Context: State governments have been powerful sites of Republican resistance to the implementation of the Affordable Care Act (ACA), the Democratic Party's signature 2010 law. By influencing how citizens experience the ACA, state-level implementation can affect the national-level political implications of the law.
Methods: I examine three largely unstudied areas of marketplace implementation: navigator laws, transitional plan termination, and rating area configurations. For each policy area, I use linear probability models to investigate the determinants of state lawmakers bolstering or eroding marketplaces.
Findings: In each case, Democrat-controlled states were more likely to bolster marketplaces than Republican-controlled states were, with decisions more polarized in those policy areas—navigator laws and transitional plan termination—and with greater potential for national-level feedback. For navigator laws, where Republican state lawmakers were most cross-pressured by national party interests and local interests, marketplace eroding policy was highly associated with strength of conservative networks.
Conclusion: Crafters of federal legislation cannot expect state lawmakers to universally implement federal law to maximize the direct benefits to their constituents. Rather, we should expect state lawmakers to, in many instances, implement federal law in ways that benefit their parties.
In 2017, following the failure of congressional Republicans to pass a “repeal and replace” bill, the Trump administration introduced several executive actions that have undermined the Affordable Care Act's (ACA) health insurance marketplaces (see, e.g., Giovannelli and Curran 2018). The administration cut the open enrollment period in half, slashed funding for organizations tasked with helping individuals enroll in health insurance plans (Bump 2017), and more dramatically, announced that the federal government would not be paying out Cost-Sharing Reduction subsidies in 2018 (Jost 2017b). The Trump administration's undermining of ACA marketplaces likely comes as no surprise to close observers of ACA implementation. Journalistic accounts have been noting for several years the subtler undermining of ACA marketplaces pursued mainly in Republican-controlled states (see, e.g. Levey 2016).1
Why would state governments choose policies that limited the choice and affordability of health insurance for constituents? In an era of extreme partisan polarization and nationalization of state politics (Hopkins 2018; Rogers 2017), I propose that state policy might be used as a tool for political competition between national-level parties and interest groups.
Political scientists have long understood that public policy can have important consequences for politics through “policy feedback” processes (e.g., Pierson 1993). But scholars have only recently analyzed the question of when strategic lawmakers might be able to use policy for political gain (Anzia and Moe 2016). Moreover, existing work has not considered explicitly cases in which state policy choices influence politics at the national level—or more generally, where policies have multilevel feedback effects.
This is potentially a serious oversight, since state policy can influence national political battles in crucial ways. Most directly, state governments determine how votes are translated into national-level representation by drawing congressional districts (Stephanopoulos and McGhee 2015). State policies, such as felon disenfranchisement, also influence who is eligible to cast ballots in the first place (see, e.g., Highton 2017). State policy can also strengthen or weaken organized groups like unions that seek to influence policy in multiple states and federally (Feigenbaum, Hertel-Fernandez, and Williamson 2018; Flavin and Hartney 2015). Finally, state policy can influence how a national law is experienced by citizens, and therefore shape the degree to which that policy produces political gains or losses. In the case of the ACA, state policies that led to poorly functioning marketplaces made the ACA more vulnerable to repeal and gave the national-level Republican Party the opportunity to lay blame on their political opponents (Arnold 1992). On the other hand, state policies that led to well-functioning marketplaces gave the national-level Democratic Party the opportunity to claim credit.
ACA marketplace implementation is a well-suited and important case for considering the factors that motivate state governments to enact policies with national-level feedback effects. The high salience and polarized nature of the ACA renders marketplace performance, and therefore state marketplace implementation policy, highly consequential for national political battles. Furthermore, state control over technical, low-salience aspects of marketplace implementation provides a golden opportunity for lawmakers to influence marketplace performance while remaining firmly in voters' “blind spots” (Bawn et al. 2012).
I propose a framework for understanding the politics of ACA marketplace implementation that accounts for implementation policies' potential to produce national-level political feedback. State lawmakers are generally likely to pass policies that benefit their political parties, meaning Democrats would tend to bolster marketplaces and Republicans would tend to erode them.2 But for certain policy decisions state lawmakers faced a tension between local interests and the interests of the national party. Republican lawmakers in particular were cross-pressured by their constituents' interest in functional marketplaces and their party's interest in undermining the marketplaces. I highlight the importance of other mechanisms like federated ideological groups that might push state lawmakers to prioritize national-level interests at the expense of local interests.
I present evidence consistent with this framework from three largely unstudied areas of ACA marketplace implementation: navigator laws, transitional plan termination, and rating area configurations.3 While these policies were generally low salience, they had the potential to meaningfully influence the performance of marketplaces.4 In each case, Democrat-controlled states were more likely to bolster marketplaces than Republican-controlled states were, with decisions more polarized in those policy areas—navigator laws and transitional plan termination—with greater potential for national-level feedback. Moreover, for navigator laws, where Republican state lawmakers were most cross-pressured, marketplace-eroding policies were more common in states with strong conservative networks.
Background on ACA Marketplaces
The marketplaces provide a number of functions within the ACA health insurance regime. First, by eliminating medical underwriting, they pool risk across consumers such that those expected to incur high costs have greater access to affordable plans. Second, by mandating that plans are standardized, categorized into tiers according to actuarial value, and sold on government-run websites, the marketplaces allow consumers to more easily compare plans. Third, the marketplaces provide a mechanism for the delivery of income-based subsidies for both premiums and cost sharing.
