As The Biden Administration Begins Unwinding Them, Medicaid Work Experiments Remain Unreasonable, Unnecessary, And Harmful

One month before the Biden administration took office, the United States Supreme Court agreed to hear the Trump administration’s appeal in Azar v Gresham and Philbrick v Azar. The consolidated cases focus on the legality of state experiments conducted under Section 1115 of the Social Security Act that condition Medicaid eligibility on work.

Given recent actions on the part of the new administration, whether the cases will proceed at this point is uncertain. But oral arguments are now scheduled for March 29. Should the cases move forward, the Court will consider the ultimate question of whether 1115 in fact authorizes the Department of Health and Human Services (HHS) to approve experiments that compel work as a condition of Medicaid eligibility.

The Biden Administration’s Moves To Unwind Medicaid Work Experiments

On January 28, the Biden administration announced its intention to begin the process of rolling back the prior administration’s approvals of 1115 waivers authorizing Medicaid work requirements. On February 12, this process began in earnest when the Centers for Medicare and Medicaid Services (CMS) took down the original January 2018 CMS letter to state Medicaid directors soliciting compelled Medicaid work experiments. Accompanying this removal were letters from acting CMS Administrator Elizabeth Richter to both Arkansas and New Hampshire (both parties to the Supreme Court appeal).

Richter notified both states that the agency now “has serious concerns about testing policies that create a risk of a substantial loss of health care coverage in the near term,” given the situation created by COVID-19 and the public health and economic crises the pandemic has triggered. This concern will not end when the public health emergency declaration does given that “the uncertainty regarding the lingering health consequences of COVID-19 infections further exacerbates the harms of coverage loss for Medicaid beneficiaries.” For this reason, the acting Administrator concluded, “allowing work and other community engagement requirements to take effect in [New Hampshire and Arkansas] would not promote the objectives of the Medicaid program.”

In addition to notifying them that the agency is terminating its earlier approval, CMS notified New Hampshire and Arkansas that it was rescinding a January 4 letter from then-CMS Administrator Seema Verma to state Medicaid agencies operating 1115 demonstrations; that letter sought to significantly lengthen and complicate the 1115 experimental termination process. The new termination procedures are unnecessary under existing federal rules governing the 1115 termination process applicable to demonstrations found either in material non-compliance or “not likely to achieve statutory purposes,” the acting administrator said. She further concluded that “[b]y proposing to delineate additional procedures for the withdrawal or suspension of a demonstration approval . . . the January 4, 2021 letter did not address CMS’s need for flexibility to make and effectuate determinations under [existing federal regulations]. The current COVID-19 pandemic and economic environment, along with an ongoing need for general CMS oversight of section 1115 demonstrations, necessitate that CMS maintain the regulatory flexibility to respond appropriately to current or changed circumstances.” Presumably this rescission of the prior administration’s effort to modify the 1115 suspension or termination process will apply to all state demonstrations that received the January 4 letter.

The Administration Has Opened A Discussion With The Supreme Court; Awaiting The Court’s Response

Notably the CMS letters terminating approval of the Arkansas and New Hampshire can be viewed as a conversation of sorts with the Court, just as they are missives to the individual states. The administration has not yet filed information directly with the Court regarding its stated position in the case. But a careful reading of the letters suggests where its thinking may lie.

In its letters, the Biden administration does not argue that HHS broadly lacks the power to approve such experiments; the administration’s position instead is that current circumstances legally prevent the agency from making the requisite finding on which the exercise of 1115 powers rests—namely, that the experiment will “promote the objectives” of the Social Security Act program that is the subject of the experiment. By not resting its decision on the ultimate, and abstract, question of agency power under 1115, the administration appears to be signaling two things to the Court. First the ultimate question of 1115 powers is inevitably a contextual one, driven by innumerable factual circumstances that must be weighed in order to determine when, and under what conditions, the HHS Secretary is justified in finding that any particular experiment will promote Medicaid objectives. Second, there simply can be no question that given the present circumstances—a pandemic, a public health emergency, economic devastation, unprecedented need, and a federal law (the Families First Coronavirus Response Act) that bars states from reducing eligibility as a condition of receiving enhanced federal Medicaid funding—now is not the time to pursue any demonstrations that carry the health risks documented by compelled work experiments.

