Interview Summary

You two were part of a research team that also included economists from MIT, Harvard, and the University of Maryland and recently published an analysis of the effects of SNAP work requirements in the American Economic Journal: Economic Policy. Adam, before we jump into the study, would you tell our listeners what SNAP work requirements are and how proponents and critics see them?

Adam –  Sure. So, SNAP’s work requirements dictate that some recipients must be working, training, or volunteering to receive benefits for more than a few months. So the policy is historically applied to childless adults who are younger than 50 and don’t have a disability, and as you mentioned, this group is often referred to as able-bodied adults without dependents and called by their acronym ABAWDs.

So, the work requirement stipulates that ABAWDs must be working 80 hours per month, participating in a qualifying training program or volunteering. If they don’t satisfy that requirement, then they’re only entitled to three months of SNAP benefits within a three-year period. Counties with high rates of unemployment may temporarily be granted exemptions from the policy by USDA’s Food and Nutrition Services as well. So in terms of kind of the proponents and opponents of the policy, those in favor of work requirements are concerned that providing government benefits discourages work. These people generally believe that those recipients of government assistance should work as a condition for that help. Their argument for the policy is that by incentivizing work, people will develop a stronger attachment to the labor market. Eventually they become self-sufficient, and they can earn enough to get by without the government assistance.

By contrast, the critics see work requirements as a policy that prevents people who are economically vulnerable from receiving food assistance, and it prevents them from accessing assistance in times when they need it the most. So, the opponents argue that if the reason people are not working is something other than the economic incentives of the policy, then the work requirements are really just going to cut people’s benefits without getting more people to work. It’s worth noting that work requirements have long been contentious and most recently were the main issue in the debt ceiling debate. So as part of that compromise, the SNAP work requirements will gradually be applied to ABAWDs up to age 55, but additional groups will now be exempt from the requirement, including veterans, the homeless, and those who are 18 to 24 who were previously in foster care.

I really appreciate how you’ve brought in the changes to the rules around work requirements under the Inflation Reduction Act. This is going to be an important policy discussion as we move forward. It’s interesting, particularly the exemption for individuals who are unhoused because of the work that shows up in the paper that you all have. So Elena, let’s now turn to you. How did you study the effect of work requirements in this paper?

Elena – As you might imagine, studying the effect of work requirements on SNAP recipients and potential recipients is actually a little bit tricky. You have to be able to find comparable groups of people, some of whom face work requirements in SNAP and some of whom don’t, but are otherwise similar enough to one another that it would make sense to compare their employment trajectories and their use of SNAP. Generally speaking, you don’t just directly want to make that comparison across people to whom work requirements apply versus people to whom work requirements don’t apply in SNAP because they’re pretty different groups. The ones who face work requirements tend to be younger, tend to not have disabilities, and tend to not have dependents like children, especially, in the household. As you can imagine, if you look at the folks who are exempted from work requirements under typical SNAP rules, there are lots of reasons why they may be less likely to hold a job than the ones who do face work requirements. They may be older, or they may be disabled. So, you can’t just attribute that difference to the work requirements themselves, which means in order to study the effect of the work requirements, you have to go looking for a context where the people who face them and don’t face them are much more similar to one another than that.

What we ended up doing is looking at essentially the same set of people as work requirements policies changed over time. In fact, what you end up doing is comparing the same person before work requirements exist to the same person after work requirements exist. Now, you might say, “Okay, but work requirements have been part of SNAP for decades. Where did you find a time when they didn’t exist?” The answer is that temporarily, many states suspended work requirements for a few years starting in the Great Recession when the job market was very weak and it was just considered too hard for people to find jobs even if they were trying. States were attempting not to be punitive during a very difficult job market. One of those states was Virginia, and so we got very detailed data on both SNAP use and employment and earnings from employment from the State of Virginia. Then we were able to trace what happened to SNAP recipients when that work requirements suspension expired in the fall of 2013. Virginia then put work requirements back into SNAP, which meant suddenly the same people who could use SNAP without work requirements just a month before were facing work requirements newly. We could compare how their employment responses changed to the changes of the employment responses of other groups who continued to be exempt from work requirements, for example, because they were just a little bit older than 50 when work requirements came back. We followed both of those age groups, comparing whether they were working and whether they were getting SNAP for several months and years after work requirements were reintroduced by the State of Virginia. That allowed us to both solve this problem that it’s generally very hard to make comparisons across groups. In this way we were comparing sort of across groups but also within the same person.