Initial health reform packages proposed by Democratic congressional leaders in 2009 included a public option as a federal backstop in the case that private insurer offerings failed to offer consumers the competition-driven choice and affordability promised by marketplace proponents. However, the public option was ultimately dropped from legislation in order to retain the votes of conservative-leaning Senate Democrats like Ben Nelson of Nebraska (Klein 2013). Absent a public option, the fate of the ACA was heavily exposed to the ability and motivation of states to promote functional marketplaces.
This exposure has proven costly. State governments, particularly those controlled by Republicans, have adopted a variety of policies that studies demonstrate have had negative effects on the marketplaces. Most notably, nonexpansion of Medicaid pushed lower-income individuals, who tend to be of lower health status, onto marketplaces, reducing the health of the enrollee risk pool and putting upward pressure on premiums (Sen and DeLeire 2016). The vast majority of Republican-controlled states also declined to establish State-Based Marketplaces (SBMs) despite the financial incentives offered by the federal government to do so. Recent work indicates SBMs tend to outperform Federally Facilitated Marketplaces (FFMs), likely in part due to greater outreach and enrollment funding leading to higher enrollment rates (Zhu, Polsky, and Zhang 2018). Beyond these higher-profile measures, state governments have also enacted several lower-salience policies eroding local ACA marketplaces (e.g., Sommers et al. 2015).
Multilevel Policy Feedback and Cross-Pressured State Lawmakers
I propose a framework that can help to illuminate patterns of state behavior across implementation decisions, as well as state politics and policy more broadly. Central to the framework is the idea that state policy choices can have implications for national-level political competition through a multilevel policy feedback dynamic.
In general, the policy feedback literature investigates “the ability of policies—through their design, resources, and implementation—to shape the attitudes and behaviors of political elites and mass public, as well as to affect the evolution of policymaking institutions, and through any of these dynamics potentially to affect subsequent policymaking processes” (Mettler and Sorelle 2014: 152). Scholars have applied a feedback lens to a wide range of policy areas from Social Security (e.g., Campbell 2003) to welfare (e.g., Pierson 1996; Soss 1999) to criminal justice (e.g., Weaver and Lerman 2010). More recently, scholars have analyzed the way in which the potential for policies to produce feedback shapes their politics (Anzia and Moe 2016). Yet, scholars have not paid much attention to the sites and levels of government at which feedback effects manifest.
While it is not generally framed this way, existing work in state health policy (including ACA implementation) demonstrates the clear consequences state policy choices can have for national-level politics. For instance, states that expanded Medicaid featured higher rates of voting relative to nonexpansion states, at least in the short term (Clinton and Sances 2018; Haselswerdt 2017). More generally, Medicaid beneficiaries participate in politics at greater rates across a number of dimensions in states with more generous programs (Michener 2018). Due to the federal structure of American politics, state policies that influence political participation influence national elections.5
I propose that, like the policies discussed above, state implementation of ACA marketplaces produced national-level feedback—and that considering these effects can help illuminate the politics of ACA marketplace implementation. There are two mechanisms by which state marketplace implementation decisions would be expected to “affect subsequent policymaking processes” at the national level. First, state implementation choices affect national-level electoral politics. Consider a hypothetical scenario in which states across the board chose marketplace-bolstering policies. As indicated by the existing literature (e.g., Dickstein et al. 2015; Huth and Karcher 2016; Sommers et al. 2015), marketplaces would generally feature greater insurer and individual participation in addition to lower prices. Recent estimates from Kogan and Wood (n.d.), comparing voting in counties with high-performing versus poor-performing marketplaces, suggest Republicans benefited from poor marketplace performance in the 2016 presidential election. Note further that these types of analyses investigating the political consequences of county-level variation in marketplace performance likely underestimate the full effect of marketplace eroding. This is because, to the degree individuals factor marketplace performance into their vote choice, they are likely to take into account broader marketplace attributes in addition to the performance of marketplaces in their own county relative to other counties.
Due to the strong association between the ACA and the Democratic Party, the Democratic Party would generally benefit from marketplace bolstering, while the Republican Party would generally benefit from marketplace eroding. Furthermore, due to the United States' two-party system and winner-take-all elections, what is good for the Republican Party will tend to be bad for the Democratic Party, and vice versa.
Second, and related to the first, marketplace-eroding implementation policy would weaken the ACA and make it easier for Republicans in Congress to repeal the law—and vice versa for marketplace-bolstering policy. Between 2011 and 2016, Republicans in the House introduced 730 bills either retrenching or repealing the ACA (Rocco and Haeder 2018).6 Marketplace struggles are often cited by opponents of the ACA as a rationale for repeal (see, e.g., Healy and Goodnough 2016). Moreover, higher-quality marketplaces might motivate beneficiaries to mobilize in support of the ACA, while lower-quality marketplaces might seed opposition or indifference to the law.7
How would these feedback effects influence the behavior of state lawmakers? Most basically, we might expect reelection-motivated state lawmakers to choose policies that benefit their broader parties (see, e.g., Mayhew 1974). Recent work suggests that voters prioritize national-level factors even in state elections, with presidential approval three times as predictive of votes for state legislative seats than state legislature approval (Rogers 2017). As a result, reelection-motivated state lawmakers have an incentive to use policy to burnish their broader party brand. To the degree that marketplace implementation policy would affect ACA repeal prospects, we might also expect Republican state lawmakers to seek to erode marketplaces and Democratic state lawmakers to bolster them.
