Over the past several decades a growing body of research has made headway in extending E. E. Schattschneider’s (Reference Schattschneider1935) insight that “a new policy creates a new politics” by treating policy not only as an outcome of politics, but also as an independent variable that itself influences the political behavior of mass publics.Footnote 1 Scholars have devised research to isolate the policy feedback effects generated by assistance for the poor, Social Security, G.I. Bill education and training benefits, and other policies.Footnote 2 Recent studies suggest, however, that new policies do not always transform political behavior, even in instances when lawmakers themselves purposefully aimed to engineer such outcomes.Footnote 3 These variations in outcomes present a challenge to our understanding of policy feedback effects.
In order to determine why a policy may generate some types of feedback effects and not others, we need to ask questions such as do political conditions affect whether specific types of feedback effects occur, for instance by offsetting or muting some outcomes but not others? What sorts of motivations trigger feedback effects? Is self-interest the sole driver or might the collective considerations associated with sociotropism play a part as well? How does the interplay between policy benefits and burdens influence feedback effects? In short, the study of “policy feedback” and its paradigmatic shuffling of causal ordering needs to be strengthened to better specify the nature of policy effects and potentially countervailing forces.
Fortunately, the analytic concepts and theories for such inquiry have been developed through the study of public opinion and political psychology. But policy feedback research has not yet taken full advantage of this research, in part because it has focused primarily on political participation.Footnote 4 In addition, most prior studies have relied primarily on cross-sectional surveys. Although policy feedback research seeks to study individual-level changes in opinion and behavior, it has tended to use data collected in a given moment.Footnote 5 Even time series data derived from several cross-sectional studies are inadequate because they do not permit scholars to control for endogeneity or selection bias, the possibility that policy beneficiaries are different from nonbeneficiaries in some manner that is either unknown or for which we lack appropriate controls. The investigation of policy feedback can be advanced, however, by using panel data that are collected over time and encompass a wide array of indicators to test competing theoretical expectations.
We develop here theoretical expectations of policy feedback by drawing particularly on important research on public opinion and political psychology. First, we focus on the political factors that influence the occurrence of policy feedback, considering for example how elevated partisanship and high levels of distrust in government may blunt the effects of new policy. Our second expectation is that policy designs may generate appreciation for personal or collective effects of policy. Features that spotlight concrete benefits of tangible importance to individuals may activate awareness of personal policy effects and motivate self-interest. Conversely, features that highlight broad collective aims may trigger sociotropic motivations, such as expectations of payoffs to the nation and the vulnerable. Third, we expect that such feedback effects emanate not only from program benefits, on which scholars have focused to date, but also from the burdens that policies impose, such as taxes and restrictions on behavior.
We test our expectations using unique data to study change over time in how individuals experience and react to policy change. We conduct a panel study of public attitudes about the Affordable Care Act (ACA) from its enactment in 2010 through 2014. This study includes indicators of individuals’ personal experiences with ACA benefits that are designed to track change over time. We find that policy effects form in response to the interplay between policy design, political conditions, and motivations, as well as benefits and burdens.
The next section draws on public opinion and political psychology research as well as recent feedback studies to specify a theoretical framework. The second section discusses our panel data and our analysis of policy effects. The conclusion outlines new directions in the study of policy feedback.
Strengthening Theories of Policy Feedback
Strengthening understanding of policy feedback, Paul Pierson argues, depends on specifying “how policies matter and under what conditions.”Footnote 6 Progress toward that goal has been made through increasingly rigorous case studies of policy effectsFootnote 7 and conceptual development and critiques.Footnote 8
Policy feedback scholarship has been particularly attentive to policy design and the resource and interpretative effects it generates. Policies convey resource effects to mass publics by providing specific benefits in the form of goods or services or by imposing burdens through extraction of material costs or regulations on behavior. These resource effects may exert a material impact or alter incentives. Interpretive or cognitive effects may be transmitted directly through the benefits and burdens or through the policy’s design features, including rules and procedures governing policies, their degree of visibility, and their delivery mechanisms. These aspects of policies may convey information and a means of interpreting political and social phenomena.Footnote 9
While a number of earlier studies emphasize the impact of policy design on feedback effects, recent research questions whether the frequency of such effects may be over-estimated.Footnote 10 Kimberly Morgan and Andrea Campbell’s panel study of the Medicare Modernization Act of 2003 (MMA) found, for example, that the law’s new prescription drug benefit failed to increase seniors’ support for the program because it offered a limited benefit that became known for the “donut hole” in its coverage.Footnote 11 Policy visibility also determines whether or not policy feedback develops. Suzanne Mettler (Reference Mettler2011) shows, for instance, that generous tax breaks for home mortgages obscure government’s role as a provider of benefits and depresses awareness of its effects.Footnote 12
Developing a research agenda to explain the incidence and mechanisms of policy feedback requires greater attention to three analytical issues. First, do political conditions influence whether or not feedback effects occur, and if so, which types matter? Second, how does policy feedback affect motivations to pursue self-interest or broader collective outcomes? Third, what are the effects of policy burdens or costs as distinct from policy benefits, which are the typical focus of policy feedback research?
We investigate these critical research questions in the context of the ACA’s implementation. In particular, we examine policy feedback effects through two stages, as indicated in figure 1. First, we explore how the ACA’s feedback effects influence individuals’ assessments of policy benefits and costs. Second, we examine how each of those factors, in turn, influences how individuals evaluate the ACA overall. Throughout these two stages, we study the impact of political circumstances, motivations, and policy benefits and burdens on feedback effects.
Figure 1 Policy feedback for mass publics: how policy affects public opinion
How Politics Competes with Policy Feedback Effects
In a recent review of policy feedback research, Eric Patashnik and Julian Zelizer challenge scholars to treat “the capacity of public policies to remake politics [a]s contingent, conditional, and contested.”.Footnote 13 Fortunately, a long tradition in public opinion research lays the groundwork for understanding how political circumstances may influence individual attitudes and evaluations of politics and policy.
Scholars know, for example, that political party identification acts as an extraordinarily powerful influence on many Americans, anchoring their choices of candidates and their attitudes toward salient policies including economic issues such as health reform and taxes.Footnote 14 In the case of highly politicized policies, partisanship may act as a “framing device” that activates “internal structures of the mind” and filters resource effects and interpretative effects. Of particular importance to feedback research, partisanship may influence the receptivity of citizens to policy effects (such as gaining insurance coverage) by blocking or muting them.Footnote 15 Partisanship may, for instance, prompt strong partisans to dismiss or adopt a critical attitude toward new legislation identified with the rival political party (e.g., “Obamacare”).
