The Affordable Care Act (ACA) has allowed researchers to examine mass policy feedback effects—how public policies affect individuals' attitudes and political behaviors—in real time while using causal models. These efforts help address criticisms of the extant feedbacks literature and have revealed new policy feedback effects and new information on the conditions under which policy feedbacks occur. The ACA case also raises empirical and theoretical questions about the types of data needed to assess feedback effects, the magnitude of policy effects required for detection, the time frame in which feedbacks occur, and the suitability of various empirical approaches for assessing policy feedback effects. Thus, the ACA not only adds an important empirical case to the study of policy feedbacks but also helps refine policy feedback theory.
For scholars of policy feedbacks, the implementation of the Affordable Care Act (ACA) held great theoretical and empirical promise. Those who study how public policies might reshape politics had an opportunity to evaluate the effects of a major new policy in real time and to employ causal methods, addressing past concerns with this body of research. The possibility that the ACA case might reveal new types of feedbacks or new mechanisms was exciting as well.
Thus far, scholars have uncovered some positive attitudinal effects: those with personal experience getting new or improved health insurance coverage are more favorable toward the law; and some positive behavioral effects: those personally affected are more likely to vote. But the effects are modest and sometimes temporary or contradictory (in some analyses recipient political participation declines). We may need to reconsider what types of social policy benefits can produce feedback effects, yielding more nuanced theory.
Concerns with Policy Feedbacks Theory and the ACA's Possible Contributions
A burgeoning literature examines public policies not just as outcomes of political processes but also as inputs. Existing policies may reshape the political environment and subsequent policy making by altering the resources and interests and mobilizing capacities of political actors at both the elite and mass levels (Béland 2010; Campbell 2012). Empirical cases of mass public reactions to social welfare policy and beyond (e.g., Simonovits et al., forthcoming) have accrued, typically showing that policies have resource and “interpretive” effects (Pierson 1993) that can be positive or negative, boosting or undermining recipients' rates of political participation or altering their preferences on public policies.
Despite theoretical and empirical advances, the policy feedbacks literature has faced both limitations and criticism. One critique warns that researchers have selected on the dependent variable, looking for cases of apparent feedbacks and then reading backward into the historical record. In many cases, we know where we have ended up and we look to program parameters and experiences to explain these patterns. The ACA promised a prospective opportunity to see what happens when Americans receive new social policy benefits—a more rigorous test in which hypothesis development preceded outcome measurement, not vice versa.
A related critique is that much policy feedbacks work has utilized observational data. Scholars have tried to strengthen causal inference, using longitudinal research designs, instrumental variables models, matching techniques, and so on, but concerns about selection and other threats to inference remain. The attitudinal and behavioral effects attributed to public policies could arise instead from preexisting characteristics of the target populations. With the quasi-experimental rollout of many of its provisions, the ACA provided an opportunity for stronger causal inference.
ACA Mass Policy Feedback Effects Thus Far
Regarding political behavior feedbacks, scholars have examined three aspects of the ACA: the Medicaid expansion, the expansion of private insurance through the Health Insurance Marketplace, and the dependent care provision allowing those under 26 to remain on their parents' insurance. Most of this work has examined policy effects on voter turnout and registration, and the findings are modest and mixed. Studies using aggregate data found that the extension of insurance coverage has increased political participation, but those using individual-level data found no effect on participation.
Haselswerdt (2017) found that aggregate turnout in House races declined less from the 2012 presidential election to the 2014 midterm in states that had expanded Medicaid, compared to those that had not. Clinton and Sances (2018) compared counties in expansion and nonexpansion states sharing a border and found that voter registration and turnout among low-income citizens under 65, the target population, increased in expansion-state counties, particularly those with a high share of eligible citizens (although the turnout effect fades over time). Courtemanche, Marton, and Yelowitz (2019) examined the effect of both Medicaid expansion and private insurance expansion via the Marketplace and the individual mandate (in place until the penalty was zeroed out by the December 2017 Tax Cuts and Jobs Act). Using individual-level participation data and estimated insurance coverage, they found that the ACA's effects were small and statistically insignificant. Chattopadhyay (2017) compared political participation among those just above and below the dependent care provision age threshold and found little effect.
