A major new study finds the biggest federal work subsidy doesn’t promote work

The most important paper in tax policy right now, explained.

The earned income tax credit (EITC), one of the largest tax benefits distributed every year, exists to do one thing in particular: encourage low-income people to work. But new research suggests: Maybe it doesn’t?

A research paper from Princeton economist Henrik Kleven, released in draft form in September after circulating as a set of slides for months, looks at the EITC’s creation and the four times it’s been expanded, including a big increase in the credit’s value in 1993.

Kleven found, consistent with several past studies, that the 1993 increase was followed by a large increase in employment. But he finds that no other increase had this effect, and that the 1993 increase might have more to do with both state and federal welfare reforms and the economic boom of the ’90s than with the EITC itself.

Kleven’s conclusion is a stark departure from the existing literature on the credit, and, if true, undermines the case for the EITC on both the left and right. Conservatives and business interests have long grudgingly tolerated or outright supported modest expansions of the EITC because of its tie to work: It isn’t welfare, but an earned benefit that helps the economy by increasing labor force participation. If it doesn’t help labor force participation, that rationale goes away.

Similarly, liberals have long supported the program as a politically realistic way to direct money to low-income people, a benefit whose impact is boosted by the additional wages recipients earn because the credit pushes them to work more. But if the conservative rationale for it is bunk, and the additional pay doesn’t result because the credit doesn’t actually spur more work, then this rationale similarly falls away.

The policy alternatives toward which Kleven’s findings push people are pretty much diametric opposites. For conservatives, the Kleven research suggests that welfare reform was more effective at pushing people into work than previously thought, and that similar welfare reforms extended to programs like food stamps might be similarly effective.

For liberals and leftists, the research suggests that the EITC’s tie to work could be severed completely without major economic damage, and the government should move toward a no-strings-attached cash subsidy for the poor, like a basic income or negative income tax.

Kleven doesn’t endorse either of these strategies — he doesn’t offer a prescriptive takeaway at all — and it’s entirely possible future studies will contradict his conclusion, just as he contradicts the conclusions of studies before him. But the study is already influencing DC policy conversations, and fits with a broader shift away from bipartisan cooperation for policies like EITC and toward more aggressive and distinct approaches in each party.

A brief summary of what the EITC does

Before we get too in the weeds, let’s review the policy we’re talking about.

The EITC is a tax credit that workers typically receive in a big lump sum when they file their tax returns in the spring. It differs from most tax credits in that it’s “refundable”: Even if you don’t earn enough money to pay income taxes, you can still get the credit in your tax refund.

The credit is tiny for childless adults, topping out at $529 for 2019, but very substantial for parents, with benefits maxing out at $6,557 for families with three or more kids. It phases in with each dollar you earn, and phases out gradually; the rates differ based on how many kids you have, and are summarized in the chart below, but a single mother of two gets 45 cents for every dollar she earns up to her first $14,570 in earnings; at $19,030 in earnings and above, the credit starts phasing out at 21.06 cents on the dollar, until it’s totally gone for a single mom making $46,073 or more.

Tax Policy Center

The result looks a bit like a trapezoid, as you can see above; the EITC is sometimes known in tax circles as a “trapezoidal” policy for this reason (tax circles are extremely fun).

The thresholds are different for married couples, but there’s still a marriage penalty: marrying a partner can bump your income up enough that you no longer benefit from the credit or losing hundreds or thousands of dollars in benefits, creating a strong reason to stay unmarried.

The EITC was first introduced in 1975 but has been expanded four times since: in 1986 (as part of that year’s landmark tax reform), 1990, 1993, and 2009 (as part of the stimulus package). Twenty nine states, plus DC and Puerto Rico, also have state EITCs that supplement the federal credit.

Kleven’s conclusion: all but maybe one EITC expansion did nothing for employment

So Kleven’s paper specifically focuses on the labor force participation of single mothers — that is, single mothers who are employed or are looking for work — because they get the best deal out of the credit. If the credit spurs anyone to work, it should be them. In particular, he compares their labor force participation rates to those for single women without children, who were mostly unaffected by EITC increases.

Henrik Kleven

He finds that from 1968 and 1992, a period that encompassed the 1975 introduction of the EITC and major expansions in 1986 and 1990, the gap didn’t change at all. You would expect single women with children to catch up if they responded to a big new subsidy by jumping into the workforce. That didn’t happen.

Kleven validates this conclusion by doing more careful, controlled event studies comparing labor force participation and employment pre- and post-reform, and comes up empty: “Apart from the 1993 reform, all estimates are either small and statistically insignificant or they are negative,” he finds. After the 1993 EITC expansion, labor force participation grew by about 3 percentage points for single mothers. But the 1975, 1986, 1990, and 2009 expansions of the EITC didn’t have any effect on labor force participation or employment, according to his analysis. Nor did state EITC enactments and expansions.

This finding directly contradicts that of an early paper on the EITC, by Georgetown’s Nada Eissa and Harvard’s Jeffrey Liebman. In that 1996 study, Eissa and Liebman found that single women with children saw their labor force participation grow by almost three points relative to single women without children after the 1986 EITC expansion. Kleven disputes that paper by noting that “most of the effect goes away between 1990 and 1991, i.e. after the 1990 reform had enacted further EITC expansions.” That suggests, he says, that fluctuations in employment in the late ‘80s were the result of something other than the EITC. (I’ve asked Eissa for comment on the Kleven paper and will update when I hear back.)