In the case of the Affordable Care Act, the partisan divide among lawmakers as it journeyed through Congress in 2009 and 2010 intensified a parallel divide in the mass public. After enactment, President Obama and his Democratic allies touted new benefits and warned of GOP sabotage, while Republican lawmakers and their backers consistently opposed the law and primed voters to focus on the burdens of its taxes and the risks of its programs. These dueling communications among political elites framed the ACA in ways that motivated individuals to cue on partisan affiliations and, in particular, triggered Republicans to become stalwart critics. If partisan polarization is as “strikingly enduring and difficult to shake” as researchers report, it may override policy feedback effects for years to come.Footnote 16
Political sophistication—specifically, political knowledge and education—may also influence the receptivity of individuals to “policy learning” and awareness of policy effects. One possibility is that individuals with high levels of political knowledge and education will be better equipped to learn from new policies, detect policy effects that benefit them and others, and change their earlier views.Footnote 17 Another possibility is that more sophisticated citizens will select information that confirms their pre-existing views and resist learning about new policies that challenges those beliefs.Footnote 18
In addition, distrust of government and the strategies of interest groups and dueling politicians to activate it can be a potent tool to short-circuit policy effects. A number of studies indicate that spotlighting the costs and risks of policy proposals can activate perceptions of risk and tap a deep well of distrust in government.Footnote 19 The contentious fight over the ACA, for instance, may prime political distrust of Washington not only among conservatives but also among groups such as independents, low-income people, and others who have little knowledge of health reform and ample unpleasant experiences with government.
The political institutional context of policy experiences may also influence whether or not individuals are receptive to feedback effects. Arrangements for policy delivery that showcase the government’s role are more likely to make citizens aware of it, whereas those that channel public authority through private organizations or the tax code—or what Suzanne Mettler identifies as the “submerged state”—may leave them believing that only the market is at work.Footnote 20 In turn, such distinctions may shape citizens’ views about the appropriate realm and scope of government activity. Those who possess private health insurance, for instance, may be less supportive of health reform than those who benefit from public programs. This is quite pertinent as the ACA largely left undisturbed individuals who already had employer-provided insurance; they remained free of direct, visible interventions that would focus their attention on the benefits of health reform. Meanwhile, individuals already included within government programs (especially Medicare and Medicaid) were accustomed to public interventions and may therefore be primed to recognize and approve of the ACA’s benefits.Footnote 21
A distinguished tradition of research on public opinion and policy feedback effects demonstrates that public evaluations are affected by the relative specificity and visibility of government programs. Free and Cantril, and Page and Jacobs revealed that when Americans focus on specific government social welfare programs, they tend to be appreciative and supportive, as “operational liberals”; by contrast, when their attention is drawn to abstractions, philosophical conservatism and distrust of government prevail.Footnote 22 This research points to the importance of policy design: evaluations of specific programs tend to spotlight the tangible impacts of government and invite operational liberalism; assessments of government in the abstract tend to elicit philosophical conservativism. In the case of public evaluations of the ACA, the potentially dueling influences of policy feedback and political circumstances may be affected by whether the targets of evaluation are tangible new health programs or references to the vague monikers “Obamacare” or “health care reform.” We expect significant feedback when the attention of individuals is focused on the tangible effects of specific policies on their own lives, muting the impact of political conditions. By contrast, when “Obamacare” or “health care reform” are the target of evaluation, we expect partisanship and political distrust to exert significant influence in place of feedback effects.
In short, we explore variation in the occurrence of policy feedback effects by examining the ACA’s impact on public opinion. The circumstances of the political context—partisanship, trust in government, political sophistication, and prior policy experiences—may disrupt or override Americans’ experiences of the law’s new benefits and its feedback effects.
How Policies Trigger Motivations
Research on political behavior and public opinion has had much to say about individual motivation. Yet, to date, feedback research has neglected to provide explicit exploration of the range of motivations that policy effects might trigger and the conditions under which they might arise. How do policy design and delivery activate self-interest and under what conditions? And, if personal gain is not sufficient, what other considerations might trigger feedback?
That self-interest would drive individual evaluations of policy effects is hardly surprising. The assumption of individualistic motivations is prevalent in studies of voting that focus on “pocketbook” calculations and spatial modelling of voting as well as in rational choice analysis of institutions.Footnote 23
Several policy feedback studies imply that self-interest compels individuals to take political action to protect programs that benefit them, whether as Social Security beneficiaries or parents of school-aged children who attempt to influence education policy.Footnote 24 Policy design is particularly significant: prior research implies that individuals are primed to focus on their personal stakes when policy design accentuates benefits. Indeed, advocates for Social Security initiated the distribution of personal statements during the 1990s to reveal the program’s future financial payoffs to Americans; such information improved individuals’ understanding of the program’s tangible benefits and boosted their confidence in and support for it.Footnote 25
Although the social sciences tend to concentrate on self-interest, research has repeatedly found evidence that individuals also weigh broader community considerations, exercising what social psychologists refer to as “sociotropism.” “Public regarding” orientations among voters drive “sociotropic” evaluations of politicians and policies.Footnote 26 On the question of whether self-interest or sociotropism dominates, a diverse body of research indicates that neither motivation is inherent but rather varies depending on policy design, relative elite conflict, and political factors that amplify or submerge its features.Footnote 27
Policy designs and their definition of target groups may prime individual motivation: some policies (or particular features of policies) may activate self-interest and others may prompt consideration of the value to the broader community. Social security successfully generates the latter, as evidenced by the fact that younger Americans are as supportive of it as seniors.Footnote 28 An important study of policy feedback by Joe Soss and Sanford Schram implicitly distinguishes between self-interested motivations, understood as “the experience of public policy as a visible and directly consequential factor in one’s life,” and sociotropic motivations—namely, the valuing of policies by individuals who are not direct beneficiaries because of “what they affirm about ‘us’ . . . . as expressions of group values [about] who we are, what we stand for, and what we expect of one another.”Footnote 29 Which motivation individuals adopt, Soss and Schram argue, is “highly contingent on [policy] visibility and proximity for mass publics.”Footnote 30
