When we think about the division of responsibilities between state governments and the federal government, is it obvious that cash transfers for the poor should be handled primarily at the state level? I’m not sure that’s true.
At the Washington Post, Ezra Klein offers thoughts on welfare reform:
In 1996, before welfare reform passed, 68 of 100 families living in poverty with children received welfare benefits. In 2010, two years into the worst economy since the Great Depression, only 27 of every 100 such families were receiving benefits. And that’s not because they were all holding good jobs or because states had somehow managed to make the grants go further. Quite the opposite, actually.
Liz Schott, a senior fellow at the Center on Budget and Policy Priorities, explains that states use about 30 percent of their block grants to fund basic assistance — what most of us would think of as “welfare.” An additional 15 percent goes to subsidized child care. Eleven percent goes to work supports. And the other 44 percent? Miscellaneous other things, including closing state budget holes. The end result is that fewer families get welfare. That’s not a “reform.” It’s a cut.
In Sunday’s New York Times, Jason DeParle looked at Arizona, which has cut its welfare caseloads in half since the recession. “The poor people who were dropped from cash assistance here, mostly single mothers, talk with surprising openness about the desperate, and sometimes illegal, ways they make ends meet,” DeParle reported. “They have sold food stamps, sold blood, skipped meals, shoplifted, doubled up with friends, scavenged trash bins for bottles and cans and returned to relationships with violent partners — all with children in tow.”
There’s much to admire about welfare reform. In particular, the bill’s authors were right that the safety net needed to do more to help people find and keep work. But since welfare reform, we have done much to encourage work through the safety net: the earned-income tax credit, the child tax credit and various work and child-care supports. And as our workfare system has grown more robust, the traditional welfare system for those who can’t find or keep work has eroded. The share of households living on less than $2 a day has doubled to 4 percent since the passage of welfare reform, according to a study by Luke Shaefer of the University of Michigan and Kathryn Edin of Harvard. [Emphasis added]
Consider the following two passages. First,
In 1996, before welfare reform passed, 68 of 100 families living in poverty with children received welfare benefits. In 2010, two years into the worst economy since the Great Depression, only 27 of every 100 such families were receiving benefits.
Second,
But since welfare reform, we have done much to encourage work through the safety net: the earned-income tax credit, the child tax credit and various work and child-care supports. And as our workfare system has grown more robust, the traditional welfare system for those who can’t find or keep work has eroded.
While there are state-level tax credits that aim to mitigate poverty, the federal EITC provided recipients with $60 billion in 2010 and the child tax credit provided $48 billion to low- and moderate-income families, per Donald Marron of the Tax Policy Center. As Ezra suggests, the expansion of these tax benefits and work supports came after welfare reform. As state governments have reduced welfare expenditures, has the federal government filled the vacuum, at least in part?
Unlike state-level welfare spending, the federal government seems to be providing benefits (a) that are less visible, as tax expenditures are seen as somewhat different from direct social expenditures; (b) that have less stigma attached to them (bracketing the question of whether this is right or appropriate); and (c) that are funded by a level of government that is not constrained by balanced budget rules, thus making countercyclical spending much easier, and that can take better advantage of more mobile tax bases (e.g., corporate income and individual income on nonwage income).
If this is indeed what happened, and I’m by no means sure that it is (indeed, I imagine the story is a complicated one), one could argue that it represents an improvement over past practice. I’m a great believer in “competitive federalism,” a regime in which exit and mobility on the part of capital and labor are used to discipline state governments. Yet there are some domains in which this kind of competition is more constructive than others. Because capital and labor are mobile, redistribution is difficult to achieve at the local level for the simple reason that people can easily leave from one jurisdiction to another. In Milton Friedman’s view, this was a good reason for keeping redistribution as local as possible, to give citizens the freedom to choose not to participate in schemes of redistribution. But if you believe that some level of redistribution is valuable, it follows that redistribution should generally happen at higher levels of government.
In the past, I’ve endorsed the federalization of Medicaid to address the problem that Michael Greve identifies with joint programs:
Equally pervasive are fiscal federalism cartels: instead of letting states fund programs from own-source revenues, under competitive conditions, Congress collects taxes from citizens across the country and then sends them to state and local governments. Medicaid, education, welfare, and countless others operate on this principle. The programs usually require states to comply with federal funding conditions and to match the federal funds, but they are driven by fiscal incentives. No state would devote 25 percent of its budget to Medicaid if it had to tax its own citizens for the cost. But states are evidently willing to tax themselves for half the cost— because if they don’t, they leave federal dollars on the table. And so every state ends up with a level of spending that no state would adopt on its own.
