Social Security and Medicare enjoy strong political coalitions within the mass public because middle-class Americans believe they derive benefits from these programs and stand alongside lower-income beneficiaries in defending them from erosion. By pooling data from nine nationally representative surveys, this article examines whether the Affordable Care Act (ACA) is cultivating a similar cross-class constituency. The results show that middle-income Americans are less likely than low-income Americans to say the ACA has helped them personally so far. On the other hand, partisanship conditions the relationship between income and beliefs about benefits likely to be derived from the ACA in the long run. In total, the results suggest that cross-class Democratic optimism about long-run benefits may enable the ACA to reap positive beneficiary feedbacks, but a large and bipartisan cross-class constituency appears unlikely. Drawing on these results, this article also makes theoretical contributions to the policy feedback literature by underscoring the need for research on prospections' power in policy feedbacks and proposing a strategy for researchers, policy makers, and public managers to identify where partisanship intervenes in the standard policy feedback logic model, and thereby to better assess how it fragments and conditions positive feedback effects in target populations.
Health insurance exchanges will be created. . . . And when this exchange is up and running millions of people will get tax breaks to help them afford coverage, which represents the largest middle class tax cut for health care in history.—President Obama, March 23, 2010
Thanks to the law, 20 million more Americans now know the security of health insurance. Twenty million. Twenty million people. . . . Your insurance is better than it was, even if you don't know it.—President Obama, March 3, 2016
All of America has protections it didn't have before.—President Obama, June 25, 2015
Political scientists devote much attention to the factors that enable a policy to make its beneficiaries into a constituency that protects the policy itself (Campbell 2003a; Patashnik and Zelizer 2013; Skocpol 1992). The basic logic model of how a policy cultivates positive feedbacks from recipients is well delineated: resource effects occur as the policy conveys money, goods, or services to beneficiaries, and favorable interpretive effects occur if benefit administration tells recipients that they are valued in the polity (Mettler 2002; Pierson 1993). If these effects are robust, beneficiaries develop political attitudes and behaviors that make the policy hard for reelection-minded policy makers to overtly cut back (Hacker 2004), with path dependence entrenching the policy further over time (Pierson 2000). Social Security and Medicare are touchstone examples of policies surrounded by such positive beneficiary feedbacks (Campbell 2003a, 2003b).
Is the 2010 Patient Protection and Affordable Care Act (ACA) cultivating this type of enthusiasm from its intended beneficiaries? As signed into law, the ACA was slated to make nonseniors with incomes up to 138 percent of the federal poverty level (FPL) eligible for Medicaid (HealthCare.govn.d.-b), to give tax credits to help uninsured Americans with incomes 100–400 percent FPL buy coverage on the ACA's exchanges (HHS 2015a), and to create consumer protections for people of all income levels (Jacobs and Skocpol 2012: 121–35). Its target population could be defined as virtually all nonseniors (Jacobs 2011: 626), and in light of the dramatic expansion of federal efforts to widen health insurance access that the ACA represents, scholars have noted that it “may, over time, spur the formation of organized support . . . from a broad coalition of Americans” (Jacobs 2014: 637).
Yet, the “proportion of Americans who believe they will benefit from the ACA has never risen substantially beyond one-third” (Brodie, Deane, and Cho 2011: 1098), and after its passage, the share of “those seeing any personal benefit actually declined, outpaced by those who perceived reforms to be essentially irrelevant” (Schlesinger 2011: 1013). This ambivalence contrasts with the “tremendous enthusiasm among the elderly” and 95 percent take-up rate that, according to Robert Ball, Social Security commissioner in 1962–73, met Medicare upon implementation in 1966 (Gluck and Reno 2001: 4). Scholars have proposed that the ACA's complicated design and politically polarizing enactment may prevent it from acquiring the type of political bubble wrap that seniors give Social Security and Medicare (Mettler 2011; Oberlander and Weaver 2015; Patashnik and Zelizer 2013). And thus far, the ACA does not appear to be a political third rail. President Trump—who took office in January 2017 with a unified Republican Congress and a vacant US Supreme Court seat—indeed campaigned on promises to swiftly repeal the ACA in the days immediately preceding the November 2016 election (Diamond 2016).
A question important to understanding the ACA's capacity to generate positive feedbacks from its vast target population is whether it is acquiring loyalty from the middle class. Scholars argue that Social Security, Medicare, and even Medicaid have resisted retrenchment not simply because economically diverse target populations are built into their policy designs but because middle-income earners believe they have a stake in these programs (Campbell 2003a: 7, 21; Grogan and Patashnik 2003; Oberlander 2003: 5, 79–81; Olson 2010: 6, 13, 28–29, 224; Skocpol 1995: 7). Indeed, some argue that what fundamentally differentiates well-secured social insurance programs like Medicare and Social Security from public assistance programs that regularly face cuts is that only the former have middle-income beneficiaries who stand alongside lower-income earners in shielding benefits from repeal (Esping-Andersen 1989, 1990; Quadagno 1991). Esping-Andersen (1989) posits that the entire “consolidation of welfare states after World War II came to depend fundamentally on the political alliances of the new middle classes” (29). Quadagno (1991) argues that what makes Social Security fairly immune to retrenchment is that it successfully “engages the middle class in a solidaristic agenda supporting the welfare state” (37).
The fact that the ACA offers benefits of some type to people across the income distribution may give it the potential to achieve what Quadagno (1991: 41) calls “middle-class incorporation” and to thus build what scholars describe as a “cross-class constituency” (e.g., Oberlander 2003: 79–80; Orloff 1993; Skocpol 1991, cited in Chattopadhyay 2015). Jacobs and Skocpol (2012: 131, 126) posit that “the most economically vulnerable people in our country are . . . winners in the reformed health care system,” but so too are “millions of middle-class Americans.” With remarks such as those in the epigraphs, President Obama publicly made clear that the ACA's target population includes the middle class. Yet, policy design does not guarantee that a program will be embraced by those slated to receive its benefits (Oberlander 2003: 67–68; Patashnik and Zelizer 2013: 1072, 1075–79). “Perceptions of material stakes” are, rather, what may determine “whether a fragile policy begins to ‘take’” (Patashnik and Zelizer 2013: 1078, emphasis added; see also Mettler 2011).
This article aims to shed light on the American middle class's perception of its stake in the ACA. Do middle-income earners believe they have personally benefited from the ACA thus far? Do they think they will personally benefit from the ACA in the future? The article investigates these questions by pooling data from nine nationally representative surveys dating February 2012 through April 2015 and comparing middle-income Americans' rates of belief that they and their families have been, and will be, helped by the ACA, to rates of these beliefs among lower-income earners. The results cast light on whether the ACA is acquiring the cross-class constituency its design seems to give it the potential to cultivate—a question of scholarly and political relevance, given the ACA's uncertain future at this writing. Drawing on the results, this article also makes theoretical contributions to the policy feedback literature by underscoring the need for research on the power of prospections to fuel feedbacks and by proposing a strategy by which researchers, policy makers, and public managers can locate the specific points at which partisanship intervenes in the standard feedback logic model and can thereby better unpack how partisanship fragments and conditions the formation of positive feedback effects in target populations.
The Middle Class's Role in Social Policy Entrenchment
The solidity of the coalition around a policy plays a large role in its endurance. If a coalition fragments or fails to consolidate, the policy may face revision (Kingdon 2003; Baumgartner and Jones 1993; Sabatier and Jenkins-Smith 1993). Work to “divide-and-conquer” coalitions is indeed a key tool in policy retrenchment (Pierson 1994: 8). Typically, interest groups, government leaders, and the public can shape policy feedbacks and compose policy coalitions (Pierson 1993, 1994). For the ACA, the health care industry, states, and the affected public are key coalition blocs (Ario and Jacobs 2012: 1855–60; Grogan 2011: 407–8; Jacobs and Ario 2012: 2603, 2606; Jacobs and Skocpol 2012: 182; Mettler 2011: 6, 32, 122; Patashnik and Zelizer 2013: 1079, 1081; Starr 2011; see also Olson 2010). For any policy, fragmentation can happen between blocs (on “producers” and “consumers,” see Pierson 1994). It can also happen within them, for example, states' divergent responses to the ACA (Greer 2011; Haeder and Weimer 2015; Nicholson-Crotty 2012; Rigby 2012; Sparer 2011). In the target population and mass public bloc, race (Gilens 1999; Quadagno 1994; Soss, Fording, and Schram 2011; Wilson 1991; Winter 2006), age (Quadagno 1989; Street and Cossman 2006), and income (Greenstein 1991; Skocpol 1991) are latent differences that politics can activate, foiling citizen constituencies around a social policy.1 This article studies whether citizens' perceptions of the ACA's personal impact fragment by income.
