Context: This article argues that the devolution of the Affordable Care Act (ACA) to the states contributed to the slow progression of national public support for health care reform.
Methods: Using small-area estimation techniques, the authors measured quarterly state ACA attitudes on five topics from 2009 to the start of the 2016 presidential election.
Findings: Public support for the ACA increased after gubernatorial announcement of state-based exchanges. However, the adoption of federal or partnership marketplaces had virtually no effect on public opinion of the ACA and, in some cases, even decreased positive perceptions.
Conclusions: The authors' analyses point to the complexities in mass preferences toward the ACA and policy feedback more generally. The slow movement of national ACA support was due partly to state-level variations in policy making. The findings suggest that, as time progresses, attitudes in Republican-leaning states with state-based marketplaces will become more positive toward the ACA, presumably as residents begin to experience the positive effects of the law. More broadly, this work highlights the importance of looking at state-level variations in opinions and policies.
While the Affordable Care Act (ACA) is widely seen as a victory for President Barack Obama and the Democrats, the election of President Donald Trump coupled with a Republican Congress put the political sustainability of the law at risk. ACA proponents hoped that constituents would rally around the new legislation after receiving personal benefits from the law (Skocpol 2010). After all, the Massachusetts health reform plan implemented in 2006, which served as a model for the ACA, enjoyed high levels of public support (Blue Cross Blue Shield of Massachusetts Foundation 2014).
While mass constituencies did mobilize when the ACA was in mortal danger of repeal, public support for the ACA has not increased as much or as quickly as some observers predicted. One reason for this mixed support for the ACA is that many Americans were satisfied with their personal health coverage prior to the ACA (Jacobs and Shapiro 1994), and public perceptions were relatively unchanged after enactment, as well as after the Supreme Court ruling in 2012 (Blendon, Benson, and Brulé 2012). While several elements of the ACA receive strong support, many Americans express their strong disapproval of the law in its entirety (Brodie, Deane, and Cho 2011). Only recently has the ACA drawn approval from a majority of Americans. A 2017 Pew poll found that, for the first time, a majority of American adults (54%) approve of the ACA, although 43% still want changes to the law (Fingerhut 2017).
Why have Americans not been quicker to support the ACA? One account suggests that individuals base their opinions on collective evaluations of how health reform influences the country instead of personal evaluations of how the law influences their family (Lynch and Gollust 2010). This is consistent with research finding that self-interest is unrelated to political preferences (Sears and Funk 1990); thus, efforts to build support by front-loading the bill with personal benefits failed because opinions are unaffected by personal gains. Another answer lies in the highly partisan nature of health care reform. The majority of Obama voters favored implementation or expansion of the ACA (78%) and supported a more activist federal government in US health care (92%) (Blendon, Benson, and Brulé 2012). According to this view, there is little room for personal benefits to influence opinion, as partisanship trumps all else. Finally, it may be that individuals have not cognitively linked the ACA to their personal situation, either because not enough time has passed for citizens to understand how the new law influences their family (Jacobs and Skocpol 2012) or because many individuals are uninformed about the daily impact of policies (Mettler 2011).
We consider another reason for the slow movement of public support: devolution of policy making to the states. The decision to devolve much of the ACA policy making to the states was strategic; without it the passage of health care reform may have been doomed, yet, decentralizing the ACA also led to significant state-led resistance (Jones 2017). How has state-level resistance influenced trends in public support for the ACA? To answer this question, we needed to extend previous studies in two ways. First, we considered dynamic analyses by moving beyond cross-sectional analyses at a single point in time. This is crucial, because panel studies allow scholars to understand how policy implementation has changed opinion. Jacobs and Mettler (2016) were the first to move in this direction by employing panel analyses from 2010 to 2014 on a nationally representative sample (see also Jacobs and Mettler 2018). They found that, while there were supportive shifts in ACA opinion, party politics and distrust for government largely explain the slow progression of ACA support at the individual level.
