House Budget Committee

Means-Tested Programs, Work Incentives, and Block Grants

Editor's Note: See the figures associated with this testimony in the attached PDF.

Chairman Ryan, Ranking Member Van Hollen, and Members of the Committee:

Thanks for inviting me to testify today. I consider it a privilege to have the opportunity to talk to members of the House Budget Committee.

In response to instructions from the Committee, I’m going to talk about three topics: trends in spending on means-tested programs; work incentives and phase out rates for means-tested programs; and block grants.

Trends in Means-Tested Spending

The lower line in Figure 1, based on a Brookings analysis of federal budget data published by the Office of Management and Budget, shows federal spending since 1962 in the ten biggest means-tested federal programs. In 2011, we estimate that about 87 percent of the spending was on entitlement programs.[1] Federal spending on poor and low-income Americans has increased enormously. Since 1980, by which time all but two of the ten programs that spent the most money in 2011 were in place, spending has increased by about $500 billion, from $126 billion to $626 billion after adjusting for inflation.

One cause of the increase in spending is that both the population and the number of poor people in the U.S. have increased over time. Thus, even if the federal government spent the same amount of money in 2011 on means-tested programs per person in poverty as we spent in 1962, spending would have increased. The solid line in Figure 1 expresses the increase in federal means-tested spending as spending per person in poverty. Expressed in this way, over the past five decades, federal spending on major means-tested programs has increased from about $516 to a little more than $13,034 per person in poverty. If we use the figure on spending per person in poverty in 1980, when most of the major means-tested programs were in place, the increase is from about $4,300 to $13,000 per person or more than $3 spent in 2011 for every dollar spent in 1980.[2] More recently, means-tested spending increased from about $477 billion to $626 billion in the three years of the Obama administration, an increase of about 31 percent. However, the recession that began in December 2007 and the increase in poverty during and following the recession is an important part of the explanation for increased means-tested spending during the Obama administration. Spending per person in poverty increased by about 9 percent as compared with the 31 percent increase in total spending during the first three years of the Obama administration. A portion of the rise in means-tested spending, which was authorized as part of the American Recovery and Reinvestment Act of 2009, began to expire in 2010.

Figure 2 shows how means-tested spending is distributed among eight broad categories of programs.[3] The categories include health, cash, food, housing, education, social services, energy, and employment and training. The figures are for 2009, the last year for which the Congressional Research Service has calculated means-tested spending within these eight categories. Not surprisingly, the figure shows that health is by far the biggest category of spending at $319 billion in 2009, about 2.5 times as much as cash programs, the second biggest category. Employment and training at $9 billion is the smallest of the eight categories. Figure 2 shows that means-tested spending, like total spending in the federal budget, is driven in large part by the rising cost of health care. In this respect, figuring out ways to control the growth of health care spending would reduce the rate of increase in both total federal spending (and debt) as well as federal means-tested spending.

A few additional points about these figures are in order. First, keep in mind that these spending data are for only the ten largest means-tested programs. The Congressional Research Service estimates that in 2009, spending on these ten programs represented about 75 percent of total federal means-tested spending.[4] If that percentage remained roughly the same for 2011, total federal means-tested spending in that year was closer to $835 billion than the $626 billion spent on the ten biggest programs.

Second, state and local governments spend their own money on many of these programs. The Congressional Research Service has estimated that state and local governments supplemented federal spending on means-tested programs by around 27 percent in 2004.[5] If we assume that the 27 percent has remained roughly constant, we can estimate that total federal, state, and local government spending on means-tested programs was probably around $1,143 billion in 2011.

On a per-person in poverty basis, that figure represents about $23,731 in spending by federal, state, and local governments. But this estimate should be considered in light of several caveats. The first is that not all of the spending on means-tested programs goes directly to individuals and families. Some of the money is spent on programs, such as the $14.5 billion spent on Title I of the No Child Left Behind Act and the $9 billion in spending on employment and training programs, that provide services rather than direct cash or in-kind benefits. Second, some of the money in programs that provide cash or in-kind benefits directly to households goes to individuals and families that are not below the poverty level. Children in families of up to 200 percent of the poverty level, for example, are eligible for Medicaid or the Child Health Insurance Program (CHIP) in almost every state.[6] Similarly, people in households with incomes up to 130 percent of poverty are eligible for SNAP benefits (Supplemental Nutrition Assistance Program, previously food stamps). In the case of the Earned Income Tax Credit (EITC), in 2010 a single mother with two children could receive benefits if the mother’s income was below $40,964, about 225 percent of the poverty level for this family. Professor David Armor of the School of Public Policy at George Mason University is in the process of using Census Bureau data and data from other sources to estimate the percentage of benefits in health, nutrition, housing, and cash means-tested programs that go to individuals or households with income above the poverty line. Although Armor’s work has not yet been published, he is finding substantial fractions of the benefits in all these programs going to individuals and families with income above the poverty level and some of it even going to those with incomes above 200 percent of the poverty level.

