"If the goal was to get rid of poverty, we failed": the legacy of the 1996 welfare reform

Presidents are usually effusive, grandiose, and triumphant when they sign major legislation that will form a huge part of their legacy. In 1996, Bill Clinton’s announcement that he’d sign a bill ending "welfare as we know it" was not that.

It was defensive and at times openly apologetic for what was about to happen.

"Some parts of this bill still go too far," he conceded. "This bill still cuts deeper than it should in nutritional assistance, mostly for working families with children."

He probably didn’t know then how true those words would one day turn out to be. Clinton, and now his wife, presumptive Democratic nominee Hillary Clinton, has run hot and cold over the years on the Personal Responsibility and Work Opportunity Act (PRWORA) — what is now colloquially referred to as "welfare reform."

Ten years after signing the bill, Bill took an enthusiastic victory lap in the pages of the New York Times. "Welfare reform has proved a great success," he declared. And at first, it seemed like he was right. Evidence seemed to indicate that it really did increase employment, and worst-case scenarios liberal critics predicted didn’t come to pass.

But another 10 years later, the mood changed again. Days before the New York Democratic primary, Hillary Clinton, who once urged her husband to sign the bill, told WNYC that "we have to take a hard look at it again." New research had showed that in fact welfare reform fell short in the depths of the Great Recession — substantially increasing deep poverty and leaving families who can’t find work without any cash safety net.

The Clintons’ statements have mirrored the overall reputation of the bill over time, especially among Democrats. In 1996 the party was all but unanimous about the fact that something had to change with welfare. There was heated debate over whether the changes were acceptable, including within the Clinton administration itself, but ultimately most congressional Democrats voted for the law on the same mixed grounds that Clinton signed it: It wasn’t perfect, but it was a necessary step toward reform.

Now it is a cross Hillary has had to bear during her presidential run, and a topic on which Bernie Sanders and his left-wing intellectual supporters could, and did, attack her mercilessly.

The idea behind welfare reform back in the '90s was that something was deeply broken about the system and needed to change. That was correct. But something else is deeply broken now, and it too needs to change.

Keeping people out of work was a feature, not a bug, of the original welfare program

Farm Security Administration/Office of War Information Collection, Library of Congress

What we talk about when we talk about welfare is the Aid to Dependent Children (ADC) program, enacted as part of the Social Security Act of 1935 in the midst of the Great Depression.

This program was initially an outgrowth of a type of program that cropped up in a number of states, known as Mothers' Pensions, in the 1910s. The programs were meant to help single mothers, but only the small fraction deemed deserving.

"That typically excluded divorced mothers and those with children born outside marriage," New York Times reporter Jason DeParle writes in his book American Dream: Three Women, Ten Kids, and a Nation's Drive to End Welfare, which informs much of the history in this article. "And it almost always excluded racial minorities." Indeed, the recipients of Mothers' Pensions were 96 percent white.

ADC too initially excluded black mothers, as did Social Security's flagship old-age pension program. It built on the Mothers' Pension system by providing funding for states to pursue their own programs, with states setting benefit levels and many of the rules. Gradually, federal oversight increased, first with a push in the 1940s and '50s by federal authorities to end racial discrimination.

By 1960, DeParle notes, the program was 40 percent black, and while the program initially targeted widows to the exclusion of divorcées and unwed mothers, almost two-thirds of recipients by that year fell in the latter category. ADC was renamed the Aid to Families With Dependent Children (AFDC) program in 1962.

The purpose of these programs, from the beginning, was to deter women from work. After all, it was originally designed for white widows, a population deemed deserving of aid and one expected to stay home and raise children rather than enter the workforce.

But work by black mothers was more common, and more expected, especially in Southern states that relied on black women's labor as maids, nannies, and agricultural workers. That fed an expectation that black women should be working even if they had children, which in turn fed an anti-welfare backlash.

The backlash was worsened by rapid increases in the rolls, beginning as racial discrimination eased in the 1950s. It was also aided by the rise of the "welfare rights movement" in the 1960s, which organized to enroll poor people then cut off from the program and won important legal victories like King v. Smith, which established AFDC as an entitlement that all eligible people must be allowed to receive:

Fang and Keane 2004

The two most vocal leaders of the welfare rights movement — sociologists Frances Fox Piven and Richard Cloward — described their goal as overwhelming the system until the government was forced to enact a guaranteed income, a policy that then gaining traction both with economists and with activists like Martin Luther King Jr.

They almost got their wish: Richard Nixon proposed a welfare reform package called the Family Assistance Plan that would have included a guaranteed income — albeit one that recipients would have to meet work requirements to get. The bill passed the House before dying due to both liberal opposition (because it was too stingy) and conservative opposition (because it was a guaranteed income).

Contrary to popular belief, Lyndon B. Johnson’s war on poverty did not, Columbia professor Jane Waldfogel writes, do much to expand AFDC. Johnson hated the idea of poverty reduction through handouts rather than work, and resisted calls from his adviser Sargent Shriver to implement a guaranteed income.

Old welfare really did keep people out of work — and those on it wanted to work

Thomas J. O'Halloran

The anti-welfare, pro–work requirement movement really came into its own in the 1980s. Ronald Reagan had campaigned against the program, railing against "welfare queens" living large and driving Cadillacs, and slashed AFDC by about a sixth upon taking office.