Fragmentation by income merits scrutiny since, by some accounts, a social policy's fate pivots on whether it secures a middle-class following and thus a cross-class constituency. Esping-Andersen (1989: 28, 31) argues that a welfare state's trajectory hinges on the degree to which people of different incomes coalesce behind its policies, proposing that “‘middle-class’ universalism” has at times “protected the welfare state against backlash” (Esping-Andersen 1990: 69). Quadagno (1991: 37) posits that revisions to Social Security from 1969 to 1972 that “incorporated the middle class into the welfare state” are what made this program fairly immune to retrenchment. Orloff (1993: 301, 181) argues that political institutions shape class coalitions but sees these coalitions as independently shaping policy, proposing that “the key proximate factor in the initiation of new social provision for the aged in Britain, Canada, and the United States was the emergence of a cross-class coalition for social spending” between the “working-class” and “upper- and middle-class reformers.”
Middle-income earners are valuable coalition members in part because “time, money, and civic skills” are critical resources for political participation, and both “money and some kinds of civic skills are closely related to SES [socioeconomic status]” (Verba, Schlozman, and Brady 1995: 271, 282). Often, the “vulnerability of programs has turned on” whether those who would cut them “could identify substantial reforms that would not generate a major public outcry” (Pierson 1994: 6). Policy history suggests that the “poor [have] provided a weak and isolated political target, unable to push back as others could” (Soss, Fording, and Schram 2011: 34). By way of contrast, the fact that “Medicare's constituency had a middle-class identity” may be a key reason that, soon after enactment, elected officials once opposed to the policy “vied to be seen as program friends . . . lest they pay a heavy price at election time” (Oberlander 2003: 5). Middle-class attachment to a policy gives it a constituency with participatory resources readily at hand.
Even a long-standing social insurance program's political footing can slip if the middle class disinvests and leaves low-income earners as its only constituents. Oberlander (2003: 50) notes that the “voice of [Medicare] beneficiaries” may be “dulled by the exit of elder participants with higher incomes into private supplemental insurance.” Similarly, while “seniors have been fairly united” around Social Security, proposals that “highlight differential class interests . . . may provide retrenchment proponents with a divide-and-conquer strategy” (Campbell 2003a: 144). Skocpol (1995: 7) argues that “Social Security has expanded and survived because it has middle-class participation and support. If middle-class Americans can be removed . . . it . . . could easily be cut back.”
To be clear, there are limits to the middle class's role in shaping policy trajectories. First, the foregoing arguments about coalitions and middle-class incorporation posit that a policy enjoys strong citizen support when lower- and middle-income earners are jointly vested as constituents (e.g., Esping-Andersen 1989: 28). On their own, middle-income earners appear neither necessary nor sufficient to compose a resilient citizen coalition.2 Second, the degree to which the middle class is pivotal to a policy's trajectory is likely contingent. In a coalition, industry stakeholders and states, and lines of citizen fragmentation besides income, can vary in political importance over time. Third, policy features that may or may not be endogenous to middle-class inclusion in a policy's target population—including funding structures (Pierson 1994; Oberlander 2003), constructions and treatment of beneficiaries (Schneider and Ingram 2005; Soss 1999), and benefit generosity (Campbell 2012; Mettler 2002)—may influence the policy's trajectory in ways that cannot automatically be attributed to middle-class political behaviors and attitudes.
These caveats noted, the literature reviewed here makes a compelling case for studying middle-income earners' views of a policy's personal impact, when the goal is to understand the policy's potential for positive beneficiary feedbacks. As the next section explains, the likelihood of a cross-class coalition merits close study in the case of the ACA, whose identity as a policy—relative to the middle-class incorporation seen in Medicare and Social Security—is far from clear.
The ACA in Social Policy Case Space: Aids and Obstacles to a Cross-Class Constituency
Quadagno (1991: 37) offers a seemingly straightforward measure of a policy's ability to acquire a cross-class following in proposing that “middle-class incorporation occurs when social benefits are based on a ‘social security' model rather than a ‘social assistance’ model”—that is, when policies are “social insurance,” not “public assistance” (e.g., Katz 2013). Esping-Andersen (1989: 24) does likewise in positing that a “social assistance state” exists wherever the “poor rely on the state, and the remainder on the market.” Yet, these proposals give imperfect guidance on the ACA's potential for a cross-class following, since the ACA mixes public and private insurance.3 By some accounts, “the reliance of private players on public funding is so paramount” post-ACA that “public-private distinctions become meaningless” (Grogan 2011: 403; see also Mettler 2011). We get more leverage for evaluating the ACA's potential for a cross-class constituency from observations on how public policy can be a “system of stratification” (Esping-Andersen 1990, chap. 3; see also Quadagno 1991), can shape coalitions by distributing benefits and costs (Mettler and Soss 2004: 61), and can do “boundary shifting” by “blurring” or “brightening” various social divisions (Alba 2005 and Wimmer 2013, cited in Marrow and Joseph 2015; on the ACA specifically, see also Marrow and Joseph 2015). Drawing on these concepts, this section outlines two aspects of the ACA that may foster a cross-class constituency and four aspects that may impede it.
Two Factors Disposing the ACA toward a Cross-Class Constituency
The first factor that may foster a cross-class constituency is that ACA subsidies, and many ACA regulations, appear squarely aimed at middle-income earners—and on balance appear roughly as valuable to the middle class as other ACA benefits are to lower-income earners. ACA subsidies help people with incomes 100–400 percent FPL buy insurance on HealthCare.gov or a state exchange, when they lack other, acceptable coverage. In 2015, a person making $11,670 to $46,680 was subsidy eligible, as was a family of four making $23,850 to $95,400 (Henry J. Kaiser Family Foundation 2014a: 2). The median US household income in 2012 was just over $51,000 (Noss 2013); thus, ACA subsidies reach “well into the middle class” (Jacobs and Ario 2012: 2604). Grogan (2011: 406) in fact describes the ACA as a policy that helps to “prop up the private sector for the middle class.”
ACA consumer protections cover all Americans (Jacobs and Skocpol 2012: 121–35); yet, some may mainly help the middle class. The ACA dependent coverage mandate, which requires that insurers let children stay on parental plans up to age twenty-six, may do little to help young adults from low-income—often uninsured—households (Kaplan 2014). The ACA's guaranteed issue mandate, which requires that insurers offer coverage to nongroup applicants regardless of health but does not limit plan price (Henry J. Kaiser Family Foundation 2012), is likely more valuable to those with more income. The “shared responsibility payment” that the ACA mandates of large employers may curb the ongoing decline in job-based insurance (HealthCare.govn.d.-a; Jacobs 2014: 635), which has particularly hurt “middle-income groups” (Ario and Jacobs 2012: 1859).
These subsidies and protections are resources, which should, per the standard feedback logic model (Mettler 2002; Pierson 1993), nurture resource effects among middle-income earners and thus perceptions of a positive “material stake” (Patashnik and Zelizer 2013: 1078) in the ACA. These effects may be modest in magnitude, since policy may only marginally raise the material welfare of people with sufficient financial means (Campbell 2002; Mettler 2002). But when assessing the ACA's potential for a cross-class constituency, the relevant question is whether ACA benefit distribution gives low- and middle-income earners a roughly equal stake in the aggregate.
It appears likely to do so, particularly due to the uneven execution of the most tangible ACA benefit for lower-income earners as signed into law: Medicaid eligibility for all people below 139 percent FPL. Pre-ACA, eligibility hinged on several factors beyond income (Olson 2010), and Medicaid may have been more stigmatized (Allen et al. 2014; Grogan 2014: 142–43; Olson 2012; but see Grogan and Patashnik 2003). The expansion thus offered low-income earners resources (insurance) and positive messages about government. Yet, National Federation of Independent Business v. Sebelius (567 U.S. 519 (2012)) made the expansion optional, and as of March 2016, nineteen states were declining it (Henry J. Kaiser Family Foundation 2016a). While these states may engage in “backstage cooperation” with other ACA provisions (Grogan 2011: 408–9), in nonexpansion states people with incomes below 100 percent FPL are eligible for neither Medicaid nor subsidies. Nearly 3 million people fall in this “coverage gap” (Garfield and Damico 2016). Low-income adults in nonexpansion states are less likely than those in expansion states to say the ACA is personally helpful (Sommers et al. 2015). Thus, due to cross-state variation in Medicaid expansion, low-income earners' assessments of being helped by the ACA are likely, in the aggregate, to be no greater, and potentially less than, those of middle-income earners.4
The second, related factor that may help the ACA acquire a cross-class constituency is that administration of the Medicaid expansion—where implemented—blurs the line between low- and middle-income ACA benefits. The ACA does track those below 139 percent FPL into public insurance and those above into private insurance, seemingly limiting middle-class interest in the Medicaid component (e.g., Esping-Andersen 1989: 24). Yet, over 35 percent of adults below 200 percent FPL may shift between Medicaid and subsidy eligibility in a six-month period, as may “50 percent, or 28 million” in a year (Sommers and Rosenbaum 2011: 228). This “churn” gives some middle-income earners a stake in this part of the ACA. And, as churn “drive[s] more integration between Medicaid and commercial insurance markets,” we may see “the wall between ‘welfare medicine’ and mainstream care . . . erode” (Ario and Jacobs 2012: 1862; see also Grogan 2011: 403).
