Context: This article provides a detailed picture of the mindset of Americans about Medicaid work requirements and the important roles that perception of deservingness and racial bias play in public attitudes.
Methods: The authors conducted a large original survey to investigate public attitudes toward work requirements. They analyzed the predictors of overall support for work requirements, correlates of who should be exempt from them, and attitudes toward work supports that make compliance with work requirements easier.
Findings: The authors found that public opinion is split relatively evenly when it comes to Medicaid work requirements in the abstract. When Americans are confronted with the complexities of the issue, important nuances emerge. The authors also found consistent evidence that support for work requirements is higher among conservatives, those who see Medicaid as a short-term program, and racially resentful non-Hispanic whites. They show that groups that have historically been framed as deserving see high levels of support for their exemption (e.g., the disabled and senior citizens). Finally, the authors found that Americans are supportive of policies that provide individuals with help when transitioning into the workforce.
Conclusions: Americans' views of Medicaid and the populations it serves are complex and continue to be influenced by perceptions of deservingness and race.
Since entering office, the Trump administration has taken a slew of actions to transform the American health care system and the social safety net. One of the major efforts in this vein has been the introduction of work requirements into the Medicaid program (Musumeci, Garfield, and Rudowitz 2018; Haeder 2019b). Not surprisingly, the Trump administration has found willing partners in red states across the nation, including early adopters Kentucky (Gangopadhyaya and Kenney 2018) and Arkansas (Gangopadhyaya et al. 2018).
Work requirements have been touted by proponents for a variety of reasons including encouraging a “culture of work,” prioritizing scarce government resources, providing a way out of poverty for beneficiaries, and undoing the disincentives inherent in public assistance programs (Feulner 2010; Mead 1989; Watson 2015). Opponents, on the other hand, have countered that taking away medical coverage runs contrary to the goal of alleviating poverty and improving health and, as a result, also impedes Medicaid beneficiaries from potentially transitioning into stable work environments (Hahn et al. 2017). The partisan battle over their implementation has been fierce, with ongoing fights in state legislatures, governors' mansions, and court houses across the country (KFF 2019).
While a number of polls have assessed Americans' thoughts on Medicaid work requirements in the abstract (Wogan 2018), our understanding of the nuances of the issue is decidedly more limited. Importantly, we argue that this battle over the introduction of work requirements, and the public's opinion on the topic, should not be considered in isolation. Instead, we believe it should be viewed in the larger context of the historic development of the American welfare state in general and the emergence and evolution of the Medicaid program in particular. That is, the conflict over Medicaid work requirements is not novel but rather is the latest episode in the struggle to define the parameters of a complex program that has been ambiguous from the very start. This conflict has been fought for decades, and it is unclear whether it will be wholly resolved in the near or even distant future.
A number of analyses have illustrated this conflict at the elite level in both the US Congress and state legislatures (Barrilleaux and Rainey 2014; Olson 2015; Oberlander 2016; Callaghan and Jacobs 2017; Shor 2018; Rocco and Haeder 2018). In this analysis, however, we focus on the conflicting Medicaid attitudes of the mass public. To do so, we rely on a large original survey weighted to population benchmarks to analyze the predictors of support and opposition to Medicaid work requirements. Critically, our analysis aims to expand beyond prior research exploring the neat black-and-white division of opposition or support in the abstract by disentangling public attitudes on the details of work-requirement program design (Wogan 2018). In particular, we focus on which groups Americans think should be exempt from work requirements and whether beneficiaries should be given additional assistance to support their transitions into the workforce.
Importantly, we also place the current debate about Medicaid work requirements in the broader context of the development of the American welfare state, particularly as it applies to health care programs. We emphasize the crucial roles that perceptions of deservingness, personal responsibility, and racial bias have played in the historic development of the Medicaid program and how they continue to shape Americans' perceptions of Medicaid today. By analyzing how Americans feel about who should be exempt from work requirements and whether beneficiaries should be given additional assistance in their transitions into the workforce, we provide a more complete and nuanced picture of the complexities of public attitudes on this policy topic while adding to the broader literature on the ever-evolving American welfare state.
In the next section, we lay out the historical developments preceding the creation of the Medicaid program. We then highlight how these legacies, particularly those connected to the concept of deservingness, were directly transferred into the Medicaid program itself. We further highlight the inherent ambiguities of the authorizing legislation that have been carried forward to this date. Last, we analyze Americans' support for work requirements, first in the abstract, and then in terms of which groups Americans think should be exempt and whether work supports should be provided to Medicaid beneficiaries. We conclude by discussing the implications of our findings for our understanding of current efforts to add work requirements to the Medicaid program and for attitudes toward the American welfare state more broadly.
The Road to the Affordable Care Act's Medicaid Expansion
In the Anglo-Saxon context, the provision of any public assistance, including access to medical care, has historically been closely tied to the concept of indigence, which finds its origins in the Elizabethan Poor Laws (Terris 1951; Stern 1946). American colonists largely followed the British example and assigned responsibility for the destitute to local governments (Huberfeld 2011). Based on their experiences from the homeland, Americans adhered strongly to the “ideology of the dole[,] . . . the widely accepted belief that relief recipients are lazy and do not want to work” (Rose 1989: 65). In line with this belief, individuals who were able-bodied, minorities, or immigrants were relegated to “indoor relief,” which generally required them to enter a workhouse or a county farm (Rose 1989; Haeder 2019a). In sharp contrast to the “undeserving poor,” the “deserving poor,” like middle-class widows and their children, whose poverty occurred through no apparent fault of their own, were generally able to obtain “outdoor relief,” receiving assistance like food, clothing, or even monetary payments while continuing to live in the community (Rose 1989). As governments slowly expanded their provision of assistance, perceptions of deservingness continued to dominate policy-making decisions. For example, the federal government's assistance programs focused on deserving groups like union veterans and widows (Skocpol 1992), the disabled in need of rehabilitation (Reed 1992), and mothers and infants (Gordon 1992; Moehling and Thomasson 2012). Finally, states and localities began to initiate small cash assistance programs for seniors, children and their widowed mothers, and the unemployable blind in the early twentieth century (Orentlicher 2015).
