In addition to reviewing the statistical measure of poverty, the panel was asked to consider issues of benefit levels for government family assistance programs—in particular, a national minimum benefit standard for the Aid to Families with Dependent Children (AFDC) program. Currently, there are large differences in AFDC benefit standards across states, and no state provides benefits as generous as the official poverty thresholds.

Federal policy makers have several times considered enacting a uniform minimum benefit standard that would provide a nationwide floor for AFDC benefits. The congressional debate over the Family Support Act (FSA) of 1988 included proposals for a national minimum benefit, but they were not accepted, largely because of the sizable estimated budgetary costs to the government. The FSA did request a study of minimum benefit standards, however, and this chapter responds to that request. We considered conceptual and statistical issues involved in setting a national minimum benefit standard for AFDC, just as we considered such issues for the poverty line.

In our review, we focused on the nature of the relationship between program benefit levels (whether in AFDC or other cash and near-cash assistance programs) and a measure of poverty (whether ours or another), and we show why that relationship is indirect at best. We also considered the relationship of the proposed poverty measure to AFDC standards of need. AFDC is unique among cash and near-cash assistance programs in that the states are required to establish a standard of need but are not required to—and often do not—use this standard to determine actual benefits. (See Appendix D for details of the AFDC program.)

We recommend (in Chapter 7) that serious consideration be given to the use of the proposed poverty measure as an eligibility standard for programs that tie eligibility for benefits and services to the current poverty measure. It might seem a logical next step to suggest a direct relationship of the proposed poverty measure to program benefits. Certainly, the existence of a poverty threshold that makes reasonable adjustments for differences in family circumstances, including differences in the cost of living across regions of the country, creates an impetus for program benefits to be related to that threshold. However, there are many factors that properly enter into a determination of benefit levels, only one of which is a poverty threshold.

At present, there is wide variation in AFDC benefits across the 50 states and the District of Columbia, and, in most states, benefits are considerably below the official poverty threshold. As of January 1994, the states' median standard of need for a three-person family was 60 percent of the corresponding official poverty threshold, and the median maximum benefit was 38 percent of the poverty threshold. ¹ The median of the maximum combined AFDC and food stamp benefit for the states was 69 percent of the poverty threshold. Looking across states, the maximum AFDC benefit for a three-person family in January 1994 varied from $923 per month in Alaska to $120 in Mississippi, with a median of $366, a mean of $396, and a coefficient of variation of 40 percent; see Table 8-1.² The maximum AFDC benefit ranged from $240 to $552 (25-58% of the poverty threshold) in about two-thirds of the states; eight states exceeded this range, and eight states fell below it.

The maximum combined AFDC and food stamp benefit for a three-person family exhibited somewhat less dispersion, varying from $1,208 in Alaska to $415 in Mississippi, with a median of $658, a mean of $675, and a coefficient of variation of 22 percent. Food stamps have this effect because of the program's benefit formula, which assumes that families will devote 30 percent of their countable income to food expenditures (see Chapter 7). Hence, an increase of $1 in AFDC benefits (or other countable income) decreases food stamp benefits by 30 cents, and a decrease of $1 in AFDC benefits (or other countable income) increases food stamp benefits by 30 cents. The maximum combined AFDC and food stamp benefit ranged from $528 to $822 (55-86% of the poverty threshold) in 39 states. Adjusting AFDC and food stamp benefit levels to take account of differences in the cost of living by state further reduces the variation, although only to a limited extent (see below).

The original Aid to Dependent Children program, the predecessor to AFDC, was enacted in 1935 as part of the legislation that instituted a national Social Security system.³ It was designed to put on a sounder footing the states' programs to provide ''mothers' pensions," but there was no intent to mandate a prominent role for the federal government. ⁴ The legislation provided that the federal government would pay 33 percent of the program's costs, with a

maximum federal payment of $6 a month for the first child and $4 for other children. In 1950 the program was amended to provide benefits to the mother herself (or another caretaker of a dependent child or children), and in 1962 the program was amended to provide benefits (at state discretion) to two-parent families in which both were unemployed. The original legislation required that the states pay one-third of the costs (i.e., it prohibited states from the common practice of laying off all their costs on local jurisdictions), and it made federal payments "conditional on passage and enforcement of mandatory State laws and on the submission of approved plans assuring minimum standards in investigation, amounts of grants, and administration" (Congressional Record, January 17, 1935:548).

Since one-half of the counties in the United States did not provide mothers' pensions at the time and there was wide variation in payments across counties within states, the legislation had the effect of reducing within-state variation in benefits. However, it had little effect on across state variation, leaving broad discretion to the states to set need standards, payments, and eligibility rules. For example, states were allowed to keep their residency requirements, and most did so until the Supreme Court in 1969 ruled them to be unconstitutional.

Historically, reformers have followed three strategies to try to establish more uniform state policies with regard to AFDC benefits (see Peterson and Rom, 1990:99-100), focusing on the matching formula, a supplementary national program, and national minimum benefit standards.

The federal matching percentage was raised from 33 to 50 percent in 1939, to be consistent with the percentage for programs to assist the needy elderly, blind, and disabled. The formula was changed several times more between 1944 and 1958. Finally, in 1965, states were given the option of switching to the matching formula adopted for the Medicaid program. This formula committed the federal government to paying at least 50 percent of the welfare benefit in every state and to paying a higher matching rate (up to 83%) in those states with lower per capita income. Currently, all states use the Medicaid matching formula for AFDC benefits. The matching percentages in fiscal 1994 varied from 50-55 percent in 19 states and the District of Columbia to 70-79 percent in 13 states (U.S. House of Representatives, 1994: Table 10-17).

The rationale for the changes in the matching formula included the desire to provide incentives for low-benefit states to raise their benefits. However, Peterson and Rom (1990) found that the differences in benefit levels across states remained essentially unchanged, with a coefficient of variation that ranged from 34 to 37 percent in each decade from the 1940s to the 1980s.

Food assistance programs in the United States were initially very

localized. Many communities did not participate in the food stamp (or commodity distribution) program, and eligibility standards varied widely among those that did. In 1970 the Food Stamp Program was effectively nationalized: a single national standard was adopted, which was higher than any then in use in the states; the Secretary of Agriculture was empowered to set national eligibility requirements; and stiff penalties could be imposed on states that did not operate a program in every county. In 1977 Congress eliminated all purchase requirements for food stamps, making them a simple supplement to cash assistance in inverse proportion to family income (as well as a benefit to working families not receiving cash assistance).

