Against a backdrop of ongoing operational challenges, insurance market turbulence, and the ever present pull of partisanship, enrollment in the ACA's programs has soared and significant variations have developed across states in terms of their pace of coverage expansion. Our article explores why ACA enrollment has varied so dramatically across states. We explore the potential influence of party control, presidential cueing, administrative capacity, the reverberating effects of ACA policy decisions, affluence, and unemployment on enrollment. Our multivariate analysis finds that party control dominated early state decision making, but that relative enrollment in insurance exchanges and the Medicaid expansion are driven by a changing mix of political and administrative factors. Health politics is entering a new era as Republicans replace the ACA and devolve significant discretion to states to administer Medicaid and other programs. Our findings offer insights into future directions in health reform and in learning and diffusion.
Since the enactment of the Affordable Care Act (ACA) in 2010, its implementation has faced a series of policy and political hurdles. One obstacle is partisanship—the “workhorse” of American politics (Bafumi and Shapiro 2009). Congressional Republicans continue to uniformly oppose the ACA and most Republicans nationwide share that antipathy. This intense, sustained partisan hostility has been particularly threatening to health reform because Republicans control Congress as well as all branches of government or at least one of the levers of power in forty-three of fifty states (86 percent).
The fiery partisan resistance is compounded by persistent—if shrinking—website snafus and administrative problems combined with nearly unrelenting press scrutiny that spotlights the failings of the program. Political opponents as well as independent experts predicted that the balky implementation would discourage enrollment; some Republican leaders went as far as to call “the rollout of the Affordable Care Act . . . an unmitigated disaster,” and insisted that the law would not work (Fox News 2014).
The double whammy of implementation snags and partisan roadblocks reinforce other well-known limits on individual engagement that worked to obstruct the diffusion of health reform across states. Gathering the kind of information that would be necessary to sort through coverage options and then sign up imposes substantial costs; prior research demonstrate that these costs often deter other forms of political and policy participation. In addition, accurate learning about policy is often skewed away from those most benefiting from new policies such as the ACA and toward the higher income strata that have less need for coverage. Moreover, negative framing (as illustrated by the attacks on the ACA) activates perceptions of risk, which have been connected in prior research to distrust and disengagement (Kahneman and Tversky 1984; Fiorina 1990; Popkin 1991; Cappella and Jamieson 1997; Kuklinski and Quirk 2000; Druckman 2004). These cognitive and communicative difficulties are intensified by the actual costs that loom over individuals who fail to enroll in insurance and face the threat of penalty.
Expectations about enrollment in the ACA should be humbled by the substantial impact that comparatively mild hurdles (such as registering to vote) exert in discouraging participation (Downs 1957). Put simply, the multiple hurdles—intense partisan polarization, implementation stumbles, and limits on individual cognition and behavior—should lead us to expect low enrollment in the ACA.
The ACA raises a general puzzle: its drive to engage individuals in enrolling for coverage should, according to prior research, be severely hampered by these hurdles and yet, large and sharply rising numbers of people are taking on taxing barriers and enrolling. During the first enrollment period that started in October 2013, 7.3 million people decided to sign up and pay their premiums on insurance exchanges. During the next year, exchange sign-ups rose by 27 percent to 9.3 million and by year 3, sign-ups for the exchanges rose to 12.7 million (actual enrollment may decline to approximately 10 million because individuals fail to pay their premiums or other issues) (US Department of Health and Human Services 2014 and 2015; Galewitz 2016). Over the same period, the state-based expansion of Medicaid experienced a dramatic growth of coverage for low-income Americans, with 14.5 million individuals gaining Medicaid coverage from 2013 to 2015 (CMS 2016).
Why is enrollment in the ACA as high as it is despite menacing conditions that generally discourage participation? Moreover, what accounts for the variation across states in enrollment given the federal government's incentives and mandates? With enrollment exceeding expectations, we need to explain why ACA participation is more extensive in Vermont and Massachusetts than in Alaska and Oklahoma even though the latter states exhibit a far larger pool of uninsured to tap.
The differential impact of federal government incentives on states to implement the ACA poses important puzzles for students of health policy and analysts of diffusion. On the one hand, the ACA fits within past research on vertical diffusion—the federal government's use of financial incentives and mandates to coax states into implementing its new policies (Allen, Pettus, and Haider-Markel 2004; Karch 2006; Shipan and Volden 2006; Shipan and Volden 2008).1 On the other hand, however, the implementation of the ACA has occurred against a backdrop of formidable obstacles that may disrupt common patterns in policy diffusion and produce striking variations across states in the successful implementation of the ACA's programs.
In the discussion below, we explore general explanations for political and policy development and their feedback effects. We then use these bodies of research to structure our analysis of state ACA enrollment in Medicaid and the exchange. Our analysis confirms earlier findings that party control of state government drives early state decisions about the type of exchange to establish and whether or not to pursue Medicaid expansion, but it also reveals that partisanship is less influential in capturing the unique variations in state enrollment in each program. For the exchange, we find that enrollment is lower in states with high unemployment and higher in states that are more supportive of President Obama, after controlling for party control of government. For Medicaid, we find that variations in enrollment are driven by new policy and strengthened administration, Democratic control of the legislature, and high unemployment. Although our research was conducted before the Republicans gained unified control in Washington and pursued their plans to repeal and replace the ACA, our findings about the variations in ACA enrollment offer important insights into the future of health reforms that are devolved to the states.