The design of the ACA may trigger self-interest when it spotlights immediate, tangible benefits for targeted populations and increases the probability that individuals will experience them directly and be able to detect the stakes for themselves or their family. The ACA established salient provisions for seniors on Medicare (new prescription drug benefits) and for low- and middle-income Americans up to 400 percent of the federal poverty line, who became eligible for subsidies to purchase private health insurance. These policy features may generate self-interested reactions because they distribute, to use Soss and Schram’s phrasing, “visible and directly consequential” benefits, which create disparate and concentrated stakes across distinct demographic groups. For example, higher income individuals may be motivated to become especially opposed to the higher costs of new benefits, as the most affluent three percent pay more in Medicare taxes and well-established businesses pay new fees. Less affluent Americans, by contrast, may be particularly favorable to the insurance subsidies and new prescription drug coverage because they help them to gain access and afford coverage. The ACA’s discrete stakes across income levels may be replicated for other demographic groups: women stand to benefit more than men from concentrated and salient new preventative services (such as free mammograms) and regulatory protections against insurers dropping their coverage if they are diagnosed with breast cancer or other afflictions; people of color compared to whites may be particularly welcoming of new coverage given their disproportionately high rates of uninsurance; and seniors may be especially worried about reductions in their existing programs (such as the ratcheting back of the special treatment for “Medicare Advantage,” an alternative to traditional Medicare).Footnote 31
In addition, the ACA’s policy design may activate sociotropism, to the extent that it distributes information and generates symbols and narratives about the value of new benefits to the community as a whole, or to “deserving” populations of people. It may activate broad collective considerations by expanding access to medical care to disadvantaged groups through the expansion of Medicaid for all low-income adults up to 138% of the federal poverty line, or by advancing the national interest in lowering national health care spending, which if left unaddressed puts a drag on wage increases and the international competitiveness of U.S. businesses. The ACA also established new rules for private insurers with broad benefits to the country: insurers are required to cover all individuals without regard to gender or previous illness, they are restricted in how much they can increase premium prices based on age and other characteristics, and they are compelled to devote at least 80% of premiums to medical care and improving health or issue refunds, which totals approximately $3 billion.
Whether or how individuals respond to policy effects does not emanate, however, from a mechanical decoding of interests. If so, we would see stronger support for and enrollment in the ACA among those who are eligible. The triggering of self-interest or sociotropism is conditioned not only by policy design but also by the general political environment. Acrimonious political debate may highlight new benefits or, more likely, emphasize risks associated with higher costs, reduced coverage for some, and increased paperwork.Footnote 32
In short, personal motivations may be triggered by particular policy designs and how they intersect with individuals’ life circumstances. The increased visibility of targeted, tangible benefits may activate self-interest while sociotropic considerations are more likely when information and narratives effectively convey broad considerations of the collective good and the needs of vulnerable populations.
The connection of policy design to motivation is not, however, simply a byproduct of whether or not benefits are distributed; new burdens may themselves be decisive. Research in social psychology and political communications finds that individuals, under certain circumstances, tend to be more sensitive to loss than gain.Footnote 33 The inclination of individuals to focus on their personal self-interest can extend to their sensitivity to risk and the possibility for loss.
The Policy Effects of Benefits and Burdens or Costs
When Schneider and Ingram called attention to how policy design affects democracy, they pointed out that policies vary in their allocation of not only benefits (subsidies, tax breaks, and advantageous rules), but also burdens (costly taxes, restrictive rules, and punishments).Footnote 34 To date, however, empirical analysis of feedback effects has concentrated primarily on benefits, particularly in the social welfare arena.Footnote 35 Missing in the literature is broader attention to the policy effects of costs, whether in the form of taxation, sanctions, regulations, or rules that restrict liberty and impose obligations. We need to identify the political circumstances that offset or enable the impact of policy burdens on citizens’ attitudes about policies.
Although research on the feedback effects of burdens is limited, other fields of American politics can contribute to building a useful analytic framework for new research, starting with studies of taxation. Scholars find that taxes that most spotlight costs—for example, income, payroll, and property taxes—are especially likely to trigger public resentment. Their design and administration requires recurrent payment of visible and, in the case of income taxes, large lump sums that both fuel intense opposition among the most politically active (the affluent and business community) and heightens the mass public’s vulnerability to being primed by organized groups to perceive burdens and to paying more than one’s “fair share,” even among the less affluent who pay disproportionately less. Although President George W. Bush’s tax cuts disproportionately favored the most affluent one percent (who received 36% of the cuts), majorities mistakenly concluded that they would benefit and they supported the cuts.Footnote 36 Liberal lawmakers confront the mismatch between the salience of taxes versus the opaqueness of benefits: when Americans are asked if they get their “money’s worth” from their tax dollars, it is not surprising that many fail to perceive benefits because it is difficult to identify what their money pays for in concrete terms. In short, whether taxes exert a persistently negative impact on public opinion or not depends on their policy design (specifically their degree of visibility), their immediate and tangible impact on the public, and how these compare to the visibility and impact of program benefits.Footnote 37
Research on public opinion has demonstrated, however, that the public can also be motivated by a pragmatic or “operational” liberalism if policy designs and political elites showcase specific government programs from education to medical care that address citizens’ concrete needs in everyday life. Rigorous studies reveal that the public supports such programs and is willing to pay taxes (even higher taxes) for them. Support for using taxes to pay for core government functions extends to pluralities and, on some issues, even majorities of the affluent and those who identify as Republicans.Footnote 38
Political conflict often focuses on the policy design of burdens and benefits. Conservatives accentuate the direct, immediate, and widespread cost of taxes and structure their proposals for tax reductions to alleviate these “burdens.” By contrast, progressives link taxes to popular programs and frame opposition to taxes as forcing a trade-off that puts valued benefits at risk. Policies that spotlight benefits can generate feedback effects that overshadow or diffuse public perceptions of associated burdens or costs.
In addition, the struggle over policy designs of benefits and burdens has important implications for individual motivation. Profiling burdens tends to accentuate self-interest while accenting benefits may trigger either self-interest or sociotropism. The policy design of burdens and benefits can influence how ordinary citizens assess the “cost-benefit” ratio of particular policies and their supportive or critical assessments of them.