The effect of these changes was to reduce the variation across states in the combined value of AFDC and food stamp benefits. However, there was no incentive for low-benefit states to raise AFDC benefits per se; rather the provision that food stamp benefits increase (decrease) by 30 cents for every dollar decrease (increase) in cash benefits in effect rewarded states that kept cash benefits low and penalized states that increased them.

The strategy of legislating a uniform minimum benefit standard for AFDC has never achieved legislative success. In an early discussion of needed reforms in public assistance programs, Leon Keyserling's Conference on Economic Progress (1959:58) urged that "minimum uniform standards among the States should be set by the Federal Government." In 1965, the Office of Economic Opportunity (OEO) proposed a negative income tax program with a single nationwide payment schedule as part of its first "national anti-poverty plan." As part of its second plan in 1966, OEO again proposed a negative income tax program with a single nationwide payment schedule; besides being available to all poor persons without regard to demographic category, this proposed program would have gradually replaced existing public assistance programs (including AFDC) by 1972. In that same year the Advisory Council on Public Welfare (1966:15,22,117) recommended a "minimum standard for public assistance payments below which no State may fall." It proposed (p. xii) that the ''Federal Government … set nationwide standards, adjusted by objective criteria to varying costs and conditions among the States, and assume the total cost of their implementation above a stipulated State share."

In 1967 President Johnson proposed that states be required to pay 100 percent of their standard of need, but he did not propose any specific minimum benefit standard. The proposal was rejected in the House Ways and Means Committee. The President's Commission on Income Maintenance (1969:7) recommended a "universal income supplement program financed and administered by the Federal Government." Concerning benefit levels for this income supplement program, the Commission stated (p. 59) that "attempts to reflect different costs of living in different areas would involve many difficulties and so a uniform National supplement is recommended."

The Family Assistance Plan (FAP) put forward by the Nixon Administration in the late 1960s provided for two kinds of programs, each with national minimum benefit standards: a program for low-income elderly, blind, and disabled (which subsequently became the Supplemental Security Income [SSI] program), and a program for all families with dependent children, regardless of work status. The proposed FAP AFDC minimum benefit was $1,600 a year for a family of four (about 40% of the official poverty line at that time)—a level that would have raised benefits in 16 states. FAP passed the House in 1970 but died in the Senate: conservatives questioned the adequacy of the work incentives; liberals criticized the national minimum benefit as inadequate.

The Carter Administration's Better Jobs and Income Program, proposed as legislation in 1977, also included a national minimum benefit for a program that would have combined AFDC, SSI, and food stamps; the minimum was set at $4,200 for a family of four whose head could not be expected to work (about 70% of the poverty line). This proposal died in Congress. In 1979 a scaled-back plan was introduced that proposed a national minimum benefit for AFDC at about 75 percent of the poverty line. The House passed this plan with the minimum benefit lowered to 65 percent of the poverty line (which would have raised benefits in 13 states). This bill died in the Senate.

As noted above, proposals for a national minimum benefit were originally considered for the 1988 FSA. In 1987 the House Ways and Means Committee approved a minimum benefit standard, but opposition from southern Democrats on the grounds of increased costs to their states resulted in stripping this provision from the legislation. The FSA instead mandated a study of minimum benefit standards.

Today, the de facto national minimum level of available benefits for AFDC recipients is the maximum food stamp allowance combined with the maximum AFDC benefit in the lowest benefit state. In January 1994, this amount was $415 per month for a three-person family, or 43 percent of the corresponding poverty threshold. Hence, the issue of a national minimum benefit standard for AFDC really comes down to an issue of raising this de facto standard. Arguments for adopting such a nationwide minimum benefit standard for AFDC have been made on the basis of equity: that low-income families with children should not be disadvantaged simply by reason of their state of residence. In addition, others have argued that differences in benefits encourage low-income families to migrate from low-benefit to high-benefit states. There have been studies of the migration effects of AFDC, but they suffer from serious data and methodological problems. The results suggest that there is an effect on the migration behavior of low-income families but that the effect, for a number of reasons, is quite weak (see below).

We considered the issue of a national minimum benefit standard in somewhat broader terms, asking the question of how or if the proposed poverty measure could or should be linked with benefit levels for a program such as AFDC or a combination of AFDC and other cash and near-cash assistance programs. We first broached this issue in Chapter 7 , in which we discussed the possible use of the proposed poverty measure for programs that already relate eligibility to the current measure. We pointed out some of the reasons that program agencies might want to make the link less direct, for example, by setting eligibility cutoffs at a fraction of the poverty thresholds. Here we explore more fully the reasons that a program benefit standard could differ from a poverty standard and, more generally, why the design of an assistance program could deviate from the goal of helping everyone who is classified as poor.

A note on terminology: When we speak of a "benefit standard" in the context of AFDC, we mean what is referred to in that program as the "maximum benefit" in contrast to either the ''need standard" or the "payment standard." A family must have gross income below 185 percent of the need standard to be eligible for AFDC; it must also have net or countable income below 100 percent of the payment standard. A number of states have a payment standard below their need standard, and some states cap the maximum benefit at a lower level than the payment standard (see below).

The measurement of poverty or need does not necessarily imply anything about the extent to which need can or should be alleviated through government assistance programs. There are five key issues that separate measurement of need and alleviation of need: budget constraints, both overall and from competing demands on funding resources; strategies and preferences for targeting program benefits; interactions among programs; behavioral responses to program incentives; and, finally, cost-sharing provisions for federal-state programs.

Scarce budget resources may well limit the extent to which benefit standards can approach the poverty threshold, particularly in entitlement programs, such as AFDC, that must provide benefits to all eligible applicants. Both globally and in the United States, the areas with the greatest poverty are typically the areas that can least afford high benefits. For example, in some African countries, such a high proportion of the population is poor (by any standard) that very few resources are available internally to alleviate poverty.