Explanations for Strenuous Participation
It may appear to be simple common sense for America's low- and middle-income people to sign up for new health benefits: better coverage of needed medical care and, for most people, lower costs—Medicaid is free for those under 138% of the federal poverty line, and 86% signing up on exchanges in 2015 received subsidies to purchase insurance (KFF 2014; Pear 2015).
Extensive research points to a number of reasons, however, that individuals and groups fail to invest in gaining the necessary information and refrain from taking action despite intuitive payoffs. One barrier is a lack of adequate interest and information to act—even if participation is a “good deal.” Awareness of potential payoffs from participation often elude those who would most benefit, and the information that is available is routinely incomplete, imperfect, and clouded by uncertainty about its veracity (Converse 1964; Delli Carpini and Keeter 1996).
Opponents of change work hard to fend off threats to the status quo that serves them. Depending on the repertoire of available tools, they may use intimidation through negative sanctions and physical retribution, and sow doubt through misinformation in hopes of raising the costs of change and creating a distraction from the payoffs (Tilly 1978; Gaventa 1982). Opponents of reform in America can tap a deep well of distrust in government that they can activate by spotlighting costs and risks against incumbent officeholders and new programs (Cappella and Jamieson 1997; Hetherington 1999; Hetherington and Globetti 2002). There is also a repertoire of administrative tools to dampen participation in American social policy. Poor relief as far back as sixteenth-century England literally branded those who sought assistance. In America, the process of receiving welfare is notoriously convoluted, demeaning, and punitive in order to dampen requests for aid from eligible people (Soss 1999; Soss, Fording, and Schram 2011). Enrollment in Medicaid, which is administered by states, has long exhibited substantial gaps between eligibility and actual enrollment in part because states developed laborious and confusing steps that deterred participation (e.g., Selden, Banthin, and Cohen 1998).
Decisions by Americans to seek coverage in the new ACA programs faced intense and familiar barriers. After two years of enrollment, public knowledge remained limited, especially among the most affected—about half of the uninsured reported not knowing when to sign up for insurance or about the payoff of the new benefits for them (KFF 2015, fig. 16). What awareness and knowledge that did exist was compromised by aggressive efforts by reform opponents to highlight website problems (even if resolved) and to prime political distrust of Washington; this has contributed to disapproval of the ACA by a consistent three-quarters of Republicans and nearly half of independents (KFF 2015, fig. 2). Moreover, the ACA funded states to designate “navigators” who would work with target populations to sign them up for new programs, but some southern states that harbored the largest uninsured populations obstructed the efforts of navigators (Seitz-Wald 2013; Emanuel 2014). All of these hurdles to seeking the payoff of health coverage are higher for vulnerable populations where awareness and knowledge are low and distrust of the government is particularly high (Doty, Rasmussen, and Collins 2014).
Despite the imposing impediments, large numbers of Americans decided to invest the time and effort to sign up for insurance on exchanges and to register for the expanded Medicaid program. This rising coverage did not occur equally across the country; it rose more sharply in some states and less in others. What accounts for the higher participation in ACA coverage and the variations across states?
Four accounts may explain both the increasing number of Americans who enrolled in ACA programs despite impediments as well as the variations across states.
Partisanship is the “workhorse” of American national and state politics (Bafumi and Shapiro 2009), sorting lawmakers and voters into consistent blocks. The rising partisan polarization in Washington has now spread to states; the partisan conflict over salient policies in states can over-ride or significantly offset financial and other attractions of new federal programs (Deering and Shelly 2009; Layman et al. 2010; Shor and McCarty 2011; Doan and McFarlane 2012). Party exerts a strong and sustained hold on government officials who face elections because of party activists: they insist on fidelity to single issues and reward or punish lawmakers who compromise these core policy goals by withholding campaign contributions, volunteer help, and support during the nomination process. Constituency may modify, however, the pull of partisanship: legislators are beholden to narrow districts that might be particularly ideological while governors are elected by broader state constituencies.
The ACA poses a stark test of partisanship given the law's generous financial terms. On the one hand, the federal government offered large grants to plan and set up state insurance exchanges as well as new Medicaid funding to cover all of the costs for extending coverage for the first three years and, afterward, 90 percent of expenses. In addition, the federal government's new funding of coverage reduces enormous and rising pressure on state budgets to pick up the cost of “uncompensated care” for the uninsured; states that refused to adopt the new funding from the expanded Medicaid program are experiencing rising rates of hospital closures (Angeles 2010; Conlan and Posner 2011; Reiter, Noles, and Pink 2015). This kind of exceptionally generous financing typically energizes the pressure groups at state capitols that stand to gain and convinces lawmakers to accept the new federal support (Thomas and Hrebenar 1999; Singhal 2008).