He was buffeted by a growing community of conservative think tanks, and most influentially by political scientist Charles Murray, who first made his name with his 1984 book Losing Ground, which argued that abolishing welfare entirely would actually reduce poverty.

Reputable social scientists dismissed the book, but its racially charged conclusions struck a chord. In his book proposal for Losing Ground, Murray wrote, "Why can a publisher sell it? Because a huge number of well-meaning whites fear that they are closet racists, and this book tells them they are not. It's going to make them feel better about things they already think but do not know how to say." He wasn’t wrong.

With enemies like that, many on the left were understandably suspicious of attacks on welfare. But the program really did have deep problems. The three most common criticisms made of AFDC were:

1. It caused poor adults who could work to not work.
2. It caused dependency; rather than using it as a temporary safety net, some people embraced it as a way of life.
3. It encouraged having children out of wedlock and discouraged marriage.

The first of these claims was definitely true, the second was kind of true, and the last might have been true but the effect was very, very small.

Very few social policy analysts would contest that AFDC discouraged work. For one thing, it often featured a 100 percent phaseout rate: Each dollar you earned meant one less dollar in benefits.

"I'd always talk about it in intro econ classes," Hilary Hoynes, an economist at UC Berkeley, says. "Students come in thinking welfare recipients are lazy, and then you talk them through the basics of the economics and describe the incentives of the program, and people were like, 'Of course, why would you work in this program?'"

Ironically, one thing Reagan did upon taking office was increase the phaseout rate, therefore making AFDC an even worse work deterrent than it had been before. "Reagan's cuts cut all the working people off and made it a program for people who weren't working," the Center on Budget and Policy Priorities' Liz Schott explains.

Russell Safe Foundation

Ethnographic work by Kathryn Edin (now at Johns Hopkins) and Laura Lein (now at University of Michigan) in the early 1990s suggested that AFDC both provided inadequate benefits for the poor, leaving them mired in poverty, and provided significant work disincentives. The phaseout rate, they concluded, was a major deterrent.

"Most of the welfare-reliant mothers we interviewed had an accurate view of the benefits they would lose by going to work," Edin and Lein write in their book Making Ends Meet. "They made reasonable assessments of how much they would need to earn to offset the added costs of work."

The families they spoke with concluded that low-wage jobs would at best leave them no better off than they were with welfare, that it might leave them worse off if they lost their job and were forced to reapply for welfare, and that the jobs they could realistically get would not prepare them for better, higher-paying work in the future.

But because of the stinginess of AFDC, Edin and Lein found that welfare mothers still needed to find other sources of income. Only about 64 percent of their income came from AFDC, food stamps, and other safety net programs. Twelve percent came from off-the-books work (which wouldn't reduce welfare benefits) and underground work (selling sex, drugs, etc.), and 17 percent came from friends and family members. Only 2 percent came from formal work.

Some economists I talked to were more skeptical that the phaseout rates were the major problem — but still agreed that AFDC discouraged work. "Congress lowered the phaseout rate from 100 percent to 67 percent in 1967 and work levels among AFDC recipients went up only modestly," Robert Moffitt at Johns Hopkins notes. The bigger work disincentive, he says, was the mere fact that people were getting money with no strings attached.

Even though few welfare recipients were formally working, research indicated they actually wanted to be working. "Research indicated, including research that we conducted in Chicago, that AFDC recipients preferred work to welfare and would readily accept jobs that will not result in slipping deeper into poverty," according to William Julius Wilson, a professor and sociologist at Harvard University.

What about the claim that AFDC caused dependency? The most influential work on that question was done by Harvard’s Mary Jo Bane and David Ellwood in the 1980s. They found that while most people who entered AFDC left fairly quickly, a minority stayed for long periods, averaging eight years. Though a minority of those entering, because of its longer duration this group made up the majority of the AFDC caseload at any given time.

Moreover, plenty of research has documented that there was a correlation between mothers receiving AFDC and their daughters receiving it later; the literature is split on whether this was causal (the mom getting welfare actually made the daughter more likely to take it up) or whether it was wholly explained by the fact that children of poor parents tend to be poor too.

The best, most recent research I’ve seen on this question, from the University of Kentucky’s James Ziliak, Robert Paul Hartley, and Carlos Lamarche, finds that at least some of the relationship was causal.

Reasonable people can disagree about how bad it was that a majority of AFDC recipients at any given time were in the midst of a years-long spell. After all, that was the original intent of Mothers’ Pensions: that the money could support mothers indefinitely, that they’d never need to work. But it’s hard to deny that the program did lead to millions of people living off welfare for years at a time, without working.

As for the claim that welfare discouraged traditional families, the evidence there is shakiest. Christopher Jencks, a professor of social policy at Harvard's Kennedy School of Government, tells me the argument that welfare discouraged marriage "was probably a little true, but not very." The effect sizes just weren't big enough.

And the phenomenon seemed to be largely about women preferring to rely on their welfare checks than on the earnings of their male partner — it was about independence as much as anything.

