Abstract
Context: Twenty states are pursuing community engagement requirements (“work requirements”) in Medicaid, though legal challenges are ongoing. While most nondisabled low-income individuals work, it is less clear how many engage in the required number of hours of qualifying community engagement activities and what heterogeneity may exist by race/ethnicity, age, and gender. The authors' objective was to estimate current levels of employment and other community engagement activities among potential Medicaid beneficiaries.
Methods: The authors analyzed the US Census Bureau's national time-use survey data for the years 2015 through 2018. Their main sample consisted of nondisabled adults between 19 and 64 years with family incomes less than 138% of the federal poverty level (N = 2,551).
Findings: Nationally, low-income adults who might become subject to Medicaid work requirements already spent an average of 30 hours per week on community engagement activities. However, 22% of the low-income population—particularly women, older adults, and those with less education—would not currently satisfy a 20-hour-per-week requirement.
Conclusions: Although the majority of potential Medicaid beneficiaries already meet community engagement requirements or are exempt, 22% would not currently satisfy a 20-hour-per-week requirement and therefore could be at risk for losing coverage.
An increasing number of states are proposing work requirements—often called “community engagement” or “personal responsibility” requirements—as a condition of Medicaid eligibility (Academy Health 2019; Medicaid.gov2019). These policies require nondisabled adults to complete a minimum number of hours of work or other approved community engagement activities—such as education, job searching, volunteering, and caregiving—each week to receive health coverage through Medicaid (Carroll 2018). This study uses national data to analyze levels of time spent on qualifying community engagement activities among low-income adults who might become subject to Medicaid work requirements. To our knowledge, this is the first study to assess low-income adults' levels of community engagement at the national level.
In January 2018, the Centers for Medicare and Medicaid Services (CMS) issued guidance for Section 1115 waivers that would allow states to impose work requirements as a condition of Medicaid eligibility, with exemptions for certain populations. At the time of this writing, 10 states received approval to implement work requirements (Arizona, Arkansas, Indiana, Kentucky, Michigan, New Hampshire, Ohio, South Carolina, Utah, and Wisconsin), while several others have pending waiver applications or are considering work requirements (KFF 2020). Of the 10 states with approved waivers, 8 have expanded Medicaid under the Affordable Care Act (ACA) to all individuals with family income below 138% of the federal poverty level; 6 expanded in 2014 (AR, AZ, KY, MI, NH, OH), 1 in 2015 (IN), 1 in 2020 (UT), and 2 are not participating in the ACA Medicaid expansion (SC, WI).
Approved states that have begun to implement work requirements are facing numerous legal challenges. Beginning January 2020, Michigan implemented its requirement that adult Medicaid recipients below age 63 complete a minimum 80 hours per month of community engagement activities to maintain coverage. A lawsuit is attempting to block Michigan's work requirement, but the requirement is in effect at the time of this article's writing. Indiana implemented a minimum 20 work-hours requirement for adults below age 60 in July 2019, with a phase-in of required hours from July through September; however, the state announced in October 2019 that it would temporarily suspend enforcement. Arkansas's approved waiver took effect starting June 1, 2018, and required Medicaid beneficiaries ages 19–49 to meet community engagement requirements of at least 20 hours per week, but a US District Court judge set aside the state's waiver and stopped implementation of the work requirement in March 2019. In February 2020, a federal appeals court unanimously upheld the lower court's ruling striking down work requirements for Arkansas's Medicaid recipients; the three-judge panel for the US Court of Appeals stated that the work requirements were not consistent with the primary objective of the Medicaid statute, which is to provide health care coverage for the poor (Goodnough 2020).1 Other states proposed broader requirements for all nonelderly adults, including Kentucky, New Hampshire, and South Carolina, although federal judges blocked the implementation of work requirements in June 2018 for Kentucky and in July 2019 for New Hampshire. Recently, Arizona and Indiana have suspended or postponed the implementation of work requirements until the lawsuits challenging the requirements are resolved.
In most of the 10 approved states, work requirements can be met through paid employment, job training, job searching, volunteer work, education, or caregiving for someone elderly/disabled. Most states exempt beneficiaries who are disabled, full-time students, parents with children below age 6 years, and pregnant women. See table 1 for a summary of eligible activities and exemptions in states with approved waivers; appendix table A1 (online appendix) provides a more detailed version.