On the other hand, the sharp partisan divide that enveloped the ACA since congressional debates in 2009 and 2010 when no Republicans voted for its passage has persisted at the national level in Congress and in states (Oberlander 2011; Jacobs and Callaghan 2013; Rigby and Haselswerdt 2013; Frean, Gruber, and Sommers 2016; Jacobs and Skocpol 2016). Although Republican governors in Ohio, Arizona, and a few other states pressed for the adoption of Medicaid's expansion to secure federal financing and respond to pressure from medical providers, the political influence of party propelled many Republican governors and nearly all GOP-controlled legislatures to resist health reform. With Republicans enjoying one-party control in twenty-four states, and wielding a veto through their control over at least one branch of government in nineteen more states, elite partisanship generally obstructed the ACA's implementation. States controlled by Democrats—such as Connecticut and Rhode Island—adopted the expansion of Medicaid, and most established state exchanges; among states where Republicans leaders controlled all levers of power, most have rejected state exchanges (with the exception of Nevada, Michigan, Idaho, and Arkansas) and two-thirds (17 of 24) refused to approve the new Medicaid benefits.
Party control of government conditioned ACA enrollment. Republican control ruled out the option of state insurance exchanges and often precluded the opportunity to seek expanded Medicaid coverage, while individuals in states run by Democrats enjoyed more opportunities to seek new health benefits. In addition, some Republican-controlled states actively discouraged their residents from registering on federal insurance exchanges (Seitz-Wald 2013).
There are some indications, however, that there has been divergence in the motivation of Republican policy makers. Governors who represent the entire state and a wide range of interests (including businesses, medical care providers, and uninsured who seek medical care funded by the state) may diverge from Republican legislators who represent a narrow district and may be especially attuned to the Tea Party and other conservative activists (Skocpol and Williamson 2012). For instance, Republican governors in Ohio and Arizona pushed their states to adopt Medicaid expansion even though the GOP-controlled legislatures and conservative activists opposed the reforms. While Republicans from both branches of government are prone to be less supportive of health reform than their Democratic counterparts, GOP governors appear more willing to accept the federal government's generous financial terms and implement reform than their legislative counterparts.
Presidents enjoy unrivaled public attention and can influence, if they are popular, how Americans evaluate government policy as well as members of Congress and other politicians. A long line of research finds that presidents with strong support in elections or favorable approval ratings affect public support of members of Congress from the president's party; conversely, antipathy to the president prompts Americans to vent their anger against them by, for instance, voting against the president's partisans in Congress (Fiorina 1983; Abramowitz 1985; Cover 1986; Gronke, Koch, and Wilson 2003).
Presidents exert an independent effect that “cues” the public, beyond the influence of party affiliation. As one study explains, “Citizens look to the President for a cue as to how the country is doing, how things are going in Washington, and how well Congress is doing its job” (Gronke, Koch, and Wilson 2003: 805). The effect can be significant: the president can cue Americans as they assess new policy initiatives and competing candidates. Cueing on presidential support is a cognitive shortcut for Americans with low levels of information; it gives them some basis to assess complicated government programs (Popkin 1991).
Public support for President Obama may have had particular relevance to the public's evaluations of the ACA and commitment to enroll in its programs. Rarely has a new program been as closely associated with a president as the ACA, which is widely referenced as “Obamacare.”
The cueing account anticipates that states with higher levels of support for Obama will experience higher enrollment. Given the barriers to enrolling in the exchanges and Medicaid, Democrats or independents who harbor reservations about the ACA but support Obama may be more willing to take on the taxing barriers to enroll in the ACA's programs. This dynamic may even apply to a relatively small number of Republicans who are “weak partisans” but respect the president. In these cases, Americans are following the cue of President Obama despite their partisanship or, in the case of Democrats, their reservations about health reform.
Feedback Effects and Institutions
A growing body of research reveals that established government programs are not just a product of politics but also influence politics. The “feedback effects” of established programs generate resources and motivation for potential program beneficiaries to become aware of new programs and invest the time in enrolling in them and then supporting them. The passage of Social Security in 1935, for instance, influenced the decisions of senior citizens to learn about the program and join it as well as to increase their political engagement, which produced the higher levels of voting that are evident today (Campbell 2003; see also Mettler and Soss 2004).
In addition, historical institutionalists find that existing administrative capacity fosters “path dependence” that predisposes policy developments along established paths (Skocpol and Ikenberry 1983; Skocpol 1992; Pierson 2000). In particular, hierarchical structures that facilitate coherent direction, as well as well-trained and rule-guided civil servants, enhance the confidence and capability of lawmakers and citizens in pursuing new programs. For instance, the Medicare program of hospital insurance for seniors was constructed in the mid-1960s on the existing and widely respected Social Security system for administering the financing and delivery of benefits (Jacobs 1993). Research on state health policy reveals that competent administrative capacity generates support among policy makers for creating effective eligibility determination, enrollment processing, provider reimbursement, and quality assessments (Gold, Sparer, and Chu 1996; Holahan et al. 1998; Miller 2005).