Paul J. Richards/AFP/Getty Images

As the welfare backlash grew in the 1980s, the federal government was beginning to issue waivers to states to allow them to experiment with various welfare-to-work approaches. A number of governors, notably Gov. Tommy Thompson (R-WI), seized the opportunity, experimenting with work supports, time limits, work requirements, and other approaches that would inspire the latter federal reforms.

The waiver programs were often run as actual experiments, and the results were encouraging. "Out of 11 state programs studied, nine raised employment and earnings, albeit modestly," DeParle recalls.

For example, the organization MDRC produced a hugely influential study in 1993 of a welfare-to-work program in Riverside County, California, which found that the Riverside approach (which emphasized getting a job above all else, including education and training) raised earnings more than 50 percent. That fueled a burgeoning consensus that liberals should accede to work requirements and accept some form of "workfare."

(Subsequent research suggested the Riverside finding was a fluke, the product of an already stronger local economy rather than any reforms.)

The liberal pro-reform line was that assisting the poor was worthwhile but that AFDC had unacceptable work disincentives, and that a public jobs program would be superior. Mickey Kaus at the New Republic was the most vocal journalistic exponent of this view, while Ellwood's 1988 book Poor Support laid out the most sophisticated, detailed plan for achieving a liberal version of reform, including greatly expanded health benefits, child care, and guaranteed jobs, alongside time limits on receipt of actual cash welfare.

Arkansas Gov. Bill Clinton was strongly associated with this approach, putting together "the Clinton plan" through his role as chair of the welfare task force at the National Governors Association. The plan sought $1 billion to $2 billion more a year in spending, in exchange for tougher work requirements. Every governor in the country save one endorsed it.

The plan eventually morphed into the Family Support Act, which provided additional matching funds in exchange for sending a certain fraction of the recipients to work programs. Reagan signed it into law in 1988, but it didn't stop the welfare rolls from continuing to grow.

How Democrats came around to the idea of reforming welfare — and why Bill Clinton signed it

"Welfare should be a second chance, not a way of life. In a Clinton administration, we're going to put an end to welfare as we know it," said then-Gov. Clinton in October 1991.

From the get-go, he was passionate about expanding time limits and work requirements and trying to dramatically reduce the welfare rolls.

Still, a number of liberal welfare experts who joined his administration, including Georgetown law professor Peter Edelman, the Kennedy School's Bane, and Ellwood, would be tasked with developing the Clinton welfare plan, alongside the more vociferously anti-welfare aide and speechwriter Bruce Reed.

Almost immediately, it became clear that the idealized grand bargain, in which every welfare recipient would get a subsidized job if needed, along with child care and transportation assistance and everything else they might need to thrive in the workforce, wasn't going to fly. The costs were too much for Congress to stomach, even more so after Republicans regained power in 1995.

"By just mailing checks, the government spent an average of about $5,000 per family each year," DeParle notes. "A work slot (with child care for just one child) would cost about $11,700. The bill for 2 million of them would raise welfare's annual costs by more than $13 billion, nearly 50 percent. Nothing like that seemed remotely possible."

Congressional Republicans began putting together proposals that took the time limits that Ellwood had wanted to pair with very generous government support, and just imposed them without any of that help on the side. In the official House GOP plan, states could totally cut off recipients after three years in a work program.

Under a conservative alternative known as the Real Welfare Reform Act, women under 26 with children out of marriage would be stripped of all AFDC, food stamps, and housing assistance. Just like that.

And, DeParle reports, the Clinton administration task force charged with building out a plan was growing more supportive of hard time limits, much to Ellwood's concern. An idea he saw as part of a comprehensive expansion of the safety net was being transformed beyond recognition.

Eventually, Clinton's team put out a plan in June 1994. It limited cash benefits to two years and required participants to work after that; if they couldn't find work, a government job would be provided. To keep costs down, it only applied to people born after 1971: those 23 and under, at the time.

Richard Ellis / AFP / Getty

Once the GOP was back in control of the House and Senate, Speaker Newt Gingrich quickly settled on the idea of block-granting AFDC. Instead of the federal government matching state spending, the feds would grant a set amount of cash every year, which states could then use as they liked to implement a welfare programs.

States would be given huge amounts of discretion as to whether to use those funds for job training, child care, etc. This allowed Congress to avoid getting too specific and gelled nicely with the Republicans' general support of localized governance.

Initially, the Republican plan didn't appear to have much in the way of work requirements — something Democrats, of all people, attacked them for, calling them "weak on work." To get around that, the bill got work requirements but also something known as a "caseload reduction credit," which let states count people who left welfare as working whether they really had a job or not. Just by cutting the rolls, states could meet the requirement.

The Democratic opposition continued to buy into Republican policy ideas, with House Minority Leader Dick Gephardt proposing a four-year cold turkey time limit for aid and Senate Minority Leader Tom Daschle proposing five years. And Bill Clinton was sounding sympathetic to even Republican proposals.

Wendell Primus, a deeply respected appointee to the Department of Health and Human Services who vociferously opposed to cuts to welfare, produced an estimate suggesting the Republicans' Senate bill would push 1.1 million people into poverty. Clinton saw the study, but nonetheless privately suggested he'd sign such a bill anyway.