The feedback logic, in this way, aligns with other important reasons why state lawmakers might choose marketplace-eroding or -bolstering policies. For instance, state lawmakers might choose policies consistent with national party goals as an expression of ideology, or simply to be “team players.” In addition, state lawmakers in Republican-controlled states might choose marketplace-eroding policies in response to the demands of anti-ACA constituents, and vice versa in Democrat-controlled states. Indeed, existing studies have emphasized the importance of partisan control of office in predicting ACA state implementation across several policy areas (e.g., Beland, Rocco, and Waddan 2016; Hertel-Fernandez, Skocpol, and Lynch 2016; Jacobs and Callaghan 2013; Jones, Bradley, and Oberlander 2014; Rigby and Haselswerdt 2013).
Yet, despite all of these factors pushing in the same direction, we see variation in implementation policies within partisan control of state government, particularly on the Republican side (see table 1). Republican-controlled states did not universally erode ACA marketplaces across each policy dimension. This suggests that state lawmakers were, on some policy choices, cross-pressured. While state lawmakers do have an incentive to promote national party brand, they also have an incentive to respond to local interests. Indeed, Hertel-Fernandez, Skocpol, and Lynch (2016) highlight cross-pressured Republican state lawmakers in their study of state Medicaid expansion. Similarly, in determining whether to bolster or erode marketplaces, Republican lawmakers, for certain policies, faced a tension between producing positive feedback for the party by eroding the ACA marketplaces and responding to local interests in functional marketplaces.
With all the factors pushing state lawmakers to align with their parties on ACA marketplace implementation, there are two important reasons (besides the very existence of variation in policy decisions within party control of government) to think lawmakers were cross-pressured. First, the logic of retrospective voting suggests voters would punish incumbents for adverse outcomes like expensive or low-quality health insurance options. This expectation stems from a well-developed literature demonstrating that voters tend to reward incumbents for strong economic performance and punish incumbents for weak economic performance (Healy and Malhotra 2013). Scholars have also demonstrated voter responsiveness to a number of other performance indicators like student test scores (Berry and Howell 2007), natural disaster assistance (Gasper and Reeves 2011), and, at the local level, road quality (Burnett and Kogan 2017). Voters similarly might reward or punish incumbents on the basis of their access to affordable, high-quality health insurance options. Importantly, retrospective voting in the context of the ACA would not necessarily depend on the traceability of outcomes (Arnold 1992) to policy choices—only on voters' assessments of how they are doing.
Second, state lawmakers might discount national-level feedback effects. Anzia and Moe (2016) point out that even where lawmakers have the opportunity to use policy for political gain, collective action problems can prevent them from doing so, since individual lawmakers may benefit from a policy's feedback effects regardless of whether they contribute to its passage. The extent of the collective action problem is unclear in the case of ACA marketplace implementation, since state lawmakers may be rewarded by voters for representing national party interests. However, there remain clear externalities to the behaviors of individual lawmakers. Republicans as a whole benefit from weak marketplace performance produced by marketplace-eroding policies (regardless of whether they enact marketplace-eroding policies), and vice versa for Democrats. These externalities therefore might lead state lawmakers to privilege local interests over national party interests.
Applying the multilevel feedback framework thus suggests two central hypotheses. First, the partisan division of a policy would be a function of the policy's feedback potential. Policies with greater potential to produce political feedback are more likely to be polarized, and vice versa. Second, mechanisms that push state lawmakers to adopt policies that benefit the national-level party are likely to be particularly important for those policy areas on which state lawmakers are cross-pressured.
In this general case, Republican state lawmakers were more likely to be cross-pressured since producing positive feedback for the party required eroding local marketplaces. However, there was a clear mechanism pushing Republican state lawmakers to erode marketplaces: cross-state conservative groups like American Legislative Exchange Council (ALEC), State Policy Network (SPN), and Americans for Prosperity (AFP). Due to their federated structure and investments in state politics, cross-state conservative groups were well-positioned to coordinate resistance to the ACA in the lead-up to the opening of marketplaces (Skocpol and Hertel-Fernandez 2016) and had a long-term goal of ACA repeal that aligned with the interests of the national Republican Party (ALEC 2011).
Often working in concert, these groups use a number of mechanisms to influence state policy. Among other things, ALEC disseminates model bills that members are encouraged to introduce and support. By subsidizing the crafting of legislation (Hertel-Fernandez 2014), ALEC reduces the cost of state lawmakers to erode ACA marketplaces. But, as Hertel-Fernandez, Skocpol, and Lynch (2016) show, the power of these federated conservative organizations goes well beyond writing model bills. AFP uses its vast resources to influence primary and general elections, encouraging the rise of sympathetic politicians and credibly threatening incumbents (Skocpol and Hertel-Fernandez 2016). Think tanks associated with the SPN disseminate studies and analysis supporting proposed policies—and attacking alternatives. Existing work suggests these groups were highly engaged, and often effective, in resisting the successful implementation of the ACA (Hertel-Fernandez, Skocpol, and Lynch 2016; Jones, Bradley, and Oberlander 2014). In particular, Hertel-Fernandez, Skocpol, and Lynch (2016) highlight the crucial role these groups played in pressuring state-level Republicans to neglect local interests by not expanding Medicaid.