States with a history of competent administration and adopting generous health policy created a trajectory of policy change that facilitated the adoption of the expanded Medicaid program and the establishment of state-based exchanges. Research on path dependence suggests that states with a history of effectiveness and health reform both boost the confidence of lawmakers in the feasibility of policy change and distribute tangible benefits in a way that activates organized groups and voters to press for further reform. In addition, states with well-formed health reform trajectories motivate lawmakers and civil servants to invest in competent and coherent administration. In the case of the ACA, prior research suggests that these states will use available federal funding to strengthen their administrative capacity to subsidize the cost of enrolling by distributing information, paying “navigators” to provide assistance, and staffing call centers. As a result, states with higher capacity are expected to handle enrollment in a smooth and consistent manner that enables more individuals to sign up for ACA coverage; states with weak administration obstruct enrollment by leaving individuals to absorb the costs of signing up.
State Economic Circumstances
The economic prosperity of states conditions the generosity and scope of policy making, according to a number of studies. Research finds that states that are affluent with robust employment are, in general, more innovative and receptive to developing new federal initiatives, while less well-off states are less inclined to adopt new fiscal commitments—even if they are heavily subsidized by the federal government (Welch and Thompson 1980; Davies and Derthick 1997: 229–31; Rigby and Haselswerdt 2013). In the health policy arena, a notable body of research similarly reports that prosperous states are more likely to innovate, adopt Medicaid programs, and implement managed care models (Grogan 1993; Satterthwaite 2002; Miller 2005).
The economic circumstances within a state might also condition ACA enrollment across states. Specifically, a state's affluence as well as its unemployment rate might influence the generosity and scope of state policy makers, and thereby factor into their willingness to adopt Medicaid reform and back the operations of insurance exchanges.
The implication, then, is that weaker state economies—as evident in higher unemployment rates or lower personal income—will tilt against ACA implementation because policy makers and other elites conclude that they are unable to afford even modest investments, while better-off states will be more confident and able to invest in adopting health reform. This drag on coverage may be particularly apparent in the insurance exchanges; even with government subsidies, the costs on economically strapped individuals to enroll may prove too high.
There is, however, some uncertainty about the application of past research to the ACA. The reformers who designed the ACA and worked to implement it challenged the finding of prior research that reported a positive relationship between economic well-being and the adoption of new policy. They created unusually generous financial terms precisely to entice states that were struggling to generate employment and lift personal income to expand Medicaid (Jacobs and Skocpol 2016).
What explains the number of individuals in each state that decided to enroll in the new insurance exchanges and expanded Medicaid program despite intimidating obstacles? We investigate four sets of hypotheses to explain variation in enrollment across states.
Individuals in states with Democratic governors will enroll in Medicaid and the exchange at higher rates than those in states with Republican governors.
Individuals in states with more extensive Democratic control of the state legislature will enroll in Medicaid and the exchange at higher rates than those in states with more Republican control.
Individuals in states with higher levels of public support for President Obama will enroll in Medicaid and the exchange at higher rates than those in states with lower levels of support for Obama.
Individuals in states with histories of strong administrative capacity will enroll in Medicaid and the exchange at higher levels than those in states with weaker administrative capacity.
States with greater control over their insurance exchanges will have higher levels of enrollment in the exchange.
States with greater progress implementing Medicaid reform as stipulated by the ACA will have higher enrollment in Medicaid.
State Economic Circumstances
Individuals in states with higher personal income will enroll in Medicaid and the exchange at higher rates than those in less affluent states.
Individuals in states with higher levels of unemployment will be less likely to enroll in insurance exchanges but more likely to enroll in Medicaid.
Studying ACA Enrollment
The absolute number of Americans who enrolled in the ACA conveys the sheer magnitude of participation across the country—even in conservative states such as Texas where the third largest sign-up in the insurance exchange was registered.
The use of absolute numbers, however, conveys an incomplete picture and omits the context of participation. Texas had a large number of individuals enroll in the exchange, but it was also was home to one of the largest populations of uninsured people (5.8 million in 2013). We need measures of state progress in implementing health reform that can control for the context of insurance availability. This will allow us to assess each state's relative exertion and opportunity to sign up the uninsured.
Explaining Relative Enrollment
Variations across states in ACA enrollment are revealed by studying the decisions of individuals to sign up for coverage. We track state variation using relative measures of enrollment in the ACA's two primary programs to expand coverage: the insurance exchange and the expansion of Medicaid. We created a relative measure of exchange enrollment based on the number of individuals who signed up with a qualified health plan (QHP) in each state through the end of the 2015 open enrollment period as a proportion of the total number of uninsured in each state in 2013.2 Similarly, we calculated the change in Medicaid enrollment from the pre-ACA monthly average Medicaid and CHIP enrollment for July–September 2013 and the end of the 2015 open enrollment period, and then divided this amount by the total number of uninsured in each state. Our measures reflect developments as of February 22, 2015—the end of the 2015 open enrollment period.3 (See appendix for detailed information on the construction of our enrollment measures as well as our independent variables.)
Our measures of relative enrollment, which are presented in figures 1 and 2, show that Texas and other Republican-controlled states which boast large absolute numbers of exchange enrollment trail Vermont, Massachusetts, and other states in covering their population of uninsured. For the exchange, well over half the uninsured in Vermont and Massachusetts enrolled while less than a quarter signed up in Oklahoma, Texas, and Alaska. In terms of increased Medicaid coverage since the ACA's passage, three-quarters of the uninsured enrolled in Massachusetts, Kentucky, and Oregon while only a half or less enrolled in Texas, Oklahoma, and Alaska.4 These are startling differences in the relative progress of states in reaching their uninsured that require rigorous and comprehensive explanation.