Four Factors Operating against a Cross-Class ACA Constituency
Yet, four features sharply differentiate the ACA from the social insurance traits of Medicare and Social Security and may impede a cross-class constituency. The first is that, despite churn, the ACA does track citizens into different benefits at a given point in time (Oberlander and Weaver 2015: 55; Oberlander 2012: 2166, cited in Patashnik and Zelizer 2013: 1079). As noted, tracking is legislated by income at 139 and 100–400 percent FPL. De facto tracking also happens by age. For example, ACA benefits mainly help nonseniors (Jacobs 2011: 626), and seniors fear the ACA may damage Medicare (Bradley and Chen 2014; Campbell 2011: 966; Jacobs and Mettler 2011: 926–27; Kelly 2015; Skocpol 2010: 1291). Tracking also happens based on access to job-based health insurance. Those with adequate job-based coverage do not qualify for ACA subsidies; they mainly benefit from the ACA's fairly intangible consumer protections.5
Such tracking may impede middle-class incorporation. In a social insurance state, “all strata and classes are incorporated under one universal insurance system” even if “benefits are graduated according to . . . earnings” (Esping-Andersen 1989: 26). Social Security is an example of a program with graduated benefits that nonetheless softens income differences among seniors (Campbell 2003a: 7, 145). In contrast, policy designs that “legislat[e] distinct programs for different class and status groups” may “consolidate divisions among wage earners” (Esping-Andersen 1989: 23).
The second factor that may hinder a cross-class constituency is that ACA subsidy administration brightens income differences within America's economically diverse middle class, particularly the line between middle and upper middle incomes. The ACA leaves many people whose incomes exceed 400 percent FPL “in the uncomfortable middle: not poor enough for help, but not rich enough to be indifferent to cost” (Thomas, Abelson, and McGinty 2013). The subsidy cutoff at 400 percent FPL may feel “especially arbitrary to people whose incomes vary from year to year” and to people who are “‘just right over that line’” (Thomas, Abelson, and McGinty 2013).
The third factor that may work against a cross-class ACA constituency is that the ACA's redistributive nature may be more visible than that of Social Security. The ACA is more redistributive than Social Security (Patashnik and Zelizer 2013: 1080), perhaps marking “‘the federal government's biggest attack on economic inequality’” in thirty years (Leonhardt 2010, quoted in Skocpol 2010: 1290). And, Social Security became redistributive gradually, via 1939 and 1950s amendments (Béland 2005: 98, 130, 140). Béland (2005: 98, 108) notes that political elites “never officially acknowledged the shift toward income redistribution initiated in 1939,” instead continuing the “official discourse concerning ‘earned rights’” from a “contributory program.” Even after elites started framing it as an antipoverty program in the 1960s, few in the mass public saw its redistributive character (Béland 2005: 140).
The ACA's redistributive efforts may be far more visible to the middle class for two reasons. First, the ACA's cost distribution may brighten the line between middle- and upper-middle-income earners. Second, and simultaneously, ACA benefits for middle-income earners are “submerged” in an absolute sense and relative to Medicaid (Mettler 2011). Regarding cost distribution, only the top 2–3 percent of earners face most of the ACA's taxes on citizens (Jacobs 2011: 626; Jacobs 2014: 634)—singles who earn $200,000 or couples who earn $250,000 or more annually (Jacobs 2014: 634; Skocpol 2010: 1290). Yet, costs do thereby fall on the highest middle-income earners—“affluent but not super-rich families” (Jacobs and Skocpol 2012: 134).
Further, this cost distribution resembles that of the spurned 1988 Medicare Catastrophic Coverage Act (MCCA). The MCCA imposed an $800 surtax on “only the top 5 percent of upper income elderly” (Street 1993: 439) but was repealed in 1989 after backlash from “many middle income elderly [who] believed that they, too, would be responsible for the full $800” (Street 1993: 439; see also Oberlander 2003: 66–69). Far from fostering a coalition between low- and middle-income earners, the MCCA's work brightening the line between the top 5 and bottom 95 percent of income holders led middle-class seniors to see this policy as one in which the “principles of social insurance and middle-class incorporation were undermined” (Street 1993: 440). Oberlander and Weaver (2015: 55, 49) posit that the ACA's design falls between the MCCA and Medicare, rather than resembling the former but note that the ACA's imposition of “costs on concentrated groups” is a feature that risks generating “self-undermining feedbacks.”
Heightening the chance that the ACA's cost distribution will evoke a middle-class response reminiscent of the MCCA is the fact that the benefits the ACA offers the middle class are less visible than those it gives low-income earners.6 Both ACA consumer protections and subsidies are what Mettler (2011) calls “submerged” (Béland 2010: 579; Jacobs 2014: 637; Patashnik and Zelizer 2013: 1079); thus, middle-income earners may be unaware of them (Campbell 2011: 967; Mettler 2011: 38; National Association of Insurance Commissioners 2012). President Obama's March 3, 2016 remarks in the epigraphs above suggest concern that this is the case. In contrast, Medicaid has a strong identity as a government program (Mettler 2011: 38), meaning that this redistributive (if unevenly implemented) piece of the ACA is quite visible. Mettler (2011: 118) points to this contrast clearly in noting that the ACA “enlarges visible governance . . . by making more Americans eligible for Medicaid” but otherwise “perpetuates the submerged state, channeling . . . customers toward insurance companies.” With submerged policies, “many Americans express disdain for government social spending, incognizant that they themselves benefit” (Mettler 2011: 6).
In principle, most adults report being very (44 percent) or somewhat (31 percent) favorable to ACA efforts to “provide financial help to low and moderate income Americans who don't get insurance through their jobs to help them purchase coverage” (Henry J. Kaiser Family Foundation 2011a). Yet, the MCCA “illustrates both how contingent middle-class support can be and how readily class fragmentation can be activated” (Quadagno 1991: 53), and it is unclear to what degree middle-income earners believe they are among the ACA's beneficiaries. Media portrayals of ACA beneficiaries appear unlikely to foster such an understanding (Chattopadhyay 2015), and in an October 2016 poll 50 percent of respondents said the ACA helps “lower-income people,” versus 19 percent saying it helps themselves and/or their families (Kirzinger, Sugarman, and Brodie 2016). Especially in “low information environments” like that around the ACA (McCabe 2016: 865), submerged benefits may also be quite susceptible to framing efforts aimed at “exacerbating citizens' images of themselves as aggrieved, involuntary financial supporters of benefits for others, never beneficiaries themselves” (Campbell 2011: 968; see also Jacobs and Mettler 2011: 931).
The fourth factor that may avert a cross-class constituency is that both elite and mass attitudes toward the ACA strongly fragment by party. Among elites, partisan conflict is evident in the fact that Congress passed the ACA without one Republican vote (Patashnik and Zelizer 2013: 1079–80) and in the over sixty attempts to repeal the ACA since its enactment (Obama 2016: 530). These events occurred despite the similarity many ACA provisions bear to past Republican proposals (Quadagno 2014). Partisan conflict is also institutionalized in the aforesaid state rejections of the ACA Medicaid expansion, as these decisions correlate highly to state governments' partisan makeup (Jones, Bradley, and Oberlander 2014: 127; Rigby 2012; but see Jacobs and Callaghan 2013). Morone (2016: 828) calls such challenges to the ACA “an extreme instance of partisan politics,” noting that “there is no precedent—in health or social policy—for a conflict that has gone on for years after the president signed major legislation” (837).