Employing analyses over time is beneficial but not enough. Understanding how devolving policy making to the states also rests on considering the role of state policy design in explaining shifts in ACA attitudes. There is a growing recognition that policy decisions and, ultimately, implementation have the potential to shape attitudes and political behaviors of target populations, such as beneficiaries, as well as other members of the public (e.g., Jacobs and Mettler 2018; Maltby 2017). Even with this recognition, however, scholars almost exclusively focus on national policy sentiment toward health care reform and essentially ignore variations in opinion shifts across the states (for an exception, see Michener 2018). Only by shifting the analyses to the states are we better able to understand how variations in policy designs influenced public support for health care reform, which in turn informs us about the success of subsequent policy decisions and health policy outcomes.
Using small-area estimation techniques, we generated quarterly state opinion toward the ACA from the fourth quarter of 2009 to the second quarter of 2016 on five topics. We measure general favorability toward the ACA, opinions about the law's impact on the country and one's family, understanding of the ACA's impact, and support for keeping or expanding the law.1 After presenting the measures, we report our statistical results. We focused on the role that state policy decisions, including the type of health insurance marketplace and the timing of gubernatorial announcements about the structure of the state's marketplace, played in influencing state opinion trends.
We found variation in opinions toward the ACA across time and states. Moreover, the models suggest that the variation in opinion trends across the states is partly explained by state-level policy making. Our results show that public support for the ACA increased after gubernatorial announcement of state-based exchanges: general public support for the ACA increased by 3%, and positive perceptions of the ACA's impact on the country and the family increased by almost 2%. In contrast, the adoption of federal or partnership marketplaces had virtually no effect on the public's perceptions of the ACA and, in some cases, even decreased positive perceptions.
Our results have two implications for our understanding about the ACA and health care reform. First, they suggest that, as time progresses after announcement of state-based exchanges in Republican-leaning states, attitudes may become more favorable toward the ACA, presumably as residents begin to experience the positive effects of the law. This may help quell fervent Republican opposition to the ACA at the national level. Second, our results suggest that Obama's decision to give states significant leeway over the type of marketplace exchange undermined ACA favorability among the public and, thus, is partly responsible for the continued threat of repeal at the national level.
Policy Feedback and the ACA
While we were interested in describing state trends in ACA opinion, answering our question—of whether Obama's decision to give states flexibility in ACA policy making resulted in stagnant national support of health care reform—rests solely on the idea that state-level policies influence ACA attitudes. The way in which policies influence attitudes is called attitudinal policy feedback. Policy characteristics are important for determining the mechanism through which policies influence opinions (Pacheco and Maltby 2017). Policies that are proximate and visible are most likely to shape public opinion (Soss and Schram 2007). Proximate policies directly influence individuals, while more distant policies are largely hidden. Visible policies receive large amounts of media and electoral attention. Proximate policies influence mass opinions through implementation (Soss 1999), while visible policies influence preferences through the information environment (Brewer 2003). Of course, it is possible for both implementation and the information environment to influence policy feedback simultaneously; these processes are not mutually exclusive, and we suspect that both mechanisms account for the effects of the ACA on state opinion.
Classic studies on attitudinal policy feedback concentrate on redistributive policies, primarily at the national level.2 Mass feedback effects from these types of policies tend to be short-lived or nonexistent since most people do not come into direct contact with the policy (Soss and Schram 2007). The exception here is the vast work on the “thermostatic model,” which found that, as policy output increases, the public's preferences for more policy output decreases; as policy output decreases, the public's preferences for more policy output increases (e.g., Erikson, MacKuen, and Stimson 2002; Wlezien 1995). The result is a reactive system of governance where public opinion and policy constantly adjust and readjust to each other over time. For example, Erikson, MacKuen, and Stimson (2002) found that the public becomes more liberal in response to conservative policies (and vice versa). There is evidence that the thermostatic model of responsiveness occurs in the states (Pacheco 2013b).