Our measures of relative enrollment point to the influence of partisanship. The highest scores in figure 1 are among states with strong Democratic Party control (Vermont and Massachusetts), and the lowest are among states with unified Republican Party control (Nevada). Similarly, in figure 2, the highest scores are in Democratic strongholds such as Oregon and Massachusetts, and the lowest scores are found in Republican-controlled states such as Nebraska and Utah.
There are also signs, however, that partisan control of state government may not fully account for the wide variation in relative enrollment across states. Republican states such as Florida and Wisconsin score quite well on the exchange measure, for example, and a few states with Democratic control—such as Hawaii—do comparatively less well, finishing outside the top 15 on both measures.
What explains the variation in individual decisions across states as reflected in relative enrollment? We examine four explanations, creating distinct independent variables for each.
The first explanation is focused on partisanship and is measured with two key variables. One is a dummy variable for whether or not the state has a Democratic governor; the other is a three-point variable capturing full, partial, or no Democratic control of the state legislature. Our distinct measures for each lawmaking branch allow us to account for differential institutional interests regarding health reform. Governors—including some Republicans—may have focused on the state's broad interests in securing the generous federal government financial support to reduce budgets and in supporting medical providers who pressed for widening coverage. Alternatively, legislators focused on narrower regional and political interests, with Republicans opposing reform in response to intense pressure by conservative constituents (including well-organized and vocal Tea Party groups from their districts). Prior research has also used split-party measures to track the differential political incentives of state lawmakers (Ferguson 2003; Miller and Blanding 2012; Rigby and Haselswerdt 2013).
The second explanation is presidential cueing and measures the support for Obama through his vote share in each state during the 2012 presidential election. Presidential vote share has been used previously to account for public support of presidents and is a useful measure here given the close ties between President Obama and the ACA (Cover 1986; Abramowitz 2008; Shelly 2012; Barrilleaux and Rainey 2014).
The next explanation is institutional and focuses on two potential effects. Administrative capacity differentially equips states to implement health reform and help individuals enroll but, unfortunately, past research has not produced an appropriate measure of administrative capacity for us to borrow. In this context, we rely on a rough gauge of state capacity to handle insurance oversight, which is a component of exchange operation, and to aid the poor and vulnerable developed in prior work (see Jacobs and Callaghan 2013; Callaghan and Jacobs 2016). The variation in our measure has a strong degree of face validity as it corresponds with expected differences across states—from the weak resources of Alabama to the greater capabilities in Maryland, Massachusetts, and Connecticut (Gold, Sparer, and Chu 1996).
Another potential institutional effect relates to policy feedbacks and, specifically, the impact of states adopting insurance exchanges and the expanded Medicaid program on enrollment. Our measure of insurance exchanges used a three-point scale to score greater state involvement: No points were assigned where states played little or no significant role and defaulted to the federal exchange; one point was granted to states that formed a “partnership” with the federal exchange to share responsibility; and two points were awarded for states that established their own exchange.5 In addition, we designed a six-point scale to measure greater state progress toward implementing steps to adopt Medicaid expansion: zero points were awarded for no progress; one point was granted to states that obtained planning grants; two were assigned when the governor made a public announcement of support; three points were given for the formal submission of a waiver proposing a Medicaid reform alternative; four points were awarded if the state implemented the alternative and it was considered compliant with federal law; and five points were awarded to states that implemented the program stipulated by the ACA. Higher scores on our measures for exchange and Medicaid reform indicated that states supplied more resources to individuals who might consider ACA enrollment.
The fourth explanation focuses on economic circumstances and is measured in two ways. The first is with state affluence which is measured by real, price-adjusted estimates of per capita personal income for states. These data were obtained from the Bureau of Economic Analysis and are updated to the most recent year available, 2014.6 The second measure is the unemployment rate in each state. This measure is a twelve-month average of the unemployment rate in each state in 2014 and is taken from the National Conference of State Legislatures.
We also tested for a fifth explanation related to groups external to the state that aggressively targeted ACA enrollment through a dummy variable for whether or not a state was an Enroll America state. This variable was excluded from final analysis because it failed to exert a statistically significant effect on enrollment.7
Method of Analysis
We study the impact on relative exchange enrollment and relative Medicaid enrollment by the sets of variables we outlined: party control of state government, Obama vote share, the institutional effects of administrative capacity and establishing new ACA programs (insurance exchanges and new Medicaid benefits), economic affluence and unemployment. Specifically, party control, Obama vote share, administrative capacity, state affluence, and unemployment are theorized to exert direct effects on enrollment as well as to work through the intervening variables of established exchanges and expanded Medicaid programs.