For AFDC, the states with low-benefit standards tend to be the states with higher poverty rates and with lower per capita incomes and, hence, with less ability to provide assistance to their needy families. Thus, maximum benefits in January 1990 were negatively correlated with the 1989 state poverty rate

(correlation coefficient, –.55) and positively correlated with the 1989 state per capita income (correlation coefficient, .67).⁵ However, there is considerable variation in benefit levels among the states that is not explained by differences in income. Peterson and Rom (1989) and Plotnick and Winters (1985) show that differences in AFDC benefits across states relate to a variety of political, ethnic, and economic differences.

For the nation as a whole, it would be hard to argue that the United States lacks sufficient revenue-generating ability to provide assistance to families below the poverty level. But the country's funding resources are not unlimited, and there are many demands on them. Assistance programs must compete with all other uses of taxpayers' funds.

In order to maximize the effectiveness of limited funds and achieve other policy goals, there may be reasons to target assistance payments to particular groups, even though simple measurement of need would not necessarily identify them as unique. For example, because of the long-term social cost of children growing up in economic deprivation, it may be sensible to concentrate assistance dollars on poor families with children, even though other groups have need that is just as great.

There are many examples of targeting in current programs. The Earned Income Tax Credit (EITC) was originally targeted to working poor families with children and was recently expanded to cover childless workers as well (see Appendix D). Food stamps offers another example of targeting, in that the program is designed to provide a more secure safety net for the elderly and disabled than for other people. This feature operates through the definition of countable income, which permits more generous deductions for households with elderly and disabled members in determining eligibility and benefits. Also, there is a higher asset limit for households with an elderly member (see Appendix D).

Another approach would be to concentrate scarce assistance dollars on the poorest families (the "worst off" among the poor), even though helping the families closest to the poverty line (the "best off" among the poor) would achieve the fastest reduction in measured need. In other words, although the strategy of helping the poorest poor will not produce as large a reduction in measured need per dollar spent as helping other poor people, it may be the best strategy to reduce poverty.

The existence of multiple assistance programs can affect the level of the benefit standard that makes sense for any one of them. For example, AFDC interacts with food stamps and public housing, and it makes little sense to think of an AFDC benefit standard in isolation from these programs (or in isolation from such programs as the EITC and enforcement of child support). However, given the different ways in which eligibility and benefits are calculated, it is not easy to determine an appropriate adjustment to AFDC benefit levels to take account of program interactions.

In the case of AFDC and food stamps, for example, one could certainly argue for excluding food costs from the AFDC benefit standard because of the almost universal provision of food stamps (and school meals) to AFDC families. As noted above, it would also be to a state's financial benefit to reduce its AFDC benefit standard by as much as the value of the Thrifty Food Plan because the Food Stamp Program will provide higher benefits than otherwise would have been the case. However, only in the case of states with very low AFDC benefit standards will the Food Stamp Program in fact make up the entire difference for recipients. This occurs because the program assumes that, after deductions, 30 percent of countable income including AFDC benefits is available for food consumption and, hence, reduces food stamp benefits accordingly. As a hypothetical example, consider a state that wants to provide combined AFDC and food stamp benefits at the level of the official poverty threshold. The deductions in the Food Stamp Program make it difficult to calculate by how much the state should reduce its AFDC benefit standard, but it can be demonstrated that not to reduce the AFDC standard at all may overcompensate recipients by as much as 10 percent relative to the poverty threshold, while to reduce the AFDC standard by the full amount of the Thrifty Food Plan may undercompensate recipients by as much as 17 percent.⁶

Program interactions virtually dictate that designers of assistance programs use complicated models to evaluate likely program effects. Some models are designed to point out odd interactions of such program features as maximum benefit levels and tax rates on other income by estimating the benefit package

that would accrue to a specific type of family at a particular income level. Other models use microsimulation techniques operating on large-scale household databases to project the effects on program costs and caseloads of specified program features, given the distribution of the population and estimates of the likelihood of participation and other behavioral effects (see Citro and Hanushek, 1991; Lewis and Michel, 1990).

Human beings participate in programs, and programs undeniably affect their behavior. Some effects are intended, others are unintended; some effects are positive, others are negative.

Some programs have an explicit goal of providing a positive incentive: for example, the federal government subsidizes student loans to encourage more young people to obtain the economic and other benefits of a college education. As another example, the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) seeks out poor pregnant women, mothers, and children to provide food supplements with the goal of healthier pregnancies, healthier babies, and, ultimately, healthier children and adults.

Other programs have a primary goal of providing income support to needy people. Such cash and near-cash assistance programs as AFDC and food stamps must contend with the fact that economic support has negative incentive effects to the extent that recipients are encouraged to rely on the program and not take steps to become self-supporting. Research on AFDC has examined incentive effects in the areas of work effort, family structure, and migration.

Both economic theory and empirical research indicate that such programs as AFDC adversely affect the work choices of the eligible population. These programs provide a "guaranteed" base income to those who do not work; the resulting "income effect" allows individuals to work less. These programs also impose taxes on earned income. Since workers' net wages are now lower, the "substitution effect" encourages them to decrease the number of hours worked as it is relatively less expensive for them to do so. The combination of these provisions results in an unambiguous decrease in the aggregate number of hours of market work by the eligible low-income population.⁷

Extensive research has been undertaken to estimate the magnitude of the

reduction in work resulting from the disincentives embedded in assistance programs (see Danziger, Haveman, and Plotnick, 1981, and Moffitt, 1992, for extensive literature reviews on the subject). Although evidence shows that AFDC reduces the number of hours worked by single mothers, the estimates of those reductions vary among studies—from 1 to 10 hours per week. Moffitt (1992), in his review of the literature, concludes that "there is still considerable uncertainty regarding the magnitude of the effects."

Moffitt (1992) points out that there is very little research on the effects of in-kind assistance on labor supply.⁸ He also notes the importance of exploring the effects of multiple assistance programs; however, these effects are difficult to model, and little work has been done in the area.

Much of the literature on family structure focuses on whether AFDC encourages the formation of single-parent families headed by women. Since benefits are targeted to mothers with children and no spouse present, they may provide incentives to delay marriage or remarriage, to obtain a divorce, or to have children outside of marriage. Early work looking at the effect of AFDC on the increase in female-headed families is extremely mixed (see the summary in Groeneveld, Hannan, and Tuma, 1983). Studies in the 1980s, however, show more consistent evidence of an effect (see Danziger et al., 1982; Ellwood and Bane, 1985; Hoffman and Duncan, 1988; Moffitt, 1992). There is also some evidence of an effect of AFDC benefit levels on the probability that a female head lives independently rather than in a larger family (see Ellwood and Bane, 1985; Hutchens, Jakubson, and Schwartz, 1989).

