Consequences of Work Requirements in Arkansas: Two-Year Impacts on Coverage, Employment, and Affordability of Care

Abstract

In June 2018, Arkansas became the first state to implement work requirements in Medicaid, requiring adults ages 30–49 to work 20 hours a week, participate in “community engagement” activities, or qualify for an exemption to maintain coverage. By April 2019, when a federal judge put the policy on hold, 18,000 adults had already lost coverage. We analyze the policy’s effects before and after these events, using a telephone survey in late 2019 of 2,706 low-income adults in Arkansas and three control states, compared to data from 2016 and 2018. We have four main findings. First, most of the Medicaid coverage losses in 2018 were reversed in 2019 after the court order. Second, work requirements did not increase employment over 18 months of follow-up. Third, 30- to 49-year-olds in Arkansas who had lost Medicaid in the prior year experienced adverse consequences – 50 percent reported serious problems paying off medical debt; 56 percent delayed care due to cost; and 64 percent delayed medications due to cost. These rates were significantly higher than among Arkansans who remained in Medicaid all year. Finally, awareness of work requirements remained poor, with more than 70 percent of Arkansans unsure whether the policy was in effect.

INTRODUCTION

The Centers for Medicare and Medicaid Services (CMS) has authorized 10 states to implement “community engagement” requirements in Medicaid, as of April 2020.¹ Commonly known as work requirements, these policies require Medicaid beneficiaries to spend a certain number of hours each month working or in other activities such as community service or job training, or qualify for an exemption. Then, beneficiaries must report those activities or their qualifying exemption to the state to maintain coverage. The effect of these policies on long-term employment and the ability of beneficiaries to navigate the reporting requirements are important questions, with implications for the 20 states with waivers approved or under consideration by CMS.

The first and only state to fully implement this policy to date was Arkansas, which starting in June 2018 required adults ages 30–49 to file monthly online reports on their employment or other activities to maintain coverage. By early 2019, 18,000 adults had been removed from the program by the state for non-compliance – most of whom never submitted any reports to the state.^2,3 An evaluation conducted using data from late 2018 revealed a significant increase in the uninsured rate in the target population, with no change in employment or community engagement.⁴ In April 2019, a federal judge halted the state’s work requirements, ruling that the program did not satisfy the primary objective of Medicaid and should not have been approved by CMS. Since then, a federal appeals court upheld the ruling and a separate March 2020 federal ruling blocked Michigan’s community engagement policy, but several other states continue with their plans for work requirements.^5,6 In defending Arkansas’s work requirements, CMS and state officials argued that the program needed to be evaluated over a longer follow-up period to determine whether it increased employment, and speculated that many of those initially losing coverage would obtain private insurance later on.⁷

Our objective was to examine the second-year impacts of Arkansas’s work requirements. We assessed changes in coverage and employment in late 2019, 18 months after the policy was first implemented and 9 months after it was blocked by the courts. We also examined the implications of Medicaid disenrollment in Arkansas on affordability of care and medical debt, as well as awareness of and experiences with the state’s work requirements among low-income Arkansans.

METHODS

Study Design and Setting

We conducted a random-digit dialing telephone survey to measure outcomes in late 2016 before the implementation of work requirements in Arkansas, in late 2018 while the policy was in effect, and in late 2019 after the policy had been reversed by the courts. We compared outcomes among Arkansans ages 30–49 to outcomes among Arkansans ages 19–29 and 50–64 (who were not affected by the 2018 policy), as well as among adults in three comparison states: Kentucky, Louisiana, and Texas. These states were originally chosen to study various approaches to Medicaid expansion due to their geographic proximity to each other and similar demographic profiles. In 2018, Kentucky’s planned work requirement was blocked by the courts before implementation, and in 2019, the new governor withdrew the state’s proposal.⁸ Louisiana and Texas have not implemented work requirements.

Our baseline data from 2016 and 2018 come from a survey instrument that has been previously compared to data from the American Community Survey and Behavioral Risk Factor Surveillance System, which showed similar estimates for coverage and access to care over time in these states.⁹ This project was approved by the institutional review board of the authors’ institution.

