expenditures from states' own funds, and requires that states report certain data on recipients and expenditures. Many requirements are either difficult for the federal government to enforce or can be avoided by states using exemptions and exceptions allowed by the law.

The Changing Incentives for TANF-to-SSI Transfers

The Supplemental Security Income program was established in 1972, and first benefits were paid in 1974. From the beginning the program created incentives for states to promote transfer of persons receiving AFDC to SSI because states paid as much as half of AFDC benefits but the federal government pays all of the basic SSI benefit. (About half of the states supplement the federal payment, but most such supplements are small.) When a member of a family receiving AFDC moved to SSI, total family income generally increased because the consequent reduction in their AFDC payments was more than offset by the gain from SSI.

The TANF program's progenitor, the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA), enhanced states' incentives for promoting TANF-to-SSI transfers by changing the terms of the federal/state assistance partnership. AFDC was financed with a federal matching grant, and each dollar of benefits paid from state funds was matched with at least a dollar of federal funds, up to as much as $4 for states with low per capita income. PRWORA substituted a block grant for the matching grant. Under AFDC, transfer of an AFDC recipient to SSI states saved only their share of benefit expenditures. In contrast, under the TANF block grant, all of the savings from such transfers accrues to the states. States have responded to the altered terms of federal participation and to recession-related fiscal problems by continuing a long-term decline in real (inflation-adjusted) benefits that predates PRWORA. Declining real benefits combined with increasing obligations for recipients to participate in work and work-related activities have made the transfer from TANF to SSI more attractive for recipients with disabilities.

The incentives created by the change in welfare financing are not trivial. The shift to the block grant increased the states' fiscal gain from moving a recipient from TANF to SSI by a factor of at least two (for states with high per capita income) to as much as four (for low-income states). In addition to the incentive effect of the block grant, the 1996 reforms increased requirements for rates of participation by recipients in work or work-related activities. States anticipated that achieving the new targets would be both costly and administratively difficult, because recipients who had not previously engaged in work-related activities typically faced more barriers to work—including disabilities—than did those who were already engaged. This, too, encouraged removing persons with disabilities from the TANF rolls to SSI. The TANF High Performance Bonus added to the TANF-to-SSI incentive. The bonus (now defunct) was awarded to states on the basis of job placement, employment retention, and earnings gains for adult TANF recipients, without adjusting for disability (Wiseman 2004). If adult TANF recipients with disabilities are less likely to become employed, to keep jobs once they have them, or to experience substantial earnings gains once employed, moving them to SSI improved a state's chance for a performance bonus. The shift to the block grant also raised the marginal cost to states of increasing TANF benefits and the fiscal gain from reducing them. On average (weighted by the TANF caseload), state TANF benefits fell by 7 percent in real terms between 1996 and 2003. Only 19 states increased even nominal TANF benefits over this interval; 5 actually reduced them.¹

Federal SSI benefits are indexed. Therefore, as TANF benefits have declined in value, the gain to TANF families from transferring an adult or child to SSI has grown. Table 1 illustrates the change for a one-adult, two-child family in which either the adult—generally the mother—or a child moves to SSI. In the illustration, the family has no income other than TANF and, potentially, SSI; we have not included food stamps, which if treated as cash would narrow some of these differentials.² The table shows weighted national averages of AFDC (1996) and TANF (2003) benefits, both including and excluding California. California is important because of the size of the state's caseload (accounting for 22 percent of the national TANF caseload in 2003) and because after 1996 the state adopted a dual payment structure in which cases exempt from work requirements—including cases with adults with disabilities or adults caring for children with disabilities—received a substantial benefit increase over nonexempt cases. (At the same time, California introduced regional cost-of-living differentials; data in the table are for Los Angeles.) If the California family illustrated in Table 1 were not exempt, the benefit would have been $704, not $786, in 2003. Therefore we report national averages with and without California, as well as comparable benefits for New York (represented by New York City) and Texas.

