Table of Contents | Previous | Next |
Chapter Six
USE OF OTHER GOVERNMENT PROGRAMS
6.1. BACKGROUND
Many families affected by welfare reform are likely to remain eligible for other U.S. government programs in the "safety net." The largest such programs are the Food Stamp Program (FSP) and Medicaid. Low-income families also often qualify for housing assistance and various other nutrition programs, such as the School Lunch Program, the School Breakfast Program, and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC).
The quantity of research analyzing the effects of welfare reform on the use of these government programs varies widely. Most of the random assignment studies estimate food stamp impacts. This is probably because, in many states, the same administrative data system is used to track both welfare receipt and food stamp use. Thus, administrative data on food stamp receipt are nearly as readily available to researchers as administrative data on welfare use. Aggregate caseload data and survey-based information on food stamps are also available, but they have been less utilized for econometric studies than the corresponding welfare data.
Fewer random assignment studies, and only one econometric study, have considered the effects of welfare reform on Medicaid enrollment. Moreover, most of the random assignment studies that address the question do so using survey data, which generally include fewer observations than the administrative data used to analyze other outcomes. Only a few studies consider participation in other government programs, and those that do are based entirely on survey data.
Before discussing the results from studies that attempt to estimate the causal effects of welfare reform on the use of these various government programs, we provide some background on the programs, recent trends in participation, and, where relevant, provisions of PRWORA that affect those programs directly. The amount of detail we provide on each program is roughly proportional to the number of research studies available to synthesize.
6.1.1. The Food Stamp Program
The FSP provides low-income households with coupons–or more recently, electronic benefits accessed by means akin to a debit card–that can be used to purchase food. Households must have incomes less than 130 percent of the federal poverty line (FPL) to qualify. In addition, the household’s net income–equal to its income less certain deductions–must fall below the FPL. The household must also satisfy asset tests.
Benefit levels are uniform in the continental United States. A family of three with no other source of income is eligible for $341 in monthly benefits. For families with other income, including income from TANF, benefits are reduced by 30 cents for every dollar of net income. As a result, a three-person household with income equal to half the FPL ($610/month) would be eligible for at least $235/month in food stamps. A three-person household with income equal to the FPL would be eligible for somewhere between $89 and $201 in food stamps (Zedlewski and Gruber, 2001).
Unlike cash aid programs, eligibility for food stamps is based on household composition rather than family structure. As a result, the FSP serves a larger client base than the TANF program. Of the 7.3 million households (which included 17.1 million persons) served by the FSP in 2000, 26 percent also received assistance from TANF (USDA, 2001a).
Although TANF households generally are eligible for food stamps, in 1999 only 81 percent of TANF households received FSP benefits (USDHHS, 2000). Non-TANF households must go through an eligibility determination process that requires substantial documentation of income sources. When families leave welfare, they may lose their eligibility for food stamps. If they become ineligible because of income gains, they stop receiving food stamp benefits. If they are unaware that they may remain eligible for food stamps, or if they are unable or unwilling to take time off from work or to document their sources of income, they may also stop receiving food stamps.47
Beyond these potential indirect effects on FSP use, PRWORA directly affected FSP eligibility for two groups of recipients: legal immigrants and able-bodied adults without dependents (ABAWDs). Legal immigrants who entered the United States after the passage of PRWORA were barred from receiving most kinds of aid, including food stamps. Legal immigrants present prior to the passage of PRWORA were to be dropped from the FSP rolls within a year, although the 1997 Balanced Budget Act repealed the latter provision. PRWORA limited ABAWDs to three months of FSP receipt in any 12-month period, unless they were working or participating in an approved work program. States can apply for waivers to exempt from the time limit ABAWDs who live in areas with high unemployment rates or insufficient jobs.
Recent trends in the food stamp caseload resemble recent trends in the welfare caseload. During the late 1980s, the program served about 19 million people per month. By 1994, the number of persons served by the FSP grew to 27.5 million; by 2000, it had fallen to 17.2 million, the lowest level since 1978. In relative terms, the FSP caseload fell 37 percent between 1994 and 2000.
Although participation rates for immigrants and ABAWDs fell substantially after 1996, they account for a small proportion of the overall caseload decline because they account for a relatively small share of all food stamp households. The group accounting for the largest share of the decline was single-parent families, which includes a number of families leaving cash aid (USDA, 1999).
Leavers studies show that former welfare recipients leave the FSP at a high rate. Among 13 cohorts tracked in twelve USDHHS-funded leaver studies, the fraction of leavers receiving food stamp benefits in the first quarter after leaving welfare ranged from 23 to 78 percent (USDHHS, 2001a). Six of the studies had rates ranging from 35 to 54 percent, whereas two were lower and four were higher. Data from the National Study of America’s Families are similar, showing that only about one-third of recent welfare leavers continued to receive food stamps (Zedlewski and Gruber, 2001). Acs et al. (2001) suggest that, among long-term welfare recipients, these rates did not change much during the 1990s.
Although much of the decline in FSP use results from a decrease in eligibility arising from gains in income, over half of the decrease stems from a decrease in take-up rates on the part of income-eligible households (USDA, 2001b). National Study of America’s Families data show that about one-half of all leaver families have post-welfare incomes less than the FPL (Loprest, 2001), which implies that more than half of all leaver families remain income-eligible for food stamps. However, among leavers with incomes between 50 and 100 percent of the FPL, only 45 percent continued receiving food stamps. Among leavers with incomes below 50 percent of the FPL, only 51 percent received food stamps (Zedlewski and Gruber, 2001). Low take-up among the former group may stem from the relatively low benefit payment for which they would be eligible (Blank and Ruggles 1996). However, low take-up among the latter group represents nonparticipation by families who would be eligible for a relatively large payment.