Scholars have speculated on the mechanisms linking the ACA and political behavior, although existing data do not permit their rigorous evaluation. One possibility is that, in providing health insurance, the ACA may improve physical health, which is associated with greater political participation (Burden et al. 2017; Gollust and Rahn 2015; Pacheco and Fletcher 2015). Or it may boost mental health diagnosis and treatment, as did the Oregon Health Plan lottery, or reduce stress and anxiety by improving low-income families' financial stability (Baicker et al. 2013), allowing them to pursue the “luxury good” of political participation (Rosenstone and Hansen 1993).
A second mechanism could be political engagement, including political interest, knowledge, and efficacy (Verba, Schlozman, and Brady 1995) or positive “interpretive effects” (Pierson 1993). Gaining insurance through the ACA might increase recipients' sense of stake in public affairs (Clinton and Sances 2018) or make recipients feel more incorporated into the polity as deserving citizens (Pierson 1993) or more grateful to government in a way that enhances civic engagement (Mettler 2005).
A third possible mechanism is mobilization. Under the 1993 National Voter Registration Act, social assistance agencies, including the health exchanges and Medicaid offices, are required to offer voter registration, which may explain increased turnout in the Medicaid expansion states (Clinton and Sances 2018). After the 2016 election ushered in unified Republican control of government and threats to repeal the ACA, grassroots groups emerged to defend the law (Gose and Skocpol 2019; Meyer and Tarrow 2018). Policy threat can boost participation among social program recipients (Campbell 2003), although even here the evidence is mixed; for example, previous Medicaid cuts, such as Tennessee's 2005 rollback, resulted in greater turnout declines in the counties with the largest disenrollment (Haselswerdt and Michener 2019).
A new feedback effect that has emerged in the ACA case is backlash. The ACA was debated, passed, and implemented during a time of great partisanship; there is a 60– to 70–point gulf between Republicans and Democrats in favorability toward the legislation. So strong are partisan feelings that take-up varies by party identification: not only do many Republicans oppose the law even if they might benefit from it (Kliff 2016), but also Republicans in need of insurance are less likely to sign up if they encounter the government interface (healthcare.gov) than a private one (healthsherpa.com) (Lerman, Sadin, and Trachtman 2017). Some of the increased political participation after ACA implementation apparently comes from those who opposed the reform (Haselswerdt 2017; McCabe 2016). In fact, Fording and Patton (forthcoming) show that public backlash in conservative Medicaid expansion states induced lawmakers to impose new forms of conditionality on Medicaid receipt, such as work requirements, and that such policies are spreading even to states that did not expand Medicaid to begin with.
Policy feedback scholars have also examined whether implementation of the ACA would change attitudes toward the law. Perhaps favorability would rise once people gained insurance through its provisions (Jacobs and Mettler 2011) and once others realized that the worst predictions of the law's detractors did not materialize. At the same time, there were reasons to believe implementation would not change attitudes. Earlier reforms of welfare and Medicare had failed to alter attitudes, among either beneficiaries or the broader public (Morgan and Campbell 2011; Soss and Schram 2007). Partisanship often overwhelms personal experience to begin with as a factor in public preferences, and the highly partisan environment surrounding the ACA may have heightened that effect (Patashnik and Zelizer 2013). In addition, the law's complex and often hidden design elements might undermine the possibilities for attitudinal change (Chattopadhyay 2018, 2019).
Prior to implementation, symbolic factors such as partisanship, racial attitudes, and government trust dominated ACA attitudes. During the 2009–10 debate, party identification was more important in determining support or opposition to the reform than were demographic factors suggesting a material stake, including age, income, or race (Kriner and Reeves 2014). Racial attitudes were also highly correlated with ACA support (Henderson and Hillygus 2011; Tesler 2012). Panel data from the 2010–14 preimplementation period similarly showed that Republicans and those with low trust in government were more likely to say that the ACA was increasing their tax burden (Jacobs and Mettler 2016, 2018).
