[ Main page of Report | Contents of Report ]
The econometric models support the claim that state fiscal capacity exerts positive effects on total spending for social programs as well as spending for major components of social welfare budgets, such as cash assistance, medical assistance, and other public welfare programs. However, the models raise two sets of new questions:
The econometric models, therefore, fail to fully explain important differences across program areas and states, though the models narrow and focus the explanatory problems.
The remainder of this analysis explores these unexplained differences among states and program areas in three ways. First, we examine the estimated state effects from the econometric models among the six case study states, all of which have high social needs and low fiscal-capacity. How do these state effects relate to each other? Do we find clusters of states with similar spending patterns?
Second, after we identify state differences in spending patterns based on the econometric models, we determine whether those findings are consistent with changes in state spending after 2000, the last year in the Census data, in our six case study states. These states, along with nearly all states in the U.S., experienced fiscal difficulties after the economic downturn of 2001 and 2002. And we consider whether our statistical findings are consistent with or depart from the spending decisions these poor states made in the years after 2000, including the fiscal crises. Although we have no comparable Census data for these years, we can examine changes in roughly comparable spending based on administrative data on program expenditures and on reports of decisions from interviews and documents obtained during site visits in middle and late 2003.
Third, we use the site visits and especially the discussions with state officials to understand how spending choices are made in different program areas and what factors influence those choices. We want to understand what decision-making processes might generate the statistical relationships found in the econometric models, how and whether these processes vary across program areas, and what state characteristics affect those processes. Succinctly stated, we found that spending decisions in different program areas are made in different ways; that these different modes of decision-making respond to different state characteristics; and that these differences in processes and influences might help explain why certain variables are important, or unimportant, in the econometric analyses and why poor states vary in their spending patterns. These findings might also help us understand some of the trends in state spending on social programs noted in subsection IIIA. And, they hold important implications for the broader questions posed in this study about federal influence over state spending, packages of social programs, and effects of economic cycles.
[ Go to Contents ]
The division among the poor states we noted in section III was also found among the six poor states selected for site visits. Exhibit IV-1 shows the estimated state effects for cash assistance, Medicaid, non-health social services, public hospitals, and non-social welfare. Again, these state effect scores were scaled to equal zero for the mean state effect and 100 (or -100) for a state effect one standard deviation above, or below, the mean.
|Cash Assistance||Medicaid||Non-health Social Services||Public Hospitals||Non-social Welfare|
New Mexico and Arizona were higher than the southern states on cash assistance spending but lower on Medicaid and public hospitals. Mississippi, Louisiana, South Carolina, and West Virginia were all low on cash assistance as well as spending on non-social welfare. Yet they were relatively high on payments to medical vendors or public hospitals. Spending on non-health services was less clearly related to other expenditures. It was high in New Mexico but low in Arizona; high for West Virginia but lower in Mississippi, Louisiana, and South Carolina.
These basic differences among poor states also seemed to be reflected in spending changes from 1977 to 2000, as displayed in Exhibit IV-2. Because the spending patterns of the two western states generally differed from those found among the four southern and border states, we divided the states into those two groupings for the purpose of comparing trends. Thus, the left side of Exhibit IV-2 shows the changes in spending on cash assistance, Medicaid, and non-health social services for New Mexico and Arizona, while the right side traces those changes for Mississippi, Louisiana, South Carolina, and West Virginia.
Trends in Spending on Social Welfare Programs Among Six Poor States, 1977-2000
|Arizona and New Mexico
|Louisiana, Mississippi, South Carolina, and West Virginia
|Non-health Social Services
||Non-health Social Services
The most consistent pattern in spending changes across all states was the strong growth of Medicaid in the early 1990s. Most of the states continued to show growth or at least stability for the remainder of the decade. The two exceptions were Arizona and Louisiana, which showed some decline in real spending on Medicaid in the late 1990s.
Most states also showed a marked decline in spending on cash assistance from the middle 1990s onward, due presumably to a combination of welfare reforms and economic growth. However, the western states differed from the others in that this decline came after a substantial increase in spending on cash assistance in the early 1990s. The southern states experienced no such increase. In fact, Mississippi and Louisiana were already seeing declines in real spending on cash assistance before the bigger drops in spending of the middle and late 1990s. Changes in non-health social services spending were typically less dramatic. Most states showed overall though slow increases in spending over these years, with the western states and West Virginia showing more volatility than the southern states.
Perhaps the most striking finding from these data involves the different linkages between spending trends in each state. In Arizona and New Mexico, spending changes in cash assistance, Medicaid, and non-health social services showed a rough correspondence, with slow change before the 1990s, strong growth in the early 1990s, and some stalling or even a decline in real spending in the late 1990s. Arizona, for example, showed strong increases in all three forms of spending in the early 1990s, reversals in the late 1990s, and an uptick in 2000. In contrast, spending changes in other states showed less linkage or even negative relationships. In Mississippi and South Carolina, Medicaid and non-health social services grew rapidly while spending on cash assistance grew slowly or fell. West Virginia and Louisiana showed little linkage in spending trends across functional areas.
Thus, the dynamic patterns of spending reflected some of the state differences we found in the estimated state effects. Arizona and New Mexico treated all three major functional spending areas in roughly similar ways-to some extent, like the wealthier states did. Their spending on cash assistance was not extremely low, nor did it show secular decline. The other states showed a sharp divergence between long-run staticity or decline in cash assistance spending and rapid growth in spending on Medicaid, starting in the early 1990s. In these latter states, other public welfare spending also grew, though usually more fitfully and less strongly. Thus, at least among the southern states, little evidence of a linked package of social welfare programs emerged. In fact, different program areas showed some indications of negative relationships over time.
[ Go to Contents ]
Although we have no Census data after 2000, we can draw from the site visits and administrative spending data on selected programs to determine whether these six poor states continued or departed from prior trends and differences. Subsection IVB examines recent changes in spending among the six states and finds, for the most part, that the state trends and differences noted in the Census data through FY 2000 continued with some modifications after that year. Before we trace those spending changes, however, we must put them into context by briefly discussing the most important change in the recent fiscal environment, the state fiscal crises and especially the revenue shortfalls of 2001 and thereafter.
