The Wyden-Smith tax bill under consideration in the House has rekindled a debate about the Child Tax Credit (CTC) and work incentives. We, along with our colleagues, Angela Rachidi and Matt Weidinger, recently released an analysis of the incentives built into one overlooked feature of the CTC reforms proposed in the bill—the so-called “look-back” provision. This policy change would allow low-income families to receive a CTC in years when they do not work, so long as they worked in the previous year. They could determine their eligibility for the refundable part of the CTC—the part that doesn’t offset income taxes—using either their earnings in the year for which they are filing their taxes or in the preceding year. If a family had no earnings in the tax year, but earned $15,000 the year before, they could receive the refundable portion of the CTC they’d have gotten if they had earned $15,000 again.

As we explained and modelled in our analysis, and as our colleague, Kyle Pomerleau, has also explained, the look-back provision will make working more attractive to some families and less attractive to others. Our analysis is, to our knowledge, the only one that attempts to determine the implications of these partly offsetting incentives. While we continue to refine our analyses, some of the objections that critics have directed at our paper compel us to respond in order to correct inaccurate claims and clarify where legitimate uncertainty exists.

Our paper explains that the look-back provision makes it less attractive—even if only by a couple thousand dollars—for parents to work next year if they are currently working and otherwise planned to work the following year too. Their income in the absence of work next year will be higher than it would have been absent the look-back provision. At the same time, it makes working this year more attractive—even if only by a couple thousand dollars—for parents who weren’t going to work this year or next year. They can go without earnings next year and receive the refundable part of this year’s CTC again. Their income in the absence of work next year will be higher as a result of working this year. (We acknowledge that this simple model of work incentives could be refined by considering a full life-cycle model of labor supply decisions by parents; however, we were restricted to considering a simple two-period model given that the existing CTC was not implemented until 2018 and COVID-19 caused major labor market disruptions beginning in 2020.)

Empirically, whether the look-back provision encourages or discourages work, on net, depends on the relative prevalence of people who work year in and year out in the absence of the provision and people who would not work this year or next in the absence of the provision. It also depends on how much the “return to work” declines for the former group and increases for the latter group. Finally, it depends on how sensitive employment decisions for both groups are to changes in the return to work. Our paper transparently lays out our assumptions, which are based on empirical evidence generated by us and by 30 years of economic research. We find that about 350,000 parents per year would leave employment as a result of the look-back provision. Roughly 200,000 parents per year would enter employment, so on net, the provision would reduce employment by 150,000 workers per year.

It’s worth clarifying explicitly that our paper is intended to model the long-term effects of a permanent look-back provision, not what would happen in 2024 or 2025 per se if Wyden-Smith became law. The current CTC will revert to its less generous and less costly pre-2018 form at the end of 2025, meaning that CTC reform will occupy much of Congress’s time that year. Proponents of a CTC expansion generally would prefer a child allowance of the sort that would remove any work requirement from CTC eligibility. They likely view a look-back as an improvement to the CTC even if it falls short of their ideal, and they might be expected to push for it in 2025 as part of a second-best option. Our intent is to show how the look-back provision would likely affect employment if it were ensconced in law over the long term. (That is one reason that analyses of the impact of Wyden-Smith in 2024 and 2025, such as the Joint Committee on Taxation estimate of revenue effects, may differ from ours. However, it is not clear how or whether JCT modeled the employment effects of the look-back provision.)

Is $2,000 Really Enough to Make People Stop Working?

One critique that has been leveled at our analysis strikes at the very idea that anyone would quit their job for a measly $2,000 (though notably, many parents would face larger work disincentives than that, including sole-earner married parents who would face a work disincentive of $4,000 on average). While that may seem implausible to you, the American economy includes people with varying attachments to work. Does it seem implausible that there are people who are currently not working but thinking about it who would reach their tipping point if another $2,000 were offered them? If such people exist, then there also exist people who are near their tipping point and considering leaving employment.