The ACA is a rich testing ground for the feedback effects of burdens in the form of new taxes. Throughout the law’s enactment, conservatives worked to mobilize anti-tax resentment. Republican leaders warned of the burden of a “government takeover” and the conservative Wall Street Journal editorial page warned that taxes “will need to rise precipitously” to pay for the new law.Footnote 39 By contrast, the ACA’s advocates accentuated its benefits and attempted to obscure its taxes. The debate over repealing the ACA has followed similar themes: conservatives flagged its burdens while progressives invited loss aversion in response to the ending of benefits to millions.
Theoretical Expectations: Policy Feedback Effects on Public Opinion
Drawing on the public opinion and political psychology literatures to advance policy feedback research produces three theoretical expectations.
Policy feedback effects are influenced by the interplay between features of policy design and the political context as defined by partisanship, trust in government, political sophistication, and prior policy experiences. These political circumstances may affect whether or not policy affects public opinion.
The combination of policy design and elite framing of policies may trigger divergent motivations. Self-interest is apt to prevail when benefits are tangible and personally experienced as meeting recipients’ concentrated needs (such as making insurance affordable). Sociotropic considerations are likely when policy designs and public promotions focus on broadly-shared problems or collective gains (such as making insurance a right enjoyed by all).
The perception of costs or burdens may be heightened when policy designs spotlight taxes or punitive regulations and obscured when costs are submerged or when tangible benefits are highlighted.
In short, feedback effects are contingent on the interplay between policy design, political conditions, and individual motivations.
Studies of a single case limit scholars’ ability to draw generalizations, but a long analytic tradition uses intensive case studies to develop research questions and preliminary theoretical expectations for future research to investigate.Footnote 40 For instance, Andrea Campbell (Reference Campbell2003) used Social Security to generate questions and theoretical insights for policy feedback research.Footnote 41 The case of the ACA offers analytic advantages because it encompasses a variety of regulations and benefits with distinct policy designs. It supplies tangible benefits to multiple, specific target populations (youth, lower- and middle-income people, and seniors), not just one such population. Moreover, the ACA relies not just on one policy design (as some prior studies have done) but on a range—from visible new benefits (expanded coverage of senior prescription drug benefits) to regulations to prevent insurance companies from refusing coverage to individuals with pre-existing medical conditions. Because we are studying responses by the same panel of individuals, we can measure perceptions and experiences of distinctive ACA features over time and analyze how they affect support for health reform overall.
Data and Methods
We test these expectations by analyzing three waves of a nationally representative panel of individuals who were asked identical questions. Panel data permit us to examine how the ACA, as it unfolds, influenced individual experiences with new programs and evaluations of the law.
The Advantages of Panel Research
The panel study approach to examining public reactions to the ACA’s implementation makes it possible to correct a methodological drawback of most prior research on policy effects—selection bias. Previous empirical studies of policy effects mostly consisted of in-depth case studiesFootnote 42 and comparisons across several policies.Footnote 43 These studies relied on cross-sectional data, making it difficult or impossible to determine whether observed attitudes and behavior actually result from the policy intervention or instead emanate from pre-existing characteristics that are not known or cannot be controlled for in statistical analysis. Only a few feedback studies to date have utilized panel data.Footnote 44
Our panel approach collected in-depth data from the same group of individuals over time soon after passage of ACA in 2010 and continued through the implementation of key provisions in 2012 and 2014. The first wave in fall 2010 surveyed 1,200 adults; this included 1,000 in a national random sample plus an oversample of 200 individuals who were between the ages of 18 and 64 and living in low-income households with incomes under $35,000. We returned to these same 1,200 individuals with the same questionnaire in 2012 after the National Federation of Independent Business v. Sebelius Supreme Court decision, and again in 2014, one year after the health insurance exchanges began and 9 months following the start of the Medicaid expansions. In each case, the interviews were conducted during the election season in September and October, when health reform was consistently likely to receive heightened attention. The Survey Research Institute at Cornell University conducted the 2010 survey using landlines only; Abt SRBI conducted it in 2012 and 2014, using both mobile phone numbers and landlines to contact participants. (Fuller details are available in the online appendix.)
We retained subjects over time through regular communications and incentives. Footnote 45 Seventy-nine percent (828 out of 1,054) of our 2014 panel also participated in the 2012 or 2010 surveys and, of particular importance for our analysis below, 66% (792 individuals) of the 2010 participants cooperated with the 2014 survey. We also used survey weights to match representative demographic targets and to produce nationally representative samples, which allow us to generalize from our panel to the adult population in the United States.
In creating the survey instrument, we drew on extensive research to distinguish between two types of questions: those that measure subjective states of mind and those that gather data on factual experiences. Our question about respondents’ feelings about health reform, for example, was designed to measure their subjective attitudes about favoring or not favoring the new law. By contrast, we also designed questions to gather data about the actual experiences of respondents, which “in theory . . . could be verified.”Footnote 46 Respondents were asked to report whether they were insured or not; we used self-reporting about insurance status in 2010, 2012, and 2014 to track whether members of our panel gained insurance, lost it, or experienced no change. In addition, we designed batteries of questions to gather data on whether specific ACA benefits (including subsidies to buy insurance and new prescription drug benefits for seniors) exerted an “impact . . . on you and your family.” Whether respondents used the tangible new benefits created by the ACA is, at least theoretically, objectively verifiable: it targets a definable experience within a specific time period—the past two years—that respondents are asked to report on. Survey research uses similar types of questions to ask respondents to report on their demographic and economic status, the medical service or government benefits they received, and their impact on them. Policy feedback research has used such measures, as well.Footnote 47
Using the same survey instrument with a stable panel over time diminishes the risk that respondents’ answers are simply a by-product of how a question is framed. While responses to survey instruments at one point in time may be influenced by question wording, changes in responses over time to identically worded questions by the same individuals are more apt to reflect genuine reactions to new experiences.
In sum, our panel data allow us to: (1) examine whether changes in policy evaluations result from partisanship or from new experiences with specific ACA benefits and costs, (2) probe the underlying mechanisms that facilitate policy effects, and (3) trace the effects of new attitudes on general public evaluations of the ACA.