Proponents of these policies argue that the incentive to work will provide financial and health benefits for new workers, increase labor force participation, and shift some individuals from public to private coverage (Price and Verma 2018). On the other hand, implementing these requirements may be administratively costly (GAO 2019) and could reduce participation in Medicaid through restrictive eligibility criteria that some Medicaid enrollees do not meet, social stigma, or additional bureaucratic obstacles to enrollment (Hahn et al. 2018; Huberfeld 2018). For example, approximately 18,200 people benefitting from Arkansas's Medicaid expansion lost coverage as of October 2018, mostly due to not reporting any activities whatsoever to the state (Rudowitz, Musumeci, and Hall 2019). In its own waiver application, the state of Kentucky projected a 15% drop in adult Medicaid enrollment by the fifth year of implementation of Medicaid work requirements, the equivalent of nearly 100,000 people losing coverage for a full year (Meier 2017; Solomon 2018). Moreover, there are many reasons that individuals otherwise eligible for Medicaid may not be able to work, including but not limited to substance use disorders (Meyer 2018), lack of economic opportunities in their area, and lack of transportation options.
Little is known about the extent to which low-income individuals currently engage in qualified community engagement activities and how this differs in terms of race, age, and gender. While earlier studies have evaluated sociodemographic and labor market characteristics of Medicaid recipients or low-income individuals (Carroll 2018; Gangopadhyaya and Kenney 2018; Gangopadhyaya et al. 2018; Garfield et al. 2019; Greene 2019; Silvestri, Holland, and Ross 2018; Sommers et al. 2018; Tipirneni, Goold, and Ayanian 2018; Wen, Saloner, and Cummings 2019), there has been little research examining time spent on other activities that will also satisfy the new requirements, such as caregiving for the elderly/disabled, volunteering, and education. Also, most studies have not quantified whether the total hours spent on such activities would satisfy commonly proposed thresholds.
Some notable exceptions include a study of current community engagement levels among Medicaid enrollees in Kentucky (Venkataramani et al. 2019) and a study that assessed changes in all qualifying activities, including work and other community engagement activities among low-income adults in Arkansas after the first year of Medicaid work requirements implementation in the state (Sommers et al. 2019). Sommers et al. (2019) conducted a telephone survey of low-income adults in Arkansas and three control states and found that the work requirements led to substantial losses in insurance coverage but no significant change in employment or community engagement among low-income adults. Venkataramani et al. (2019) and Sommers et al. (2019) provide important information on the potential effects of work requirements, but study results were limited to a single or small number of states. We find that there is substantial heterogeneity in low-income adults' levels of community engagement across states, so single-state or regional studies may not be reflective of the whole country. To our knowledge, this is the first study to use national data to analyze levels of time spent on qualifying community engagement activities among low-income adults who might become subject to Medicaid work requirements.
Data and Methods
Data
Our data source was the American Time Use Survey (ATUS), a nationally representative survey that collects data on time spent on various activities each day (Hofferth, Flood, and Sobek 2019). ATUS respondents were randomly selected from a subset of households that completed their final interview for the Current Population Survey (CPS). The response rate is 48%, and survey size is approximately 10,000 individuals per year. With the use of sampling weights to correct for nonresponse bias, the ATUS is nationally representative of the noninstitutionalized, civilian US population.
