Let’s start with a table. (With apologies to our mobile audience…)
Evidence on Extensive Margin Employment Elasticities for Single Mothers
| Reported | Revised | Consistent w/ 0.75? | Notes | |
| Reviews or Based on Reviews | ||||
| Corinth et al. (2021, rev 2022) | 0.75 | — | p. 22; based on midpoint of McClelland and Mok (2012), p. 5 (below); unpublished | |
| Hotz and Scholz (2003) | 0.7-1.2 | 0.4-1.7 | Y | reported: Hotz and Scholz (2003), p. 173-174; for range, see McClelland and Mok (2012), p. 41 (below); see first 5 entries under Original Studies, below; revised: see notes for Eissa and Liebman (1996) and Keane and Moffitt (1998) below |
| Eissa, Kleven, and Kreiner (2004) | 0.4-1.7 | — | Y | “The estimated participation elasticities have been in the interval from 0.35 to 1.7, with a central value of about 0.7.” p. 798; single mothers |
| Eissa and Hoynes (2006) | 0.7-1.2 | — | Y | “The range of the implied labor force participation elasticity with respect to net income across all studies is quite narrow–between 0.69 and 1.16.” p. 98; EITC-eligible single mothers |
| McClelland and Mok (2012) | 0.3-1.2 | 0.4-1.7 | Y | reported (based on review): p. 41; p. 5 — “low-income taxpayers eligible for the EITC”; p. 14 — “low-income single women with children who are the focus of the EITC studies discussed below”; unpublished revised: see Winship (2022), p. 5 |
| Chetty et al. (2013) | 0.3-0.4 | 0.4-0.7 | Y | reported: Table 1; reanalyses of Eissa & Liebman and Meyer & Rosenbaum (below); single mothers revised: see Winship (2022), p. 4; upper bound from Corinth and Meyer (2021, rev 2022) |
| Nichols and Rothstein (2016) | 0.7-1.0 | — | Y | review: “…consensus estimates of the extensive-margin elasticity around 0.7 to 1.0”, p. 198; EITC recipients |
| Goldin, Maag, and Michelmore (2021, rev 2022) | 0.3-1.0 | — | Y | draft of 9/3/21: “…the extensive margin elasticities found in this literature (which approach 0.70 to 1.00 for single mothers)…” draft of 10/7/21: “…the extensive margin elasticities found in this literature (which range from 0.30 to 1.00 for single mothers)…” draft of 10/27/21 and final draft: “Extensive margin elasticities with respect to changes in the net-of-tax return to work, calculated based on changes in the EITC tax schedule, range from 0.30 to 0.45″ |
| Bastian (2022) | 0.6-0.7+ | — | Y | “A large literature on lower-income unmarried mothers and the EITC has typically found strong labor supply elasticities, around 0.60-0.70 on the extensive margin, though sometimes even larger.” (p. 11-12, draft dated June 12, 2022); unpublished |
| Average of Reviews (excl Corinth et al.) | 0.85 | |||
| Reported | Revised | Consistent w/ 0.75? | Notes | |
| Original Studies | ||||
| Dickert, Houser, and Scholz (1995) | 0.9 | — | Y | see Hotz and Scholz (2003), p. 173; single-parent families |
| Eissa and Liebman (1996) | 1.2 | 0.4 | lower | reported: see Hotz and Scholz (2003), p. 173; single mothers revised: Winship (2022), p. 4, revising Chetty et al. (2013), Table 1, which put it at 0.30 |
| Keane and Moffitt (1998) | 1.0 | 1.7 | Y | reported: see Hotz and Scholz (2003), p. 173; single mothers revised: Corinth et al. (2021, rev 2022), p. A-21 |
| Meyer and Rosenbaum (2001) | 1.1 | 0.7 | Y | reported: Hotz and Scholz (2003), p. 174, report 0.7, citing Meyer and Rosenbaum (1999); M&R (2001) revised to 1.1 (p. 1092); single mothers revised: Corinth et al. (2021, rev 2022), p. A-24, correcting Chetty et al. (2012) |
| Hotz, Mullin, and Scholz (2002) | 1.0-1.7 | — | Y | see Hotz and Scholz (2003), p. 174; last draft of paper was 2010, unpublished; former & current welfare recipients |
| Grogger (2003) | 0.6 | — | lower | see Lippold (2019), p. 10; female-headed families |
| Gelber and Mitchell (2012) | 0.3-1.2 | — | Y | Table 2 (single women, not just single mothers) |
| Chetty, Friedman and Saez (2013) | 0.4 | — | much lower | p. 2715-2717; EITC-eligible families w/ children (incl. married) |
| Lippold (2019) | 0.4-1.0 | — | Y | p. 43-45 (lower-income parents, not just single mothers); unpublished |