Despite the usefulness of such contextual information, trends and patterns among welfare leavers do not establish whether welfare reform has caused FSP use to fall. As with other welfare-related outcomes, changes in the economy and other antipoverty policies, such as the EITC, may also underlie the recent declines in the FSP caseload. In Section 6.2 below, we review the evidence from a number of random assignment and econometric studies on the extent to which welfare reform has caused the decline in the FSP.
6.1.2. Medicaid and the State Children’s Health Insurance Program
The Medicaid program, which enrolls roughly 40 million Americans at an annual cost of about $176 billion, serves three largely distinct populations: the elderly, the blind or disabled, and poor families with children. Poor families with children account for about two-thirds of Medicaid beneficiaries but only about 25 percent of annual Medicaid expenditures (Health Care Financing Administration, 2000). Since this is the group most relevant to welfare reform, we restrict our attention to it below.
Originally, poor families with children received coverage under Medicaid primarily by qualifying for cash welfare. However, beginning in the mid-1980s, expansions of coverage to poor pregnant women and poor children not receiving welfare weakened the link between the two programs. The largest expansions provide eligibility to children depending on their age and their family’s income. By 1990, the states were required to provide coverage for all children under age six who lived in families with incomes less than 133 percent of the FPL. They were also required to cover all children born after September 30, 1983, in families with incomes below 100 percent of the FPL (Congressional Research Service, 1993). As a result, by 1997, all children under age 15 in families with incomes below the poverty line were eligible for Medicaid, regardless of whether their families received (or even qualified for) welfare. Many states exercised the option to extend coverage to older children and to children in families with higher incomes.
The Family Support Act of 1988 extended Medicaid to families that left welfare because of employment or earnings via Transitional Medical Assistance (TMA). Under FSA, TMA was available for one year after leaving welfare. Many states have extended the eligibility period under their TANF programs.
PRWORA officially ended the link between welfare and Medicaid by eliminating automatic Medicaid eligibility for welfare families and establishing a new eligibility category. Under section 1931 of the Social Security Act, states are required to extend Medicaid coverage to families who would have qualified for welfare under the AFDC eligibility rules that the state had in place as of July 1996. Section 1931 also provides states with the flexibility to expand Medicaid to cover more low-income families, including two-parent working families, at the state’s option.
Health care coverage was further expanded by the 1997 Balanced Budget Act, which established the State Children’s Health Insurance Program (SCHIP). SCHIP is designed to provide coverage for children in families with incomes up to 200 percent of the FPL who otherwise are ineligible for Medicaid (HCFA, 2000). About 30 states have chosen to cover children in families with incomes up to at least this level.
When Medicaid was tied to welfare, enrollments were fairly stable. Between 1975 and 1985, the number of poor children and adults covered by Medicaid grew from just over 14 million to just over 15 million. Enrollment grew to 17.2 million by 1990, then to 24.8 million by 1995 (Gruber, 2000). The number of covered persons in poor families fell between 1995 and 1997, which is the most recent year for which consistently defined data are available.48 Coverage of poor children fell by 12.4 percent between 1995 and 1997; coverage of poor adults fell by 11.2 percent (Committee on Ways and Means, 2000, Table 15-14).
The nationwide trend in enrollments masks substantially different trends among the states. While the overall decline in enrollment among persons in poor families was 5.3 percent between 1995 and 1997, ten states experienced increases in enrollment. At the same time, nine states experienced decreases of 10 percent or more (Ku and Bruen, 1999).
Leavers studies generally show low rates of Medicaid coverage among persons leaving welfare, but like the nationwide trend data, there is a fair amount of heterogeneity across different locales. Nine state-specific studies funded by USDHHS provide coverage rates for family heads in their first quarter after leaving welfare. Of these, eight report coverage rates ranging from 42 to 69 percent. Wisconsin showed higher coverage, at 80 percent (USDHHS, 2001a, Appendix B). Data from the National Study of America’s Families show that 56 percent of leavers are covered by Medicaid in the first six months after leaving aid (Garrett and Holahan, 2000).
Children are more likely than adults to retain eligibility, but the differential varies greatly by site. In Wisconsin, 86 percent of children were covered by Medicaid in the first quarter after their families left welfare, compared with 80 percent of adults. In Missouri, 42 percent of adults were covered, but 85 percent of children retained eligibility (USDHHS, 2001a). The National Survey of America’s Families shows that 70 percent of children in families that had been off welfare for six months or less retained their Medicaid coverage (Garrett and Holahan, 2000).
Since Congress specifically sought to ensure that families would not lose health care coverage as a result of welfare reform, there has been substantial interest in explaining recent declines in Medicaid enrollments. Some analysts have suggested that the strong economy has played a role, whereas others have pointed to confusion on the part of recipients about the new relationship between welfare and Medicaid programs (Kenney and Haley, 2001). Others have suggested that states’ TANF diversion programs may have deterred some families from completing their Medicaid applications (GAO, 1999c; Dion and Pavetti, 2000; Guyer, 2000). In Section 6.2, we synthesize the evidence on the extent to which welfare reform has caused the recent declines in Medicaid enrollments.
6.1.3. Other Programs
Although there has been relatively little research into the effects of welfare reform on participation in other nutrition programs or subsidized housing programs, it is useful to summarize some of the most important programs here. After the FSP, the largest nutrition programs are the National School Lunch Program (NSLP), the School Breakfast Program, and WIC. The school nutrition programs provide meals to primary and secondary school students; students from families with incomes below 130 percent of the FPL receive their meals for free and students from families with incomes between 130 and 185 percent of the FPL receive their meals at a reduced price. The WIC program provides vouchers for pregnant women, new mothers, and children up to age 5 that can be redeemed for specific food items that contain nutrients often lacking from the diets of low-income families.