Comparison of the 1996 and 2003 benefit amounts reported in Table 1 for families receiving only AFDC/TANF benefits indicates that average real TANF benefits declined for the nation as a whole over this period (from $494 to $478, in 2003 dollars) but increased in California (from $699 to $786). Table 1 also shows the consequence for the family's total income—AFDC/TANF assistance plus SSI—of moving either one child or the adult to SSI. The outcomes for the adult and child cases are not identical because many states' AFDC/TANF benefit structures provide more money for adults than for children. The illustration makes four important points.

• In both 1996 and 2003, moving one family member to SSI substantially increased total family income. In New York City in 2003, for example, transferring one adult to SSI increased family income by almost 92 percent. Elsewhere, the gain was greater. On average nationwide, moving a child or the adult from TANF to SSI would more than double the cash income of a family wholly dependent on TANF.
• Moving the adult generally added more to family income than did moving a child.
• The change was worth more in 2003 than in 1996 because the erosion of TANF benefits has enhanced the relative attractiveness of SSI.
• The percentage gain in family income was particularly large in a low-benefit state like Texas.

Beyond the downward trend in benefits, the TANF programs in many states continue to require work or involvement in work-related activities, and all but seven states have lifetime time limits on financial support. Although many persons with disabilities are exempted from activity requirements, time limits, or both, the award of such exemptions is in many states highly discretionary and therefore a source of uncertainty. From the perspective of persons with disabilities, SSI offers more money and far greater security.

The Outlook

There is no near-term prospect for a change in the trend of these incentives. The Administration's budget proposal for federal fiscal year 2007 provided level funding for the TANF block grant. As caseloads have declined, states have reallocated TANF grant funds to uses other than cash benefits. In fiscal year 2003, basic assistance (that is, grants to families receiving TANF) accounted for just 35.2 percent of state and federal expenditures from TANF funds (federal and state contributions mandated by maintenance-of-effort requirements), down from more than 90 percent before welfare reform in 1996.³ The benefits from these nonassistance expenditures are widely distributed and enjoy a growing constituency. At the same time, real funding for other programs with objectives and clientele that overlap with those of TANF and for which TANF funds can be used has also declined, and states are showing growing interest in using TANF funds for expanding prekindergarten education and other child welfare programs. These competing uses increase pressure to reduce TANF cash benefits, or at least not to change them in nominal terms. As a result, the gap between TANF and SSI benefits is likely to grow.

The consequences of welfare reform and TANF operations for persons with disabilities and the demand for SSI is a long-standing matter of concern to the Social Security Administration as well as to policy analysts outside the agency (see, for example, Garrett and Gleid 2000; Stapleton and others 2001/2002; Kubik 2003; Lee and others 2004; Schmidt and Sevak 2004; Nadel, Wamhoff, and Wiseman 2003/2004; Ziliak 2004). However, it is difficult to assess the extent to which TANF and SSI interact in practice because of data problems (to be discussed below) on both sides. In a recent report (GAO 2004), the Government Accountability Office poses, but does not for the most part answer, questions about the number of adults and children who move from the TANF program to SSI and possibilities for intercepting adults in the application process who are capable of employment. GAO recommends that the Social Security Administration work with TANF offices to develop screening tools, assessments, or other data that would identify TANF recipients with impairments who, though potentially eligible for SSI, may also be capable of working. In its response, the Social Security Administration concurred with the recommendation (GAO 2004, 30). However, a key question remains: How significant is the TANF/SSI connection? In the next two sections, we show the connection to be large.

SSI Receipt within TANF Families

Available data on certain features of the AFDC/TANF caseload lead us to concentrate much of our analysis on trends over the fiscal year 2000–2003 period (note that all caseload data are for fiscal years). Given the dramatic developments in the caseload, it is important to place this focal period in a longer-term context. Chart 1 shows the national AFDC/TANF caseload from fiscal year 1990 through fiscal year 2003 and makes obvious a point that is well known—the number of families receiving AFDC/TANF benefits fell precipitously after 1994. This decline slowed with the onset of recession in 2000/2001. Therefore, when we look at TANF after 1999, we are considering TANF "after the fall," with a much diminished, residual caseload of about 2 million families. Even if the rate of transfer of adults and children from TANF families to SSI increased between 1997 (the first year of TANF's implementation) and 2003, the number of families involved could well diminish given that the base—the number of families receiving assistance in a typical month—had shrunk by almost half by 2003.