Extensive research has been done on the effect of AFDC on illegitimacy. The work has studied whether the existence of public assistance increases the chances that babies will be born to unmarried women since a woman no longer needs a husband to help support a child. The work has also considered whether the existence of public assistance increases the likelihood that a woman will have a child in order to become eligible for benefits at all or have another child in order to receive additional benefits. The evidence on this issue in the literature is inconclusive: some studies find effects for some groups (e.g., white or black teenagers), and others find no effects for the same groups (see Duncan and Hoffman, 1990; Ellwood and Bane, 1985; An, Haveman, and Wolfe, 1991; Lundburg and Plotnick, 1990; Plotnick, 1990).

The extent to which the wide variation in AFDC benefit levels across states influences patterns of interstate migration is of

particular relevance to the issue of a nationwide minimum benefit standard. Hence, we considered it in some detail.

Substantial cross-state differentials in AFDC benefits have existed since the inception of the AFDC program, but they have created greater policy concern since residency requirements were ruled unconstitutional in 1969. In particular, policy makers (particularly in high-benefit states) have worried that their states attract welfare recipients, thereby increasing the burden on taxpayers. A simple comparison of the expected income available to AFDC-eligible families in highand low-benefit states clearly indicates that such families can receive more income in a high-benefit state, which should create an incentive for them to relocate. Since the same states have generally remained high- or low-benefit states, if such migration occurs, it should have been steadily occurring in about the same regional patterns throughout the past 25 years. There are, however, at least three reasons why such an effect might be small or not exist at all.

First, moving costs money. Not only are there actual transportation costs associated with moving, but families that migrate will often have to pay a security deposit for a new apartment, experience some transitional time during which they are neither working nor on AFDC, and bear the myriad of costs associated with relocation to a new city and residence. Low-income families may be least able to bear these moving costs.

Second, families—and particularly low-income women—may lack information about their income opportunities in distant state locations. States do not generally advertise their AFDC benefit levels, and unless women have other sources of information (such as friends or relatives in another location), they may have only a hazy idea about alternative benefit levels.

Third, relocation decisions are affected by many things other than income expectations. In particular, especially for low-income women with children, there may be substantial nonmonetary costs to moving. The presence of family and friends in their current location may provide many benefits: friends and family can provide free baby-sitting services, can be a source of shared resources in hard economic times, and can be an important source of psychological support. In addition, women with children might be quite risk-averse about relocating their children to an unknown low-income neighborhood, with concerns about school, crime, and gangs. For many women, these nonmonetary costs might be large enough that they completely swamp any differences in expected income levels.

These arguments indicate that the expected effects of AFDC benefit levels on migration behavior among low-income women with children are probably small, at least in part because this is a population that one would expect to be less mobile than many others. On the margin, however, one may still expect that AFDC benefits would have a positive effect on migration probabilities.

In order to measure the size of any welfare-induced migration, one ideally

would have longitudinal data that track family location decisions. The data would also contain information on women's expectations and their economic opportunities in alternative locations, including not only what they know about alternative AFDC benefit levels in different locations, but also what they know about wage and employment opportunities. The need to have control variables available on non-AFDC economic opportunities is particularly important, since state AFDC benefit levels are positively correlated with state income and wage levels. (This is not surprising, given that only high-income states can afford to pay high AFDC benefits and that only states with high wage levels can pay high AFDC benefits without creating large work disincentives.) Finally, the data would contain information on whether women have friends, family, or any other source of support or contacts within alternative state locations. (For example, knowing if a woman or her parents have ever lived in another state would be one way of controlling for the nonmonetary costs of choosing a different location.)

Unfortunately, a national data set with such information does not exist. The empirical research has been based on much more limited data, and, as a result, the quality of most the analyses is suspect (see Moffitt, 1992, for a review of the literature). Despite the problems, however, two conclusions are warranted: migration rates among AFDC recipients are quite low, and there is a small positive effect of AFDC benefits on the probability of migrating to (or not migrating out of) a high-benefit state (see, e.g., Blank, 1988; Clark, 1988, 1990, 1992; Gramlich and Laren, 1984; however, Peterson and Rom, 1989, find larger effects). The results are convincing not because any one of the studies is very well done, but because studies done in different ways with very different types of errors all point in the same direction.

The research suggests that welfare-induced migration should be a second-order concern for policy makers. For states that have large populations very close to each other, large benefit differences may indeed induce a migration flow. However, on average, the effects of AFDC benefits on migration are small and movement among the AFDC population is infrequent.

The fact that different states have had long-term AFDC benefit differentials that are very large and have been very large for many years is perhaps further evidence that migration effects are hard to discern in the data. Although states may talk about this problem, high-benefit states have not been concerned enough about it (with a few exceptions) to cut their benefits relative to other states.

There is a lot that is not known about migration effects. There is little or no evidence on the propensity to use AFDC by recent migrants in comparison with natives in a state; on the comparative duration of their AFDC spells if they do become recipients; or on the effects of AFDC benefits on inducing families not currently eligible for AFDC to migrate to a state (i.e., whether people think about "potential safety net" issues). In addition, the growing

number of foreign immigrants into the United States may be affected differently by this issue than are native-born U.S. residents.

Overall, it is clear that cash and near-cash assistance programs have negative effects on such behaviors as work effort, although the effects may often be small. These incentive effects can cause a disjuncture between measured need and the amount of assistance program dollars required to alleviate that need. That is, if people who are provided benefits that are intended to move them out of poverty respond in such ways as working less, the net effect will be to leave them in poverty and to require even more funds to move them out of poverty. Because of such behavioral effects, it is misleading to describe the aggregate "poverty gap" (i.e., the difference between the poverty line and a family's resources, aggregated over all families) as the dollar amount that the government would have to spend to eliminate poverty. Because of behavioral responses, an expenditure that should decrease the poverty gap to zero will almost surely fall short. Moreover, the incentive problem is even broader in that program benefits (whether from public assistance programs or social insurance programs, such as unemployment compensation) may lead to reduced work by beneficiaries, even though the poverty gap or other poverty statistic is not affected.