Unlike the trend in food stamp participation, which was similar to trends in welfare use, participation in these three smaller nutrition programs grew through most of the 1990s. The number of children receiving free or reduced-price school lunches rose from 11.6 million per month (during the school year) in 1990 to 15.5 million in 2000. The number of children receiving free or reduced-price school breakfasts rose from 3.5 million per month in 1990 to 6.3 million in 2000. The number of persons served by the WIC program rose from 4.5 million in 1990 to 7.4 million in 1997, then fell to 7.2 million in 2000. The costs of operating the school nutrition programs in 2000 (including the tiny Special Milk Program) amounted to $7.6 billion; the cost of the WIC program was $3.97 billion (USDA, 2001d). The General Accounting Office (GAO) (1999) interprets increases in school nutrition programs as evidence of program substitution, that is, as evidence that families are turning to programs other than the FSP in order to satisfy their demand for nutrition assistance.
Housing assistance of various kinds is provided through a number of different programs that are run by several different agencies (Olsen, 2001). However, most rental assistance funds targeted to very-low-income households are administered by the Department of Housing and Urban Development. Such aid is provided in one of two forms: project-based assistance or household-based rental subsidies (Committee on Ways and Means, 2000). The number of renters assisted by these programs grew more or less steadily during the last decade, from 1.6 million households in 1990 to 2.1 million households in 2000.
6.2. Random Assignment and Econometric Studies of the Effects of Welfare Reform on Use of Other Government Programs
As noted above, there are relatively fewer random assignment and econometric studies examining the effects of welfare reform on the use of other government programs. Thus, unlike in the previous chapters, we combine the discussion of random assignment and econometric studies, breaking the discussion up by the three categories of other government programs presented above. The limited amount of information from the random assignment studies regarding subgroup differences in the outcomes covered in this chapter is summarized in Appendix A.
6.2.1. Food Stamps
Of the welfare-related outcomes considered in this chapter, the Food Stamp Program is the most studied. Most of the random assignment studies analyze the effects of welfare reform on food stamp use. In addition, four econometric studies address the question using national data from either administrative sources or the CPS.
Random Assignment Studies
Table 6.1 presents results from the random assignment studies. The only study focusing on financial work incentives that provides information about food stamp usage is the WRP Incentives-Only program, since MFIP involved a food stamp cashout. The WRP Incentives-Only program had no significant effect on food stamp use, just as it had no significant effect on welfare use.
| Food Stamp use | ||||||
|---|---|---|---|---|---|---|
| Name | Cases served | Data | Measure | Control mean |
Impact | % |
| A. Programs that focus on financial work incentives | ||||||
| WRP-IO | Single-parent recipients and applicants | A | Ever received FS, last 3 mos. of FU | 50.1 | 1.0 | 2.0% |
| B. Programs that focus on financial work incentives tied to hours of work | ||||||
| New Hope | Poor families employed FT at RA | A | Months receiving FS, year 1 of 2-yr FU | 5.3 | -0.3 | -5.7% |
| A | Months receiving FS, year 2 of 2-yr FU | 4.5 | 1.0** | 22.2% | ||
| Poor families not employed FT at RA | A | Months receiving FS, year 1 of 2-yr FU | 7.5 | -0.1 | -1.3% | |
| A | Months receiving FS, year 2 of 2-yr FU | 5.2 | 0.4 | 7.7% | ||
| C. Programs that focus on mandatory work-related activities | ||||||
| LA Jobs-1st GAIN | Single-parent recipients and applicants | A | Received FS, Q8 | 64.5 | -4.2*** | -6.5% |
| Atlanta LFA | Recipients and applicants | A | Received FS, Q8 | 76.9 | -1.2 | -1.6% |
| Grand Rapids LFA | Recipients and applicants | A | Received FS, Q8 | 67.3 | -5.8*** | -8.6% |
| Riverside LFA | Recipients and applicants | A | Received FS, Q8 | 54.4 | -7.6*** | -14.0% |
| Portland | Recipients and applicants; no cases with substantial barriers | A | Received FS, Q8 | 63.3 | -4.6*** | -7.3% |
| Atlanta HCD | Recipients and applicants | A | Received FS, Q8 | 76.9 | -1.0 | -1.3% |
| Grand Rapids HCD | Recipients and applicants | A | Received FS, Q8 | 67.3 | -3.8** | -5.6% |
| Riverside HCD | Recipients and applicants | A | Received FS, Q8 | 57.6 | -5.5*** | -9.5% |
| Columbus Integrated | Recipients and applicants | A | Received FS, Q8 | 64.0 | -6.0*** | -9.4% |
| Columbus Traditional | Recipients and applicants | A | Received FS, Q8 | 64 | -4.0*** | -6.3% |
| Detroit | Recipients and applicants | A | Received FS, Q8 | 81.7 | -3.5*** | -4.3% |
| Oklahoma City | Applicants | A | Received FS, Q8 | 56.0 | 0.4 | 0.7% |
| IMPACT Basic Track | Recipients and applicants-basic track | A | Received FS, Q4 | 61.8 | 1.0 | 1.6% |
| D. Programs that focus on financial work incentives and mandatory work-related activities | ||||||
| WRP | Single-parent recipients and applicants | A | Ever received FS, last 3 mos. of FU | 50.1 | -1.6 | -3.2% |
| TSMF | Recipients | A | Monthly FS receipt over 4-yr FU | 69.1 | -0.9*** | -1.3% |
| Applicants | A | Monthly FS receipt over 1-yr FU | 67.6 | -0.8 | -1.2% | |
| A | Monthly FS receipt over 2-yr FU | 61.2 | -2.3*** | -3.8% | ||
| FIP | Recipients | A | FS receipt, Q4 | 79.3 | -1.9* | -2.4% |