He got temporarily saved from himself by Gingrich's own blunders. In November and December 1995, Gingrich packaged welfare reform together with a broader balanced budget package that included substantial cuts to Medicare, proposed in the midst of two government shutdowns over broader spending-level issues. When Clinton vetoed the spending cuts, he vetoed the attached welfare reform. In January, he vetoed it on its own.

Humiliated, Gingrich initially didn't want to try for reform again, DeParle reports, until Jimmy Hayes, a House member who switched in 1996 from the Democrats to Republicans, told him that forcing Clinton to veto or sign a standalone welfare bill would put the Democrats in a deep political bind. If Clinton vetoed, he'd be painted as soft on welfare. If he signed, Democrats would rebel. So in July, the House and Senate passed a standalone bill and sent it to the president.

DeParle reports that the administration was split over whether to sign, but more appeared to support a veto than not. Secretary of Health and Human Services Donna Shalala wanted a veto, Secretary of Labor Robert Reich wanted a veto, Treasury Secretary Robert Rubin wanted a veto, Secretary of Housing and Urban Development Henry Cisneros wanted a veto, Chief of Staff Leon Panetta wanted a veto.

David Ellwood, who had resigned from the administration in 1995, urged a veto. Same for political aides George Stephanopoulos and Harold Ickes. On the pro side were Reed, Al Gore, and Rahm Emanuel, arguing that the president couldn't afford to be seen keeping a failed system going.

Sen. Daniel Patrick Moynihan (D-NY) famously predicted "children sleeping on grates, picked up in the morning frozen."

Clinton signed the bill, in the process renaming AFDC the Temporary Assistance for Needy Families (TANF). In response, HHS officials Primus, Peter Edelman, and Mary Jo Bane resigned in protest. Edelman, a longtime personal friend of the Clintons, wrote a long piece for the Atlantic calling welfare reform the "worst thing Bill Clinton has done."

At first, welfare reform seemed to be working as intended

Clinton and congressional Republicans had picked just about the best possible time to enact a law like this. The late '90s were an era of extreme labor market tightness that has not been seen since. Unemployment dipped below 4 percent at times. Even many of the least-skilled former AFDC recipients could find work.

Add to that the fact that Clinton had pushed through a huge expansion in the earned income tax credit (EITC), which made low-wage work considerably more attractive, and you had a scenario where the aftermath of welfare reform could see poverty actually fall.

And fall it did. Here's how the anchored supplemental poverty measure — a more accurate metric for deprivation than the official poverty measure — changed from 1993 to 2002:

And welfare enrollment fell dramatically:

Congressional Research Service / Gene Falk

What’s more, it appears that the relationship was at least somewhat causal. Welfare reform really did appear to be decreasing caseloads and increasing earnings and employment. Typically, researchers found that the EITC and the growing economy did more, but welfare reform played a role all the same. "Some women who were on welfare really did have enough work skills to get a decent job at wages above the minimum wage, and they are probably better off off welfare," Moffitt says.

For example, in one influential paper the University of Chicago's Jeffrey Grogger estimated that time limits caused about one-eighth of the decline in welfare use and 7 percent of the rise in employment among single mothers from 1993 to 1999 — less than the EITC, but still significant.

More dramatically, in 2004 Hanming Fang (now at Penn) and Michael P. Keane (now at Oxford) estimated that work requirements and time limits explained 68 percent of the decrease in welfare usage and 27 percent of the increase in workforce participation among single mothers. Once again, EITC was a bigger factor in workforce participation, but welfare reform still mattered.

Most impressively, in 2000, University of Wisconsin's Rebecca Blank (later Obama's acting commerce secretary) and the University of Michigan’s Robert Schoeni found a rise in total income and decline in poverty for most affected by the 1996 reforms.

"Welfare reform is now widely viewed as one of the greatest successes of contemporary social policy," Jencks and Scott Winship wrote in a 2004 piece.

But later research cast doubt on welfare reform’s efficacy

Gregory Rec/Portland Press Herald via Getty Images

Asked again earlier this year about that article, Jencks told the Washington Post’s Max Ehrenfreund simply: "I was wrong." In that, he’s joined by most social policy experts I consulted for this piece. While conservatives still defend the 1996 reforms, there’s a broad consensus outside the law’s core supporters that something went deeply wrong, and that the initial assessments were way too optimistic.

For one thing, subsequent research cast doubt on the original pro-reform studies. The best, most recent review of the literature I've seen comes from the University of Kentucky's James Ziliak, a veteran researcher on welfare policy issues. It was finalized in late 2015 and incorporates not just the initial rush of research into welfare reform post-1996 but studies since the Great Recession began as well.

Ziliak agrees with prior research that while you can certainly debate the relative importance of welfare reform compared with the EITC and the business cycle, TANF and its associated rules almost certainly caused a decrease in the welfare rolls and increased employment. He also finds a relative consensus that it increased earnings.

For most family structure–related outcomes — teen births, marriage, etc. — the evidence was too limited and mixed to draw firm conclusions.

But Ziliak also concludes that welfare reform reduced the incomes of the poor, when you take both the increase in earnings and the decline in transfer income into account, and especially if you look at the poorest of the poor. In other words, transitioning from welfare to work may have gotten people jobs, but those people didn't actually come out ahead monetarily. Welfare would've paid better than work did.

Specific studies finding this include one by Ziliak himself, with Kentucky colleagues Christopher Bollinger and Luis Gonzalez, which concludes, "The earnings gains among the low skilled a decade after the implementation of TANF and expansions of the EITC have been more than offset by losses in transfer income and have left the most vulnerable single mothers either running in place or falling behind."

Bitler, Hoynes, and Penn's Jonah Gelbach looked at the Connecticut Jobs First waiver experiment, a pre-TANF reform effort with many similar characteristics to federal welfare reform, and found reductions in total income for participants at the bottom at the income distribution. That study is particularly compelling since it was an actual experiment with random assignment.

Ziliak also points out flaws in some of the earlier research suggesting welfare reform reduced poverty. Blank and Schoeni’s paper, for instance, excluded those reporting no earnings and income, which might have made the overall population appear better off than it really was.

"While the official poverty rate fell among single mother families in the 5 years after reform, this is less in evidence over the long term, especially for broader definitions of income that include in-kind transfers like SNAP [Supplemental Nutrition Assistance Program, also known as food stamps] and tax payments and credits," Ziliak writes in an email. "We saw especially an increase in so-called deep poverty, the fraction living below 50 percent of the poverty line."

"If the goal of welfare reform was to get rid of welfare, we succeeded," the University of Wisconsin’s Timothy Smeeding notes. "If the goal was to get rid of poverty, we failed."

Welfare reform almost certainly increased deep poverty

There's the rub. Deep poverty figures tell a much less rosy story than those for overall poverty. That suggests, as Hopkins's Moffitt has argued, that welfare reform was part of a shift away from aid for the poorest of the poor and toward the highest-earning of the poor: those with jobs, who benefited from higher minimum wages and EITC.

When you look at overall poverty from 1993 to 2012 (again, measured accurately rather than through the official measure), you see that it fell substantially in the 1990s and then stayed down.

That was true even in economic downturns. In 1993, as the nation recovered from the early '90s recession, the poverty rate was nearly 21 percent. In 2012, in the wake of the worst recession in post-WWII history, it spiked up to 16 percent, a huge improvement.

The same can't be said of deep poverty. That fell in the '90s too — but by a comparatively tiny amount. And by 2012, it was back fairly close to its 1993 level:

Tellingly, the deep poverty rate in 1996, when welfare reform passed, was 4.6 percent. Since the end of the late-'90s boom in 2001, the rate has never been that low again.

If you try to isolate the effects of welfare reform, it appears that if anything it probably increased deep poverty in the US. The most disturbing evidence in this regard comes courtesy of the University of Michigan's Luke Shaefer and Johns Hopkins's Kathryn Edin, who have documented an increase in the share of Americans living on $2 a day or less in cash income.

Using data from the Survey on Income and Program Participation (SIPP), they found that the share of households with less than $2 per day, per person, shot up from 1996 to 2011, from 1.7 percent of households with children to 4.3 percent. That's a 153 percent increase.

Shaefer and Edin, 2013

The growth is much smaller if you throw food stamps, tax credits, and housing subsidies into the mix, but it's still an increase of more than 45 percent: from 1.1 percent of households to 1.6 percent. That just underscores Edin and Shaefer's main point, which is that more and more families are being forced to get by without a reliable source of cash income.

And cash matters. You can't pay the rent with food stamps. You can't buy clothing for your children, or refill a subway card, or pay the car bill, or refill your gas tank either. You can't eat housing subsidies (and very few of the poor get them, in any case).

Shaefer and Edin are clear that they view this development as, in large part, a result of welfare reform. "The percentage growth in extreme poverty over our study period was greatest among vulnerable groups who were most likely to be impacted by the 1996 welfare reform," they note. Households headed by single women saw a larger increase in extreme poverty. Households with children (the only ones eligible for AFDC) saw an increase more than twice as large as the one households without children experienced.

And TANF did much less to reduce extreme poverty than AFDC did. In 1996, AFDC brought 1.15 million households above the $2-a-day threshold. In 2011, it was only bringing about 291,000 above it — and this was after more than a decade of population growth.

This conclusion is backed up by other data Shaefer and Edin present. Food stamp recipients must verify their income to the program, and misreporting can result in criminal charges, increasing the data's credibility. In 1996, 316,000 SNAP applicants reported having no income. In 2012, 1.2 million did. "Our SIPP estimates fall behind SNAP administrative data estimates as of 2012," Shaefer and Edin write, "suggesting that our SIPP estimates are on the low side."

Shaefer and Edin 2015

Shaefer and Edin also did extensive fieldwork talking to families meeting the $2-a-day threshold, reporting their findings in the book $2 a Day. They found, again and again, that the effective end of cash welfare was forcing families into desperate measures to provide for themselves.

Many people they talked to would sell plasma, one of the few ways to make quick cash when work can't be found. Others would sell their food stamps for cash — a rare practice generally, but common among the extremely poor. The one thing they couldn't do is rely on cash welfare.

"TANF is virtually dead in all of these places. It's absolutely striking that every one of our families is categorically eligible for TANF and none of them are receiving it," Edin told me. "Where it's really dead is in the imaginations and thought processes of the poor. This is not seen as a fallback. In most cases, it doesn't occur to people to apply."

The $2-a-day research has provoked a fair share of pushback from critics, but none of their arguments really refute the central conclusions. The most common argument is that SIPP underreports income, exaggerating the rate of extreme poverty.

But SIPP, Edin and Shaefer argue, suffers less from this problem than other surveys, and the fact that SNAP's data tracks SIPP so well is further evidence that something real is going on here.

Moreover, for this to explain the rise in extreme poverty, underreporting would have to increase over time — and between 1996 and 2005, it actually fell for SIPP.

In a review of Edin and Shaefer's book, Christopher Jencks from Harvard presents still more evidence of this increase in deep poverty and deterioration in living standards for the poorest of the poor. He charts how the economic resources of Americans at the 50th, 10th, 5th, and 2nd percentile of the income distribution have changed from 1967 to the present:

New York Review of Books / Christopher Jencks

While people at the middle, and even the middle poor, saw their resource levels stagnate from 1999 onward, they plummeted for Americans at the 2nd percentile. This, he concludes, "support Edin and Shaefer’s claim that the poorest of the poor were a lot worse off in 2012 than in either 1996 or 1999."

Other analyses of SIPP have come to similar conclusions as Edin and Shaefer. A 2011 paper by Hopkins's Moffitt, Mathematica's Yonatan Ben-Shalom, and University of Wisconsin Madison's John Karl Scholz estimated changes in post-transfer deep poverty. They found that 4.5 percent of families were under half of the poverty line in 1993, but 6.6 percent were in 2004, a nearly 50 percent increase.

Edin and Shaefer were also building upon a rather large literature showing an increase in the number of "disconnected mothers": single moms who were neither receiving cash benefits nor in the workforce. The Urban Institute's Pamela Loprest and Austin Nichols estimated that while in 1996, one in eight low-income single mothers were disconnected — having no earnings and no TANF or Supplemental Security Income benefits, and not in school — one in five were in the years from 2004 to 2008:

Loprest 2012

Similarly, Wisconsin's Blank and Michigan's Brian Kovak (now at Carnegie Mellon) find that the share of single mothers with no earnings or welfare not in school doubled from 10 to 20 percent from 1990 to 2005. If you include women with very low earnings and no Social Security Insurance (SSI) income, the rate goes from 12 to 22 percent. Many of these families still received SNAP benefits, but that’s about it.

The bottom line is that a large and growing literature finds, consistently, that deep poverty defined in a variety of ways increased after the introduction of welfare reform. The increases are particularly striking among single mothers, the main group benefiting from AFDC.

It's hard to interpret this evidence as saying anything other than that welfare reform decreased living standards for the most vulnerable members of American society.

Block granting: where even conservatives think reform fell short

Welfare reform has also made welfare less responsive to recessions. This is a disturbing development given the central role that safety net programs play in assisting families and boosting the economy when unemployment rises and work is hard to come by. Because programs like food stamps are entitlements, they are guaranteed to everyone who qualifies for benefits.

As incomes fall in recessions, more people qualify and automatically start getting assistance. That's a form of automatic stimulus that is vital in helping counteract downturns and helping families cope in the meantime.

AFDC was a program like that. As an entitlement, it took all eligible people, and as the number of people qualifying rose in recessions, its ranks swelled. It had a lot of problems, but it did help counteract bad business cycles.

TANF, by contrast, is a block grant. It gives a set amount of money to states, every year, regardless of how many people are eligible for benefits. That means its ranks don't swell in recessions:

Center on Budget and Policy Priorities

This means that recessions now cause a greater increase in deep poverty than occurred before welfare reform. Two recent papers by UC Davis's Marianne Bitler and UC Berkeley's Hilary Hoynes broke down how various programs have affected the "cyclicality" of poverty during the Great Recession: that is, the degree to which poverty changes in reaction to changes in unemployment. For instance, SNAP (food stamps) reduced cyclicality by giving unemployed people boosts in spending power and helping them get out of poverty.

But TANF, unique among the programs analyzed, increased cyclicality, especially for those in deep poverty. It made the deeply poor more vulnerable to changes in the business cycle. "Those at the very bottom," Hoynes says. "That's the safety net that's gone."

Harry Holzer, a labor economist at Georgetown who's widely cited on low-wage work issues, notes that "welfare reform was based on a strong assumption that almost all of the poor could get jobs. … That model really didn't work well in the Great Recession."

If that were the only problem with the block grant structure, that'd be bad enough. But the problems run much deeper than that. For one thing, the actual size of the block grant, in inflation-adjusted terms, has declined dramatically. Since 1997, the federal contribution has been frozen at $16.5 billion. But $16.5 billion in 1997 was worth a lot more than $16.5 billion is worth today. The Congressional Research Service finds that inflation has eroded a third of the value of the block grant:

By contrast, from 1997 to 2013, EITC and child tax credit payments grew by more than 50 percent. That's what's supposed to happen as the economy grows. The erosion of TANF money is legitimately an outlier, unlike what's happening to any other major safety net program.

It gets worse. Welfare reform didn't just turn AFDC into a block-granted program; it also gave states huge amount of flexibility in how to use that money. And because there's little in the way of incentives for states to use it for actual cash assistance, or even work programs, it's being plundered for use in barely related pursuits, like administration of the child protection system.

In 2014, just 26 percent of TANF spending went to "basic assistance" — cash welfare — and another 24 percent went to work programs and child care, according to a Center on Budget and Policy Priorities analysis. A third went to activities well outside the intended function of welfare reform.

For example, Michigan has used the money for college scholarships, and Louisiana has used it to fund anti-abortion crisis pregnancy centers.

Center on Budget and Policy Priorities

The availability of the money as a kind of slush fund for states — if only they don't use it on actual welfare — additionally creates an incentive for states to discourage potential beneficiaries from applying. In Georgia, applicants received flyers with slogans like "TANF is not good enough for any family" and "We believe welfare is not the best option for your family." Applicants were rejected for missing appointments or for filing fewer than 24 job applications in a week.

Rae McCormack, one of the people living on less than $2 a day profiled by Edin and Shaefer, reported being told by a caseworker, "We don't have enough to go around for everyone. Come back next year" — even though caseloads in Ohio are very low.

"You set up a system that incentivizes welfare for states, not people," Shaefer told me. "States can keep their caseloads low and redirect the money to what they would've spent on otherwise."

This has prompted a backlash among even many conservatives. Peter Germanis, a veteran of the welfare reform battles from his time at the Heritage Foundation and the Reagan White House, has become an outspoken critic of TANF because of the perverse incentives created by the block grant.

"When it comes to the TANF legislation," he writes, "Congress got virtually every technical detail wrong. … Congress gave states too much flexibility and they have used it to create a giant slush fund."

Other conservatives have told me they agree. Lawrence Mead, a political scientist at NYU and one of the intellectual godfathers of welfare reform, still considers TANF a success but finds the Germanis critique compelling.

"There are clear-cut abuses and problems in TANF regarding its implementation," he says. "The problems are clear, and the three that stand out are the failure to allow people to apply for aid; the atrophy of the work programs; and the diversion of funds to other programs. Those were not intended in TANF, and they should be stopped. We should go back to a program that does provide aid to the needy, even if it does require work."

Robert Doar, now at the American Enterprise Institute and formerly commissioner of welfare programs for Michael Bloomberg in New York, agrees. "If you were going to tweak TANF going forward, tightening up the restrictions on the spending of TANF dollars so that states were encouraged to spend them on TANF families or TANF-like families, I'd support that," he says. But he argues that the matching-fund structure of AFDC, where states that spent more got more federal money in turn, created the opposite bad incentive, toward expanding caseloads.

Brookings's Ron Haskins, who helped draft the welfare reform legislation as a House committee staffer and is still a defender of it, nonetheless concedes that the block grant went awry, and it makes him wary of plans like House Speaker Paul Ryan's to block grant food stamps, Medicaid, and the rest of the federal safety net as well.

"The TANF experience has got to give you pause," he says. "If you give states flexibility, they're going to use it, and they're going to use it in ways you didn't anticipate."

How Democrats have turned on welfare reform, and where we go from here

These critiques of the law attracted little notice outside of academia and think tanks until the 2016 primaries. But because Bernie Sanders opposed welfare reform while Hillary Clinton claimed to have rounded up votes for it, the new skeptical consensus became a significant liability for Clinton.

Sanders himself cited the Edin/Shaefer research, saying, "Since [welfare reform] was signed into law, the number of families living in extreme poverty has more than doubled."

"The federal safety net for poor families was torn to shreds by the Clinton administration in its effort to ‘end welfare as we know it,’" Michelle Alexander wrote in her essay in the Nation "Why Hillary Clinton Does Not Deserve the Black Vote."

A recent dispute between Sanders-sympathetic Demos researcher Matt Bruenig and Clinton ally and Center for American Progress head Neera Tanden, which ended with Bruenig getting dismissed from his position, centered on his allegations that Tanden supported welfare reform.

Tellingly, Tanden did not defend welfare reform. She denied ever supporting it — just like Clinton herself declined to defend the law's record to WNYC, rather trying to explain how it failed.

"I would put most of the responsibility on the Bush administration and on governors and on the failure to be able to get some of what was tried to have more modulation when there were downturns in the economy," she said.

At the same time, the leftward Democratic Party movement on welfare reform has not led to calls to return to AFDC. The 100 percent phaseout rates, combined with sub–poverty level benefits, are not good policy. And programs that are widely hated by the public are not good long-term policy.

There's widespread consensus, including by experts on the left, that resurrecting the program is not an option. "AFDC was a deeply flawed program, and it was particularly deeply flawed because it was so disliked across the board," Shaefer says. "No one sensible is saying that we should go back to the old system," Harry Holzer adds.

So what do we do instead?

Almost everyone across the political spectrum agrees that something has likely gone wrong with the way the TANF block grant is handled. So the simplest way to address welfare reform's failures would be to try to reform TANF to function more effectively.

The Obama administration has a proposal to do just that in its latest budget, requiring that at least 55 percent of state and federal funds go to work activities, child care, and cash assistance. Obama would also create a permanent TANF fund for recessions, to help make the program better at fighting downturns, and would, for the first time ever, increase the size of the TANF block grant.