| Schanzenbach and Strain (2020) | 0.9-1.5 | — | Y | Corinth et al. (2021, rev 2022), p. A-21; single mothers and single mothers w/ hs or less; unpublished |
| Bastian (2020) | 0.6 | — | lower | Appendix D; single mothers |
| Bastian and Jones (2021) | 0.3 | ~1.5 | Y | reported: p. 8 (EITC-eligible women, not single mothers) revision: Winship (2022), p. 5 (EITC-eligible single mothers) |
| Bastian and Lochner (2021) | ~1.5+ | Y | see Winship (2022), p. 5; single mothers | |
| Michelmore and Pilkaukas (2021) | 0-0.4 | — | much lower | p. 914; single mothers w/o college degree |
| Hoynes and Patel (2018, rev 2022) | 0.1-0.4 | 1.0-1.4 | Y | reported: Appendix Table 7; single mothers revised: Corinth and Meyer (2022), Table 1 |
| Kang (2022) | 0.1-0.2 | — | much lower | p. 30; single mothers or single mothers w/ hs or less; unpublished dissertation |
| Zheng (2022) | 0.5 | — | lower | p. 17; single mothers; unpublished dissertation |
| Kleven (2023) | -1.5-1.6 | — | Y | Table 3; low-income single mothers; unpublished |
| Average of Original Studies | 0.80 |
The table summarizes the evidence in favor of a key assumption we and our colleagues make in our recent analysis of the Wyden-Smith bill—an assumption that has been subjected to unwarranted dismissal. The methodology at issue involves our “labor supply elasticities.” Our choice of elasticities drives our estimate of how large the employment effects would be if the so-called “look-back” provision of Wyden-Smith’s child tax credit (CTC) reform were permanent policy. (That provision would let people use either earnings from the most recent tax year or the one before it to determine the refundable child tax credit they receive. It would encourage work among some families while discouraging it among others.)
Which are the proper elasticities to use in this kind of analysis is an incredibly wonky, but central point of contention. Don’t worry—we’re not going to get into the weeds here. (For that, you can peruse our other writing on the debate.) But you need to understand that we are on the correct side of the debate, despite what a number of prominent critics have claimed without always engaging the evidence. We’ll circle back to our table after providing some background.
In economics, “elasticities” simply describe the percentage change in one thing caused by a one-percent change in another thing. The elasticity at issue indicates the percent change in employment due to a one-percent change in the return to work. A large elasticity (positive or negative) means that work decisions are more sensitive to changes in the payoff to working. Elasticities close to zero indicate that work decisions are little affected by changes in the return to work.
We employ two different elasticities—one for single mothers, and one for everyone else. Here’s how it works, for now taking up the case of single mothers. We want to estimate the number of single mothers who would work two years in a row in the absence of the look-back provision but who will work only one year in its presence. Our target population is single mothers who worked in both 2018 and 2019. For each of them, we estimate the percent change in the return to working both years versus only working one year, due to the policy change. That just involves comparing how much better off they’d be working both years in the presence of the look-back to how much better off they’d be working both years in its absence. We multiply the number of single mothers who worked in both 2018 and 2019 by the average percent change in the return to work. Finally, we multiply that product by the labor supply elasticity we assume for the target population.
A number of critics have claimed that the labor supply elasticities we use are “outliers”—that our elasticities imply that employment is much more sensitive to changes in the return to work than most research suggests. If that were true, then our estimates of employment losses (and employment gains) from implementing the look-back would be too large.