Sample and Survey

Our survey was conducted via landlines and cellular telephone, in English and Spanish, from November 6, 2019 to January 3, 2020. For brevity, we refer to data from this wave of the survey as “2019 data;” less than two percent of respondents were interviewed in early January 2020, and excluding those observations has little effect on the results. Survey respondents were U.S. citizens ages 19–64 who reported 2018 household incomes below 138 percent of the federal poverty level; we defined the sample based on the prior year’s income, since if work requirements led more people to get jobs in 2019, then this would bias a sample defined by 2019 income. Most respondents (95.7%) were contacted via random digit-dialing; a subset (4.3%) of respondents were contacted by the survey vendor after participating in a different survey, in order to improve efficiency in screening for the target population and facilitate oversampling of Arkansans ages 30–49. These data were then combined with data from 2016 and 2018.⁴ The average 3-year response rate (according to the American Association of Public Opinion Research RR3 definition) was 15%, which is similar to or better than other rapid turnaround survey data sources used for timely policy analysis of Medicaid and the Affordable Care Act.^10,11 We used survey weights based on demographic benchmarks from the Census Bureau and National Health Interview Survey to yield results representative of the low-income population in each state.

Outcomes

The primary outcomes were the percentage of respondents who were uninsured, the percentage of respondents with Medicaid coverage, and the percentage of respondents who worked at least 20 hours a week.

Each respondent’s coverage status was categorized into one of seven mutually exclusive categories using the following hierarchy: Uninsured, Medicaid, Medicare, Marketplace, Employer-Sponsored Insurance, Non-Group, and Other. Because Arkansas’s Medicaid expansion in 2014 used Medicaid funds to purchase private plans on the ACA marketplace for enrollees, we combined Marketplace coverage and Medicaid coverage into a single category for analysis.

Secondary outcomes included the percentage of respondents with employer-sponsored insurance, the number of hours worked per week, and the percentage of respondents not engaged in any activity (including exemptions) that would meet the work requirement. We asked individuals about their participation in the following activities that would meet the state’s work requirement: 80 hours of work per month, job search, job training, or community service. We also assessed the following potential exemption categories: pregnant women, persons with disabilities, full-time students, and persons caring for a child or other household member.

To examine the potential consequences of disenrolling from Medicaid after work requirements were implemented, we assessed the amount of medical debt and the percentage who reported that their medical debt was a serious problem for them. We also assessed the percentages who reported delays in care or who had skipped taking prescribed medications due to cost. To identify individuals who had disenrolled in the prior year, respondents were asked whether they had ever lost Medicaid or Marketplace coverage in the previous 12 months.

Finally, we asked respondents in Arkansas several questions about their awareness of and experience with work requirements.

Statistical Analysis

Our main study design was a triple-difference approach, using variation in year, age, and state to identify the impacts of Arkansas’s work requirement policy and its subsequent reversal by the court. Our model tested whether the change in outcomes among Arkansans ages 30–49, relative to the change among Arkansans ages 19–29 and 50–64, was larger than the comparable change in the three comparison states. The policy estimate for 2019 is the coefficient on the three-way interaction of indicators for Arkansas, age group 30–49, and the year 2019. Previous research has assessed stability in trends of coverage and employment before the implementation of work requirements for our study’s relevant state and age groups using the American Community Survey, showing no signs of divergence prior to the implementation of work requirements in 2018, which offers support for our study design.⁴

For coverage outcomes, we analyzed repeated cross-sectional data from the three waves of our survey – 2016, 2018, and 2019 – which each contain a distinct sample of respondents. We estimated linear regression models, clustering standard errors by state-age group (20 clusters total, with age groups 19–29, 30–39, 40–49, 50–59, and 60–64). Models adjusted for sex, race/ethnicity, education, interview language (English or Spanish), marital status, residence area (urban or rural), and year, as well as interactions between year and covariates to replicate previously published estimates with this data for 2016 and 2018 (see Appendix for regression models).¹²

For employment outcomes, respondents in 2018 were asked about 2017 and 2018 employment, and respondents in 2019 were asked about 2018 and 2019 employment. This approach of asking about previous and current year employment was used because we did not have baseline employment data from before 2018. Due to the longitudinal nature of the data and non-independent observations across time, we used a multilevel mixed model with random effects for age groups in each state and for each respondent.

For both coverage and employment models, we conducted a sensitivity analysis including only respondents ages 30–49.

For affordability outcomes after disenrollment, we calculated survey-weighted means and medians to compare results among Arkansans ages 30–49 who had disenrolled from Medicaid/Marketplace coverage in the past year to Arkansans ages 30–49 who had remained enrolled in Medicaid/Marketplace coverage. We asked respondents to estimate their amount of medical debt, and responses were coded as discrete categories: $0, $1-$1,000; $1,001-$2,000; $2,001-$5,000; $5,001-$10,000; and more than $10,000. To compute the mean and median debt, we conservatively used the lower bound of the applicable range for each respondent to impute a specific debt amount. Some respondents offered a specific debt value, as opposed to a category; when available, the specific dollar value amount was used in our calculations. We also calculated average medical debt using values top-coded at $20,000, to test the effect of potential outliers.

