For the past two years, economist Jacob Bastian has been the main researcher dedicated to trumpeting the virtues of a child tax credit (CTC) expansion. Writing first as an economist from academic perches at Rutgers and Princeton, later as an affiliate of moderate think tanks like the Niskanen Center and the R Street Institute, and now tweeting steadily as the senior economist in President Biden’s Council of Economic Advisers, Bastian has pushed back against research concluding that CTC expansions would encourage large numbers of single parents to leave the workforce.
Critiques of those arguments have shown that his findings, ironically, reinforce the claims of his opponents in the CTC debate. His latest effort is no different. More than that, though, his paper, when reanalyzed, shows just how important work-incentivizing safety net reforms have been over the past 40 years. That is the other side of showing that recent proposed CTC reforms would discourage work.
Bastian’s new paper is his latest attempt to refute an important finding of my colleagues Kevin Corinth and Bruce Meyer and their coauthors. The CTC expansion temporarily enacted in 2021, which Democrats sought to make permanent, would have given the same CTC to parents regardless of whether or not they worked. That was a break from CTC policy before and since 2021, which has required that parents have earned income and which has reserved the full CTC for those with sufficient income tax liability. Corinth et al. (2022) concluded that making the 2021 expansion permanent would reduce parental employment by almost 1.5 million.
The Corinth et al. finding hinges on a measure of the sensitivity of single mothers’ decision to work or not work in response to changes in the payoff to working. Corinth, Meyer, and their coauthors semi-famously assumed a “labor supply elasticity” of 0.75 to quantify this sensitivity, and Bastian and others have unsuccessfully tried to argue that that is far too high. (In truth, the research of even several of these critics, Bastian included, is consistent with an elasticity this high.)
In the latest paper, Bastian tries to show that if you assume this degree of sensitivity and then simulate what would happen to single mothers’ employment if both the CTC and the earned income tax credit (EITC) were eliminated, you get employment declines that cannot be believed. The paper amounts to arguing that thirty years of research on the sensitivity of work decisions to incentives—a body of work to which Bastian has made key contributions—must be deeply flawed, so we cannot trust empirical analyses that cast CTC expansion in a poor light.
To make this case, Bastian needs a few ingredients to quantify his argument: the number of working single mothers (separately depending on whether or not they are eligible for the EITC), the average change in the return to work they would face if the EITC were eliminated (separately for the two groups), the same average change if the CTC were eliminated, and labor supply elasticities for both groups of single mothers. Bastian wishes to discredit the Corinth et al. elasticities, so he uses them (0.75 for EITC-eligible single mothers, 0.25 for other single mothers). He has estimates of the return-to-work changes from his own research. If one has the number of working single mothers (in each category), their average change in the return to work from eliminating the EITC and CTC, and the relevant elasticity, one can multiply these terms together to get the number of single mothers in each category who would stop working.
However, rather than use the number of working single mothers in these calculations, Bastian uses the number of all single mothers (working or not). That is directly opposed to what he says is appropriate in a forthcoming paper that will appear in the National Tax Journal. In that paper, he applies his elasticities and average change in return to work estimates only to working parents, rightly saying “only working parents can stop working.” (p. 18) Corinth et al. do the same. By basing his calculations on all single mothers, Bastian overstates the employment declines from eliminating the EITC and CTC.[1]
Because of this and other problems, the calculations Bastian makes to argue that Corinth et al.’s analyses are implausible bear no relevance to the question.[2] We can do the calculations correctly using Bastian’s own 2017 numbers from his forthcoming NTJ paper.[3] He finds 6.8 million working EITC-eligible single mothers and 1.9 million other working single mothers with adjusted gross income under $80,000.[4] It looks like there are perhaps another 1.0 million working single mothers with AGI over $80,000, and another 2.4 million without earnings.[5] That makes a total of 12 million single mothers.
Bastian wants to argue that the elasticities used by Corinth et al. are implausible. If we use those elasticities and apply Bastian’s estimate of the average change in the return to work from eliminating the EITC (24.7 percent for EITC-eligible single mothers) to the number of working EITC-eligible single mothers, we find that 1.26 million EITC-eligible single mothers would stop working if the EITC were eliminated. (Eliminating the EITC obviously would not affect single mothers not eligible for it.)