In addition, the nice thing about designing the study this way is that we got to study not just those people who got on SNAP knowing that they would face work requirements and so thinking it was worth all the administrative hassle to get on SNAP in the first place because they expected to be able to meet the work requirements and continue to receive SNAP. But we were also able to study those people who knew at the time that they got on SNAP that they probably wouldn’t be able to meet work requirements if those work requirements existed. But because work requirements were suspended at the time, that actually didn’t discourage them from signing up in the first place because they knew that they were going to be able to stay on SNAP for a longer time while work requirements continued to be suspended.

Great, thank you. I realize this is a critical part of your study to be able to find the exact cohort and the fact that you were able to follow those individuals over a period of time compared to what some other studies have done, where they look at the sort of a sample of people that changes on an annual basis or that’s more cross-sectional. I really appreciate the great care that you all put in to discerning who that sample should be, and that offers up really clean ways of understanding the effect of a ABAWD work requirements, so thank you for that. Adam, what did your research show about who’s right about the effect of work requirements? Can you help us understand that debate a little bit more carefully given the findings of your results.

Adam – Let me begin by kind of summarizing our main results. So, we found no evidence that work requirements led more people to work. The same number of SNAP recipients worked whether work requirements were in effect or whether they were not, and the large size of the administrative data that we had access to allowed us to be very confident from a statistical perspective that any effect on employment that’s positive is likely to be extremely small.

We also didn’t find that earnings increased on average either. However, what we did see was a substantial drop in the number of ABAWDs enrolled in SNAP. We found that work requirements cut enrollment by more than half among ABAWDs, and this is a dramatic reduction in food assistance. This drop in SNAP enrollment was largely driven by people who were already on the program when the work requirements turned back on in 2013, as Elena was just describing, but we also found that many people who were newly enrolled exited the program sooner, and then some potential SNAP recipients did not enroll at all compared to what we would’ve expected in a world without work requirements. So, putting all of that evidence together, overall we found no evidence in support of the arguments that are made by those who favor work requirements.

Great, thank you. Elena, do work requirements have any other effects on potential SNAP recipients?

Elena – Well, Adam already covered the two primary study outcomes that we looked at and that policymakers usually refer to, which is are people staying on SNAP and continuing to receive benefits and are people working? But in principle separate from just getting a larger number or a larger fraction of people working, economic theory might predict that work requirements could increase income, either through the amount of work or the hourly wages among the SNAP recipients who were going to be working anyway, right? We find no effect on who’s working, in other words, how many people are working. But in principle you might say, “Okay, maybe those who were going to be working anyway are now more attached to the labor force. They’re working more hours or more weeks out of the year than they would’ve been absent the work requirements.” As Adam said, we actually on average did not find evidence that work requirements increased incomes, and so we can’t really say that work requirements improved labor force attachment on this dimension. We ran this analysis lots of different ways, and in the vast majority of our analysis versions, we found just no change in income. But if you were very motivated to cherry pick a couple of analyses in support of work requirements and ignore the totality of the evidence in the paper, then you can find a couple of versions where there’s sort of suggestive evidence that a small fraction of potential SNAP recipients, maybe something like 10 to 15%, might have had some income increases that could be attributable to work requirements.