Regarding the mass public, a burgeoning literature suggests that the ACA's partisan history may limit its capacity for positive beneficiary feedback effects (Jacobs and Mettler 2016; McCabe 2016; Morone 2016; Nyhan, Reifler, and Ubel 2013; Oberlander and Weaver 2015). Jacobs and Mettler (2016: 919, 921) found strong Democrats 52 percent more likely than Republicans to be “impressed with the ACA's impact” and found that party identification even outstripped trust in government in shaping health reform attitudes. Oberlander and Weaver (2015: 38) note that partisan differences exist “not just in whether the law is viewed favorably, but also (and less plausibly) in self-reported perceptions of whether the ACA has helped or hurt the . . . respondent.” They posit that partisan attitudes are “obscuring self-reinforcing policies and benefits” in the ACA (58)—an assessment that gains support from McCabe's (2016) finding that partisan reasoning shapes ACA views even among people who have used ACA benefits.
In comparison, while neither Social Security nor Medicare had bipartisan support at inception, both had convincing, bipartisan legislative support at final vote (Morone 2016: 834). Describing Medicare's first sixty days, Robert Ball (1966: 475, 477) noted that state agencies “performed their parts well,” that 90 percent of seniors took up the voluntary part of Medicare, and that there was “very little negative reaction on the part of the nearly 15 million persons who had the $3 deducted from their social security, railroad retirement, or civil service annuity checks.” These remarks could reflect efforts to “sell” Medicare in its early days, but Ball and others echoed them decades later (Gluck and Reno 2001: 9, 17, 28). Social Security faced stronger Republican challenge: “For the first fifteen years . . . the program was in a precarious condition . . . under almost constant attack” (Peterson 1999: 33). Yet, this opposition fell away. The US House of Representatives passed a 15 percent benefit increase in 1952 without hearings (Peterson 1999: 34). And while President Eisenhower distanced himself from such expansions (Béland 2005: 128), he tacitly gave Social Security his approval and thus a “‘bipartisan legitimacy’” that the ACA still lacks (Milkis 2012, quoted in Patashnik and Zelizer 2013: 1083).
With these considerations in mind, what are middle-income earners' views of the ACA's personal impact, and how do they compare to those of lower-income earners? A search of the Roper iPOLL database in June 2016 found nine Gallup surveys (not a panel), available for download, that asked national adults two questions apt for the present study.7 Analyses use a data set formed by pooling analytic subsets of these nine surveys. Descriptive statistics for the pooled data set are given in the appendix. Analyses used survey weights provided by the data sources.
The first question of interest examines the ACA's personal impact thus far: “As you may know, a few of the provisions of the healthcare law have already gone into effect. So far, has the new law . . . Helped you and your family, Not had an effect, or has it Hurt you and your family?”8 The second question taps into expectations of the ACA's future personal impacts: “In the long run, how do you think the healthcare law will affect your family's healthcare situation? Will it . . . Make things better, Not make much difference, or will it Make things worse?”9
The conceptual difference between assessments of a policy's impact so far and in the long run is not strictly a matter of time. Sentiments about the long run may reveal whether a program is conceivably “proximate” or instead perpetually “distant” in a respondent's eyes (Soss and Schram 2007; Campbell 2012: 340), in a way that statements about a policy's impacts to date do not.10 For instance, a young person may not say that Social Security has been personally beneficial so far but may expect it to be personally beneficial in the long run—not simply because that person expects to be sixty-five someday but because nonseniors see seniors as “ourselves in a few years”—the “ultimate ingroup” (Winter 2006: 405). People who do not think the ACA has helped them so far may think it could help them later, either as the remainder of its benefits gradually roll out or as one undergoes changes in health, insurance, or job status (e.g., Jecker 1993). Thus, both retrospective and prospective assessments of policy flows merit study.
Responses to one or both questions should signify whether people believe the ACA has given or will give them valued resources—tapping into what Patashnik and Zelizer (2013: 1078) call perceived material stakes. Responses should thus also signal whether the ACA is fostering resource effects among intended beneficiaries—one of the first steps in the chain of events leading to positive feedbacks according to the standard feedback logic model (see Mettler 2002: fig. 1).
To compare low- and middle-income (and upper-income) Americans' responses to these two questions, analyses used a four-point income group measure predefined in the data sets. This measure sorts respondents into those with monthly household pretax incomes <$2000, $2,000–4,999, $5,000–7,499, and ≥$7,500. The lowest-earning group is thus those with annual household incomes below $24,000 before taxes. The lower-middle-income group has pretax annual household income between $24,000 and approximately $59,999. The upper-middle-income group has pretax income between $60,000 and approximately $89,999. The highest-earning group comprises those with pretax annual household incomes of $90,000 and above. As noted, the median household income in the United States in 2012 was just over $51,000 (Noss 2013). Thus, it indeed makes sense to consider the second and third income categories as middle income.
As noted, incomes of 100–400 percent FPL in 2015 translated into $11,670–46,680 for an individual and $23,850–95,400 for a family of four (Henry J. Kaiser Family Foundation 2014a: 2). Thus, some single householders in the two lower-earning groups would qualify for ACA subsidies, as would families of four in, essentially, the middle two groups and some in the top earning group. In 2015, 138 percent FPL was approximately $16,243 for an individual, and $33,465 for a family of four (HHS 2015b). Thus, the bottom two income groups should contain respondents eligible for ACA Medicaid.
Multivariate analyses controlled for respondent partisan identification, ideology, race, age, sex, marital status, and whether the respondent has children. As noted, ACA views sharply differ by party. Republican identification and leaning relate inversely to several measures of ACA support in the public (Knoll and Shewmaker 2015: 98; Lewis, Dowe, and Franklin 2013: 143; see also Henderson and Hillygus 2011: 952) and to beliefs that the ACA has been personally helpful (Sommers et al. 2015: 1016; McCabe 2016: 866). Ideological conservatism also relates inversely to ACA support (Knoll and Shewmaker 2015: 98; see also Henderson and Hillygus 2011: 952).
It is relevant to control for race given evidence that white Americans are less likely than African Americans to believe that the ACA will improve their own medical care (Lewis, Dowe, and Franklin 2013: 143; see also Sommers et al. 2015: 1016). It is relevant to control for age—particularly to run the analyses with and without seniors—since, as noted, seniors worry that the ACA may undermine Medicare (Bradley and Chen 2014; Campbell 2011: 966; Jacobs and Mettler 2011: 926–27; Kelly 2015; Skocpol 2010: 1291) and “have been consistently less likely to see it as beneficial to them” (Brodie, Deane, and Cho 2011: 1098). Scholarship has not uncovered consistent differences in ACA views among men and women, but it is relevant to control for sex since some ACA provisions may particularly benefit women (Jacobs 2014: 635).
Existing research also has not highlighted marital status or the presence of children as salient lines of division in attitudes toward the ACA, but these traits signal the size of the respondent's household and thus bear on whether a respondent of a given income may qualify for either ACA tax credits or Medicaid. An uninsured respondent who is part of a family of four would qualify for an ACA subsidy even with a household income of over $95,000, while a single householder with this income would not. Further, marital status and children may impact a person's views of the ACA's personal relevance. Being married may give a person access to job-based insurance from a spouse that an unmarried person does not have (see, e.g., Bradley, Neumark, and Motika 2012), making ACA provisions to expand insurance access less valuable to married respondents. Respondents may have self-interest in the ACA due to either children's presence or absence. Parents may value the dependent coverage mandate, while low-income, childless adults may value their new eligibility for Medicaid (Henry J. Kaiser Family Foundation 2013a: 1).
It is not possible to control for all factors that may be relevant to individuals' perceptions of ACA benefits. Only one of the nine surveys queried respondent insurance status, so insurance status is not controlled for even though the uninsured have shown “strong approval for the law and interpreted its impact in . . . personal terms” (Lewis, Dowe, and Franklin 2013: 142). The ACA's ban on insurance denials due to preexisting conditions makes the law valuable to those in poor health (Jacobs 2014: 633), and concerns with personal medical costs impact health reform attitudes (Henderson and Hillygus 2011). However, it is not possible to control for either health status or concern over health expenses. The age group measure, which is controlled for (coded 1 for 18–29, 2 for 30–49, 3 for 50–64, and 4 for people 65 or more years of age), may at least signal the probability of health-related concerns. Finally, support for universal health care and for the ACA decreases as racial resentment increases (Henderson and Hillygus 2011; Knoll and Shewmaker 2015: 98; see also Tesler 2012), but the surveys studied do not offer a means to measure racial resentment.
Multivariate analyses clustered standard errors by state. Analyses included fixed effects for the surveys from which the pooled data come, thus controlling for cross-survey differences and for the fact that the surveys occurred at various points during the ACA's segmented implementation (see HHS 2015a). The key measure of interest is the four-point income variable described above. Analyses used an indicator for each income category.