In the time diary portion of the ATUS interview, survey respondents sequentially report activities conducted between 4 a.m. on the day before the interview and 4 a.m. on the day of the interview. Advantages of the short 24-hour recall period include more accurate data, lesser likelihood of the presence of social desirability bias, and a larger sample size. However, because there is day-to-day variation in how people spend their time, the ATUS reference period (one day) may not reflect the respondent's long-run time use (Frazis and Stewart 2010). For example, a respondent reporting their time use for a weekend or holiday may report fewer hours worked than if the same individual were to be reporting their time use for a weekday. We therefore restricted our sample to respondents reporting their time use for weekdays and nonholidays to understand a typical working week's profile. We converted the respondent's daily measures to their weekly estimates with a bounding exercise—the lower bound treated the sampled weekday's hours as one fifth of their weekly hours (which assumes they worked a standard 5-day week), the upper bound treated weekday hours as one seventh of weekly hours, and the middle ground (our preferred estimate) treated weekday hours as one sixth of weekly hours. The reason that the middle ground (treating weekday hours as one sixth of weekly hours) is preferred is because recent national statistics show that the average American workweek is close to six days (Saad 2014; Villapaz 2014). Additionally, we analyzed the results of the ATUS weekend responders, which indicate that work and other community engagement activity levels are roughly half of what they are for the weekday sample (appendix table B1 [online appendix]).
The Census Bureau classifies reported activities into 17 major categories and dozens of subcategories; we identified four groups of activities as those that would qualify as community engagement in most of the 10 states with approved waivers: (1) work and work-related activities, (2) volunteering, (3) education, and (4) caregiving for the elderly/disabled.
Sample Selection
For our main analysis, we used the 2015–18 ATUS and restricted the sample to adults aged 19 to 64 with household income below 138% of the poverty level who reported their time use for weekdays and non-holidays. We were most interested in those with incomes below 138% of the poverty level since this is usually the population eligible for Medicaid expansion under the ACA.2 Our final sample size was 2,551.
We also stratified our analytical sample into two groups: respondents who would potentially be exempt from work requirements based on criteria by states with approved waivers (because they were either full-time students, disabled, or had children under age 6) and those who would potentially be nonexempt from work requirements. We identified disabled individuals in the ATUS as those who responded that they were “not in the labor force because of disability.”
Our nonexempt sample most closely matched the population subject to work requirements in the majority of the approved work requirements states. However, there is some variation in exempt populations (appendix exhibit A1 [online appendix]), so we conducted sensitivity analyses in which we modified our primary sample based on rules proposed by different states. These modifications are described in the “Sensitivity Analyses” subsection. ATUS data limitations prevented us from excluding other exempt populations from our study sample: pregnant women (exempt in all 10 states), designated caregivers for disabled individuals (exempt in all 10 states), recipients of other public program assistance (exempt in 8 states), those being treated for substance use disorder (exempt in 7 states), the homeless (exempt in 3 states), the recently incarcerated (exempt in 3 states), and victims of domestic violence (exempt in 2 states).
Defining Community Engagement in the ATUS
We identified activities in the ATUS that most closely match community engagement activities in the 10 states with approved work requirements waivers: work and work-related activities, volunteering, education, and caregiving for the disabled/elderly.
Work and Work-Related Activities
All 10 states' proposals allow employment, vocational training, and job search. This definition aligns well with the “work and work-related activities” category in ATUS (which includes working, other income-generating activities, job searching, and interviewing). We found that there was a strong positive correlation (0.80) between the number of usual hours worked and reported in the CPS and time spent on work and work-related activities reported in the ATUS; this provides additional confidence in our ATUS estimates.
Volunteer Activities
Community service and unpaid volunteer work qualify as community engagement in nine states' proposals. The description of volunteer activities in the state waivers matches closely with the “volunteer activities” category in ATUS (which includes fundraising, social service and care, indoor and outdoor cleanup, etc.).
Education
Eight states (AR, AZ, IN, KY, MI, NH, OH, SC) allow time spent on education, so we included the ATUS “education activities” category (which includes taking classes, research/homework, etc.) in our analysis.
Caregiving for the Elderly and Disabled
Three states' proposals (IN, KY, NH) qualify caregiving for disabled/elderly adults. The ATUS reports time spent caring for and helping household adults and nonhousehold adults (specific activities include physical care, providing medical care, obtaining medical care, housework/cooking/shopping assistance, etc.). Unfortunately, the survey does not distinguish between care provided to elderly/disabled adults and care provided to nonelderly/nondisabled adults. However, a previous study used a smaller ATUS module which did make this distinction and found that there was no statistically significant difference between care provided to “all adults” and care provided to “infirm elderly adults” (Kydland and Pretnar 2019). This increases our confidence that “caring for and helping household adults and nonhousehold adults” does capture time spent caring for the elderly and disabled. We did not include time spent caring for and helping household or nonhousehold children, as childcare would not qualify as community engagement anywhere.