Measures and Variables
We developed dependent and independent variables for our theoretical expectations. Our models focus on explaining the impact of the ACA’s specific programs on individuals’ assessments of benefits and burdens in 2014 compared to 2010, and in turn, on their evaluations of the ACA as a whole.
What we are explaining . We conduct two parallel types of analyses, each consisting of two stages. Our first model examines the determinants of changing individual assessments between 2010 and 2014 of the ACA’s reported impact on “access to health insurance or medical care supported or provided by government” for the respondent and his or her family. This dependent variable is measured on a five-point scale, from “none” to “a great deal.”Footnote 48 We use ordered logistic regression for the analysis because the dependent variable is measured along ordered points.
Next, we examine whether the assessment of ACA benefits, in turn, influences the public’s overall evaluation of the health reform law. In this analysis, this dependent variable is based on respondents’ evaluations of a “major health reform bill enacted in 2010” on a 7-point scale from strongly unfavorable to strongly favorable.Footnote 49 We use structural equation modeling with standardized coefficients and present the results using a path model. We explain overall attitudes about health reform by treating the ACA’s impact on access as a mediating variable as well as by examining its determinants. This approach allows us to expand our investigation of the mechanisms facilitating policy effects; the first stage in our structural equation model examines the drivers of the view that the ACA expanded access and the second stage investigates the influences on overall assessments of the law.
Our second pair of analyses explores the determinants of changing individual assessments of costs or burdens imposed by the ACA, specifically with respect to the law’s taxes. In particular, we examine the determinants of changes between 2010 and 2014 in individual assessments of their taxes: “Do you think that the new health care law enacted in 2010 has increased the taxes that you pay, decreased the taxes that you pay, or has it had not impact on the taxes that you pay?” Respondents who said “increased” or “decreased” were then asked to specify either “a lot” or “little.” We constructed a 3-point scale to measure beliefs about the ACA’s impact on taxes: 1 for “increased taxes a lot,” 2 for “increased taxes a little,” and 3 for “no impact” or “decreased” either “a lot” or “a little.”Footnote 50 Because the dependent variable relies on ordinal categories, we use ordered logistic regression.
Finally, as in the first pair of analyses, we explain individuals’ overall evaluation of the “major health care bill enacted in 2010,” treating tax attitudes as a mediating variable along with the same roster of independent variables. Once again, we use structural equation modeling with standardized coefficients and display our results using a path model.
Explanatory variables. We developed a set of theoretically-significant independent variables to account for public assessments of the ACA’s benefits and costs. In each analysis, we included the value of the dependent variable for the 2010 wave. This allowed us to control for the inertial elements of opinion and focus on the determinants of changed assessments by 2014. This is a unique and critical component of our analysis of policy feedback.
For our analysis of political conditions, we included partisan identity, which was measured along a 7-point scale from 1 for “strong Republican;” 4 for “independent;” and 7 for “strong Democrat.” The points in between indicate individuals who “lean” toward one of the parties or have a “weak” affiliation. We also included political trust and measured it with the longstanding question, “How much of the time do you think you can trust the government in Washington to do what is right: just about always, most of the time, or only some of the time?” We coded the responses in the direction of “most of the time,” indicating greater trust. Finally, the institutional context of policies was assessed using dummy variables to measure the type of health coverage individuals had in 2014: insurance purchased on their own or coverage by a government program (Medicare, Medicaid, or other). Coverage with a “health plan through employer” is the reference category.Footnote 51
These political and policy variables allow us to examine the theoretical expectations we raised earlier. Building on research about the tendency of Americans to be “operational liberals” in appreciating specific policy provisions, we expect that individual assessments of the benefits and costs of the ACA may grow more salutary over time even after controlling for partisanship and trust in government. The public’s tendency to appreciate tangible programs may be enhanced by the institutional context of policy: We anticipate that individuals with explicitly government-provided coverage may be more prone to adopt more positive assessments of benefits and costs than those with employer-provided benefits, who may be less aware of the government’s role.
By contrast, when individuals are asked to provide overall evaluations of “health care reform,” an abstract assessment rather than one tied to specific programs and benefits, they may be inclined toward “philosophical conservatism.” When evaluating the general idea of health care reform, we expect that partisanship and distrust of government may overwhelm the policy feedback effects.
To study policy feedback effects, we measured insurance status in 2014 relative to 2010 to examine whether gaining or retaining insurance improved assessments of the ACA’s benefits. In particular, this independent variable measures whether an individual is “covered by any form of health insurance or health plan.” We created three dummy variables: one for those who were uninsured in 2010 and have become insured by 2014, another for those uninsured in both years, and another for those with coverage in 2010 who lost it by 2014. (The excluded category indicated that they were insured in both years.) In addition to this direct measure of policy experiences, we examine the self-reports by the respondents of the impact on themselves and their families of two specific ACA provisions: “help for seniors to pay for prescription drugs” and “tax credits and other subsidies to help people pay for health insurance,” with each scaled toward greater impact.Footnote 52
We also created independent variables to investigate how motivations may be triggered by policies, potentially weakening or strengthening assessments of benefits and costs. Indicators of self-interest include respondent usage of programs that help seniors with prescription medications and provide subsidies to purchase private insurance for themselves and their families. Further indicators of self-interest include demographic measures of groups that were most expected to be aided by the ACA in terms of gender (the higher value is assigned to female), race and ethnicity (non-white), year of birth (younger), and income (under $35,000).Footnote 53 Conversely, if such groups express preferences that would appear to contradict their own self-interest (e.g., affluent individuals offering positive evaluations of the ACA), that would provide evidence of sociotropism and broader, more collective considerations. We included another indicator of sociotropic motivation: evaluations of the effect of reform on “the United States as a whole,” measured on a 7-point scale from making things “much worse” to “much better.”Footnote 54
As controls, we include two measures of political sophistication. A 5-point scale of political knowledge is based on the number of correct answers respondents gave about general politics and health policy.Footnote 55 The second measure consists of a 5-point scale indicating level of formal education.Footnote 56
Explaining Policy Feedback from Benefits and Costs
Our analysis studies both ACA benefits and costs and finds a similar theme: policy effects result from the interplay of policy design, political conditions, and motivations. When the ACA’s tangible benefits or taxes are highlighted, policy effects withstand adverse political circumstances. By contrast, when the focus shifts to general evaluations of health reform and philosophical attitudes are triggered, political conditions can overwhelm the impact of tangible policies. Moreover, tangible benefits and costs tend to generate self-interested responses to policies; when the public is instead directed to focus on the ACA’s broad commitment to improving the situation for the country, individuals tend to adopt a more sociotropic mindset.
The Policy Effects of ACA Benefits
We begin our analysis by regressing the 2014 dependent variable for ACA benefits—namely, the law’s impact on expanding access to health insurance and medical care—on the insurance status variables that measure change from 2010 to 2014, independent variables for 2014, and the lagged dependent variable for 2010. We present our results in table 1; the coefficients are in odds ratio format, which means that coefficients over 1 indicate a positive effect and those under 1, a negative effect. Column 1 shows the effect of usage of specific ACA provisions and column 2 presents the effect of changed insurance status as the law was implemented. Column 3 tests the impact of insurance type along with the experiences of the ACA provisions.
Table 1 Determinants of ACA impact on access to health insurance or medical care (ordered logistic regression, coefficients in odds ratio format)a
+p<.10, *p<.05, **p<.01. Standard errors are in parentheses. Exact question wording is in the online appendix.
Source: U.S. Public Policy Study, Wave 1-3.
Dependent variable is the perception in 2014 that the ACA had an impact for respondent and family on access to health insurance or medical care supported or provided by government. Coded toward more impact (“a great deal”).
Notes: aCoefficients in odds ratio format; coefficients over 1 indicate a positive effect and coefficients under 1 indicate a negative effect.
b The excluded reference category is insured in 2010 and 2014.
c The excluded reference category is health coverage from employer
Our first observation is that the lagged dependent variable—the public’s assessment of the ACA’s impact on access to health care in 2010—was a significant influence in 2014 in each of the models (odds ratio of 1.24, p<.05 in column 1). This finding means that the probability of considering the ACA as having “a great deal” of impact on access in 2014 was 1.24 times greater among those who held that same view in 2010 than among those who reported it to have a lesser impact. This is not surprising and indicates that public opinion tends to have an inertial or “lock-in” quality: once formed, attitudes are resistant to change. By holding constant attitudes about the ACA’s impact on access in 2010, we are able to study the determinants of changed assessments of access in 2014. This is a unique and critical component of our analysis of policy feedbacks.
The most striking substantive pattern is that policy feedback effects are occurring. Even with potent controls, individuals who experience tangible changes in their insurance coverage or medical care adopted a more salutary view of the law’s impact on access. Several of these policy effects result from self-interested motivations. As expected, we find that those who have experienced personal or familial impacts of the law’s specific features are more likely to report that it has also enhanced their overall access to health coverage by 2014 compared to 2010. Column 1 shows that usage of specific ACA provisions influenced assessments of access to medical care. People who report benefitting more from ACA subsidies (as compared to those who did not report a benefit) are 1.62 times more likely to see the ACA as exerting a greater impact on widening access to health care (an odds ratio of 1.62, p<.01, column 1). The use of subsidies, as reported by individuals on our panel, is consistent with the aggregate pattern of a sharp decline in the uninsured rate, which fell among 19- to 64-year olds from 20% in 2013—when the subsides began—to 13% in 2015.Footnote 57
In addition, individuals who report a personal experience of the ACA’s provision of assistance to seniors for prescription drugs are 1.29 times more likely to assess its impact on access to health care favorably (1.29, p<.05 in Column 2). This feature of the ACA is credited with saving an average of $1,407 annually for 8.2 million affected seniors between 2010, when its implementation began, and 2014.Footnote 58
We continue our investigation of how self-interest operates through policy feedback in column 2, specifically through resource effects. Most strikingly, we find that the experience of gaining health insurance between 2010 and 2014, compared to having coverage in both years, more than doubled the odds that individuals would credit the ACA with a greater impact on widening access (odds ratio of 2.72, p<.05, column 2). Not surprisingly, individuals who remain uninsured are more likely to have shifted in the opposite direction (.15, p<.01, column 2; odds ratios below 1 are negative), though it is important to note that only a small percentage of the population lost coverage during this period.Footnote 59
Sociotropism has only modest effects, most notably in column 2. Individuals who perceive the law as making things “much better” in “the United States as a whole” as compared to “somewhat better” are 1.1 times more likely to credit the ACA with widening access (1.1, p<.10). Beyond this result, though, we find little evidence of sociotropism.
Political circumstances do not override the policy effects of the ACA’s tangible new benefits to widen access. Most impressive, party identification does not wash out the ACA’s policy effects: in fact, partisanship does not register a significant impact in any of the models in table 1. This does not mean, of course, that partisanship is absent, but rather that it was not a notable driver of change in attitudes about access, or put differently, it did not become significantly more important between 2010 and 2014 in shaping such views. Similarly, the feedback effects held while controlling for political trust and political knowledge, and like partisanship, neither of them registered a significant impact.
Finally, in column 3, we find that policy design, specifically the institutional arrangements for policy delivery, highlights government’s role. As expected, receiving health coverage from government (as compared to an employer) has a positive and significant effect on improving evaluations of the ACA’s impact on access in 2014 compared to 2010 (4.12, p<.01). This suggests that individuals who benefit from government coverage are an impressive 4.12 times more likely to assess the ACA to have a greater impact on access than those with employer-provided coverage. The reasons for this might be two-fold. To begin with, those who are accustomed to perceiving government’s role in subsidizing health care may be primed to notice new forms of such provision, whereas those who have employer-provided coverage may be less likely to notice them. In addition, since many of the newly covered individuals gained Medicaid coverage (11 of the 22 million covered by the ACA in 2014), they or their friends and family may have been particularly aware of the ACA’s impact. These policy effects withstood control variables.
To sum up, the ACA is yielding tangible and discernable effects in the lives of Americans and their families, and over time, those personal experiences are in turn changing their assessments of the law’s impact on their own and their families’ access to health coverage. Self-interest and the institutional setting for receiving coverage mediate the ACA’s feedback effects. The extent of such appraisals varies with individuals’ prior coverage status and type of coverage: those who gain coverage or who receive government help are more likely to notice the impact of the ACA. Conversely, for Americans who were already insured prior to the enactment of the law or enjoy employment-based coverage, their attitudes remain unaffected by the law’s implementation in 2014 compared to 2010.