The Great Depression significantly altered the public assistance framework in America, reformulating the concept of federalism and increasing federal involvement in social policy through shared-governance programs (Stevens and Stevens 2003; Berkowitz 1991; Rose 1989; Haeder and Weimer 2015b). Yet, crucially, it left much of the underlying ideology surrounding deservingness in the provision of public assistance untouched. The Roosevelt administration initiated a number of temporary work programs, some of which included a relief component (Macmahon, Millett, and Ogden  1971; Mott and Roemer 1948; Gangopadhyaya et al. 2018; Haeder and Weimer 2015a).1 However, most of these programs generally focused on the group of individuals considered to be deserving: white males willing to work (Rose 1989, 1993; Mettler 1998). At the same time, the federal government also followed the example of previous programs of state and local governments by codifying and expanding benefits for vulnerable groups who were not expected to participate in the wage economy. These included children and their mothers, the aged, the blind, and the disabled through the Social Security Act (Stevens and Stevens 2003; Lenroot 1936; Berkowitz 1980). As Stevens and Stevens (2003) put it, some groups among the poor were not social deviants, and the Social Security Act “provid[ed] for those made dependent through no fault of their own.”
While limited in its practical impact, another crucial expansion of welfare benefits and federal commitment came via the Kerr-Mills Act of 1960, which expanded assistance to the aged (Huberfeld 2011) and soon after, in 1962, to disabled individuals older than 21 and the blind older than 65 (Fine 1998). Yet the failure of the program to enroll a significant number of beneficiaries played a large role in the eventual passage of Medicare and Medicaid in 1965 (Moore and Smith 2005; Oberlander 2003). Not surprisingly, the Medicaid amendments to the Social Security Act themselves carried forward the preexisting determinations of deservingness by requiring coverage for participants in previous federal programs—that is, those receiving Old Age Assistance, Aid to the Blind, Aid to Families with Dependent Children (AFDC), and Aid to the Permanently and Totally Disabled (Engel 2006; Gilman 1998). Medicaid's focus on public assistance also became apparent in its institutional arrangement, as it was created as a shared-governance program for recipients (not “beneficiaries,” like their Medicare cousins) and was housed in the federal Social and Rehabilitation Services agency and state welfare departments (Engel 2006). Yet the program was also full of ambiguities, “a conservative, sparse, uneven, and stigmatized program, but . . . it was potentially sweeping at the same time” (Olson 2010: 26). On the one hand, this paradoxical nature led Congress to quickly reign in the expansive programs in California and New York (Grogan and Patashnik 2003a; Haeder 2019a; Stevens and Stevens 2003). Yet, at the same time, it extended coverage for certain deserving groups like children through the Early and Periodic Screening, Diagnostic, and Treatment program (Engel 2006; Iglehart and Sommers 2015; Grogan and Patashnik 2003a) and the aged, blind, and disabled via the federal Supplemental Security Income program (Grogan and Patashnik 2003a; Grogan and Park 2018; Thompson 2015). Importantly, the latter, “highly sympathetic groups,” gained automatic eligibility to Medicaid (Grogan and Patashnik 2003a: 829).
From the 1980s forward, Congress repeatedly expanded Medicaid to children, pregnant women, and mothers through creative entrepreneurship, unorthodox lawmaking, and pilot and optional programs (Grogan and Patashnik 2003a; Tanenbaum 1995; Gilman 1998; Thompson 2012). Crucially, advocates for these expansions generally did not base their argument for extension on questions of deservingness, but rather focused on health outcomes and cost effectiveness (Tanenbaum 1995). Seniors saw increased protections to prevent spousal impoverishment (Grogan and Patashnik 2003a, 2003b), and Medicaid strengthened its position as the primary payer for nursing home care (Olson 2010). Tellingly, the parts of the Medicare Catastrophic Coverage Act that survived were those requiring Medicaid to pay Medicare premiums for dual eligible beneficiaries (Coughlin, Ku, and Holahan 1994; Grogan and Patashnik 2003a). AFDC populations, on the other hand, saw repeated and considerable retrenchments in the AFDC program itself, which led to a corresponding reduction in their ability to enroll in Medicaid (Grogan and Patashnik 2003a), and were subject to cost-control measures through managed care (Grogan 1997).
Medicaid in the Age of the Affordable Care Act
As illustrated above, through repeated expansions and contractions, by the mid-1990s, Medicaid had essentially been transformed into two programs (Tanenbaum 1995). On the one hand, there was the long-term care program for the elderly and for the disabled—a core social entitlement with positive social constructions of deservingness, vividly defended by President Clinton as such (Grogan and Patashnik 2003b). On the other hand, there was the acute health care program for poor families, with limited benefits and limited elite or mass-level support.
With the passage of the Affordable Care Act (ACA) in 2010, Democrats hoped to wash away many of these carefully crafted groupings separating the deserving from the undeserving. Specifically, by expanding eligibility limits and standardizing them across the nation, the Obama administration and its allies in Congress hoped to dramatically expand the pool of eligible recipients and break down some of the stigma associated with the program (Iglehart and Sommers 2015).2 Yet the intense political conflict that surrounded both the passage (see Haeder 2012) and implementation (see Haeder 2020), including dozens of attempts to repeal and replace the ACA (Rocco and Haeder 2018) and the incomplete expansion of Medicaid (Barrilleaux and Rainey 2014; Olson 2015; Oberlander 2016; Callaghan and Jacobs 2017; Shor 2018), seem to indicate a lack of general consensus on the issue.
Of course, the ACA has been significantly altered by the Supreme Court's decision to make the Medicaid expansion optional (Haeder and Weimer 2013; Richardson 2019; Banks 2017). In response, the Obama administration took advantage of the ACA's vast delegation of executive powers to entice recalcitrant states to expand their Medicaid programs (Callaghan and Jacobs, 2014; Jacobs and Callaghan 2013; Thompson 2013; Thompson, Gusmano, and Shinohara 2018; Richardson 2019; Haeder and Yackee 2020). While it was willing to accept a number of conservative policy preferences like premiums and healthy behavior incentives, it held steadfast in opposing partial Medicaid expansions, Medicaid block granting, and the implementation of Medicaid work requirements (Grogan, Singer, and Jones 2017; Dinan 2014).