| A | FS receipt, Q8 | 66 | -2.1* | -3.2% | ||
| Applicants | A | FS receipt, Q4 | 44.9 | -0.2 | -0.4% | |
| A | FS receipt, Q8 | n/a | ||||
| E. Programs that focus on other individual reforms | ||||||
| F. Programs that focus on TANF-like bundle of reforms (time limits with financial incentives, work-related activities, or both) | ||||||
| EMPOWER (a) | Recipients | A | Monthly FS receipt, months 1-36 | 47.2 | -1.3* | -2.8% |
| IMPACT Placement Track |
Recipients and applicants-placement track | A | Received FS, Q4 | 66.2 | -7.0*** | -10.6% |
| VIP/VIEW | Recipients | A | Welfare receipt in Q8 | 70.1 | -2.6*** | -3.7% |
| ABC | Recipients and applicants | S | Percent receiving FS at spring 1997 interview | 69.5 | -4.2* | -6.0% |
| FTP | Recipients and applicants | A | Avg. percent receiving FS, year 2 | 60.6 | -0.9 | -1.5% |
| A | Avg. percent receiving FS, year 3 | 48.8 | -0.4 | -0.8% | ||
| A | Avg. percent receiving FS, year 4 | 40.7 | -0.7 | -1.7% | ||
| JOBS First | Recipients and applicants | A | Ever received FS, Q7 | 61.6 | 3.1** | 5.0% |
| A | Ever received FS, Q8 | 58.8 | -0.6 | -1.0% | ||
| A | Ever received FS, Q16 | 42.5 | -3.3** | -7.8% | ||
|
NOTES: For full program names and citations,
see Table 3.4. Abbreviations: A=administrative data; S=survey data;
FU=follow-up; HH=household; Q=quarter; FS=Food Stamps; FT=full-time. |
Of the programs focusing on financial incentives tied to hours of work, food stamp impacts are available only for New Hope, because the SSP programs were conducted in Canada. The New Hope program resulted in increased food stamp use at the end of the second year among families employed at the time of random assignment. Since the same group was less likely to use welfare at that time (see Table 4.1), this suggests that these New Hope participants partially replaced their reduced welfare payments with increased food stamp benefits. However, among families not initially employed full time, New Hope had no significant effect on food stamp use, just as it had no effect on welfare use.
Most of the programs that focused on mandatory work-related activities resulted in significant declines in food stamp use. The four exceptions were the Atlanta LFA, Atlanta HCD, Oklahoma City, and IMPACT Basic Track programs. The average reduction in FSP use across the thirteen programs was 3.5 percentage points, which amounts to 62.5 percent of their average 5.6 percentage point reduction in welfare use. This suggests that about 65 percent of the families leaving welfare discontinued their food stamp benefits.49
The programs combining financial work incentives and mandatory work-related activities had effects on food stamp use that are similar to their effects on welfare use. WRP, which involved both a weak financial incentive and a delayed work requirement, had no significant effect on either. Reductions in food stamp use among TSMF recipients are similar to their reductions in welfare use. For the most part, the reductions in food stamp use are slightly smaller, although the earlier cohort of TSMF applicants actually reduced their food stamp use by a slightly greater amount than their welfare use. An exception to the general rule is FIP, where food stamp use fell despite increases in welfare use.50
Among programs that focused on TANF-like reform bundles, Indiana’s Placement Track recipients had lower food stamp use than the controls. The reduction in food stamp use was less than the reduction in welfare use in the fourth quarter of the program but was slightly greater by the eighth quarter. Arizona’s EMPOWER program and Virginia’s VIP/VIEW program also had somewhat larger effects on food stamps than on welfare use. The food stamp results for Delaware’s ABC program cannot be compared to the corresponding welfare-use results because the measures differ across the two outcomes.
In the pre—time limit period, FTP and Jobs First had effects on food stamp use that were similar to their effects on welfare use, although the food stamp impacts are insignificant. In the post—time limit period, food stamp use was lower in the treatment group compared with the control group for both programs, but only the Jobs First impact was significant, and then only in quarter 16. All the post—time limit impacts were substantially smaller than the corresponding impacts on welfare use.
The changes in program impacts between the pre— and post—time limit periods suggest that reaching the welfare time limit had relatively little effect on food stamp use. In FTP, food stamp use actually rose by 0.5 percentage points between years two and three. In Jobs First, food stamp use fell by 3.7 percentage points between quarters seven and eight, which amounts to about 30 percent of the corresponding decline in welfare use. Apparently, most of the families who exhausted their welfare benefits continued their food stamp use, at least initially. Although the impact of FTP stayed roughly constant through year four, the impact of Jobs First became more negative.
Econometric Studies
The four econometric studies of the effects of welfare reform on food stamp use are summarized in Table 6.2. The studies by Wallace and Blank (1999) and Currie and Grogger (2001) employ models that are comparable to the static models of welfare caseloads summarized in Chapter 4. They also employ similar policy measures to capture the effects of welfare reform waivers and TANF. Figlio, Gunderson, and Ziliak (2000) and Wilde et al. (2000) employ lagged dependent variable specifications similar to the welfare caseload model of Figlio and Ziliak (1999).