The few studies of attitudinal policy feedback on non-redistributive policies suggest that policies can influence stereotypes, perceptions of what is harmful, and support for analogous policies in the future (Pacheco 2013a). The latter is most relevant for our article since we argue that the decision to decentralize ACA decisions muted national public support for the ACA, thus keeping the prospect of repeal alive. The broader implication beyond health policy is that policy makers can design policies that move toward eradicating social inequalities in ways that are beyond their immediate, quantifiable effects.
Regarding health care reform, the ACA presents a unique opportunity to reexamine and expand on prior work on attitudinal policy feedback, since the ACA is proximate and visible to the American public. Jacobs and Mettler (2011) even considered the question of the long-term attitudinal effects of the ACA to be the lingering puzzle in health care reform. Yet, most research on state-level ACA policy feedback has concentrated on participatory effects as opposed to attitudinal effects. For instance, Haselswerdt (2017) found a “backlash” effect for states that decided to expand Medicaid under the ACA; conservatives had higher rates of political participation in US House races in 2014 compared to 2012, presumably as a reaction to health care reform. In addition, Clinton and Sances (2018) found that Medicaid expansion had a short-term positive effect on registration and participation.
Two highly relevant reports about the ACA and public opinion are Morgan and Kang (2015) and Brodie, Deane, and Cho (2011). Focusing on the national level, Morgan and Kang (2015) found that support for spending on health care decreased after the adoption of the ACA, which is consistent with theories of thermostatic responsiveness (e.g., Wlezien 1995). Unlike Morgan and Kang (2015), we focused on subnational ACA opinion instead of national support for health care spending. Because we are looking at opinion toward a policy as opposed to spending, we anticipate positive policy effects much more in line with what Pacheco (2013b) found regarding antismoking legislation and public opinion. This is predicated on the idea that individuals living in states that could immediately benefit from the ACA (e.g., those that chose a state-based exchange) are likely to increase support for the ACA.
At the regional level, Brodie, Deane, and Cho (2011: 1102) found that political characteristics of a region affect attitudes toward the ACA, arguing that national measures of public opinion “mask important regional (and state-level) distinctions in views.” While Brodie and colleagues focused solely on 2011, our research is dynamic, thus allowing us to understand how shifts in public opinion respond to state-level policy decisions. Overall, then, we expand on this work by looking at how state variations in the type and timing of ACA decisions influence a broad range of state ACA attitudes over time. In so doing, we not only expand on the policy feedback literature but also explore the consequences of policy devolution for national public opinion on health care reform.
Data and Methods
To measure state-level opinions toward the ACA, we gathered monthly data from national surveys by the Kaiser Family Foundation, Gallup, Pew, and CBS/New York Times. We selected these surveys for several reasons. First, the questions have similar wording across surveys; this increases our confidence that changes in opinion are not due to shifts in question wording. Second, by combining questions across surveys, we increased the amount of information and the reliability of our estimates, particularly for the less populated states. Pooling surveys also increased the information that we had over time. In the end, we collected data on anywhere from 40,000 to 117,000 respondents from 2009 to 2016, depending on the question. This tracks opinion a few months before the ACA became law through the beginning of the 2016 presidential election (see table 1 for specific questions and data sources).
We compiled survey data on ACA attitudes across five different topics to get a fuller picture about how opinions vary across the states and respond to policy decisions. However, we were agnostic regarding the theoretical expectations about how state-level policies influence each of these outcomes. First, we compiled data on overall ACA favorability from the question, “As you may know, a health reform bill was signed into law in 2010. Given what you know about the health reform law, do you have a generally favorable or generally unfavorable opinion of it?” Although the exact text varies somewhat from one survey to the next, this phrasing from Kaiser Family Foundation polls represents the most common wording of questions. Higher values represented more favorable opinions. This question is the most frequently asked and analyzed by scholars, yet it is also limited in a number of ways. Its general nature does not assess what specifics about the law a person favors over others, and it is difficult to understand a respondent's rationale for support. For instance, one cannot understand if the rationale is based on self-interest or sociotropic concerns.