Medicaid in the Trump Era
Since taking office, however, the Trump administration has reversed course by promoting the introduction of work requirements into the Medicaid program and actively solicited states to take them up on the offer. Lacking congressional majorities to obtain statutory change, it has instead sought to do so via Section 1115 demonstration waivers, which have historically been used to expand coverage and benefits (Thompson and Burke 2007; Rosenbaum 2017; Callaghan and Jacobs 2014). These waivers allow states to make changes to their Medicaid programs that temporarily omit certain requirements of the Medicaid statute to test new approaches to providing coverage to their populations. These demonstrations are intended to serve as laboratories for developing more effective policy solutions by individual states that can then be applied nationally. As mentioned above, Section 1115 waivers were used by the Obama administration as a bargaining chip to entice conservative states into expanding their Medicaid programs under the ACA (Grogan, Singer, and Jones 2017). Critically, the Trump administration's efforts represent a new direction in the use of these demonstration waivers, moving away from a strategy to encourage small-scale policy innovation or to entice reticent states into expanding access, and toward a new effort to emphasize deservingness and work (Gangopadhyaya et al. 2018).
Of course, work requirements have long been a favored policy solution of conservatives, who have cast their support as evidence-based policy making. Indeed, the Centers for Medicare and Medicaid Services (CMS) has argued that work requirements are “likely to assist in improving health outcomes; . . . address behavioral and social factors that influence health outcomes; . . . incentivize beneficiaries to engage in their own health care and achieve better health outcomes; and . . . familiarize beneficiaries with a benefit design that is typical of what they may encounter in the commercial market and thereby facilitate smoother beneficiary transition to commercial coverage” (Neale 2018). It deserves note that this argument, as described above, defies a broad scholarly consensus on the efficacy of work requirements and its interrelationship with health coverage (Waters and Escoriaza 2018).
However, there is a general legal consensus that CMS does not have the regulatory powers to approve work requirements given the current statutory restrictions on its 1115 waiver authority (Hinton et al. 2019). Indeed, the current CMS leadership has at times openly acknowledged this limitation and has indicated that statutory changes may be necessary to implement work requirements in Medicaid (Dickson 2018). Recent verdicts striking down work requirements in Kentucky and Arkansas serve to emphasize the questionable legal nature of the administration's approach (Scott 2019). Nonetheless, the Trump administration appears strongly committed to moving forward with the implementation of work requirements (Goldberg 2018). As CMS administrator Seema Verma put it, “Let me be clear to everyone in this room, we will approve proposals that promote community engagement activities,” that is, participation in work, training, educational, or volunteer opportunities (Verma 2017).
The push toward work requirements by CMS has been welcomed by a number of states. By June 2019, more than a dozen Republican-led states had taken advantage of the opportunity and submitted waivers to CMS that contain some form of work requirement (KFF 2019). These include states like Arkansas, Indiana, Kentucky, New Hampshire, and Wisconsin. This pursuit of work requirements has been particularly common in red states that have expanded Medicaid, which work by Fording and Patton (2019) attributes to the desire of Republican governors to reap the political benefits from liberals and independents for expanding Medicaid while simultaneously instilling an expectation of work to prevent backlash from conservatives.
Interestingly, each state is putting its own stamp on proposed work requirement policies—with significant differences emerging among the states in terms of work effort required, what activities count toward compliance, and exemptions (KFF 2019). Two early adopters, Arkansas and Kentucky, serve as useful examples of this variation in program design. In Kentucky, the Bevin administration implemented “community engagement” requirements within the Medicaid program that required beneficiaries to show a work effort of at least 80 hours per month (Gangopadhyaya and Kenney 2018). Certain groups were exempt from these requirements, including children younger than age 19 and adults older than age 65. Similarly, those receiving disability benefits, pregnant women, the medically frail, and primary care givers were also exempt. Implementation costs for the state were estimated to amount to hundreds of millions of dollars, with more than 100,000 beneficiaries losing coverage. In nearby Arkansas, implementation appeared to be particularly punitive and geared toward reducing enrollment in the Medicaid program (Gangopadhyaya et al. 2018; Wilson and Thompson 2018). Indeed, the state's approach has been characterized by critics as a deliberate attempt to reduce enrollment by establishing large administrative burdens for beneficiaries seeking to comply with expectations. These include exclusively relying on an online portal that is unavailable for large parts of the week in a state that is already subject to very low internet access rates. Not surprisingly, large coverage losses have been the result (DeMillo 2018). As mentioned above, both efforts have been subject to adverse judicial rulings.
The Continued Struggle over Deservingness and Personal Responsibility
As the description of the history of the Medicaid program above illustrates, perceptions about deservingness and personal responsibility have consistently influenced policy decisions about public assistance. Not surprisingly, racial undertones have also contributed. With regard to health care, these factors have worked to sort populations between (Medicare vs. Medicaid) and within (the aged vs. children and mothers vs. the poor) programs. This sorting, and the ensuing benefits and burdens based on perceptions of deservingness and political power, have been predictable (Schneider and Ingram 1997; Ingram and Schneider 2005). Critically, while program changes have occurred over time, social constructions have been remarkably persistent and self-reinforcing once implemented into public policy (Schneider and Ingram 1997; Ingram and Schneider 2005). The ACA's Medicaid expansion is the latest illustration of this point. Yet even at the elite level, a lack of consensus about the addition of work requirements to the Medicaid program has become apparent. Although written nearly twenty years ago, Grogan and Patashnik's astute observation holds true today: “The underlying conflict over Medicaid's status in the US welfare state had not been wholly resolved” (Grogan and Patashnik 2003a: 824).