In sum, the logic of multilevel policy feedback suggests we should expect Democrat-controlled state governments to bolster marketplaces, and Republican-controlled state governments to erode them. For Democratic state lawmakers, there is not generally a tension between what is good for constituents and what is good for the national party, since both benefit from marketplace bolstering. The national-level Republican Party benefits from marketplace-eroding policies, but state-level Republicans might be cross-pressured on certain policies. Where Republicans are cross-pressured, the strength of conservative networks might be a key predictor of whether states enact marketplace-eroding implementation policies. In the following section, I introduce three marketplace implementation decisions and map them onto this theoretical perspective.
The ACA includes funding for organizations and individuals—so-called navigators—to assist consumers with enrolling in and using health insurance. Navigator laws, laws that restrict the activities of health navigators, were enacted in a number of states in 2013 and 2014 legislative sessions. The restrictions included limitations on advice navigators can provide,8 in-state residency requirements that prevent national groups from serving as navigators, prohibitions on receiving insurer compensation (which generally disqualifies health care providers from serving as assisters), and requirements that navigators carry certain types of insurance.9
While proponents of navigator laws argue that they are necessary for consumer protection, most ACA advocates believe these laws have little purpose besides hindering outreach and enrollment efforts (Jost 2013). Indeed, evidence suggests restrictive navigator laws have reduced ACA awareness and enrollment rates (Sommers et al. 2015). Since enrolling a large number of individuals is key to the long-term sustainability of marketplaces, these laws would tend to erode ACA marketplaces.
The case of navigator laws pitted local interests in well-functioning marketplaces against the national Republican Party's interest in the erosion of the ACA. Navigator laws would weaken ACA marketplaces and therefore produce positive feedback for the national Republican Party both by improving electoral prospects and enhancing prospects to repeal a law that key policy demanders within the party opposed. Yet, eroding marketplaces through navigator laws required actively moving the status quo policy in a way that would produce local costs in the form of greater difficulty finding health insurance and more poorly performing marketplaces (Sommers et al. 2015). Furthermore, due to navigator lows' salience, state lawmakers had little to gain among voters by signaling opposition to the ACA. These factors all suggest navigator laws would be uncommon in Democrat-controlled states and might only be enacted in states with strong elements of conservative networks promoting the laws.10
Transitional Plan Termination
The second state marketplace implementation decision considered is whether states terminated or extended transitional plans after marketplace opening. The grandfathering clause of the ACA stipulates that individuals enrolled in noncompliant plans prior to 2010 would be permitted to remain on those plans as long as they continued to be offered (CCIIO n.d.). This clause, however, did not apply to plans initiated between the passage of the law in 2010 and the opening of marketplaces in 2014 (“transitional” plans). Due to political pressure, the Obama administration announced in 2013 that it would pass the buck to states to decide whether to extend transitional plans (Jost 2017a).
Terminating transitional plans, while imposing salient costs on the younger, healthier individuals who enrolled in these plans, would be expected to bolster the long-term sustainability of marketplaces. This is because marketplace sustainability requires enrollment of a balanced pool of healthier and less healthy individuals. Allowing healthier individuals to remain in a separate risk pool would tend to increase premiums on the marketplace and increase the risk of a premium “death spiral” (Cutler and Zeckhauser 1998). Existing analysis suggests these decisions mattered for marketplace enrollee composition, with nonterminating states tending to feature less healthy enrollees on average (Huth and Karcher 2016; Semanskee, Cox, and Levitt 2016).
Due to its positive effects on the marketplaces, namely, in putting downward pressure on premiums and promoting long-run sustainability, termination of transitional plans would produce positive feedback for the national-level Democratic Party. However, unlike for navigator laws, in this case bolstering local marketplaces required state lawmakers to actively move the status quo policy. In addition, bolstering marketplaces required imposing salient, concentrated costs on constituents whose plans would be terminated. Thus, in this area, we can expect a tension for Democrat-controlled state governments as opposed to Republican-controlled state governments. For Republicans, not terminating transitional plans avoided political costs while contributing to ACA marketplace erosion, which benefits the party. Democrats, though, had to choose between imposing a salient cost on transitional plan enrollees, on the one hand, and undermining long-run marketplace stability, on the other.
Under what conditions would states terminate transitional plans? First, we might expect Democrat-controlled state governments to be more likely to invest in long-run marketplace robustness by terminating transitional plans in states where the Republican coalition was weaker and so did not pose a strong electoral threat. Second, we might expect the decision calculus to depend on the degree to which state governments were politically invested in the ACA marketplaces, which can be measured by whether they were on track to establish an SBM. State governments without SBM's might be less likely to terminate transitional plans in order to bolster the marketplaces, since they were less likely to receive credit for marketplace functioning (or blame for marketplace dysfunction) (Arnold 1992).11
Rating Area Configurations
The final ACA marketplace implementation policy I consider is the configuration of rating areas within states. States were required to set geographic rating areas within which specific plan premiums would vary only by defined age and smoking bands—defining the geographic level of risk pooling. While the literature on optimal rating area configurations is sparse, the existing work suggests that rating area configurations should aim for large enrollee populations with low heterogeneity of health risk, or projected health spending (Dickstein et al. 2015). This allows insurers to spread risk across a large number of individuals while not encountering too much potential variation in expected costs depending on who enrolls in their plans.