Yet a decision not to provide any type of assistance because there may be some undesired behavioral responses on the part of some program participants is extreme. Such programs as AFDC have their success stories as well as their problems, and, as noted above, behavioral effects, when present, are often small. More fundamentally, there are groups in need—such as children—who are not responsible for their situations but who will suffer if benefits are curtailed.

A program designer, then, faces the very difficult task of specifying benefit levels and other program provisions so as to assist needy people but in a manner that does not encourage behaviors that increase program costs or may otherwise be viewed as dysfunctional and, if possible, that encourages functional behaviors. The task is made more difficult by the fact that research findings on incentive effects are often incomplete or inconclusive.⁹ There is also the problem that other aspects of the environment may undercut efforts to provide incentives for increased self-support: for example, there may not be jobs available in the private sector for welfare recipients.

Issues of program incentives have been at the center of policy debate for AFDC, which is directed to families whom the public would like to see increasingly responsible for their own support. Consequently, there has been considerable experimentation with changes in benefit levels and formulas for

calculating disposable income to try to develop effective incentives for recipients to become self-supporting through work and to encourage family stability and better parenting. To date, results show limited effects on such behaviors as work effort from changes in benefit levels and the tax rate on earnings. The evidence is not yet in on more recent state initiatives, such as not increasing benefits when another child is born or reducing benefits if parents do not stay in school or fail to have their children vaccinated. It is important, moreover, to note that other programs besides AFDC raise concerns about incentives. For example, assistance programs for retired or disabled people, such as Social Security and SSI, have negative effects on work effort (see Hurd, 1990; Quinn, Burkhauser, and Myers, 1990; Rust and Phelan, 1993; Wise, 1992).

In the United States, federal-state cost-sharing provisions have important effects—not always intended—on program benefit levels and the possibilities for changing those levels. For AFDC, the federal government historically has tried to provide incentives to low-income, low-benefit states to raise benefits by picking up a higher share of assistance program costs in these states. However, there has been little effect on states' behavior: low-benefit states have generally opted to minimize their own budget outlays rather than to raise benefits, and, hence, the variation in benefits across states has remained high (see Peterson and Rom, 1990). Similarly, states have taken advantage of the fact that the Food Stamp Program, for which benefits are funded entirely by the federal government, will partly make up for lower AFDC benefit standards.

The current situation in which low-income, low-benefit states receive higher rates of federal reimbursement makes it difficult to devise a politically palatable scheme for raising AFDC benefits to some national minimum standard. A review of one such proposal by the Congressional Budget Office (CBO) (1989a), the Partnership Act of 1987 (introduced in the 100th Congress as S. 862 but never enacted), starkly illuminates the problems.

The Partnership Act proposed to expand the federal role in financing AFDC and Medicaid and to pay for these expansions by eliminating a number of grant-in-aid programs (e.g., Community Services Block Grants and Urban Mass Transit Research). The act provided for a national minimum AFDC benefit standard that, when combined with food stamps, would ultimately reach 90 percent of the federal poverty line for families with no other income. At the same time, the federal matching rate for AFDC benefits up to the minimum standard would be raised to 90 percent.

The evaluation of the federal and state costs of this proposal found that it was not cost-neutral overall as it was intended to be. Rather, if the program had been fully phased in by 1994, CBO (1989a) estimated net costs to the federal government of $38 billion and net savings to the states of $22 billion,

for a net increase in federal-state government expenditures of $16 billion. More important, CBO found that the effects would be very uneven across states: states in the South would actually incur net costs rather than savings, and savings would be highest for wealthier states. Low-income, low-benefit states would have to pay more than better off, more generous states to bring their benefits up to the federal minimum AFDC standard; moreover, those states, with their already higher matching percentages, would gain less from the increased federal matching rate than better off states. Hence, there would be little incentive for low-income, low-benefit states to support this type of proposal.

Of course, there are other ways in which to combine a national minimum benefit standard for AFDC with a provision for federal-state cost sharing. However, the current structure of the program makes it difficult to devise a scheme that does not increase overall program costs or that does not disadvantage some states relative to others.

This brief review of some key factors that enter into the design of assistance programs—funding constraints, considerations of the target population, program interactions, incentive effects, and federal-state cost-sharing provisions—makes it clear why it is difficult to link poverty thresholds directly to benefits. To those who are involved in evaluating and designing government assistance programs, our observations will come as no surprise and indeed may seem obvious. Yet we believe it is worth underscoring the point that measuring need, by determining how many people have resources below a reasonable poverty standard, is different from determining the proper societal response to that need.

Many factors properly enter into a determination of program benefit standards, including judgements about the extent to which society is prepared to allocate scarce resources to supporting low-income people and the mix of goals that society wants government assistance programs to serve. The critical role of such judgements is the reason that a panel such as ours, chosen for expertise in measurement issues, cannot make recommendations about appropriate benefit levels for specific assistance programs. Ultimately, the determination of appropriate programs and policies to alleviate poverty involves political choices—namely, the consideration of competing public objectives against the constraints of scarce public resources within the framework of a nation's social and political climate and belief system.

However, the fact that we do not make a recommendation about national minimum benefit standards for AFDC (or other programs) should not be taken to mean that there is no argument for such a benefit standard. On one hand, it is clear that the states differ in their preferences for spending on public

assistance, and these preferences should be given weight in any national policy making. On the other hand, there are equity problems in providing needy families with very different levels of public support on the basis of where they happen to live when their economic problems arise. From this perspective, while the proposed poverty measure (or any standard of need) cannot be used by itself to determine benefit levels, it does have a role to play in the policy debate.

In most government assistance programs, the benefit standard (i.e., the maximum amount of benefits provided to people with no other income) and the need standard are one and the same: people who are eligible because their countable income falls below the benefit standard are in turn entitled to receive benefits up to the amount of the standard. ¹⁰ As noted above, the standard for a particular program reflects judgements about a variety of factors, including appropriate levels of need, constraints on available funds, and the desire to provide positive incentives to recipients.