Is the charge true? It is not. At all. The controversy has revolved around our elasticity estimates for single mothers, so that’s what we’ll focus on here.
We chose our elasticity for single mothers of 0.75 based on reviews that we have both conducted of 30 years of research on labor supply elasticities. This is the same elasticity used by one of us (Corinth) in a 2021 study assessing the temporary 2021 expansion of the CTC. That study estimated that making the expansion permanent would reduce employment by 1.5 million workers.
Corinth and his coauthors arrived at 0.75 because it was the midpoint of a range of estimates provided by the Congressional Budget Office for “lower-income taxpayers eligible for the EITC [Earned Income Tax Credit].” Corinth et al. offered several other justifications for using 0.75, citing reviews of the literature and individual studies. After the inaugural round of criticism, they even showed that when calculated correctly, the estimates from studies by prominent advocates of a CTC expansion—two of whom had publicly dismissed the 0.75 estimate—were at least as large. Subsequently, another prominent critic’s claim that 0.75 was absurdly large was undermined when one of us (Winship) pointed out that the naysayer’s own research implied elasticities of between 0.6 and 1.5.
Despite our attempts to engage with our critics, the perception persists that 0.75 is much too large an elasticity to use when assessing the effects of CTC reforms on single mothers. Rather than go through another elaboration of our case, we present our table at the top of this post, summarizing the research on elasticities for single mothers and low-income families. We divide the research into summaries of the literature—some of them comprehensive, some informal in the course of discussing related studies—and original analyses (or calculations from original analyses). We show reported elasticities, and when we and others have revised those reported estimates (here, here, and here), we show those too. Scroll back up and take another gander—we’ll wait.
A few words are in order. First, we report the average elasticities for the reviews (excluding Corinth et al.) and for the original studies. To compute the averages, we use the midpoint when a study reports a range. The average for the eight reviews is 0.85, and the average for the 18 original studies is 0.80. After incorporating subsequent revisions to several studies’ estimates, the Corinth et al. estimate of 0.75 is consistent with the conclusions of all eight reviews, meaning that it is within the range of studies in the review (generally on the low end of the range). The 0.75 estimate is consistent with all but one review without incorporating any revisions.
Of the 18 original studies, 0.75 is consistent with 11 of them, meaning that it is within the range found in a study, near what the study finds, or lower than in the study. Our preferred elasticity is often substantially lower than studies report or on the low end of the range of estimates. Our 0.75 elasticity is somewhat higher than the elasticities in four of the studies, and substantially higher than those reported in only three studies. Without revisions, 0.75 is consistent with ten studies (again, often lower than what a study reports), somewhat higher than in three, and substantially higher than in five.
The review includes studies by a number of economists that most would regard as being at the top of the field. It includes two John Bates Clark medalists (given to the top economist under 40), Congressional Budget Office researchers, the senior economist of the Council of Economic Advisers, and the president of the University of Oregon. Of the 26 papers listed, 16 include at least one author who signed a public letter supporting the permanent expansion of the CTC in 2021. Two papers include the organizers of that letter. Average the estimates of the 16 and it comes to 0.89. Remove two unpublished dissertation chapters from the table and the average elasticity for the remaining 16 original studies is 0.86.
One criticism of our analysis might be that the research we have reviewed has generally considered EITC-eligible single mothers rather than single mothers generally. Unlike Corinth et al., we have used the 0.75 estimate for all single mothers. The elasticity for single mothers with higher income may very well be lower than the elasticity for EITC-eligible single mothers. However, given that it’s likely the appropriate elasticity for EITC-eligible single mothers exceeds 0.75, the elasticity for other single mothers may not be far from 0.75. At any rate, we estimate that single mothers not eligible for the EITC (or not eligible if they worked) are no more than a quarter of single mothers as a whole. If the appropriate elasticity for them were, say, 0.4, and the appropriate elasticity for EITC-eligible single mothers were 0.85, the elasticity for all single mothers would be no lower than 0.74.
We hope this review has clarified that whatever criticism might be levied at our analysis of the Wyden-Smith look-back, the charge that we use outlier elasticities is not a valid one.