At the time of our site visits, all six states were experiencing difficult fiscal problems. Yet these problems varied between rich and poor states as well as among these six states. The crisis was driven largely by a sharp drop in revenues in 2001 and 2002 (Boyd 2003). The declines were greatest among the comparatively wealthy states in the Northeast and Far West. They were less severe in the Middle Atlantic and Rocky Mountain states, while the declines were smallest among states in the Great Lakes area, the South, and the Plains states.
One behind this geographical variation was the source of states' tax revenues. Because the greatest decline in revenues occurred in individual and corporate income taxes, states that relied heavily on consumption taxes saw no extreme declines in revenues. In general, the poor states in our sample relied more heavily on sales taxes. For example, in FY 2003, Arizona, Mississippi, and New Mexico each received about half of their tax revenues from sales taxes. Of these six states, only West Virginia, which received 34 percent of its tax revenues from this source, relied less on sales taxes than the median state (which received 37 percent of its tax revenues from consumption). In addition, smaller revenue declines occurred in states without many wealthy residents, since much of the revenue decline stemmed from losses in income from capital gains after the stock market bubble of the 1990s burst in late 2001. Because these low fiscal capacity states did not reap large revenue gains from the bubble, they did not greatly suffer from its abrupt termination.
Exhibit IV-3 shows changes in tax revenues in these six poor states, and for all states, for the years 1998 to 2003, standardized to equal 100 during the revenue peak of 2001. Though four of the states experienced declines in tax revenues after 2001, two states continued to see growth. Thus, except for the two western states, which tracked national averages closely, the poor states, when compared to the rest of the country, saw smaller increases in revenues in the late 1990s as well as smaller declines after 2001.
Some short-run factors alleviated state revenue problems in poor as well as wealthy states. Two of the most important were money from the tobacco settlement and state fiscal relief from the federal government in 2003. Nonetheless, social welfare programs in low fiscal capacity states were hit hard in other ways, largely as a result of increasing needs. Food Stamp recipients-one measure of need roughly comparable across states-grew an average of 38.5 percent among the six poor states between FYs 2000 and 2003, while the average growth among all states was 24.9 percent. Low fiscal capacity states also saw no significant decline in TANF caseloads between 2000 and 2003, while higher fiscal capacity states did (Gais, Burke, & Corso, 2003).
Unemployment rates among the six poor states generally did not rise faster than the rates in the U.S. as a whole, but their rates were already higher than the U.S. average, as Exhibit IV-4 demonstrates. Unemployment levels rose in most of the six states between 1999 and 2002. South Carolina, Louisiana, and Mississippi saw the largest increases. The southern states also experienced relatively early increases in unemployment: Mississippi and Louisiana's unemployment rates began to rise in 2000 and South Carolina's as early as 1999.
Unemployment Rates in the Six Poor States and in the U.S. as a Whole, 1995-2003
Other factors compounded the short-run fiscal problems in these states while alleviating them in others. For example, although West Virginia experienced no decline in revenues, neither had it participated in the boom of the 1990s. It, thus, had few reserves to draw on in managing its fiscal crisis. In fact, West Virginia's fiscal problems were compounded by unfunded liabilities in the state's workers' compensation and retirement programs-spending that was enforced by the courts. South Carolina also saw slow growth in revenues for some years, compounded by its large increases in unemployment; indeed, budget shortfalls and cuts have plagued the state since 2000.
New Mexico was exceptional in suffering little from the recession. The state was able to rely on revenues from minerals, oil, and gas as well as many services. Because it had tried to keep 5 percent of recurring appropriations in reserves, the general fund entered the recession with an operating reserve and had no deficit, even in fiscal year (SFY) 2004.
Despite this variation, these six states have generally faced substantial increases in needs since the beginning of the decade, while their tax revenues grew slowly or even declined. All of the states were, thus, under some fiscal pressure, and how they handled that pressure revealed much about influences on state choices in social welfare programs.
The site visits largely indicated that the spending trends and state differences found before 2000 continued even through the recession. Perhaps the clearest example was the continued growth of Medicaid in these states. In most of the six poor states, the growth in spending on Medicaid before 2000-indicated by increases in the Census Bureau's Vendor Payments for Medical Care category-continued through 2003, despite budget pressures in the states.
Exhibit IV-5 illustrates changes in per capita spending on Medicaid, including federal as well as state expenditures, from federal FYs 1997 to 2003. These data come not from the Census Bureau but from Medicaid administrative data on program spending. The data series include the six case study states plus the mean of all states in the U.S. as well as the mean of the six poor states.
Per Capita Spending on Medicaid in Six Poor States, 1997-2003 (federal and state)
As the graph shows, spending on Medicaid among the six poor states typically grew faster than the average per capita spending among all states in the U.S. Although not shown in the graph, the average per capita spending among the six poor states was slightly below the national state average from 1997 through 2000. By 2001, however, the poor state per capita average was higher than the national average and remained so through 2003.
Four of the states (i.e., South Carolina, Arizona, New Mexico, and West Virginia) showed continued or even accelerated growth in nominal spending after 2000. Growth was especially strong in New Mexico and Arizona. Only Louisiana, which had spent the most per capita on its Medicaid program in 2002, substantially reduced its spending in 2003, while Mississippi showed little change.
Spending growth in Medicaid in recent years was especially high among the two western states, Arizona and New Mexico. Although Arizona's per capita spending remained well below that of other states throughout this period, New Mexico's per capita spending on Medicaid exceeded that of the other five states and was well above the national average by 2003. The large regional differences among these states in their per capita expenditures on Medicaid thus declined in recent years.
The reasons for this continued growth in spending and growing convergence among the poor states were many, but some factors were widespread. The economic cycle and the growing costs of pharmaceutical drugs and home/personal care services were widely cited in site visits as forcing up spending. Another major force was the expansion of Medicaid enrollments in recent years as a byproduct of child health insurance program (CHIP) outreach activities in the late 1990s and early 2000s. In one state, for example, administrators thought that two Medicaid participants were found and added for every new CHIP enrollee.