Our study relies on path analysis with standardized coefficients to investigate the effect of our explanatory variables on exchange and Medicaid enrollment. We use path analysis and structural equations to estimate the links (or paths) among variables in order to examine the underlying causal connection between our measures (Asher 1983: 30; Bohrnstedt and Knoke 1994). This approach is particularly helpful because we are interested in studying the paths that run from our four sets of independent variables to our dependent variables (exchange and Medicaid enrollment), either directly or indirectly through the intervening (or endogenous) variables—state decisions regarding the type of exchange to establish and whether to adopt Medicaid expansion (Asher 1983; Bohrnstedt and Knoke 1994; Cook, Jacobs, and Kim 2010).
In short, relative exchange enrollment is modeled as a function of exogenous variables (party control, Obama vote share, administrative capacity, state affluence, and unemployment), and an endogenous variable (exchange type). Similarly, relative Medicaid enrollment is modeled as a function of the same exogenous variables but replaces exchange type with Medicaid status as the endogenous variable.
Understanding State Variation
Our analyses partly confirm earlier findings that pointed to the impact of partisanship. What is striking, however, is that partisanship is neither a consistent nor a dominant influence on exchange and Medicaid enrollment across states. Other political and economic factors are crucial to understanding ACA implementation.
Only two factors account for the variations in exchange enrollment across states, as shown in figure 3. First, unemployment exerts a negative and significant influence on relative exchange enrollment. Figure 3 shows that a one-unit increase in unemployment decreases exchange enrollment by .30 standard deviations. While this finding is consistent with our hypothesis and prior research, we found little support for our other measure of economic circumstances—affluence. Affluence appears to be unrelated with either exchange type or enrollment.
The second and strongest influence on state exchange enrollment is presidential cueing. As anticipated by prior research, Obama vote share has a positive and significant effect. Specifically, a 1 percent increase in Obama vote share increases relative exchange enrollment by .43 standard deviations (see fig. 3). This effect remains after controlling for the partisanship of the governor and legislature, neither of which reach statistical significance. This suggests that presidential cueing has an independent effect beyond the partisan control of the lawmaking branches.
Finally, our institutional variables exert little impact on exchange enrollment. Administrative capacity is a positive and significant predictor of exchange type, but both exchange type and administrative capacity are insignificant predictors of exchange enrollment.
What happens when we focus on Medicaid enrollment in figure 4? Medicaid enrollment is driven by a different dynamic. Democratic legislatures exert a positive and significant influence. As expected, Democratic bodies expand enrollment but only when analyzing Medicaid enrollment.
Second, in exploring economic circumstances, we once again find that unemployment is a significant predictor and that state affluence is an insignificant predictor of relative enrollment. Critically, however, the sign on unemployment flips and the measure is a positive and significant predictor of Medicaid enrollment as was expected in our hypotheses. Where unemployment deterred Americans from paying for insurance on exchanges, it encouraged them to seek Medicaid coverage, which does not charge a premium.
Finally, we find considerable support for the institutional account when studying relative Medicaid enrollment. State decisions to adopt the Medicaid expansion are a positive and significant predictor of relative Medicaid enrollment: states that take a one-point increase (on a six-point scale) toward further implementing Medicaid produce a .48 standard deviation increase in Medicaid enrollment. This finding should be intuitive: states that implement Medicaid expansion are able to enroll more people. More intriguing, however, is the positive and significant effect for administrative capacity. While administrative capacity plays no role in determining exchange enrollment, states with stronger administrative capacity do have higher levels of Medicaid enrollment, perhaps reflecting the fact that they are better equipped to use their capacity to locate and enroll newly eligible individuals.
Partisanship, administrative capacity, and unemployment swamp presidential cueing. Obama vote share fails to exert a significant effect.
The New Politics of Health Reform
For six years, the ACA's journey through Congress to President Obama's desk and then the pitted political and judicial fight over implementing the law has appropriately dominated the study of health politics. A new politics of health reform is, however, dawning that pits the normalization of the ACA against Republican efforts to change the law. On the one hand, the ACA has started its normalization into American health care—it has provided health coverage to millions of Americans and millions more have been impacted by its insurance regulations, prescription drug coverage, and other provisions (Congressional Budget Office 2016). On the other hand, this growing normalization will be interrupted by the efforts by the Trump administration and Republican majorities in Congress to replace the ACA.
Our research was conducted before the Republican replacement was pursued and its parameters introduced. Our research does point to three themes that may color the debates over replacing the ACA and the new directions of health reform. First, the ACA's new benefits have created new expectations among Americans and new business for stakeholders, which will create pressure for risk aversion among politicians. Enrollment in insurance exchanges and Medicaid has, as we demonstrated, increased throughout the country—even in regions where opposition was strongest. There is no case in US social policy of entirely terminating a comparatively inclusive program impacting massive numbers of Americans.8 The political backlash for doing so would be intense.
Indeed, there are telltale signs of the active or tacit acceptance of parts of the ACA. Political attacks and much of the critical press coverage focus on a relatively small part of the ACA's policy interventions—the 9 percent of the country in the individual insurance market (as of 2007) who were the original targets of the insurance exchanges (US Census Bureau 2008). Few politicians and press reports are making it a priority to take away the new prescription drug coverage for seniors, the new protections for all Americans against insurance companies excluding the ill or charging them exorbitant rates, the new safeguard against elder abuse, and other provisions. Another tacit sign of the law's normalization: the arrival of the familiar dynamics of interest group politics (Lowi 1979). Pressure groups representing insurers, medical device manufacturers, and others have shifted their fire from seeking repeal to seeking a better deal (Ario and Jacobs 2012). The eventual Republican plan to replace the ACA will be a function, in part, of the new political reality created after 2010 and the implementation of health reform.