The general ACA question also does not assess how respondents view the potential impacts of the ACA. In this study, we assessed impact of the ACA on the country by compiling answers to two questions about how the ACA has currently or will eventually influence the country with higher values indicating a more positive outlook. To examine whether individuals believed the ACA was personally beneficial, we compiled responses to a question on how the ACA will affect one's family in the future, with higher values indicating a more positive influence.
The fourth question tapped into respondent understanding of the ACA, by asking whether respondents understood the impact of the ACA on their family; higher values indicate understanding. Finally, we assessed opinions about future policy changes. While exact wording varies slightly across surveys, the question we used to measure attitudes about the ACA's future generally matched this phrasing from the April 2011 Kaiser Family Foundation poll: “What would you like to see Congress do when it comes to the health care law? (1) They should expand the law. (2) They should keep the law as is. (3) They should repeal the law and replace it with a Republican-sponsored alternative. (4) They should repeal the law and not replace it.” Responses to this question indicated the percentage of individuals who want to expand or keep the law and allowed us to assess if the strategy to front-load the bill with personal benefits paid off in increased support for additional health care reform and government intervention.
Multilevel Regression and Poststratification
We used an increasingly popular small-area estimation technique called multilevel regression and poststratification (MRP) to estimate state opinions toward the ACA (Gelman and Little 1997; Park, Gelman, and Bafumi 2004, 2006). The MRP approach uses national surveys to produce accurate estimates of public opinion at lower levels of aggregation, such as by state (Lax and Phillips 2009) or congressional district (Warshaw and Rodden 2012). The use of multilevel modeling and poststratification overcomes two major problems that arise when trying to measure subnational opinion from national surveys: decreased reliability in the less populated states, and nonrepresentativeness of state samples. Multilevel modeling increases the reliability of less populous units via “shrinkage toward the mean.” Poststratification corrects for nonrepresentativeness due to sampling designs by adjusting estimates using census information to be more representative of subunit populations.
To estimate state opinions, we modeled survey responses using standard predictors of MRP: gender, race, age, education, region, state, and state presidential vote share (Lax and Phillips 2009). For the poststratification stage, we obtained population frequencies from the Census Bureau's public-use microdata samples for 2010 (US Census Bureau 2010). While others have used MRP to estimate ACA favorability across the states (Barrilleaux and Rainey 2014; Rigby and Haselswerdt 2013), these measures do not vary over time; our approach requires time-varying measures to understand how state-level decisions and the timing of these decisions influence potential shifts in ACA opinions.
We followed the lead of others and added a time component by using a three-quarter moving average to estimate quarterly opinion toward the ACA (Pacheco 2011). Our approach to adding a time component is identical to that we used in Pacheco and Maltby 2017, where we showed that this approach is reliable and valid. While we mirror our 2017 approach, our data here are more extensive in three ways. First, in 2017 we measured quarterly state trends using only the Kaiser Family Foundation survey; here we have included additional surveys, thus increasing the amount of information to create the measures. Second, in 2017 we measured quarterly state trends in ACA favorability from the second quarter of 2010 to the second quarter of 2014. Here we extend this data set well into 2016. Finally, our data here include trends on four additional topics that go beyond general favorability toward the ACA. This allows us to take a more comprehensive look at opinions toward the ACA and better understand how state policies contribute to different facets of ACA opinion. We used analyses of variance (ANOVA) to understand the variation in public attitudes toward the ACA across states and time and to inform our empirical approach with the time series analyses.
This study is also theoretically distinct from the Pacheco and Maltby 2017 study. In 2017 we were interested primarily in how shifts in state opinions toward the ACA influenced the diffusion of ACA decisions; the variable of interest was the policy. Here, in contrast, we are interested in how variations in the type and timing of state ACA decisions influenced shifts in ACA opinions; our variable of interest is state ACA opinion. In this way, our research speaks most directly to the effect that Obama's policy decisions had on national public support for health care reform.
