I. Introduction

As the “first national effort to ensure access [to health insurance] to the nonelderly population,” ^{Reference Jacobs1} the Affordable Care Act (ACA) may well be “the largest change in social welfare policy since the Great Society and perhaps the New Deal.” ^{Reference Jacobs2} Among its aims are insuring a large share of the uninsured, ^{Reference Obama3} cutting insurance disparities along the lines of health status, ⁴ age, ⁵ and race/ethnicity, ^6,7 and reducing households’ exposure to the financial vulnerabilities that illness entails. ^{8,Reference Himmelstein, Warren, Thorne and Woolhandler9} Thus, there is much to motivate research into the ACA’s potential for “durability” ^{Reference Pierson10} —meaning both its avoidance of outright retrenchment, ¹¹ and fidelity to its policy aims over time. ^{Reference Hacker12}

While some features of the ACA’s “design” ^{Reference Campbell13} may dispose it to reap “positive feedback processes,” ^{Reference Pierson14} as part II notes, other features may work against this prospect. Further, a policy’s design is not the only determinant of its trajectory, ^{Reference Schneider and Ingram15,Reference Patashnik and Zelizer16} and it is hardly certain the ACA will generate such feedbacks. ^{Reference Jacobs and Mettler17,Reference Sage18,19} As the ACA is implemented, it is surrounded by informal tools of political opportunity and challenge. ²⁰ How its beneficiaries are portrayed is one such tool.

Past research on American social policy indicates that portrayals of a policy’s beneficiaries may relate to the responses the policy generates among those beneficiaries, the mass public, and political elites—actors all critical in policy implementation and durability. ²¹ At the heart of much of this work is Schneider and Ingram’s concept of the “social construction of target populations,” meaning the “characterizations or popular images of the persons or groups whose behavior and well-being are affected by public policy.” ²² Portrayals’ dynamics are complex. Policies may enjoy political stability when portrayals help underscore beneficiaries’ breadth, ^{Reference Skocpol and Skocpol23,Reference Skocpol24} and “[a]rguments or images that spotlight social groups may activate stereotypes and prejudices” that “then become the dominant guideposts for the evaluation of public policy.” ^{Reference Nelson and Kinder25} Simultaneously, policies curry strength and guard beneficiary interests when they encourage beneficiaries to see themselves as “politically meaningful” ²⁶ groups, ^{Reference Mettler27,28,29} and may lack political endurance when this is not the case. ^{30,31,Reference Soss, Fording and Schram32} Beneficiary mobilization is “not an automatic process,” ³³ and “[l]egislative achievements remain…prone to being dismantled” ^{Reference Mettler34} when intended beneficiaries have “barely offered more than tepid support.” ³⁵ Such dynamics may matter greatly to the ACA as a law forged amid partisan strife ³⁶ and repeatedly subject to Supreme Court scrutiny. Yet, no study to date has systematically described and analyzed ACA beneficiary portrayals. This paper does so. It first seeks descriptive answers to three questions about portrayals that the literature connects to policy durability: (1) Do portrayals depict ACA beneficiaries as economically heterogeneous? (2) Do portrayals focus attention on groups that have acquired new “political relevance” ³⁷ due to the ACA? (3) What themes that have served as messages about beneficiary “deservingness” in past American social policy are most frequent in ACA beneficiary portrayals? It then assesses how the portrayal patterns uncovered in answering these questions may work for and against ACA durability, finding reasons for confidence and caution.

Specifically, this paper reports on portrayals of two ACA “target populations” before and during the first four months of the ACA’s first open enrollment period: those who are (or would be) newly eligible for Medicaid pending state expansion, and separately those eligible for the ACA’s insurance subsidies. A focus on these groups is logical since two major ways that the ACA aims to cover the uninsured are by offering (1) Medicaid to people up to 133 percent (“effectively 138 percent” ³⁸ ) of the federal poverty level (FPL) who were not previously eligible for the program, and (2) subsidies to buy insurance to the uninsured at 100–400 percent FPL. ^{Reference Jacobs39} While not the only means by which the ACA addresses uninsurance, these two mechanisms’ importance is underscored by their discussion among academics, ⁴⁰ research institutes, ^{Reference Price and Saltzman41,Reference Dorn, Holahan, Carroll and McGrath42,Reference Shartzer, Kenney, Long, Hempstead and Wissoker43} and courts, ^{Reference Gostin and Garcia44,45} by the striking cross-state (and cross-time) variation in Medicaid expansion and exchange facilitation arrangements (see Table 1), ^46,47,48,49 and by their inter-dependence. ⁵⁰

Table 1. State Medicaid expansion and exchange decisions as of fourth quarter 2015.

Note: Medicaid decision data in this table are from http://kff.org/health-reform/state-indicator/state-decisions-for-creating-health-insurance-exchanges-and-expanding-medicaid/ and are as of November 2, 2015. This KFF webpage also provides data on exchange arrangements, but exchange data in this table are from page 56 of the report at http://www.ncsl.org/portals/1/documents/health/health_insurance_exchanges_state_profiles.pdf, and are as of October 20, 2015. While assembling the newspaper text data for this article, the author referred to Medicaid expansion decision data from June 2014 from http://kaiserfamilyfoundation.files.wordpress.com/2014/03/current-status-of-the-medicaid-expansion-decisions-healthreform1.png, and exchange arrangement decision data from May 2014 that were available on page 56 of the report previously found at http://www.ncsl.org/Portals/1/Documents/Health/Health_Insurance_Exchanges_State_Profiles.pdf.

This paper also reports on portrayals of two groups informally crafted by the ACA in this period: those gaining health insurance (insurance-gainers) due to the ACA, and those losing it (insurance-losers). The latter emerged as a category during HealthCare.gov’s rollout, as well illustrated by a document entitled “‘Because of Obamacare…I Lost My Insurance”’ that circulated among ACA opponents in Congress in November 2013. ^{Reference Lipton and Stolberg51} For brevity, this paper refers to the newly Medicaid-eligible, subsidy-eligible, and insurance-gainers collectively as ACA “beneficiaries,” in the sense used by others. ⁵²

The portrayals reported come from an original dataset of newspaper article text applicable at the state level from August 1, 2013 through January 31, 2014, gathered retroactively on February 3, 2014. (Citations for the newspaper articles used to build this paper’s data are not included in the bibliography but are available by request). The ACA’s first open-enrollment period officially ran October 1, 2013 to March 31, 2014, with a brief Special Enrollment Period thereafter. ⁵³ The data thus begin two months before, and run four months into, this first open enrollment period. For this reason and others that part IV notes, the timeframe studied is a critical one in the ACA’s implementation.

To preview the results, the data suggest that portrayals underplay beneficiaries’ economic heterogeneity, pay little attention to young adults—a key group the ACA makes politically relevant—and emphasize themes of workforce participation, economic self-sufficiency, and insider status, giving little attention to health status, age, gender, and race/ethnicity. These three descriptive findings—assessed through the lens of existing literature—suggest reasons for both expectancy and caution regarding the ACA’s durability. Favoring durability is the finding that beneficiaries are depicted in terms associated with deservingness in past American social policy—particularly being cast as workers and insiders. Yet, there are three broad reasons for caution. First, ACA insurance-losers are also portrayed as deserving. Second, it is unclear how the portrayal patterns found may impact the durability of the ACA’s efforts to cut insurance disparities by age, health status, and especially race/ethnicity. Third, portrayals’ strong casting of beneficiaries as workers, and limited attention to beneficiaries’ economic heterogeneity and to young adults, may do little to help the ACA cultivate beneficiaries’ political engagement.

II. Background: motivations and foundational literature

A key part of the ACA’s design that may support durability is that it is in many ways what Skocpol terms a “[c]ross-[c]lass [s]ocial [p]olic[y].” ⁵⁴ Skocpol and others ^{55,Reference Jacobs and Skocpol56,57} note that the ACA is a “[r]edistributive [r]eform” ^{Reference Skocpol58} in granting “subsides to millions of working people of modest means.” ⁵⁹ It also includes features focused on cutting disparities. ^60,61 Simultaneously, subsidies go to middle-income earners. ^62,63 President Obama has appealed to the mass public for “‘help in spreading the word”’ on how the ACA “‘can make a difference in your lives and the lives of your families,”’ ⁶⁴ and the ACA by some estimates may benefit the “vast majority of Americans.” ⁶⁵ Cross-class policies tend to enjoy notable spans of political success. ^{66,Reference Katz67} They are administered in non-stigmatizing ways, ^68,69 foster political participation by beneficiaries of all income levels, ⁷⁰ and arrange at least latent coalitions of low- and middle-income earners. ^{Reference Weir, Orloff, Skocpol, Weir, Orloff and Skocpol71,Reference Skocpol and Skocpol72} Scholars argue that the latter’s stake as beneficiaries bolsters Social Security, ^{73,Reference Skocpol and Skocpol74} Medicare, ^{Reference Oberlander75} and Medicaid. ^{Reference Olson76} Neckerman, Aponte, and Wilson argue that programs “available to working- and middle-class segments of society as well as to the poor, are more likely to attract political support, and therefore more generous and stable funding.” ^{Reference Neckerman, Aponte, Wilson, Weir, Orloff and Skocpol77} Greenstein argues that policies are politically stronger when—like the ACA in some respects—they “do not separate those modestly above the poverty line from those well below it” but rather “combine these two groups.” ^{Reference Greenstein, Jencks and Peterson78}

Yet, another part of the ACA’s design may work against  beneficiary mobilization and mass support: many of its benefits are what Mettler calls “submerged.” ⁷⁹ Unlike Social Security, whose “stakes are high and clear,” ⁸⁰ submerged benefits “leave much of the public, even beneficiaries themselves, unaware of government’s role, incognizant of…what is at stake… and unengaged and unlikely to take action.” ⁸¹ Individuals’ “awareness of government activity” is key in fostering positive policy feedbacks; ⁸² but, “unlike visible policies that…attract group loyalty, policies of the submerged state are…too hidden…to generate such affiliations.” ⁸³ Mettler notes that if a policy “neglects to reveal to citizens what government will do or has done on their behalf, it may….prove unsustainable, as opponents dismantle it or undermine implementation.” ⁸⁴ Mettler ⁸⁵ and Patashnik and Zelizer ⁸⁶ explicitly worry that these concerns apply to the ACA.

Given these cross-cutting considerations, influences on the ACA’s trajectory other than its design may be pivotal. Political context matters to policy durability. ⁸⁷ An informal but potent influence on a policy’s course may reside in how beneficiaries are discussed and portrayed. ^{88,Reference Winter89,90} Certainly, non-portrayal factors also matter, including political institutions, ^{Reference Weir, Orloff, Skocpol, Weir, Orloff and Skocpol91,92,93} coalitions, ^{Reference Weir, Orloff, Skocpol, Weir, Orloff and Skocpol94,Reference Finegold, Weir, Orloff and Skocpol95,Reference Myles, Weir, Orloff and Skocpol96,97} “voluntary groups,” ⁹⁸ regional economies, ^{Reference Quadagno, Weir, Orloff and Skocpol99,100} and others articulated by Pierson, ¹⁰¹ Greenstein, ¹⁰² and Patashnik and Zelizer. ^{103,Reference Patashnik and Zelizer104} Yet, as discussed below, portrayals relate to a policy’s reception among intended beneficiaries, the mass public, and elites. And critically, portrayals may impact beneficiaries’ participation in political institutions and coalitions, ^{105,Reference Ingram, Schneider, Schneider and Ingram106,Reference Soss107} thus setting—like prior policies do ¹⁰⁸ —the terms on which these other influences on a policy’s course operate. For all of these reasons, ACA beneficiary portrayals deserve study.

Portrayals also merit attention because three factors make the ACA vulnerable to the politics of target population constructions. First, the fact that the ACA is not universal, and does redistribute, makes it susceptible to a point made by Nelson and Kinder and echoed by Iyengar: ^{Reference Iyengar109} “[g]roup-centrism is pervasive” in American politics. ¹¹⁰ “Many policies are…criticized on those grounds in public debate,” ¹¹¹ potentially influencing mass opinion on the policy. ¹¹² And, this criticism of policy can entail criticism of beneficiaries that discourages their program-specific and broader political participation. ¹¹³

Second, scholars suggest that beneficiary portrayals can play a role in bringing about an eventual, and potentially bipartisan, unravelling of policies embraced by only one party. For instance, Soss, Fording, and Schram argue that “[f]rom the 1960s on, conservatives brandished the troubled lives of the poor as evidence of liberal [policy] failure and as a wedge issue,” ¹¹⁴ and that parts of the 1996 Personal Responsibility and Work Opportunity Reconciliation Act responded to these portrayals, ¹¹⁵ replacing Aid to Families with Dependent Children with Temporary Assistance for Needy Families. There are likely examples with the left’s and right’s roles reversed in other policy areas. The ACA has faced strong Republican challenge from inception, ^116,117,118 though the law also is not beloved by all Democrats. ¹¹⁹ Skocpol and others argue that “no one…should think…the battle is over” ¹²⁰ simply because the ACA became law. ^121,122

Third, as explained below, benefit take-up may be sensitive to portrayals, and take-up is critical to the ACA’s implementation. The possibility that “customers would not come” was a main concern of an executive branch official involved in implementing the ACA federal exchange. ^{Reference Stolberg and Shear123} Low rates of Medicaid and subsidy take-up may especially hurt poorer states’ economies, ¹²⁴ as these benefits are key routes by which the ACA stands to put money into states as others send it to the federal level. ^{125,126,Reference Dubay, Holahan, Long and Lawton127,128} Thus, intended enrollees not taking up these benefits is arguably functionally akin to states leaving “money on the table.” ^{Reference Nicholson-Crotty129}

A closer look at existing research unpacks how beneficiary portrayals may relate to a policy’s reception by beneficiaries, the public, and elites; how portrayals may bear on people’s thoughts about a policy; and why media portrayals are particularly important in these dynamics.

III. Descriptive questions

A vast range of social constructions merit study and could shape an analysis of target population portrayals. The analyses here do not start from hypotheses about the content of ACA beneficiary portrayals. Rather, the goal is to first objectively describe those portrayals, with an eye toward answering the three questions stated in the introduction, and then assess the possible pros and cons of the answers uncovered for durability. Existing literature explains how these three questions are useful in understanding portrayals’ possible bearing on a policy’s strength, and offers guidance for detecting the answers to these questions in textual portrayal data.

1. Are ACA beneficiaries portrayed as economically heterogeneous?

As part II notes, a large set of scholarship argues that policies known to carry “cross-class” benefits enjoy broader mass support, ^{232,Reference Wilson, Jencks and Peterson233,234,235,236,237,238} and broader beneficiary political participation, ²³⁹ than those appearing to benefit only poorer Americans. (And policies seen as narrowly benefitting the wealthy may also be unpopular. ²⁴⁰ ) References to the range or mix of income levels to which a program gives benefits signal that a policy’s beneficiaries are economically heterogeneous. ²⁴¹ Portrayals advancing the “political isolation of the poor…from the working and middle classes” would do the opposite. ²⁴²

2. Do portrayals of beneficiaries highlight groups to which the ACA gives new political relevance?

Another question in the literature, and one highly relevant to the ACA, ²⁴³ is whether a policy and the politics around it encourage beneficiaries to “identify themselves as [its] valued recipients.” ²⁴⁴ Policies can “stimulat[e] brand new social identities and political capacities,” ²⁴⁵ and gain political protection in the process, as documented in Campbell’s study of how Social Security made seniors “[o]nce the age group least likely to participate in politics…the most active.” ²⁴⁶ “Without a government policy…there [wa]s no politically meaningful senior ‘group.”’ ²⁴⁷ But “[o]nce governmental benefits [we]re conferred on the basis of age…the group ha[d] political relevance and [wa]s ripe for mobilization by policy entrepreneurs, interest groups, and political parties.” ²⁴⁸ Similar potential exists in the ACA in several ways. ^{249,250,251,252} To take one example, young adults may remain on parental insurance (if existent) up to age 26 under the ACA, ^{Reference Sommers, Buchmueller, Decker, Carey and Kronick253} and are coveted enrollees in the insurance options the ACA organizes. ^254,255 While engaging this group is not—as part VIII discusses—the only way the ACA may “make citizens,” ²⁵⁶ tracing mention of young adults is a start at gauging whether beneficiary portrayals are helping to foster newly “[p]oliticized social identities.” ²⁵⁷

3. What themes that have served as messages about beneficiary “deservingness” in past social policy are most frequent in ACA beneficiary portrayals?

A question scholars have scrutinized for virtually all policies is whether beneficiaries are cast as “deserving” or “undeserving.” ^{Reference Schneider and Ingram258,259,260,261} Divisions between these categories “run like fault lines through the entire history of American social provision.” ²⁶² Policies whose beneficiaries are understood as deserving tend to be popular with the mass public, ²⁶³ and those beneficiaries tend to exhibit “no moral qualms about defending the program.” ²⁶⁴ Policies whose beneficiaries are constructed as undeserving tend to be unpopular with the mass public ²⁶⁵ and beneficiaries themselves. ²⁶⁶ Before this question can be broached for the ACA, however, it is necessary to answer a more preliminary one: what themes that have served as messages about beneficiary “deservingness” in the past are seen in ACA beneficiary portrayals?

Past research indicates that messages about deservingness may take the form of portrayals that associate beneficiaries with “social groups” ²⁶⁷ or “subgroups,” ²⁶⁸ use “[p]ositive” or “[n]egative” constructions ²⁶⁹ or stereotypes of those groups to deem them (un)deserving, and operate on mass, elite, and beneficiary attitudes through framing mechanisms. ²⁷⁰ Citing Gordon, ^{Reference Gordon271} Soss, Fording, and Schram observe that “[e]valuations of deservingness [historically] fell along many dimensions…often rooted in differences of gender, race, ethnicity, and religion.” ²⁷² Alleged behaviors are another metric in evaluating deservingness. ^{Reference Kim273} Past research indeed highlights several themes by which portrayals may send messages about deservingness.