When Political Factors Matter
Now we turn from examining the policy effects of specific ACA benefits (the first stage of our model in figure 1 and the study of operational liberalism) to examining the public’s overall evaluation of health reform, the second stage of our model and the context in which philosophical conservatism may emerge.Footnote 60 We use path analysis to disentangle the complicated set of paths that drive public evaluations of health reform overall. Our analysis relies on structural equation modeling with standardized coefficients.Footnote 61 Attitudes about health reform are modeled as a function of exogenous variables (political conditions and policies delivering self-interested and sociotropic payoffs) and an endogenous variable (assessment that the ACA expanded access). We also control for the 2010 values of both the mediating and dependent variable, including them as lagged variables so that we can consider changing evaluations. We present a path model that highlights the theoretically most important relationships and focus on the “total effects” on the dependent variable.Footnote 62 (A fuller analysis of our SEM results for the direct, indirect, and total effects are available in the online appendix.)
Three main findings emerge from figure 2. First, sociotropic motivation generates the most powerful effect on changes in evaluations of the law. Individuals who perceive the ACA as making things better for “the United States as a whole” have become more supportive of it between 2010 and 2014 (Total Effect, .46, p<.01).
Figure 2 Impact on access and overall evaluations of the health reform bill
Second, however, the ACA’s specific benefits that triggered the public’s self-interest when evaluating the widening of access (refer to table 1) are washed out as influences on evaluations of the health bill (“ACA Support”) in 2014. In particular, subsidies and prescription drug coverage have significant effects on the intervening variable of access but their total effects on overall evaluations of health reform are not statistically significant. Neither does the experience of gaining insurance between 2010 and 2014 exert a meaningful effect on 2014 evaluations of the ACA.
Self-interest arose from negative policy experiences. Individuals who lost their health insurance between 2010 and 2014 and those who remain uninsured became a bit less supportive of the ACA over time (Total Effects of -.05, p<.05 and -.13, p<.01, respectively). There is a negative and marginally significant effect for subsidies.
The power of political context is the third major finding in figure 2. The total effects of trust in government (.10, p<.01) and party identification (.18, p<.01) are highly significant influences on health reform support. Put differently, those who are distrusting of government and who affiliate with the Republican Party evaluate the ACA unfavorably even if its specific provisions have yielded a positive and discernable impact on their access to health care. This is consistent with prior survey research on public reactions to vague or general government programs.Footnote 63
In short, the lodestar of social science research and a foundational assumption of feedback research—self-interest—is largely overwhelmed by powerful political factors. Evaluations of the ACA overall are driven by trust in government and partisanship rather than new benefits that have tangible effects on the health care of themselves and their families.Footnote 64 Although political attitudes exerted powerful effects, the strongest impact emanated from sociotropism: individuals who perceive the ACA’s national contribution (quite apart from the experiences of themselves and their families) grew more supportive of it.
The Policy Effects of ACA Burdens or Costs
What are the feedback effects of burdens imposed by policies? Do new costs and, specifically, the taxes paid by individuals to finance a policy provoke a backlash against it or are the public’s evaluations mitigated by political factors and self-interest and sociotropic considerations? We begin by examining assessments of tax burdens and then we consider the mechanisms driving these appraisals and their consequences for the public’s overall evaluation of the ACA.
A different mix of dynamics influenced assessments of the ACA’s impact on taxes as compared to new benefits and access. In table 2, we continue to find that the lagged dependent variable is statistically significant in all the four models. This permits us to focus on the determinants of changed attitudes after 2010.
Table 2 Determinants of moderated views of the tax impact of health reform, 2010–2014 (ordered logistic regression, coefficients in odds ratio format)a
+p<.10, *p<.05, **p<.01, ***p<.001.
Source: U.S. Public Policy Study, Wave 1-3.
Dependent variable is the perception in 2014 that ACA had decreased the taxes respondent pays or had no impact as opposed to increasing them a little or a lot (coded toward decreased). The question wording is: “Do you think that the new health care law enacted in 2010 has increased the taxes that you pay, decreased the taxes that you pay, or has it had no impact on the taxes that you pay?”
Notes: aCoefficients in odds ratio format; coefficients over 1 indicate a positive effect and coefficients under 1 indicate a negative effect.
b The excluded reference category is insured in 2010 and 2014.
c The excluded reference category is health coverage from employer
In contrast to our analysis of the policy effects of benefits, sociotropism is the dominant motivation driving feedback effects toward taxes. Individuals who perceived the ACA to be having a positive effect on “the United States as a whole” have greater odds of reporting, in 2014 compared to 2010, that the tax impact on themselves in their family declined or had no effect (column 1, odds ratio of 1.22, p<.05, similar results in the other columns). Interestingly, the self-interest measures fail to yield a significant impact on the assessment of the ACA’s tax burden, with one exception: those who experienced subsidies to help pay for insurance have greater odds of viewing their taxes as increasing (column 1, .76, p<.05). We suspect that these individuals, while they appreciate the effect of the subsidies, might be cognizant of the tax penalty they would have incurred if they had failed to sign up for coverage, and resent it.
Political circumstances bear quite different effects on assessments of tax burdens as compared to ACA benefits. In contrast to our examinations of attitudes about access, trust in government and, especially, partisanship do yield consistent significant effects in shaping views about ACA taxes. Of particular note, in column 2, the odds of perceiving that taxes declined or remained the same due to health reform are greater among Democrats (1.17, p<.10) than Republicans; individuals who expressed greater trust in government (1.69, p<.10) also expressed a more moderate appraisals of ACA taxes. Put differently, this means that Republicans have 83% lower odds of moderating their views of ACA taxes over time than do Democrats, and those with less trust in government have 31% lower odds of doing so compared to more trusting individuals.
Political circumstances appear to be overriding some policy feedback effects. In particular, Republicans and those who distrust government became more convinced that the ACA was increasing their tax burden, regardless of their experiences with the law’s policy provisions.
One of the most startling findings in table 2, however, is that partisanship and, especially, Republican warnings about the ACA ratcheting up of taxes did not drown out sociotropism.
Individuals who appreciated the ACA’s effects on the United States discounted the tax burdens that accompany it, after controlling for party identification.