Respondents also indicated that Medicaid programs in their states were limited to begin with. Large parts of their budgets were driven by mandatory services rather than optional services or optional groups. Major cutbacks in eligibility were, thus, often impossible without federal cutbacks.
Yet some of the growth resulted from major program expansions since 2000. The most striking change occurred in Arizona. State voters overrode years of legislative opposition to Medicaid expansion in 2000 by enacting a citizen's initiative, Proposition 204 that greatly expanded Medicaid coverage to include households up to 100 percent of the federal poverty level. Enrollment grew from 575,000 in January 2001 to 902,500 in January 2003. The state also took advantage of the Health Insurance Flexibility and Accountability (HIFA) initiative, which allowed eligible working parents of children enrolled in the SCHIP program to qualify for coverage. Louisiana expanded eligibility, as well, for children, aged, and disabled in FY 2003, and it increased eligibility among pregnant women to include all those under 200 percent of the federal poverty level.
Some of the growth in Medicaid spending seemed attributable to a lack of fiscal pressures in some states. New Mexico, as noted already, was little affected by the recession and was able to cover its substantial increases in Medicaid costs with its large operating reserves. Officials attributed some of the program growth in recent years to delayed efforts to institute cost-saving measures, such as preferred drug lists.
When cuts were made, they were generally reductions in services or reimbursements, not new restrictions on eligibility. In Louisiana, Medicaid prescriptions were limited to eight per month; upfront payments to hospitals for outpatient services were reduced; and subsidies for the state's charity hospitals were cut. South Carolina lowered ceilings on the number of office visits, prescription drugs, and home health visits, and it reduced reimbursement rates for physicians, though not for hospitals and nursing homes. In general, however, these cuts were marginal and were not expected to greatly slow Medicaid spending.(30)
Spending on cash assistance has declined sharply in most states over the last several years, especially since the implementation of welfare reform (Boyd et al., 2003). As noted above, the declines in cash assistance spending through 2000 were steep for wealthy states but smaller for poor states.
Although the categories we combined and labeled Cash Assistance under the Census Bureau definitions might include state SSI and federal SSI that passes through state accounts, state and local general assistance programs, and some other programs, for these six poor states, most of the spending in the cash assistance category seems to be TANF benefits.(31) Before 2000, per capita spending on cash assistance dropped in all of these six states. Some of the declines, especially among the southern states, began long before state welfare reforms were implemented, with the western states showing more volatility.
The graph at the top of Exhibit IV-6 shows changes in per capita TANF benefits from 1997 through 2002. Among the six poor states, cash assistance spending typically changed little, continuing the dominant trend among states in the lowest fiscal capacity quartile before 2000. New Mexico was an exception among the poor states: its cash assistance spending continued to fall. But declines after 2000 were not found among the other low fiscal capacity states. Louisiana, Mississippi, and South Carolina saw little change through 2002. Some of the poor states saw increases in cash assistance spending: our site visits indicated that Louisiana's spending on TANF benefits increased dramatically between FY 2002 and 2003, and Arizona's spending on assistance increased, as well. West Virginia was more volatile, dropping steeply through 1999, rising in 2000, and then falling and rising again. Thus, through 2002, most of the poor states either stayed close to the already low levels of spending on cash assistance they had reached by 2000, or their spending began to creep up. The regional differences between states in their cash assistance spending, evident in the 1977-2000 Census data, continued through the early 2000s, as South Carolina, Louisiana, and Mississippi's TANF per capita expenditures remained lower than the other states through 2002.
Changes in TANF Spending and Caseloads, 1997-2002
TANF Per Capita Expenditures
Number of TANF Cases Per Capita
The flat spending levels for TANF benefits among the poor states contrasted with changes in other, mostly wealthier, states, where drops in spending on assistance continued past 2000. In the nation as a whole, the average state's TANF per capita spending declined between 2000 and 2002, as shown in the graph at the top of Exhibit IV-6, freeing some money for services and other programs. The trend in Census data toward convergence in per capita spending on cash assistance between rich and poor states before 2000 thus also appeared in the expenditure data for TANF cash assistance, even after 2000.
Changes in TANF spending roughly reflected changes in TANF caseloads among the poor states, which are displayed in graph at the bottom of Exhibit IV-6 as the number of TANF cases per capita. After the large and widespread caseload declines through 1999, caseloads in most of the six states began to hold steady or climb. New Mexico and, to a lesser extent, Louisiana were exceptions; their caseloads fell through 2002. The other states either showed little change or drifted upwards. West Virginia's caseloads rose sharply from 2000 through 2002, while Arizona, Mississippi, and South Carolina's cases grew more slowly and later.
One reason spending on assistance tracked caseload changes was that, with few exceptions, these six states made few major changes in their cash assistance benefits, eligibility criteria, or other policies since 2000. Exhibit IV-7 shows AFDC/TANF benefits levels (based on the amount a three-person family would receive if it had no other income), adjusted for inflation, between 1980 and 2001. These six states rarely changed their benefit levels, producing a long-run downward trend in benefit levels due to inflation. Exceptions included Mississippi, which used its TANF surplus to raise its maximum benefits for a three-person family in the late 1990s, the state's first increase in benefits in more than two decades. West Virginia increased its benefits in 2000 to pay recipients the equivalent of a minimum wage for their hours in the state's work experience program (required under the rules of the Fair Labor Standards Act). And New Mexico raised and then lowered its maximum benefits between 1998 and 2000. As these exceptions indicate, most of the change among these states was upward, not downward, with respect to cash assistance benefits.