As noted earlier, all analyses were survey-weighted to reflect the target population in each state.

Limitations

Our study has several limitations. First, our survey response rate was 15%, which is lower than the response rates for government surveys. However, we use demographic weighting to address nonresponse bias, and our survey instrument has been previously compared to government survey data with similar results.⁹ Respondents may be prone to survey biases, such as recall bias and social desirability bias, which may lead to inaccurate estimates of the amount of medical debt or over-reporting of current employment status. Another bias in our survey appears to be due to a telescoping phenomenon where people recall higher employment rates in the past, likely beyond the 12-month period about which they were being asked. For this reason, we also tested our employment using only current-year results (without the lookback question), and our findings for 2018 vs. 2019 employment were similar as in our main model. Some survey items were answered by only a small percentage of respondents, such as questions asked of those who reported disenrolling from Medicaid or Marketplace or those not meeting the work requirement.

Importantly, we cannot determine which coverage losses and affordability changes were due directly to the work requirement policy, since there was substantial confusion about the policy among respondents and we did not have administrative enrollment data to which we could link our results. In models comparing access and affordability challenges for people losing Medicaid in Arkansas, presumably not all of these individuals lost coverage due to work requirements (as all states have a baseline level of churning). Thus, our results for these outcomes are descriptive, rather than the stronger quasi-experimental analysis for our insurance coverage and employment outcomes.

In addition, our survey asked individuals whether they had disenrolled from Medicaid or Marketplace in the past 12 months, (i.e. since November/December 2018). This time period imperfectly overlaps with the state’s disenrollment of individuals not meeting work and reporting requirements from August through December 2018. As a result, our estimate of the number of individuals who disenrolled from Medicaid or Marketplace after work requirements went into effect is imprecise.

While our quasi-experimental analysis for the study’s primary outcomes is a strong design that uses both within-state and out-of-state control groups, the study is still potentially subject to unobserved time-varying confounders that differ across state and age groups. Finally, in terms of external validity, Arkansas’s work requirements implementation, as well as the characteristics of its residents, may differ from other states.

RESULTS

Study Sample and Descriptive Statistics

Our overall sample included 8,661 respondents (2,706 from 2019; 3,004 from 2018; and 2,951 from 2016), and each year of the sample was independent from the other years. We oversampled in Arkansas and in the 30–49 year old age group, the group targeted by work requirements; overall, 60 percent of the sample came from Arkansas. Exhibit A1 presents demographic characteristics of the respondents.¹²

Coverage

The percentage of Arkansans ages 30–49 with Medicaid/Marketplace coverage dropped from 70.5% in 2016 (pre-work requirements) to 63.7% in 2018 (during work requirements), and rose to 66.1% in 2019 (work requirements no longer in effect) (Exhibit A2).¹² The uninsured rate for Arkansans ages 30–49 rose from 10.5% in 2016 to 14.6% in 2018, and then back down to 12.5% in 2019. Meanwhile, the uninsured rate for adults ages 30–49 in our comparison states was fairly stable all three years.

Exhibit 1 shows regression estimates of coverage outcomes as a function of the work requirement policy changes, using a triple difference model. Work requirements in 2018 led to a significant drop in Medicaid/Marketplace coverage of 13.2 percentage points for Arkansans ages 30–49, relative to other age groups and states. But in 2019, after the court reversed the policy, Medicaid/Marketplace coverage rates for the treatment group compared to control groups did not differ significantly, suggesting coverage losses from work requirements were mostly reversed after the court order. For the uninsured outcome, the policy estimate of work requirements led to a significant increase in the uninsured rate of 7.1 percentage points for Arkansans ages 30–49, relative to other age groups and states, consistent with previous research.⁴ But in 2019, after the court reversed the policy, uninsured rates for the treatment compared to control groups did not differ significantly, suggesting that overall uninsured rates returned to pre-work requirement levels.

Exhibit 1: Changes in Health Insurance Associated with Arkansas’s 2018 Medicaid Work Requirements.

Source: Authors’ analysis of cross-sectional survey data from low-income adults ages 19–64 in Arkansas, Kentucky, Louisiana, and Texas from 2016, 2018, and 2019 (N = 8,661).