Doar has an even simpler proposal: He'd require states to examine all cases of food stamp recipients reporting no income and work to enroll them in TANF. That, he argues, would effectively tackle cases of extreme poverty.

Most left-of-center experts I spoke to, however, wanted to go further. Their answers generally involved the creation of one of two new programs for low-income Americans. The first is a child allowance. The second is a subsidized jobs program, like the one David Ellwood originally wanted as part of welfare reform.

A child allowance is a universal benefit paid out equally to all families with children. It's a very common policy across the developed world, with countries like Germany, France, Sweden, and Ireland all boasting versions. In a US context, that would provide a source of cash to families with children who lost one when AFDC went away — but because it's universal, it would likely avoid becoming as politically toxic as AFDC did.

The Center for American Progress endorses a version of this through its proposal to make the $1,000 child tax credit fully refundable and add a $125-a-month ($1,500-a-year) young child tax credit, also fully refundable. Rep. Rosa DeLauro (D-CT), along with Nancy Pelosi and Democrats’ ranking member on Ways and Means, Sander Levin, has introduced legislation creating the latter.

Wisconsin’s Timothy Smeeding, Jane Waldfogel at Columbia, Shaefer, and a number of other poverty researchers are putting together an even more ambitious proposal that would pay out $250 a month per child under 6 and $200 a month per kid ages 6 to 18.

"There's still an incentive to work," Smeeding argues. "We're talking about something that's modest and that's low. … It's not a lot of money, but it does provide a reliable steady floor."

The main argument against that plan would be that it does hurt the incentive to work. That's what's implied by Moffitt's argument that the income effect — the raw increase in income generated by welfare — was the primary reason AFDC reduced work. "You can't give people money without causing them to work less," Mead insists.

Hoynes, though slightly skeptical of a child allowance on cost grounds, disputes this. "Poor folks' income effect responses are not very large, because they're really poor, and they need the money," she argues. That suggests a child allowance wouldn't provide much of a work disincentive.

The other main category of solution involves subsidized or public-provided jobs. This was part of the original liberal image of welfare reform and responds to Americans' wariness of assisting poor people who aren't working. It got new traction in the wake of the recession, when the stimulus package authorized a TANF emergency fund that states used for subsidized job programs.

Though the program wasn’t rigorously evaluated, retrospective analyses suggested that the fund was effective at creating work at low cost: $1.3 billion in federal funding led to more than 260,000 new jobs, a ratio of about $5,000 per job created. The fact that the subsidized jobs programs were set up and implemented quickly was also encouraging, suggesting that large-scale direct job creation is administratively possible in the modern era.

The idea could get fairly wide buy-in. Brookings and AEI recently convened a bipartisan working group of 15 poverty scholars — including Ellwood, Doar, Haskins, Mead, and Waldfogel — attempting to arrive at a common approach to combating poverty. The resulting report was very enthusiastic about subsidized jobs, and particularly positive toward the TANF emergency fund.

"I actually support that in principle," Mead says. "If you want to require work, you must be sure work is available, but in principle that would have to mean some kind of job guarantee." This could be useful not just for the single mothers AFDC targeted but for underemployed men as well.

William Julius Wilson, at Harvard, is also a strong believer in subsidized jobs, suggesting former Rep. George Miller's (D-CA) 2011 Local Jobs for America Act as a worthy model.

"When I talk about public sector jobs, I mean the types of jobs provided by the Works Progress Administration [WPA] during the Great Depression," Wilson says. "Work that would help to improve the infrastructure in our communities, including state and local park districts that suffer from lack of upkeep and limited hours; cleaning playgrounds, beaches, and other recreational areas; working in public libraries to keep them open into the evening and on weekends; increasing the number of safe, stimulating day care centers for toddlers and preschoolers; improving and increasing the number of after-school programs for school-age children."

Sheldon Danziger, a leading left-of-center welfare expert who's now serving as president of the Russell Sage Foundation and who was in the Brookings-AEI group, also places heavy emphasis on subsidized work. "We don’t have enough experience with guaranteed public sector work. I would certainly experiment with that," he says. "I would also welcome public subsidies for private sector and nonprofit employers, summer jobs for youth, etc."

One thing that's clear is that any solution would cost money. The TANF Emergency Fund might've been cheap, but there's no guarantee that it would scale up at the same cost per job. Miller's bill allocated $75 billion. A child allowance paying $2,500 per year ($208 per month) would cost $109.3 billion a year on top of existing child tax credit spending.

A comprehensive plan from the Community Advocates Public Policy Institute in Wisconsin, incorporating a subsidized jobs program and expanded EITC, among other measures, would cut poverty by at least half. It'd also cost $332 billion to $399 billion per year.

To be clear, the US could afford a policy that big. Even though $400 billion a year sounds like a lot, it is a little over 2 percent of GDP; the US could grow its government by that amount and still remain way smaller than European welfare states, cutting poverty dramatically. The question is whether, in an era of divided government where the Republican Party has a strong ideological commitment to lower spending, a governing coalition will ever be willing to make that large a commitment to fighting poverty.

Watch: We know how to end poverty. So why don't we?