With consensus toward Medicaid elusive among the elite, we assess whether a consensus has emerged at the mass level. Specifically, we explore public attitudes toward the Trump administration's efforts to promote work requirements in state Medicaid programs—analyzing not only the general tilt of public opinion on the topic but also how Americans feel about the specifics of program design. Crucially, our approach allows us to determine empirically whether Americans continue to think about Medicaid as two separate programs as it has historically been framed, one broad-based social program for long-term care for the elderly and disabled and one welfare acute health care program for the poor. While our empirical focus is on Medicaid work requirements, we argue that Americans' opinions on this issue are reflective of their broader opinions of the Medicaid program and the perceived deservingness of certain groups.
Of course, there are also more practical reasons to explore how Americans feel about Medicaid work requirements. Better understanding public opinion in this area could help policy makers considering work requirements to understand what the public thinks generally about both their implementation as well as the specifics of program design. To the extent that opinion reveals itself to consistently support or oppose the exemption of certain groups from work requirements or particular work supports to make compliance more feasible, it could guide politicians in their efforts at program design and reform moving forward. More generally, public opinion about work requirements may spill over into perceptions about the ACA's Medicaid expansion and either support or undermine its entrenchment (Patashnik and Zelizer 2013; Jacobs and Mettler 2018).
Additionally, understanding opinion in this area is important because the addition of work into Medicaid could transform opinions more broadly toward the program. While prior work by Soss and Schram (2007) in the area of welfare reform found that the implementation of work requirements and time limits did little to change public opinion toward Temporary Assistance for Needy Families (TANF), there are reasons to suspect that opinion could be more responsive to policy change in the area of Medicaid. Unlike TANF, which most Americans view as a distant but visible program that does not directly impact most Americans' lives, with only 1.2 million families receiving benefits (Falk and Landers 2019), Medicaid use is widespread, with more than 70 million Americans on Medicaid or the Children's Health Insurance Program and roughly 10.9 million gaining access to the program under the ACA (Skopec, Holahan, and Elmendorf 2018). Critically, recent work supports the idea that transformations of the Medicaid program could prove consequential. Research on the Medicaid expansion under the ACA has demonstrated important examples of policy feedback, resulting in surges in political participation and positive attitudes toward the program (Lerman and McCabe 2017; Hopkins and Parish 2019; Haselswerdt 2017). Conversely, others have found that Medicaid disenrollments may have demobilizing effects (Haselswerdt and Michener 2019). Thus, gaining an initial understanding of public opinion toward Medicaid work requirements is vital, and an important first step in gauging how program change could influence future feedback.
Data and Expectations
To assess how Americans feel about work requirements for Medicaid beneficiaries, we developed an original survey that was administered through Qualtrics.3 Qualtrics respondents come from large, online, opt-in panel services that recruit respondents to surveys and provide incentives based on the amount of effort required and the population being sampled. Qualtrics invited 10,362 Americans to take this survey, and we obtained an effective sample size of 1,377.4
While our analysis relies on a nonprobability opt-in panel, prior research suggests that high-quality nonprobability samples like ours produce similar results to probability-based surveys (Kennedy et al. 2016).5 Furthermore, while probability samples are preferred for descriptive work, opt-in samples are appropriate when modeling relationships between variables, as we are doing here (Baker et al. 2013; Coppock and McClellan 2019; Levay, Freese, and Druckman 2016; Kennedy et al. 2016). Our data are also weighted to reflect national population benchmarks on gender, race, income, and education, which we drew from the US Census 2017 Current Population Survey. Although the unweighted data are not far off from these population benchmarks, table A1 in the appendix shows that the weights we use improve representativeness.6
In our analysis of Medicaid work requirements, we rely on a wide range of explanatory measures. We account for political Ideology via a seven-point scale ranging from “extremely liberal” to “extremely conservative.” Work requirements for Medicaid recipients have been a polarizing issue, with conservative elites strongly in support and liberal elites in opposition. These cues should be reflected in mass opinion on the matter (Hacker and Pierson 2019). We hence expect more conservative respondents to be more supportive of work requirements.7
Yet, we argue that ideology does do not tell the entire story here. Social connections, personal experience, or plain self-interest can serve as important mediators of policy attitudes (Grogan and Park 2017a). These connections should play an important role in how individuals feel about Medicaid work requirements, as health is an incredibly personal issue. Personal experiences can bridge the divide between abstract government programs and the highly personal and individual experience recipients of government programs have. We thus expect that individuals who have either been on Medicaid themselves or who have members of their immediate family on Medicaid should also be more hesitant to support work requirements. Of course, individuals who are currently receiving their insurance coverage through Medicaid may be particularly affected. We therefore also include an indicator to account for whether respondents are currently part of the Medicaid program.
How an individual views the Medicaid program itself should also play an important role in one's perception of whether work requirements would make for a better program. We include two measures of individual beliefs about Medicaid as a program. First, we asked respondents if they opposed or favored the Medicaid program in general. Individuals with favorable or strongly Favorable Views of the Medicaid program should be less likely to support the addition of work requirements to the program. Second, we asked respondents whether they consider Medicaid to be a Limited-Time Program to temporarily help people in times of need; individuals who endorse this position should favor pushing people into the workforce using work requirements.
Perceived deservingness, envy, and resentment may also play an important role in how individuals feel about government benefits (Cramer 2016; Brown 1991). With many Americans struggling to make ends meet, those who are not receiving the same public assistance may grow to resent those who do. To capture this possibility, we asked whether respondents believed that Medicaid recipients were deserving or undeserving of the health benefit. Individuals who consider Medicaid Beneficiaries as Deserving should consistently be opposed to work requirements (Tanenbaum 1995). How individuals obtain their own insurance may also affect their perceptions of work requirements, particularly as private and commercial coverage has grown less generous over time (Mettler 2019; Michener 2019). This should apply to individuals who receive their insurance coverage through their employer. After all, they are working to obtain their benefits; why shouldn't their fellow Americans do the same? Additionally, the same may hold true for those on Medicare, which is often conceived as an “earned” benefit (Brown and Sparer 2003). Finally, those uninsured may exhibit resentment toward those who are able to obtain publicly financed insurance coverage (Brown 1991). As such, we include a binary indicator for the type of insurance each respondent has, with specific measures for holding Employer-Sponsored Insurance, receiving Medicare, or being Uninsured. As mentioned above, we also include whether an individual is receiving insurance coverage through Medicaid.