Time Series Approach
To explore the role of the type and timing of ACA exchange decisions on public opinion, we used an error correction model, which allows for the estimation of both short- and long-term effects of independent variables and tells us how quickly a system returns to equilibrium or the overall mean after being disrupted. The dependent variable captured the changes in state opinion toward the ACA and a lagged dependent variable was included to account for time dependence.
The main independent variables were time-varying measures that combine the type and timing of the marketplace decision by governors. For example, the “state announced state-based marketplace decision” variable was coded 0 in all time periods in which the governor did not make an announcement and a 1 in time periods after a governor announced a state-based market. This variable was also coded 0 in states where governors announced other types of marketplace decisions. We had three of these types of variables to correspond to each of the three marketplace types. We included both the differenced and lagged versions of these variables to capture short- and long-term impacts on public opinion. Additionally, we included a counter variable that captured the time since the state announced its decision, regardless of the type of decision. We included this to account for any learning that occurred among the public prior to shifts in opinion.
Given the nature of the data, there is complexity in estimating models that account for both unit heterogeneity and autocorrelation (see Beck and Katz 1995, 2011; Bell and Jones 2015; De Boef and Keele 2008). Proper identification of the modeling strategy requires an understanding of the source of variation in the dependent variable, as well as the limitations of the data. The source of variation for the dependent variables differs, with some (ACA favorability) having a majority of variance between states and others (ACA understanding) having a majority of variance within states. In addition, there are 50 states (n = 50) across 27 quarter-years (t = 27), suggesting that t is large enough to capture dynamic patterns and that asymptotics exist in both n and t (for more information, see Wooldridge 2010).3
One concern with time series cross-sectional data is that systemic factors or unit specific forces that are not included in the model cause omitted variable bias. Time trends can be used to account for systemic factors, while fixed unit effects account for heterogeneity. These solutions are atheoretical and may lead to an overparameterization of the model (e.g., Achen 2005). Nonetheless, we included fixed unit effects (e.g., state dummies) to account for unit heterogeneity and fixed time effects (e.g., quarter dummies) to account for national-level systemic factors. We also included panel-corrected standard errors as suggested by Beck and Katz (2011).
Trends in State ACA Opinions
Figure 1 gives a descriptive glimpse into the dynamic properties of state opinion toward the ACA from 2009 to 2016.4 As shown in figure 1, state opinion toward the ACA varies across both time and states, but the amount of variance differs across question type. While favorability toward the ACA is generally low at the national level, a closer look at the data in figure 1A shows that in some states (e.g., California) most residents favor the ACA while in others (e.g., West Virginia) support is much lower than the national average. There is also movement in ACA favorability, with some states declining in support and others experiencing bouts of increased favorability. Analysis of variance (ANOVA) showed that the majority of variance is between states (95%) rather than within states. These results are consistent with those of Pacheco and Maltby 2017.
As shown in figure 1B and 1C, the public generally has a more positive outlook about the ACA's ultimate impact on the country versus for themselves or their family. Yet, both responses appear to have a negative, downward trend over time, with occasional upticks. Many states (e.g., Illinois and Florida) experienced temporary increases in positive views between 2011 and 2012, while others (e.g., Texas and New York) experienced a general decline across the entire time span. ANOVA indicated that the majority of variance for the impact of the ACA on the country is between states (87%), while variance is evenly split across states and time for attitudes about the impact of the ACA on one's family.
Figure 1C shows significant movement over time in the public's understanding of the ACA, but little variance in trends across states. Respondents for all of the states indicated increasing understanding of the ACA until the late 2010s. At that point, the decline in understanding among respondents was minimal in some states (e.g., Florida) and substantial in others (e.g., California and West Virginia). ANOVA showed that most of the variance in opinions on understanding of the ACA occurred across time (98%), with parallel trends across the states. Finally, figure 1C shows trends in views about keeping or expanding the ACA; ANOVA showed that 76% of the variance occurred between states. The most interesting aspect of this topic is the decline in support for the ACA in late 2012 and into 2013, which occurred for nearly all states.