After implementation began, modest evidence of increased support among those benefiting from the law's provisions—an attitudinal policy feedback—emerged. Fewer survey respondents said the law had no effect on health care access (Jacobs and Mettler 2016). The gap between Republicans and Democrats in ACA favorability was smaller among those who gained insurance through the Marketplace than among those with employer-based insurance (McCabe 2016). Between 2010 and 2017, Medicaid expansion made lower-income Americans more favorable toward the ACA, with effects stronger among nonwhites and Democrats (Hopkins and Parish, forthcoming). Those with personal or family experience gaining insurance, using subsidies, or getting prescription drug help as senior citizens were more likely to say the ACA has had a favorable impact on health access (Jacobs and Mettler 2018). Those buying insurance in the Marketplace were more positive toward the ACA than those who remained uninsured, as were those in their early sixties whose insurance premiums were newly capped by the law (Hobbs and Hopkins 2019). At the same time, those purchasing insurance on the exchanges who experienced local premium spikes became less favorable toward the ACA (Hobbs and Hopkins 2019).
The political environment mattered—pro-ACA announcements by governors in one state increased public support for the law in nearby states (Pacheco and Maltby 2017)—as did political threat: pooled 2009–17 data showed that after the 2016 election made Republican repeal threats credible, ACA approval was higher (and support for repeal lower) in Medicaid expansion states compared to nonexpansion states, especially among lower-education nonsenior adults (Sances and Clinton 2019).
Thus, evidence both from panel surveys (Jacobs and Mettler 2016) and causal analyses (Hobbs and Hopkins 2019; Sances and Clinton 2019) shows that personal experience altered attitudes toward the ACA. In some instances, the benefits conferred are tangible, visible, and large enough (Citrin and Green 1991) to enhance political participation and to induce protective stances, especially in the face of policy threat. And the ACA has highlighted a previously unrecognized phenomenon, political backlash, useful for explaining greater participation by a law's opponents. That said, in many instances these feedback effects are small in magnitude, contradictory in direction, and in some cases, temporary.
Why Haven't More Feedbacks Emerged?
Scholars have begun to speculate about the modest size of policy feedbacks arising from the ACA. The ACA's hidden design elements—using private insurance to spread coverage in the Marketplace, in the dependent coverage provision, in Medicaid managed care—makes it difficult for recipients to connect their health insurance experience to government activity, undercutting attitudinal or behavioral change (Chattopadhyay 2018, 2019; see also Béland, Rocco, and Waddan 2019). The facts that important policy decisions were devolved to the state level and that some state implementation choices increased public support for the ACA while others decreased it suggest that federalism can undermine policy feedbacks (Pacheco and Maltby 2019). Another possibility is that, while the vociferous public debate over the ACA may have enhanced feedback effects by informing people about the law, it may also have heightened the influence of partisanship and motivated reasoning over personal experience as a factor in attitudes and behaviors around the law (Patashnik and Zelizer 2013). Because the ACA's target populations—low-income or young in many cases—are marginal voters to begin with, the benefits may have been insufficient to push them permanently over the participatory hump, explaining why some of the participation increases have been only temporary (Clinton and Sances 2018). Oberlander and Weaver (2015) argued the ACA has suffered “self-undermining policy feedbacks” for reasons like those above, which concern beneficiaries themselves, as well as additional factors, such as concentrated losses and festering grievances among significant groups like taxpayers and employers, persistent political incentives for key elites to criticize the law's provisions and characterize its beneficiaries as undeserving, and the law's vulnerability to legal challenge.
These observations strike me as correct, but I believe ACA feedback effects have been modest for additional reasons. Some of these factors are specific to the ACA or to health policy. Yet others have larger theoretical and empirical implications for policy feedbacks scholarship.