Means-tested spending has increased enormously, no matter how it is measured. Although there have been some periods of comparatively rapid growth, such as during the recession of 2007 to 2009, Figure 1 shows that spending has grown almost every year for the last five decades. The increase in spending has been the most rapid in health programs, but cash, nutrition, and several other types of spending have also increased rapidly. Similarly, spending per person in poverty has also increased substantially, although not quite as rapidly as total spending.

Work Incentives and Benefit Phase Outs

The impact of welfare benefits on work incentive has always been a contentious issue. Common sense tells us that if able-bodied people get welfare benefits without doing anything in return, their incentive to work and achieve self-sufficiency will be diminished. This common sense view is also supported by a host of research studies. Reviews of the empirical evidence on this issue have consistently shown that welfare reduces work effort.[7] To reduce such work disincentive, most means-tested programs have phase out rates because program designers want to maintain a financial incentive for benefit recipients to work. The hope is that by reducing welfare benefits by less than a dollar for each dollar of earnings, recipients will have at least some incentive to work or work more. The ideal outcome would be to design benefits so that an extra dollar of earnings would always produce a net income increase that is as close to the amount of earnings as possible. The lower the phase out rate, the greater the increase in net income and therefore work incentive. However, lower phase out rates make means-tested programs more expensive. There is a clear tradeoff between program cost, benefit phase out rates, and work incentive.

The difficult problems posed by phase out rates and work disincentives is greatly complicated by the fact that all families with earnings are subject to taxation of their earnings and some families receive more than one means-tested benefit. Consider some of the possibilities: workers are subject to the roughly 15.3 percent FICA tax[8] from their first dollar of earnings; they could face an EITC phase out of up to 21 percent; families with housing benefits face a marginal tax rate of 30 percent on their earnings; and so forth. Considering all of the effects on net income and work incentives simultaneously strains the ability to understand just how much net income would change at a particular point in a person’s earnings curve. Figure 3 is taken from a 1995 report from the Congressional Research Service. Although the specific phase out rates portrayed in the figure are somewhat out of date, a mere glance at the figure conveys the immense complexity of trying to figure out the net impact of so many different phase out and phase in rates operating simultaneously. The Congressional Budget Office is now completing a similar report on marginal tax rates in the tax and transfer system which goes into great detail in showing the actual marginal tax rates faced by individuals and families with various characteristics. Some of the rates are very high and under some circumstances an extra dollar of earnings can result in net income increases of 50 cents or less.

Problems maintaining work incentives are an inevitable consequence of means-tested programs. It would be possible to reduce, but not eliminate, the work disincentive effect of the current system if all benefits could be combined and then phased out at a single phase out rate. However, there are many problems with creating such a system. For one thing, the current benefits system is a combination of cash (the EITC, the Child Tax Credit, TANF, and Supplemental Security Income) and in-kind benefits (primarily SNAP and other nutrition programs, housing, Medicaid and SCHIP, and home heating). Perhaps the in-kind benefits could be paid as cash, but that would cause problems with various interest groups such as the National Grocers Association that would fight against cashing out SNAP benefits. Democrats might oppose converting benefits to cash because providing a lump sum cash payment would make the high level of benefits paid to some families more transparent than under a system when some of the benefits are paid in kind, thereby raising objections from Republicans who would likely argue that the system is too generous and should be cut. Moreover, the administrative complexity of such a system might make it very difficult to operate. Yet another problem is that an all-cash system could greatly increase the number of means-tested benefits families receive (although they would be combined into one benefit). As surprising as it might seem, under the current system few families actually receive all the means-tested benefits for which they qualify. A recent study sponsored by the Department of Agriculture showed that only 72 percent of people qualified for SNAP benefits actually receive them and that in some states the rate is below 60 percent.[9] Both the Current Population Survey and the Survey of Income and Program Participation show that random samples of Americans receive relatively few of the benefits for which they are qualified.[10]

Given the difficulty of phasing out means-tested benefits and maintaining work incentives, an approach to this issue taken in the TANF program established by the 1996 welfare reform legislation is important to consider. Regardless of benefit phase out rates, a matter that was left up to states by the 1996 law, the federal statute requires state programs to have two features that directly address work incentive. The first is that all state programs are required to have strong work requirements. Specifically, at any given moment 50 percent of TANF recipients must be involved in work activities that are tightly defined in the legislation. States that do not comply are fined. As part of the work requirement, states are required to impose financial sanctions on recipients who do not comply with the work requirement. The combination of work requirements imposed on both states and individuals backed up by financial sanctions serve to motivate states to adopt demanding programs and recipients to prepare for and look for work, usually in the private sector. In addition, the TANF legislation imposes a five-year limit on benefit receipt, sending a strong signal that benefits are not permanent, as they had been under the Aid to Families with Dependent Children program that TANF replaced. With strong work requirements and time limits, the work incentive created by benefit phase out rates is much less important. Soon enough, individuals must work regardless of the financial work incentives.

Despite these strong pro-work features of the TANF program, it would be a serious mistake to think that American social policy depends exclusively on these essentially negative inducements to work. Beginning roughly in the mid-1980s, Congress created or reformed a host of programs that supplemented the income of poor and low-income working families, especially single mothers. These reforms included:

  • Expansion of Medicaid and CHIP benefits so that all children in families under 200 percent of poverty are eligible for coverage in most states
  • Several expansions of funding for child care and reform of child care programs to give states more flexibility in use of child care subsidies to help working families
  • Several reforms of the SNAP program making it easier for working families to receive food subsidies
  • Numerous expansions of the EITC; the maximum EITC benefit in most states is greater than the average value of their TANF benefit
  • Creation of the refundable child tax credit that, like the EITC, provides a cash benefit to low-income families with earnings.

Taken together, these work support benefits constitute the nation’s most successful method of attacking poverty among families with children.[11] The combination of increased work by poor mothers following welfare reform and benefits from the work support system resulted in substantial declines in poverty among children in female-headed families. Even today, after two recessions, the poverty rate among children living in female-headed families is lower than it was before welfare reform and the work rate among single mothers is still higher than before welfare reform.[12]

The TANF experience demonstrates that using phase out rates to increase work incentive can be trumped by strong work requirements and a comprehensive work support system.

Block Grants

Block grants provide states with a sum of money to accomplish broad policy purposes which are specified in the authorizing language. Block grants can be constructed so that they achieve a major goal of state policy and a major goal of federal policy.[13] States are always pleased to accept federal dollars, of course, but they also want flexibility with how the dollars are to be spent. Thus, states are doubly pleased if the block grant specifies the broad purposes of the federal grant and leaves it to state government to decide how best to achieve those purposes. From the federal perspective a major potential advantage of block grants is that spending can be controlled. In the case of open-ended entitlement programs such as Medicaid and SNAP, everyone who meets program qualifications has a legal right to receive the benefit. By contrast, in programs with capped spending such as housing programs and the major child care programs, local authorities or states receive a fixed amount of money and individuals are not entitled to receive the benefits. Most block grants, including the TANF program, the Child Care and Development Fund, and the Social Service Block Grant, have fixed funding. In all three of these cases, federal spending has increased slowly if at all in recent years and then only when Congress explicitly authorized and appropriated the additional funds. Given the enormous and growing deficit that afflicts the federal government, the possibility of spending control in major areas of social policy through the use of block grants should not go unnoticed.

The history of federal block grants shows that it is rare for the federal government to provide states with funds to achieve broad social goals without some strings attached. In the case of TANF, for example, the block grant came accompanied by substantial requirements for data reporting, work requirements that states had to follow, and many other strings. These requirements were negotiated with states in marathon sessions that resulted in requirements that states felt they could live with. If Congress is to create additional block grants, it would be advisable to negotiate the terms of the block grant with states. In the case of TANF, Congress worked with the National Governors Association, the National Conference of State Legislators, and the American Public Human Services Association to find mutually acceptable provisions on work requirements, data reporting, and other details.

The general issue of block grant requirements is especially important because of the need for accountability in spending federal funds. Under the Single Audit Act, all federal grants of over $100,000 given to states must be audited under widely accepted audit standards. But accountability for spending goes far beyond ensuring that funds are spent on activities for which they are intended. Rather, recent years have seen increased emphasis on showing whether federal funds are spent on state programs that actually achieve their purposes. Especially in education programs and welfare programs designed to encourage work, high quality program evaluations, usually involving random assignment designs, are the order of the day. Both the Bush and Obama administrations placed great emphasis on the importance of evidence-based policy.[14] Evidence-based policy is especially important today because the nation’s major social intervention programs in preschool, the public schools, delinquency, employment and training, and many other areas usually do not have significant impacts on the social problems they were designed to address.[15]

To continue this growing federal practice of insisting on program accountability, block grants should include, in addition to financial accountability, two types of mandatory reporting. First, all programs receiving block grant funds should be required to report a standard set of data on program participants such as number and characteristics of people served, type of treatment, length of treatment, and, where possible, evidence of program success. Second, the secretary of the federal agency administering the program should be provided with funds to conduct high-quality evaluations of selected programs to determine if particular approaches or program models, as well as the specific characteristics of program models, are effective in producing the desired program outcomes.