Notes: Data points indicate the coefficients from a triple-difference model, comparing adults in the target age range for Arkansas’s 2018 work requirements (30–49) vs. other age groups, in Arkansas vs. other states, for each year. 2016 represents the baseline comparison year. The bars indicate the 95% confidence intervals for each estimate (based on state-age group clustered standard errors). Models were survey weighted and adjusted for year of interview interacted with the following covariates: sex, race/ethnicity, marital status, education, interview language (English or Spanish), and place of residence (urban or rural). Medicaid and Marketplace coverage were combined as a single outcome, since in Arkansas most Medicaid expansion beneficiaries are enrolled in free Marketplace coverage.

There was a fairly large but non-significant change in employer-sponsored insurance associated with work requirements in 2018, and a smaller non-significant change in 2019. Results were similar using models including only respondents ages 30–49 (Exhibit A3).¹²

Employment and Community Engagement

The share of respondents working at least 20 hours a week ranged from 36.7% to 48.4% depending on the year, age group, and state (Exhibit A4).¹² Mean hours worked per week ranged from 14.7 to 19.5. Combining all qualifying working and community engagement activities, plus potential exemptions, more than 95% of the target population was already meeting the requirements or should have qualified for an exemption. Note that there are two estimates for 2018, since we asked 2018 respondents to report current-year employment and we asked 2019 respondents to report previous-year employment. The look-back employment measure is higher for all groups, and this is likely due to the telescoping bias discussed in the Limitations section.

Exhibit 2 presents estimates for the changes in employment outcomes associated with work requirements from our regression models. Comparing Arkansans ages 30–49 to other age groups and other states, we find no significant changes in employment, number of hours worked, or community engagement status between 2018 (during work requirements) and 2019 (after work requirements put on hold). Results were similar using models including only adults ages 30–49 (Exhibit A5).¹²

Exhibit 2: Changes in Employment Outcomes Associated with Arkansas’s 2018 Medicaid Work Requirements.

Source: Authors’ analysis of longitudinal survey data from low-income adults ages 19–64 in Arkansas, Kentucky, Louisiana, and Texas from 2018 and 2019 (N = 3,004 individuals for 2017–2018, and 2,706 individuals for 2018–2019, minus item non-response for each outcome).

Notes: Each person in the sample provided information on these outcomes for two consecutive years (asked about these outcomes for last year and then again for this year). Data points indicate the coefficients from a triple-difference model, comparing adults in the target age range for Arkansas’s 2018 work requirements (30–49) vs. other age groups, in Arkansas vs. other states, for this year vs. the prior year. The bars indicate the 95% confidence intervals for each estimate (based on state-age group clustered standard errors). Given the repeated observations for each person, we used a survey-weighted random effects model and adjusted for sex, race/ethnicity, marital status, education, interview language (English or Spanish), and place of residence (urban or rural). Individuals were identified as not meeting work requirements if they did not report working at least 20 hours per week, disability status, full-time student status, participation in job training, actively seeking work, community service, pregnancy, or caring for a child or household member who could not care for himself or herself.

Affordability and Access to Care After Disenrollment

Of Arkansans ages 30–49 who reported disenrolling from Medicaid or Marketplace at any point in the past year (N = 117), 55.6% re-enrolled and were covered by Medicaid or Marketplace by the time of the survey, 8.2% had employer-sponsored coverage, 10.1% had other insurance, and 26.1% were uninsured. Exhibit 3 presents results for medical debt and financial barriers to care, comparing Arkansans ages 30–49 who reported disenrolling from Medicaid/Marketplace coverage in the prior year to those who maintained their Medicaid/Marketplace coverage. Arkansans ages 30–49 who had lost Medicaid/Marketplace coverage frequently reported adverse consequences – 49.8% percent reported serious problems paying off medical debt, 55.9% percent delayed needed care due to cost in the past year, and 63.8% percent delayed taking medications due to cost. These rates were all significantly higher than among Arkansans who had maintained their Medicaid/Marketplace all year (P < 0.01).

Exhibit 3: Medical Debt and Access to Care After Medicaid/Marketplace Disenrollment Among Low-Income Arkansans Targeted by Work Requirement (Ages 30–49).

Source: Authors’ analysis of 2019 survey data from low-income Arkansans ages 30–49 who reported disenrolling from Medicaid or Marketplace coverage in the past year (N = 117) vs. those in Medicaid/Marketplace coverage without any disruption (N = 376).