Statistical Analysis
We first examined levels of community engagement for our main analytical sample as well as each of the alternate samples described above. Our measures of community engagement were (a) the average number of hours per week spent on qualifying community engagement; and (b) the percent of the sample that would meet a 20-hour weekly requirement, the percent that would not meet a 20-hour requirement, and the percent that would be exempt from work requirements based on most states' criteria.
To understand whether specific demographic groups would be at higher risk of losing Medicaid eligibility under new work requirements, we also conducted subgroup analysis. Specifically, we estimated the above measures separately for men versus women, childless adults versus parents, labor force participants versus those outside the labor force, employed versus not employed, insured versus uninsured, residents of Medicaid expansion states versus nonexpansion states, and residents of states with approved Medicaid work requirements waivers versus states without such waivers, since these groups would likely have different rates of enrollment in Medicaid under a work requirement. The sample size for the insured versus uninsured analysis was smaller (N = 633) because insurance information is available only for the 29% of respondents whom we can link to the March CPS. These descriptive means allow us to assess potential disparities in the likely effects of the new policy, without adjustment for factors that may themselves contribute to the disparities; this is consistent with recommendations from the Institute of Medicine for this kind of analysis (Lê Cook, McGuire, and Zaslavsky 2012).
Next, we used the ATUS data to assess the association between community engagement activities and sociodemographic characteristics among low-income adults using a multivariate logistic regression model. We estimated three separate regressions for the following binary outcome variables: (1) respondent is exempt from work requirements, (2) respondent is not exempt but meets work requirements, and (3) respondent is not exempt and not meeting requirements. We converted odds ratios to marginal effects for ease of interpretation. Independent variables were sex, marital status, parental status, age, educational attainment, race/ethnicity, and census region. Statistical significance was defined using a two-tailed p-value of 0.10.
All estimates accounted for the ATUS survey weights and complex survey methodology. This study used publicly available data and was deemed nonhuman subjects research by the Indiana University internal review board.
Sensitivity Analyses
We conducted sensitivity analyses in which we modified our analytical sample in the following ways:
Age-Based Sensitivity Check
Three states (KY, NH, SC) proposed work requirements for all nonelderly, nondisabled beneficiaries; MI would exempt those above age 62; IN and UT would exempt those above age 59; and four states (AR, AZ, OH, WI) would exempt those above age 49. Our primary sample included adults below age 65 and a sensitivity analysis studied adults below age 50.
Sensitivity Check Based on Student Status
Full-time students would be exempt from work requirements in eight states (AR, AZ, IN, KY, MI, OH, UT, WI), so our primary analysis excluded full-time students. We conducted sensitivity analyses in which we also excluded part-time students, since they would be exempt in OH, UT, and WI.
Sensitivity Check Based on Parental Status
Our primary sample excluded parents with children below age 6 because they are exempt from all 10 states' proposals. We also conducted sensitivity analyses in which we excluded parents with children below age 18, as 6 states would exempt this group (AR, AZ, KY, OH, SC, WI).
Some states' proposals do not consider caregiving and education as qualifying community engagement activities. Specifically, two states (UT, WI) would not count education, and seven states (AR, AZ, MI, OH, UT, SC, WI) would not count caregiving for elderly/disabled adults. We therefore conducted a sensitivity analysis excluding time spent on caregiving for elderly/disabled adults and education from our list of qualifying community engagement activities.
Limitations
This analysis is descriptive and correlational, thus no causal inference should be drawn that these represent any effects of community engagement requirements. Our results do not indicate whether individuals could easily modify their time spent on various activities in response to the new requirements. Our analysis is also unable to assess whether individuals could document these activities adequately to satisfy various state reporting requirements. Our empirical approach involves multiplying daily reported work hours by an estimated number of days in the week, which provides an unbiased estimate of average hours per week but does not capture the true variance over individuals. If a significant portion of our sample experiences fluctuation in weekly hours worked (Butcher and Schanzenbach 2018; Karpman, Hahn, and Gangopadhyaya 2019), our approach may not accurately capture the number of individuals marked as falling over or under the 20-hour threshold.