Separate State Programs

The TANF caseload plotted in Chart 1 includes cases in Separate State Programs (SSPs). This inclusion turns out to be relevant to assessing the TANF/SSI overlap. SSPs are assistance programs that are administered by TANF agencies but are paid for wholly from state funds. When such programs are conducted in a manner consistent with federal regulations, money that states spend on SSPs counts toward the federal maintenance-of-effort requirements under which states must sustain a certain level of contribution to costs of TANF and approved related activities.

States use SSPs for a variety of purposes, including the provision of supplemental transportation, housing, and child care services for families receiving TANF cash benefits. However, SSPs are also attractive to states as alternative vehicles for income support because families receiving cash benefits through SSPs are not subject to TANF's work participation requirements. Complying with those requirements is costly, and the rules are difficult to apply to some families. Requirements for adults in two-parent families are particularly stringent; as a result, it is common for two-parent families to be assisted in SSPs. Significantly, this is also true for families with adults who are incapacitated. Some states specifically assign families with adults who are considered candidates for SSI to SSPs during the often lengthy SSI application process. Maryland, for example, moves TANF families with an SSI candidate to an SSP called the Disability Entitlement and Advocacy Program, which pays a contractor specifically to pursue the SSI award on the disabled recipient's behalf (Kaplan 2000, 4). If the application fails, the family is returned to TANF. Although the proportion of total TANF-related cash assistance cases attributable to SSPs remains quite small (just 6.8 percent in fiscal year 2003), the fact that virtually all of these cases include adults means that they account for a larger share of adult recipients—almost 15 percent in fiscal year 2003, up from 9.1 percent in fiscal year 2000. In contrast to what is typical of SSP families, the regular TANF caseload includes a significant and growing proportion (over one-third, in 2002) of child-only cases, in which for various reasons adults are not counted as beneficiaries and are therefore not subject to TANF participation requirements (Gibbs and others 2004).

The distinction between recipients of cash benefits in SSPs and those formally "in TANF" is not a matter of substance, because states can vary work requirements among TANF families and thus achieve the same outcome as is accomplished with SSPs. The distinction is instead principally a matter of strategy with respect to federal activity requirements. Recipients are unaware of the difference, and official caseload statistics (such as those underlying Chart 1) combine the counts.⁴ Benefit amounts received by SSP participants do not differ from those received by similarly constituted families formally in TANF. States must transmit the same information about SSP recipients to the Office of Family Assistance (the agency responsible for TANF within the Department of Health and Human Services' Administration for Children and Families) that they do for families counted as being in TANF. As noted above, SSP participants differ from TANF participants principally in that states do not subject SSP participants to the same work requirements imposed on TANF participants. Our review of states' SSP rules suggest that these differences in work requirements typically are linked to special situations such as families with two parents at home, adults needed as caretakers, and so on. Since SSP recipient families are otherwise like TANF recipient families and generally see themselves as "in TANF," it is important to combine them in assessing the prevalence of SSI receipt among families receiving welfare today. Thus, for the remainder of this analysis, we refer to such families as TANF/SSP recipients.

The Prevalence of Supplemental Security Income

To assess the prevalence of SSI receipt among TANF/SSP recipient households, we use data from the annual TANF/SSP case characteristics surveys for fiscal years 2000–2003 from the Office of Family Assistance.⁵ By fiscal year 2003, approximately one in eight (12.6 percent) TANF/SSP cases nationwide were in a household that included an adult SSI recipient; 1 in 20 (4.8 percent) included a child SSI recipient; 1 in 6 (16.1 percent) had an adult, a child, or both; and 1 in 80 (1.2 percent) had both (Table 2). Throughout the period, in each month more than 256,000 cases were linked to an adult SSI recipient, slightly less than 100,000 families included a child recipient, and well over 327,000 included an adult or a child SSI recipient.