Notes: All four outcomes are significantly different between the disenrolled group and the continuously enrolled group (P < 0.01). All results are based on survey-weighted chi-square tests. The outcome “Having Serious Problems Paying Off Medical Bills” is calculated as a percentage of the total sample (including those with no medical bills).

Arkansans ages 30–49 who disenrolled in the prior year had an average medical debt of $2,261, compared to $1,752 for those who reported continuous enrollment (P = 0.35, results not shown). Results when top-coding debt values above $20,000 were marginally significant (P = 0.07). The median medical debt amount was $1,001 for those who disenrolled, compared to $0 for those who did not (P < 0.05 using quantile regression).

Awareness of and Experience with Work Requirements

Exhibit 4 describes Arkansas residents’ experiences with work requirements, as of late 2019. Of Arkansans ages 30–49 with Medicaid/Marketplace coverage in the past year, 34.9% had heard nothing at all about the work requirements policy. Multivariate analysis showed that those without a high school degree, men, unmarried individuals, and those living in an urban setting were significantly less likely to have heard of the policy, relative to those with a high school degree, women, individuals who were married or partnered, and those living in rural areas, respectively (Exhibit A8).¹²

Exhibit 4:

Awareness of and Experience with Work Requirements Among Arkansas Residents, 2019

Variable	Percent
Heard of Work Requirements (“A lot” or “A little,” vs. “Nothing at all”)
All Arkansas Residents (N = 1,700)	53.0%
Arkansas Residents Ages 30–49 with Medicaid/Marketplace in past year (N = 493)	65.1%
Main Source of Information about Work Requirements
Arkansas Residents Who Had Heard of Work Requirements (N = 921)
Letter or call from state or Medicaid office	46.2%
Family or friend	16.9%
Social media or internet	14.6%
TV or newspaper	13.1%
Health care provider, hospital, or doctor’s office	11.3%
Community organization, church, or place of worship	4.6%
Knowledge of Current Status of Work Requirements (as of Nov.-Dec. 2019)
All Arkansas Residents (N = 1,700)
Correctly said the requirements were not in effect	5.7%
Incorrectly said the requirements were in effect	23.5%
Not sure	70.8%
Arkansas Residents Who Had Heard of Work Requirements (N = 921)
Correctly said the requirements were not in effect	7.6%
Incorrectly said the requirements were in effect	40.0%
Not sure	52.5%
Preferred Method of Reporting Work and Qualifying Activities
All Arkansas Residents (N = 1,700)
Internet, using a smartphone or a smartphone app	32.6%
Calling a telephone number	27.8%
Paper form in the mail	14.5%
In person at my doctor’s office or Medicaid office	13.9%
Internet website, using a computer	11.3%

Source: Authors’ analysis of 2019 survey data from low-income Arkansas residents (N = 1,700).

Notes: The subsample of Arkansas residents 30–49 with Medicaid/Marketplace in past year (N = 493) is comprised of those reporting current Medicaid/Marketplace coverage and those reporting disenrollment from their Medicaid/Marketplace coverage over the past year. All analyses were survey weighted.

Of those who had heard of work requirements, 46.2% reported that their main source of information about the policy was the state or Medicaid office (Exhibit 4), while smaller numbers (ranging from 11 to 17%) had heard from friends or family, medical providers, news sources, or the internet.

Knowledge about the status of work requirements was poor, with 70.8% of all Arkansas residents unsure whether the policy was currently in effect and only 5.7% correctly answering that the requirements were not in effect as of late 2019.

When asked for their preferred method of reporting work activities to the state, only 11.3% of Arkansans said an internet website using a computer, which was the state’s only available option for most of 2018 (before adding a telephone option in December of that year). The most popular option was a website using a smartphone or smartphone app (32.6%), followed by telephone (27.8%).

Given that many who disenrolled in 2018 never reported any information to the state despite the state’s outreach efforts,¹³ we also asked Arkansans who dropped out of Medicaid (N = 110) whether they had changed their phone number or address in the past year, which may be why state outreach efforts were not received. 35.6% reported a change in contact information, and fewer than half (45.6%) of those individuals said they had notified the state Medicaid office about the change (Exhibit A9).¹²

Of the small percentage of Arkansas residents who were not meeting the work requirement or an exemption to the policy (N = 106), 28% reported that they would like to start working if a job were available. When these individuals were asked about whether various state services would help them find a job, 80% said job training or more education, and 72% said transportation to and from work.

