States are called “laboratories of democracy” because when certain states adopt new policies, they demonstrate the efficacy (or lack of efficacy) of those policies to other states. As an AEI essay noted, “A state-based process facilitates gradualism and, therefore, feedback and institutional learning.” Such interstate comparisons and feedback-based learning are possible because of America’s federalist system, something conservatives often exalt.

But do states actually learn from the successes and failures of policies in other states? A policy initiative by the Trump administration provides a test case. Under previous administrations, attempts by states to use 1115 waivers to impose work requirements on Medicaid recipients were never approved. However, under the Trump administration, all states could impose work requirements on Medicaid’s able-bodied and non-elderly working-age population. Twelve states were approved to do so, although only Arkansas managed to implement it for about a year. While the Biden administration has since rescinded waiver permission in four of these states, this short-lived, Trump-era policy still provides a natural experiment for understanding whether work requirements, well, work.

The results are in. A recent paper in the Health Affairs Journal, “Medicaid Work Requirements In Arkansas: Two-Year Impacts On Coverage, Employment, And Affordability Of Care,” by Benjamin Sommers, Lucy Chen, Robert Blendon. John Orav, and Arnold Epstein shows that Medicaid work requirements did not promote employment. The researchers conducted random phone dial surveying in late 2016 before a work requirement for Medicaid recipients was levied. They did so again in 2018 while the work requirement was in place, and then again in 2019 after a federal judge put the work requirement policy on hold. Results for these different periods were compared between age groups and similar states. Arkansas, for example, was compared against Kentucky, Louisiana, and Texas — similar states which ended up not implementing a work requirement.

One of the outcomes the authors measured was the percentage of respondents who worked at least twenty hours a week (which was the expanded work requirement for Medicaid eligibility in Arkansas). After organizing recipients based on the type of medical coverage they had and controlling for exempted populations like the disabled, the team used a triple-difference approach to look at changes in outcomes between the Arkansas and comparison state populations for the 30-49 age group. They found “no significant changes in employment (as measured by working more than twenty hours a week), community engagement status (as measured by meeting or not meeting the work requirement), or number of hours worked between 2018 and 2019.”

Proponents of work requirements cite welfare reform in the 1990s when people left welfare and employment went up. A recent letter signed by many of these proponents uses the increase in employment after the 1996 welfare reform bill as support for work requirements on welfare and against extending the child tax credit. But this reasoning has problems; it’s very hard to make a causal statement based on a time series analysis since many variables affect employment. What is needed is a counterfactual. Recent Medicaid waivers created a good counterfactual because it’s possible to compare the states that received waivers and those that did not.

There is a slew of reasons why work requirements are bad policy. As welfare reform expert Peter Germanis puts it, “work requirements are dead.” They impose administrative burdens and encourage recipients not to apply for and participate in social insurance programs (a widespread phenomenon) by making it difficult to enroll in those programs. Work requirements are ineffective at actually driving labor force engagement because recipients are generally already employed. Finally, they often feature ineffective exemptions that punish the disabled, the elderly, residents of high-unemployment areas, and others unable to work. Furthermore, Arkansas residents in the 30-49 age group who were thrown off Medicaid experienced difficulty paying off medical debt and delayed necessary medical care within the past year at significantly higher rates than those who had sustained coverage, confirming our argument that the absence of a robust social insurance net leads to reliance on subpar loans. In fact, when these Medicaid work requirements began, Ed Dolan, a Senior Fellow at the Niskanen Center, predicted they wouldn’t work, citing much of the existing evidence base. 

The laboratories of democracy have thus confirmed what we already knew: Work requirements don’t in fact work, and we have the Trump administration to thank for reminding us!

Matt Darling is the Employment Policy Fellow at the Niskanen Center.

Audrey Xu is a poverty and welfare policy intern at the Niskanen Center and a rising senior at Rutgers University.