AFDC/TANF Maximum Benefit Levels in Six Poor States, Adjusted for Inflation, 1980-2000
Only West Virginia made substantial cuts in its assistance program through other policy changes: it reduced its earned income disregard and child support pass-through in 2002.(32) But the remaining states made few changes in other TANF policies, including their income disregards, time limits on assistance, and sanctions for noncompliance with work activities. The interviews with state officials revealed little interest in making these rules less generous during a recession.(33)
Because Medicaid spending was not cut substantially in these states and, in fact, continued to grow in most states and because spending on cash assistance programs remained essentially flat, offering few opportunities for savings, fiscal pressures during state budget crises fell mostly on non-health social services programs. This outcome was especially true because other budget pressures to maintain or increase spending in non-social welfare were strong in these states. The greatest external pressures came from education programs, which were the top priorities of elected officials in South Carolina and Louisiana and were high priorities elsewhere. However, states had some flexibility in dealing with these pressures due to two factors: the large "surpluses" in TANF spending that states had accumulated since 1997 and the wide array of programs that might be funded under TANF non-assistance.
One way in which states responded to fiscal pressures was to spend down their TANF surpluses. Most poor states were fiscally conservative in 1997 and 1998 in using their federal TANF dollars and typically spent less than three-fourths of their grants on current program needs. Their total spending grew rapidly, but the surpluses they built gave them a cushion they drew on to support their basic assistance programs in later years.
Exhibit IV-8 tracks these changes by displaying the percentage of federal TANF grants spent by states from 1998 through 2002, averaged for each quartile of state fiscal capacity. State spending of federal dollars in the poorest quartile as a percentage of states' annual grants was low in 1998, only 57 percent. All but the wealthiest quartile of states increased their spending of federal TANF funds in 1999, at least partly in response to the flexibility offered by the federal government's "non-assistance rule" (see below; also see Plein (2001)). Spending continued to grow after 1999 among states in the poorest quartile until average spending of federal TANF funds exceeded 111 percent of states' annual TANF grants in 2002. Between 2001 and 2002, when the recession began to hit most states, TANF spending increased in all quartiles, probably reflecting growing fiscal pressures due to revenue shortfalls and growing needs.
Average of TANF Spending as a Percentage of the State's Annual Grant, by State Fiscal Capacity, 1998-2002 (federal dollars only)
Carry-over funds accumulated for most states, though they were especially large among the poor states, despite the small size of their grants relative to their needy populations (Gais & Weaver, 2002). The federal TANF law required states to use their carry-over funds only for "basic assistance," or benefits to meet a family's "ongoing basic needs," such as food, clothing, and shelter or supportive services such as transportation or child care for families whose heads were unemployed (U.S. House of Representatives 2000, p. 355). Because most low fiscal capacity states had small cash assistance programs-and states tended to refrain from expanding such benefits-such states were more constrained in using their surpluses than were others. Nonetheless, the TANF surpluses relieved fiscal pressures on the states in our sample; they typically used their surpluses to fund their current basic assistance needs, thereby freeing current TANF grant funds for other, non-assistance programs.
A second source of flexibility was the wide range of services and programs that could be funded with federal TANF and required state MOE funds. Especially important was the broad definition of TANF non-assistance, which was clarified by the federal government in its rulemaking in 1999. Non-assistance could include any non-recurrent, short-term benefits, work subsidies, and supportive services to employed families. Exhibit IV-9 shows that per capita spending on non-assistance was generally lower and grew less rapidly among the six poor states when compared to the average for all states in the U.S. These differences in per capita spending on TANF non-assistance between the poorest states and other states seemed to continue the trends we noted in the Census data on state spending on non-health social services during the late 1990s. Nonetheless, TANF non-assistance spending rose in most of these states after 1999, most notably in West Virginia. Thus, even the poor states, with their relatively small TANF grants, had a new source of funding for non-health social services.
Per Capita Spending on TANF Non-assistance, Six Poor States and U.S. State Average, 1997-2002
How did the states use their flexibility under TANF to deal with their fiscal crises? And what happened to their spending on non-health social services? Considerable variation occurred among these six states, but a few generalizations applied to most.
First, we found a growing variety of programs funded with TANF dollars, as administrators and elected officials sought funding for high-priority services. Administrators viewed as critical many of these programs that otherwise faced cuts. South Carolina and West Virginia were using a growing share of their TANF grants to support child welfare programs, including protective services, foster care, emergency shelters, and others that administrators regarded as involving high stakes. TANF thus allowed these states to mitigate cuts in funding for their child welfare programs, despite large reductions in state matching funds under Title IV-E. Still, cuts occurred in South Carolina in staffing and payments for protective services, foster care, and adoption in the late 1990s and early 2000s. These cuts suggest an intriguing puzzle for fiscal federalism theory because major parts of IV-E used the same attractive open-ended match rate as Medicaid. To make room in TANF for these child welfare funding needs, which administrators viewed as involving "life and death" issues, they eliminated a number of employment services that had been funded under TANF.
The greater importance of TANF dollars in funding a wider array of programs also seemed to be pushed by drops in federal aid. For example, the nominal and constant-dollar decline in federal support for the Social Services Block Grant has left some of these states scrambling for dollars for child welfare, child care, and a number of other basic services-a deficiency that TANF non-assistance dollars seemed to address.
Other changes in the composition of TANF non-assistance spending resulted less from stress than surplus. Louisiana's large TANF surplus led the state to set up a nearly universal pre-Kindergarten (pre-K) program with TANF dollars. But even though the TANF surplus declined substantially in recent years-and although the state's transportation, child care, and employment services programs were poorly funded-the popularity of the pre-K program prevented legislators from cutting it or even requiring the program to be supported with general revenue funds. Another state noted how it drew on TANF dollars to launch a school readiness program. Several states also began to use a wider range of programs to meet their MOE requirements. Because of cuts in state programs previously included under MOE, state TANF agencies searched for current public expenditures in other agencies and public institutions that might fit within the fairly broad definition of MOE spending, even if the programs had had little or no connection with TANF in the past. Such processes led, for example, to counting several university scholarship programs as MOE dollars.(34)
We found, thus, an expanding range of programs supported under TANF and its MOE requirements, sometimes because money could be found nowhere else and sometimes because the programs, often educational in nature and serving a larger population, were more politically popular than programs that exclusively served the poor. We are not suggesting that this expanded range of programs represents "supplantation," or improper shifting of program funds from state to federal sources. Yet it does suggest that states have learned to exercise greater flexibility in using the block grant, just as they did after the promulgation of the non-assistance rule in 1999. Gauging the aggregate effects of these program shifts was difficult. To the extent, however, that TANF and MOE funds were redirected to support child welfare, education, and other programs, fewer funds were perhaps available for income and work supports.