The adults and children counted in Table 2 are the beneficiaries of SSI awards in times past. We now look at the incidence of new awards to persons receiving TANF/SSP assistance.

SSI Awards to TANF Recipients

Separate State Programs complicate efforts to gauge the importance of TANF in providing intermediate support for persons applying for SSI. The reason is that current procedures for recording SSI data do not treat receipt of assistance in an SSP as equivalent to being in TANF. The sequence of questions asked during an SSI eligibility interview focuses on information on "income based on need (IBON) that is wholly or partially federally funded." Such income is subsequently identified by code "F" in the applicant's Supplemental Security Record (SSR). Type F income includes many things (payments for foster care, various types of refugee cash assistance, and so on), but a limited review of data in the Modernized SSI Claims System indicates that TANF accounts for 95 percent or more of that income, and therefore the presence of such income may be taken as a TANF proxy. However, this marker is not available for adults and children in cases counted as part of an SSP. SSI program eligibility rules do not count assistance based on need (ABON) as income if it is wholly state funded. Separate State Programs fall into this category, so persons who are members of families receiving such benefits cannot be identified in the Social Security Administration's data as TANF related. Not all states have SSPs, but it is possible to use data on rates of TANF-related SSI applications in states that do not have this innovation to adjust data for the undercount in states that do. Before introducing this adjustment, we review the prevalence of reported Type F income for SSI awardees aged 64 or younger.

Incidence of SSI Awards Without Adjustment for Separate State Programs

Given that the data on case characteristics from the Office of Family Assistance are for the 4-year period fiscal years 2000–2003 and that reliable data on SSP recipiency are available only for this period, for comparability we concentrate here on awards made within this interval. Our incidence estimates are based on the 10 percent SSI applicants and awards sample commonly used within the Social Security Administration's Office of Policy for program analysis. The general tabulation of our extract, converted to an estimate of all TANF-related cases for each of the 4 years and compared with aggregate SSI applications for persons who, at the time of application, were aged 64 or younger, is shown in Table 3. (Note that tabulations of data from the Supplemental Security Record are for calendar, not fiscal, years to allow matching our data to published SSA reports from the Social Security Administration on total SSI awards.)

An average of 677,000 awards were made per year for all ages under 65. About three-fourths of these awards (514,125) were to working-age adults (ages 18–64). Awards to children and adults increased by 24 percent and 11 percent, respectively, over the 2000–2003 period. TANF does appear to provide important intermediate support for children and adults who later receive SSI: over the entire period, an average of 28.7 percent of all awards to children were to children living in families associated with TANF (by the presence of Type F income); for adults, the corresponding figure was 9.5 percent. Although the absolute numbers of TANF-associated awards to children went up over the period, the percentage of awards to both children and adults associated with TANF fell.

Adult TANF recipients are predominantly women (mothers). Less than 10 percent of adult TANF recipients in fiscal year 2003 were men (DHHS 2004, Tables 19 and 20). The connection between TANF and SSI is more clearly revealed if we consider awards to women by age. Over the 4-year period 2000–2003, more than a quarter of awards to women aged 22–39 were linked to TANF (Table 4). Here, too, we see a downward trend: in each age category as well as overall, the proportion of awards going to women in families apparently receiving TANF generally declined.

Incidence of SSI Awards with Adjustment for Separate State Programs

Given the definition of Type F income, the recent growth in TANF Separate State Programs means that the estimates of TANF-related SSI awards in Tables 3 and 4 are biased downward, and the understatement is growing over time. To correct for this distortion, we exploit the facts that our SSR extract provides awards by state and that not all states have SSPs. Our approach, described in detail in the appendix, is to regress annual SSI awards for states on state TANF plus SSP recipiency and to study how the incidence of awards varies with the use of SSPs. We use the results to adjust our estimates of the proportions of SSI applicants who come from TANF families in states with SSPs. Our estimates indicate that, by 2003, states' use of SSPs led to an aggregate understatement of TANF-related SSI awards of 26 percent for adults and 24 percent for children.