We do not yet know what the administration will formally communicate to the Court, nor what the Court will do. But these letters offer clues. Should the Justices decide to go forward, however (which they could well do given the fact that Arkansas and New Hampshire both are parties to the case and their demonstrations have not yet been officially terminated) the Court could decide that work experiments are in fact lawful, thereby giving future administrations the power to require work as a basic condition of eligibility. Thus, the outcome of Gresham and Philbrick remains enormously important, since the Court’s decision could go well beyond the procedural grounds on which the unanimous decision by Court of Appeals for the DC Circuit rested.  As one of us recently has written, the Court appears to have done an unusual thing on appeal, posing a question broader than the one considered by the lower court.

The Nature Of 1115 And The Authority It Confers

Whatever the Court eventually does, answering the broader question the Court has posed means considering the nature of Section 1115 itself. The law does not empower HHS to run alternative versions of Congressional programs. Instead, 1115 is an experimental authority. As such, the power granted under 1115 is tied to the reasonable exercise of basic research norms: the development of hypotheses based in evidence, sound research design that allows hypotheses to be tested through research methods, and evaluation. These norms are found in the law itself, which requires evaluations, as well as in implementing rules.

The reasonableness of a research hypothesis is tied to whether the assumptions and evidence on which the hypothesis rests are themselves reasonable. In the case of Medicaid compelled work experiments, this means examining the evidence available to CMS when in January 2018 the agency hypothesized that compulsory work—20 hours per week, either paid or voluntary—would improve the health of working age adults. It is also worth considering what we have learned since that time.

The Evidence That CMS Ignored When Inviting Work Requirement Demonstrations

Following Congress’s repeated refusals in 2017 to enact Medicaid work requirements as an option for states, the Trump administration began to formulate a policy of demonstrations involving compelled work as a condition of eligibility for nonexempt people. Its policy was announced by then-CMS administrator Seema Verma in November 2017 at a state Medicaid directors conference. On January 11, 2018, CMS issued a State Medicaid Directors Letter describing how this new policy could be tested as an 1115 experiment. As of that time, several states already had submitted experimental proposals.

By the end of 2018, the administration had approved work requirement experiments in seven states. The White House further emphasized its enthusiasm for work requirements in April 2018 through an executive order requiring agencies to expand them in other benefit programs as well.

To approve Medicaid and Children’s Health Insurance program (CHIP) demonstration projects, federal regulations require states to identify goals of the demonstration, relevant research hypotheses that will be examined, and plans to evaluate the projects. The Secretary can then approve demonstrations, if, in his judgment, they are “likely to assist in promoting the statutory objectives of the Medicaid or CHIP program.” The hypotheses CMS and states have identified for work experiments have shifted over time, but when CMS first invited work requirement experiments, improving the “health and well-being” of those who would theoretically gain work was the clear focus. For example, CMS’ approval of Arkansas’ program in March 2018 asserted, “The demonstration is likely to assist in improving health outcomes through strategies that promote community engagement and address certain health determinants.”

But the evidence showed that the hypothesis itself was not reasonable, and thus, testing whether work requirements for Medicaid would improve health was not reasonable research. By the time CMS released its January 2018, it was clear that compelled work does not lead to meaningful improvements in employment, nor does it improve health. Furthermore, by January 2018 a substantial body of evidence showed that the predictable harm from large-scale loss of access to health care would vastly outstrip any possible benefits associated with increased income for a few. The Trump Administration disregarded this body of research, proceeding as if it did not exist, thereby placing the entire undertaking outside the scope of reasonable social welfare experimentation and experimental norms.

Evidence From Other Programs

Much early research came from welfare-to-work experiments, especially cash assistance experiments in the 1990s and early 2000s funded by HHS. Evidence of the harm associated with work requirements in the Supplemental Nutrition Assistance Program (SNAP) was also available before 2018. The Medicaid work requirements endorsed by the CMS were based on those implemented in cash assistance and SNAP, and CMS pointed to no evidence showing that outcomes would be any better for Medicaid. Indeed, the Temporary Assistance for Needy Families (TANF) and SNAP work requirements included funding for employment training or work supports (such as child care), but CMS refused to permit Medicaid funding to be used for such supports; thus, the Medicaid work requirements were, even at their origin, less likely than their TANF and SNAP counterparts to help promote employment or health outcomes.

In August 2017, the Cochrane Collaborative published a review based on an analysis of 12 randomized controlled trials—the gold standard of research—of welfare-to-work initiatives, such as work requirements, incorporating 27,482 participants. The studies reviewed by the Collaborative, a non-partisan organization that conducts rigorous systematic reviews of research regarding the effectiveness of medical and public health interventions, were completed between 1994 and 2011. The review concluded that welfare-to-work programs have no meaningful effects on the mental or physical health of either parents, who were subject to the requirement, or their children. The programs also failed to produce long-lasting effects on employment or incomes of recipients. The authors theorized that the lack of meaningful impacts on health might have been due to the lack of lasting changes in income.