In determining marketplace rating areas, states could choose configurations based on 1) counties, 2) zip codes, or 3) Metropolitian Statisical Areas (MSA). Alternatively, they could default to the federal standard, which would set each MSA as a rating area, with all non-MSA territory constituting an additional rating area (so the total number would be the number of MSA's in a state, plus one). Nondefaulting states could configure rating areas by counties, zip codes, or MSA's, but if the total proposed number exceeded the number of MSA's plus one they were required to apply for approval from the federal government.
The effort states spent determining rating areas varied considerably. In several states, choosing rating area configurations was an intensive, analytical process. For instance, in California, where the rating area configuration was determined through legislation, the Department of Insurance produced and disseminated to the legislature an actuarial study arguing that their proposed plan would minimize disruption to consumer rates. On the other end of the spectrum, a number of states did not—at least based on what is discernable from public information—spend any resources evaluating rating area configurations options, and simply defaulted to the federal standard.
The case of rating areas is similar in some ways and different in others from the previous two. The basic framework remains. With the power to configure rating areas, state governments could choose either a configuration well suited to the health geography of their states, thus bolstering the local ACA marketplace and producing positive feedback for Democrats—or one ill-suited to the health geography of their states, thus eroding the local marketplace and producing positive feedback for Republicans.
However, unlike in the case of navigator laws, where eroding required erecting burdensome regulations, suboptimal rating area configurations do not impose clear costs to states (beyond the effect on the marketplace itself). Unlike in the case of transitional plan termination, which required taking low-cost plans away from constituents, there were not clear costs to marketplace bolstering. Moreover, unlike in the other policy areas discussed, rating area configuration decisions had to be approved federally if they departed significantly from Health and Human Services standards. This raised the transaction costs to trying to use rating areas to erode marketplaces, in addition to lowering the likelihood of successful erosion. This meant there was minimal scope for state-level Republicans to produce positive feedback for the national party. Given these factors, I would hypothesize a reduced role of partisanship and conservative networks.
Data and Methods
Testing the hypotheses outlined above requires a data set linking ACA implementation policy choices to various state characteristics, including partisan control of government and strength of conservative networks, in addition to a set of control variables to address potential confounding. Policy choice data was drawn from several sources. The navigator law outcome variable was produced for a report from the Commonwealth Fund, while transitional plan termination data came from healthinsurance.org (Norris 2016). Rating area configuration data was drawn from the Centers for Medicare and Medicaid Services (CMS 2018).
I use a measure of conservative network strength at the state level developed and applied by Hertel-Fernandez, Skocpol, and Lynch (2016) in their paper on Medicaid expansion. As discussed in that paper, the measure includes four components. The first component accounts for the strength of ALEC, measured by the share of state legislators who were ALEC members as of 2013 as well as how many of the state's top four legislative leaders were affiliated with ALEC. The second component accounts for the presence of SPN-affiliated think tanks, and is measured as the relative budget of SPN-affiliated think tanks to the budget of think tanks on the center left and left. The third component accounts for the cross-state lobbying efforts of representatives of the Foundation for Government Accountability (FGA), an SPN-affiliated think tank formed to lobby on issues of health and welfare in the states. This component is measured by the activity of FGA in a state regarding Medicaid expansion, which is likely highly correlated with activity on the marketplace implementation policies analyzed here. The fourth component accounts for the strength of AFP, recording whether AFP had an office during the ACA implementation period and the length of time the office had existed beforehand. Appendix A displays the distribution of the measure across the states.12
For each of the policies, I first inspect cross-tabs of policy choices by state control of government. Second, I estimate linear probability models to more systematically test which factors were predictive of policy choices across the three areas.13 In regression models, I incorporate a number of variables that prior studies have shown to be predictive of ACA implementation policy, including policy legacies (Beland, Rocco, and Waddan 2016; Jacobs and Callaghan 2013), administrative capacity (Jacobs and Callaghan 2013), and ideology (Shor 2018).
Cross-tabs displaying state government choices across the three policy areas outlined above are presented in table 1. As expected, the vast majority of the variation in navigator law enactment is within Republican-controlled and divided states. On the other hand, not a single Republican-controlled state terminated transitional plans, while there was some variation in policy choices among divided and Democrat-controlled states. Finally, very few states defaulted to the federal standard when it came to rating area configurations, suggesting minimal marketplace erosion through this policy mechanism.14
I turn next to estimating linear probability models of policy choices. Table 2 presents results from estimating linear probability models using a number of model specifications with navigator law enactment as the dependent variable, and state-level attributes as independent variables. The model featuring only state control of office explains just 16% of the variation in navigator law enactment. Adding conservative network strength to the model improves explanatory power markedly, with conservative network strength predictive of navigator law enactment.15 Column 3 adds SBM establishment to the model, which I find to be negatively associated with navigator law enactment.16 This suggests that states investing politically in marketplaces by establishing SBMs were less likely to erode those marketplaces.
Column 4 estimates a model featuring several additional covariates to account for confounding and test some alternative hypotheses. First off, to the degree that conservative network strength or SBM establishment are associated with general state conservatism, results could alternatively be driven by legislator ideology or constituent preferences regarding the ACA. To address this concern, I include measures of state-level favorability toward the ACA as of 2012, as well as the mean estimated ideology in each party across state legislative chambers from Shor and McCarty's American Legislatures Project (Shor and McCarty 2011).17 The measure of state-level ACA favorability comes from Barrilleaux and Rainey (2014) and is generated by applying a multilevel regression and model to Kaiser Health Tracking Poll data.18 Strikingly, the results indicate neither ACA favorability nor legislator ideology are (conditional on other covariates) strongly related to navigator law enactment.19
I also include measures of ideological dispersion of state legislators in each of the parties, testing Anzia and Moe's (2016) argument that ideologically heterogeneous coalitions are less likely to produce positive feedback. While the direction of the coefficients on measures of dispersion are consistent with Anzia and Moe (2016), with more heterogeneous Republican coalitions less likely to pass navigator laws, and the opposite for Democrats, estimates are not statistically significant.20 Finally, I include measures of pre-ACA Medicaid generosity and administrative capacity from Callaghan and Jacobs (2016, 2017). Neither is a statistically significant predictor of navigator law enactment. In the fully specified model, the only statistically significant predictors of navigator law enactment are conservative network strength and SBM establishment.
I turn next to transitional plan termination. Linear probability models, presented in table 3, demonstrate that a significant portion of the variation in transitional plan termination can be explained by a simple model featuring control of state office (Republican control being associated with extension of transitional plans). Adding conservative network strength only increases explanatory power marginally. While the coefficient on the conservative network strength variable is significant in this model, it is not significant in the model including SBM establishment (column 3), suggesting a weak association. On the other hand, including SBM establishment increases model fit substantially, with SBM establishment significantly associated with transitional plan termination. This suggests that, as expected, Democratic lawmakers in SBM states were more willing to incur short-term costs to bolster marketplaces in the long run.
Similar to the prior analyses, I do not find ACA favorability or legislator ideology to be significantly related to the outcome. However, transitional plan termination was more common in states with ideologically heterogeneous Republican coalitions. One explanation for this result is that heterogeneous Republican coalitions posed less of a threat to a government controlled by Democrats, making them more willing to take on the risks of terminating transitional plans.
I turn finally to rating area configurations. Recall that, since state lawmakers had less scope to influence marketplaces using rating area configurations, I expected less polarization. As a first cut at exploring the political determinants of rating area configurations, I code the outcome variable based on whether or not states defaulted to the federal standard. While this is not a precise measure of marketplace eroding, it signals a lack of interest in actively promoting optimal rating areas. Regression results, presented in table 4, demonstrate that, while Democratic control is associated with nondefault, the relationship is weak and not statistically significant. Indeed, none of the variables is significantly associated with rating area defaulting in any of the four models.
The measure of eroding based on defaulting to the federal standard is a rough measure, though. It is likely that for some states the federal standard was decently suited to the state's health geography, while other states may have actively chosen poor rating area configurations. As a robustness check, I compute a measure of the quality of a state's rating area configuration based on the degree to which rating areas reduced the pooling of highly heterogeneous health risk. I code health risk scores at the county level using measures published by Blue Cross Blue Shield (BCBS 2017) reflecting the actual health spending of enrollees, run an analysis of variance (ANOVA), and compute the corresponding F-statistic for each state. Higher F-statistics indicate that a greater proportion of the total variation in health risk is accounted for by the rating area divisions, which would tend to have positive effects on the marketplaces (Dickstein et al. 2015).21
This measure is also imperfect. Most problematically, it cannot be applied to states like Vermont that use one rating area for the whole state, or to states like Florida that classify each county as a separate rating area, reducing sample size considerably. Despite these problems, I recover similar results, presented in appendix B, using this measure as with the simple measure based on states defaulting to the federal standard.22
Patterns of state marketplace implementation policy across the three policy areas explored support the hypotheses put forward, and the corresponding theoretical framework. Passing navigator laws would produce positive feedback for the Republican Party, but required actively making policy that would be costly to many constituents. I hypothesized that conservative networks would be key to pressuring state lawmakers to enact these laws. Indeed, in this policy area, strength of conservative networks was an important predictor of whether states eroded marketplaces.
With respect to transitional plans, eroding the marketplaces was much easier. Terminating transitional plans (the marketplace-bolstering policy) required a change from the status quo, and provoked a backlash from consumers whose plans would be canceled. As a result, variation in this case generally occurred among Democrat-controlled states. Moreover, the evidence suggests Democrat-controlled states were more likely to bolster the marketplaces—producing positive feedback for the national party—where they had invested in marketplace performance by establishing SBM's, and where the Republican coalition posed less of a threat.
While investigating rating area configurations poses some measurement problems, the results, on balance, suggest that factors like control of state office, strength of conservative networks, and prior SBM establishment played a weaker role than in the other implementation areas examined. The potential feedback effects were minimal in this case, since configurations had to be approved federally. In addition, the fact that rating area configurations were determined bureaucratically perhaps limited the power of the cross-state conservative groups, which tend to exert the most influence over state legislators (Hertel-Fernandez, Skocpol, and Lynch 2016).
The analysis also provides evidence that the role of ideological homogeneity in state party coalitions plays less of a role in determining whether state-national feedback is produced than it does in cases where strong within-state feedback effects would be expected (Anzia and Moe 2016). This makes theoretical sense. If ideological homogeneity leads rank and file state lawmakers to invest additional authority in state party leaders (Aldrich and Battista 2002), it likely also facilitates the enactment of policies that are politically beneficial in those states. Investment of greater authority in state party leaders is likely to be less consequential in an environment where state parties can free-ride off of policies passed in other states that produce political benefits for the national-level party.
There are several limitations of the study that I address here. First, I do not measure variation in the strength or preferences of concentrated local interests like health insurance companies with a stake in these policy decisions. Indeed, Hertel-Fernandez, Skocpol, and Lynch (2016) argue that Medicaid expansion in Republican-controlled states depended on the relative power of local business groups versus cross-state ideological groups, with local Chambers of Commerce tending to support expansion. While I do not deny that local interests likely matter in the cases I study as well, they seem less relevant than in the case of Medicaid expansion. One reason is that in several of the cases I examine there are likely competing local groups, as opposed to a unified local front. For instance, laws restricting publicly funded navigators benefited competing private health insurance navigators but likely hurt health insurers to the degree they reduced enrollment. Similarly, transitional plan termination benefited insurers with a large portfolio of transitional plan enrollees but likely hurt health insurers committed to the marketplace. Perhaps due to divided local interests, combined with the relatively lower stakes of these policies, I do not find evidence of state Chambers of Commerce taking clear stands on the implementation issues I study. Additionally, it seems unlikely that the variation in strength and preferences of local interests is both meaningful enough and sufficiently associated with the factors I study to drive the findings.
Second, this study does not address the degree to which state lawmakers intentionally eroded or bolstered marketplaces to produce certain national-level feedbacks. The proposed theory is concerned with the conditions under which state lawmakers would enact policy that advantages their party, as opposed to the intentions of state lawmakers. While the intentions of state lawmakers will likely remain unknown, conservative groups have not been shy about their willingness to undermine ACA marketplaces as a step toward repealing the law.23
Third, there are limits to the inferences that can be drawn from observational data. While the evidence is consistent with the proposed theoretical framework, the design does not permit strong causal claims. More specifically, there is always the potential for omitted variable bias. One alternative explanation for the patterns uncovered in the empirical analysis is that the measures of conservative network strength are simply serving as proxies to other factors like ideology. Relatedly, there is the concern that the strength of conservative networks in a state is in itself endogenous to preexisting factors that themselves are associated with implementation policy decisions.
Yet, the evidence suggests these confounding factors are not driving results. Inspecting table 2, the coefficient associated with conservative network strength is larger in the fully specified model featuring measures of ideology and ACA favorability than in the model featuring only state control of government, SBM establishment, and conservative network index. If conservative networks were taking hold generally in those ideologically conservative states predisposed to erode ACA marketplaces, we would expect that controlling for ideology would reduce the magnitude of the conservative network coefficient. Beyond the empirical evidence, patterns of state passage of navigator laws provide reason to believe that state ACA marketplace implementation is not driven by principled expression of ideology. In particular, the marketplace-eroding policy required writing additional government regulations aimed at consumer protection, which is not generally associated with conservative principles. Finally, if preexisting factors associated with conservative network strength were driving results, we might expect conservative network strength to predict implementation policy across each policy dimension—but, the measure is only strongly predictive of navigator laws.
State policy can have important consequences for national-level competition between political parties and interest groups (Feigenbaum, Hertel-Fernandez, and Williamson 2018; Flavin and Hartney 2015; Stephanopoulos and McGhee 2015;). Yet, political scientists have not addressed the question of how these dynamics might influence state policy choices. This article provides an early step to begin to answer this question.
With state policy increasingly nationalized (Hopkins 2018) and polarized (Grumbach 2018), we should expect state lawmakers to generally adopt policies that benefit their national-level parties. However, national-level groups and parties face limitations in using state policy to promote their broader political interests. In particular, state lawmakers might be cross-pressured by national party interests and local interests. In these cases, producing positive feedback is likely to depend on other mechanisms like federated policy networks.
Evidence from ACA marketplace implementation lends support to this theoretical framework. Democrat-controlled states generally bolstered marketplaces, while Republican-controlled states generally eroded marketplaces. However, comparing patterns across multiple implementation policies deepens this story in two important ways. First, polarization in implementation was stronger for policies with greater potential for national-level feedback. Second, cross-state variation in the strength of conservative groups played a more important role where eroding marketplaces required a locally costly departure from the status quo.
Theoretically, this work brings together two areas of political science—federalism and policy feedback—in a way that should be fruitful for future research. Recent literature in federalism has emphasized the increasing degree to which states act not as separate sites of governing authority, but rather as alternative venues of partisan contestation (Bulman-Pozen 2013). Moreover, scholars have shown that policy increasingly diverges based on state control of government (Caughey, Warshaw, and Xu 2017; Grumbach 2018). In addition, at least on the Right, organized networks focused on influencing state policy have grown in strength over time (Skocpol and Hertel-Fernandez 2016). These trends are important in their own right, but they also have serious implications for policy feedback.