The SSP adjustment makes a significant difference (see Table 5, which combines these estimates with the preadjustment data in Table 3). We estimate that over the period 2001–2003, failure to account for SSPs produced an undercount of the proportion of SSI awardees coming from TANF households of about 6.3 percentage points for children and 2.5 percentage points for adults. The figures for adults are aggregate, covering men and women in all adult age groups. Given that SSPs are commonly used specifically for cases in which adults are believed to be good candidates for SSI awards, the undercount produced by the SSP option probably accrues principally among younger women. Looking back at the unadjusted data on awards to women in Table 4, the aggregate adjustment suggests that if the revised data were detailed enough to support division by age and sex, the numbers for women associated with TANF would increase substantially for women younger than age 40, and as many as a third of SSI awards to women aged 22–39 would be to TANF recipients.

The Bottom Line: Incidence of SSI Awards Before and After the Advent of TANF

The last step in our analysis is to compare flows from AFDC to SSI before TANF with the results we have developed for calendar years 2001–2003. The SSR extract allows us to replicate our procedure for identifying SSI awards to recipients of public assistance for pre-TANF years and to compare the incidence of awards with adjusted rates for 2001–2003. We choose 1991–1993 and 1994–1996 for this purpose; these intervals cover the rapid caseload increase (1991–1993) and decrease (1994–1996) before implementation of the 1996 reforms under PRWORA (see Chart 1 for corresponding fiscal year data).

Our analysis indicates that the incidence of SSI awards among TANF recipients has been much greater in recent years than it was during the early 1990s (Table 6). Award rates for children rose from an average of 0.92 per month per 1,000 child TANF recipients in 1991–1993 to 1.28 per 1,000 in 2001–2003. The change was even more dramatic for adults. On average in each month of 1991–1993, 1.55 TANF-linked SSI awards were made per 1,000 recipients. By 2001–2003, the average rate was slightly over 4 per 1,000 per month. Although the number of SSI awards associated with welfare went down between 1991–1993 and 2001–2003 by 42 percent for children and 25 percent for adults, the caseload fell substantially more, so incidence rose.

These results do not necessarily mean that states are now more aggressively pursuing SSI eligibility for their TANF clients; it could be that families with disabilities are still likely to get into TANF and so the decline in recipiency among this group has been smaller than the decline in the use of welfare among families without such barriers.

Conclusion

A significant proportion of each year's SSI awards to disabled people aged 64 or younger go to TANF/SSP recipients, and many families that receive support from TANF/SSP include adults, children, or both who receive SSI. Given the Social Security Administration's efforts to improve eligibility assessment for applicants, to ensure timely access to SSI benefits for individuals who qualify, and to improve prospects for eventual employment of the disabled, there is definitely a basis for working with TANF authorities both nationally and locally on coordinating services and on smoothing the process of assessing SSI eligibility.

In February 2006, the President signed the Deficit Reduction Act of 2005. Among other things, this legislation reauthorized TANF through fiscal year 2010, but with some rules changes that are important in light of the analysis presented in this article. The new law substantially increases effective federal requirements for work participation by adult TANF recipients and mandates that adults in Separate State Programs be included in participation requirements beginning in fiscal year 2007. Thus, Separate State Programs will no longer provide a means for exempting from work requirements families in the process of applying for SSI, and the increased emphasis on work participation may result in more SSI applications from adult TANF recipients.