First, we may be looking at the wrong political acts. Most analyses of behavioral feedbacks have focused on voter registration and turnout, the acts that are the most common and have the best data availability. But voting is driven not just by resources but especially by civic duty (Verba, Schlozman, and Brady 1995), so it may not be the most sensitive instrument for assessing a resource effect arising from a new social policy benefit. The vote is also a blunt instrument with little information-carrying capacity (Verba, Schlozman, and Brady 1995). Political acts that are more policy specific, such as contacting elected officials or protest, may be better measures of policy feedback. Anecdotal evidence suggests that Republicans' ACA repeal attempts spurred protest activity (e.g., Gose and Skocpol 2019), but we lack the data to systematically assess that possibility, hence scholars' emphasis so far on voting.
Second, it may be that the ACA is simply less capable of producing policy feedback effects than are other policies. To Chattopadhyay's points about the ACA's fractured design undermining possible feedback effects, I add the visibility of fellow recipients. That the ACA provides different policy fixes for different types of people getting health insurance from different sources undermines the proximity and visibility that seem to fuel some feedback effects. Consider senior citizens receiving Social Security and Medicare. They can readily recognize one another. They are also numerous, located everywhere, and in some places even live together. But ACA beneficiaries cannot identify each other. Some get health insurance because they are newly eligible for Medicaid, some because they were previously eligible but newly signed up, some because they are newly purchasing private plans on exchanges, with or without government subsidies, and some because they can stay on their parents' private health insurance until age 26. Such individuals are scattered everywhere, concentrated nowhere. There are few mechanisms—either through the programs themselves or through mobilizing organizations, which barely exist—for such individuals to recognize one another and work together. Such organizations are crucial for both asking people to participate and explaining to them the stakes of public debates. Even if mobilizing organizations were to emerge, they would have difficulty identifying potential members. Health politics may be unique because health has a private cast that may undermine public efforts at mobilization and because health concerns often do not align with other forms of identity (Carpenter 2012). On top of those complicating factors, the ACA's fractured policy design heightens informational and mobilizational barriers.
A third and related point is that the ACA may be less likely to create feedback effects because its benefits are too modest and its policy mechanisms too indirect. Consider Social Security: The program's retirement benefits are large enough to solve, mostly, the problem of senior poverty (although pockets of poverty remain, especially among women, ethnic and racial minorities, and older seniors). They also address the underlying problem of retirement security directly, with cash delivered to the household budget. Or consider food stamps: They vary in size with income, and even the maximum benefit was never intended to cover 100% of recipients' food purchases. The mechanism, however, is direct: an EBT card that is used to purchase food, alleviating hunger.
In contrast, consider health insurance: What people really want is health security—access to affordable health insurance and medical care—but the ACA falls short of its promise for many, solving so little of the underlying problem.1 One shortcoming is the magnitude of its benefits: the subsidies on the exchanges are too small to make insurance affordable for many, and the law has failed to stem rising deductibles and underinsurance in the employer market. Inadequate health insurance means continued health insecurity for many Americans. The other problem is the indirectness of the mechanism, a phenomenon inherent to health insurance, not just the ACA. Compared to a direct mechanism like injecting cash or cash equivalents into the household budget, health insurance provides financial security more indirectly. Ideally health insurance is complete enough not to cripple households' budgets with unaffordable out-of-pocket costs; ideally health insurance provides sufficient access to health care to support work or whatever activity makes for the household budget. But both of these linkages between health insurance and financial security are probabilistic. The more a program falls short of its goals, whether through the inadequacy of the benefit or the indirectness of its mechanism, the less its feedback-generating capacity.
Fourth, it is worth underlining the effects of heightened partisanship and polarization on the ACA's ability to generate feedback effects (Oberlander and Weaver 2015; Patashnik and Zelizer 2013). Ordinary people, for whom politics and policy are a sideshow in life, need help interpreting political events and policy changes. High levels of partisanship and polarization mean the public has been continually bombarded with conflicting messages on the ACA from elites, including a highly critical stream from the political Right. The benefits of the ACA are thus disputed in the public realm, likely undermining the development of support among beneficiaries, whose personal experience is tarnished, and among nonbeneficiaries, who, lacking personal experience, were suspicious to begin with. It is easier for a reform to become the new normal when elite messages are more consistently supportive.