| Other controls | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Data | Sample population | Begin | End | Outcome | Dep. var. | Policy var. | Coeff. (s.e.) | % effect | Includes LDV's? | Economy | Demogr. | Fixed Effects | Policy |
| Wallace and Blank (1999) | FS caseload | Total population | 80 | 96 | annual FS caseload/total population | Log | Any waiver | -0.032 (0.017) | -3.2 | no | U, U-1, U-2, median wage, 20th percentile wage |
%elderly, %immigrants-1, %immigrants-2, %female heads, %non-marital births | S, Y, State time trends |
B, party control |
| 80 | 98 | monthly FS caseload | Log | Any waiver (12 lags) | -0.025 (0.000) | -2.5 | U (12 lags) | S, state-specific month dummies | ||||||
| TANF (12 lags) | -0.137 (0.000) | -13.7 | ||||||||||||
| Currie and Grogger (2001) | CPS micro data | HH w/ incomes < 300% of FPL | 80 | 98 | Any FS last year | Level | Any waiver | -7.800 (3.500) | -5.4 | no | U | A, E, R, # kids, MSA | S, Y, State time trends |
B, EBT, mean recertifica-tion interval |
| TANF | -16.200 (5.900) | -11.2 | ||||||||||||
| Figlio, Gunderson, and Ziliak (2000) | FS caseload | Total population | 80 | 98 | annual FS caseload / population | Log | Any waiver | 1.014 (0.747) | 1.0 | yes (4 lags) | U (4 lags), EG (4 lags) | S, Y, State time trends |
B, EBT, ABAWD waiver | |
| Wilde, et al. (2000) | FS caseload | Total population | 80 | 98 | annual FS caseload / population | Log | Any waiver | 0.621 (0.757) | 0.6 | yes (4 lags) | U (4 lags), EG (4 lags) | S, Y, State time trends |
B, EBT, ABAWD waiver, political climate | |
| NOTES: Abbreviations: LDV=lagged dependent variable;
s.e.=standard error; U=unemployment rate;U-1=lagged unemployment rate;
EG=employment growth; A=age, E=education, R=race, B=maximum welfare
benefit, MW=minimum wage; ABAWD=able-bodied adults without dependents; EBT=electronic benefit transfer system; EITC=Earned Income Tax Credit; S=state; Y=year. |
The estimates of Wallace and Blank and Currie and Grogger are fairly similar, despite the fact that they are based on different data. Wallace and Blank estimate that waivers reduced food stamp use by 2.5—3.2 percent, depending on whether they use monthly or annual food stamp caseload data. They estimate that welfare reform reduced food stamp caseloads by about 14 percent. Currie and Grogger’s estimates, which are based on CPS data, indicate that waivers reduced food stamp use by 5.4 percent and that TANF reduced food stamp use by 11.2 percent. Both studies suggest that welfare reform accounts for a substantial fraction of the decline in the FSP caseload.
Figlio et al. and Wilde et al. both include four lags of food stamp caseloads in their model and report that waivers had no significant effect on food stamp use. These models are subject to the same technical and interpretational problems as the lagged dependent variable models of welfare caseloads, as discussed in Chapter 4. As in the studies of welfare caseloads, the lagged dependent variable models yield results that run contrary not only to the results from the other econometric studies, but also to the results from most of the random assignment studies.
6.2.2. Medicaid
Random Assignment Studies
Only seven of the random assignment studies report information on Medicaid receipt.51 Their results are summarized in Table 6.3. Much of the information about Medicaid use comes from surveys rather than administrative data. As a result, sample sizes are smaller and significance levels are correspondingly lower.52
| Medicaid coverage | ||||||
|---|---|---|---|---|---|---|
| Name | Cases served | Data | Measure | Control mean | Impact | % |
| A. Programs that focus on financial work incentives | ||||||
| MFIP-IO | Urban single parents recipients | S | Adult covered by Medicaid 36 mos. after RA | 66.2 | 4.2 | 6.3% |
| B. Programs that focus on financial work incentives tied to hours of work | ||||||
| C. Programs that focus on mandatory work-related activities | ||||||
| LA Jobs-1st GAIN | Single-parent recipients and applicants | A | Months of Medicaid coverage, years 1 and 2 | 21.2 | -0.9*** | -4.2% |
| D. Programs that focus on financial work incentives and mandatory work-related activities | ||||||
| MFIP | Urban single-parent recipients | S | Adult covered by Medicaid at 36-month interview | 66.2 | 6.4* | 9.7% |
| TSMF | Recipients | A | Adult average monthly Medicaid eligibility over 4-yr FU | 67.6 | -1.7*** | -2.5% |
| Applicants | A | Adult average monthly Medicaid eligibility over 1-yr FU | 70.2 | 0.6 | 0.9% | |
| A | Adult average monthly Medicaid eligibility over 2-yr FU | 61.2 | -1.0 | -1.6% | ||
| FIP | Recipients | A | Any paid claims, Q4 | 89.5 | -3.1*** | -3.5% |
| A | Any paid claims, Q8 | 76.5 | -1.7 | -2.2% | ||
| Applicants | A | Any Medicaid action, Q4 | 58.1 | 2.4 | 4.1% | |
| A | Any Medicaid action, Q8 | n/a | ||||
| E. Programs that focus on other individual reforms | ||||||
| F. Programs that focus on TANF-like bundle of reforms (time limits with financial incentives, work-related activities, or both) | ||||||
| ABC | Recipients and applicants | S | Percent receiving Medicaid at interview | 76.7 | -2.0 | -2.6% |
| FTP | Recipients and applicants | S | Respondent covered by Medicaid at 4-year survey | 36.8 | -2.6 | -7.1% |
| JOBS First | Recipients and applicants | S | Respondent covered by Medicaid at 3-year survey | 60.4 | 9.1*** | 15.1% |
| NOTES: For full program names and citations, see
Table 3.4. Abbreviations: A=administrative data; S=survey data; FU=follow-up;
HH=household; Q=quarter. * = statistically significant at the 10 percent level; ** = statistically significant at the 5 percent level; *** = statistically significant at the 1 percent level. |
For the most part, these programs had effects on Medicaid qualitatively similar to their effects on welfare use. Both MFIP-IO and MFIP increased Medicaid coverage at the 36-month mark, although only the estimate for regular MFIP is significant. This is consistent with these programs’ effects on welfare use. Both programs increased welfare use significantly in year three, and since welfare recipients were automatically entitled to receive Medicaid at the time, Medicaid eligibility should have risen as well.