DISCUSSION

Arkansas was the testing ground for the nation’s first work requirement in Medicaid in 2018, and subsequently, it became a de facto test of the effects of reversing this policy when a federal judge blocked the requirement in 2019. In this analysis of new survey data from Arkansas and several comparison states, we find that the state’s 2018 coverage losses in the work requirement’s target population (low-income adults ages 30–49) were largely reversed by the end of 2019 after the court order. Meanwhile, at 18 months of follow-up after initial implementation of the work requirement, we found no evidence that low-income adults had increased their employment or other community engagement activities either in the first year when the policy was still in effect, or longer-term after the policy was blocked.

While coverage rates had recovered, our analysis of various affordability measures indicates that disenrollment during the study period was associated with adverse outcomes in terms of poorer medication adherence, delayed care, and medical debt that averaged over $2,200 and was a “serious problem” for nearly half of respondents. This is consistent with prior evidence on the harmful effects of churning in coverage,^14–16 though for these outcomes, we are using a descriptive analysis only and are not able to definitively tie churning among all of these adults to work requirements.

Meanwhile, Arkansas Medicaid beneficiaries and other low-income adults remain confused about work requirements, with 70% unsure if the policy was in effect, and only 6% correctly aware that the policy had been put on hold. This raises concerns that even though the policy is not in effect, it may still be leading some potentially eligible individuals not to apply for or renew their Medicaid coverage due to concerns about the work requirement. Despite low levels of awareness about the policy’s legal status, coverage rates nonetheless rebounded after the work requirement was no longer in effect.

Overall, these findings are consistent with qualitative data from Arkansas,¹³ as well as New Hampshire,¹⁷ showing that misinformation and confusion are major barriers to implementing work requirements in Medicaid. Our findings suggest that some of the logistical decisions made by Arkansas policymakers made it harder for people to report data to the state, as more than 85% of respondents reported they would have preferred a method of reporting work activities to the state other than a computer web-based system, which was Arkansas’s exclusive approach for the first several months of the policy. Smartphone, telephone, mail, and in-person options would have facilitated data reporting for many.

As noted previously, over 95% of Arkansas beneficiaries in our survey already met the state’s Medicaid work requirements or should have been eligible for an exemption. This suggests that barriers to reporting data to the state – rather than not meeting the requirements themselves – was the main cause for coverage losses in 2018.^3,13 For the small number of Arkansas residents in our survey who were not employed but wanted to work, two potential state services were identified by respondents as factors that would most help them find a job – job training/education and transportation to work. The lack of Medicaid payment for job supports in these areas further complicates efforts to obtain employment under work requirements, and may explain in part why we found no changes in this outcome. While Arkansas has tried offering referrals to work supports in the past with little success, other states have had more success with more intensive outreach programs, such as Montana’s Health and Economic Livelihood Partnership Link (HELP-Link), a targeted outreach program to unemployed Medicaid beneficiaries that entails individual career counseling on the phone, in-person job training, tuition assistance, and transportation support (and notably, no reporting requirement or disenrollment penalty for non-employment).^18,19

Finally, though not directly assessed in our analysis, any consideration of work requirements in Medicaid should consider administrative costs as well. Implementation of this policy cost an estimated $26.1 million, with 17% of the cost paid for by the state and 84% paid for by the federal government.²⁰

In conclusion, our study showed that Arkansas’s work requirements led to coverage losses associated with significant negative impacts on medical debt and affordability of care, without improving employment. Significant confusion and misinformation about the policy still remain, even though the policy is no longer in place. Thus, we find no evidence that the policy succeeded in its stated goals of promoting work and instead find substantial evidence of harm to health care coverage and access. Our results should provide a strong note of caution for federal and state policymakers considering work requirement policies in the future.

Exhibit 5:

Summary of Study Outcomes and Main Findings

Policy Outcome	Measures	Main Findings
Employment & Community Engagement	* Any employment * Hours worked * Any community engagement activity	* No significant changes in any employment or community engagement outcome
Health Insurance Coverage	* Medicaid/Marketplace Coverage * Employer Sponsored Insurance (ESI) * Uninsured	* Significant drop in Medicaid and rise in uninsurance in 2018 while work requirements in effect * Most coverage losses reversed in 2019 after court blocked work requirements * No significant change in ESI in either year
Access to Care and Medical Debt	* Cost-related delays in care * Cost-related delays in medications * Serious problems paying off medical bills	* Loss of Medicaid/Marketplace coverage in prior year associated with more serious problems with medical debt and more cost-related delays in care

Source: Authors’ analysis of 2016, 2018, and 2019 survey data from low-income residents ages 19–64 in Arkansas, Kentucky, Louisiana, and Texas (N = 8,661)