Fifth, and most important for policy feedbacks theory, is the role of time. Feedback effects may be modest thus far because of the relatively short time the ACA has been in existence. If the linkage between the ACA and increased political activity is the resource of improved health then it could take years for improved health to materialize, as the Oregon Health Plan experience suggests (Baicker et al. 2013).
More profoundly, thinking about time raises some serious questions about the use of causal models—or at least the types of causal models we have been using—to detect feedback effects. The ACA has facilitated the causal analysis of policy feedback effects, which I applaud. But many causal models presume immediate effects: recipients get a new benefit, and now they instantly have new interests, or they instantly receive and internalize “interpretive” messages about their worth as citizens that should influence their attitudes and behaviors in the short term. That is, many extant causal models presume a type of flip-the-switch effect. But is that what happens? Do we really think that people adopt a new set of interests, or recognize a new set of citizenship messages, that quickly? Some factors can change in the short term, such as the resource factors feeding into political participation. But other factors underlying political participation and especially political attitudes are less conducive to immediate change. Something else to ponder: we know that aggregate opinion change typically comes about through cohort replacement, not individual-level change, which suggests the flip-the-switch effect may be rare indeed.2
Moreover, the short time frame of many causal models measures only the “feed,” not the “back,” and hence does not encapsulate a complete expression of policy feedbacks theory. If we think that feedbacks happen in a cyclical, iterative fashion, with policies changing attitudes and behaviors, which in turn reverberate through the political system to produce new policies, then causal models may capture only one-half of one iteration of the cycle, too short term to capture the full phenomenon. In addition, some mechanisms connecting a policy with political behaviors and attitudes may take longer to materialize than others. Consider my analysis of the role of Social Security in boosting senior citizens' political participation, which showed that these effects grew iteratively over decades (Campbell 2003). It took time for the resource effect to grow, as more seniors were eligible for retirement benefits and as they grew more generous. It took time for mobilizing entities to focus on seniors as a political constituency worthy of outreach efforts. It took even more time for seniors' outsized sense of political efficacy to develop, as they observed their bursts of participation aimed at protecting the program recognized and rewarded by politicians. Clearly causal models deserve a place in policy feedbacks work—we must know that an effect we observe is truly due to the policy itself and not competing factors—but we must also recognize the limits of causal models and utilize them in conjunction with other methods of assessing the relationships between public policies and public behaviors and attitudes over time.
Future Research in Mass Policy Feedbacks
The ACA has added an important case of policy feedback, providing a new example of threat as a motivator and revealing a new phenomenon—backlash—to look for in other policy areas. We see yet again the toll that obscured, complicated policy designs (especially privatized designs) take on individuals' ability to recognize the government role in their public policy experiences and to defend it. And we have the welcome extension of causal models to this empirical area, where they have been sorely needed.
But the ACA case also shows that scholars of policy feedback still have work to do. We need to think more deeply about what causal models imply about the timing and nature of policy feedbacks. We need more data, both survey data and qualitative data, that follow individuals' program experiences and evolving thinking and behavior and that follows mobilizing organizations and their strategies. As always, we need more analyses that show both that policies influence attitudes and behaviors and that, in turn, those altered attitudes and behaviors reshape the political environment and influence subsequent rounds of policymaking. In this regard I commend the work of Fording and Patton (forthcoming), which shows that Medicaid expansion by Republican governors angered Republican voters in their states (the feed), which induced those governors to impose work requirements to retroactively limit Medicaid expansion (the back). And “the back” has continued: states newly adopting Medicaid expansion have decided to impose work requirements from the outset, and even states that never adopted Medicaid expansion have decided to impose work requirements on their existing programs. This scholarship confirms the value of the policy feedback approach, which I hope the ACA—the most important social policy change in a generation—will continue to foster.
I thank Jon Oberlander, Julianna Pacheco, and Jacqueline Chattopadhyay for their very helpful comments on an earlier version of the manuscript.
Thanks to Jon Oberlander for his comments on this paragraph.
Many thanks to Julianna Pacheco for this observation.