The second feature of the new politics of health reform is the emerging effects of new policies across states (Mettler and Soss 2004; Jacobs and Mettler 2011). Our evidence is clearest with regard to Medicaid enrollment where stronger administrative capacity and progress toward adopting Medicaid expansion contribute to higher enrollment. This particular dynamic is not, so far, evident with regard to insurance exchanges; the decision of most states to default to Washington may account for this difference. Our quantitative research does indicate, however, that Republican control of state government does not deter exchange enrollment, despite media speculation that it does. The broader point is that the ACA's concrete programs gained traction to the point that partisanship did not—prior to 2017—suffocate state implementation and enrollment.
The dynamics of policy evolution before the Republican replacement of the ACA are likely to be imprinted on their plans. For instance, Republican decisions to grant states greater discretion and control over budget and benefit structure will produce national variations. Based on our findings, we expect administrative capacity and relative progress implementing the ACA may continue to characterize state decisions about whether and how to use their new flexibility.
Finally, state innovations—from liberal approaches in Massachusetts and elsewhere to the “private option” in Arkansas and the “consumer-driven” model in Indiana—are likely to contribute to state-to-state learning and diffusion after 2017. Even with the repeal and replacement of the ACA, the performance and budgetary gains of states implementing the ACA introduced a new chapter to America's history of state experimentation and diffusion of effective programs across the US, or what is known as “horiztonal diffusion” (Volden 2006; Gilardi, Füglister, and Luyet 2009; Shipan and Volden 2012). In particular, the ACA's early outcomes suggest that developing administrative capacity and choosing to implement the Medicaid expansion facilititate wider coverage of the uninsured. Research on “horizontal diffusion” suggests that states with relatively low enrollment of the uninsured may look to the successes of other states to try to secure the budgetary benefits and, possibly, the political rewards of improved coverage. In short, increased state discretion in health care under President Trump could accelarate this learning and diffusion, with states building on what has worked under the ACA to effectively and affordably cover their residents.
Much remains for future research. Whether and to what extent the ACA's particular provisions lock in or are watered down or terminated are uncertain on the eve of Donald Trump's inauguration. While the ACA will experience significant change, policy continuity is also likely. Attention to policy feedback—how the ACA's implementation impacted the expectations and preferences of Americans and stakeholders—offers an analytic framework for studying the mix of change and continuity (Jacobs and Mettler 2011).
Major social policy reform in the United States—whether Social Security in 1935 or Medicare and Medicaid in 1965—often triggers a cycle of intense struggles to establish new programs. Take Social Security: its formulation and enactment enjoyed some bipartisan support, but its passage ignited decades of pitched ideological battles that included efforts to repeal it outright or to undercut it by converting it into a voluntary program (as Ronald Reagan proposed for decades). What is significant is that the contention at the birth of Social Security and Medicare was followed by a normalization of politics: partisan acceptance, quiet interest group negotiation, the diffusion of successful innovations, and the formation of supportive voting blocs and organizations (Jacobs 1993; Brown and Jacobs 2008).
Direct comparisons with the ACA will be incomplete and inaccurate after the Republicans repeal and replace the ACA. But these comparisons were always imperfect because the ACA's benefits are more diffuse and not limited to a single program.
Nonetheless, it appears that aspects of normalization have begun in the form of interest group acceptance, state borrowing of innovations, and the activation of advocates for specific programs. The pressure of nomalization is evident in the Republican rejection of the outright termination favored by its most conservative supporters and its pursuit of a replacement plan that acknowledges the new reality of public and stakeholder expectations. Perhaps the most salient lesson of American political development and the history of social policy reform is that the new path of health reform inaugruated in 2010 and recast in 2017 set the conditions for future debate. Over time, the 2017 replacement may well stand out for significant reversals of the ACA as well as the Republican Party's acceptance of health reforms that it had firmly rejected. Future debates over health reform will respond to the new policies and incorporate Republican ideological concessions to government's expanded role in health policy.
1. There is also growing evidence of horizontal diffusion in the implementation of the ACA. While not a focus of our analysis here, we conclude by speculating about horizontal diffusion as part of the new politics of health reform.
2. Our measure of the number of uninsured is incomplete but provides the most current available data. The Small Area Health Insurance Estimates (SAHIE) offers more accurate measures of the number of uninsured, but it will not be available for analysis until later in 2016.
3. The ACA explicitly bars the undocumented from enrolling in the insurance exchanges. Our study makes no effort to confirm this legal requirement. In terms of our measures of Medicaid enrollment and the number of uninsured, there are no reliable and readily available data that differentiate the undocumented from citizens.
4. It is important to note that our estimates are rough gauges for relative enrollment. Although Massachusetts appears to cover 100 percent of its population in figure 2, this does not presume that there are no uninsured in the state. Instead, it reflects the timing and imperfections of data collection.