We might say, well, if jointness is the problem, why don’t we have state governments assume exclusive responsibility for Medicaid, education, and welfare? This is (in part) the idea behind block granting programs, to transition away from jointness. The danger, however, is that some state governments will manage these new responsibilities irresponsibility, and essentially dare the federal government to not bail them out in the face of looming fiscal disaster. So we’re faced with a difficult question: in which domains are state governments likely to engage in a race to the top and in which domains are they likely to engage in a race to the bottom? My (subjective) sense is that most policy domains are likely to benefit from competition among the states, but that Medicaid and (possibly) welfare are domains in which the federal government should take the lead. Some will object that this will leave the states with only modest responsibilities. I disagree, particularly since I envision that the locus of responsibility for infrastructure, education, and labor market regulation should shift almost entirely to the states.
To return to Ezra’s post, he references a paper by Shaefer and Edin. I believe he’s referring to “Extreme Poverty in the United States, 1996 to 2011,” which concludes as follows:
In sum, we estimate that, as of the beginning of 2011, about 1.46 million U.S. households with about 2.8 million children were surviving on $2 or less in income per person per day in a given month. This constitutes almost 20 percent of all non- elderly households with children living in poverty. About 866,000 households appear to live in extreme poverty across a full calendar quarter. The prevalence of extreme poverty rose sharply between 1996 and 2011. This growth has been concentrated among those groups that were most affected by the 1996 welfare reform. Despite the presence of a substantial in- kind safety net, a significant number of households with children continue to slip through the cracks. And it is unclear how households with no cash income—either from work, government programs, assets, friends, family members, or informal sources—are getting by even if they do manage to claim some form of in-kind benefit.
While the best source of data available for this study, the SIPP does likely suffer from some under-reporting of income by respondents. However, under-reporting likely does not explain the dramatic increase in extreme poverty over our study period. Further, under-reporting of income itself suggests adverse outcomes, such as engagement in the underground economy (Edin & Lein, 1997). Finally, when SIPP calendar weights become available for the full study period, adding annual estimates of extreme poverty will be an important addition to this analysis.
When we consider SNAP benefits as equivalent to dollars, this reduces the number of extremely poor households with children by about half. We estimate that SNAP currently saves roughly 1.4 million children from extreme poverty. In addition, many of the households in extreme poverty are accessing public health insurance for at least one of their children, and about one in five have a housing subsidy. These in-kind safety-net programs are playing a vital role, and are probably blunting some of the hardship that American children living in extreme poverty would otherwise face. However, it would be wrong to conclude that the U.S. safety net is strong, or even adequate, when one in five poor households with children are living without meaningful cash income. [Emphasis added]
This is an important and sobering report. Yet it is important to note that it focuses on cash income and that engagement in the underground economy often reflects high implicit marginal tax rates, as Edelman, Holzer, and Offner argued in Reconnecting Disadvantaged Young Men. That is, participation in the underground economy is not necessarily exogenous; survey respondents might choose to underreport income, in the (mistaken) belief that they might endanger a source of income that is not counted against income limits for access to various transfers.
I also recommend reading the work of Bruce Meyer. In a recent paper, Meyer and Robert M. George examine underreporting, including underreporting in the SIPP (which, I should stress, Shaefer and Edin acknowledge as a potential problem):
A number of studies have documented significant underreporting of food stamps in large national surveys such as the CPS or the Survey of Income and Program Participation (SIPP). Several studies estimate underreporting by using administrative microdata that is directly linked to survey data. In perhaps the most comprehensive of these matching studies, Marquis and Moore (1990) show that 23 percent of survey respondents in four states, who were food stamps recipients according to administrative microdata, failed to report participation in the 1984 SIPP. Using a subset of these data, Bollinger and David (1997) find a nonreporting rate of 12 percent. Bollinger and David also conclude that higher income recipients are more likely and female recipients are less likely to fail to report receipt. Taeuber et al. (2004) examine FSP administrative records in Maryland linked to the national 2001 Supplementary Survey (American Community Survey), finding that about 40 percent of recipients do not report receipt.