People who leave employment do not necessarily lose all their income. By leaving work, they may qualify for safety net benefits (or additional safety net benefits) that will partly compensate them for the lost earnings. People leaving work might move in with their parents to save on expenses, or they might move in with a romantic partner (or let their partner move in). It does not take a large fraction of workers to fall into these camps to produce a large absolute estimate of employment losses. Our figures imply that 1.0 percent of working families with children (and 3.0 percent of working single mothers) would stop working each year as a result of the look-back provision’s incentives. You might think that’s not a big enough number to worry about, but that’s a different critique than saying no one will be affected.

The Ping-Pong Fallacy

Whether our estimates are right or not depends on our assumptions. One simplifying assumption that we made, which could have been made more explicit, is that people have more or less stable plans for whether or not they will work in future years and that they are more or less able to implement their plans. If people knew with certainty which years they would work over, say, the next decade, in the absence of the look-back provision, then they could optimally change those ten-year plans to suit their preferences after the look-back is implemented.

It’s not that the same 550,000 people (350,000 “exiters” plus 200,000 “entrants”) would ping-pong in and out of the labor force every year over those ten years. Different people would move in and out in different years. Some people might take a year off in one year when they wouldn’t have. For instance, perhaps someone planned to work ten years in a row, but instead they work six, stop working for a year, and then work another three years. Or perhaps they lose their job, begin collecting unemployment insurance, and then sit out the next year even after the UI runs out. Conversely, some people might work in one year when they wouldn’t have.

Alternatively, some people might change their decision in multiple years. After all, there are, of course, people only loosely attached to the labor force who currently ping-pong in and out.

Work as Insurance

It’s true, of course, that people can’t perfectly know their intentions for working over a period of time, and even if they could, they wouldn’t be able perfectly to implement those plans. If our assumption that people can reliably predict their employment is off by a lot, that would be a problem for our results. For example, Pomerleau emphasizes the insurance value of working this year even if people are pretty sure they will work next year. What if something happens and they can’t work? By working this year, they ensure that they will at least have some income next year, equal to this year’s refundable tax credit (along with Unemployment Insurance and other existing safety net benefits). In the presence of uncertainty, there is always some value to working this year even if you worked in the previous year.

Empirically speaking, however, the insurance value of guaranteeing the CTC for next year is likely quite small. The less likely it is that parents are unexpectedly unable to work the following year, the smaller the benefit from guaranteeing the CTC. We estimate that among single mothers who worked in 2018, just 5.6% did not work in 2019, and among sole-earner parents in married couples who worked in 2018, just 1.6% did not work in 2019. Even if we assume all of the non-work in 2019 was involuntary, risk-neutral parents would value the insurance policy at only 5.6% or 1.6% of the CTC payment they guaranteed, and thus reduce our return to work changes by these percentages. Of course, parents are likely risk averse, and research suggests that workers may be willing to pay around a 60% markup for Unemployment Insurance. But even this would reduce our return to work changes by only 9.0% for single moms and 2.6% for sole earner parents in married couples. This would translate into a less than one-tenth reduction in the number of parents exiting employment due to the one-year lookback provision.

Do We Overstate the Sensitivity of Work Decisions to Changes in Incentives?

Finally, a key set of assumptions in our analyses involves the extent to which people will stop or start working when the reward to working changes. These assumptions are expressed in the “labor supply elasticities” that we use to get our estimates of employment declines and increases. We have shown elsewhere that our elasticities are firmly in line with the results of 30 years of research on the topic.

Our estimates of the impact on employment of the Wyden-Smith look-back provision remain the only ones available. Our critics have offered objections that we do not believe alter our conclusion, and they have tended to assume that any employment effects are negligible. We have stated the assumptions embedded in our modeling transparently, and we believe they are backed by evidence. It is essential to evaluate proposed policies rigorously in order to determine the best ways to build a safety net that reduces hardship while still encouraging upward mobility. Weakening these aspects of the safety net could prove harmful to children in the long run, and we and our colleagues have offered a number of alternative antipoverty policies that would take seriously this risk.