https://substack.com/app-link/post?publication_id=492324&post_id=201217544&utm_source=post-email-title&utm_campaign=email-post-title&isFreemail=true&r=rovhk&token=eyJ1c2VyX2lkIjo0NjUxMDE4NCwicG9zdF9pZCI6MjAxMjE3NTQ0LCJpYXQiOjE3ODEwMDcwNzMsImV4cCI6MTc4MzU5OTA3MywiaXNzIjoicHViLTQ5MjMyNCIsInN1YiI6InBvc3QtcmVhY3Rpb24ifQ.kwnHyzGqjHD15zr4vvGIdpR42fprnEEwS8iGulwFBtM
~~ recommended by newestbeginning ~~
Cherry-picked studies and unrealistic assumptions define a new government report
Last week, the Department of Health and Human Services, under the Office of the Assistant Secretary for Planning and Evaluation (ASPE) released a new “study” on the effects of work requirement policy on employment and poverty. You might not have heard of it, but it represents a potentially important step in the federal government pushing a policy that will hurt many vulnerable people. At a time when the administration appears to be at war with social science, why did it need this report?
To answer this question, we looked more closely. We are academics at the University of Colorado and Harvard who study administrative burdens in the social safety net, particularly the effects of work requirements policies. We are gravely concerned that this ASPE study misrepresents the state of scholarly evidence on the impacts of work requirements and that this will lead to poorly informed and punitive policy decisions in the future. So, we are writing this post to set the record straight.
The report makes two big, dubious claims:
“Rigorous studies find that work requirements are most likely to increase employment when they combine clear expectations with supports that reduce the cost of finding and maintaining work.”
“Under varying conditions, including job availability and implementation quality, the policy could reduce poverty by 1.6 to 2.9 million people.”
The report comes as the Centers for Medicare and Medicaid Services CMS unveils regressive new policies on work requirements. It seems designed to provide a patina of intellectual credibility to a policy that most researchers are dubious of. In a press release, CMS toutedechoed the findings of this study to arguestipulate that Medicaid work requirements — which states must impose upon enrollees who have coverage through the ACA’s Medicaid expansion by January 2027 — “could reduce poverty by as much as 2.9 million people.”
To reach these conclusions, the ASPE report briefly summarizes select literature on the efficacy of work requirements. ASPE commissioned a more comprehensive — but still incomplete — literature review from the contract research organization Mathematica, which is linked as a technical appendix.
The report then offers two estimates of the potential impact of Medicaid work requirements on poverty. The assumptions embedded in these models are not supported by either a generous reading of the report’s own literature review, or by our much more comprehensive literature review here.
Glaring omissions from the literature review
We see three big omissions from this ASPE report:
Medicaid work requirements were already attempted in the first Trump administration and the research is clear that these work requirements had no positive effect on employment and made participants worse off.
The study cherry-picks only a small amount of the full literature on the effects of work requirements and *happens* to only discuss papers that find work requirements to have a positive effect on employment when the literature actually tilts the other way – most studies find no effects of work requirements on employment.
One consistent finding across dozens of papers on work requirements is that they reduce program participation among vulnerable individuals and households. This finding is completely ignored.
Omission 1: The biggest omission from the ASPE Study is any mention or discussion of the most directly related research — studies of the last time Medicaid work requirements were attempted, which occurred in the first Trump administration. In 2018, the administration proposed to let states use demonstration waivers to opt into Medicaid work requirements, which had never been part of the program’s design. But, these waiver-based work requirements were only ever implemented in this period in one state: Arkansas.
Other states had planned to adopt similar policies but were stymied by litigation on the issue (or, in the case of New Hampshire, suspended implementation due to large anticipated coverage losses). This single-state implementation provided researchers a nice “natural experiment” they could use to study the effects of this policy by comparing changes in outcomes of Medicaid participants in Arkansas to other states in this time period.
The Arkansas work requirements were first enforced in June 2018 and were in effect for less than a year before a court suspended the policy; but, this period provided sufficient experience for a small and consistent evidence base. A novel survey fielded in late 2018 found considerable confusion among enrollees about whether the new requirements applied to them and documented significant coverage losses. These coverage losses also substantiated subsequent research using Census data.
The findings of these studies were consistent with the state’s own administrative data. Arkansas had been able to automatically exempt or qualify (that is, deem in-compliance) about two-thirds of people potentially subject to the requirements, but the majority (about 75%) of the people who needed to report on their compliance or exempt status failed to do so; ultimately about 18,000 adults lost coverage. This happened despite the fact that Arkansas used a purely attestation-based model; no one had to produce pay stubs or doctor’s notes — the types of documentation that CMS will require states collect when they can’t use data they already have to exempt or qualify an enrollee.
A follow-up to the first survey, fielded in late 2019, found that Arkansans who had lost Medicaid while work requirements were in effect commonly reported serious problems paying off medical debt and delaying care and medication due to cost.
Importantly, all three peer-reviewed studies consistently found that there was no statistically significant effect of work requirements on employment by those likely to be subject to the policy. In part, this is likely because the overwhelming majority of Medicaid enrollees either already meet the work requirements or meet one of the enumerated exemptions.