Similarly, L.A. Jobs-First GAIN decreased Medicaid use, much as it decreased welfare use. The program decreased welfare use by 1.12 months and Medicaid use by 0.89 months, over the 24-month follow-up period. (Freedman et al., 2000b, Table 4.1, p. 74). This suggests that roughly 80 percent of those who left welfare as a result of the program also discontinued their Medicaid receipt.
In the Michigan TSMF program, Medicaid eligibility fell for ongoing recipients by about the same amount as welfare use. For the applicant groups, however, the effects of the program on Medicaid were insignificant, even though the program reduced welfare use among the early cohort of applicants by a significant amount. FIP is again an exception to the general rule, resulting in decreased Medicaid use despite increases in welfare use. Delaware’s ABC program had no significant effect on Medicaid, just as it had no significant effect on welfare use. FTP had little effect on Medicaid at the time of the survey, despite reducing welfare use. Jobs First substantially increased Medicaid coverage at the time of the three-year follow-up at the same time that it substantially decreased welfare use. Jobs First extended TMA eligibility one additional year, which may account for this pattern, although the impact of this particular policy component can not be isolated in the Jobs First experimental design. The only other program that extended TMA was ABC, and that program produced no change in welfare use or Medicaid use.
Econometric Studies
The single econometric study to examine the effects of welfare reform on Medicaid caseloads is summarized in Table 6.4. That study, by Ku and Garrett (2000), uses state-level administrative data on nonelderly, nondisabled adults and children receiving Medicaid. The authors fit separate models for adults and children and include an extensive set of policy variables and controls in both specifications. They also include state and year dummies to control for unobservables.
| Other controls | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Data | Sample population | Begin | End | Outcome | Dep. var. | Policy var. | Coeff. (s.e.) | % effect | Includes LDV's? | Economy | Demogr. | Fixed Effects | Policy |
| Ku and Garrett (2000) | State-level Medicaid caseload | Adults in 44 states | 84 | 96 | annual non-disabled adult recipients/annual adult population | Log | Any waiver | -0.008 (0.022) | -0.8 | no | U, earnings | %men insured, %FHHH, %Hispanic |
S, Y | B, size of Medicaid expansion, medically needy program, UP program |
| Earnings disregards | 0.110 (0.082) | 11.0 | ||||||||||||
| Family cap | -0.042 (0.032) | -4.2 | ||||||||||||
| Percent exempt from JOBS | 0.114 (0.083) | 11.4 | ||||||||||||
| Ku and Garrett (2000) | State-level Medicaid caseload | Children in 44 states | 84 | 96 | annual non-disabled child recipients/annual child population | Log | Any waiver | -0.002 (0.021) | -0.2 | no | U, earnings | %men insured, %FHHH, %Hispanic |
S, Y | B, size of Medicaid expansion, medically needy program, UP program |
| Earnings disregards | -0.031 (0.081) | -3.1 | ||||||||||||
| Family cap | -0.024 (0.031) | -2.4 | ||||||||||||
| Percent exempt from JOBS | 0.153 (0.080) | 15.3 | ||||||||||||
| NOTES: Abbreviations: LDV=lagged dependent variable; s.e.=standard error; FHHH=female-headed households; U=unemployment rate; B=maximum welfare benefit, S=state; Y=year. |
The models provide some evidence that welfare waivers affect Medicaid use. Although the authors find that earnings disregards and family caps had no significant effects on Medicaid use, they find that fewer exemptions from JOBS work requirements had a negative and marginally significant effect. Since waivers generally reduced the fraction of the welfare caseload exempt from work requirements, this suggests that welfare waivers may have reduced Medicaid receipt among both adults and children.
6.2.3. Other Programs
Only four sets of random assignment studies have analyzed the effects of welfare reform on participants use of other nutrition programs: the NEWWS programs, L.A. Jobs-First GAIN, Indiana’s IMPACT evaluation, and Delaware’s ABC evaluation. The results of the effects of these programs on participation in the various nutrition programs are summarized in Table 6.5. All these results are based on survey data.53 To our knowledge, there are no econometric studies that analyze the effects of welfare reform on school nutrition programs or the WIC program.