Another limitation is that our exclusion criteria reduced our analytical sample to just over 2,500 individuals, which may mean we were underpowered to detect some demographic differences in our logistic model. It is also unclear how closely our categorization of activities will match those implemented by states under their community engagement criteria (see the “Defining Community Engagement in the ATUS” section above). Finally, all of our data come from surveys and are self-reported; results may therefore be subject to social desirability bias and recall bias.
Results
Table 2 displays demographic and economic characteristics of the overall analytical sample, and separately for potentially exempt and potentially nonexempt respondents (N = 2,551). The nonexempt sample was more likely to be male, childless, of older age, and have higher educational attainment.
Average Number of Hours Spent on Community Engagement
Figure 1 displays the average number of weekly hours respondents reported spending on all activities that may qualify as “community engagement” under state proposals. Overall, low-income adults spent 25.1 hours per week on average on community engagement activities (95% CI = 23.9, 26.3); respondents who would be nonexempt spent 29.7 hours per week (95% CI = 28.1, 31.4), while those who would be exempt spent 19.8 hours per week (95% CI = 18.1, 21.5).
The results presented in Figure 1 are based on our middle estimate in which we multiplied the number of reported weekday hours by 6. For the nonexempt sample, our lower bound (which multiplied weekday hours by 5) and upper bound (which multiplied weekday hours by 7) were 24.9 and 34.7 hours per week, respectively. Appendix figure B1 (online appendix) displays the lower and upper bounds for each activity of community engagement.
Figure 1 and appendix figure B1 (online appendix) also show that, on average, low-income adults spent a modest amount of time on community engagement activities other than work. The nonexempt population spent 0.7 (lower bound) to 1.0 hours (upper bound) per week caregiving, 0.9 to 1.2 hours per week on education, 0.7 to 0.9 hours per week volunteering, and 22.6 to 31.6 hours per week working. Sensitivity analyses showed that the average levels of community engagement were remarkably similar even when we restricted our study sample to the below-50 population, all nonstudents, US citizens, and those with income below 50% of the poverty level (appendix table B2 [online appendix]).
We also conducted tests to assess heterogeneity by sociodemographic and geographic characteristics. We found that among the nonexempt low-income population, women had statistically significantly lower community engagement levels than men; the uninsured, those not in the labor force, and those not employed also had statistically significantly lower community engagement levels than their respective counterparts (appendix table B3 [online appendix]). However, we found no statistically significant differences in levels of total community engagement between childless adults and parents, between residents of Medicaid expansion and those of nonexpansion states, and between residents of states that have approved Medicaid work requirements waivers and those of other states.
Percent of Low-Income Nonelderly Adults that Would Not Meet a 20-Hour Requirement
Figure 2 presents the percent of our low-income sample that would meet a 20-hour weekly community engagement requirement (the standard requirement set in most states with approved waivers). We found that 47% of low-income nonelderly adults would be exempt from the requirement. (Appendix table A2 presents reason for exemption; 24% of this sample had children under age 6, 18% were disabled, and 10% were full-time students [online appendix].) Another 31% of low-income nonelderly adults would not be exempt but would meet a 20-hour community engagement requirement, while 22% would not be exempt or meet the requirement. Appendix figure C1 (online appendix) presents similar estimates based on the lower and upper bounds of weekly hours. The bounds are relatively tight; between 30% and 32% of low-income nonelderly adults would meet the 20-hour requirement, while 21% to 23% would not meet the requirement. Low-income nonelderly adults who are not exempt and not meeting requirements work an average of 2.8 hours per week and would thus need to work an additional 17.2 hours per week to meet a 20-hour requirement.
Figure 2 also presents results separately for men, women, childless adults, and parents. We found that women were statistically significantly more likely to be exempt but also less likely to meet a 20-hour requirement than men (p < 0.10); overall, 23% of women would not be exempt or meet a 20-hour requirement versus only 20% of men. Finally, we found that 26% of childless adults would not be exempt or meet a 20-hour requirement versus only 18% of parents, a statistically significant difference (p <0.01). Appendix figure C2 (online appendix) presents additional heterogeneity tests. Notably, those who are uninsured, not employed, and not in the labor force are at higher risk of not meeting a 20-hour requirement.