Appendix

As indicated in this article, Separate State Programs are relevant to assessing the overlap between the Temporary Assistance for Needy Families and Supplemental Security Income programs, but the Social Security Administration's current procedures for recording data do not identify awardees receiving benefits in SSPs as being linked to TANF. In the first years of TANF's operation, considerable confusion existed over what states could do with SSPs, when such outlays amounted to income support, when SSP expenditures counted toward maintenance of effort, and what reporting was required on the use of funds in this way. These requirements are now well established, and states have increased their use of SSPs. In fiscal year 2000, 24 states had one or more SSPs, and in 17 of them at least one of the SSPs was specifically designed for two-parent families. By fiscal year 2003 this number had increased to 31, with 24 states providing benefits to two-parent families in this way.

The recent growth in TANF Separate State Programs means that the estimates of TANF-related SSI awards based only on the presence of federally funded income based on need (Type F income) in SSI recipients' Supplemental Security Records are biased downward, and the understatement is growing over time. We correct for this understatement by exploiting the facts that our SSR extract provides awards by state and that not all states have SSPs. We regress annual SSI awards for states on state TANF plus SSP recipiency and study the way in which the incidence of awards varies with use of SSPs. Given states' overall TANF + SSP recipiency, we expect that TANF-related SSI awards should decline as the number of people in SSPs grows. Our statistical model incorporates a lag between award rates and the TANF/SSP caseload. Before presenting the model, we justify the lag.

The Lag Between Application and Award

TANF is important to SSI policy in part because the program supports families awaiting action on their SSI application. Awards take time, and a sense of this lag is provided by our SSR extract. We have calculated the time between the date of most recent application and first SSI payment for all awards over the calendar year 2000–2003 period. The results for adults and children appear in Chart A-1.

In Chart A-1, each line identifies the cumulative proportion of the year's award recipients who are accounted for by an elapsed time between application and award date that is equal to or less than the specified number of months. Thus in the adult chart, reading upward from the 12-month tick mark, we find that approximately 55 percent of awards in 2003 were to persons who had applied for SSI no more than 12 months before. The fact that none of the lines reach 100 percent at the 36-month point indicates that some SSI awards were obtained only after more than 3 years of review and, presumably, adjudication.

Note that children's awards are typically made much more quickly. By 2003, almost 90 percent of awards to TANF-related children were made within 12 months of application. In contrast to the results for adults, elapsed time between application and award for children has been falling over time. Nevertheless, TANF-related awards for children and adults in any fiscal year are clearly a function of both current and past TANF and SSP caseloads, and our model must accommodate this.

The Model

When TANF recipients are placed in Separate State Programs, the ability to identify the case as related to TANF is lost because the Social Security Administration's procedures do not count such income as Type F. To gauge the significance of this effect, we estimate the coefficients of ta regression of TANF-related SSI awards, by state, on states' TANF and SSP caseloads. The regression (equation 1) appears in the box below; we estimate coefficients separately for adults and children.

We expect that "steady state" TANF-related SSI awards will be positively related to caseloads, so the sum (β₁ + β₃) should be greater than 0. We have argued, however, that the Social Security Administration's procedures for recording data fail to identify persons in SSPs. Therefore, when comparing states with and without SSPs, states with SSPs should have a lower incidence of awards linked to TANF, which implies that (β₂ + β₄) is less than 0. Furthermore, because requirements apply only to adults and because adult TANF benefits often exceed benefits for children, the states have far greater incentives to place families with adults with disabilities in SSPs than they do for families with children who are SSI candidates. Thus, although we expect states with SSPs to have depressed rates of identified TANF-to-SSI transfer for both adults and children, the effect should be stronger for adults. Since the 4 years we consider clearly involve states' learning about the SSP option, the coefficients are likely to change. Accordingly, we estimate separate equations for each year rather than pooling our data.