In Arkansas, the 2019 follow-up survey found that 95% of the target population was already meeting the requirement in some way. The study using Census data, which had more statistical power to detect changes (due to a larger sample size) similarly found no effect. While the Congressional Budget Office often doesn’t unpack their analyses as much as analysts would like, it is telling that they stated in a June 4 letter to members of lawmakers that “[f]ew of those disenrolled from Medicaid because of the [work requirements] policy would have access to and enroll in employment-based coverage,” suggesting that even Congress’s scorekeepers don’t believe that the policy will move meaningful numbers of people into good jobs with robust benefits.
Despite vocal championing by the first Trump administration, these earlier, opt-in work requirements were ultimately jettisoned by the courts because they were deemed at odds with the core purpose of Medicaid: providing medical care to the needy. In court opinions, the administration was taken to task for failing to consider whether and how work requirements would adversely affect coverage. Simply put, the judge held that if Medicaid’s statutory reason for existing is to provide health insurance to low-income people, there was no way to understand work requirements as advancing that goal. (The reason work requirements are permitted — indeed, required — now is that the underlying statute was changed).
Omission 2: The second omission from the ASPE report is a fuller literature review beyond just the experience of Medicaid work requirements in Arkansas. This study focuses only on evaluations of work requirements under the TANF (Temporary Assistance for Needy Families) program and housing benefits. Many of these evaluations were conducted in the 1990s, impacted a different population than those who will be affected by Medicaid work requirements, and the work requirements themselves were very different from those being carried out under Medicaid right now.
“Remember the 90s” is a great policy if you are trying to evoke millennial nostalgia, but not a great ethos for an evidence-based public policy.
For example, Dr. East has worked since fall 2023 to create and update a summary of the literature on work requirements in SNAP with Dr. Lauren Bauer at the Hamilton Project at the Brookings Institution. The studies summarized here focus on a much more similar population, time period, and policy design as the current Medicaid work requirement expansions, relative to the literature reviewed in the ASPE Study. The authors summarize their findings as follows:
Our conclusion from a review of the literature on work requirements is that the best evidence shows they do not increase employment. Moreover, this research finds work requirements cause a large decrease in participation in SNAP. This is concerning because many SNAP recipients, especially those subject to the time limit work requirements, have little safety net to rely on besides SNAP.
Neither this summary, nor the papers it reviews, are included in the ASPE primary report. The technical appendix includes some more contemporary work reviewed by Bauer and East, raising the question of why such work was not of enough interest to the authors to make it into the main report.
Omission 3: The final omission is a lack of discussion of the impacts of work requirements on program participation, which the literature is quite clear about. Work requirements can reduce participation for a number of reasons besides increases in employment: people may want to work but can’t find a job that meets the strict hours requirement, people may satisfy the work requirements or be exempt from them but struggle with the red tape imposed by the work requirements to document work effort or exemption reasons.
The best causal studies of SNAP work requirements find participation drops of 15 to over 50% after work requirements are imposed. Experience with Medicaid work requirements is more limited, but a report by the Urban Institute and the Robert Wood Johnson Foundation projected disenrollment rates of 19-37% among working adults subject to the new requirements and more frequent eligibility checks. Recent work by one of the authors of this piece finds that the red tape alone explains a large portion of the disenrollment.
By reducing access to Medicaid, work requirements are likely to increase poverty. In 2024, Medicaid was the third most impactful anti-poverty program after Social Security and Medicare according to the Census Bureau. Recent estimates suggest work requirements could reduce Medicaid enrollment by 3-7 million people by 2028, driving the majority of coverage losses which could total up to 10 million. It’s hard to see how moving millions of people away from an effective anti-poverty program does anything other than worsen poverty.
 |
Flawed models out-of-step with the report’s own literature review
After conducting this flawed literature review, the authors claim to more concretely “estimate” the potential effects of the new Medicaid work requirements policy. They do so by modelling two scenarios under the strong assumption “that applicable adults who do not already meet the requirements (i.e. those not working or studying sufficient hours) increase their work hours to comply.”
The first scenario relies further on the assumption that “all eligible adults enroll, and immediately find adequate employment to meet the requirements.” This is a facially absurd approach; the model assumes sunshine and rainbows, so it yields sunshine and rainbows. Assumptions are important in modeling, but they must be at least grounded by reality if the model is to be taken seriously.
The second scenario attempts to be more realistic by adjusting for incomplete take-up of Medicaid, incomplete take-up of work, and other (less immediately apparent) factors. While not as fanciful as the first model, the resulting estimates are still out-of-step with the most recent and rigorous scholarship on work requirements. Neither model contemplates — or is designed to contemplate — how many working or exempt people might lose coverage because of new paperwork burdens.
In both unrealistic scenarios, the authors estimate that the average annual earnings increase would be $16,780 in households where adults (immediately, frictionlessly) respond to the new policy. This estimate is not only contradicted by the omitted literature described above, but also by the research cited in the report’s own technical appendix. The largest estimated effect of work requirements on long-term earnings included in the review was less than one-tenth of that size ($1,506 in 2018 dollars). This estimate is from a single 1994 cash welfare pilot in Minnesota that, for myriad reasons, might not be expected to generalize to Medicaid work requirements 33 years later.
Critically, the authors of the report — much more so than the authors of the CMS press release linking back to it — offer some high-level clarity about what their report is and what it is not, writing that their results “are simulated calculations, not necessarily what is or will occur in the real world.” Indeed.
Given the research evidence and policy details, we conclude that, in the real world, Medicaid work requirements are much more likely to worsen poverty than to reduce it.
No comments:
Post a Comment