[1] Housing, Title I grants to local education agencies, and half of Pell Grants were counted as non-entitlement spending. Thus, $545.0 of total spending of $626.2 or 87 percent was entitlement spending in 2011.

[2] The number of people in poverty in 2011 will not be published by the Census Bureau until next fall. Based on a model designed by Richard Bavier, a former senior official at OMB, that has successfully predicted the poverty level for the past several years, poverty increased by .6 percentage points in 2011, bringing the poverty rate to 15.7 percent in that year.  According to the American Community Survey, the population of the United States in 2011 was 306 million. Thus, around 48,042,000 people were poor in 2011. This is the figure used to compute means-tested spending per person in poverty for 2011.

[3] Karen Spar, “Federal Benefits and Services for People with Low Income: Programs, Policy, and Spending, FY2008-FY2009” (R41625), Washington, DC: Congressional Research Service, January 2011, p. 9.

[4] Karen Spar, “Federal Benefits and Services for People with Low Income: Programs, Policy, and Spending, FY2008-FY2009” (R41625), Washington, DC: Congressional Research Service, January 2011, p. 12.

[5] Domestic Social Policy Division, “Cash and noncash Benefits for Persons with Limited Income: Eligibility Rules, Recipient and Expenditure Data, FY2002-FY2004” (RL33340), Washington, DC: Congressional Research Service, March 2006.

[6] According to the Kaiser Family Foundation, only North Dakota, Oklahoma, Arkansas, and Idaho set the child eligibility level for Medicaid and CHIP below 200 percent of poverty; half the states are at 250 percent or above. See

[7] Robert Moffitt, “Incentive Effects of the U.S. Welfare System: A Review,” Journal of Economic Literature, 1992, 30(1): 1-61; Sheldon Danziger, Robert Haveman, and Roert Plotnick, “How Income Transfers Affect Work, Savings, and the Income Distribution: A Critical Review,” Journal of Economic Literature, 1981, 97(3): 975-1028.

[8] The FICA tax for Social Security and Medicare is split between employer and employee payments totaling 15.3 percent of earnings (up to a limit of about $106,800 in the case of the Social Security Tax). However, economists assume that if the portion of the tax paid by employers were not required by federal law, employers would increase wages by an amount equal to the tax.

[9] Karen E. Cunnyngham, “Teaching Those in Need: State Supplemental Nutrition Assistance Program Participation Rates in 2009,” Washington, DC: U.S. Department of Agriculture, December 2011.

[10] Janet Currie, “The Takeup Rate of Social Benefits,” Los Angeles, CA: UCLA, 2004.

[11] A study in the 2008 edition of the Ways and Means Committee’s Green Book (see Appendix E, Table E-31) shows that, primarily because of increased work rates, the poverty rate of families headed by never-married mothers

before any government taxes or transfers, fell from 48.3 percent to 39.6 percent between 1989 and 2006. Thus, due to work effort by these mothers, the poverty rate before taxes and transfers had fallen by nearly 20 percent. When support for these working families from the work support system were added to income in 2006, the poverty rate fell from 39.6 percent to 26.1 percent or by an additional 35 percent. The combination of increased work rates and an improved work support system has had a major impact on poverty among families headed by never-married mothers.

[12] Poverty rates for families headed by single mothers decreased between 1996 and the start of the Great Recession in late 2007, from 35.8 percent to 30.7 percent. Even with increased hardship and unemployment during the recession, the 2010 rate of 34.2 percent is still below the 1996 rate. According to Brookings tabulations of data from the Current Population Survey, the average employment-to-population ratio for never-married mothers in the five years before welfare reform in 1996 was 44.6 percent. The ratio increased every year through 2000, increasing to 65.6 percent in that year, an increase of 47 percent compared with the five years before welfare reform. Even in 2010, after two recessions, the rate was 58.7 percent, still over 30 percent above the pre-welfare reform level.

[13] Pietro S. Nivola, Jennifer L. Noyes, and Isabel V. Sawhill, “Waive of the Future? Federalism and the Next Phase of Welfare Reform,” Brookings Institution, Welfare Reform and Beyond Policy Brief #29, March 2004.

[14] Ron Haskins and Jon Baron, “Building the Connection between Policy and Evidence,” London: NESTA, September 2011.

[15] Isabel V. Sawhill and Jon Baron, “Federal Programs for Youth: More of the Same Won’t Work, Youth Today, May 2010, p. 21.