The case that welfare reform is more important than EITC

That leaves 1993. Kleven is skeptical here too. The EITC expansion that year, spearheaded by newly elected President Bill Clinton, was larger for families with two kids or more than for families with one kid — but employment increased for families with one child as well. More importantly, he finds that when you control for state unemployment and state welfare waivers that placed new restrictions on access to welfare from 1994-1996, any employment effects of the EITC go away.

The conclusion: welfare reform pushed a bunch of people off welfare and into work, and the EITC didn’t do much of anything for employment, not least because it’s tucked in the tax code and hard to understand.

This is at odds with the prior literature on the EITC and welfare reform, which I summarized here and generally put greater weight on the EITC than welfare reform for the increase in single mothers’ employment during the 1990s. A 2003 paper by UChicago’s Jeffrey Grogger, for instance, found that time limits imposed by welfare reform were less important than the EITC both in increasing employment and in reducing welfare usage.

Kleven’s paper has not been peer reviewed and most economists I’ve contacted who’ve written papers with different conclusions haven’t had time to read and process it. Some of the difference appears to come down to what, precisely, Kleven’s models control for that others do not.

Kleven allows the effect of welfare reform in his model to vary by year and state, in part to try to capture variations in how states implemented the policy (which gave governors and legislators a lot of latitude). But that also adds a ton of new variables to the model, which sometimes leads to over-controlling: accidentally controlling away for effects that are actually the result of the EITC but get picked up anyway by building such a large model with so many inputs.

That makes it a bit hard as a non-economist to be sure whom to believe here. I’ll put it this way: before I read Kleven’s paper, I was maybe 90 percent sure that the EITC reliably induces single mothers to work more. Kleven’s taken that number down to about 60 percent certainty. I’m inclined to trust the literature before him slightly more. But I’m also much more prepared than I was before to believe that literature is wrong, and the sheer comprehensiveness of Kleven’s paper lends it credibility to me.

For his part, Kleven credits some of the paper’s power to its simplicity. “Everything’s told through very simple graphs,” Kleven told me. “When you look at the data, and you just plot the raw data and graphs, it’s just not clear that anything is happening or anything sharp is happening around [non-1993] reforms.”

What does this mean for policy?

As Kleven is quick to state, his paper being correct needn’t damn the EITC. “If the EITC doesn’t have labor supply impacts, that could imply its an even better policy,” he says. “It’s a pure money dump on the working poor that doesn’t come with any labor supply distortions.”

But that raises an obvious question: why should the working poor be the only ones to get the money dump? This is the argument that EITC critics from the left like the People’s Policy Project’s Matt Bruenig have raised. If the EITC doesn’t actually encourage work, there’s no real rationale for the “phase-in” side of the policy. We should just give, say, mothers with two kids $6,500 cash no matter what she earns at first, and then tax it away as she earns more. It should be more like food stamps, not a subsidy to work.

Some politicians are on board with this approach too. Rep. Rashida Tlaib (D-MI) has introduced the LIFT+ Act, a variation on Kamala Harris’s LIFT Act plan to expand the EITC. Tlaib’s variant drops the phase-in and just offers $3,000 per adult, phasing out with income but with no work requirement.

But there’s also material for conservative welfare reform defenders, like the Joint Economic Committee’s Scott Winship, who have been insistent that the 1996 federal law and contemporaneous state experiments spurred additional employment. They typically haven’t dismissed the possibility that the EITC did as much or more — but if Kleven is right, then the conservative case for welfare reform on the grounds that it promoted work in a population that was overly dependent on government aid gets stronger. This doesn’t bear directly on whether welfare reform reduced poverty, as Winship and others argue, but it’s an important part of the conservative argument.

The current major conservative push on welfare issues is to adopt tougher work requirement for food stamps, which some conservative states have already done. You could imagine the Kleven paper bolstering the argument for these, on the grounds that if work requirements effectively increased employment when applied to cash welfare, they should do the same if applied to food stamps. The Heritage Foundation’s Robert Rector (one of the most vociferous and dogged opponents of all welfare programs in all of Washington) and Jennifer Marshall called work requirements for food stamps and public housing the “unfinished work of welfare reform” in a 2013 essay.

There are many reasons to be wary of such a policy: even if food stamps deter work somewhat (and they probably do), one of the reasons Bill Clinton signed the welfare reform bill was that he thought through Medicaid and food stamps, the poorest of the poor would be guaranteed food and health care, even if they didn’t work. This, he reasoned, was a basic humanitarian floor, even if giving cash was too much. But the logic of welfare reform, bolstered by the Kleven paper, does push in the direction of applying work requirements to an ever-increasing set of programs.

Sign up for the Future Perfect newsletter. Twice a week, you’ll get a roundup of ideas and solutions for tackling our biggest challenges: improving public health, decreasing human and animal suffering, easing catastrophic risks, and — to put it simply — getting better at doing good.