Analysis Justifying Trump Administration’s Medicaid Work Requirement Rule Is Deeply Flawed

The 2025 reconciliation law instituted a harmful policy taking Medicaid coverage away from people not meeting a harsh work requirement, and the Trump Administration recently finalized a rule that will make the policy even more harmful.^[1] In particular, the new rule makes it harder for people to qualify for an exemption from the work requirement due to medical frailty, which includes: blindness or disability; substance use disorder; disabling mental disorder; physical, intellectual, or developmental disability that significantly impairs a person’s ability to perform one or more activities of daily living; and serious or complex medication condition. As a result, people with these conditions will likely experience significantly more coverage loss.^[2]

Major federal rules require a Regulatory Impact Analysis (RIA) to help policymakers and the public understand the benefits and costs of the rule and determine whether the benefits are likely to justify the costs. The RIA for the Medicaid work requirement rule is deeply flawed: it greatly inflates the benefits of the rule while ignoring most of the costs.

The Analysis Ignores Impacts of Coverage Loss and Disregards Research and Evidence

The RIA ignores societal costs of taking away Medicaid coverage. Without health coverage, people are more likely to die, be in poorer health, and accumulate medical debt.^[3] Omitting the impacts of coverage loss, which are large and strongly supported by the research literature, is indefensible and runs counter to federal guidance for regulatory analysis.^[4]

The RIA ignores all studies assessing the impact of Medicaid work requirements. There are three peer-reviewed studies analyzing the impacts of Medicaid work requirements in Arkansas, the only state to follow through with taking coverage away from people who didn’t meet the harsh requirements, disenrolling 1 in 4 enrollees subject to them.^[5] The studies found that Medicaid work requirements had no impacts on employment or hours worked but led to large-scale disenrollment and increased uninsurance. Rates of coverage loss were similar in other states that attempted Medicaid work requirement policies. New Hampshire halted its policy after being on the verge of disenrolling about 1 in 3 Medicaid expansion enrollees, implying an even higher rate of coverage loss than in Arkansas.^[6] Similarly, Michigan was set to disenroll nearly 1 in 3 enrollees subject to its requirement before its program was blocked by a court ruling.^[7] Instead of engaging with the evidence, the RIA states, “We have not relied on these previous demonstrations for data or assumptions used in this analysis.” It’s striking that an RIA for the Medicaid work requirement would ignore studies of Medicaid work requirements.

The RIA assumes the work requirement will increase employment and result in low disenrollment, contrary to evidence. The research on Medicaid work requirements finds no impact on employment and much higher rates of disenrollment than assumed in the RIA. The RIA’s estimate of net Medicaid coverage loss of 3.3 million people in 2034 is far below the Congressional Budget Office (CBO) estimate of about 5.7 million, in large part because CBO incorporates the relevant research and data that the RIA ignores.^[8]

Analytically Unsound Assumptions Drive RIA Results

In addition to the obvious flaws discussed above, the RIA relies on assumptions that are weakly supported, arbitrary, or unrealistic. These assumptions are complex and technical but important drivers of the RIA’s results.

The RIA assigns implausibly large net social benefits to employment. The RIA assumes that the societal benefits and costs of the Medicaid work requirement derive overwhelmingly from the impact on employment and other forms of community engagement (e.g., volunteering, education). The RIA values the time benefit of work as the additional output of the work done by people subject to the requirement (the marginal product of labor), which it estimates by using their assumed pre-tax compensation of $25 per hour as a proxy.

But it values the time cost of the work requirement using a different approach: a person’s net private return to working, or the wage a person requires to choose to work. The RIA approximates this amount using a Trump Administration estimate of the economy-wide average effective marginal tax rate^[9] and subtracts $12 per hour of purported taxes and transfers from its assumed $25 per hour rate of pre-tax compensation. The result is an estimated $13 per hour time cost of work, only about half of the $25 per hour estimated time benefit.

The RIA counts the entire $12 per hour gap between the two approaches as a net social benefit, an assumption that is inconsistent with federal guidelines and best practice for benefit-cost analysis, which call for careful examination of the affected population’s employment decisions and market conditions.^[10] Treating the full gap as a net social benefit is also an extreme assumption given economic research showing that such gaps can arise for many reasons.^[11]

Even aside from the RIA’s thin conceptual basis, the calculated gap is more than double what it should be, according to Department of Health and Human Services (HHS) estimates of population-specific effective marginal tax rates for a population similar to the one subject to the work requirement (i.e., low-income households, mostly without children).^[12]

The RIA is even weaker in its valuation of non-work activities such as volunteering and education. It assumes the same inflated net benefit of $12 per hour as it did for its valuation of work, even though the conceptual basis for such a valuation does not apply. For example, the value of volunteer time depends on the volunteer activities and whether the person derives satisfaction from them, while the value of education depends on the type of education and impacts on earnings.^[13]

The RIA overestimates “human capital” benefits of employment, such as work knowledge and relationships, for former Medicaid enrollees who it assumes become employed and remain working after they are no longer enrolled. The RIA bases its estimate on a study of a 1990s Canadian policy that offered extra income for people who chose to take up full-time work, a policy that is poorly suited for understanding the Medicaid work requirement, which will block or take away health coverage from people not working (or participating in other qualified activities) at least 80 hours per month.^[14]

The Canadian study itself stated that due to the policy’s unusual design, a more fulsome model is needed to evaluate “even minor variants” of an income subsidy program (much less a completely different policy like the Medicaid work requirement). Moreover, the study showed benefits falling rapidly after 15 months and having “no lasting impact.” But the RIA’s assumptions imply that benefits persist well beyond what was found in the Canadian study. Finally, the RIA combines its overestimates of human capital benefits derived from the Canadian study with a largely arbitrary assumption about how many former Medicaid enrollees would benefit.

The RIA inflates budgetary savings by unrealistically assuming immediate, fully phased-in impacts. This assumption is inconsistent with the policy design of the work requirement and runs counter to state experience and research.^[15] By assuming immediate, fully phased-in impacts, the RIA overestimates cumulative ten-year budgetary savings without increasing its maximum annual estimate of coverage loss.