One is through messages about whether beneficiaries engage in work and efforts toward self-sufficiency. Work strongly, perhaps centrally, ²⁷⁴ distinguishes the deserving in American social policy. ^{Reference Piven and Cloward275,276,277,278,Reference Skocpol and Skocpol279,Reference Ikenberry, Skocpol and Skocpol280,281} Thus, “citizens and taxpayers” are considered deserving, ²⁸² as are the “working poor.” ^283,284,285 Schneider and Ingram’s “[a]dvantaged” category consists of those who are favorably constructed and includes “business, science, the military, [and] the middle class.” ²⁸⁶ Veterans are usually seen as deserving, ^{Reference Amenta, Skocpol and Skocpol287,288,289,290,Reference Jensen, Schneider and Ingram291} as are those pursuing education. ^292,293 While “‘the rich”’ and “‘Wall Street bankers”’ are often viewed unfavorably, ²⁹⁴ customers or “‘clients”’ ^295,296,297 in the economy are constructed favorably. Especially in neoliberalism, the “competent and self-reliant market actor—working, investing, choosing, and assessing returns—is…synonymous with the good citizen.” ²⁹⁸ Those unable to “look out for their own interests” ^{Reference Schriner, Schneider and Ingram299} or achieve “economic self-reliance” ³⁰⁰ are cast as the opposite. Indeed, Americans across the income distribution often see welfare recipients as undeserving, ^{301,302,Reference Quadagno303,304} along with the non-working “nonelderly, nondisabled poor,” ³⁰⁵ criminals, ³⁰⁶ and those with “‘vicious habits”’ ³⁰⁷ —such as alcohol or drug use ³⁰⁸ —believed to self-inflict dependency. ³⁰⁹

Whether beneficiaries are ‘‘insiders” or ‘‘outsiders” has also served as a signal of deservingness, with “ingroups” being considered deserving. ³¹⁰ (Soss, Fording, and Schram in fact suggest that the “line between community members and ‘outsiders”’ historically formed a “moral distinction” alongside deservingness, drawn between people. ³¹¹ ) Certain words or phrases can convey assertions about whether beneficiaries are “insiders.” Winter notes that use of the “first person” and phrases like “our elderly” by elites discussing a policy signals that beneficiaries are “ingroups,” ³¹² as do comments associating the policy with “‘everyone,”’ the “national good,” and referencing “‘Americans…pull[ing] together for the common good.”’ ³¹³ Ono and Sloop make similar observations about words such as “Californians” in newspapers in some circumstances. ³¹⁴ While this article studies media remarks, Winter’s notes on the language of elected officials, and Ono and Sloop’s insights, are instructive in explaining one way to track references to “insider” status in the portrayal data.

Race is a uniquely strong cue in American politics, ^{Reference Carmines and Stimson315,316} and one that may be invoked obliquely. ^{Reference Mendelberg317,Reference Mendelberg318,319} Schneider and Ingram observe that racial and ethnic minorities have frequently been negatively constructed ³²⁰ as compared to Americans of European origin. ³²¹ Other scholarship gives examples of these constructions, ^{322,323,324,325} and research finds beliefs about beneficiary race or ethnicity often twined with beliefs about whether benefits are deserved—a pattern that scholars have interpreted as evidence of “symbolic racism,” ^{Reference Rabinowitz, Sears, Sidanius and Krosnick326} “racial resentment,” ^{Reference DeSante327} and “welfare racism.” ³²⁸ Winter finds favorability to Social Security as an “earned” benefit linked not so much to its universalism but its association with “whiteness.” ³²⁹ Gilens and others present evidence that opposition to welfare is rooted in an interaction of negative racial stereotypes with beliefs that African Americans are beneficiaries. ^330,331,332 Soss, Fording, and Schram report that welfare is more punitive where many beneficiaries are racial minorities. ³³³ Asian Americans, ³³⁴ Latinos, ^{Reference Ana335} and immigrants ^{Reference Hochschild, Chattopadhyay, Gay and Jones-Correa336,337,Reference Schneider, Ingram, Schneider and Ingram338} have often been defined as undeserving in being defined as “foreigners” ³³⁹ or “outsiders.” Soss, Fording, and Schram’s “Racial Classification Model” ³⁴⁰ posits that “race can operate below the level of conscious reflection” ³⁴¹ as a “mental structure” ³⁴² in all individuals’ beliefs and decisions. ³⁴³ In researching race and policy, it is important to study “how minorities are portrayed,” ³⁴⁴ and the many “implicit” forms racial references can take. ^{345,346,347,348} Such analysis is, however, beyond the methodology of the present article. As a start at preliminarily assessing whether race/ethnicity is salient in ACA portrayals, this article tracks the frequency of explicit references. (See the Appendix for a detailed methodological explanation.)

Portrayals of beneficiary gender have also carried implications about deservingness in U.S. social policy. Public assistance beneficiaries may be seen more sympathetically when female. ^349,350 The United States once “came close to forging a maternalist welfare state” dedicated to “help[ing] adult American women as mothers or as potential mothers,” ³⁵¹ and mothers tend to be positively constructed. ³⁵² In contrast, “able-bodied” men on public assistance are especially considered to be the “‘unworthy’ poor.” ³⁵³ That said, women’s somewhat more favorable treatment in social policy is historically more typical for white than non-white women, ^{354,355,356,357,358} middle and upper income than poor women, ³⁵⁹ and married than unmarried mothers. ^360,361,362 Thus, it is relevant to track women’s prevalence, and references to motherhood and families, in the data.

Messages about deservingness have also been carried in portrayals of vulnerability, especially rooted in age and health. Children ^{Reference Skocpol and Skocpol363,364,365,366} and elderly people are treated as deserving, ^367,368 as are the disabled, ³⁶⁹ with some caveats. ^{370,Reference Piatak371} Families are viewed favorably, ³⁷² but single-parent households less so. ^373,374,375

Table 2 summarizes the groups and behaviors used in past constructions of deservingness or lack thereof as suggested by the aforesaid literature, roughly identifying how groups, or groups qua behaviors, have been used as foils against each other in politics according to that literature. It is similar to Schneider and Ingram’s table of the positively and negatively constructed and includes groups (e.g. “The Rich,” “dependents,” Wall Street/“big” business) that they define, ³⁷⁶ but unlike their table ignores power. It is also similar to Kim’s table of traits and behaviors used as, essentially, signals of (un)deservingness, and attributed to people and groups in politics and media in the past. ³⁷⁷

Table 2. A summary of the literature: Past divisions of deserving and undeserving.

Table 2 has caveats. First, constructions of deservingness change over time. ^378,379 Second, there are historical exceptions to several groups’ placement, including immigrants, ^{Reference Newton, Schneider and Ingram380} African Americans and Whites, ^381,382 Asian Americans, ³⁸³ veterans, ³⁸⁴ and seniors. ³⁸⁵ Third, race/ethnicity references can overlap with insider/outsider references, ^386,387 and are also “enmeshed” ³⁸⁸ with ideas about work. ^389,390,391 And as noted, women’s and mothers’ perceived deservingness may vary by race, income, and marital status. Fourth, “[t]he ACA is disrupting long-standing patterns of American politics” ³⁹² and so could alter the table.

IV. Data

This paper aligns with past research on portrayals in studying news, ³⁹³ newspapers, ³⁹⁴ and newspaper text. ^395,396 Druckman finds that “television news and newspapers…do not drastically differ in terms of content” and that “newspapers, and not television news, play a significant…role in informing the electorate.” ^{Reference Druckman397} Others concur. ^{Reference Chiasson398} Schneider and Ingram identify text as a source of portrayal data, ³⁹⁹ and Oberlander also indicates that the “vocabulary” used in describing a policy’s beneficiaries is a key influence on how they are politically understood. ⁴⁰⁰

Newspaper articles discussing those eligible for Medicaid and/or subsidies under the ACA were collected through LexisNexis Academic, retroactively on February 3, 2014. Articles identified as relevant to a state include those in newspapers based in that state, and those that mention that state printed in out-of-state papers, according to LexisNexis. The articles should thus reflect not only what residents may have read from state-based outlets, but what they (or elites or journalists) may have read about the state in other outlets. Complete search procedures are available on request. The search returned 8019 articles. The average number per state is 160.4 ( $\text{SD}=155$ ). California had the most (897), Delaware and Nebraska the fewest (18 each).

As noted, the articles run August 1, 2013 through January 31, 2014, starting two months before the first ACA open-enrollment period officially began and running through its first four months. This timing alone makes the period studied significant. At this time, many Americans first interacted with a major ACA policy output—insurance exchanges. Important pieces of the ACA that give it cross-class reach also took effect in this period. In these months, the ACA touched people with incomes well above the federal poverty line, with the initiation of subsidies on January 1, 2014, ⁴⁰¹ and touched those just above and far below the poverty line with the Medicaid expansion also intended for January 2014. ⁴⁰² This conveyance of “resources” likely helped the ACA begin to be “woven into the fabric of people’s lives” in a way that it was not in its initial years. ⁴⁰³ Also in these months, the ACA became squarely applicable to working-age Americans, beyond the young and elderly as in earlier years. ⁴⁰⁴

The months studied are also prime for research on beneficiary portrayals because in these months, news stories about people signing up for insurance under the ACA were salient insofar as HealthCare.gov was salient. New York Times articles from the last quarter of 2013—not drawn from the articles coded to build the text data for this article, any overlap being by chance—remarked on the “political crisis caused by…[the website’s] disastrous debut,” ^{Reference Pear405} noting that President Obama “had to stand in the Rose Garden to apologize for a broken Web site.” ^{Reference Johnson and Reed406} The website’s problems prompted several, visible, executive branch actions from October to December 2013. ^{Reference Goldstein407,408,Reference Shear and Pear409} The timeframe studied here coincides with and extends just beyond these events. This salience means that the portrayals studied here are ones many citizens likely encountered and possibly “received.” ^{Reference Zaller410}

V. Methods

The Appendix provides a full description of the methods used to analyze the articles. In brief, the analysis included three phases. First, the author and graduate assistants manually coded select sentences in each state’s article batch for descriptions of each of the four groups: the newly Medicaid-eligible, the subsidy-eligible, insurance-gainers, and insurance-losers. For the first and third groups, coding excluded descriptions of people eligible for Medicaid prior to the ACA, joining it by what is known as the “woodwork effect.” ^{Reference Blumenthal and Squires411} For all groups, references to national-level information were coded, but final coded material should otherwise describe only residents of the state at hand, not of other states. Consequently, when it appears in the data, a term like “Californians” should come from a California article and thus signal an “insider” portrayal in the sense discussed in part III. Second, I harvested the words thus coded for each of the four groups, placing them in four documents (“bags of words” ^{Reference Grimmer and Stewart412,Reference Lucas, Nielsen, Roberts, Stewart, Storer and Tingley413} ) for each state. I then assembled four “corpus[es],” ⁴¹⁴ one consisting of the 50 documents (one per state) describing the newly Medicaid-eligible, a second the 50 describing the subsidy-eligible, a third the 50 describing insurance-gainers, and a fourth the 50 describing insurance-losers. I ran each corpus through basic automated textual processing in R, making the unconventional choice not to reduce words to their roots and not to remove terms like “the,” numbers, and punctuation ^415,416 in the interest of preserving information relevant to portrayals. I then generated four “document-term matri[ces]” (“DTMs”), ^417,418 one consisting of the terms describing the newly Medicaid-eligible, one the terms describing the subsidy-eligible, one the terms describing insurance-gainers, and one the terms describing insurance-losers. A DTM reports each unique term in a corpus and its frequency by document. ⁴¹⁹ Third, I manually assigned each term a “tag.” Tags mainly stem from the literature reviewed in part III. For instance, “his” and male names such as “jonathan” were tagged Male. The term “our” was tagged Strong Insider given Winter’s discussion of “our elderly.” ⁴²⁰ A “dictionary” ⁴²¹ lists the terms receiving each tag. Dictionary 1 (see online Appendix 1) contains 40 tags, which appear in the left-most column of Table 3. I created five alternative dictionaries (described in online Appendix 2) since terms can be differently tagged based on analyst interest. Results reported in detail are those from using Dictionary 1. Part VI mentions two other dictionaries (3 and 5), described in the Appendix.

Table 3. Proportion of tagged terms receiving each tag, using Dictionary 1.

Note: Where no terms in the DTM received the tag, “no terms” appears; 0.000 is non-zero.

To check if results would systematically differ using a database other than LexisNexis, an assistant gathered articles for August 1, 2013 through January 31, 2014 from NewsBank for ten states. Articles were analyzed, as above, for portrayals of the subsidy- and newly Medicaid-eligible. As part VI reports, results differ little from those the LexisNexis articles yield.

VI. Results

Before answering the three questions set forth in the introduction and part III and assessing their implications, I report factual results for each DTM. Results are reported for the 50 states aggregated together. Since some articles appeared in the article batches of more than one state, the aggregate results up-weight portrayals that were more widespread. Although part IV makes clear that data were gathered at the state level, I do not assess cross-state portrayal variation. The ten-state NewsBank check is adequate only to tell us whether articles obtained through LexisNexis versus NewsBank would generate different results at the aggregate level. The NewsBank results suggest the answer is no; but, it is still possible that articles found through LexisNexis versus NewsBank carry systematically different portrayals at the state level. Whether this is the case requires study beyond this paper. By reporting results for the aggregate level, I report findings that the NewsBank check suggests would replicate reasonably well if articles were gathered through a database other than LexisNexis.

NewsBank check for the newly Medicaid-eligible and subsidy-eligible

As noted, analyses of portrayals of the subsidy- and the newly Medicaid-eligible were also performed using articles located through NewsBank for ten states. I coded the DTMs obtained from this work using Dictionary 1, albeit that a small number of terms had to be assigned a tag here for the first time by virtue of not appearing in the LexisNexis articles. (Dictionary 1 in online Appendix 1 does not include these new terms; they are available by request.)

Figure 1. Top three tags as shares of tagged terms, by group.

Table 4. Top three tags for each ACA group.

In the resulting Medicaid DTM, there are 747 unique terms and 4641 terms total, 66.5 percent of which were left untagged. For the 1556 terms tagged, top tags were Strong Insiders (20 percent), followed by Workers (19 percent) and Poor/Very Poor (18 percent). The top three tags are thus the same as in the Dictionary 1 LexisNexis results, although their ordering differs.

In the subsidy DTM from this work, there are 733 unique terms and 5287 terms total, 71 percent of which were left untagged. For the 1512 terms tagged, the top tags are Workers (25 percent), Families (13 percent), and Poor/Very Poor (12 percent). These results too are very similar to the Dictionary 1 LexisNexis results. Full NewsBank results are available on request.

VII. Discussion: answers to the three descriptive questions

We can now answer the descriptive questions about portrayals outlined above, and then discuss their implications in part VIII. To aid the discussion, Table 4 recaps the top three tags (using Dictionary 1) in portrayals of the newly Medicaid-eligible, subsidy-eligible, insurance-gainers, and insurance-losers. Figure 1 illustrates their proportions as a share of tagged terms.

1. Are ACA beneficiaries portrayed as economically heterogeneous?

Terms signaling income heterogeneity among ACA beneficiaries were tagged Cross Class. (See online Appendix 1 for details for Dictionary 1.) This tag is absent from Table 4. Under any dictionary, this tag accounts for less than 2 percent of tagged terms describing the newly Medicaid-eligible, and less than 1 percent of tagged terms describing insurance-gainers (and insurance-losers). It accounts for 15 percent of terms in the subsidy DTM when using Dictionary 3 but virtually the same share as in Dictionary 1 (2.4 percent) using the other dictionaries.

By definition, the newly Medicaid-eligible are those 138 percent FPL or below, so it is logical that Cross Class terms are infrequent in portrayals of this group. But the infrequence of cross-class terms in portrayals of the subsidy-eligible—at least absent strong assumptions about numbers—may signal an under-representation of the de jure cross-class nature (100–400 percent FPL) of this group in media portrayals. The infrequence of Cross Class terms in portrayals of insurance-gainers may also understate the prevalence of middle income earners in this group. Government data on the incomes of those receiving ACA benefits do not yet appear to exist, but estimates suggest that, among previously uninsured adults gaining insurance after the ACA, over 40 percent have family incomes ranging 139 percent to 399 percent FPL, and over 10 percent have family incomes at 400 percent FPL or more. ^{Reference Shartzer, Long and Zuckerman422} Such figures reinforce arguments by Skocpol and Jacobs and others that ACA beneficiaries are economically diverse. ⁴²³ Yet especially for insurance-gainers, portrayals do not appear to underscore this economic heterogeneity, and so likely failed to help address indications that “many low to middle income Americans remained unaware of the law’s key features and were skeptical that it will actually help them” as of 2013. ⁴²⁴ Thus, portrayals may fail to help bestow on the ACA the political strength many find around cross-class policies.

An important caveat is the possibility that cross-class portrayals abound but are missed by the methods here. This would happen if cross-class markers largely surface in numbers rather than words, as the dictionaries (albeit #3) leave many numbers untagged since they are difficult to interpret without context. (And even Dictionary 3 leaves numbers lacking “$” signs largely untagged.) Fully detecting cross-class portrayals may require different methods.

2. Do portrayals of beneficiaries highlight groups to which the ACA gives new political relevance?

As noted in part III, the ACA gives new political relevance to being a young adult under age 26. The U.S. Department of Health and Human Services (HHS) reports that 28 percent of those who selected an insurance plan on ACA exchanges in the first open enrollment period were aged 18 to 34; ⁴²⁵ (age 18 to 26 data appear unavailable). Young adults were thus indeed a non-trivial share of exchange-based enrollees. Yet, references to young adults make up less than 2 percent of tagged terms in all four DTMs (albeit 2.2 percent of the subsidy DTM and 2.6 percent of the insurance-gainer DTM using Dictionary 3). Their infrequent mention suggests that ACA beneficiary portrayals from these months are unlikely to advance a perception—among young adults themselves, the mass public, or political elites—that young adults have what Campbell would call an “identity as a program clientele.” ⁴²⁶

The methods could miss some age references strictly carried in numbers. It could also be that portrayals of young adults were higher in 2010, when eligibility to remain on parental insurance took effect. ⁴²⁷ But even if this is so, the fact that attention to young adults apparently did not persist into 2013–2014 may be a finding in support of Jacobs’ warning that the ACA “may fall short of the mobilizing impacts documented by scholars of policy feedback.” ⁴²⁸

3. What themes that have served as messages about beneficiary “deservingness” in past social policy are most frequent in ACA beneficiary portrayals?

Table 4 suggests that messages about workforce participation, economic self-sufficiency, and insider status, are the key themes or routes through which portrayals may be signaling beneficiary (un)deservingness in the case of the ACA. As detailed above, the Workers tag is the most frequent tag in descriptions of the newly Medicaid-eligible and the subsidy-eligible, and the second most frequent in descriptions of ACA insurance-losers. The Poor/Very Poor tag is one of the three most frequent in portrayals of both the newly Medicaid-eligible and the subsidy-eligible, and the Customers/Clients tag is one of the three most frequent in portrayals of ACA insurance-losers. The Strong Insiders tag is among the three most frequent in portrayals of the newly Medicaid-eligible, ACA insurance-gainers, and ACA insurance-losers.

Other signals about deservingness discussed in part III—such as messages about health, age, gender, and race/ethnicity—appear relatively infrequent in portrayals of ACA beneficiaries, as seen in Table 3. Regarding health, the Bad Health tag applies to less than 4 percent of terms in each DTM, and the Good Health/Able Bodied tag to less than 1 percent of terms in each DTM.

Age-related tags are absent from Table 4. Beyond Young Adults, the Near Elderly tag makes up less than 1 percent of tagged terms in each DTM when using Dictionary 1, even though this group may particularly stand to benefit from ACA regulations. ^429,430 Adults is the most frequent age tag in portrayals of the newly Medicaid-eligible (7.4 percent of tagged terms) and of insurance-gainers (3.2 percent) using that dictionary. And for the subsidy-eligible, Adults and Young Adults appear rarely and with near equal frequency (respectively 1.4 and 1.9 percent of tagged terms).

References to gender are also infrequent. Within the references that exist, across all DTMs, portrayals are somewhat more focused on women than men (see Table 3 for details). Women’s slightly greater frequency in portrayals appears commensurate with women’s actual prevalence as ACA beneficiaries. At least of those who selected an insurance plan through ACA marketplaces during the first open enrollment period, an estimated 54 percent were female. ⁴³¹

Explicit references to race also appear infrequent. As seen in Table 3, across all four DTMs, both White and non-White references are less than 2 percent of tagged terms (and are absent in descriptions of insurance-losers). While potentially related to salience, ^432,433 it is not possible to assess whether portrayals of race/ethnicity are proportionate. Explicit mentions of race/ethnicity appear most in descriptions of insurance-gainers, who are portrayed as more often non-White (1.6 percent of tagged terms) than White (0.2 percent of tagged terms). In California, which as noted accounts for many of the articles and thus plays a large role in the data, a minority (35 percent) of marketplace enrollees were non-Hispanic White. ⁴³⁴ Nationally, HHS reports race/ethnicity information for people who selected insurance on federally-facilitated marketplaces as of May 1, 2014. ⁴³⁵ In those data, of the 69 percent who reported a race, 63 percent identified as White. ⁴³⁶

There are also methodological caveats to assessing the frequency of references to race/ethnicity. The methods may undercount explicit references since none of the dictionaries break out terms tagged as Surnames (see the Appendix) by national origin. ^{437,Reference Enos438} Second, as noted, the White and non-White tags will miss implicit race references—“words that have racial associations but are not racial nouns or adjectives.” ⁴³⁹ Third, as part III notes, scholars find messages about race/ethnicity interlaced with messages about insiders/outsiders and messages about work. As Table 4 underscores, the Workers tag and Strong Insiders tag are among the most frequent in portrayals.

VIII. Implications for durability

Are the patterns these questions have uncovered favorable to the ACA’s durability, as defined at the outset of the paper? This section offers two broad answers for consideration. On the one hand, portrayals appear conducive to durability in that they depict beneficiaries in terms associated with deservingness in past U.S. social policy. On the other hand, there are three reasons for caution. First, portrayals also cast insurance-losers as deserving. Second, it is unclear if the portrayals found will aid, or allow “drift” ⁴⁴⁰ away from, the ACA’s goals of cutting insurance disparities. Research is needed on the implications of the (non)-role of race/ethnicity, along with age and health, in portrayals, and the prominent role of the insider theme. Third, the frequent portrayal of beneficiaries as Workers, coupled with the inattention to young adults and to beneficiary economic heterogeneity, may do little to foster beneficiary political engagement.

IX. Conclusions

If the ACA achieves the goals noted in the Introduction, it may extend the American welfare state’s “scope of risk protection,” ⁵²⁴ and by “reduc[ing] vulnerability” even help address entrenched inequalities. ⁵²⁵ This article therefore takes the premise that the ACA’s durability merits scholarly interest, and—motivated by past research on social policy, welfare state retrenchment, and social constructions that suggests that a policy’s durability relates to depictions of its beneficiaries—takes the premise that ACA beneficiary portrayals merit systematic study. This article first describes those portrayals by answering three questions that the literature connects to durability: (1) Are beneficiaries portrayed as economically heterogeneous? (2) Do portrayals highlight groups to which ACA gives new political meaning? (3) What themes that have served as messages of “deservingness” in the past riddle ACA portrayals? It finds the answer to question (1) to be “no,” with the caveat that the methods may miss some cross-class portrayals. The answer to question (2) also appears to be “no,” with the caveats that young adults may have received more attention in months prior to those studied, and that the methods may miss some references to age. In response to question (3), portrayals most often highlight themes of work, economic self-sufficiency, and insider status. Health status, age, gender, and race/ethnicity appear to receive little attention.

The article then assesses how the patterns uncovered in answering these descriptive questions may work for and against durability. While “[i]t is hazardous to make predictions about how policies will evolve, especially about those with as many moving parts as the ACA,” ⁵²⁶ these findings—assessed through the lens of existing literature—suggest reasons for both confidence and caution about whether portrayals are likely to bolster the ACA’s strength in resisting direct retrenchment and in diligently pursuing goals such as disparity reduction. To recap part VIII, a key result favoring durability is that beneficiaries are depicted in terms associated with deservingness in past U.S. social policy, in being portrayed as workers and insiders. But, there are three potential reasons for caution. First, ACA insurance-losers are also portrayed as deserving. Second, the durability of the ACA’s disparity cutting efforts may merit watching, with an eye to the unclear impacts that the relative absence of age, health, and race/ethnicity in portrayals, and the possibly adverse impacts the Strong Insider theme’s salience in portrayals, may have on this goal. Third, the Worker theme’s salience may, despite its ties to deservingness, join limited attention to young adults and to beneficiaries’ economic heterogeneity, as a portrayal pattern that does little to help the ACA politically mobilize its beneficiaries.

This article is the first to quantitatively describe and analyze media portrayals of ACA beneficiaries. Of studies located through JSTOR and other search strategies in May 2015 (details available by request), only a handful appear likely to speak to ACA beneficiary depictions, ^{Reference Mariner527,Reference Oliver528,Reference Sours529,Reference Hussey and Pearson-Merkowitz530,Reference Leimbigler531} and inspection makes clear that the first three do not. Hussey and Pearson-Merkowitz state that ACA “discourse…did not portray blacks as the likely beneficiaries of national health insurance or health care reform, while it did focus on illegal immigrants.” ⁵³² Their paper points to two New York Times articles exemplifying immigrant-focused discussions, but not to systematic studies of ACA beneficiary portrayals, and the paper itself examines a different question. ⁵³³ Leimbigler’s work shares the present article’s interest in studying how social constructions relate to the ACA’s political strength, ⁵³⁴ discusses political engagement, ⁵³⁵ and in studying “markets” and “rights” frames may trace language similar to that tagged Strong Insider, Cross Class, and Customer/Client here. ⁵³⁶ But Leimbigler’s work studies elite framing, and differs from this article in its goals, ⁵³⁷ methods, ⁵³⁸ and outputs. ⁵³⁹ Barry and colleagues study media messages during ACA implementation but not beneficiary portrayals. ^{Reference Barry540}

Early beneficiary portrayals—such as those studied here—may have uniquely powerful implications, ^541,542 and data of the type gathered here may be a foundation for broaching large questions falling beyond this article, such as whether the ACA is understood as “public assistance” or “social insurance,” a ubiquitous distinction. ^{543,Reference Nelson and Gordon544,Reference Stone, Schneider and Ingram545} As noted, the data may be a basis for studying how the ACA may alter inequalities in political participation. ⁵⁴⁶ The type of data built here may also be a start at answering Jacobs and Mettler’s call for research on the “structural framing” around the ACA, ⁵⁴⁷ meaning “institutionally-based communications of social welfare policies that…frame concrete programmatic returns to individuals.” ⁵⁴⁸

It is important to stress, however, that the research carried out here does not claim to describe framing effects around the ACA or its beneficiaries. It reports the frequency of tags devised to organize carefully coded but atomized terms; “framing effects depend more heavily on the qualities of frames than on their frequency of dissemination.” ⁵⁴⁹ This article does not (and cannot) weight ACA beneficiary portrayals by the degree to which the public, beneficiaries, or policymakers find them convincing. These are empirical questions about which we do not yet have data.

Several other methodological limitations were stated above and are not repeated here. I rather highlight three points about external validity. First, as noted, results pertain to the 50 states in the aggregate; sub-national results may differ. Second, visual images—not studied here—may bear as much or more on beneficiary constructions as do numbers and words. ^{550,551,552,553,554}

Third, the degree to which the results generalize outside the months studied is an open question. One consideration is that the ACA’s implementation timeline may make it logical for the media to generate different portrayal patterns at different times. Another is the partisan discord in the timeframe studied. In November 2013 for instance, the “‘Because of Obamacare…I Lost My Health Insurance”’ booklet met with a White House document entitled “‘G.O.P. Obstruction/Sabotage—Affordable Care Act.”’ ⁵⁵⁵ Representative Frederick Upton (R-MI) charged President Obama with misleading Americans on being able to keep their insurance, and Representative Michael Doyle (D-PA) retorted “‘[d]on’t pretend you care about the American people’s health care here….You just want to repeal the Affordable Care Act.”’ ^{Reference Parker and Pear556} Whether partisan discord impacted media portrayals is a question beyond this paper. So is the question of whether any such impact makes the portrayals found here unique to these months, or rather predictive of future portrayals.

Looking forward, tight public budgets may lead to a sharpening of divisions between those deemed “worthy and unworthy” ⁵⁵⁷ within a policy’s beneficiaries. ^558,559 It will be important to continue to track ACA beneficiary portrayals in the coming years, using multiple methods, to update the answers to the questions broached here, and to continue to assess how the politics around this new policy are shaping its capacity to insure the uninsured, mitigate health-based financial vulnerabilities, cut inter-group disparities, and improve health.

Note

This work was supported, in part, by funds provided by the University of North Carolina at Charlotte. For research assistance, I thank Marc Botero, Brittany Bumgarner, Kim Hill, Jenny Kaemmerlen Kabool, Charlie Lamprecht, Tracy Martin, Allison McMurry Cordell, Erika (Palmer) Ruanne, JoEllen Pope, and Ashlyn Shrewsbury. For information about LexisNexis indexing, I thank Amanda Binder. A course on “Text Mining and Text Analysis with R” taught by Scott Moser in 2013 provided instructive guidance on performing automated text analysis in $\boldsymbol{R}$ . Early versions of this paper were presented at the annual meetings of the American Political Science Association (August 2014), Midwest Political Science Association (April 2014), Northeastern Political Science Association (November 2013), and Southeastern Conference for Public Administration (September 2013). For helpful discussions and comments, I thank other panel participants at these conferences, Bill Brandon, Erik Bucy, Eileen Burgin, Sally Cohen, Nicole Krassas, Ted Marmor, Carl Snook, Robert Sprinkle, Deborah Stone, John Szmer, and Miya Woolfalk. I also thank two anonymous reviewers for helpful suggestions, including but not limited to the suggestion to examine portrayals of those gaining insurance versus those losing insurance under the ACA. Errors are my own. For references citing iPOLL as the distributor, the following acknowledgement applies, as stated by the Roper Center website: ‘‘The survey results reported here were obtained from searches of the iPOLL Databank and other resources provided by the Roper Center for Public Opinion Research, University of Connecticut.” The author’s access to Roper iPOLL was obtained through UNC Charlotte. Online Appendices 1, 2 and 3 are available as supplementary material on Cambridge Journals Online. Data are available from the author.