Figure 1 reports the share of respondents in each income group who say the ACA has helped them so far (vs. hurt or had no impact) among people under age sixty-five. Results were similar among respondents of all ages. One takeaway is that a minority in any group feels helped by the ACA to date; thus, the high rates of aggregate ambivalence about the ACA's personal impact noted by others (Mettler 2011: 89, 90, 110; Schlesinger 2011: 1013; Sommers et al. 2015: 1013) are not masking strikingly high rates of enthusiasm by one income subgroup. The other takeaway, however, is that there is variation by income in these views. In fact, the relationship between income group and a belief that the ACA has been personally helpful thus far is U-shaped. While just under 22 percent of nonsenior respondents in the lowest income group say that the ACA has helped, only 13.7 percent of nonseniors in the lower-middle category believe so, as do only 10.4 percent of nonseniors in the upper-middle category. In the top income category, 13.7 percent of nonseniors say that the ACA has helped them/their families.
Figure 2 gives the share of nonsenior respondents in each group who expect that, in the long run, the ACA will make their family's health care situation better. (Again, results for people of all ages are similar.) The U-shape pattern is more shallow here. Just over 31 percent of nonseniors in the lowest income group believe the ACA will improve their situation in the long run, compared to just under 23 percent of lower-middle-income nonseniors, just under 21 percent of upper-middle income, and just over 22 percent of high-income nonseniors.
Do these patterns hold in multivariate analyses? I describe views of the ACA's personal impact to date first, and views of its long-run personal impacts second. Discussions of partisanship first focus on Democrats and Republicans and turn to Independents thereafter.
Democrats versus Republicans
Table 1 gives the results of modeling the probability that a respondent says that the ACA has helped his or her family thus far as a function of income category, controlling for the measures noted above. Looking at the first set of results in table 1, Americans with pretax monthly household incomes of $2,000–4,999, $5,000–7,499, and ≥$7,500 are each significantly less likely than those with incomes less than $2,000 to believe the ACA has helped them and their families so far. The results are similar when analyses are restricted to nonseniors. Middle-income earners appear less likely than people in the lowest-earning group to feel the ACA has been personally helpful to date. Thus, multivariate analyses suggest a U-shape relationship between income and perceptions of the ACA's personal impact thus far, similar to that shown in figure 1.
The models that interact income with partisanship shown in table 1 give insight into if and how this result differs for Republicans, Democrats, and Independents. Figure 3 illustrates the key takeaways from the fourth set of results regarding views of the ACA's personal impact thus far among nonseniors. This figure is for hypothetical non-Hispanic white, married men with children, of median ideology, age 30–49. The first takeaway is that Democrats/Democrat leaners are more likely than Republicans/Republican leaners to believe the ACA has been personally beneficial so far; the gap between Democrats and Republicans is statistically significant at each of the four income levels.11
The second, arguably more interesting, takeaway from figure 3 is that nonsenior Democrats/Democrat leaners are part of the reason for the U-shape relationship between income and beliefs that the ACA has been personally beneficial thus far. An accompanying contrast plot (data not shown) indicates that nonsenior Republicans/Republican leaners earning $2,000–4,999, $5,000–7,499, and ≥$7,500 are each statistically less likely than nonsenior Republicans/Republican leaners earning under $2,000 to say the ACA has helped them personally thus far. Democrats exhibit a different pattern, but middle-income ambivalence still appears. Nonsenior Democrats/Democrat leaners earning $2,000–4,999 per month are statistically no less likely than nonsenior Democrats earning less than $2,000 to say that the ACA has helped them so far, nor are nonsenior Democrats earning ≥ $7,500. However, upper-middle income nonsenior Democrats—those making $5,000–7,499—are statistically less likely than the lowest-earning nonsenior Democrats to say that the ACA has helped them personally so far. Thus, asked whether the ACA has been personally helpful to date, nonsenior upper-middle-income earners report less sanguine views than nonsenior low-income earners, even among Democrats.
A different story unfolds when examining beliefs that the ACA will benefit one's family in the long run (table 2). The first set of results in table 2 shows that, among respondents of all ages, those making $2,000–4,999 per month before taxes are less likely than those making $2,000 or less to expect the ACA will help them or their families in the long run, while those making $5,000–7,499 and those making ≥$7,500 exhibit no difference from low-income earners in this expectation. Among nonseniors, those making $2,000–4,999 and $5,000–7,499 are both significantly less likely than those making $2,000 or less to think the ACA will be personally helpful. The multivariate results thus echo the story told by figure 2: middle-income earners are less likely to expect the ACA to be personally helpful in the long run than people in the lowest income group.
Yet, this aggregate pattern masks a dramatic difference in how income relates to ACA prospections among Democrats versus Republicans. Figure 4 illustrates the marginal effects generated from the fourth set of results in table 2, again for hypothetical non-Hispanic white, married men with children, of median liberalism, aged 30–49. This figure thus describes results based on the model run on nonseniors, although a similar figure is obtained from the third set of results in table 2, for respondents of all ages. As with figure 3, we can unpack figure 4 by first looking at differences by party and then at differences by income within each party.
In terms of party, the difference in figure 4 between nonsenior Democrats/Democrat leaners and nonsenior Republicans/Republican leaners—the former being more likely to expect the ACA to be helpful—is significant at all income levels. In terms of income, nonsenior Republicans/Republican leaners earning $2,000–4,999, $5,000–7,499, and ≥$7,500 are each significantly less likely than nonsenior Republicans/Republican leaners making under $2,000/month to expect the ACA to be personally helpful in the long run. However, among nonsenior Democrats/Democrat leaners, the top three income groups are each statistically no less likely than those in the lowest earning group to say the ACA will be personally helpful in the long run—pointing to a nascent cross-class constituency among Democrats, rooted in ACA prospections. We obtain similar evidence of a potential cross-class constituency rooted in prospections among Democrats, and no such pattern among Republicans, if we look at all ages.
In sum, among both Republicans and Democrats, middle-income earners split away from low-income earners in their views on whether the ACA has been personally beneficial so far, but upper-income Democrats to some degree stand with low-income Democrats in saying the ACA has been personally beneficial so far, and low-, middle-, and upper-income Democrats all equally expect the ACA to be personally beneficial in the long run.
Independents: Leaving Democrats on Their Own
In their views of the ACA to date and in the long run, Independents show some statistically significant differences from Republicans but on balance look more like Republicans than Democrats in the relationships they exhibit between ACA benefit assessments and income. Returning to figure 3, Independents earning below $2,000 per month are more likely than Republicans/Republican leaners earning the same to believe the ACA has been personally beneficial, and this difference is statistically significant. However, for the top three income groups, the difference between Independents and Republicans/Republican leaners is insignificant. Thus, middle- and upper-income Independents appear to think like Republicans regarding the ACA's personal impacts so far. Differences by income among Independents on this question show some traces of similarity to patterns seen among Democrats when we focus on the top versus bottom earning groups, but middle-income versus low-income earners resemble the pattern seen among Republicans. Specifically, Independents in the top earning group are statistically no less likely than those in the lowest earning group to say they have benefited from the ACA so far, at least among nonseniors. However, Independents from the two middle-income groups are statistically less likely than Independents from the lowest earning group to say the ACA has helped them so far.
In their prospections about the ACA's personal impacts, Independents show optimism that statistically is unified across income groups as it is for Democrats, but at a markedly lower level. And while lower-middle- and upper-income Independents are not as pessimistic as Republicans of the same income, upper-middle- and low-income Independents are. Returning to figure 4, nonsenior Independents earning $2,000–4,999 and earning ≥$7,500 are significantly more likely to expect the ACA to be personally beneficial in the long run than are Republicans/Republican leaners at these income levels. However, among the lowest income earners and those earning $5,000–7,499, Independents show no statistical difference from Republicans/Republican leaners in their expectations of the ACA's personal impacts. The top three income groups among Independents are each statistically no less likely than the lowest earning group to expect the ACA to be personally helpful in the long run, resembling Democrats in that all income groups exhibit statistically equal levels of optimism. Yet, Independents' level of optimism is substantially lower than that of Democrats. Further, among people of all ages, Independents do not join Democrats in exhibiting cross-class optimism about the ACA's long-run personal impacts.12 In sum, Independents' propensity to believe the ACA will be personally beneficial in the long run is statistically indistinguishable from that of Republicans for multiple income groups, and while nonelderly Independents' expectations for the ACA's long-run personal impacts may be internally united in a way that somewhat resembles those of Democrats, Independents' levels of optimism are lower. This result aligns with external evidence that Independents on balance hold negative ACA views (Kirzinger, Sugarman, and Brodie 2016).