| Nutrition program participation | ||||||
|---|---|---|---|---|---|---|
| Name | Cases served | Data | Measure | Control mean | Impact | % |
| A. Programs that focus on financial work incentives | ||||||
| B. Programs that focus on financial work incentives tied to hours of work | ||||||
| C. Programs that focus on mandatory work-related activities | ||||||
| LA Jobs-1st GAIN | Single-parent recipients and applicants | S | Ever participate in school meal program during FU | 66.3 | 2.4 | 3.6% |
| Atlanta LFA | Recipients and applicants | S | Ever participate in school meal program during FU | 86.2 | 1.8 | 2.1% |
| Grand Rapids LFA | Recipients and applicants | S | Ever participate in school meal program during FU | 67.1 | 1.2 | 1.8% |
| Riverside LFA | Recipients and applicants | S | Ever participate in school meal program during FU | 78.1 | -1.8 | -2.3% |
| Portland | Recipients and applicants; no cases with substantial barriers | S | Ever participate in school meal program during FU | 66.1 | -1.6 | -2.4% |
| Atlanta HCD | Recipients and applicants | S | Ever participate in school meal program during FU | 86.2 | 3.4** | 3.9% |
| Grand Rapids HCD | Recipients and applicants | S | Ever participate in school meal program during FU | 67.1 | -1.5 | -2.2% |
| Riverside HCD | Recipients and applicants | S | Ever participate in school meal program during FU | 81.4 | 0.4 | 0.5% |
| Columbus Integrated | Recipients and applicants | S | Ever participate in school meal program during FU | 75.6 | -1.4 | -1.9% |
| Columbus Traditional | Recipients and applicants | S | Ever participate in school meal program during FU | 75.6 | -0.9 | -1.2% |
| Detroit | Recipients and applicants | S | Ever participate in school meal program during FU | 60.2 | 1.2 | 2.0% |
| Oklahoma City | Applicants | S | Ever participate in school meal program during FU | 59.6 | -2.1 | -3.5% |
| D. Programs that focus on financial work incentives and mandatory work-related activities | ||||||
| E. Programs that focus on other individual reforms | ||||||
| F. Programs that focus on TANF-like bundle of reforms (time limits with financial incentives, work-related activities, or both) | ||||||
| IMPACT | Recipients and applicants-basic and placement tracks | S | Participating in school lunch at time of two-year survey | 60.9 | 2.7 | 4.4% |
| S | Participating in school breakfast at time of two-year survey | 47.3 | -2.5 | -5.3% | ||
| S | Participating in WIC at time of two-year survey | 23.2 | -1.7 | -7.3% | ||
| ABC | Recipients and applicants | S | Participating in school lunch at time of Spring 1997 survey | 52.3 | -0.9 | -1.7% |
| S | Participating in school breakfast at time of Spring 1997 survey | 45.0 | -0.1 | -0.2% | ||
| S | Participating in WIC at time of Spring 1997 survey | 35.4 | -2.3 | -6.5% | ||
| NOTES: For full program names and citations, see
Table 3.4. Abbreviations: A=administrative data; S=survey data; FU=follow-up;
HH=household; Q=quarter. * = statistically significant at the 10 percent level; ** = statistically significant at the 5 percent level; *** = statistically significant at the 1 percent level. |
Of the 18 estimates included in Table 6.5, all but one are fairly small and insignificant. The one exception involves the Atlanta HCD program, where School Lunch use rose by a statistically significant 3.4 percentage points. Since we would expect roughly 1 in 20 statistical estimates to be significant at the 5 percent level even if the programs truly had no effect, these results suggest that welfare reform has had little if any effect on participation on these other nutrition programs. They suggest that, in these programs, at least, families do not use school nutrition programs as a substitute for the FSP.
The only random assignment studies to include information about housing subsidies are L.A. Jobs-First GAIN, the NEWWS programs, EMPOWER, FTP, and Jobs First. All information comes from surveys in which participants were asked whether they lived in public housing projects or received rent subsidies. The results are summarized in Table 6.6.
| Public housing | Subsidized housing | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Name | Cases served | Data | Measure | Control mean | Impact | % | Control mean | Impact | % |
| A. Programs that focus on financial work incentives | |||||||||
| B. Programs that focus on financial work incentives tied to hours of work | |||||||||
| C. Programs that focus on mandatory work-related activities | |||||||||
| LA Jobs-1st GAIN | Single-parent recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 16.1 | 0.5 | 3.1% | 8 | 1.8 | 23.4% |
| Atlanta LFA | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 33.1 | 1.3 | 3.9% | 19.4 | 1.1 | 5.7% |
| Grand Rapids LFA | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 9.4 | -0.7 | -7.4% | 11.5 | -1.6 | -13.9% |
| Riverside LFA | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 6.3 | -0.4 | -6.3% | 7.9 | -0.5 | -6.3% |
| Portland | Recipients and applicants; no cases with substantial barriers | S | Lived in public/subsidized housing at end of 2-year FU | 18.6 | 1.1 | 5.9% | 17.0 | -6.5** | -38.2% |
| Atlanta HCD | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 33.1 | -1.0 | -3.0% | 19.4 | 4.4** | 22.7% |
| Grand Rapids HCD | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 9.4 | 0.3 | 3.2% | 11.5 | -0.5 | -4.3% |
| Riverside HCD | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 7.3 | -2.1 | -28.8% | 8.3 | -1.0 | -12.0% |
| Columbus Integrated | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 27.6 | 0.5 | 1.8% | 8.2 | -1.1 | -13.4% |
| Columbus Traditional | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 27.6 | -0.5 | -1.8% | 8.2 | -2.5 | -30.5% |
| Detroit | Recipients and applicants | S | Lived in public/subsidized housing at end of 2-year FU | 9.4 | 0.0 | 0.0% | 1.8 | 0.2 | 11.1% |
| Oklahoma City | Applicants | S | Lived in public/subsidized housing at end of 2-year FU | 12.6 | 0.3 | 2.4% | 9.5 | -0.3 | -3.2% |
| D. Programs that focus on financial work incentives and mandatory work-related activities | |||||||||
| E. Programs that focus on other individual reforms | |||||||||
| F. Programs that focus on TANF-like bundle of reforms (time limits with financial incentives, work-related activities, or both) | |||||||||
| EMPOWER (a) | Recipients | S | Lived in public or subsidized housing at wave 1 interview | 29.1 | 0.6 | 2.1% | |||
| FTP | Recipients and applicants | S | Lived in public or subsidized housing at 4-year interview | 22.1 | -1.3 | -5.9% | |||
| JOBS First | Recipients and applicants | S | Lived in public or subsidized housing at interim interview | 48.4 | -4.4 | -9.1% | |||
| NOTES: For full program names and citations,
see Table 3.4. Abbreviations: A=administrative data; S=survey data;
FU=follow-up; HH=household; Q=quarter. * = statistically significant at the 10 percent level; ** = statistically significant at the 5 percent level; *** = statistically significant at the 1 percent level. (a) Phoenix site only, cash assistance. |
All but two of the estimates are insignificant.54 Of the two significant estimates, one is negative and the other is positive. These estimates provide little reason to think that these types of reforms affect whether welfare recipients live in public or subsidized housing.