Second, we found that state flexibility under TANF allowed policymakers and especially administrators to weigh and adjust programs already funded under the block grant. In general, programs were cut that served goals considered less critical or that were thought to have less immediate impacts. In most states, for example, few child care programs were cut. Child care programs were perceived in most of these states as already under-funded, yet valuable, because the child care subsidies constituted one of the few significant services with immediate and widespread importance for low-income families. These programs were also regarded as serving multiple goals. Transitional child care encouraged people to move off of cash assistance, and child care subsidies for foster parents, many of whom worked, were considered essential to the foster care program. Some discretionary cuts in child care were imposed, though since most of these states did not put large amounts of state money into their child care programs, only the two states that had put larger sums of TANF or state money into their subsidy programs in the 1990s (Arizona and Louisiana) made major reductions in their child care spending. For all these reasons, child care programs outside Arizona and Louisiana did not bear the brunt of cuts even through FY 2004.
Some programs, however, were considered less essential, though states varied in what they thought essential. Pregnancy prevention programs, fatherhood programs, parenting classes (e.g., in Mississippi, involving a charitable choice initiative), and job services contracts were often seen as inconsequential in the short-run to meet basic performance criteria and fared less well in these states. In South Carolina, for example, many job development positions and their functions were eliminated in FY 2004. There were exceptions. A TANF advisory board in West Virginia recommended the elimination of the state's "marriage bonus," a cash grant to people who wed while on assistance, but the state eventually decided to keep the provision. Also, Louisiana saw its TANF priorities as including pregnancy prevention programs. For the most part, however, TANF services were often weighed in terms of their immediate importance. In several states, agency administrators also said they used performance measures to decide which programs to cut and which to sustain-or which contracts to eliminate.
Third, we found that administrative expenses, especially staff, were often severely cut in social service agencies. South Carolina's Department of Social Services saw a 26 percent reduction in staff between FYs 2001 and 2003. Arizona has made major reductions in its human service workforce since the beginning of the decade, despite the state's large increase in Medicaid workloads after the passage of Proposition 204. Louisiana has refrained from cutting staff much in the last year, but it reduced positions in its state administered welfare system substantially in FY 2002. Plans to improve information systems were widely postponed or cancelled altogether in most of the six states. These administrative cutbacks seemed greater among poor states than among others: TANF administrative costs per capita declined 22 percent among these six states between FYs 2000 and 2002, while the average decline in all states was only 5 percent over the same period.
Administrative expenses have been reduced in other funding streams, too, or at least they have failed to keep up with caseloads. Food Stamp administrative expenses were generally split equally between the federal and state governments. But despite substantial increases in the number of households on Food Stamps in recent years, states were sometimes unwilling to pull down additional federal dollars to support Food Stamp Program (FSP) administration. As Exhibit IV-10 indicates, federal dollars for administrative expenses per FSP household recently declined. For example, although South Carolina, Mississippi, and especially Arizona have seen large increases in their Food Stamp caseloads in recent years, they have either refrained from increasing or have cut their state matches and, thus, the amounts they get from the federal government to support the program's administration.
Federal Share of State Administrative Expenses in the Food Stamp Program in the Six Poor States (dollars per FSP household)
5. Influences on Spending and Explanations of State Differences
These six states varied in their spending patterns, both before and after FY 2000, though we found that some regional differences in spending patterns of poor states before 2000 have declined in strength in recent years. For the most part, spending trends evident among the poor states before 2000 continued after that year: medical assistance expanded; cash assistance spending remained low and fairly static; and non-health social services programs grew slowly or, in states where fiscal problems were most acute, not at all.
What accounts for these patterns of change and differences across states? We explored several factors in the site visits. But our major finding was that the state differences and the trends in various program functions stemmed in part from differences in how states made decisions in major program areas. Each of the major functional areas in social services- cash assistance, Medicaid, and non-health social services-were dominated by a distinct mode of decision-making. These modes varied by who participated in or controlled the decisions, the frequency of major choices, and the criteria brought to bear on decisions. These distinct decision-making styles might help account for different decision outcomes and changes over time.
[ Go to Contents ]
The site visits indicated that decisions about Medicaid were more likely than other social program areas to involve active and highly organized constituencies, especially service providers, and to engage the attention of elected officials, including legislators and sometimes governors and their staff or top appointees. Medicaid budgets and policies were thus highly salient to elected officials and often assessed as much for their economic effects on health industries and jobs as for their effects on clients. Our respondents singled out nursing homes and hospitals as having especially powerful lobbies in several of our states. Developmental disability programs were also viewed as having effective advocates among institutions as well as community-based providers. Doctors were typically seen as less effective, and their reimbursements were more likely to be cut or less likely to be increased. Drug companies, dentists, and others also became active from time to time in state legislatures if and when relevant cuts or expanded services were under consideration. Even child advocacy organizations in these states were more interested and active on CHIP and Medicaid issues than in other social programs; indeed, some were service providers themselves because many were directly involved in CHIP outreach.
per 100,000 population, 2001
|Percentage of Nursing Facility Residents
Using Medicaid, 2001
The direct involvement of service providers or industry representatives in the budget process might help account for some of the geographical differences found in state spending on payments to medical vendors. Southern and rural states have more hospitals and nursing homes per capita than other states, and the health industry plays a larger role in these states' economies. As Exhibit IV-10 demonstrates, Mississippi and Louisiana had about twice the hospital beds per unit population than did Arizona and New Mexico, and nursing home industries in southern states depended more on Medicaid patients than in the western states. The nursing home data suggest another important difference between poor southern and border states and others: although hospitals and other health care institutions in many states tend to split between some that rely heavily on Medicaid and many that do not, such divisions are less common among poor rural states (i.e., most institutions get substantial support from Medicaid). Most births, for example, are paid for by Medicaid in these states. In such states, Medicaid is more likely perceived as a universalistic program-one that helps many, if not most, communities and populations.