The model could be limited to variables 1–4, that is, current and immediate past caseloads, because the data in Chart A-1 imply that current awards are generally for people who applied within the past 2 years. However, we have not modeled the state's decision to adopt the SSP option. If states creating SSPs are states that are exceptionally prone to promoting application for SSI, then the error term in the equation is correlated with two of the right-hand variables (those related to numbers of cases in SSPs) and the coefficient estimates will be biased. To capture this state effect, we have added a fifth variable that reflects typical flows from AFDC to SSI in the state in the period running up to the passage of the Personal Responsibility and Work Opportunity Reconciliation Act. This variable—the "predicted adult/child SSI awards in state s in year t" variable—is defined as the product of the TANF + SSP caseload in state s in period t multiplied by the "predicted award rate" p_s (the average over the 1994–1996 period of the ratio of AFDC-related SSI awards to the AFDC caseload in state s). Including this variable causes all other effects to be defined relative to what would have occurred had award rates among TANF recipients been the same as award rates before TANF. "State effects" should be captured here.

Estimation Results

The estimation results are reported in Table A-1 (for adults) and Table A-2 (for children). Because the regressions involve counts of recipients in the current and preceding year, we can estimate the coefficients of the model only for the last 3 of the 4 years for which we have data.⁶ We report estimated coefficients and the weighted mean value of p_s for each year (mean p changes slightly because of changes in the distribution of cases across states).

These results support several inferences. First, as was reasonably anticipated, an increase in total cases raises predicted awards. This result is best simulated by assuming that current and previous years' caseloads go up; such a change affects predicted awards by the two coefficients β₁ + β₂ plus the consequence judged from past experience. This "predicted" component is β₅ times the prediction rate for the state (p_s). Tables A-1 and A-2 show this sum as the weighted mean value of the prediction rate. The implication is that nationwide an enduring (for at least a year) increase of 100 TANF cases added 2.95 adult SSI awards per year in 2001, 4.42 in 2002, and 4.20 in 2003.

Second, SSPs make a difference. For both children and adults, having recipients in SSPs produces a statistically significant reduction in the number of SSI awards that SSA's recording procedures connect to TANF.

Third, the sum of coefficients on the SSP caseloads for the present and preceding years exceeds in absolute size the corresponding sums for combined TANF and SSP caseloads (including the predicted awards factor). The implication is that knowing that an adult or child is in a Separate State Program raises the likelihood of that person's receiving an SSI award over what is associated with knowing that the person receives support through TANF.

Finally, for all 3 years of data, a child's being in an SSP seems to count for less than it does for adults. This finding is consistent with our hypothesis that because children are not the likely targets of SSP initiatives, the marginal effect on predicted awards of knowing that children are in SSPs should be smaller in absolute magnitude than the corresponding effect for adults. Nevertheless, the consequence of enrollment in SSPs for predicted TANF-related awards is substantial for both adults and children.

These regressions are rather unsophisticated. In particular, we have not made any adjustment for the obvious problem of heteroscedasticity in the disturbance terms (in 2003, adult TANF recipients in California received 64 times as many SSI awards as did recipients in South Dakota). For reasons that are unclear, the fit of the 2001 regressions appears inferior to what is achieved for 2002 and 2003. Nevertheless, these simple models provide the basis for a rough calculation of the degree to which current awards are understated because of the SSP problem.

The Adjustment

We use the regression results to estimate for each state with an SSP what TANF-related awards would have been had all the SSP cases been left in TANF. We use the estimated coefficients for SSP_s,t and SSP_s,t-1 to calculate awards missed and then add this number to awards actually reported to create an estimate of actual TANF-related awards for each state with an SSP. These totals are then combined with reported awards for states without SSPs to produce an estimate of TANF-related awards nationally in the absence of SSPs. The difference between this prediction and actual reported TANF-linked SSI awards is our estimate of the undercount. These results are summarized in Table A-3. The message of the regressions is that by 2003 states' use of SSPs led to an understatement of TANF-related SSI awards of 26 percent for adults and 24 percent for children.

Table 5 in the body of this article combines these estimates with the preadjustment data in Table 3. The SSP adjustment makes a significant difference. As reported there, we estimate that failure to account for SSPs produces an undercount of approximately 6.3 percentage points in the proportion of child SSI awardees coming from TANF households and of 2.5 percentage points in the number of adults over the calendar year 2001–2003 period.