Measuring State ACA Decisions
We focused on the type of marketplace states adopted and the timing of gubernatorial announcements of their state's marketplace structure. Our data on the type of marketplace adoption on the month and year in which governors announced their decision came from policy briefs provided by the Kaiser Family Foundation. Regarding the structure of the marketplace, states could choose to control all aspects of the health insurance marketplace (state marketplace), share control of the marketplace with the federal government (partnership marketplace), or cede all power in marketplace creation to the federal government (federal marketplace). Most states opted for either a state-only or federal-only marketplace (16 states and 27 states, respectively); only seven states chose a partnership model. California was the first state to announce the structure of its marketplace in September 2010. All 50 states had announced their marketplace structure by May 2013.
Governors are the most powerful person in state government (Beyle 2004; Jones, Bradley, and Oberlander 2014) and instrumental in determining which marketplace their state opted for and when (Rigby and Haselswerdt 2013). Republican-led and conservative states tended to delay decisions and default to a federal marketplace, while Democrat-led and liberal states were more likely to establish their own insurance marketplace (Jones, Bradley, and Oberlander 2014; Rigby and Haselswerdt 2013). Even still, by announcing the marketplace structure for their state, governors played a critical role in legitimizing the ACA by signaling to the public that health care reform in their state will move forward, especially since the announcements received high media attention at both the national and local levels (Gollust et al. 2014).
Table 2 presents some descriptive statistics showing the bivariate relationship between support for the ACA and state marketplace type, showing average ACA attitudes across state marketplace types. States that opted for the state-based exchange had residents more supportive of the ACA compared to other states: 50% of respondents in state-based exchanges favored the ACA across the time period compared to 45% in states with partnership exchanges and 41% in states with federal exchanges. Residents in state-based exchanges also had more positive perceptions of the policy's effect on the country: 42% of respondents in states with state-based exchanges answered that the country would be better off compared to 37% in states with partnership exchanges and 34% in those with federal exchanges. Residents in states with state-based exchanges were also more likely to support expanding or keeping the ACA compared to residents in states with partnership or federal exchanges. There were fewer differences across states in understanding of the ACA over the time period.
Time Series Results
Error correction model results are shown in table 3.5 Gubernatorial announcements of the state exchange matter in the long term for overall ACA favorability and positive perceptions of the ACA toward the country and the family.6 State announcements of state-based marketplaces in the previous quarter had positive and statistically significant effects on general ACA support, beliefs that the ACA will make the country better off, and beliefs that the ACA will make one's family better off. According to the error correction model, general public support for the ACA increased by 3% gradually over subsequent quarters after gubernatorial announcement of a state-based exchange.7 Likewise, the model suggests that positive perceptions of the ACA's impact on the country and the family increased by almost 2% gradually over the subsequent quarters after gubernatorial announcement of a state-based exchange. Of course, most states that announced marketplace decisions early on selected for the state marketplace and tended to be liberal. However, the model suggests that, even in Republican-leaning states, as time progressed after state-based marketplace announcement, attitudes became more favorable toward the ACA, presumably as residents began to experience the positive effects of the law. We discuss this possibility at length in the discussion section.
We saw virtually no effect of gubernatorial announcements of federal or partnership marketplace types on ACA opinions, with two exceptions. First, we found an immediate increase in ACA favorability after announcement of the federal marketplace, yet this effect was quite small. Specifically, the model indicates that the percentage of residents who favored the ACA increased by almost 1% in the quarter immediately after gubernatorial announcement of the federal marketplace. We suspect that this may be a temporary legitimation effect as opposed to long-term influence of policy feedback, particularly since we did not see positive influences in any of the other outcomes. Second, we found a small yet negative impact of policy decision on ACA understanding, but only in the short term (significant at the 0.10 level). The model indicates that announcement of partnership and federal marketplace types decreased understanding of the ACA immediately in the next quarter. More specifically, gubernatorial announcement of the partnership marketplace decreased the percentage of residents who understood the ACA by just over 1%; we saw a similar negative impact on ACA understanding among residents in states that selected a federal marketplace. This is not completely unexpected. HealthCare.gov had some technical problems in the early stages of the ACA, which likely prevented residents from feeling like they truly understood the policy.