AFDC is unique in that federal legislation requires each state to establish a standard of need for families with children who have no other means of support, and, in a separate process, to determine a payment standard, which may be lower than the need standard.¹¹ Both the need standard and the payment standard restrict eligibility for benefits (see below). Furthermore, states may set a maximum benefit amount that is below both the standard of need and the payment standard.

One might surmise that the need standard, as distinct from the payment standard or maximum benefit, is supposed to represent a type of poverty concept. In this case, one might want to consider the use of the proposed poverty measure as the standard.¹² The use of the proposed measure would reduce the current wide disparity in need standards among the states, while recognizing geographic cost-of-living differences. However, it is not clear that the states have typically interpreted the need standard as a poverty concept. Indeed, the role of the need standard in the AFDC program seems

murky at best, given that states can, and often do, set benefit standards that fall below their need standards.

Since the provision for separately determined need standards exists in the AFDC program, however, we believe it useful to consider the issues involved in the possible use of the proposed poverty measure by the states for this purpose. We begin by describing the basic regulatory framework within which AFDC has operated. We then describe methods of setting need standards that were used in the 1970s and 1980s, current differences in standards and equivalence scales among the states and their relationship to the current poverty line, and trends in need standards and maximum benefits over time. Finally, we discuss the potential relevance of the proposed poverty measure to AFDC need standards. We conclude by encouraging the states to give serious consideration to linking their AFDC need standard to the proposed poverty measure. On balance, that measure would be advantageous for this purpose, although it may need to be modified in some respects.

AFDC is a state-administered program with funding provided by both the states and the federal government through a matching provision (see Appendix D). In order to qualify for federal funding, a state must establish a standard of need that defines in monetary amounts the basic needs the state wishes to recognize as appropriate for an assistance standard of living—although neither the components of the standard nor the methods for setting the standard are prescribed by federal law or regulation. The state must apply this standard uniformly and statewide in determining financial eligibility for assistance, but it may vary the standard to account for family size or composition, area cost-of-living differentials, or other factors.

States may adopt lower payment standards and maximum benefit amounts than their need standards by such methods as paying a percentage of the difference between the family's income and the need standard, paying a percentage of the need standard, or capping benefits at a specified amount. Recently, a number of states have altered their benefit provisions to satisfy budget constraints and to try to induce recipients to adopt preferred behaviors. As examples, some states no longer provide an additional benefit for an additional child or they condition benefit amounts on such actions as the recipient's obtaining immunization shots for his or her children.¹³

Over the years, amendments to the law, court decisions, and federal regulations have formally reaffirmed the states' autonomy in deciding AFDC benefit levels. In particular, the 1967 amendments to the Social Security Act affirmed the right of states to set benefit maximums and to apply ''ratable reductions" in order to set benefits lower than the standard of need. The 1967 amendments included a provision to require states to update their need standard to reflect cost-of-living increases since the standard was adopted; however, states were not required to pay benefits consistent with these increases (for an account of the results of this provision, see Rabin, 1970).

Although the states have very wide latitude in setting their need standard and benefit levels, federal regulations have always been more specific about the resource side of the ledger for determining AFDC eligibility and benefits (see U.S. House of Representatives, 1994:327-331; Solomon and Neisner, 1993). Currently, to receive AFDC payments, a family must pass two income tests. First, the family's gross income cannot be higher than 185 percent of the state's need standard and the family's net or countable income must not exceed 100 percent of the need standard or payment standard, whichever is lower.¹⁴

In 1980, Urban Systems Research & Engineering, Inc. (USR&E) completed a study for the Social Security Administration of AFDC standard setting practices, which included a survey of all 50 states and the District of Columbia and case studies of 11 states. USR&E was critical of state practices with regard to standard setting. In part, this criticism stemmed from the viewpoint expressed in the USR&E study that a standard must be "normative" or "absolute," in the sense that an expert standard of need should be developed for each budget component—independent of expenditure patterns—and then priced out. But as we discuss throughout this report, there are other types of poverty or need standards that merit serious consideration, with advantages and disadvantages. However, USR&E seems justifiably to have concluded that relatively few states in the 1970s were following good standard setting practices, in the sense that they developed their need standard as the result of a well-documented, carefully worked-out process or periodically reviewed their standard to determine whether it should be updated or redefined.

USR&E classified the methods originally used by the states to derive their need standard; see Table 8-2.

Market Basket Pricing Studies "The market basket approach, which involves the specification and pricing of every component of need, is the traditional method for conceiving and measuring absolute need, and historically it

has been the most popular basis for AFDC need standards" (Urban Systems Research & Engineering, 1980:8); 21 states reported using this approach. However, only three states had standards that were based on pricing studies conducted in the last 10 years (i.e., in 1969-1979), and only one state had updated its standard regularly on the basis of repeated pricing studies to account for cost-of-living increases. USR&E criticized (perhaps too harshly) the practice in the more recent market basket studies of using expenditure surveys to determine the shelter component of the need standard rather than developing a normative standard for shelter and then pricing it out.

Expenditure Surveys Six states reported basing their standard on expenditure surveys that were limited to AFDC recipients. USR&E properly criticized this approach as tautological, in that the population for determining the "standard" was based on current program participants.

Lower Level Budget Five states reported adapting the lower level family budget of the Bureau of Labor Statistics (BLS) as the basis for their need standards, and all of these states had done so as of 1969 or later. USR&E noted correctly that the BLS budgets represented a combination of normative standards and actual expenditure patterns. The states using the BLS lower level budget generally deleted categories they deemed "inappropriate," either on judgemental grounds (e.g., alcoholic beverages) or on grounds that other programs covered the expenditure (e.g., medical care). However, only two of the five states had regularly updated their need standard.

Other One state used a multiplier approach similar to the Orshansky method for deriving the poverty line; six states used a combination of methods; and twelve states used completely arbitrary methods or methods that could not be ascertained in the USR&E survey.

The Congressional Research Service regularly tracks changes in the level of the states' need standards and benefit levels (see, e.g., Solomon, 1991; Solomon and Neisner, 1993), but little information was obtained about standard setting practices in the 1980s until recently. In 1992, the Center on Budget and Policy Priorities completed a study for the Administration on Children and Families of AFDC standard setting practices in the late 1980s. This report (Larin and Porter, 1992) was prepared to fulfill the requirement in the 1988 Family Support Act that the states evaluate their AFDC need standard at least once every 3 years and report the results to the Secretary of the Department of Health and Human Services (HHS).