5. Nevada, New Mexico, and Oregon received scores of “1”: they approved state exchanges and assumed significant responsibilities while also deferring to the federal exchange's website to handle much of the process of eligibility determination and enrollment.
6. We also tested state revenue per capita in place of economic affluence but ultimately excluded it because it performed similarly. Revenue per capita never reaches statistical significance in the exchange path analysis and in the Medicaid path analysis; it is an insignificant predictor of Medicaid enrollment and has an insignificant total effect, although it does have a marginally significant influence on Medicaid status.
7. Tests determined that the Enroll America variable had no statistically significant direct or total effect on relative exchange enrollment or relative Medicaid enrollment. It was, however, a negative and significant predictor of exchange type.
8. Pensions for Civil War veterans and their dependents reached millions of Americans but its scope was limited to those with a connection to the Civil War. The program died off with them (Skocpol 1992).
Relative Exchange Enrollment DV
Our exchange enrollment measure is captured using the equation (QHP/Uninsured pre-ACA). The numerator of the equation reflects the number of individuals determined eligible to enroll in a plan through the exchange marketplace who have selected a plan from 11/15/14 to 2/22/15. The number uninsured in each state is based on American Community Survey Estimates for 2013 and population data originally from the census bureau's March 2014 Current Population Survey as reported by Kaiser.
Relative Medicaid Enrollment DV
Our Medicaid enrollment measure is captured using the equation (Medicaid Change/Uninsured pre-ACA). Our Medicaid data is state reported and recorded by Kaiser and represents the difference between total enrollment in March 2015 and Pre-ACA monthly average Medicaid and CHIP enrollment for July-September 2013. Please note that not all changes in enrollment in Medicaid may be related to the ACA. The number uninsured in each state is based on American Community Survey Estimates for 2013 and population data originally from the census bureau's March 2014 Current Population Survey as reported by Kaiser.
This variable is a dummy variable which is scored as a “1” if the governor is a Democrat and “0” if the governor is not a Democrat. It is updated to 2015.
1 point = full Democratic control of both houses of the state legislature.
.5 points = Democratic control of only one house of the state legislature . . . includes splits.
0 points = full Republican control of the state legislature
Results updated to 2015.
Obama Vote 2012
This variable is taken from data made available by the Federal Election Commission and reflects the percentage of the popular vote won by President Obama in the 2012 presidential election against Mitt Romney.
State affluence is a measure of state per capita personal income dollars for the most recent year available, 2014. This information is available through the Beaureau of Economic Analysis.
State unemployment is a measure of the average unemployment in each state that was created by averaging state unemployment in each month of 2014. This data is available through the National Conference of State Legislatures.
This variable scores states along 3 points with higher scores reflecting increased state control over the exchange. Zero points were awarded for a federal exchange, 1 point was awarded for a partnership exchange, and 2 points were awarded for a state exchange. Nevada, New Mexico, and Oregon, which operate federally supported state-based marketplaces, are coded as 1’s.
Stage 1: Level 1 Grant Application
1 = States who applied for a level one grant from the federal government to implement ACA reform
0 = No level one grant application by state
Stage 2: Public Statement in Support of Reform from Governor
1 = Clear statement from governor in support of reform
0 = No indication of gubernatorial support for ACA related Medicaid reform including outright opposition to reform
Stage 3: Medicaid Alternative Approval Sought/Non-ACA Alternative Implemented
1 = State pursuit of federal approval for an ACA alternative or outright implementation of Medicaid expansion not categorized as ACA compliant
0 = No alternative sought or implemented
Stage 4: Federally Approved ACA Medicaid Alternative Implemented
1 = State implementation of a federally approved ACA alternative
0 = No alternative implemented
Stage 5: Implementation of Federally Stipulated Medicaid Reform
1 = State implementation of Medicaid reform as stipulated under federal law
0 = No state implementation of ACA compliant Medicaid reform
This variable is scored 0–5 with states awarded points based on the highest stage they have achieved towards reform. The stages do not represent a temporal sequence as states may vary in the order of action; instead, they measure state progress in completing increasingly important steps. Thus a state with no level-1 grant but which has implemented Medicaid reform would get a score of 5 despite skipping stage 1. A state that has completed none of the stages is coded as a 0.
Administrative capacity is a cumulative measure based on the following: 1 point in the small group market if states guarantee and issue all products, have special rules for groups of one, have state-imposed limits on rating, have health insurance subsidies and up to 1 point for state authority to review rates. The same number of points was available for the same subjects in the individual market and 1 point was also added for guarantee issuing some products. One point was also awarded for the presence of a high-risk pool program, the high-risk pool being open to Health Insurance Portability and Accountability Act–eligible individuals, the high-risk pool being open to medically needy individuals, and the high-risk pool being open to health coverage tax credit–eligible individuals. One point was also awarded for state-mandated coverage of infertility treatment and eating-disorder coverage in the individual and small group markets. Three points were awarded for a state having an all-payer claims database, up to 1 point was given (based on Kaiser-established categories) for federal hazard preparedness funding, and 1 point each were given for state conversion coverage in small firms, state restrictions against balance billing in preferred provider organizations, and out-of-network providers in health maintenance organizations.