Political support for Medicaid was more than simply a matter of organized providers. It was also one of the few major programs where legislators knew some of the clients and viewed them sympathetically. Officials regarded many nursing home residents as coming from middle or working class families. Also, one legislative aide noted that experienced legislators in his state knew every one of the families in their districts who relied on the Continuum of Care program under Medicaid that paid for expensive and intensive services to families caring for dependents with multiple and severe problems. The prevalent view that these and other health-related problems were not the clients' fault, the fact that some clients relied so intensively on these programs that they often contacted legislators if problems occurred, and the widespread distribution of such cases throughout the state were all characteristics that enhanced political support in state legislatures, which often dominated budget decisions, especially in the southern states.
Federal requirements and incentives were also critical. Many of these poor states had minimal Medicaid programs, covering relatively few optional services and populations. Thus, increases in federal minimum requirements, which grew dramatically in the early 1990s, strongly affected these states. Also, unquestionably, the poor states are willing to spend money on Medicaid due to the high federal match rate for low fiscal capacity states and the ability of states to use a number of strategies to maximize their draw of federal dollars under Medicaid. The attractive "prices" of Medicaid in these states were indeed important factors in maintaining state support for the program, especially during a recession. Medicaid match rates, which were based on states' per capita personal income, ranged among these states from more than 3:1 for Mississippi to 2:1 for Arizona, and they were even higher for CHIP. Health care administrators and other advocates used these rates widely as arguments in budget battles that any cut in state spending produces a much larger cut in total spending.
But though the fiscal formula might prove a necessary condition for Medicaid's fiscal robustness, it seems to be an insufficient condition. The same match was available for child welfare programs, and the results were much less expansive. As a matter of fact, child welfare programs have had such problems obtaining state matching funds in recent years, some state administrators have used their discretion over the TANF block grant to support such programs, even though the fiscal price of TANF dollars was much higher. Attractive matching formulae would seem to exert a contingent effect on spending: they can constitute powerful political arguments in budget battles when the arguments are backed up by strong, organized constituencies.
That states, and especially poor states, support Medicaid strongly because they can use certain strategies to maximize their draw of federal resources might also be argued. Again, the site visits revealed that poor states employ these strategies, and they surely exert some impact on support for the program. The two most common methods of maximizing federal dollars among these states relied on Medicaid expenditures for disproportionate share hospital (DSH) programs and upper payment limit (UPL) programs (for details on these expenditure programs and their use in maximization strategies, see Coughlin and Zuckerman, 2002). For example, Louisiana has relied a great deal on DSH programs to increase its federal payments under Medicaid, while South Carolina has relied more on UPL. DSH and UPL programs have allowed states to make large Medicaid payments to health providers, payments which the state has in turn used to claim federal matching dollars. Sometimes these state payments to providers under DSH or UPL are made in response to large payments by providers to the state (through, for example, intergovernmental transfers or donations). In such instances, the health care providers are usually reimbursed for their donations or intergovernmental transfers by the state, while the state gets, in effect, a higher federal match rate as it uses its transactions with health care providers to pull down federal dollars.
Yet such strategies do not account for much of the Medicaid program in these states. Nor do they provide much help in explaining recent program expansions.
As Exhibit IV-11 demonstrates, DSH payments constitute a small part of the Medicaid budgets in most of these states. Only Louisiana and South Carolina rely more than the national average on DSH payments as a percentage of their total Medicaid expenditures, and some of the states receive little DSH money. Also, these payments have declined over the years as the federal government imposed limits on their use in the middle 1990s, so such strategies cannot account for the growth in spending among poor states on Medicaid in the late 1990s and early 2000s. Again, Medicaid maximization strategies might well contribute to the growth of the program, but by themselves they cannot explain its dynamics and widespread support. Such strategies are probably more powerful when they are backed up by more concrete motivations, such as important constituencies. For example, the extensive use of DSH in Louisiana might reflect formalized concerns about supporting parts of the state's health industry: Louisiana has a statutory provision that requires it to keep rural hospitals viable.
DSH Payments as a Percentage of Total Medicaid Spending, Six Poor States and Average for All States, 1993-2003
Other factors might have increased support for Medicaid. In most of the states, the agencies that administered Medicaid were separated from those that administered welfare and non-health social services, and that separation insulated health programs from the often negative views about social service agencies held among legislators. As Exhibit IV-12 shows, the most consistent divide in agency responsibilities in these six states was between health programs and all non-health social services and income support programs. Only in West Virginia did the same agency that handled "welfare" also manage Medicaid.
|TANF||Child Care||Child Welfare||Medicaid||SCHIP|
|West Virginia||Department of Health and Human Services||Department of Administration|
|Arizona||Department of Economic Security||Arizona Health Care Cost Containment System|
|Louisiana||DSS and Governor's Office||Department of Social Services||Department of Health and Hospitals|
|Mississippi||Department of Human Services||Office of the Governor||State and Public School Employees Health Ins Mgt Brd|
|New Mexico||Human Services Department||Children, Youth, and Families Department||Human Services Department|
|South Carolina||Department of Social Services||Department of Health and Human Services||Department of Social Services||Department of Health and Human Services|
Perhaps reflecting while also reinforcing the strong political support for health programs, political and administrative leadership were frequently more stable among health agencies and legislative committees dealing with health issues, while other human service agencies often saw rapid turnover. Sometimes, as in South Carolina, the decades-long involvement of a single senator was seen by respondents as a critical factor in ensuring stable support for Medicaid. But though that situation might not be the case in every state, such stability in leadership was much less common in other social program areas.
[ Go to Contents ]
We found a different type of decision-making in TANF cash assistance programs. Medicaid was a high visibility program that sometimes commanded the attention of governors and often drew the attention of legislators and organized lobbyists, but TANF cash assistance programs drew attention only sporadically from legislators and from few others outside the state bureaucracies. Although many changes were made in Medicaid rates, service coverage, eligibility levels, and other elements in annual state budget processes, policies governing cash assistance changed much less often.
As already noted, most of these states had low maximum benefit levels, especially in the southern states. Other rules governing cash assistance also tended to be stricter among these states. Time limits were more likely to be shorter than required by federal law. Sanctions were usually stricter, applying to the entire household. Earnings disregards were in all but one case time dependent (i.e., they became less generous or disappeared altogether after 4 to 12 months of working while on assistance.) That provision is atypical among the non-poor states, most of which had earnings disregards that remained the same indefinitely.
These benefit levels and other rules were changed infrequently. Until 2000, Mississippi had not changed its nominal benefits for more than 2 decades. Nor had most of these states revisited many of their newer, welfare-reform rules-such as sanction policies or earnings disregards-that they had established when they enacted their AFDC waivers or TANF programs in the middle 1990s. West Virginia was, as noted, one exception. It did reduce its earnings disregard in the face of budget pressures but took that action only after a special advisory board was established to recommend changes. In the normal course of budget politics, these rules were not reconsidered. In fact, there was considerable reluctance to revisit them.
The rules, benefit levels, and eligibility standards that applied to cash assistance were usually established by legislation, and in these states, some deference was shown to the legislature on such matters. One state, where the agency administering TANF benefits believed it had the authority to change the maximum benefits, refused to do so on its own even though agency officials believed an increase was overdue. State decision-making about cash assistance was thus dominated by the legislative process, and changes occurred occasionally rather than regularly as a part of the budget process.
The lack of change in cash assistance policies in some of the case study states appeared to reflect a reluctance to engage in a policy discussion that was often ideologically and racially divisive. The intertwined issue of race and welfare was brought up during a number of the site visits, although usually indirectly. For example, officials spoke of the poor reputation of the state departments of social services (DSS)-"welfare agencies"-in the state legislatures, and how some of this poor reputation was attributable to negative views of their clients. In some instances, the political problems faced by cash assistance programs appeared to be reinforced by restrictive policies, which limited the number of participants in the program to the neediest families in the state. According to a DSS report from a state with low maximum benefits, only 2 percent of families in the state received cash assistance, and 86 percent of these recipients were African American.
As noted in our literature review, several studies of social welfare spending found that welfare spending was influenced by the racial composition of the state population or the state welfare caseload.(35) Consistent with this literature, among the six states studied, the lowest levels of spending on cash assistance, the lowest benefit levels, and the strictest regulations were found among three states with the highest proportions of assistance caseloads composed of African Americans. This relationship held more generally with poor states, not just those in our sample. As Exhibit IV-13 demonstrates, the relationship between the proportion of African Americans in a state and the estimated state effects for cash assistance was negative for states in the lowest quartile for fiscal capacity. The proportion of African-Americans in a state may be a proxy for other demographic and economic variables, and it is interesting to note that his relationship failed to appear among wealthier states (those in the first three quartiles of fiscal capacity). There also was little relationship between the proportion of African Americans in a state and estimated state effects for either Medicaid or non-health social services. This supports our belief that there are distinct decision-making modes for each of the major functional areas in social services - cash assistance, Medicaid, and non-health social services.
Relationship Between State Effects for Cash Assistance and Average Percentage of African American Population, Lowest Quartile for State Fiscal Capacity
[ Go to Contents ]
A third mode of decision-making was prominent when we examined choices about non-health social services in these states, including TANF non-assistance, child care subsidies, and child welfare. Although decisions about Medicaid funding were strongly affected by provider interests and state funding needs and decisions about the critical parameters in cash assistance were influenced by political ideologies, choices regarding non-health social services more likely reflected a sort of technocratic politics. Decisions were made on the basis of assessments by state administrators about the relative merits of different programs and their centrality in achieving major goals.
However, these decisions were constrained by the availability of resources, and in recent years, those resources have been limited. Administrators were given flexibility in deciding what services to support, but the flexibility often looked like a delegation by the legislature, and sometimes the governor, of difficult choices between competing needs. That is, in recent years, many of the social service agencies among these six states were instructed by their legislatures not precisely where to reduce services but how much they had to cut from their total budgets. There were exceptions: sometimes governors or legislators championed particular services or programs, and these programs served as additional constraints. In these six states, such programs were usually some relatively popular educational initiative, such as Louisiana's pre-K program.
This manner of decision-making in part resulted from the low political salience of these non-health social service programs. Although many of the services, such as child care or child welfare, involved private providers, the providers were rarely organized and active in advocating these programs in state legislatures. Nor were child and other advocacy organizations as involved with these programs as they were in Medicaid or CHIP.
This political flexibility was augmented in recent years by the availability of the TANF block grant and fairly loose constraints as to what service and benefit programs might be included under TANF non-assistance and the state's MOE. In fact, sometimes this flexibility was as much a "burden" as a "blessing," as one administrator told us. Because the state legislature viewed the social services agency as having so much discretion in managing the block grant, agency officials perceived the legislature as unconcerned about imposing major cuts in state funding of agency programs because "they thought we could cover critical shortfalls with our 'slush fund'" (i.e., TANF block grant).
As noted above, within these constraints, administrators often made decisions based on judgments about which services were critical to major agency goals, which programs were successful in terms of performance standards, which programs were needed to create or maintain a coherent package of benefits or services, which programs involved high stakes (e.g., life or death issues), and how and whether programs can be administered with fewer staff. These were unenviable decisions in most of the states because state funds were scarce, TANF surpluses declined rapidly, and some block grants that provided greater per capita support to poor states, such as SSBG, were reduced in size. But, unquestionably, state officials had more control over the mix of services the state could offer than they had just a decade ago.
[ Go to Contents ]
Our analyses of estimated state effects and site visits to poor states suggest several basic conclusions:
These different modes of decision-making might help account for some of the findings from the 50-state models of various forms of social welfare spending, though we cannot say so with certainty. Cash assistance, for example, was found to be strongly related to unemployment, a finding which makes sense in light of our conclusion that policies were infrequently changed in this area, except during major periods of reform, as in the early and middle 1990s, and spending changes were usually driven by caseloads. Medicaid spending was negatively related to state population density, a finding that made sense if health providers in rural states were particularly active politically in supporting Medicaid because they constituted a larger part of the economy or they depended more on public funds to sustain themselves. Our conclusions might also help explain the fact that unemployment failed to increase spending on non-health social services and that federal grants to non-social welfare tended to increase spending on such functions. The site visits suggested that overall state revenue levels strongly affected spending on these non-health social services, which were more discretionary in the short-run than other programs. And it was reasonable to assume that these overall revenues were negatively related to unemployment rates and positively related to sources of federal dollars.
More generally, our findings suggest the importance of state and local constituencies in understanding why some programs thrive, even during difficult budgetary conditions, and why others fail to thrive, even when the different programs offer similar fiscal federalism incentives. In these states, constituencies-mostly service providers-were active and strong on health issues, while they were largely absent from state budgetary politics on other social welfare matters, a fact consistent with spending trends among poor states as well as with state differences. If constituencies are, in fact, critical and if they continue to be skewed toward the health industry, low fiscal capacity states might face a long-term squeeze on non-health social services unless social welfare spending can obtain a larger share of the state budget. Given the strong pressures from other program areas outside social welfare, from education, to prisons, to highways and roads, the latter option is unlikely.(36) Thus, non-health social services might well continue to constitute a smaller and smaller share of state budgets.
[ Go to Contents ]
(30) In 2004, after the case studies were completed, Mississippi enacted program changes that may eliminate eligibility for 65,000 Medicaid recipients, mostly elderly and disabled people. Although most of these recipients were eligible for Medicare benefits, some estimates indicated that about 5,000 were not. However, the strong political reaction to the cuts led the state House and its leadership to seek to rescind the changes soon after passage. As of mid-June 2004, the final resolution was still unclear.
(31) Of these six states, only South Carolina offered more than a negligible SSI state supplementation benefit (U.S. House Ways and Means Committee 2000, pp. 266-267).
(32) "Earned income disregard" refers to the amount of earnings excluded when determining a family's eligibility for cash benefits or the size of the benefit. Under AFDC, states disregarded $120 plus one-third of the remaining earnings for the first four months on a job when calculating benefits. For the next 8 months, the disregard was reduced to $120, and thereafter (after 12 months) to $90. Most states increased their earned income disregards under TANF; most (i.e., 30 out of 51) also eliminated the time-dependency of earnings disregards in calculating benefits. Of the six states in our sample, earnings disregards in calculating benefits are reduced after a person is on assistance for several months in Louisiana, Mississippi, New Mexico, and South Carolina (Administration for Children and Families 2003). "Child support pass-through policies" refers to the amount of child support money a state passes along to a family on welfare. Although the family receives TANF benefits, the state might retain current support and arrearages it collects up to the cumulative amount of TANF benefits that have been paid to the family. Under AFDC, states had to pay, or pass through, the first $50 of collections to the family.
(33) Time limits either restrict the number of months a family may receive assistance before work requirements begin, or they limit the number of months a family may receive assistance regardless of their employment status. Arizona, Mississippi, South Carolina, and West Virginia required family heads to work immediately, and New Mexico required them to work after 3 months on assistance-all lower than the 24 months required under federal law. Arizona, Louisiana, and South Carolina also had special "intermittent" time limits that precluded families' receiving assistance for more than 24 months during some overall time period (i.e., either 60 or 120 months). "Sanctions" refer to the loss of benefits if the family head or heads fail to comply with TANF work requirements. Full family sanctions (i.e., the complete elimination of the benefit) are imposed at the first violation in Mississippi and South Carolina for 1 to 2 months, and partial sanctions lasting at least 3 months are imposed for the first violation in Louisiana and West Virginia. The more typical initial sanctions, sometimes full though usually partial, are imposed until compliance (i.e., there is no minimum period) (Administration for Children and Families 2003).
(34) TANF Maintenance of Effort (TANF-MOE) requirements were, in general, not particularly constraining for these six low fiscal capacity states. TANF-MOE provisions impose financial penalties on states for failing to spend state funds on low-income children and families at a level equal to at least 80 percent of their FY 1994 level (or 75 percent if they meet the minimum work participation rates). However, in low fiscal capacity states, the required TANF-MOE spending levels tend to be low. For example, in the six poor states, the minimal (75 percent) MOE requirements constituted, on average, only 30.5 percent of the states' federal TANF grants (FY 2001). For all states, the same average was 61.4 percent. In wealthy states, such as New York and Massachusetts, the MOE spending requirements were equal to their TANF grants (i.e., these states had to spend state funds at least 100 percent of their TANF grant to satisfy the MOE requirement). Since TANF grants were also typically smaller on a per capita basis in the low fiscal capacity states, TANF-MOE requirements were generally easier to meet in poor states than in wealthy states.
(35) As noted in the literature review in the first section of the report, this literature includes studies by Brown (1995), Plotnick and Winters (1985; 1990), Gais and Weaver (2002), and Kousser (2002).
(36) Although social welfare spending, based on the Census data, increased more rapidly than non-social welfare spending in the late 1980s and early 1990s in the six poor states analyzed here, little change occurred in the ratio during the middle and late 1990s. The average percentage of total state and local expenditures going to social welfare, not including the public hospital category, was about 18 percent from 1977 to 1989. It then rose to 26 percent by 1993, after which it drifted slightly downward to just under 25 percent in 2000. The highest percentages were in the southern and border states, which ranged between 27 percent (West Virginia) to 31 percent (Mississippi) in 2000. Arizona (14 percent) and New Mexico (21 percent) invested much smaller parts of their total budgets to social welfare spending.
Top of Page | Contents
Main Page of Report | Contents of Report
Human Services Policy (HSP)
Assistant Secretary for Planning and Evaluation (ASPE)
U.S. Department of Health and Human Services (HHS)