If we apply Bastian’s estimates of the average change in the return to work from eliminating the CTC (9.9 percent for EITC-eligible single mothers, 4.5 percent for other single mothers) but use Corinth et al.’s elasticities, we get an additional 530,000 single mothers who would stop working.[6] The total comes to about 1.8 million single mothers, or 15 percent of them. Bastian, however, reports 22 percent using his unusable numbers, overstating the actual number by nearly 50 percent.[7]
What kind of employment rate would be implied by a drop of 1.8 million single mothers? Bastian’s Figure 2 displays the trend in single mothers’ employment rate from 1940 (about 44 percent) to 2022 (72.3 percent). I estimate that in the absence of the EITC and CTC, the 2022 rate would be 59.2 percent.[8] Eyeballing Bastian’s Figure 2, that is roughly the level of single mother employment that existed in 1980 or 1984, before the EITC was significantly expanded for the first time in 1986 and before the CTC existed. Bastian, however, using his bad numbers, claims that a 0.75 elasticity means that single mother employment would fall to 50.1 percent “a level not seen since 1950.” (p. 4)
Bastian cites the research of Schanzenbach and Strain (2020) to argue that a reasonable guess for the true employment effect of eliminating just the EITC would be a 9.0-percentage-point reduction among single mothers. Using Bastian’s 2017 numbers as above (with the Corinth et al. elasticities) and applying them to Bastian’s 2022 employment rate, the estimated reduction in employment is 9.5 percent of single mothers. In his paper, Bastian mis-estimates this as 15.6 percent.
Thus, the conclusion that an elasticity of 0.75 cannot be reasonable is completely reversed when the problems in Bastian’s analysis are corrected.[9] An elasticity of that size is entirely compatible with the changes in employment we have seen among single mothers.[10] The proper conclusion is that the EITC and CTC have contributed enormously to increasing employment among single mothers, which has contributed enormously to reducing their poverty.
Bastian’s paper, properly reanalyzed, thus reinforces several conclusions about safety net policy. First, it bolsters the findings of Corinth et al. that a permanent expansion of the CTC would reduce employment by 1.5 million parents. Second, it reinforces the assumption used by Corinth et al. that single mothers are highly responsive to changes in the return to work. Third, it demonstrates the important consequence of the National Academy of Science’s mis-modeling of a child allowance in its influential 2019 report on child poverty. Finally, it shows that pro-work safety net reforms have increased the employment of single mothers markedly over the past 30 years. (In that regard, you really should look at Bastian’s Figure 2, which shows the employment rate of single mothers jumping about 14 percentage points from 1994 to 2000, around the time of state and federal welfare reforms and a large EITC expansion. This was a leap with no precedent, which has not been repeated since, which did not occur among single childless women, and which was not fully reversed even during the Great Recession.)
We should not depart from the course of building a pro-work safety net, which has occurred alongside declining poverty rates among children and single-parent families. Corinth and I, with our colleagues Angela Rachidi and Matt Weidinger, have argued that the House-passed tax bill under consideration in the Senate unfortunately includes a provision that would set back some of this progress. Bastian’s modeling experiment only reinforces this conclusion.
[1] To calculate an employment effect, we multiply an elasticity by an average percent change in the return to work and then by some population total. The question is whether the population total should be just employed people or all people. The denominator of the elasticity is the percent change in the return to work, so when we multiply the elasticity by the average percent change in the return to work, the denominator goes away. That leaves just the numerator, which is the percent change in employment. That can be expressed as another fraction— the change in employment divided by the baseline number of employed people. If we multiply that by the number of employed people in a group, then the denominator goes away again and we just have the change in employment. If, instead, we multiply by the number of people (employed and not employed), no terms cancel out, and the change in employment is overstated by the ratio of people to employed people.