Sensitivity analyses showed that the percent of adults who would not meet a 20-hour requirement was remarkably similar even when we excluded caregiving and when we restricted our study sample to US citizens, all nonstudents, and those with income below 50% of the poverty level (appendix figures C3–C5 [online appendix]).
Sociodemographic Correlates of Community Engagement
Finally, we used multivariate logistic models and the ATUS data to assess correlations between community engagement and sociodemographic characteristics. Table 3 displays the results of these regressions, which largely confirm the pattern seen in our unadjusted stratified analyses. We examined three outcomes: probability of exemption from work requirements, probability of not being exempt but meeting a 20-hour requirement, and probability of not being exempt and not meeting a 20-hour requirement. Women and individuals aged 45–64 were at significantly higher risk of not being exempt and not meeting a 20-hour requirement, compared to men and younger adults (p < 0.05). Individuals with some college (p < 0.01) and college-level education (p < 0.01) were more likely to meet a 20-hour requirement than those whose educational attainment level was less than high school.
Discussion
In this analysis of national survey data, we estimated that nonelderly low-income adults without disabilities spend an average of 29.7 hours per week on qualifying “community engagement activities.” The results show that 47% would be exempt from work requirements according to the criteria proposed by most states with approved waivers, 31% would exceed the 20-hour cutoff that several states are employing as a prerequisite for Medicaid coverage, and 22% would not be exempt or meet the requirement.
On the one hand, this indicates that most potential beneficiaries are already meeting the requirement. However, many will still need to navigate the administrative process in their state to maintain their coverage (Carroll 2018). Arkansas, for instance, was using electronic wage data and other information to satisfy the requirements on behalf of many beneficiaries, without any new reporting from the enrollees; but reporting rates among the remaining population were low (Alker and Clark 2018; Hill and Burroughs 2019). Moreover, nearly 33% of nonexempt Arkansas residents were unaware of the work requirements policy, and half were unsure whether the requirements applied to them (Sommers et al. 2019). On the other hand, 22% of adults in our sample are not exempt and are not yet meeting the requirement; this population could be at risk for losing coverage. We note that this figure is higher than other published estimates. For instance, Sommers et al. (2019) found that in Arkansas only about 3% of low-income adults subject to work requirements were not exempt or meeting the 20-hour requirement. Another study of Kentucky's Medicaid demonstration waiver program found that only 15% of enrollees would not be exempt or meet community engagement requirements (Venkataramani et al. 2019). One reason for the discrepancy between our findings and these earlier works may be because earlier studies focus on single states and specific age groups (e.g., Sommers et al. [2019] examine adults aged 30–49 in Arkansas), whereas we consider adults aged 19–64 nationwide.3
A recent report by the Kaiser Family Foundation using an alternative data source concluded that most nonelderly Medicaid adults already work or face substantial barriers to work, leaving a very small share of adults to whom these policies are directed (Garfield et al. 2019), similar to findings from Kansas, which is considering work requirements (Sommers et al. 2018). However, without attention to the number of hours worked, these results may understate the potential impact of the new requirements. Our findings suggest that about 22% of low-income, nondisabled adults may risk coverage loss unless they increase their total community engagement hours.
While advocates argue that the new incentives will increase work and other activities, it is unclear whether there are adequate job opportunities for individuals not currently meeting the requirements to do so in the future. Critics of the ACA warned that the Medicaid expansion would lead many low-income adults to stop working since they would no longer require jobs to obtain health insurance; research indicates that has not occurred (Gooptu et al. 2016; Kaestner et al. 2017; Moriya, Selden, and Simon 2016). Whether the opposite set of incentives under work requirements will be adequate to change employment behavior in this population remains to be seen, and early findings from the first year of Arkansas's work requirement suggest that the new policy has not significantly impacted levels of community engagement (Sommers et al. 2019).
