Skip to main content

Testimony: Robert Greenstein Before the House Budget Committee Hearing on Strengthening the Safety Net

Thank you for the invitation to testify today.  I am Bob Greenstein, President of the Center on Budget and Policy Priorities, a policy institute located here in Washington.  I also served many years ago as Administrator of the Food and Nutrition Service at USDA, which operates the Supplemental Nutrition Assistance Program (SNAP, then called food stamps) and other domestic food assistance programs.

My testimony covers several related issues: a look at the functioning of the safety net as a whole; a more intensive focus on Medicaid and SNAP; and a discussion of changes in welfare over the past 15 years and their implications.

I.        The Safety Net

The safety net is far from perfect and contains areas that merit strengthening.  Yet as a result of a series of mostly bipartisan decisions over several decades, it is functioning far better than is often understood.

Let’s start with its effect on poverty.  Analysts across the political spectrum agree that to measure the safety net’s impact on poverty, one cannot use the “official” measure of poverty — which completely ignores SNAP (formerly known as the Food Stamp program), the Earned Income Tax Credit (EITC), rental subsidies, and the like and also fails to adjust for taxes that are withheld from paychecks and that families thus can’t spend.  If one tries to examine poverty trends over the past half century using the “official” measure, serious distortion occurs — the sharp decline in cash welfare assistance pushes poverty rates up, while the expansions (compared to the 1960s, 1970s, or 1980s) in food stamps and the EITC are ignored.  This makes it look like no progress on poverty occurred because those safety net measures that contracted are counted in the official poverty measure while those that expanded are not.  This distortion is one of the reasons that analysts prefer the broader measure.

In the mid-1990s, a National Academy of Sciences’ expert panel recommended use of such a broader measure of poverty, and the Census Bureau now issues several alternative, broad poverty measures.  Under the measure most closely resembling the NAS recommendation, the poverty rate stood at 15.5 percent in 2010.  Yet under the same measure, the poverty rate without the safety net — that is, the poverty rate based on household incomes before government assistance is counted — was 29 percent.  In other words, the safety net cut poverty nearly in half compared to what it otherwise would be.

To be sure, without the programs, some people might have worked more in 2010 (although it is hard to see where the additional jobs would have come from given the depressed labor market).  The impact of the safety net on poverty has been studied extensively, however, and the finding cited above about the safety net’s impact in reducing poverty is strongly corroborated by a comprehensive review and synthesis of the research literature on the effects of the safety net on poverty conducted recently by some of the field’s leading scholars and published by the National Bureau of Economic Research (NBER).  In this study, the scholars examined the impact of the safety net on the amount that people work, and found the safety net’s overall impact on work to be small.  They also found that, after taking this behavioral effect into account, the safety net lowers the poverty rate by approximately 14 percentage points.  In other words, one of every seven Americans would be poor without the safety net, but is above the poverty line because of it.[1]

One also can look at the Census data on how many people individual programs lift out of poverty.  In 2010, for example, the Earned Income Tax Credit and the Child Tax Credit lifted about 9 million people in low-income working families above the poverty line, including 5 million children.  SNAP lifted about 4 million out of poverty.

Among the most striking figures are those that track poverty rates over the last few years.  Given the depth and severity of the economic downturn (sometimes called the Great Recession), one would have expected poverty to have soared.  It didn’t.  The Census Bureau’s broad poverty measures show relatively modest increases in poverty, which stands in sharp contrast to the deep plunge in the economy and the doubling of the unemployment rate.  (The Census measure most closely based on the National Academy of Sciences’ recommendations showed a poverty rate of 15.3 percent in 2007, 15.7 percent in 2009, and 15.5 percent in 2010.  Other broad measures of poverty showed somewhat more of an increase but one that still is modest given the economy’s marked downturn.)

Why didn’t poverty rise more as unemployment spiraled upward?  The answer shows up in the data.  The “automatic stabilizer” response of programs like SNAP and unemployment insurance, supplemented by the temporary increases in assistance in various safety net programs that were provided under the Recovery Act, counteracted most of the increase in poverty that would otherwise have occurred.  This is a substantial accomplishment, and one that speaks well for our nation.

Questions raised about the safety net include its impact on dependency, the degree to which dollars go for administrative costs rather than benefits or services, and its effect on the long-term fiscal problem.  Another issue concerns its effect on deep poverty — people with incomes below half of the poverty line.  Let me examine each of these issues in turn.

Over the past several decades, we have moved heavily toward what analysts call a “work-based safety net.”  Cash welfare assistance for families without earnings has diminished greatly, while support for the working poor and near poor through the EITC, the Child Tax Credit, Medicaid, and SNAP has grown.  The results are notable.  Even though 2010 was a year of economic distress, with an average unemployment rate of 9.6 percent, it was marked by the following results:

  • Some 91 percent of all spending in 2010 on federal entitlement benefits went to people who either are not expected to work because they are 65 or older or disabled, or were members of working households (with work defined as a household with a member who worked more than 1,000 hours during the year).[2]
  • Image
    Moreover, seven of the other nine percentage points of entitlement benefits went for unemployment insurance that people typically must have a significant work history to qualify for, Social Security survivor benefits for widows and orphans of deceased workers, Social Security benefits for retirees aged 62-64, or medical care.

These data are consistent with the findings of the NBER study cited above that presents a comprehensive review of the research literature and finds that today’s more work-based safety net does not have a large effect on work effort.  It also is worth noting that the study reported that the research literature shows that the effects of SNAP on work effect and earnings are “small” and often statistically insignificant, and the effects of Medicaid on work “appear to be minimal.”  (The findings on these programs differ significantly from the findings regarding the old welfare program prior to welfare reform.)  Some studies also find increases in work effort when eligibility for programs such as Medicaid is extended to children or adults in low-income working families so families don’t need to be non-workers receiving cash welfare assistance to qualify.  These findings demonstrate, contrary to popular impression, how strongly intertwined the safety net has become with work, as well the important role it plays in providing help when the economy falters and jobs are scarce.

In other words, concerns that the safety net is leading millions to become dependent and cease working are not borne out by the research.  (Nevertheless, we should explore ways to encourage more people nearing retirement age to work longer.  The challenge there is to find ways to do so without impoverishing people in that age bracket who cannot work or cannot find a job because of their occupational background and skills or due to health reasons.)

A second question regarding the safety net regards the ratio of benefits and services to administrative costs.  Some administrative costs are necessary in any program to assure program integrity (i.e., that the people served are truly eligible and that appropriate amounts are paid).  Here, too, the safety net performs well.  A recent examination of six major means-tested programs — Medicaid, SNAP, Supplemental Security Income (SSI), Section 8 housing vouchers, school meals programs, and the EITC — finds that in each of these programs, between 90 percent and 99 percent of expenditures go for benefits or services that reach the intended beneficiaries.  Federal and state administrative costs combined account for only 1 percent to 10 percent of program costs, with most of the administrative costs occurring at the state level.[3]


A third issue — and a particularly important one in the current budgetary context — involves the safety net’s cost trajectory.  The nation faces a serious long-term fiscal problem, as a result of a large projected imbalance between revenues and expenditures.  Under current policies, expenditures will climb as a percentage of Gross Domestic Product (GDP), while revenues remain at levels that are low relative to need, given the aging of the population and continuing increases in health care costs throughout the health care system.

This raises an important question: will means-tested programs rise in cost as a share of GDP and thereby contribute to our long-term fiscal problems?

As is well known, Medicaid is projected to rise in cost, for several reasons.  Health care costs throughout the entire U.S. health care system — in both the public and private sectors — have been growing faster than GDP for several decades.  Medicaid isn’t the cause of this systemwide cost growth, and over the past decade, Medicaid costs per beneficiary have been rising more slowly than per-beneficiary cost under private insurance.  Furthermore, Medicaid costs per beneficiary (adjusted for differences in health status) are substantially lower than those under private insurance because Medicaid pays providers lower rates and has lower administrative costs.  But systemwide health care costs increases, driven in part by medical advances that improve health and lengthen life but add to costs, will push up health care costs across the board.

A second reason that Medicaid costs will rise is the aging of the population.  Older people have much higher average health care costs than younger people do.  Elderly and disabled beneficiaries account for 25 percent of Medicaid beneficiaries today but 68 percent of program costs.  As the population ages, the number and share of beneficiaries who are elderly will rise, increasing program costs.

Yet another reason that Medicaid costs will rise is the continued erosion of employer-based health coverage.  Over time, fewer low-income people are able to get coverage through their (or a family member’s) employer, causing more of them to turn for coverage to Medicaid.

Finally, the coverage expansions in the Affordable Care Act — both in Medicaid and for subsidies to help near-poor and many middle-income families afford coverage in the new health insurance exchanges — will raise expenditures for means-tested health care expenditures, even though the Congressional Budget Office (CBO) projects that these expenditure increases will not add to deficits (because the costs are offset under the Affordable Care Act primarily through savings in Medicare and new revenues).

For these reasons, if one looks at total means-tested program costs, they appear to remain high in the years to come.  But if one examines costs for means-tested program other than health care programs, the picture changes dramatically.  Means-tested programs outside of health insurance will decline in cost as the economy recovers and are not projected to rise in future decades as a percentage of GDP.  Here are the data, all of which come from the official historical tables on federal budget expenditures and CBO’s projections of future expenditures.

  • Image
    In fiscal year 2011, total federal expenditures for means-tested entitlement (or mandatory) programs outside health care programs equaled 2.0 percent of GDP.  This was about 50 percent higher than the average for the prior 40 years — which was 1.3 percept of GDP.  The costs of these programs have risen significantly in the last few years.[4]
  • But the recent increases are largely driven by the economic downturn and temporary program expansions under ARRA.  The CBO projections show that total expenditures for means-tested entitlements outside health care will decline steadily as a share of the economy as the economy recovers, falling to 1.3 percent of GDP by 2020 and thereafter.  (These figures do notassume sequestration.)
  • In other words, by 2020, total means-tested entitlement expenditures outside health care, measured as a share of GDP — including expenditures for SNAP, the EITC, and the other programs — will return all of the way to their prior 40-year average.

These figures do not include low-income discretionary programs.  Under the Budget Control Act’s caps, non-defense discretionary spending will fall over the decade to its lowest level as a share of GDP since 1962 (and probably since 1931).  As a result, some decline in low-income discretionary programs appears inevitable.  Thus, totalexpenditures on low-income (or means-tested) programs outside health care — including low-income discretionary programs — are expected to decline over the coming decade to a level below their average over the prior 40-year period.  As a result, this part of the budget isn’t contributing to the long-term fiscal problem.

Increases in SNAP Costs

There has been particular misunderstanding of what has happened with expenditures for SNAP and the reasons for that.  SNAP costs have grown substantially over the past decade — by more than can be explained just by the economic downturn.  This has led some to assume that eligibility expansions are largely responsible and that as a result, SNAP expenditures and participation will continue to grow.  Careful analysis indicates, however, that both such assumptions are incorrect.

As in many other areas of budgetary analysis, the year picked as a “starting point” for an analysis matters.  For SNAP, choosing a starting point that only looks back ten years provides a somewhat distorted picture of program growth.  This is because SNAP participation and costs had plummeted at that point, in part due to a large decrease in the proportion of eligible families receiving SNAP.  The 1996 welfare law was intended to encourage work, but due to problems in state administrative systems in the first years of the welfare law, many families moving from welfare to working-poor status were cut off SNAP when they left welfare, even though they remained eligible for SNAP.

This was a result contrary to what Congress intended.  Aggravating this problem, some states instituted administrative practices in those years that had the unintended effect of making it harder for many working-poor parents to participate (largely by forcing them to take too much time off from work for repeated visits to SNAP offices at frequent intervals, such as every 90 days, to reapply for benefits).  This prompted many analysts, like my fellow panelist today Ron Haskins, to call for reforms that would improve access to SNAP for low-income working families and led both the Clinton and Bush administrations to act to address this problem.  There was bipartisan consensus that having a policy where you needed to be on welfare to readily receive food stamps, and encountered difficulty receiving food stamps if you left welfare and worked for low wages, would reduce work incentives and was contrary to welfare reform goals.  Congress enacted significant, although relatively modest, changes in 2002 and 2008 to lessen barriers to SNAP participation among the working poor, as well as modest improvements in benefits that largely helped low-wage workers and their families.

My point here is that there are threemain reasons for the large increase over the past decade in SNAP expenditures and cost — the economy, the temporary increase in SNAP benefits enacted as part of the Recovery Act, and a large increase relative to ten years ago in the percentage of individuals eligible for food stamps that actually receive them.  This percentage fell from 75 percent in 1994 to 54 percent in 2002, but is back to 72 percent today.[5]   Of particular note, the percentage of eligible individuals in low-income working families receiving SNAP has been rising steadily from 43 percent in 2002, to about 60 percent in 2009, the highest percentage on record. 

The temporary SNAP benefit increase also has had a large effect, accounting for more than 25 percent of the growth in the program between 2007 and 2011.  It expires in November 2013.

These three factors dwarf the impact of eligibility changes.  For example, CBO estimates that what is referred to as “broad categorical eligibility” accounts for less than 2 percent of SNAP costs.

What lies ahead with regard to SNAP costs?  Consistent with past experience, SNAP caseloads and expenditures will decline as unemployment and poverty fall.  The SNAP “participation rate” — the percentage of eligible households that receive benefits — will likely decline some as well.  The research literature shows that the percentage of eligible households who actually apply for and receive SNAP benefits is greater when the benefits are larger and lower when the benefits are smaller.  The temporary increase in the SNAP benefit level enacted as part of ARRA has almost certainly increased the SNAP participation rate, and after the temporary increase ends in October 2013, the participation rate is likely to decline.

Where all this leads is depicted in the graph below.  The graph shows actual SNAP costs, as a share of GDP, from 1995 to the present, along with CBO’s projection of costs as a percentage of GDP through 2022.  As the graph indicates, by 2018, costs are expected to decline all of the way back to their 1995 level as a percentage of GDP, and then to edge below that. 


A Bleaker Part of the Story

I conclude this part of my testimony with a troubling set of data that are a source of serious concern — data on deep poverty, particularly among children.

Using Census data and counting non-cash benefits as income as most analysts favor, the number (and percentage) of children living below half of the poverty line increased by 650,000 between 1995 and 2005, even before the economy turned down.[6]   The welfare law has played a significant role here.  On the one hand, changes wrought by the 1996 welfare law, in combination with expansions in the EITC and, in the late 1990s, a very strong labor market, pulled more single-mother families into the labor market and raised many of their incomes.  On the other hand, the welfare changes also deepened poverty among another group of single-mother families — including some mothers with less education and skills and more physical, mental health, or other problems.  A number of families were lifted out of poverty by increased earnings supplemented by the EITC.  But some other families fell deeper into poverty as a result of having neither earnings nor cash assistance (or earnings that were more than offset by the loss of cash assistance).

That this was the primary cause of the increase in deep poverty is seen in the data.  In 1995, cash assistance provided by the old AFDC program lifted 2.2 million children out of deep poverty (i.e., lifted them above 50 percent of the poverty line).  It raised 62 percent of the children who otherwise would have been below half of the poverty line above that level.  By 2005, cash assistance provided under the Temporary Assistance for Needy Families (TANF) block grant lifted only 650,000 children above half of the poverty level, or just 21 percent of those who otherwise would be in deep poverty. 

Number of U.S. Households Living Below World Bank Measure of Poverty in Developing Nations:
Living on Less Than $2 a Day, Per Person
  Cash Income Cash Income plus Food Stamps
1996 636,000 households with
1.4 million children
475,000 households
Start of 2011 1.46 million households with
2.8 million children
800,000 households
Source:  Shaefer and Edin, “Extreme Poverty in the United States,” 1996 to 2011.

The same phenomenon is reflected in a recent study by poverty researchers that finds that both the number of families and the number of children who live below a standard that the World Bank uses to measure serious poverty in third-world countries — living on less than $2 per person per day — has doubled since 1995.[7]

These findings are of particular concern, especially in light of emerging research which shows that among low-income families, the level of family income when a child is young affects his or her school achievement and may well influence later employment and earnings as an adult.  Research finds, for example, that for a child up to age 5, an additional $3,000 in annual family income — whether in earnings or unearned income like government benefits — is associated with a substantial increase in earnings in adulthood (the leading study finds a 17 percent increase in earnings in adulthood and an average of 135 hours of additional work per year.[8] )  Allowing more young children to grow up in deep poverty not only causes immediate hardship but appears also to have long-term adverse effects on earnings and employment when the children grow up.

The research literature also shows that programs lifting the incomes of young children’s families can boost children’s achievement in school.  Gordon Berlin, president of MDRC, the leader in evaluation of welfare-to-work programs, has reported that “We have reliable evidence involving thousands of families in multiple studies demonstrating that ‘making work pay’ [through assistance that supplements low earnings] causes improvements in young children’s school performance.”  Researchers Greg Duncan and Katherine Magnuson report beneficial effects on young children in low-income families from increases in family income, whether the increase comes from earnings or other sources such as government assistance.  These findings underscore the importance of programs such as SNAP, the EITC, and the Child Tax Credit as work supports.

It bears noting that were it not for the SNAP program, the rise in severe poverty would be substantially worse.  The data show that SNAP cuts nearly in half the percentage of children living below the World Bank poverty standard (i.e., living on less than $2 per day).  In 1996, a number of the Republican members of the Ways and Means Committee responded to criticisms that the welfare reform legislation would severely harm poor children in part by noting that food stamps would remain as a floor under poor children.  The importance of retaining the SNAP program structure alongside the TANF block grant cannot be overstated.

II.         Low-Income Programs, the Budget, and Medicaid and SNAP

The nation is on an unsustainable fiscal course.  But we also have levels of poverty and inequality significantly higher than those in most other western industrialized nations.  Furthermore, poverty — and especially deep poverty — among children may have negative long-term effects on the economy as well as on the children themselves.

The report issued in late 2010 by Erskine Bowles, Alan Simpson, and a majority of Bowles-Simpson commission members emphasized, as one of its core or guiding principles, that deficit reduction should not increase poverty or harm the disadvantaged.  Their report did notcall for reductions in any low-income program outside Medicaid.  Last year, a group of Christian leaders ranging from the Catholic Bishops’ Conference and the Episcopal Church to the Salvation Army and the National Association of Evangelicals issued a call for policymakers to safeguard the poor in deficit reduction and to draw a “circle of protection” around programs targeted on them.

I would urge policymakers to seriously consider these principles.  In this spirit, I would like to discuss two proposals that I believe would cause substantial harm — the proposal to block-grant Medicaid and cut its funding severely, and the proposal to do the same with regard to the SNAP program.


The Medicaid proposal reflected in the budget resolution the House adopted last month would reduce federal Medicaid funding by 34 percent by 2022, and based on CBO analysis of a similar proposal in last year’s House budget, by nearly 50 percent by around 2030.  (These reductions would be in addition to the effect on Medicaid of repealing the health reform law.)  The depth of the cuts would grow larger each year because funding would no longer keep pace with health care costs or with Medicaid enrollment growth as the population ages and employer-based health coverage continues to erode.[9]

Could cuts of this magnitude be absorbed through state efficiencies without harming low-income children, parents, and elderly and disabled people?  The answer is almost certainly no.

Medicaid already is considerably more efficient and less costly than private health insurance, primarily because it pays providers less (and also has lower administrative costs).  As the graph at right shows, Medicaid costs per enrollee are well below those for private insurance.  In addition, Medicaid costs per beneficiary have been rising more slowly in recent years than private-sector health costs have.

Moreover, state Medicaid programs already seek out savings.  Most states already contract with private managed-care companies to operate their Medicaid programs for parents and children.

Is Medicaid coverage overly broad so that large savings could be obtained?  That is hardly the case.  Today, a working-poor parent loses Medicaid eligibility in the typical (or median) state when his or her income reaches just 63 percent of the poverty line, or $12,027 for a family of three — which is well below full-time minimum wage earnings.  And an unemployed parent loses eligibility in the typical state when her income reaches just 37 percent of the poverty line.  In addition, most states cover elderly and disabled people only up to a level between about 75 percent and 100 percent of the poverty line, and in most states, adults who aren’t raising minor children and aren’t elderly or disabled are not eligible for Medicaid at all, regardless of how poor they are.[10]

This is the program from which a cut of about one-third by the tenth year, and one-half by the twentieth year, would be extracted.  The effects would be substantial, as leading non-partisan institutes have concluded.

In its analysis of the Medicaid proposal in the House budget, the Congressional Budget Office wrote that “the magnitude of the reduction in spending…means that states would need to increase their spending on those programs, make considerable cutbacks in them, or both.  Cutbacks might involve reduced eligibility for Medicaid and CHIP, coverage of fewer [health] services, lower payments to providers, or increased cost-sharing by beneficiaries — all of which would reduce access to care”.[11]   In other words, unless states came up with rather massive new sums to offset the very large losses in federal funding, they would be compelled to institute deep cuts that would harm low-income families and individuals.

In addition, the Urban Institute analyzed the effects of the very similar Medicaid proposal in last year’s House budget.  Urban Institute researchers examined a range of scenarios, in which they assumed various levels of savings through state actions to save money without casting low-income people off the program.  The researchers estimated that under the most optimistic scenario (which assumed states could somehow reduce growth in Medicaid expenditure costs per beneficiary all the way to the rate of per-capita GDP growth), 14 million low-income people who otherwise would have Medicaid coverage would be made ineligible for it by 2021.  The researchers also estimated that if states protected the elderly and people with disabilities, the enrollment of children and adults would drop by 27 million.  (The decline would be greater because elderly and disabled beneficiaries have higher health care costs and are thus more costly to insure.)  Most of those who would lose coverage would be expected to become uninsured.

The Urban Institute researchers estimated that provider reimbursements would be cut sharply as well; the researchers’ best estimate was cuts in the range of 30 percent.  These cuts would be made to provider payment rates that already are significantly below the reimbursement rates in Medicare and private insurance (and that have already been cut in recent years by states coping with budget shortfalls).  Provider cuts of that magnitude would almost certainly cause the withdrawal of many providers from Medicaid and thereby reduce access to care for those low-income people whom Medicaid still covered.

Grim as these figures are, they understate the likely effects in years when the economy weakened, unemployment and poverty mounted, and more people lost employer-based health coverage along with their jobs.  The Urban Institute has estimated that a one-percentage point increase in unemployment results in 1 million more low-income children and non-elderly adults receiving Medicaid (and CHIP) coverage; Medicaid responds automatically when this occurs.  Under a block grant, however, it no longer would; as a result, in recessions, the ranks of the uninsured would swell.  And states would be in no position to fill the gap, as their tax revenues fall in recessions.

A block grant also would fail to allocate resources equitably across states when state economies grow or contract at different rates, when the share of the population that is elderly increases at different rates, or when employer-based coverage erodes at a different pace.  The states where need increased more rapidly would be placed at a particular disadvantage.

A block grant also would fail to respond automatically to unanticipated costs resulting from the onset of an epidemic (such as a severe flu epidemic), the onset of new diseases (as was the case with HIV-AIDS), or medical breakthroughs that improve health and save lives but at considerable cost.  Suppose a major breakthrough related to cancer, Parkinson’s, Alzheimer’s, diabetes or other diseases is found, but use of the breakthrough technologies or drugs raises costs?  Under a block grant, state Medicaid programs would have grave difficulty responding.

This is why the proposal would almost surely lead ours to become more of a two-tiered health system, where those with low incomes are denied access to various health services and medical advances while more affluent Americans receive them.  David Walker, then the Comptroller General of the United States, noted several years ago that “Medicare and Medicaid cannot grow over the long term at a slower rate than cost in the rest of the health care system without resulting in a two-tier health care system.”  The block grant proposal overlooks this crucial point.

SNAP (Food Stamps)

The SNAP program has a number of attributes.  It is well targeted: 98 to 99 percent of SNAP households have disposable income (or “net income”) below 100 percent of the poverty line.[12]  And, over the past decade, SNAP has become an important work support for households with low wages.

Even in 2010, when the unemployment rate was 9.6 percent and jobs were hard to find, the majority of SNAP households that contained a non-disabled working-age adult were working households.  Indeed, the number of low-income working households on SNAP has nearly tripled over the past decade, reflecting both a substantial increase in the percentage of eligible low-income working households that actually receive SNAP benefits and some wage erosion at the low end of the wage scale, which makes more working households eligible for SNAP.  The data also show that the majority of non-disabled working-aged adults receiving SNAP at any time who are not currently working are people who worked in the prior year or will work in the following year.

The evidence does not support the notion that SNAP suppresses work effort.  As noted earlier, the leading review of the research in the field finds little evidence that SNAP reduces work effort.  I would also note that among SNAP households that worked in the year before they began to receive SNAP, only 4 percent did not work in the following year.  This suggests that turning to SNAP when families face hard times does not lead them to cease working.

Moreover, SNAP has rather strong work requirements.  For childless adults, in fact, they are the toughest work requirements of any federal program.  People aged 18 to 50 who are not raising minor children may receive SNAP benefits for only three months (while they are not employed at least half time) out of every three years, even if they have looked diligently for work but can’t find it.   (This is a provision of the 1996 welfare law that President Clinton thought was harsh and inequitable and sought unsuccessfully to change).

I should also note that SNAP work rules and requirements complement those in TANF.  The 1996 welfare law gave states options to impose tough SNAPsanctions on TANF households who fail to comply with work or other TANF behavioral requirements — including the termination of all SNAP benefits for the entire family.  States use this flexibility to design and augment the force of their TANF work and behavioral requirements.  States also operate SNAP employment and training programs for SNAP participants who are not subject to work requirements in TANF or other programs, and states can and do impose SNAP benefit sanctions on participants who fail to comply with those work requirements.

The SNAP program does have a weakness in this area: due to limited funding, the SNAP employment and training program is able to provide work, workfare, or training slots for only a relatively modest portion of SNAP recipients who lack employment and aren’t enrolled in, or subject to, another work program or set of work requirements such as under TANF.  Under the proposed SNAP block grant, this problem could become more acute, because of the large drop in overall funding for the program.

This brings me to the proposal in the House budget to convert SNAP to a block grant and cut it at least $133.5 billion over ten years.  This proposal is a source of great concern.

SNAP funds go overwhelmingly for food purchases — nearly 95 percent of federal SNAP expenditures go directly for benefits to recipients.  Most of the remainder goes to determine eligibility, administer the work requirements and work programs, and approve and monitor compliance by retail food stores — costs that would largely remain under a block-grant.  The math here is inexorable — the only way to secure savings of this magnitude would be to cut eligibility, benefit levels, or both.

If the savings were to come entirely from eliminating eligibility for currently eligible households, more than 8 million people would need to be cut adrift from the program if the cuts began taking effect in 2013.  (If the cuts did not begin until 2016, the year in which the House budget envisions converting SNAP to a block grant, an average of almost 10 million people would have to be cut from the program in the years from 2016 through 2022 to achieve the required savings.)  States would likely cut, at least in substantial part, by lowering income (and possibly asset) limits, which would primarily remove low-income working families from the program.  A program structure that provides SNAP to families on public assistance — but denies it to many who work for low wages rather than relying on welfare — would be a program with weaker work incentives than SNAP has today.

If the savings were secured by cutting benefit levels instead, increased hunger and food insecurity would likely result.  Considerable research suggests that the SNAP benefit level may already be too low to enable many families to secure an adequate diet throughout the month.  (Many run out of adequate food toward month’s end.)  The Institute of Medicine is currently reviewing this matter and examining whether the current SNAP benefit level is adequate.  It would be dangerous to shrink benefit levels for needy children, seniors, and others.

Converting SNAP to a block grant at substantially reduced funding levels also would have other deleterious effects.

  • SNAP would no longer be able to respond to increased need during economic downturns, resulting in increased hardship and hunger in recessions.
  • Nor would SNAP be able to bolster the economy during recessions as it does today.  In studying the effect of 22 different tax and spending options to promote economic growth and jobs in a weak economy, economist Mark Zandi of Moody’s Analytics rated temporary increases in SNAP benefits first in effectiveness per dollar of cost, ahead of both unemployment insurance and all tax-cut options.  CBO also gives SNAP increases its top rating for effectiveness in a weak economy.  This is because SNAP benefits are quickly spent and injected into the economy, rather than saved.  Preventing SNAP from expanding automatically as the economy weakens by converting it to a block grant would remove what economists call an “automatic stabilizer” and hence likely make recessions somewhat deeper and longer.
  • Finally, a proposal like that reflected in the House budget would make deep poverty more widespread and severe, especially among children, who make up about half of all SNAP beneficiaries.  I noted earlier in this testimony the emerging research on the importance to children’s future earnings prospects of adequate family income and purchasing power during early childhood.  Large SNAP cuts would go in the opposite direction and make this problem more acute.

Above all else, there is the issue of children’s health.  I am old enough to remember the mid and late 1960s, when each state sets its own food stamp rules, some states cut off families at income levels as low as 50 percent of the poverty line, and some states adopted barriers that impeded participation (in some cases, with disproportionate effects on members of some minority groups).  Two teams of medical researchers conducted nutrition surveys in the late 1960s and found rates of childhood malnutrition and related diseases in some poor areas of our country that were akin to those in some third-world countries.  This led to a national bipartisan consensus — led by President Richard M. Nixon — to establish national eligibility and benefit standards for food stamps.  In the late 1970s, after the national standards had taken effect, the medical teams returned to many of the same poor areas they had studied in the late 1960s and found dramatic improvement among poor families and especially among poor children.  Child malnutrition and related conditions had become rare.  In a famous report on their findings, the medical researchers wrote.

In the Mississippi delta, in the coal fields of Appalachia and in coastal South Carolina — where visitors ten years ago could quickly see large numbers of stunted, apathetic children with swollen stomachs and the dull eyes and poorly healing wounds characteristic of malnutrition — such children are not to be seen in such numbers.  Even in areas which did not command national attention ten years ago, many poor people now have food.…[13]

The researchers credited food stamps as the single largest factor for this striking progress, concluding that “no program does more to lengthen and strengthen the lives of our people than the Food Stamp program.”  I believe this is a lesson we shouldn’t forget.

III.        Implications of Welfare Reform

The 1996 welfare law is frequently cited as a reform that transformed a key part of the safety net.  It’s often either celebrated as a spectacular success or vilified as a cruel failure.  Neither stereotype is consistent with the evidence.

Significant increases in employment among single mothers began in the early 1990s, prior to the welfare law, and continued after its enactment, with the change over the decade being quite strong.  Welfare reform was implemented at the same time that a robust expansion in the Earned Income Tax Credit was taking full effect and that the labor market was expanding rapidly, with tremendous job creation; the unemployment rate fell to 4 percent at the end of the decade.  For a number of low-income single-parent families, the result was an increase in income and a reduction in poverty.

These effects, much heralded at the time, were not the result solely (or very likely, even primarily) of the welfare law by itself.  They resulted from a combination of all three of these factors.  A highly regarded study by Jeffrey Grogger of the University of Chicago found the EITC expansion actually to have a larger effect in increasing work effort than the welfare law changes.[14]   But the two changes reinforced each other, amounting to a combination of “carrots and sticks.”

Even in the hot labor market of the late 1990s, however, some families and children experienced significantly increased hardship in the aftermath of the welfare law.  As a group, those welfare recipients who were the most employable — often with the most education, skills, and/or work experience — made the most progress.  But too often, families with the least education and job skills and/or the deepest physical, mental health, or other problems sank deeper into poverty, ending up with neither cash assistance nor earnings to support their families.  Ironically, the welfare law both reduced poverty among many of the better equipped recipients and increased deep poverty among a number of the most disadvantaged families.

The overall results were the most favorable in the hot labor market of the late 1990s, while TANF’s weaknesses have been greatest in recent years, for three reasons — the lack of availability of jobs in the economic downturn, the inability of the block grant to respond effectively to increases in need as the economy turned down, and the effect of 15 years of erosion in federal TANF funding levels that are now lower even in nominal terms than they were in 1996, and are much lower (28 percent lower) once one adjusts for inflation.

Some of the specific results are quite disturbing.

  • In 1995, for every 100 families with children living below the poverty line, 68 received some assistance from AFDC, the cash assistance program that preceded TANF.  Today, for every 100 such poor families with children, only 27 receive any cash assistance through TANF (and that includes working-poor families that receive some assistance as a supplement to their low wages).
  • For those poor families with children that do receive assistance, benefit levels have plummeted.  In the majority of states, TANF benefits now fail to lift a family with no other cash income even to 30 percent of the poverty line (before SNAP benefits).  In no state are TANF benefits sufficient to pay the rent on a modest apartment (based on HUD’s “fair market rents”) even if the entire benefit is used for rent.
  • With federal TANF funding having fallen substantially in real terms even as need has increased, many states have cut TANF employment and training programs as well, and services to help poor parents find jobs — steps counter to the goals of welfare reform.

One factor that has intensified these problems is that many states took advantage of their flexibility to move TANF block-grant funds around and have used some of the federal TANF funds to substitute for (or supplant) state expenditures, and thereby to fill state budget holes or finance new expenditures (or tax cuts) that may have little to do with low-income families.  The GAO examined a group of 10 states in 2001, and reported that five of the states in its study “used between 15 percent and 25 percent of their annual [TANF] block grants to finance state programs that had traditionally been financed with state dollars,” while another four states diverted up to 10 percent of their TANF funds in this way.[15]

This state behavior has had particularly deleterious effects on poor families in the current economic downturn: in general, once states shifted these funds, they were unable (politically) to move them back when need among families with children increased in the recession.

To reiterate, I am not saying there have not also been positive effects from the welfare changes.  The results are mixed — positive for some families, negative for others — strongest in an economy where jobs are plentiful, weakest in an economic downturn.

But if the goal is both to promote work and to maintain an adequate safety net for those lacking well-paying jobs — and especially for their children, and for low-income elderly and disabled people who cannot work — then it would be a serious mistake to convert Medicaid and SNAP to block grants as well and to shrink their funding.  Doing so would magnify TANF’s weaknesses and would substantially increase the ranks of the uninsured and the deeply poor.

Improvement in the safety net is needed, but the most important step we can take is to develop policies that are tested and proven to be effective in creating jobs for those who encounter difficulty securing employment or in helping the most disadvantaged people to improve their skills so they have a better chance of competing in a tough labor market.  Many of those who turn to the safety net are people who are currently working but earn too little (in part because of limited education and skills) or people who want to work but cannot find employment.

In conclusion, I earnestly hope that policymakers will be able to step back from the usual type of Washington debates and political battles and consider what policies would be best for “the least among us.”  I urge you to follow the Hippocratic oath and “do no harm.”  I also implore you to adopt the Bowles-Simpson principle of protecting the disadvantaged and avoiding measures that would increase poverty and hardship in a nation as abundant as ours.

Thank you.


End Notes

[1] Yonatan Ben-Shalom, Robert A. Moffitt, and John Karl Scholz “An Assessment of the Effectiveness of Anti-Poverty Programs in the United States,” NBER Working Paper 17042, May 2011.

[2] If the universe of programs examined is broadened to include major low-income assistance programs that are not entitlements — low-income housing assistance, WIC, and low-income energy assistance — the figure remains essentially unchanged: 90 percent of benefit expenditures go to people who are elderly, disabled, or members of working households.  Even if one looks solely at means-tested entitlement programs, the figure is still high, at 83 percent, meaning that even in a severe recession year, five of every six benefit dollars in means-tested entitlement programs went to people who were elderly, disabled, or members of working households.

[3] Center on Budget and Policy Priorities, January 23, 2012.

[4] Means-tested mandatory health care programs include Medicaid, CHIP, and subsidies for the purchase of health insurance under the Affordable Care Act (along with a few very small programs such as mandatory supplements to a few areas of discretionary health funding).

[5] The most recent year for which USDA publishes estimates is 2009.

[6]   In this measure, non-cash benefits are counted (and taxes subtracted), consistent with the broad measures of poverty.  Benefits from TANF, SSI, and SNAP are counted using the HHS/Urban Institute TRIM model in order to adjust for underreporting of benefits in the Census data.  If one uses a poverty definition in line with the poverty measurement recommendations of the National Academy of Sciences such as subtracting work expenses and employing a modestly updated poverty line, the estimated number of children in deep poverty rises further, to over 1 million between 1995 and 2005.  See Arloc Sherman, “Safety Net Effective At Fighting Poverty But Has Weakened For The Very Poorest,” Center on Budget and Policy Priorities, July 2009, at

[7] H. Luke Shaefer and Kathryn Edin, “Extreme Poverty in the United States, 1996 to 2011,” National Poverty Center, 2012, at

[8] Greg J. Duncan and Katherine Magnuson, “The Long Reach of Early Childhood Poverty,” Pathways, Winter 2011.

[9] The annual block grant amounts would rise only with general inflation and the percentage increase in the size of the overall U.S. population — factors that do not take into account that health care costs rise faster than general inflation or that the aging of the population will cause the elderly component of the Medicaid program, which has much higher per-beneficiary health costs, to rise more swiftly than the rate of overall U.S. population growth.  As a result, we estimate that the annual block-grant funding amounts would rise at an annual rate at least 3.5 percentage points lower than the growth rate currently projected for Medicaid (excluding the effects of the Affordable Care Act).  The proposed funding structure also does not take into account that erosion of employer-based coverage also causes Medicaid enrollment to grow over time.

[10] This would change under the Affordable Care Act, which would extend medical coverage to all non-elderly individuals with incomes below 133 percent of the poverty line.

[11] CBO, “The Long-Term Budgetary Impact of Paths for Federal Revenues and Spending Specified by Chairman Ryan,” March 2012.

[12] Moreover, 93 percent of benefits go to households with gross incomes below the poverty line, and two-fifths of SNAP households have gross incomes below half of the poverty line. 

[13] “Hunger in America: The Federal Response,” Field Foundation, 1979.

[14] Jeffrey Grogger, “The Effects of Time Limits, the EITC, and Other Policy Changes on Welfare Use, Work, and Income Among Female-Headed Families;" Reviews of Economics and Statistics, May 2003. 

[15] GAO report 01-828, August 2001.