Now, I say that you would have to ignore the totality of evidence in the rest of the paper if you wanted to run with those results because it really only was a couple out of very many analyses, and the results even there were quite weak, especially compared to the very stark findings of no effect on whether people are working and very large negative effects on people’s continued SNAP receipt.

Elena, thank you for that. Do you see any differences by subgroups? I realize that’s not maybe a part of the paper as it’s written, but knowing that different subgroups have greater unemployment, do you see any differences, say among racial or ethnic lines or along gender lines?

Elena – Well, we essentially couldn’t find effects on whether people were working as a result of work requirements for any subgroup. We are somewhat limited in our ability to do subgroup analyses because of issues like sample size and statistical power, but to the extent that we were able to cut the data, we didn’t see impacts on whether people were working for any of these subgroups, which is actually very different from what we see with people losing access to SNAP. So, we saw quite a large disproportionate impact on use of SNAP, meaning people’s ability to stay on the program and continue to receive food assistance for individuals with a history of homelessness. Further, here was some suggestive evidence of disparate impacts along racial lines, but that wasn’t statistically strong enough for me to feel comfortable claiming anything about it.

Adam, why do you think work requirements do not have much of an impact on work?

Adam – Our results suggest that the SNAP recipients that we studied likely face other barriers that are more important for employment than the work requirements policy itself. As Elena just mentioned, we found this disproportionate effect in terms of people who lack stable housing, and that can make it really difficult to hold down a job, and to be able to apply for jobs. People may also not have reliable or affordable transportation. So, without those things, you can see how it’d be very difficult for people to maintain gainful employment. Another possibility is that people’s hours may fluctuate in ways that make them ineligible in terms of meeting the requirements of 80 hours a month, and so several of those explanations come from other studies of enrollees in different safety net programs in various states. One great thing about the data we had is that we could track people’s earnings and participation in SNAP over a long period of time, but we don’t see information on how many hours they worked, for example, or what their transportation options were. Those are issues, though, that many others have noted as being important in different contexts, both using quantitative studies as well as more qualitative studies. We think this is something that future research should really focus much more on – how to quantify the importance of those different barriers that ABAWDs likely face, and then trying to figure out how to best design solutions that address them.

This does make me think about another potential issue, and that’s the benefits cliff. So, if someone is working and there is an increase in their wages, there actually could be a drop in their benefits that they receive from SNAP, and that could make this story a lot more complicated. Elena, I do have this question for you. How did you handle individuals who dropped out of having the ABAWD status, say if they had children or if a disability came up? Was this an important factor in evaluating that subpopulation?

Elena – For the validity of the study results, it was very important for us to keep a consistent sample throughout the time period that we were evaluating these outcomes over. This meant that if people were eligible for our sample definition at the beginning, then we kept them in the sample for the rest of the time period. That means that there were some people who dropped out of ABAWD status for various reasons, and actually, if I recall correctly, and Adam can jump in if this is not right, the most common way that people in the sample dropped out of ABAWD status is by having a newly documented disability. Our understanding is that what’s happening with those folks is that many of them might have been eligible for a disability documentation that would’ve made them exempt from work requirements even during the time that Virginia had suspended work requirements. But there was essentially no point in their going through the hassle and their caseworkers going through the hassle of getting that documentation because the work requirements didn’t apply to them anyway. So, when the work requirements came back, some at least of those folks got their disabilities officially documented so that they were, again, exempted from work requirements. This illustrates why it’s important for us to keep a consistent sample before and after the work requirements come back, because we want to make sure that we’re not sort of changing who’s in the comparison group, right? You don’t want apples to suddenly turn into oranges halfway through your study period.

Adam – Elena’s description was exactly right, and in terms of the magnitudes of that response – in terms of people who were exempt for a reason other than their age – it basically doubled the proportion of people who dropped out of ABAWD status. On average, about 10% of people had some exemption besides age from ABAWD status, and then we saw that the policy increased that by 5.6 percentage points, so a pretty large relative change.