Our multivariate analysis indicates that older adults, women, and those with less education are at higher risk for losing coverage as a result of not meeting the 20-hours-per-week requirement and may need additional assistance. Our finding of lower rates of community engagement among near-elderly adults (45–64) raises the possibility of negative health impacts of work requirements, as this age group likely has the greatest burden of disease and need for health care access. These results show some overlap with recent findings from Michigan that Medicaid enrollees reported being out of work or unable to work more if they were older or in poor health, though that study—unlike ours—found that men and blacks were also at higher risk (Tipirneni, Goold, and Ayanian 2018).
Losing Medicaid coverage can have detrimental impacts on low-income people's access to care, health, and financial outcomes. Sommers et al. (2019) found that the loss of Medicaid coverage associated with Arkansas's work requirement was accompanied by a significant increase in uninsurance, but no significant impact in the use of employer-sponsored insurance. While Sommers et al. found no meaningful changes in the probability of having a personal physician or cost-related delays in care in the first 3 months of disenrollment, the authors caution that longer-term assessment is essential. Previous studies of a large Medicaid disenrollment in Tennessee show that losing Medicaid coverage reduces access to care, reduces doctor and dentist visits, increases the use of free or public clinics, increases uninsured hospitalization and emergency department use, worsens self-reported health, lowers credit scores, and increases bankruptcy risk (Argys et al. 2017; DeLeire 2018; Ghosh and Simon 2015; Heavrin et al. 2011; Tarazi, Green, and Sabik 2017; Tello-Trillo 2016). Even those Medicaid enrollees who are able to find alternative insurance coverage under Medicaid work requirements may face adverse effects because insurance benefits and provider networks often differ across coverage and plan types. The literature on churning—the movement between and out of health plans—suggests that transitions between health plans reduces prescription drug adherence and erodes patients' perceived health care quality (Sommers et al. 2016).
Our analysis also provides important insights regarding currently uninsured individuals who may become eligible for Medicaid. Our subgroup comparisons indicated that low-income adults without health insurance spent fewer hours in qualifying community engagement activities than those with coverage, thus states including work requirements as part of a Medicaid expansion (such as Utah and Virginia) (Virg. H.B. 5002, 1st spec. sess., 2018) may raise particular challenges for currently uninsured individuals.
Another important implication of our results is that nonemployment community engagement activities form a nontrivial portion of time use for low-income nonelderly adults. While much of the public discussion of this policy has focused on work, 7% of our low-income sample would not meet the community engagement threshold based on employment alone, but does meet it once activities such as volunteering, education, and caregiving are included. This raises the question of how easily states will be able to verify nonemployment-based community engagement and whether this administrative challenge might lead to broader coverage losses than anticipated (Brantley and Ku 2018). However, it is also the case that most beneficiaries who meet the requirements do so based on employment or exemptions, meaning that the bulk of the published literature using other data sources (which lack detailed time-use data) are still able to provide generally useful results, though perhaps slightly underestimating the share of beneficiaries who satisfy work requirements (Garfield et al. 2019; Tipirneni, Goold, and Ayanian 2018).
Conclusion
As multiple states implement or propose work requirements in Medicaid (pending ongoing litigation), our results provide new insights into the pattern of community engagement activities among low-income adults. While most potential Medicaid beneficiaries are exempt or already meet the requirement, 21% to 23% of low-income nonelderly adults lack adequate hours spent on community engagement activities and may therefore be at risk of losing coverage.
Acknowledgments
We acknowledge the helpful suggestions and feedback of Benjamin Ukert and seminar participants at the 2019 Association for Public Policy Analysis and Management annual research meeting in Denver, CO. We are grateful to Colleen Mattingly for helpful research assistance.
Notes
Approved work requirement waivers in two other states (KY and NH) also are currently set aside by courts.
Eight of the 10 states with approved work requirements expanded Medicaid under the ACA (AR, AZ, IN, KY, MI, NH, OH, UT).
In appendix figure C6 (online appendix), we present our ATUS analysis for low-income adults aged 30–49 years in Arkansas only. We found that our results were remarkably similar to those of Sommers et al. (2019). We found that 3.1% of this sample would not be exempt or would meet a 20-hour work requirement, and Sommers et al. (2019) found that 3.3% would not be exempt or would meet a 20-hour requirement.