In early 1991 HHS sent the states a questionnaire asking for information on how each state's need standard in effect as of October 1, 1990, was developed, the relationship between the state's need standard and benefit

levels, and any changes in the need standard over the preceding 3 years. The Center on Budget and Policy Priorities analyzed the questionnaire responses (California, Minnesota, and Texas did not respond).

Larin and Porter (1992:5) conclude, as did USR&E in its earlier study, that "the majority of states cannot demonstrate that their need standards represent an amount of money necessary to purchase basic necessities." Larin and Porter document and evaluate six types of methods for setting AFDC need standards by the states in effect as of 1990 (see Table 8-2):

Federal Poverty Guidelines Fourteen states reported relating their need standard in some way to the HHS poverty guidelines.¹⁵ Of these, four states reported using the HHS poverty guidelines as is. Ten states modified the guidelines in such ways as subtracting the cost of the Thrifty Food Plan, subtracting average food stamp and Medicaid benefits, subtracting the cost of "nonessential commodities," setting their need standard as a percentage of the guidelines, or allowing their need standard to decline as a percentage of the guidelines because of not adjusting for inflation.

BLS Lower Level Budget or Living Standard Six states reported using the BLS lower level budget or living standard—last published in 1982 and developed with expenditure data from the early 1960s—as the basis for their need standards. (Another state was considering the use of a modified lower level budget for its need standard, and the welfare department in another state develops a modified lower level budget as guidance for the state legislature.) Two of the six states modified the BLS standard (e.g., by omitting men's haircuts, household supplies, and occupational costs, as well as making changes to other components of the lower level budget). These states have priced the various budget components by using BLS data or conducting local price surveys; however, none of them has adjusted the standard to keep pace with inflation.

Local Market Basket Surveys Fourteen states reported basing their need standard on local market basket surveys, but many of these states have not conducted such a survey recently.

Expenditure Ratio (or Multiplier Method) One state reported using Consumer Expenditure Survey (CEX) data to determine a ratio of all expenditures, other than housing, to apply to the cost of the Thrifty Food Plan. Housing standards were calculated separately on the basis of the actual housing costs of AFDC recipients, with three different standards used for different regions of the state. Another state reported a similar type of method, but developed its multiplier on the basis of CEX data for the lowest quintile of the household income distribution.

Legislative Determination Three states reported that need standards are set by their legislatures on the basis of budgetary considerations.

Average Payment Three states reported that they developed need standards in the early 1970s that represented average AFDC payment levels by family size. (AFDC benefits at that time were determined on a discretionary basis by caseworkers according to the particular circumstances of each recipient family.)

Unknown Methods Six states "are unable to document how their need standards were originally constructed, either because records are incomplete or lost or because their standards seem to have been set arbitrarily with no reference to living costs" (Larin and Porter, 1992:17).

Comparing Larin and Porter (1992) with USR&E (1980), one finds that many states reported using a different method in 1990 than in 1980; see Table 8-2. Only 10 states appear to have used the same method in both decades: Maine, New York, and Wisconsin consistently reported using a variation of the BLS lower level budget, and Colorado, Idaho, Massachusetts, Nebraska, Oklahoma, Oregon, and Washington consistently reported using the market basket pricing method.¹⁶ Perhaps the most important change is that 14 states now relate their need standard explicitly to the HHS poverty guidelines.

As noted above, AFDC need standards vary widely among the states (see Table 8-1). In January 1994, the need standards for the 50 states and the District of Columbia varied from $1,648 per month in New Hampshire to $320 in Indiana, with a median value of $574, a mean value of $655, and a coefficient of variation of 41 percent. The maximum AFDC benefit exhibited almost as much dispersion, although the addition of food stamps reduced the dispersion somewhat.

In a historical analysis of AFDC benefits, Peterson and Rom (1990: Table 1-1) found that a high degree of variation in benefit levels has always characterized the states. They determined that the coefficient of variation ranged

between 32 and 35 percent for the average monthly AFDC payment for 1940-1990 and between 34 and 37 percent for the maximum benefit for a four-person family for 1960-1990. The coefficient of variation was smaller for the maximum combined AFDC and food stamp benefit, ranging between 16 and 21 percent for a four-person family for 1970-1990.

In looking at the relationship of the maximum AFDC benefit to the need standard in January 1994 (see Table 8-1), 11 states paid a maximum benefit that represented 100 percent of their need standard; 23 states paid between 50 and 99 percent of their need standard (the median state paid 66 percent of its need standard); and the remaining 17 states paid less than 50 percent of their need standard, including 6 states that paid less than 33 percent of their need standard.

In looking at the adequacy of AFDC need standards and benefits against the official poverty threshold, 8 states had need standards in January 1994 that were at or above the 1993 official average weighted poverty threshold for a family of three, and 12 states had need standards that were between 70 and 99 percent of the poverty level; see Table 8-3. The remainder had need standards that were below 70 percent of the poverty level. In no state did the maximum AFDC benefit exceed the poverty level, and in only two states did the maximum benefit exceed 70 percent of the poverty level. With the addition of food stamps, the maximum combined benefit exceeded the poverty level in 2 states and was between 70 and 99 percent of the poverty level in 22 states.

In looking at the disparities in AFDC need standards and benefit levels, one obvious question is whether they are related to differences in needs and costs of living across states. We constructed an index of the adjustments by state to a national poverty threshold that would result from taking account of differences in the cost of housing. We analyzed 1990 census data to determine cost-of-housing index values by state (relative to a national value of 1.00) and then adjusted each index value downwards by a factor reflecting the proportion that shelter costs (including utilities) represent of the proposed poverty thresholds. (The methodology was the same that we used to determine adjusted cost-of-housing index values by region and size of metropolitan area for the statistical poverty measure; see Chapter 3). We also constructed an index of state median family income from 1990 census tabulations (Bureau of the Census, no date(b)) relative to 1.00 for the average median family income across the states. Not surprisingly, the index of state-adjusted poverty thresholds shows less variation than does the index of median family income; see Table 8-4. The state-adjusted poverty threshold index values range from 24 percent above to 15 percent below the national average, with a coefficient of variation of 10 percent.¹⁷ The median family income index values range from

43 percent above to 29 percent below the national average with a coefficient of variation of 17 percent.