The error correction model shows that as the time since announcement increased, state residents were unaffected by policy decisions. Time since announcement and the time dummy variables are highly correlated (r = 0.88), yet the time since announcement variable failed to reach statistical significance when the time dummy variables were not included.
Overall, then, our results suggest that ACA public opinion is influenced by state-level decisions, but with important caveats. The decision to adopt a state-based exchange increased public support and positive perceptions of the ACA in the long term, while the adoption of federal or partnership exchanges had virtually no effect on ACA support. If anything, public understanding turned negative in response to these decisions, albeit only in the short term.
Since passage of the ACA, national sentiment has been consistently underwhelming, to the surprise of many pundits and scholars. During the time period we examined, only around 45% of the country favored the ACA. However, despite such underwhelming national support, attempts to repeal the ACA have had little success. The initial attempt by the Trump administration to replace the ACA did not garner enough support even to make it to a vote. Republicans' most recent push to replace and repeal the ACA passed in the House of Representatives despite predictions by the Congressional Budget Office (2017) that the bill will leave close to 20 million uninsured by 2020. But progress stalled in the Senate. This is likely because many members of Congress—both Democrat and Republican—are hesitant to pass a bill that takes benefits away from their constituents. Additionally, although a majority of the public has never supported the ACA, a large proportion of the country wants to either keep or expand the ACA.
How can we understand both low levels of public support and resistance by the public and elected officials (both Democrat and Republican) to repeal the ACA? We need to look beyond national public opinion, as it masks significant state-level variation in ACA attitudes. In many states, most residents favor the ACA at figures much higher than the national average. This is also true for other attitudes, such as assessments of whether the ACA is personally or collectively beneficial, whether the ACA should be expanded, and how well individuals understand the ACA.
As we show in this article, we can partly explain the slow movement of national support for the ACA by state-level variations in ACA policy making. States had many choices in how to implement health care reform, including how much control they had over their health insurance marketplaces. Attitudes toward the ACA shift in response to both the type of marketplace structure that states adopt and the timing of this decision. In general, residents showed increased support for the ACA after gubernatorial announcements of state marketplace structures, which gives state officials more control over health care reform. The effects of these policy decisions had long-term impacts on how much residents favored the ACA and perceptions of its effects on the country and them personally. We found no such effect among states that decided to enact a federal or partnership exchange.
Our estimates over time provide unique data on changing evaluations of the ACA as it was implemented at the subnational level. However, our time period does not include the Trump administration, which limits our ability to discuss how a change in the administration combines with state-level implementation to influence shifts in ACA support. While the MRP approach combined with a three-quarter moving average helped solve issues of reliability, our estimates still vary in reliability in connection with state population; the most populated states have the most reliable estimates. Lastly, because of the three-quarter moving average, we potentially may have smoothed over short-term shifts in ACA opinion that occur month by month.8
Even with these limitations, our research has two implications for our understanding of health care reform and health policy in the future. First, our results suggest that, as time progressed after announcement of state-based exchanges in Republican-leaning states, attitudes may have become more favorable toward the ACA, presumably as residents began to experience the positive effects of the law. The most prominent example of this is Kentucky. For the first three years of ACA implementation, Kentucky—a decidedly conservative, Republican state—operated its own exchange called kynect. Public opinion followed suit. A 2017 Kentucky Health Issues Poll found that 44% of Kentuckians had a favorable opinion of the ACA—up from 30% in 2013. Among Republicans in Kentucky, 31% favored the ACA; this is significantly higher than the roughly 17% of Republicans who favored the ACA nationwide, according to our dataset. Unfortunately, Kentucky has decided to switch marketplace types and go back to the federal exchange. Our results suggest, however, that Kentucky's initial decision to run a state-based exchange will have a lasting positive impact on ACA attitudes among state residents. Our results suggest that if more states opt for state-based exchanges in the future, this may help quell national Republican opposition to the ACA. Unfortunately, Jones's (2017) work suggests that it is unlikely for states to switch to a state-based exchange and that, if anything, states will likely cede more control to the federal government going forward. How these potential shifts in policy design affect national public support is an area of future research.