6.3. EVALUATING THE EFFECTS OF WELFARE REFORM ON USE OF OTHER GOVERNMENT PROGRAMS
In the random assignment studies, food stamp use generally follows welfare use. In programs where welfare use rose, such as MFIP, food stamp use rose as well. In studies where welfare use fell, food stamp use tended to fall as well. FIP and New Hope represent exceptions to this general pattern. In FIP, food stamp use fell despite increases in welfare use. In New Hope, families employed full time at random assignment increased their food stamp use while decreasing their welfare use.
In programs where welfare use fell, the corresponding decrease in food stamp use was generally smaller. This is consistent with the notion that welfare reform explains part, but not all, of the observed decline in the food stamp caseload that has taken place since the early 1990s. Evidence from two of the three econometric studies that address this issue is consistent with this view as well.
None of these studies explains why a decrease in welfare use results in a decrease in food stamp use. Most families leaving welfare, and most poor families generally, are eligible for food stamps, regardless of whether they receive cash aid. Thus, there is no automatic or legal link between welfare use and food stamp receipt. Various analysts have suggested that food stamp use has fallen in response to welfare reform because of administrative practices on the part of some welfare agencies, because the transactions costs of maintaining eligibility are higher for working families, and because poor families fail to understand that they may be eligible for food stamps even if they are ineligible for, or do not elect to receive, cash welfare.
There is substantially less evidence about the effects of welfare reform on Medicaid coverage. The single econometric study on the issue finds that less lenient age exemptions for work requirements reduce Medicaid use, but those estimates are only marginally significant. The random assignment studies suggest that programs that increase welfare use increase Medicaid receipt, whereas programs that decrease welfare use also decrease Medicaid receipt. Again, FIP is an exception to this rule, as is Jobs First. Moreover, there are only seven such studies that analyze Medicaid, which is a small number on which to base any general conclusions. It is also important to keep in mind that some recipients leaving welfare for work may replace their Medicaid coverage with employer-provided health insurance. We consider the effects of welfare reform on health coverage more generally in Chapter 9 below.
The effects of welfare reform on other government programs have been studied even less. Only four sets of studies analyzed participation in school nutrition programs. One of the NEWWS programs yielded a significant increase in program use; all the other estimates were rather small and insignificant. Only the NEWWS programs and L.A. Jobs-First GAIN consider housing subsidies; of the two significant effects, one was positive and one was negative. Regarding the effects of welfare reform on the use of these other government programs, there is simply too little information available to draw any firm conclusions.
6.4. CONCLUSIONS
The results synthesized in this chapter are generally consistent with the hypothesis that welfare reform has caused part of the recent decline in food stamp use. However, they tell us nothing about the mechanisms that underlie this linkage. As noted above, the research literature has suggested that food stamp use has fallen in response to welfare reform because of administrative practices on the part of some welfare agencies, because the transactions costs of maintaining eligibility are higher for working families, and because poor families fail to understand that they may be eligible for food stamps even if they are ineligible for, or do not elect to receive, cash welfare. To ensure that poor families are more effectively covered by this important safety net program, it is important to understand the reasons why welfare reform has caused food stamp use to fall.
However, perhaps the main message from this chapter is that there is very little information available from which to draw conclusions about the effects of welfare reform on the use of other government programs. Only a few studies analyze the effects of welfare waivers on the use of school nutrition programs. Fewer still have considered housing subsidies.
In the case of the Medicaid program, this limited research base is unfortunate in light of the policy objectives of recent reforms. During the decade prior to PRWORA, the federal government expanded the eligibility criteria for Medicaid with the dual objectives of weakening the links between welfare and health coverage and providing coverage for larger numbers of poor children. In the context of PRWORA, the legislation explicitly sought to ensure that reduced eligibility for welfare would not entail reduced eligibility for Medicaid. Nevertheless, Medicaid coverage of persons in poor families fell during the late 1990s at the same time that rates of uninsurance rose. Unfortunately, the evidence that exists provides too narrow a basis from which to draw general conclusions about whether welfare reform has caused part of the decline in Medicaid.
47TANF leavers participate in the FSP at a lower rate than TANF recipients, but at a higher rate than comparable families who never received welfare (Zedlewski and Brauner, 1999).(back)
48In 1998, for the first time, capitation payments were counted as services for reporting purposes (HCFA, 2000, p. 14).(back)
49The five-year NEWWS report does not provide year-by-year impacts for food stamp use.(back)
50Information on the use of food stamps and Medicaid among FIP applicants is provided only for the first four quarters of the follow-up period.(back)
51The NEWWS programs provide impact estimates for TMA, but not for Medicaid coverage more generally. Since TMA is available only to persons leaving welfare because of employment or an earnings gain, it is a function of both Medicaid-related policy changes and the programs' impacts on employment. Moreover, TMA is only one means by which study participants may receive Medicaid coverage. For these reasons, TMA impacts are not comparable to the more general Medicaid impacts presented in Table 6.3 and are excluded from our analysis.(back)
52None of the studies provides information about SCHIP, since SCHIP was not funded until 1998, after the follow-up periods of the studies that included questions about Medicaid participation.(back)
53Estimates for IMPACT are not available separately for participants in the Basic and Placement Tracks.(back)
54Impacts from the Jobs First four-year survey were also insignificant.(back)