We then divided each state's AFDC need standard, maximum benefit, and combined maximum AFDC and food stamp benefit as of January 1994 by the appropriate state-adjusted poverty threshold index value and the appropriate median family income index value.¹⁸ If differences in the cost of living

across states (as proxied by cost-of-housing differences in the poverty threshold) are the only reason for the differences in need standards and benefit levels, then the calculation with state-adjusted poverty threshold index values should result in the same (or close to the same) dollar amounts of the need standard and maximum benefit in all states. In other words, the amounts in high-cost, high-benefit states would decrease to the mean and the amounts in low-cost, low-benefit states would increase to the mean. The same reasoning applies to the calculation with state median family income index values.

These patterns do not occur. There is only a modest effect on the variation across states in AFDC need standards when differences in the cost of living or median family income are taken out of the dollar amounts: the coefficient of variation is reduced from 41 percent to 37 percent; see Table 8-5. For maximum AFDC benefits and maximum combined AFDC and food stamp benefits, there is a somewhat greater reduction in the variation across states: the coefficient of variation for maximum AFDC benefits is reduced from 40 percent to 29-33 percent, and the coefficient of variation for maximum combined AFDC and food stamp benefits is reduced from 22 percent to 15-16 percent. However, even in the case of maximum combined AFDC and food stamp benefits, significant variation remains that cannot be explained by differences in cost of living or income levels across the states.

Equivalence scales—the proportion by which benefits to the AFDC unit are increased for each added child—also vary across states; see Table 8-6.¹⁹ Data are available on the maximum AFDC benefit by family size as of January 1994 for the 50 states and the District of Columbia, ranging from the basic two-person unit (parent or other caretaker and child) through the six-person unit

(basic unit plus four added children). At one extreme, Louisiana increases its $138 benefit for the basic two-person unit by 32 percent on average ($44) for each additional child. At the other extreme, Alaska increases its much higher benefit of $821 for the basic unit by only 12 percent ($102) for each additional child. The median value that is added on average to the basic unit benefit for each added child is 23 percent.²⁰

In looking at the shape of the equivalence scales for AFDC benefits, five states have a regular pattern whereby, within 1 or 2 percentage points, they add the same amount to the basic unit benefit for each additional child; 10 other states have a regular pattern within 6 percentage points. Ten states have a declining pattern, whereby they add progressively less for each child after the second or third. In contrast, 10 states add more for the third and fourth child than for either the second or fifth. Finally, 16 states have erratic patterns. For instance, they may add more for the third and fifth children than for the second and fourth. In this, they resemble the equivalence scale implicit in the current U.S. poverty measure, in which the second child adds 17 percent to the two-person (one-adult/one-child) poverty threshold, the third child adds 31 percent, the fourth child adds 23 percent, and the fifth child adds 20 percent.²¹

The type of equivalence scale that we recommend for the poverty measure would increase the benefit for a one-adult/one-child family the most for the second child, with declining percentages for each additional child to reflect household economies of scale. Depending on the value of the scale economy factor, our proposed equivalence scale would add an average of 27 percent (using a factor of 0.75) or an average of 22 percent (using a factor of 0.65) to the basic unit benefit for each additional child.

Looking at trends over the last two decades, it appears that relatively few states have increased their need standard or maximum benefit to keep up with inflation. Relatively few states have statutes that require them to adjust their standards for inflation, and even those states that have such requirements do not always heed them in periods of budget stringency. As of 1988, seven states had statutory requirements for adjusting their need standard to keep up with inflation, one state had a requirement to update its benefit level, and three

states had requirements to update both their need standard and their benefit levels; only one state apparently had a requirement to periodically reevaluate the need standard itself (Center on Social Welfare Policy and Law, 1988).

In the 1970s, inflation rose significantly—by 111 percent (as measured by the CPI-U). Yet only four states increased the value of their need standard in real dollar terms during this period, and the median state saw its need standard decline by 38 percent (in real terms); see Table 8-7. The decline in real terms in the value of the maximum benefit in the median state was somewhat less—26 percent; see Table 8-8. As a result, the number of states paying ''full need" doubled over the period (from 14 to 29 states), and the number paying 70 percent or more of need increased from 33 to 44 states.

In the 1980s, inflation moderated—increasing by 63 percent—and 26 states increased the value of their need standard to keep up with or exceed the rate of inflation. In the median state the need standard remained constant in real dollar terms. Many states updated their need standard after the passage of the 1981 Omnibus Budget Reconciliation Act, which included a provision that families could not be eligible for AFDC benefits if their gross income exceeded 150 percent (later raised to 185 percent) of the state need standard. In this way, states were able to avoid denying eligibility to families with earnings who would otherwise have been above the gross income cutoff although below the net income cutoff after allowable deductions. However, states did not necessarily update their benefit level to match: for example, Alabama doubled its need standard between 1980 and 1985 but did not raise its maximum benefit (see U.S. House of Representatives, 1991:601-605). Indeed, in the median state the maximum benefit declined by 22 percent. From 1980 to 1991 the number of states paying full need dropped from 29 to 16, and the number paying 70 percent or more of need dropped from 44 to 27 states.

In looking at the whole time span, the need standard declined in real terms in the median state by 33 percent from 1970 to 1991, and the maximum benefit declined in real terms by 38 percent.²² One consequence of declining AFDC benefits over the period was that food stamps (which are indexed yearly for inflation) accounted for an increasing proportion of the combined AFDC and food stamps benefit. This change reduced the financial burden on the states, since the federal government pays the full cost of food stamps.

Clearly, a persistent characteristic of the AFDC program has been the great variation in need standards (and benefit levels) among the states—variation

that considerably exceeds estimated differences in the cost of living across states. Another characteristic has been the absence in many states of systematic procedures for setting and periodically revising the AFDC need standard.

A fundamental question is whether the concept of a separate need standard makes sense: most assistance programs do not distinguish between a need standard and the maximum benefit the program will pay to participants with no other source of support. Limits on gross as well as countable income in these programs (e.g., SSI) are set as a function of the benefit standard, and such a practice could be followed in AFDC as well.

Urban Systems Research & Engineering (1980:22) argued that the AFDC need standard serves the useful function of a goal or benchmark and that need standards are not an exercise in futility: