Any way you look at it you lose: Medicaid work requirements will either fall short of anticipating savings or harm vulnerable beneficiaries

On May 22, the House of Representatives passed legislation intended to meet reconciliation targets. The largest single component of savings in that legislation is a requirement that so-called “able-bodied” adults without dependent children engage in work or other community activities for 80 or more hours per month. The Congressional Budget Office (CBO) estimate for this measure indicated that this provision would generate more than $50 billion in annual savings by the end of the ten-year budget window. The legislation as a whole is projected to cause around 11 million people to become uninsured, among whom about 5.5 to 6.3 million are projected to lose Medicaid coverage through the work requirement.

Proponents of the work requirement provision have set out two notable objectives. The first is generating federal savings to partially offset the tax cuts included in the same bill. Second, proponents argue that the provision has a moral component—”People that are able-bodied, mentally healthy and all that should be working if you’re working age,” said one representative. As Speaker Mike Johnson put it, “You return the dignity of work to young men who need to be out working instead of playing video games all day.” Those who are able to work and do not do so are “defrauding the system.” In its design, the provision embodies this moral view: it explicitly excepts from the requirement (at section 44141(9)(A)(ii)(V)) populations whose absence from work is deemed acceptable, such as caregivers, people with substance use disorders, those with disabling mental health conditions, and those with impairments affecting activities of daily living. 

As we show below, these goals—reducing the deficit and sanctioning only purported slackers—are at odds. If the requirements work as intended, targeting only able-bodied adult Medicaid enrollees not engaged in the desired activities, they will save the federal government far less money than the CBO cost estimate projects. Speaker Johnson’s archetypal young men who hang out in basements playing video games are not as common as he may imagine, and just don’t use a lot of health care services. More generally, non-working Medicaid beneficiaries without activity limitations account for very little Medicaid spending and thus, disenrolling this group would generate only modest federal savings, far less than needed to offset a significant share of the bill’s tax cuts.

That means that if the provisions are to save substantial federal money, they will have to disenroll those they promise to protect: either by disenrolling many more low-cost beneficiaries, drawing from the large group who work more than the 80 hours a month threshold, or by disenrolling much more expensive beneficiaries, drawing from the group who report activity limitations but have not yet qualified for Supplemental Security Income (SSI). Both these scenarios have significant negative consequences. Beneficiaries who are already working and are disenrolled are likely to become uninsured, losing access to health care services that many need to remain healthy and to keep working. Beneficiaries with activity limitations have very high rates of utilization of mental health and substance use services; a significant minority are taking medications used to treat psychosis or bipolar disorder. Withdrawing care from this group may save money, but it will surely harm this very vulnerable group and is very unlikely to promote community engagement.

Our analysis

We use data from the 2022 Medical Expenditure Panel Survey (MEPS) (the most recent available) to estimate how many enrollees may be subject to Medicaid work requirements. Because the 2022 panel was surveyed during the public health emergency continuous coverage period, the total count of Medicaid beneficiaries is somewhat greater in these data than would be the case in subsequent years. That means that our estimates of the savings that would be generated by ending coverage for any given group of enrollees are likely larger than would be the case today. For most of our analyses, we combine these data with the (inflation-adjusted) 2019 (pre-COVID) MEPS survey to add sample size and report averages adjusted to 2025 dollars. We repeat our analyses separately for the two data sets and find that patterns within the population are very similar. Using the combined dataset, we assess Medicaid spending patterns among enrollees (who are not parents, on SSI/Medicare, full-time students, or pregnant) who are working more than 80 hours per month, those working fewer than 80 hours per month who report an activity limitation, and those working fewer than 80 hours per month who do not report an activity limitation. The latter group is the ostensible target of the provision. To compute federal savings, we calibrate our MEPS-based estimates to match CBO’s implied per expansion adult disenrolled federal savings in 2030 (computed by dividing the $44.5 billion in estimated savings that year by the Urban Institute’s estimate of 5.9 million adults disenrolled due to work requirements). 

Consistent with prior research, we find that almost all adult Medicaid beneficiaries who are not caregivers already work or attend school full-time, and most of those who remain have activity limitations. In 2022, 77.3 million people were enrolled in Medicaid at any time in the year. Of these, 36.7 million were adults between the ages of 19 and 64 (Figure 1). One-third of this group were either parents of dependent children (under age 7 or with special needs) (17% or 6.3 million), enrolled in Medicare or SSI because of an adjudicated disability (15% or 5.6 million), or were pregnant (1% or 0.5 million). The legislation tasks Medicaid agencies with identifying these groups using existing administrative data. The legislation also excepts those who are full-time students (4% or 1.6 million) and those working more than 80 hours monthly or who meet an earnings threshold throughout the time that they are enrolled in Medicaid (about 14.4 million, or 39%). States are expected to implement processes to identify these groups using administrative data, though based on the experience of Arkansas, which used data matching in its effort to implement earlier work requirements, it is unclear how successful these efforts will be.

About 4 million people who had not been adjudicated as qualifying for SSI worked fewer than 80 hours per month and reported that they had activity limitations. If qualifying for Medicaid became more difficult, it is likely that a fraction of this group would apply and qualify for SSI.

Finally, approximately 4.3 million adults worked fewer than 80 hours per month and did not report any activity limitations. Although we refer to this group throughout as the target group, many in this group would likely be excepted from the work requirement for reasons we cannot identify in our data (such as because they are performing community service, providing caregiving to a family member with an activity limitation or a disability, or meet another exception criteria). Our data indicate that only about 300,000 people reported that they did not work because they did not want to.

Service use and spending

Next, we analyzed health care utilization and Medicaid spending among three groups of Medicaid beneficiaries.

Beneficiaries who did not meet the work hours threshold and do not report activity limitations

The group that is targeted by the mandate, those working less than 80 hours with no reported limitations, had the lowest average Medicaid spending—approximately $3,400 per person (Table 1). More than 40% of this group did not use any medical services at all (Figure 2; panel a). More than 1/4 were enrolled in Medicaid for less than six months (not shown). If federal savings were drawn entirely from this targeted group, these savings from disenrollment¹ would account for only about 1/3 of CBO’s projected savings of $44.5 billion in 2030. 

It is more likely that at least some of those disenrolled will be people who might meet exception criteria but would be dissuaded by paperwork and certification requirements, as was the experience when Arkansas, New Hampshire, Georgia, and Michigan implemented expansions incorporating work requirements. We focus on those who met the work hours threshold and on those with activity limitations who are not enrolled in SSI.

Beneficiaries who met the work hours threshold

The large group of beneficiaries who worked 80 or more hours per month had median Medicaid spending over twice as high as did those who did not meet the work threshold. About 1/3 of this group did not use any Medicaid services in a year. The highest spending levels were in the sub-group with mental health or substance use-related spending (about 1/5 of the total group) and among the small number (3% of the group) who used inpatient hospital services (Figure 2, panel b). 

This group relied greatly on Medicaid coverage to meet health insurance and health care needs. Among those meeting the hours threshold, about 2/3 were on Medicaid for the entire year and, among those who were on Medicaid only part of the year, nearly ½ spent the remainder of the year uninsured (not shown). Over 30% of people on Medicaid who met the hours threshold had a diagnosis of either hypertension or high cholesterol (not shown); prior research shows that people who cut back on the use of drugs for these conditions because of increased costs are significantly more likely to die in the near term. 

Beneficiaries who did not meet the work hours threshold but who report activity limitations

Average spending among those with activity limitations was about six times as high as for the targeted group; median spending was 42 times as high. While this group accounts for only 18% of enrollees in the three groups most likely affected by work requirements, they account for nearly half of total Medicaid spending across these populations. Within this group, the subgroup receiving care for mental health and substance use conditions accounted for approximately 60% of the full group’s Medicaid expenditures (Figure 2, panel c). Among those with mental health and substance use conditions, 21% had a prescription for either an antipsychotic or a mood stabilizer (typically used for bipolar conditions). In addition, 60% of those in this group filled a prescription for an antidepressant, and 28% filled a prescription for anxiolytics. This group relied almost entirely on Medicaid for coverage, with 73% enrolled in Medicaid for the full year.

Conclusion

Medicaid work requirements set up a classic Hobson’s choice. To the extent that the provision’s requirements primarily lead to the disenrollment of so-called “able-bodied” adults who are enrolled in Medicaid but are not working, it will save the Federal government far less money than projected—that group is relatively small, tends to have very low service use, and spends the least time on Medicaid.

Thus, if the requirements are to save money, they will have to primarily disenroll those they purport to protect: enrollees drawn from the larger group who are already working 80 hours a month or enrollees from the much more expensive group who are not working but who report activity limitations. Coverage losses in the former group would imperil their ability to continue working, the intended goal of the legislation, and prior research suggests, lead to increases in mortality. Coverage losses in the latter group would likely come at a very high price in terms of increasing morbidity and, given the high toll on families and communities of inadequately treating the mental health and substance use disorders that are prevalent in this group, could also have significant further negative effects. 

The MEPS analysis suggests that there are, indeed, some Medicaid beneficiaries (about 300,000) who could work but choose not to. They comprise fewer than 1% of all working-age adults on Medicaid and just over 1% of beneficiaries who are not on SSI/Medicare, parents of dependent children, or pregnant. People in this group—the rhetorical targets of the work requirements policy—use very few Medicaid services, so savings from disenrolling them would be small. As advocates of work requirements explicitly recognize, many other groups of Medicaid beneficiaries would experience significant and unavoidable hardship if they were disenrolled from the program, with follow-on negative effects for their local hospitals and communities. Prior efforts to surgically separate the meritoriously enrolled from the slackers have proved both ineffective and very administratively costly. Medicaid work requirements just don’t work in the way their proponents promise they do. Judging only on the criteria established by their own advocates, they will be either a fiscal or a moral failure.