Our analysis also includes measures of Trust in the Federal Government and Political Knowledge. Trust in the federal government is important to include here because individuals who trust that the federal government is making appropriate decisions, in this case allowing states to mandate work requirements, should be more supportive of their implementation. Of course, in our case, the variable is also associated with the Trump administration and its policies. Trust in government is measured with a survey item asking individuals how much of the time we can trust the federal government to do what is right. We include political knowledge in our study because individuals who better understand the political world should have a better understanding of the impact of work requirements on recipients (Musumeci and Zur 2017; Hahn et al. 2017; Pavetti 2016; Moffitt 2002).8 Political knowledge is measured with a scale developed from five questions designed to tap knowledge of the political system.
Finally, resentment may take on racial undertones (Gilens 1996, 2009; Banks 2013). Racial stereotypes have long permeated public assistance policy making and public opinion, and negative stereotyping and misperceptions about African Americans are pervasive (Gilens 1996, 2009; Snowden and Graaf 2019). These stereotypes have also long influenced the perceptions and policy making of Medicaid (Kousser 2002; Olson 2010; Leitner, Hehman, and Snowden 2018; Barrilleaux and Bernick 2003). Yet the election of the nation's first African American president further exacerbated the situation, with the politics of the ACA suffering from significant racial bias and the expansion of Medicaid particularly affected (Lanford and Quadagno 2016; Grogan and Park 2017b; Snowden and Graaf 2019; Fording and Patton 2019; Tesler 2012; Banks 2013; Henderson and Hillygus 2011; Pasek et al. 2009; Segura and Valenzuela 2010). To account for this reality, our analysis includes a measure of Racial Resentment, which is assessed in our analysis using a commonly used scale created from four questions tapping racial predispositions (Kinder and Sanders 1996). However, the scale was developed to assess the racial resentment of whites. For that reason, we estimate models separately for the population as a whole without racial resentment and again separately for non-Hispanic whites including the racial resentment measure. We suspect that higher levels of racial resentment should be associated with higher levels of support for work requirements.
In addition, we also control for a variety of other variables in our models. These include gender (a dichotomous measure with Female coded as 1); race (using separate dichotomous indicators for whether participants are Black or Hispanic); Age and its squared term, Religiosity (a 5-point scale ranging from “very inactive” to “very active”); Education (a seven-point measure capturing highest level of schooling completed); and household Income (a ten-point scale ranging from less than $5,000 to a high of greater than $150,000). Standard tests for outliers and influential observations were satisfactory and tests for multicollinearity indicated it was not a problem in our models.
Black and White: Medicaid Work Requirements
We begin our analysis by exploring public attitudes about the implementation of work requirements in the abstract. Do Americans support expanding mandatory work requirements, common in other public assistance programs, into Medicaid?9 To answer this question, we asked respondents whether or not work participation should be mandatory to obtain Medicaid benefits or whether they should remain voluntary for individuals.10 Critically, we wanted to capture informed public attitudes. For that reason, the survey includes a preamble about what Medicaid is and additional information about Trump administration's current efforts on work requirements.11
Table 1 presents the findings from our survey.12 We estimated a standard logit model where 1 indicates support for mandatory work requirements in Medicaid.13 Importantly, logit coefficients are not directly interpretable. Therefore, we also present the average marginal effect (AME) for those cases where they are statistically significant at conventional levels.14 AMEs are useful because they offer an easy interpretation of the direction and effect size of variables analogous to coefficients in traditional OLS models (Long and Freese 2014). They are also the recommended interpretation for policy-relevant work (Cameron and Trivedi 2010). For indicator variables, we calculated AMEs for changes from 0 to 1. For continuous variables, we calculated AMEs for changes by one standard deviation. Finally, for ordinal variables, we calculated AMEs for discrete changes by one unit. As mentioned above, we estimated one model for all respondents and a separate model for non-Hispanic whites that includes the racial resentment scale.
First, we note that the average predictions for support and opposition to work requirements in Medicaid are rather balanced in our data. That is, the average prediction is 0.483 in support and 0.517 in opposition. For non-Hispanic whites the numbers are 0.530 and 0.470, respectively. When it comes to individual predictors, we find statistically significant effects for four of the variables, two of which increase the probability of supporting work requirements and two having the opposite effect. Not surprisingly, individuals who are more conservative are more likely to support mandatory work requirements. Utilizing the AMEs for interpretation, a discrete change by one standard deviation in ideology toward conservatism increases the probability of supporting work requirements by 9.3 percentage points. The effect is statistically significant at the p = 0.000 level. Moreover, those who consider Medicaid as a limited-time program are 11.0 percentage points more likely to favor work requirements (p = 0.001). Other variables were found to significantly reduce support for Medicaid work requirements. For one, individuals on Medicaid are 12.9 percentage points more likely to oppose the introduction of work requirements (p = 0.033) while those who consider Medicaid recipients as deserving are 14.0 percentage points more likely to oppose work requirements (p = 0.002).
In the model that includes only non-Hispanic whites, we find the average marginal effects for the same four variables to be statistically significant as in the previous model. Effect sizes are similar. Moreover, non-Hispanic white women are 9.1 percentage points more likely than men of the same ethnic group to support work requirements (p = 0.010). Importantly, an increase in racial resentment by one standard deviation increases the probability of supporting work requirements by 8.1 percentage points (p = 0.000).
To further illustrate the effect of the individual variables, we plotted the predicted probabilities and their 90% confidence intervals in figure 1. That is, we held all variables at their mean or median value while varying the variables of interest either from 0 to 1 for indicator variables (whether an individual is on Medicaid, whether respondents thought Medicaid recipients are deserving, and whether respondents thought Medicaid is a time-limited program) or from low-to-high values for other variables (education, ideology, and racial resentment). For all dyads presented, differences of predicted probabilities are statistically different from one another at the 90% level.15
Shades of Gray: Who Should Be Exempt?