The main limitation to direct matching of survey and administrative microdata at the individual or household level is that such matches are rarely available, and when these matched data are available, it is typically only for a short time period and for a small subset of the survey respondents, such as a single state. A second approach compares reported receipt in a survey (weighted to population totals) to administrative reports of the number of recipients served or dollars distributed. Studies that use this approach also find evidence of substantial underreporting. For example, Primus et al. (1999) compare weighted food stamp dollars reported by households in the CPS Annual Demographic File (ADF) to administrative numbers. They find that the underreporting rate increased from 24 percent in 1990 to 37 percent in 1997. Bitler, Currie, and Scholz (2003) estimate food stamp underreporting rates between 1995 and 1999 of about 14 percent in the CPS Food Security Supplement and about 11 percent in the SIPP. Cody and Tuttle (2002) calculate underreporting rates for the CPS ADF that range from about 21 percent in 1991 to 36 percent in 1999.
Meyer, Mok, and Sullivan (2009) document the degree of underreporting of food stamps in several major household surveys by comparing the weighted total of reported food stamps dollars or months received in household surveys with totals made available by the U.S. Department of Agriculture, Food and Nutrition Services. A time-series for these dollar reporting rates for the CPS, the SIPP, and the Consumer Expenditure (CE) Survey is reported in Figure 1. Month reporting rates for the CPS and SIPP can be found in Figure 2. Figures 1 and 2 show that food stamps are significantly under-reported in each of these surveys. The dollar and month reporting rates are remarkably similar, suggesting that most of the underreporting is due to understating the number of months of receipt rather than dollars conditional on reporting receipt. There is other evidence that finds that monthly amounts are actually quite close to the true average for several programs and datasets. Previous research indicates that about two-thirds of the underreporting of food stamps months in surveys results from failure to report receipt at all (Moore, Marquis and Bogen, 1996).
As well as being significantly below one, the reporting rates have tended to fall over time. As shown in Figure 2, between 1987 and 2006, reporting rates for food stamp months fell in the CPS from 0.73 to 0.53. The SIPP typically has the highest reporting rate for the FSP program, and these have fluctuated but not steadily declined over time. Thus, past work suggests substantial error. However, our new evidence confirms that data quality has declined in recent years. [Emphasis added]
None of this suggests that Shaefer and Edin are wrong to suggest that the low levels of cash income among the very poor are a serious problem. Like Edward Glaeser, I tend to think that cash transfers are preferable to in-kind transfers — he makes the case particularly well:
In 1968, the last year of Lyndon Johnson’s presidency, the federal government spent $1.61 billion ($10.5 billion in 2012 dollars) on Aid to Families with Dependent Children (the predecessor of Temporary Aid to Needy Families); it spent $1.81 billion ($11.8 billion in 2012 dollars) on Medicaid and $505 million ($3.3 billion in 2012 dollars) for food and nutrition assistance. There was no Earned Income Tax Credit or housing vouchers, so the ratio of in-kind aid to cash transfer was 3 to 2.
The 2013 budget contains $293 billion for Medicaid, $112 billion for food and nutrition service (food stamps) and $28 billion for tenant- and project-based rental assistance, which includes housing vouchers. That is a total of $433 billion of in-kind transfers from these three primary programs. By contrast, the budget includes only $17 billion for the Administration for Children and Families (which administers Temporary Aid to Needy Families). The Earned Income Tax Credit paid out $59.5 billion in 2010, and Obama’s proposal would eventually increase its generosity by about $1.5 billion a year. Considering just these programs, the ratio of in-kind assistance to cash aid is now to 5.6 to 1. In 1968, the in-kind share of assistance was 60 percent; now it is 85 percent.
In Glaeser’s view, the biggest downside of this reliance on in-kind transfers is that it leads poor people to spend money on things that don’t value; and these programs are fragmented, poorly-coordinated, and plagued by work disincentives. To that end, he favors a different approach:
There is a natural solution: Combine our disparate aid efforts into a single program that delivers cash assistance and minimizes perverse incentives. We may still want some in-kind assistance, particularly for health care, and that could be handled by issuing vouchers. But for most aid, the Friedman solution of cash still seems right. By combining our aid programs and primarily giving cash, we can have a more efficient welfare system that provides more freedom and better incentives for aid recipients.
To return to an earlier theme, might it make sense for a combined aid program to be a federal program? State governments could choose to supplement said federal program, but the federal program wouldn’t be a joint program that imposes mandates on state governments. Rather, state governments would be free to design their social programs as they see fit.