From Policy Burdens to Overall Assessments of Health Reform
We now shift from the first stage of our analysis in figure 1—tracing the sources of public assessments of the ACA’s taxes—to the second stage: the impact of those evaluations on overall support for health reform in general. As in our earlier analysis of ACA support, the results reveal the potent impact of political conditions and the negative effects of self-interest in a politicized climate. At the same time, they also demonstrate the tenacity of sociotropism as a powerful antidote to political conflict.
Once again, we employ structural equation modeling with standardized coefficients and path modeling to disentangle complex relationships. Figure 3 presents theoretically relevant and significant relationships. (Full results are available in the online appendix.)
Figure 3 Effect of ACA on evaluations of tax burdens and health reform bill
Consistent with our analyses of benefits, political conditions emerge as the strongest and most significant effects on overall ACA evaluations. The totals effects of partisan identity (.18, p<.01) and trust in government (.10, p<.01) exert highly significant effects on ACA support in 2014. As we discovered in figure 2, the impact of general political conditions appears to be mirrored by broad concerns about the United States as a whole. The sociotropic measure—the ACA’s effect on the United States—is the strongest influence on increasing ACA support even in a highly politicized climate (.47, p<.01). On the other hand, consistent self-interested concerns about losing insurance coverage (-.05, p<.05) and remaining uninsured between 2010 and 2014 (-.13, p<.01) tend to depress ACA support.
In short, overall evaluations of the ACA are primarily driven not by assessments of tax burdens but by broader considerations about the United States and the political environment. More specifically, individuals who are Republican or who distrust government are less likely to favor the ACA. Those who are motivated by national considerations are more likely to favor it. Neither the law’s taxes nor tangible effects of its programs exert much influence.
New Directions for Policy Effects Research
Over the past quarter-century, policy feedback research has challenged the way scholars think about politics, demonstrating that policies once established may reshape multiple dimensions of the political universe. The first generation of empirical studies to test these claims identified the existence of feedback effects in the case of particular policies, and furthered thinking about how features of policy design may influence the mass public. Yet this body of research did not explain why such effects occur in some instances and not others nor did it account for the particular form they take. Overcoming these challenges requires new types of data to overcome selection bias as well as new empirically-grounded theory.
We contribute to overcoming these hurdles by studying policy effects over time, introducing unique panel data that limits endogeneity and deepens analysis of the underlying mechanisms of policy feedback, and by borrowing from scholarship in the field of public opinion and political psychology. Our primary focus is on stimulating new questions and theorizing about the circumstances under which policy feedback either occurs or is overwhelmed by political factors, the distinct motivations that are triggered by policy design, and the particular features of individual responses to policy burdens as well as benefits. We trace the pathways to policy evaluations, and find that policy feedback effects compete with political factors, depending on the circumstances.
Our analysis identifies four theoretically-oriented questions and findings for future exploration. First, even in the midst of a highly polarized environment that is rampant with distrust in government, some policy feedback does occur. Policy effects are most likely when individuals experience the tangible effects of specific benefits that improve their own circumstances and those of their immediate family, and become motivated by self-interest. The experiences of gaining health coverage, ACA’s subsidies to help people pay for health coverage, and the benefits to seniors to lower what they paid for prescriptions each enhanced individuals’ assessments of the law’s impact on access to health care, regardless of their partisan identity or level of trust in government. This is consistent with prior survey research on operational liberalism.Footnote 65
Second, under certain circumstances, policy feedback effects may be drowned out by the din of politics. This is most common in our analysis when the attention of individuals is drawn to abstract debates about the role and size of government and partisan identity. This became evident in our investigation of the determinants of overall evaluations of the ACA. Even the positive experiences of individuals with the law’s visible benefits were overwhelmed when public attention focus on general philosophical disputes about partisan attitudes and distrust in government.Footnote 66 The only policy feedback effects transmitted in such an environment are negatives ones: individuals who lost health coverage or remained uninsured did, understandably, come to regard the ACA more unfavorably.
Third, policy feedback effects appear to be triggered not only by self-interest but also by sociotropism even under conditions of partisanship and political distrust. Individuals who observed the ACA making a positive difference for the country as a whole—and not necessarily for themselves alone—updated their appraisals of the tax burdens of the law, recognizing by 2014 that it cost them less than they anticipated. Further, their assessments of the law’s impact for their fellow citizens prompted them to adopt a more favorable evaluation of the law over time. Indeed, this sociotropic effect registered as the strongest force at work in shaping assessments of the ACA, indicating that public-mindedness can withstand the onslaught of partisan divisiveness and loss of faith in government.
Fourth, our analysis has implications for the study of political psychology and, specifically, the impact of framing on public opinion.Footnote 67 Prior research on framing focuses on the discrete, quick, and time-bound effects of specific words and phrasings of individual speakers in particular situations. Missing in this account is the impact of institutionalized policy feedback. Established policy routinely conveys messages that chronically access individual attitudes and beliefs.
The debate over repealing the ACA provides a real-world context for our findings of contingent policy effects. The persistent conservative effort to roll back the ACA reflects the enduring presence of partisanship and government distrust, but growing public acceptance of the ACA’s new tangible benefits has posed daunting hurdles to full repeal.
Further research is needed, however. Policy feedback scholars should examine whether the effects we report are limited to the ACA or have broader reach to other policy areas. In addition, our analysis focuses on public opinion; additional research is needed to pursue the policy effects on political behavior such as increased political participation and voting. Moreover, our examination of burdens can be expanded beyond taxes to regulations and other costs of government intervention. Finally, most of our measures here assess resource effects; further research can examine when and how interpretive effects matter.
Policy feedback research has demonstrated that new laws and programs are not simply passive vehicles to absorb the political environment. Rather, new and established policy is a critical environment. The next step is to identify the contingencies and microfoundations of policy feedback effects. Whether and how policy feedbacks impact public opinion and political behavior may have significant consequences for institutional development and the practice of democracy.
Supplementary Material
Appendix A. Panel Study
Appendix B. Structural Equation Models for Figures 1 and 2
• Impact on Access and Overall Evaluations of the Health Reform Bill
• Effect of ACA on Taxes Respondent Pays and Overall Evaluations of the Health Reform Bill
The data used within the manuscript will be released upon completion of the authors’ book project. At that time, the data will be made available on the Perspectives on Politics Dataverse page, available at https://www.hhh.umn.edu/directory/lawrence-jacobs and at http://government.cornell.edu/suzanne-mettler.