[2] There are two other problems with Bastian’s calculations. Bastian thinks that the Corinth et al. study implies there are 10.0 million EITC-eligible single mothers and 1.7 million other single mothers. To get these estimates, he uses a calculation that combines a total for single mothers (11.7 million) with an estimate of 771,000 EITC-eligible single mothers who would stop working if the 2021 CTC expansion were permanent. (See his footnote 1.) He attributes this 771,000 figure to Corinth et al., but this is incorrect. Corinth et al. mention the figure in a sensitivity test (p. 27). The National Academy of Sciences (2019) estimated that expanding the generosity of the EITC would increase employment among single mothers by 771,000 (using 2016 data, p. 495). Corinth et al. extrapolate this finding to estimate that eliminating the EITC would reduce the employment of single mothers by 1.9 million. Then they show that if they use their own preferred elasticities and simulate eliminating the EITC (with their 2017 data), they find a decline in employment among single mothers of 1.7 million. The point is to show that their assumed elasticities are reasonable. There is nothing in this exercise to suggest that the decline Corinth et al. estimate from a child allowance would be 771,000. If the true number were smaller or larger, it would affect how Bastian allocates 11.7 million single mothers between those who are EITC-eligible and those who are not.
Second, the 11.7 million that Bastian uses is from 2022 data, while the 771,000 figure, as noted, is from 2016 data. Bastian’s return-to-work estimates are based on 2017 data.
[3] Ideally we would apply estimates from Corinth et al., since they use survey data linked to administrative data, while Bastian only uses survey data. However, it does not appear that Corinth et al. provide the necessary detail for conducting the analysis.
[4] Divide column 4 of Table 2 by column 7, for both EITC-receiving single mothers and other single mothers.
[5] There are 12 million single mothers in 2017 (p. 17). There are 8.6 million working single mothers with less than $80,000 in adjusted gross income. That leaves 3.4 million single parents (mothers and fathers) who are either not working or have AGI above $80,000 (p. 26). The share of single parents who are mothers is 71 percent (p. 14). Applying 71 percent to 3.4 million yields 2.4 million non-working single mothers. Subtracting 2.4 million from 3.4 million leaves 1.0 million single mothers with AGI greater than $80,000.
[6] That includes about 11,000 with AGI greater than $80,000, who I assume have an average change in the return to work of 4.5 percent and an elasticity of 0.25. Otherwise, I simply use the estimated employment decline from Table 6 of his forthcoming paper (516,000). Bastian assumes, reasonably, that the change in the return to work if the CTC were eliminated is just the CTC a parent would otherwise receive, times negative one, which is the same change in the return to work from the 2021 reform that severed the credit amount from earnings.
[7] As I have argued elsewhere, using administrative data would show more working EITC-eligible single mothers, which would increase the number of single mothers who would stop working. But it might also increase the number of single mothers generally, which would increase the denominator in estimating the percent of single mothers who would stop working. Alternatively, it might reduce the number of non-working single mothers or working non-EITC-eligible single mothers.
[8] My 1.8 million estimate uses 2017 data. To apply the results to Bastian’s 2022 employment rate, I need to compute the number of single mothers who would stop working as a share of employed single mothers (rather than as a share of all single mothers). That percentage is 18.8. Applying 18.8 percent to Bastian’s 2022 employment rate of 72.3, the counterfactual rate in the absence of the EITC and CTC would be 59.2 percent.
[9] Bastian also models estimates from Corinth and Meyer that have been incorporated into Appendix B of Corinth et al. (2022). These Appendix B results are not estimates that Corinth et al. emphasize; they are from another sensitivity test of their main results, which use a method similar to that used by National Academy of Sciences (2019) to see whether it produces a figure in the ballpark of their main results. Here are the only two discussions of this estimate from the body of the paper: (1) “If the NAS had applied to the CTC the methods it used to simulate changes to the EITC, it would have found an estimate larger than ours (as shown in Appendix B).” (p. 4); (2) “We use the NAS EITC simulations as a comparison to our child allowance simulations in two ways…. Our second approach, shown in detail in Appendix B, uses the much simpler NAS methods….the NAS EITC methods applied to the child allowance gives [sic] estimates close to our child allowance estimates. In both cases, the estimates using NAS methods are higher than ours.[emphasis added]” (p. 26-27); I have noted this before in critiquing Bastian.
[10] It’s true, as Bastian notes, that other factors contributed to the increase in employment among single mothers since the early 1980s—welfare reform, though unmentioned by Bastian, being perhaps the most important. But other changes in safety net policy, such as expansions of SNAP and Medicaid, have offset pro-employment policies.