Second, our results suggest that Obama's decision to abandon a single-payer system and instead give states significant decision-making power undermined ACA favorability among the public and thus is partly responsible for the continued threat of repeal at the national level. A simple simulation based on our results helps underscore this point. In the second quarter of 2016 (the last time point in our data set), 45% of the public approved of the ACA. Based on our model, the national figure would increase to 48% if all of the states with a federal or partnership exchange switched to a state-based exchange. That may seem like a minor effect; however, the increased support and possible long-lasting effects on the public's positive perceptions of the ACA may have helped quiet opposition at the national level. At the very least, giving states the ability to resist the ACA continued to keep healthcare reform in the spotlight and subsequently a major issue for party platforms in the 2016 presidential election.
As the effects of Obama's health care reform evolve, we urge scholars to continue to look at state-level variations in opinions and policies. The ACA's ultimate success and favorability rely on administration and implementation at both the federal and state levels. The variations in state policy designs help scholars understand the ways that policy influences public opinion and ultimately policy entrenchment. Future research should also look into how differences in self-interest, partisanship, and policy knowledge condition the effects of policy on opinion. Finally, the broader idea that policy designs regarding centralization influence trends in public support is ripe for further exploration for health care policy as well as policy feedback in general.
This work was supported in part by the Russell Sage Foundation (94-16-05). Any opinions expressed are those of the author(s) alone and should not be construed as representing the opinions of the foundation. The authors thank participants at the Russell Sage Foundation one-day conference on the social, economic, and political effects of the ACA for their valuable feedback.
The ultimate length of time that data were available for the time series varied across topic, depending on when questions were asked.
Since there are 27 time points, we are less concerned about systematic bias that may occur when including a lagged dependent variable with unit fixed effects or other non-time-varying covariates when t is finite (e.g., Nickell 1981).
We provide additional descriptive information in the online appendix (juliannapacheco.weebly.com/published-papers.html), including a table of the ACA opinions pooled, averaged over time and shown by state. We also include a series of figures that show each time trend of opinion in each state.
Inferences are nearly identical when we omit state fixed effects. The lagged gubernatorial announcement of state-based marketplace exchange variable for ACA favorability and perceptions of the law's benefit on the country fails to reach statistical significance at the 0.10 level, but only barely, when we include linear time and time squared (ACA favorability: B = 0.86, p = 0.10; benefit on country: B = 0.80, p = 0.19). It fails to reach statistical significance for perceptions of the ACA's personal effect.
The coefficient on the lagged dependent variable gives the error correction rate with a value closer to zero, indicating a slow return to equilibrium. As shown in table 3, the coefficient on the lagged dependent variable is −0.31 for general ACA favorability, −0.37 and −0.39 for perceptions of the law's benefits on the country and family, respectively, and −0.34 for expanding or keeping the ACA. This suggests that opinions for these topics were relatively slow to return to equilibrium when disrupted. Meanwhile, the coefficient on the lagged dependent variable for understanding the ACA is −0.63, suggesting a relatively quick return to equilibrium.
To obtain the long-term effect, we multiplied the coefficient on the lagged variable by 1 (since the independent variable is dichotomous) and divided by the error correction rate (De Boef and Keele 2008).
The MRP approach coupled with temporal smoothing increases the reliability of the estimates at the expense of bias. This technique may be particularly problematic for this study given that we were interested in explaining dynamics. One way to assuage this concern is to redo the analyses with estimates calculated with weighted aggregation. Our concern with the aggregation approach, however, is that the least-populated states will have too few respondents to obtain valid estimates. Nonetheless, one compromise is to redo the analyses with the most-populated states and aggregated estimates. Inferences were nearly identical when we redid the analyses with the aggregated estimates and dropped the 10 least-populated states.