Our previous analysis assessed support for Medicaid work requirements in general. Yet, the concept of Medicaid work requirements is a complex one that may mean many different things to different people. Forcing individuals into strict black-and-white answers could obfuscate much of the nuance behind people's thinking and the complex nature of work requirement design. Moreover, respondents are much more likely to receive cues from political elites about the issue in the abstract than from the underlying details. Hence, we should be able to better determine the true underlying perceptions of individuals when asking them about policy details. In our survey, we asked respondents whether certain groups of individuals should be exempted from work requirements. Based on the history of public assistance in general and the Medicaid program in particular, we reviewed the current debate in the various states contemplating work requirements and designed a set of questions to gauge public opinion about the specifics of program design and the groups to be affected by work requirements (Aron-Dine, Chaudhry, and Broaddus 2018; Gangopadhyaya et al. 2018).
We began our analysis of work requirement exemptions by exploring groups that are generally considered to elicit sympathy from Americans. In our survey, these included Seniors, the Disabled, and those suffering from Long-Term Illnesses.16 Groups in this set are often described as having social constructions that are deserving and generally powerful (Schneider and Ingram 1997). Next, we asked respondents about exemptions related to children. These included potential exemptions for Pregnant Women as well as Caretakers of Children under Age 1, Age 6, and Age 18. As illustrated above, coverage expansions have historically focused on utilitarian arguments. Children, mothers, and caretakers generally carry a deserving connotation, though they comprise a group with weak political power (Schneider and Ingram 1997). Finally, we included questions about exemption for groups often perceived to be at the low end of the deservingness scale: Individuals Living in Poverty, Individuals Living in Areas of High Unemployment, and Individuals Suffering from Addiction. These individuals are also politically weak (Schneider and Ingram 1997). Table 2a shows the results from a number of logistic regressions for each of the ten groups for all respondents while table 3a shows the analogous results for non-Hispanic whites only.
Several observations stand out. Before interpreting individual predictors of public opinion, we first focus on the average predictions at the bottom of the tables, which hold important information about the levels of deservingness Americans assign to the respective groups. That is, the average predictions allow us to rank the various groups from high levels of deservingness to low levels. First, the disabled (0.782) and seniors (0.716) hold the highest average support for exemption from work requirements, followed, with some difference in value, by long-term care patients (0.616). Second, caregivers to children follow at some distance in logical order: pregnant women (0.532), caregivers to children younger than age 1 (0.407), age 6 (0.340), and, finally, age 18 (0.173). Third, the set groups, which we described as least deserving above, score low: average predictions range from 0.327 for those living in poverty to 0.191 for those suffering from addictions. The rankings for non-Hispanic whites (table 3a) are essentially identical. Figure 2 further illustrates this ranking by graphing the predicted probabilities at mean and median values for the 10 groups for all respondents and non-Hispanic whites. The figure highlights the preeminent positions that Medicaid beneficiaries who are disabled, seniors, affected by long-term illness, and pregnant women hold in the perceptions of Americans. We note that even when the variables with statistically significant associations (described below) are set to their most extreme negative value (i.e., in opposition to exemptions), predicted probabilities remain above 0.5 for the aged, sick, and disabled. At the same time, when they are set to their most extreme positive values (i.e., in support of exemptions), predicted probabilities for the set of least deserving groups do not cross 0.5.
We next turn to individual predictors associated with supporting exemptions from Medicaid work requirements for the various groups. As logistic coefficients hold limited interpretative value, we again present the average marginal effects for those cases where they are statistically significant at conventional levels in tables 2b and 3b. For all respondents, one negative and two positive predictors immediately stand out. Perceptions of deservingness and higher levels of political knowledge are consistently associated with support for exemptions while the opposite holds for those who consider Medicaid a temporary program. Moreover, higher levels of conservatism are associated with lower support for exemptions five times (see figure 3). It also deserves noting that opinions for the top three deserving groups (disabled seniors, those with long-term illnesses, and pregnant women) do not vary by ideology. Ideology, however, plays an important role in assessing exemptions for the five groups with the lowest support for exemptions (caregivers of children younger than 6 or 18, the poor, the addicted, and those living in high-unemployment areas). Opposition to exempting pregnant women exists among those with employer-sponsored insurance, those on Medicare, and the uninsured. Interestingly, income, education, and age of the respondent do not appear to affect support for exemptions at all.
For non-Hispanic whites the results are similar, except that conservatism holds less predictive value; indeed, it only persists for those in poverty. Instead, racial resentment becomes a statistically significant predictor in five cases. We highlight the importance of racial resentment as a predictor for opposition to exemptions in figure 4. Comparing non-Hispanic whites with low levels of resentment to those with high levels of resentment, statistically and substantively significant differences become apparent for exempting individuals in areas of high unemployment and dealing with addictions as well as those caring for children younger than 18. We note that these are among the least deserving groups, which are also often associated with African Americans and other minorities (Gilens 1996, 2009; Snowden and Graaf 2019). In some other cases, like for those living in poverty, differences are substantial, but we are limited in their assessment by large confidence intervals.
Should We Lend a Helping Hand?
Our analyses have established the predictors for support of Medicaid work requirements in the abstract. We have also demonstrated the importance of deservingness, perceptions of what the Medicaid program is, and racial bias in how Americans non-Hispanic whites perceive the Medicaid program and attitudes toward exemptions. Next, we asked respondents to focus specifically on transitioning Medicaid recipients into the workforce. That is, we asked how they felt about supporting Medicaid recipients during the process of entering the workforce by offering them help with childcare, transportation, clothing, tuition, and job training. Do they prefer the approach taken by the Trump administration that provides no support at all and merely relies on sanctions to induce compliance, or do they believe that a helping hand is needed?
We again present the average predictions for the various forms of assistance to support entering the labor force in tables 4a and 4b. Predicted probabilities at mean and median values are presented in figure 5. Overall, support for three of the five forms of work support exceeds 0.500, with assistance for childcare (0.647 for all respondents and 0.682 for non-Hispanic whites), transportation (0.567 and 0.576), and especially job training (0.738 and 0.784) finding large levels of support. These findings thus appear to repudiate the approach taken by CMS and the Trump administration, which does not establish similar work supports. If work requirements are implemented, it would appear that Americans want beneficiaries to have access to support that makes compliance more feasible.