Other covariates behave as would be predicted by past research noted above. In particular, liberals are more likely than conservatives, and whites less likely than nonwhites, to say the ACA has been or will be personally helpful. Some patterns merit exploration in separate work, such as the findings that respondents with children are less likely than those without children to believe the ACA has helped them so far, and that women and married respondents are less likely than men and the unmarried to think the ACA will be helpful in the long run.
Understanding Middle-Income Earners' Ambivalence about ACA Benefits
Why are middle-income earners—including Democrats—less likely than low-income earners to think they have benefited from the ACA to date? Why do a minority of middle-income Americans, but especially Republicans, expect to benefit in the long run? While definitively answering these questions is beyond this article, existing research gives suggestive evidence on two possibilities based in the ACA's design: middle-income disappointment with ACA benefits and the limited legibility of middle-class ACA benefits. These features may partly explain aggregate middle-class ambivalence. However, to understand why these design “flaws” yield differing views of the ACA's long-run personal value within the middle class—and to assess whether raising or desubmerging benefits would ease these differences—partisan reasoning (McCabe 2016) appears to be a third, necessary part of the puzzle. The following section unpacks these explanations.
Disappointment with Benefits
Research on ACA marketplaces offers imperfect but useful insight into middle-class ACA benefit experiences, since the marketplaces are what connect people at 100–400 percent FPL to ACA subsidies. Academicians (Grande 2016; Grob et al. 2013; Haeder, Weimer, and Mukamel 2016; Housten et al. 2016; McCabe 2016) and the Henry J. Kaiser Family Foundation (2016b) have studied prices, provider networks, and the experience of shopping in these marketplaces.
Regarding price, social policies may fail to elicit positive feedbacks if their financial benefits are of insufficient magnitude to materially improve beneficiaries' lives (Campbell 2012: 339, 340; Mettler 2002: 361; Patashnik and Zelizer 2013: fig. 1). ACA subsidies may bear this shortcoming. While over 17 million people were eligible for these tax credits in 2014 (Henry J. Kaiser Family Foundation 2013b), competition among insurers on the ACA marketplaces is weak in some locations (Obama 2016: 529–530), and some consumers find the plans unaffordable (Goodman 2015; Thomas, Abelson, and McGinty 2013). People who cannot afford insurance even with subsidies may be disappointed in the ACA (Cammett, Lynch, and Bilev 2015: 951). President Obama (2016: 529) indeed noted that “congressional action to increase financial assistance to purchase coverage . . . would . . . help middle-class families who have coverage but still struggle with premiums.” By way of comparison, Peterson (1999: 33) argues that “what finally put Social Security on a secure basis were amendments adopted in 1950” that, among other things “expanded benefits by 77 percent” (also Béland 2005: 98, 121; Patashnik and Zelizer 2013: 1077–78).
Regarding network adequacy, “secret shopper” research in California's exchange—considered one of the nation's best—suggests that the ease with which one can contact health care providers and obtain timely medical care is no worse but also no better for customers of ACA marketplace plans than for those with nonmarketplace plans (Haeder, Weimer, and Mukamel 2016). More broadly, some plans were canceled after the ACA's rollout (Thomas, Abelson, and McGinty 2013), and certainly “personal experiences with changes in insurance status, not just perceptions of ACA benefits alone, affect views of the ACA's impact” (Jacobs and Mettler 2016: 918). People who lost their coverage between 2012 and 2014 became less favorable toward the ACA (McCabe 2016: 871), and among both Republicans and Democrats, drops in favorability after insurance loss were far greater in magnitude than increases in favorability after insurance gain (McCabe 2016: 873).
Regarding insurance shopping, research suggests that consumers find “insurance . . . and its regulation . . . deeply ‘confusing,’ ‘confounding,’ and ‘not understandable’” and are “often ‘at their wits’ end’” (Grob et al. 2013: 350–51). Insurance terminology is unfamiliar to many people, and health care costs can be unpredictable even for knowledgeable consumers (Grande 2016; Housten et al. 2016). Over 70 percent of uninsured adults want guidance when buying insurance (Robert Wood Johnson Foundation 2015: 9, 45). The ACA offers states federal funds to set up consumer assistance programs, but these programs vary in their consumer advocacy and fail to alleviate “mistrust” and “fear” that insurers will reject health care claims (Grob et al. 2013: 351, 353).
Benefit surety is another area in which middle-income earners of both parties may be disappointed (see, e.g., Campbell 2012: 340). If a household's actual income exceeds its anticipated income in a given year, it must pay back part or all of the tax credit received based on the original estimate (Cox et al. 2015)—a penalty that may affect up to half of subsidy recipients (Ferris 2015) and that spawned such headlines as “Nasty Tax Surprise for Obamacare Customers” (Mangan 2015).
That said, disappointment with marketplaces and subsidies has limitations as an explanation for middle-income ambivalence about the ACA—and appears insufficient to explain differences in Democratic and Republican prospections. The Henry J. Kaiser Family Foundation (2016b) finds (1) that 59 percent of people who actually enrolled in a marketplace plan in 2016 were satisfied with their premiums, and 51 percent with their deductibles; (2) that fully 75 percent of enrollees were satisfied with their hospital choices, 74 percent with their primary care physician choices, and 59 percent with their specialist choices; and (3) that majorities of marketplace enrollees found it “easy . . . to compare premiums (74%), cost-sharing (69%), and provider networks (61%), and to find a plan that met their needs (59%).” Further, McCabe (2016: 871) found both Democrats and Republicans who bought insurance through an ACA marketplace to be significantly more likely to hold a favorable ACA view than similar copartisans who had job-based insurance.
Poor Legibility of Benefits
Survey data give stronger support to the possibility that middle-income ambivalence about ACA benefits owes to their submerged form. In 2015, nearly 60 percent of uninsured eighteen- to sixty-four-year-olds either had not heard of ACA tax credits or found them confusing (Robert Wood Johnson Foundation 2015, cited in Desmond et al. 2016: 422), including 57 and 51 percent of those at 139–250 and 251–399 percent FPL, respectively (Robert Wood Johnson Foundation 2015: 9, 38). Further, between 21 and 39 percent of people may be unaware of the ACA marketplaces that administer these tax credits (Commonwealth Fund 2013a, 2013b; Henry J. Kaiser Family Foundation 2014b).
Survey data also show that low visibility indeed afflicts the ACA's consumer protections—including protections that are popular when made visible. For instance, when told that the ACA requires insurers to offer coverage to all who apply, 47 percent of the public reports feeling “very favorable” toward this provision (Henry J. Kaiser Family Foundation 2011b). In November 2011, however, 42 percent of people earning $30,000–49,999 annually and 22 percent earning over $75,000 believed that the ACA did not include this provision (Henry J. Kaiser Family Foundation 2011c).13 Similarly, 60 percent of adults feel “very favorable” toward the ACA's requirement that health plans provide consumers “easy-to-understand information” about their benefits when informed of its existence (Henry J. Kaiser Family Foundation 2011b), but 45 percent of those earning $30,000–49,999 and 27 percent of those earning over $75,000 believed in November 2011 that the ACA did not include this requirement (Henry J. Kaiser Family Foundation 2011d).14
Partisanship as Cap on Benefit Enthusiasm
Submerged ACA benefits may be gaining visibility over time (Jacobs and Mettler 2016: 918), and more middle-income earners may use marketplace plans if the ACA persists past early 2017—experience that, as noted above, correlates to favorable ACA views (McCabe 2016). Yet, it is not necessarily the case that middle-income Americans would have shown higher levels of ACA enthusiasm—particularly bipartisan enthusiasm—if experiences with marketplaces had been higher or middle-class ACA benefits more legible in the time frame studied here. Democrats and Republicans respond differently to efforts to make the submerged state legible (Shanks-Booth and SoRelle 2016), and “fact checking” beliefs about ACA provisions can deepen ACA opposition among politically knowledgeable people who see malintent in the ACA (e.g., on “death panels,” see Nyhan, Reifler, and Ubel 2013). Regarding experience, McCabe (2016: 862) points out that people who have faced health coverage changes post-ACA have thereby acquired “unmediated sources of information to form their policy preferences on healthcare reform” alongside cues from party elites. However, “partisan directional goals may still operate” when people allocate credit for positive health care changes or blame for negative ones, since individuals “have to come to their own conclusions about who or what is responsible for their healthcare situations” (McCabe 2016: 865). McCabe (2016: 861) finds that Republicans tend to blame the ACA for adverse health insurance changes, while Democrats credit the ACA for positive insurance shifts.
Implications for the ACA
History suggests that a cross-class constituency bolsters a social policy's likelihood of enduring over time (Campbell 2003a; Esping-Andersen 1989; Orloff 1993; Quadagno 1991; Skocpol 1991, 1992), and, as noted, theories about the role that stable policy coalitions play in ensuring policy endurance help explain why this is so (Kingdon 2003; Baumgartner and Jones 1993; Sabatier and Jenkins-Smith 1993). To help delineate the degree to which the ACA is acquiring a cross-class constituency, this article has investigated whether middle-income earners believe they have benefited from the ACA to date and believe they will benefit from the ACA in the long run, comparing their views to those of lower-income earners.
With only subtle differences by party, middle-income earners appear less enthusiastic about the ACA's personal impacts so far than lower-income earners, but asked about the long run, Democrats and Republicans exhibit differing patterns. Middle-income Republicans are less likely to think the ACA will be helpful in the long run than their low-income copartisans, thus showing the same pattern in retrospections and prospections. In contrast, low-, middle-, and upper-income Democrats are all equally likely to think the ACA will be personally helpful someday. While only about 32 percent of Democrats/Democrat leaners in each income group hold these prospections in multivariate analyses, this is the highest enthusiasm for ACA benefits in the results, and the only enthusiasm that is robustly cross-class. Thus, both income and party fragment the ACA's target population.
These results add to the emerging body of evidence that partisanship divides Americans' views of ACA benefits and, more broadly, that partisanship and not just policy design shapes the probability that the ACA will see positive feedback effects (Jacobs and Mettler 2016; McCabe 2016; Morone 2016; Nyhan, Reifler, and Ubel 2013; Oberlander and Weaver 2015). In addition, these results uncover what appears to be a new finding: partisanship conditions income's relationship to Americans' assessments of ACA benefits, especially their prospective assessments.
From a scholarly standpoint, this new finding may help unpack conflicting evidence on the relationship between income and ACA views in existing literature. Knoll and Shewmaker (2015: 90) note that Brady and Kessler (2010) found income to be “the defining determinant in ACA opinions” but that Lynch and Gollust (2010) found that income's relationship to ACA views “wash[ed] out in a multivariate model that includes partisanship and ideology.” The results here do not show income washed out by partisanship but, rather, that partisanship conditions income's link to ACA attitudes—especially beliefs about the ACA's long-run impacts. Democrats' prospections are nearly insensitive to income, while Republicans' prospections are not.
From a policy standpoint, this new finding suggests that fragmentation in middle-class ACA views may be enduring. That middle-income Republicans' retrospective and prospective assessments of the ACA's personal impact differ little, while middle-income Democrats' retrospections and prospections differ much, suggests that “partisan motivated reasoning” (McCabe 2016: 880) is part of why people with the same, middle-class incomes hold sharply different views of the ACA's long-run personal value. If so, then while shortcomings and poor benefit legibility may be two design-based—and thus conceivably revisable—reasons that middle-income earners show only modest enthusiasm for the ACA's impact to date, the long shadow of the ACA's partisan origins may be an unyielding barrier to bipartisan middle-class enthusiasm.
The more immediate policy implication of Americans' income and partisan fragmentation over the ACA is that the mass public appears ill-poised to coalesce against ACA dismantlement—something the 2016 election outcome cemented as a real possibility. Hours after his inauguration on January 20, 2017, President Trump signed an executive order instructing federal agencies to slacken enforcement of many ACA provisions and formally stating his goal of ACA repeal (Trump 2017). While some argue that the order is mainly symbolic (Pollack 2017), others argue that it enables administrators to enfeeble the ACA's employer and individual mandates and its de facto restrictions on the sale of substandard plans in the individual market (Chandler 2017). Such actions could retrench the ACA in concrete ways (see, e.g., Hacker 2004), and may create market uncertainty until Congress passes a replacement policy (Kodjak 2017). Further, the wide discretion that federal administrators used in implementing the ACA under the Obama administration may have set a precedent for assertive moves to revise the ACA by the Trump administration (Adler 2017).
There are caveats to trying to speculate from the results here on whether a cross-class constituency will rally against such threats to the ACA in the participatory ways that seniors rally to protest threats to Social Security (see, e.g., Campbell 2003b). This article studies ACA attitudes, which may be more sensitive to partisanship and less sensitive to policy design than are political behaviors (Campbell 2012: 337–38). People who appear indifferent to a policy may also participate in politics in support of its endurance when mobilized to do so (Arnold 1990: 68).15 In addition, Americans have engaged in unprecedented, multicity protests twice since President Trump's inauguration—on January 21, 2017, in the Women's March (Hartocollis and Alcindor 2017) and on January 28 and 29, 2017, following an executive order regarding immigration (Baker 2017).
It is also worth remembering that Social Security and Medicare both “remained politically controversial for some time after . . . enactment” (Jacobs and Skocpol 2012: 183; see also Oberlander and Weaver 2015: 38)—with the former's survival uncertain even in the middle to late 1940s (Jacobs 2011: 627; Jacobs and Skocpol 2012: 150; Patashnik and Zelizer 2013: 1077). While “the ACA's political foundations remain shaky” (Oberlander and Weaver 2015: 37), it has surely grown some political roots in the past seven years. Basic loss aversion could lead citizens to value ACA benefits “now that they represent the status quo” (Eckles and Schaffner 2010: 1). And, as noted, citizens are not the only actors in the ACA policy coalition. Private industries that worked closely with policy makers on the ACA may protest its repeal (Ario and Jacobs 2012: 1855–60; Grogan 2011: 407–8; Jacobs and Ario 2012: 2603, 2606; Jacobs and Skocpol 2012: 158; Skocpol 2010: 1290).
Yet, citizens' enthusiasm or indifference for the ACA may well “feed into elites' efforts to terminate, replace, or delay implementation of its key provisions” (Jacobs and Mettler 2011: 929). The revolt against the MCCA by nonaffluent seniors who would have gained its benefits well exemplifies that a program's intended beneficiaries can become its political foes in the presence of ambiguous information (Oberlander 2003: 67–68), and ACA challengers have worked to “generate negative interpretive feedbacks” around the law for several years (Patashnik and Zelizer 2013: 1079). Thus, the fragmentation in the ACA's target population along the lines of both income and partisanship found here bespeaks a serious weakness in the ACA's policy coalition.
Implications for Policy Feedback Research
This article draws on policy feedback theories to study the ACA. Given the ACA's unequivocal impact on American social policy—regardless of what precisely happens to the law in 2017 and after—the results return theoretical contributions to policy feedback research. They do so by underscoring the need for greater study of two broad questions in this domain of scholarship.
Can Policy Prospections Fuel Positive Feedback Effects?
The first question for feedback research—suggested by the finding that Democrats show greater prospective than retrospective ACA enthusiasm—is whether optimism about benefits not yet used can inspire participatory responses to retrenchment threats and, if so, whether enthusiasm grounded in prospections has an expiration date. To the author's knowledge, scholars have not studied this question. One reason to doubt that such prospections can single-handedly fuel positive feedbacks is that, while citizens vocally oppose cuts to benefits that they already use (Campbell 2003b), losses draw stronger reactions than gains (Kahneman and Tversky 1979), and benefits not yet used may fall more in the realm of an anticipated “gain.”16 Further, optimism that occurs mainly in citizens whose partisanship matches that of a policy's sponsors could reflect loyalties to that party (McCabe 2016), or even to a president who is the policy's most visible proponent, rather than protective feelings toward the policy itself that will last over time.
Still, prospections could aid positive feedbacks. Citizens support programs they know they may need someday. For example, middle-income seniors are “vitally concerned with preserving” Medicaid because they “view the low-income health plan as a ‘safety net’” for long-term care that Medicare does not cover (Olson 2010: 224). Citizens can also remain receptive to changes in anticipated benefits for years after a new policy is signed into law—as exemplified by the upticks in mass enthusiasm for Social Security after 1939 and 1950 amendments boosted promised benefits (Béland 2005; Patashnik and Zelizer 2013; Peterson 1999). Prospections also impact numerous political outcomes, including presidential approval (MacKuen, Erikson, and Stimson 1992; Rudolph 2002; but see Clarke and Stewart 1994; Malhotra and Margalit 2014; Nadeau and Lewis-Beck 2001), congressional approval (Durr, Gilmour, and Wolbrecht 1997), vote choice (Lacy and Paolino 1998), policy preferences (Durr 1993), and postelection consumer spending (Gerber and Huber 2009). Prospections may also shape feedbacks in as yet unidentified ways.
How and Where Does Partisanship Intervene in the Policy Feedback Logic Model?
The second question for feedback scholars concerns partisanship's place in the feedback logic model. As noted, growing evidence shows that feedbacks around the ACA hinge not just on policy design but on partisan attitudes as well (Jacobs and Mettler 2016; McCabe 2016; Morone 2016; Nyhan, Reifler, and Ubel 2013; Oberlander and Weaver 2015). The results here on partisanship's role in splitting the middle class add to this evidence, underscoring a question posed by Patashnik and Zelizer (2013), Oberlander and Weaver (2015: 39), Morone (2016), and, indirectly, Weaver (2010): Can social policies forged in a polarized context ever garner bipartisan positive feedbacks?
The results here point to the urgent need for a taxonomy by which political scientists, policy makers, and public managers can extract common takeaways from the rapidly accumulating research on how partisanship renders beneficiary feedbacks contingent. This article closes by outlining a possible strategy for how to do so: scrutinize what a given piece of evidence tells us about where partisanship intervenes in the standard policy feedback logic model. The remainder of this section offers a diagram (a revision to Mettler 2002: fig. 1) that may help facilitate such scrutiny and explains how the findings reported in several publications—including this one—can be located in this diagram and can thus offer fairly precise information on where partisanship qualifies the chain of events leading to positive feedbacks from target populations.
As noted, the logic model of how a policy cultivates positive beneficiary feedbacks is arguably best delineated by Mettler's (2002) diagram of policy begetting resource and interpretive effects, which in turn beget civic capacity, predisposition, and engagement in affected citizens. Literature points to policy design and administration as key inputs to resource and interpretive effects.17 Once resources and positive messages are conveyed, the subsequent steps in the logic model should occur. Yet, the ACA poses a puzzle, since it is generating very different assessments of its personal impacts from middle-class Democrats versus middle-class Republicans, according to the findings reported here, even though it is delivering similar resources, and positive if imperfect administrative experiences, to some middle-income earners of both parties (Henry J. Kaiser Family Foundation 2016b; McCabe 2016). This puzzle adds to indications that design and administration are insufficient to ensure positive feedbacks; partisanship conditions the process somewhere.
To underscore the analytic value of locating partisanship's intervention points in the chain of events leading to positive feedbacks, figure 5 presents a revised version of Mettler's diagram, augmented with a sketch of partisanship's conceivable pathways, depicted as dashed gray lines. There are seven routes for partisanship to condition the logic model, if partisanship acts like other overarching phenomena, such as racial considerations and social constructions, that are known to operate as “mental structure[s]” that infuse the attitudes and behaviors of policy makers, citizens, and administrators, thereby shaping benefit administration, take-up, and interpretation (Soss, Fording, and Schram 2011: 4 [quotation]; Schneider and Ingram 2005). While research amply establishes that partisanship pervades every corner of politics, many papers could be written verifying its intervention in feedback effects via these seven paths.18 The paths are labeled as Q1, Q2, and so on, to emphasize that each one's existence and mechanics are empirical questions. Numbering is for convenience and is not a claim about order; partisanship could intervene late in the logic model for a policy, even if not early on (e.g., Q6 could exist even if Q1 does not).19
The findings presented in this article suggest that, for the ACA, partisanship intervenes in the logic model via paths Q1 and/or Q2. It may also travel the other paths,20 but this article is equipped to assess only Q1 and potentially Q2 since, as noted, the outcomes studied here are assessments of resources received or expected from the ACA. That said, more research on partisanship's workings at the resource and interpretive effects juncture appears warranted. Beyond this article, other findings can also be seen as evidence that partisanship conditions positive feedback formation around the ACA via Q1 and Q2. For instance, Jacobs and Mettler (2016) and McCabe (2016) can be seen as offering evidence that a psychological mechanism—motivated reasoning by beneficiaries—carries partisanship on path Q2 by qualifying the formation of positive interpretive effects, regardless of resources conveyed. Sommers et al.'s (2015) findings on low-income earners' divergent ACA views in states with and without expanded Medicaid suggest that policy makers can carry partisanship to the model via Q1 by averting federal transfers in ways linked to party affiliation (see Jones, Bradley, and Oberlander 2014; Rigby 2012). Figure 5 renders Q1 and Q2 with thick dashed lines to underscore both the probable value of further research about these paths and this article's evidence that partisanship fragments middle-class ACA views at this juncture.
To conclude, this article finds that the ACA lacks a cross-class constituency grounded in assessments of its benefits to date. Further, this article points to a need for more research on the role of prospections and partisanship in shaping beneficiary feedback effects. Such work may help us understand mass public responses to unfolding ACA revision and repeal efforts, particularly income- and party-based fragmentation that may characterize those responses. Such work may also help us understand citizen feedbacks around other social policies forged in a polarized context.
I thank Dan Carpenter, Jennifer Hochschild, Suzanne Leland, Eric Patashnik, and two anonymous reviewers for helpful comments related to this work, and Madison Esterle for research assistance. Any errors are my own. Sources citing iPOLL, RoperExpress, and other databases at Roper were obtained from searches of the iPOLL Databank and use of resources provided by the Roper Center for Public Opinion Research at Cornell University. An earlier version of this article was presented at the 2017 Southern Political Science Association meeting.
1. Skocpol's (1992: 149) notion of “fault lines” well captures how lines of difference can operate.
2. Policies targeted to the poor may be vulnerable to retrenchment (Skocpol 1991; Soss, Fording, and Schram 2011) but can endure over time (Greenstein 1991), and middle-class programs can face criticism and revision (Greenstein 1991; Oberlander 2003: 24, 62–66, 79–80, 165–67; Olson 2010: 137–40; Pierson 1994: 6, 53, 58; Skocpol 1995: 7).
3. Grogan and Patashnik (2003: 68) also see the social insurance versus public assistance distinction as inadequate for many policies, such as those “targeted [but] with middle-class inclusion.”
4. Notably, King v. Burwell (135 S. Ct. 2480 (2015)) averted cross-state variation in ACA subsidies.
5. I thank an anonymous reviewer for emphasizing this distinction and the idea of “tracking.”
6. Oberlander and Weaver (2015: 39) in fact note that two ways that the ACA differs from Social Security and Medicare are in the “fragmentation” and “visibility” of benefits.
8. Response order, for both this question and the next, was rotated across respondents. This question was prefaced by a prompt indicating that the next few questions would ask about the ACA. The May 2014, October 2014, and April 2015 surveys used one different word: “As you may know, a number of the provisions” (codebooks accompanying Gallup Organization 2014e, 2014f, 2015).
9. This wording was used in all but the earliest survey. The February 2012 survey asked, “Now suppose all of the provisions of the healthcare law go into effect in the next few years. In the long run, how do you think the healthcare law would affect your family's healthcare situation? Would it . . . Make things better, Not make much difference, or would it Make things worse?” (codebook accompanying USA Today,2012).
10. Soss and Schram's (2007) typology aims to explain feedbacks in mass publics, not strictly in target populations. However, “proximity can refer to the likelihood that other members of the public identify with a program or believe that they might benefit from it someday” (Campbell 2012: 340), and proximity assessments do vary at the individual level (Soss and Schram 2007: 121).
11. Contrast plots illustrating the statistical significance details reported throughout this section are available on request. Margins plots such as figures 3 and 4 do not themselves indicate statistical significance. On margin and contrast plots, see www.stata.com/manuals13/rmarginsplot.pdf.
12. Running the analyses on all ages reveals that, among Independents, the top two income groups are each no less likely than the lowest earning group to expect the ACA to be helpful in the long run, but Independents earning $2,000–4,999 per month are statistically less likely than those earning less than $2,000 to expect the ACA to be personally helpful in the long run.
13. These figures are calculated by Roper iPOLL's iPOLL+ crosstabs feature based on data from Henry J. Kaiser Family Foundation (2011c) and are weighted. No value is reported for people earning $50,000–74,999 annually due to data limitations.
14. As above, these figures are calculated by Roper iPOLL's iPOLL+ crosstabs feature based on data from Henry J. Kaiser Family Foundation (2011d) and are weighted. No value is reported for people earning $50,000–74,999 annually due to data limitations.
15. I thank Eric Patashnik for pointing me to Arnold's (1990) observation in another context.
16. See also Campbell's (2003b: 30, 42) observations on “policy opportunity.”
18. Answers also likely lie in existing research that is simply not framed in these terms.
19. For example, Republicans are less likely to approve of public programs, even controlling for the number of such programs personally used (Mettler and Stonecash 2008: 284). This finding suggests that partisanship can intervene at Q6 or Q7 to avert bipartisan participation in favor of a specific policy, even if that policy's resource flows raise civic capacity in people of both parties.
20. Maximally, it may be that each of the seven intervention points merits individual study in basic research and its own “task force” during and after a policy's implementation.