Similarly, a comprehensive review by LaDonna Pavetti, published in 2016, indicated that work requirements created, at best, short-lived, modest increases in employment and almost no improvements in the poverty status of program recipients. Indeed, deep poverty often increased because so many recipients lost benefits. In 2016, the Congressional Research Service concluded that experiments testing the effectiveness of mandatory work requirements for cash assistance “did not produce evidence that welfare-to-work programs per se could reduce poverty, as often incomes of participants remained well below the poverty line.”

By 2017, it was obvious that the main consequence of work requirements was a reduction in the number of people receiving assistance. Even without reviewing research studies, federal officials could have simply checked administrative caseload data from the U.S. Department of Agriculture (USDA) for states that reinstated work requirements in SNAP for “able-bodied adults without dependents” ages 19 to 49. In SNAP, these adults are subject to termination if they do not work 80 hours per month and are not exempt due to disabilities that prevent them from working. Federal law permits states to pause work requirements during challenging economic conditions.

For example, when Arkansas reinstated SNAP work requirements after September 2015, SNAP participation fell by 58,894 persons (12.8 percent) in one year. (These data also include parents and children who were not subject to work requirements; the reduction for non-parent adults was much larger.) In comparison, neighboring Louisiana did not impose work requirements then and its SNAP participation rose by 18.9 percent in the same period, based on analyses of administrative data.

In addition, CMS could have considered what happened when work requirements were first introduced in SNAP in 1996. In 2002, a USDA synthesis of studies in four states found that although many of those who had left food stamps were employed at some level, “relatively large portions of [able-bodied adults without dependent] leavers in these studies were poor and food insecure with hunger,” indicating that the work requirement had not achieved its aim of self-sufficiency.

Undermining The Positive Effects Of Medicaid Expansion

Evidence available before January 2018 also indicated that expansions of Medicaid were improving health outcomes, so that loss of coverage was likely to harm health rather than improve it. Medicaid expansions had been associated with a wide range of improved health and health care access outcomes in studies published at that time, including reduced rates of mortality. By the end of 2017, research studies indicated that Medicaid expansions led to better access to general medical care, preventive health care (without increasing risky behaviors), opioid abuse treatments, and cancer care, as well as reductions in racial disparities in care.

By the time CMS announced its demonstration project guidance, clear and compelling evidence showed that work requirements in other programs produced substantial harm by reducing benefits but had almost no effects on recipients’ health or economic status. CMS and the states applying for demonstration waivers chose to ignore those findings to advance an unfounded storyline that work requirements would improve “health and well-being,” ignoring predictable consequences for the large number of people who would lose Medicaid coverage.

As for evaluation, there was none other than the landmark, privately funded evaluation undertaken by Dr. Benjamin Sommers and colleagues, discussed below. Indeed, even seven months into a demonstration that cost thousands of eligible people their coverage until it was halted by the courts, CMS and Arkansas still had failed to put an evaluation plan into place. The evidence suggests that to this day, no evaluation design exists.

Recent Evidence On Work Requirements

Since 2018, further research has continued to demonstrate the harm of work requirements in Medicaid and SNAP, while CMS has produced no empirical evidence of benefits. Although CMS approved numerous Medicaid work requirement waiver projects, only Arkansas was actually implemented to the point that beneficiaries were required to demonstrate their work activities or lose benefits. Arkansas implemented its program from June 2018 through March 2019. Implementation was halted by federal court decisions that invalidated CMS’ approval; the Supreme Court is now considering the case. Similar cases led to invalidation of Medicaid demonstration projects in other states before they were fully implemented.

The Evidence From Arkansas

Benjamin Sommers and his coauthors examined the brief use of Medicaid work requirements in Arkansas. Their conclusions reflected the previous evidence. A 2019 analysis found that work requirements led to a large reduction in Medicaid coverage and an increase in the ranks of those uninsured, without significant improvements in employment. A 2020 follow-up study continued to find no improvements in employment among those 30 to 49 years old, the group initially subject to the requirement. Meanwhile, work requirements caused a 7.1 percentage point increase in uninsurance and a 13.2 percentage point drop in coverage via Medicaid or the Marketplace.

A large share of those who lost Medicaid experienced problems accessing health care, such as difficulties paying medical costs, delays in taking prescription medications due to cost, and delays in obtaining health care. Sommers found that 95 percent of those subject to work requirements were either already meeting them or eligible for an exemption, implying that many lost coverage due to being stymied by complex reporting requirements.

More Evidence From SNAP

A larger body of research about SNAP work requirements complements these findings. Several studies have examined what happened to employment outcomes when work requirements for SNAP were stopped or restarted as economic condition changed. During the Great Recession, Congress halted work requirements nationally for much of 2009 to 2010. Most states gradually reinstated work requirements in subsequent years.

The research about SNAP work requirements demonstrate that the failures found in Arkansas’ Medicaid program were not the result of start-up jitters or chaotic implementation in one state. Multiple studies conducted using data from states across the country, collected over many years, find harmful effects of work requirements on participation and little or no benefit for employment.

Ku and Brantley published two peer-reviewed studies that demonstrate the harm of SNAP work requirements. One study, based on analyses of administrative data, found that SNAP work requirements consistently caused over one-third of those subject to the requirement to rapidly lose benefits when they were re-introduced between 2013 and 2017.

The other study examined data from the Census Bureau’s American Community Survey and found that, although SNAP work requirements are targeted at “able-bodied,” non-disabled adults and exempt those unable to work due to disabilities, implementation actually led to drops in participation for people who report having a disability That is, although work requirements were supposed protect those with disabilities, in reality many low-income disabled people lost benefits, contrary to the legislative intent. It is therefore not clear that similar exemptions in Medicaid work requirement programs would effectively shield those with disabilities. This study also found that SNAP work requirements have larger impacts on Black compared to non-Hispanic white adults, demonstrating that work requirements can exacerbate racial inequities.

Several additional studies have examined the consequences of SNAP work requirements across the nation using rigorous econometric methods. For example, Jeehoon Han (2019) found that suspending work requirements had no significant impact on whether adults were working and led to a very small (1.3 percent) decrease in average number of hours worked. A similar study by Brian Stacy and coauthors (2018) found no impact on either whether adults were working nor on the number of hours worked.

Timothy Harris (2019) found a marginally significant, small (0.8 percent) increase in employment for male workers only, with no effect for female workers. Harris estimated that for every five people who lost SNAP assistance when the work requirement was imposed, only one gained employment. Finally, Gray and associates (2020) found that SNAP work requirements led to a 52 percent reduction in participation but no appreciable increase in employment or earnings; they also found larger drops in participation among homeless enrollees and those with very low incomes.

FGA’s Arguments Supporting Work Requirements Based On Flawed Evidence

Echoing familiar arguments, the conservative organization Foundation for Government Accountability (FGA) filed an amicus brief to the Supreme Court in Gresham claiming that work requirements helped “millions of individuals out of the welfare trap,” a statement freighted in bias and one that assumes people will not work unless forced to do so while characterizing means-tested assistance programs as inherently bad. To back up its claim, the FGA ignored independent, high-quality research and primarily relied on its own reports, which asserted that employment improved when work requirements were imposed. Critically, these reports lacked comparison groups that examined areas or times without work requirements, a fundamental element of sound research.

Because it ignored the fact that employment tends to be cyclical, especially for low wage earners paid on an hourly basis, FGA utterly failed to demonstrate that employment gains were better when work requirements were used than when they were not in place. Equally important, FGA’s evidence ignored the harms caused by work requirements that penalized people caught in this cycle, stripping them of access to basic benefits when they hit one point of the cycle.

Work requirements are premised on the assumption that many who receive benefits choose not to work. As Ku and Brantley noted in March 2017, the reality is that most adult Medicaid beneficiaries work. However, for low-skill adults, work is often fluctuating or seasonal, meaning many people who are working may not consistently meet work requirements. The COVID-19 pandemic has put the vagaries of employment for low-skills workers into sharp relief: many, particularly those in certain industries like the restaurant or hospitality sector, have lost work and need assistance.

Over many years, and in study after study, researchers have carefully analyzed such policies using more rigorous methods, including randomized experiments and sophisticated econometric studies. We therefore know—and we knew in 2018—that work requirements lead to steep participation declines with no measurable gains and much exposure to harm.

The Wrong Prescription

The desire to help unemployed people find work is understandable. But this does not make reasonable experiments that condition health care on work and that fail to provide adequate training or supports, like child care or transportation, that might help unemployed people find or keep jobs. Indeed, CMS explicitly stated that Medicaid funds could not be used to pay for training or related supports for the unemployed.

Recent evidence confirms that work requirements do nothing to address barriers to consistent employment, while imperiling the health of low-income adults and families. The hypothesis that health would be improved by taking away health care of those who did not meet work and reporting requirements was driven not by evidence or reasonable assumptions but instead by ideology. The experiments contradicted evidence readily available at the time, and more recent evidence continues to refute them.