Table 4a illustrates the results of another series of logit models for all respondents where 1 indicates support for each assistance activity. The patterns from previous analyses are repeated. Individuals who consider Medicaid as a temporary program are opposed to providing work supports while those who consider Medicaid recipients as deserving are in favor. Higher political knowledge is associated with more support in three cases but negatively signed for clothing assistance. We also find higher levels of conservatism to be negatively associated in two of the five cases while enrollment in Medicare is negatively signed twice. Women are also substantially more supportive of childcare support while the opposite holds for those with high levels of church attendance. Once more, we find no evidence of association for personal connections to the Medicaid program. Table 4b presents the analogous results for non-Hispanic whites, which are essentially similar. However, we note that racial resentment is only statistically significant for job training while predicted probabilities at mean and median values (figure 6) show no differences between those with high and low levels of resentment.
Medicaid work requirements have become a dominant issue in health care policy since the Trump administration took office. The complexities illustrated by our in-depth analysis of what Americans think about the issue are remarkably policy relevant. Notably, states seeking to implement work requirements are confronted with the very same nuances we analyzed here (Haeder 2019b). While not as publicly prominent as the largely failed repeal of the Affordable Care Act (Rocco and Haeder 2018), work requirements still hold tremendous implications for states, beneficiaries, and the broader health care system. Increases in administrative burden from poorly designed and administered work requirements could lead to large coverage losses, disruption of access to vital services, and poorer health outcomes (Waters and Escoriaza 2018; Herd and Moynihan 2018). These losses may come with important implications for mobilizing Americans to participate in their democracy (Haselswerdt and Michener 2019). Importantly, these efforts seem to be part of a larger policy project that seeks to decrease public assistance through the back door. Yet enrollment losses go beyond the experience of individual beneficiaries because they put even greater burdens on crucial safety-net health care providers like Federally Qualified Health Centers and essential hospitals to care for the uninsured (Haeder 2019a; Rosenbaum 2018).
As our analysis shows, Americans are largely split on work requirements in the abstract. Yet important nuances emerge when Americans are confronted with the various groups who benefit from the program. Not surprisingly, with abundant partisan cues on the issue, conservatism serves as an important predictor for support of requiring work as a trade for program eligibility, as do perceptions about the nature of the Medicaid program as a temporary program. In a few cases, higher levels of religiosity and status as Medicare beneficiaries are associated with higher levels of support. The latter may be a result of the long-established bifurcation between the two programs, one “earned” and one “handed out,” while the former may be connected to the importance many religious groups place on work. On the other hand, opposition to work requirements consistently holds for those with higher levels of political knowledge. Surprisingly, we find little association for personal connections to the program, except when asked about work requirements in the abstract. Importantly, our analysis shows that perceptions of deservingness and racial bias continue to shape Americans' opinions of the Medicaid program itself as well as its beneficiaries. These findings are in line with the ambiguous legacies of the Medicaid program and the provision of public assistance more broadly.
We believe this latest point to be especially relevant for future attempts to reform the US health care system, particularly those that seek to expand insurance coverage to the remaining uninsured. As described above, the politics of the ACA have been influenced by racial biases, both against former President Obama as well as potential beneficiaries. Given the long, interconnected history of perceptions of deservingness and racial resentment, it seems unlikely that any future reform would not trigger similar backlashes, albeit perhaps somewhat more muted if passed under a non-Hispanic white president. This, of course, raises significant political concerns for those seeking to move toward more universal coverage. Importantly, policy retrenchments are equally subject to these biases, as shown in our analysis here, holding the potential to further exacerbate health inequities when access is reduced. We note, however, that Americans appear to be much less punitive than the Trump administration and many red states eagerly seeking to transform their Medicaid programs. That is, we find considerable support among Americans for efforts to support beneficiaries and to make finding and maintaining work more feasible including transportation assistance, help with childcare, and access to job training.
Our analysis also points to a need for additional research on health reform during the Trump era. Even as considerable research has demonstrated that Medicaid affects registration, turnout, participation in civic associations, and policy attitudes (Haselswerdt and Michener 2019; Michener 2017, 2018; Hopkins and Parish 2019; Clinton and Sances 2018), we need to better understand whether and how the Trump administration's systematic changes of the health system in the United States have changed how Americans participate in the political process and their attitudes about the future of the health care system in the United States. Of course, these changes are likely to interact with the lingering effects of the coronavirus pandemic as well as issues of racial equity and justice. Building on our work here, we believe that important future contributions can be made by studying public attitudes about other Trump administration health efforts including repealing the ACA, combating the opioid epidemic, reducing prescription drug prices, and responding to disease outbreaks like the coronavirus.
Additional work should also be done to study how the framing of work requirements by politicians influences support for their implementation in the states. Notably, how does the decision to frame work requirements in terms of fairness, cost savings, or deservingness influence public support? In addition, future research should test if and how opinion on work requirements varies when state-specific Medicaid program names are used. While we do provide respondents with background information about Medicaid to gather informed opinions, past research by Grogan and Park (2017, 2018) suggests that opinion responses can vary based on connection to the program, and using state-specific names (i.e., Medi-Cal or BadgerCare) could increase connectiveness and alter responses. That said, our responses on items like Medicaid favorability, program effectiveness, and personal connection are very much in line with findings from the Kaiser Family Foundation that utilized state-specific program names (KFF 2017, 2018).
Of course, there are several limitations of this analysis that need to be acknowledged. First, LGBT individuals are excluded from our sample. To the extent that LGBT individuals think differently than non-LGBT individuals about work requirements, our results could be biased. Next, considerable research demonstrates that question wording and the amount of information provided to participants can shape survey responses (Schuldt, Konrath, and Schwarz 2011; Smith et al. 2018). It remains possible that altering the amount of information provided to participants about Medicaid and work requirement design could alter public attitudes. Similarly, perceptions of deservingness may be shaped by what type of Medicaid recipient the respondent has in mind (Huber and Paris 2013). As such, additional research that tests the robustness of our findings using alternative information patterns and respondents' perceptions would be beneficial to the literature. Finally, it is critical to acknowledge that the status of work requirements is very much in flux. With Republican states around the nation considering their implementation, ongoing court battles over the legality of work requirements, and pending court decisions over the fate of the broader health care landscape, attitudes on these work requirements could continue to evolve. If they do, patterns of opinion identified here may not hold into the future. Therefore, continued monitoring of public opinion in this area is warranted.