If nationally organized groups have sway in statehouses, they are likely to promote policy with feedback effects that improve their national position. Moreover, state lawmakers may not provide much resistance, since evidence suggests they are evaluated by voters primarily based on national-level politics (Rogers 2017). I argue that in the case of the ACA these forces led to the erosion of state marketplaces in Republican-controlled states, which generally produced adverse outcomes for constituents, but positive outcomes for the Republican Party.
This work has important implications for policy makers. The framework and analysis suggests crafters of federal legislation cannot expect state lawmakers to universally implement federal law in order to maximize the direct benefits to their constituents. Rather, as a result of the greater role of national-level political forces at the state level, we should expect state lawmakers to, in many instances, implement federal law to maximize benefits to their party.
Implementation of highly polarized national law is only one of several potential mechanisms of state-national policy feedback that scholars might investigate. Future work might apply the framework developed here to state policy decisions in areas like labor, energy, voting rights, and criminal justice that are likely to have meaningful political effects at the national level.
Thanks to Sarah Anzia, Eric Schickler, Paul Pierson, Alex Hertel-Fernandez, Sarah Gollust, Laura Stoker, three anonymous reviewers, and seminar participants at UC Berkeley and the American Political Science Association 2018 meeting for helpful comments. Thanks also to Alex Hertel-Fernandez and Janet Weiner for kindly sharing data. I also acknowledge support from the National Science Foundation Graduate Research Fellowship Program under DGE 1752814. Any errors are my own.
Interestingly, analysis from the Kaiser Family Foundation suggests that Trump Administration policies have led to reduced enrollment in Federally Facilitated Marketplaces (FFMs), but not in the generally Democrat-controlled State-Based Marketplaces (SBMs).
This political logic refers in particular to the Obama era, and has likely changed under the Trump administration.
While the effects of these decisions have been studied, little attention has been paid to their determinants.
Transitional plan termination was relatively higher salience for those constituents whose plans were terminated.
Unless of course the increase is only observed for state-level elections, which is not borne out in the studies referenced.
The American Health Care Act of 2017, which would have repealed key provisions of the ACA, was passed in the House despite low public approval, but was narrowly defeated in the Senate.
Due to the importance of the ACA for key “policy demanders” (Bawn et al. 2012) in the national-level parties (Rocco and Haeder 2018), I consider effects of state policy decisions on repeal prospects to be a subset of effects on the parties.
This particular class of laws was preempted by a federal court ruling in 2016 (St. Louis Effort for AIDS v. Huff; media.npr.org/documents/2014/jan/missouriorder.pdf).
Roll-call votes for these laws are available for many of the states that passed navigator laws. However, these data are very noisy, since in many cases language concerning navigators was inserted into broader pieces of legislation. Moreover, final roll-call votes were not taken in the majority of states where laws were proposed but not enacted, restricting the scope of comparisons that could be made.
Indeed, ALEC's dissemination of a “model bill” (Hertel-Fernandez 2014) suggests the involvement of conservative networks in advancing navigator laws.
By the time state governments were making decisions about transitional plans, they would have already determined whether or not to establish an SBM.
Some of these measures are recorded after ACA implementation was underway, raising concerns of posttreatment bias, where the measure itself is a function of the outcome variable. That said, it seems unlikely that how a state was implementing the ACA would exert a strong influence on ALEC membership.
Results are robust to using a generalized linear model like logit, but linear probability estimates are presented since they are easier to interpret.
Of course, it is possible that states actively selected ill-suited rating area configurations to erode marketplaces. I address this concern using regression analysis (see Appendix B), where I am able to investigate a continuous measure of rating area quality that I cannot capture using cross-tabs. Note, however, that of the 7 states that defaulted to the federal standard, 5 (Alabama, New Mexico, North Dakota, Oklahoma, and Texas) featured rating area quality scores below the median, while 3 were among the 10 lowest-scoring states.
Since Virginia is a positive outlier on the conservative network measure and passed a navigator law, results might be driven by this single case. However, the finding is robust to excluding Virginia from the data.
Similar factors likely influenced both navigator law enactment and SBM establishment, complicating interpretation of these models. Results are generally robust to excluding SBM establishment from models.
Results are also robust to using mean estimated ideology across the state legislative chambers (versus by party).
See github.com/carlislerainey/aca-opinion/blob/master/README.md for more information on the measure.
Results are robust to substituting Obama's 2012 vote share for ACA favorability.
Lack of statistical significance should not be considered evidence against this theory, especially given the low sample size and resultant low power in the present analysis.
Defaulting to the federal standard is associated with more poorly rating area configuration (p < .05, correlation coefficient = .23), lending support to the validity of the measures.
Note that in the fully specified model (column 4) for the rating area quality-robustness check (appendix B) having a more conservative state Republican party coalition is associated with lower-quality rating area configurations. However, in this model (but not in others), strength of conservative networks is also associated with higher-quality rating area configurations, suggesting potentially spurious associations. Instability of estimates depending on model specification reflects the relatively small sample size (reduced to 40).
ALEC's 2011 State Legislatures' Guide to Repealing ObamaCare includes a section titled “Decline to Build the ObamaCare Edifice,” which recommends states reject grants to establish marketplaces and decline to enact ACA rulemaking.