Employment Effects of the Earned Income Tax Credit: Taking the Long View

An Overhyped Tax Credit

The earned-income tax credit is one of those bipartisan policy idols that everybody worships but nobody examines closely for results. Now comes A study showing that its employment benefits are grossly overstated.

Congress created the EITC in 1975 as temporary assistance to increase the marginal incentive to work. The idea, inspired in part by Milton Friedman’s negative income tax, was that tax benefits would increase as workers earned more and phase out after they hit a certain threshold (initially $4,000).

We supported the credit as an alternative to raising the minimum wage because it doesn’t reduce incentives to hire low-skilled workers. But politicians expanded the credit and have kept raising the minimum wage. (Memo to those who want a carbon tax to replace some other tax.) One in five households now claims the credit, and government payments exceed cash welfare by six times.

Complicated rules on eligibility and credit size have encouraged fraud. According to the IRS, a quarter of the $69 billion in EITC payments last year were “improper.” Yet politicians from both parties continue to pump it as a labor inducement. As Princeton economist Henrik Kleven shows, there’s scant evidence it boosts employment.

The biggest EITC expansion occurred in 1993 when Congress doubled the maximum to about $2,152 for taxpayers with one child and tripled it for those with two or more. Most economists have attributed the subsequent spike in labor participation and employment among single mothers to the credit. But Mr. Kleven concludes that the real labor stimulator was welfare reform aided by a booming economy. It’s odd, he notes, that none of the other four EITC expansions in history has been linked to higher labor participation among single mothers. So what was special about the 1993 expansion?

Some economists have posited that the 1993 expansion was simply larger than the others, though this doesn’t explain why employment increased most among women with more children. Following the 1993 expansion, employment increased by 10 percentage points among single mothers with one child, 15 points among those with two, 20 among those with three and 30 points among those with four or more. Mr. Kleven says this pattern is more consistent with state and federal welfare reforms in the early 1990s that reduced government cash benefits most among women with more children.

Before Congress passed welfare reform in 1996, 113 state and county waivers were approved to establish work requirements and family size caps and time limits on benefits. A 2016 study found Connecticut’s welfare waiver induced many women to work. Yet Mr. Kleven finds in his study that state EITC expansions have not produced a similar effect.

“Overall and contrary to consensus, the case for sizable extensive margin effects of the EITC is fragile,” Mr. Kleven concludes. His study adds to the economic research finding that the EITC phase-out causes steep infra-marginal tax rates as workers make more money.

None of this may bother Republicans or Democrats who want to redistribute more income, but they shouldn’t claim these welfare tax payments are an incentive to work.

Reconciling their work with Kleven’s“…Column (3) adds fixed effects for the interaction between year and presence of children. Recall that the fundamental identification approach in studying the EITC is the interaction of year and whether a household had children present, so the inclusion of these fixed effects absorbs most of the variation in the EITC, leaving only the differences across family sizes for those with 20 children to identify the effects. When these controls are added, the EITC’s impact on annual employment is no longer significant in either the low-education sample or among all observations, though the impact on worked last week is still significant in this specification. Following Kleven, column (4) adds a three-way fixed effect, interacting presence of waivers with presence of children with year. Once this additional control is added, coefficients on the EITC variable lose statistical significance in all four specifications. It is unsurprising that these fixed effects that absorb most of the variation used to identify the EITC’s impact also absorb its impact on employment. Instead of employing a differences-in-differences approach, Kleven (2019) estimates an event study which allows the coefficient on indicators for presence of welfare waivers (which vary at the state and year level) to vary by children-times-year. We exactly replicate his findings (column 3 of Table 6) in Appendix Table 5. Kleven shows estimates on the year-3 and year-10 post-EITC expansion coefficients; we also include year 5. We present estimates separately for the low-education sample and the overall sample of unmarried women, and for annual employment as well as whether the respondent worked in the prior week. When waivers*children*year fixed effects are included, year 3 effects are small and not statistically significant, though those in years 5 and 10 are. When this three-way fixed effect is omitted, we find consistent, positive impacts of the EITC on employment….”

new Strain pushes back at Kleven's argument that EITC gains were largely driven by welfare reform. He finds that states without early welfare reform had strong EITC impacts, suggesting welfare reform didn't do all the work, on average employment increase by ~ 5 percentage points 5 years after an EITC expansion (See Figure 3, F “All Federal Reforms”) “…When studying all federal expansions together, we find that a $1,000 increase in the size of the maximum credit is associated with a 3.2 to 3.7 percentage point increase in employment among unmarried mothers with low levels of education. When examining both federal and state expansions together, we find a 2.9 to 3.3 percentage point increase in employment…”

Looking at a series of EITC expansions, “…. We find robust evidence that four of the five credit expansions increase employment among unmarried mothers. The exception is the 2009 expansion, which was targeted on households of a specific size and occurred during a period of historically weak labor demand following the 2008 financial crisis and Great Recession. We also look more closely at the 1993 expansion and conclude that this generous credit expansion increased the extensive margin of labor supply separate and apart from any pro-work reforms to state welfare systems that occurred at the same time. Finally, we study the EITC holistically, pooling all five expansions into one event-study framework and using continuous variation in the size of the maximum EITC benefit to estimate its employment elasticity. To our knowledge, the only other paper to look comprehensively at all five federal EITC expansions is Kleven (2019), which comes to different conclusions. Specifically, Kleven finds little evidence for an EITC extensive margin effect. We find the opposite, reconcile our findings with Kleven, and argue that our analytic approach is most appropriate….”