Finally, when we asked respondents about their attitudes about work requirements, we specifically asked them the following question based on a number of previous polls conducted on the issue: “As you may know, the Trump administration is allowing some states to change their Medicaid program to require most adults without disabilities to be working or looking for work to get health insurance through Medicaid. Some states have made this mandatory while others have made this voluntary. In your opinion, should participation in these work requirement programs be voluntary or mandatory for individuals on Medicaid.” It is worth noting that the wording we (and other polls) used inevitably may predispose individuals to be drawn into the usefulness of work requirements, that is, many respondents may become favorably disposed. In our view, this perception may be in line with the underlying American ethos toward public assistance and the value of work. Yet it deserves noting that this value framework, supported even by many Democrats, is clearly to the right of most Western democracies. Unquestionably, it may also disguise some nuances in the minds of respondents about who qualifies as disabled or suffering from long-term illness, for example.
Even with these limitations, our research still provides important information about public attitudes on an emerging policy debate currently going on in the country. Our findings suggest that even as work requirements could fundamentally transform Medicaid across the country, the underlying politics of the American welfare state remain unchanged, with support driven by partisan cues, perceptions of deservingness, and racial predispositions.
Mothers and infants, of course, also benefited from the temporary Federal Emergency Maternal and Infant Care Program, a grant program that provided medical care to most of the families of US soldiers during World War II (Dutton 2007).
It is worth noting that the ACA, as passed, technically does not eliminate the historical categorical linkages in public law but rather adds the additional novel income-based pathways. Neither have we altered eligibility determinations for any other assistance programs to come in line with those of the ACA-adjusted Medicaid program (Orentlicher 2015).
Qualtrics respondents are obtained using respondents from the survey panels of Research Now and Lucid. Supporting the quality of our sample, new research by Coppock and McClellan (2019) indicates that data from Lucid—which Qualtrics used to select participants—tracks well with population benchmarks and “is suitable for evaluating many social scientific theories.”
Our survey also included an oversample of 754 LGBTQ individuals for a separate research project. We exclude this LGBTQ oversample from our analysis here, making our effective sample size 1,377. Because our goal in this study is to generalize to the US adult population, we exclude all oversampled respondents from our analysis.
Interestingly, experimental comparisons by Pew have found similar results between nonprobability and probability-based samples in the area of health, with estimates of health coverage for a variety of nonprobability samples close to a probability benchmark sample. The research also found that the nonprobability samples had high levels of predictive accuracy in the area of health coverage (Pew Research Center 2016).
The institutional review boards at Texas A&M University, West Virginia University, Pennsylvania State University, and Utah Valley University approved the study.
We note that our findings are essentially similar when we estimate our models with a partisan identification variable.
Some recent research (see Lupia 2016) has questioned the value of general political knowledge scales in place of issue-specific knowledge scales, arguing that low scores on the former do not necessarily reflect an inability to competently carry out political tasks. However, we use the standard political knowledge battery to remain consistent with most research in the field of political science and, because our study provided information to respondents about Medicaid and work requirements, making an issue-specific knowledge test impractical.
Specifically we asked, “We would now like to ask you some questions about Medicaid. Medicaid is a health care program that provides insurance coverage to millions of low-income Americans as well as to children, pregnant women, and many individuals with disabilities. The program is funded jointly by the states and the federal government.” We reviewed a number of national polling organizations' wording on the issue, including the Kaiser Family Foundation, Politico/Harvard T. H. Chan School of Public Health, Politico/Morning Consult, and Center for American Progress (Wogan 2018). We settled on this word choice to appropriately reflect what populations Medicaid covers and how it is funded. We believe that the wording choice is ideologically moderate. Findings from other polls varied widely from 39% to 72% in support of work requirements (Wogan 2018). Our results put us squarely in the center of the distribution.
Specifically we asked, “As you may know, the Trump administration is allowing some states to change their Medicaid program to require most adults without disabilities to be working or looking for work in order to get health insurance through Medicaid. Some states have made this mandatory while others have made this voluntary. In your opinion, should participation in these work requirement programs be voluntary or mandatory for individuals on Medicaid?”
By providing respondents with information about Medicaid and work requirements, we help to reduce unreliable responses from uninformed participants in our study. This is important because work requirements are a policy that affects many Americans, regardless of their prior knowledge of recent efforts, and we want an accurate picture of what respondents think of the policy itself. That said, to the extent that uninformed opinions about work requirements vary from informed attitudes, our results could vary from studies that ask similar questions without providing a preamble of relevant information. Exploring these nuances is an important direction for future research.
Differences between our overall sample size of 1,377 and sample sizes across tables reflect small amounts of missing data.
We estimated a number of different specifications to account for the potential that the opinions of individual respondents may be associated with actions taken by the states they reside in. The nonlinear nature of logit models does not allow for the inclusion of fixed effects due to the commonly known incidental parameter problem (Neyman and Scott 1948). We hence also estimated linear probability models with state-fixed effects and hierarchical models with random intercepts as well as random intercepts and random slopes for a number of specifications. These alternative specifications returned essentially similar results for the coefficients of interest and indicated that hierarchical modeling was inappropriate.
Average marginal effects are obtained by “comput[ing] the marginal effect of xk for each observation at its observed values xi, and then comput[ing] the average of these effects” (Long and Freese 2014: 242).
To assess whether predicted probabilities are statistically different from one another, we used Stata's mlincom command (Long and Freese 2014).
We note that the definition of terms such as the disabled may differ significantly among individuals. However, for the purposes of our analyses, determining the effect of perceptions of deservingness, congruence among respondents is not required. That is, we simply need the respective group to elicit a certain set of perceptions for each individual respondent.