WHITE RACIAL ATTITUDES AND PUBLIC OPINION

Racial attitudes among White adults have been shown to exert a powerful influence on contemporary American policy attitudes across several policy domains, using experimental and survey-based evidence, and accounting for several non-racial political and demographic factors (see Hutchings and Valentino Reference Hutchings and Valentino2004 for a review). Kinder and Sanders (Reference Kinder and Sanders1996) find that racial divisions are particularly pronounced on issues with explicit racial content. In surveys from 1986–1992, Black respondents are more than 40 percentage points more supportive than White respondents on racial policies from school desegregation, to preferential hiring, and federal spending on programs that assist Black individuals (Kinder and Sanders Reference Kinder and Sanders1996, Table 2.2).

However, the influence of racial attitudes on political preferences also extends into policies without explicit racial content—those with “unstated” or “covert” racial content (Kinder and Sanders Reference Kinder and Sanders1996, 31). For example, Kinder and Sanders (Reference Kinder and Sanders1996) find that Black adults are more than 30 percentage points more supportive of food stamps than White adults, about 20 percentage points more likely to oppose capital punishment, and nearly 20 percentage points more supportive of welfare policies. Importantly, Kinder and Sanders (Reference Kinder and Sanders1996) find that these racial differences, indeed, appear to be related to race—they subsist even when accounting for class differences between White and Black respondents.

To explain racial divisions on non-racial issues, White (Reference White2007) argues that these issues can become racialized in contexts where implicit racial cues prime out-group racial resentment among White individuals. Tesler describes this pattern as the “spillover of racialization” (Tesler Reference Tesler2012; Reference Tesler2016; Tesler and Sears Reference Tesler and Sears2010). Issues without explicit racial content can become racialized when coded racial language, Black imagery, and source cues—such as race as a salient social background characteristic of an elite issue advocate—are connected to the policy in mass communications (Tesler Reference Tesler2012; Reference Tesler2016). Moreover, the relationship between racial attitudes and views on non-racial policies generally seems to be growing. Enders and Scott (Reference Enders and Scott2019) suggest that racial resentment has become an increasingly strong predictor of a wide range of non-racial attitudes since the late 1980s.

BLACK IDENTITY AND PUBLIC OPINION

In addition to racial resentment among White adults, racial considerations among African Americans have also shaped the way in which Black individuals participate in politics and influence voting decisions, party identification, and public opinion. Black adults are highly unified politically, with more than 80% identifying with the Democratic Party. Dawson's (Reference Dawson1994) concept of linked fate, or the Black utility heuristic, attributes this unity to the shared historical experiences of discrimination, segregation, and extensive in-group mechanisms for communication, which have made it more efficient for African Americans to use the welfare of the racial in-group as a shortcut for understanding how political outcomes will affect the individual. Beyond linked fate, more recently, scholars have argued that non-traditional ideologies and strength of identity can help to further explain unified Black political behavior (Dawson Reference Dawson2001; Gay et al. Reference Gay, Hochschild and White2016; Hutchings and Jefferson Reference Hutchings and Jefferson2014). In particular, Hutchings and Jefferson (Reference Hutchings and Jefferson2014) find that Black adults’ preference for a larger role for government, as well as the importance of being Black to an individual's identity predict Democratic party identification.

Strength of in-group identification also helps to explain unified preferences of Black adults across several policy areas with racial content (Dawson Reference Dawson1994; Kinder and Winter Reference Kinder and Winter2001). However, in-group identification cannot explain Black preferences across all policies (Kinder and Winter Reference Kinder and Winter2001; Tate Reference Tate1993). Just as the “spillover of racialization” into preferences among White voters is limited to certain contexts, the activation of racial identity for Black public opinion only occurs when certain conditions are met. Specifically, White (Reference White2007) argues that explicit racial cues that directly implicate “Blacks” or “African Americans” can prime racial identity among Black individuals.

RACIALIZATION OF HEALTH CARE ATTITUDES IN THE OBAMA ERA

Both the theory of racialized spillover and strength of Black in-group identification can help explain why President Barack Obama's tenure in office has corresponded with persisting and sometimes increasing racial divisions in public opinion across a wide range of racial and non-racial policy areas. Hutchings (Reference Hutchings2009) found that the racial gap in opinion on racial policies in 2008 remained about the same as it was in 1988. While strength of identification with one's own racial group among Black adults may not generate the equivalent force for public opinion as the strength of hostility or resentment toward an out-group, both concepts may work in opposing ways to contribute to increasing divisions in public opinion between racial groups.

In non-racial policy areas, scholars have argued that Obama's race has served as a source cue that heightens the impact of racial attitudes among White adults on preferences when non-racial policies are closely connected to Obama as an actor (Tesler Reference Tesler2012; Reference Tesler2015). For example, Tesler (Reference Tesler2012; Reference Tesler2015) finds that racial attitudes predict support and opposition for tax policies that are attributed to Obama in experimental settings. In addition, Tesler finds that among individuals interviewed in the period soon after Obama's announcement of support for same-sex marriage, racial attitudes became more predictive of opposition to the issue.

This “spillover of racialization” became especially pronounced and well-documented on the ACA. In 2009–2010 during the passage of the law, Black respondents were more than 40 percentage points more supportive of the law than White respondents in public opinion polls. This gap is more than 10 percentage points greater than the racial division on health care during Bill Clinton's presidency (Tesler Reference Tesler2012).

It is not surprising that racialized spillover occurred on health reform given Obama's close connection with health reform during his term in office. Even prior to his election as president, Obama spoke in-depth about health care reform on multiple occasions, including two campaign events early in 2008. First, a significant portion of the Democratic primary debate held on February 26, 2008 in Cleveland, OH put a spotlight on Hillary Clinton and Obama's health care plans.Footnote ¹ By 2009–2010, President Obama treated health care as a centerpiece of his first-term agenda—supporting Congress in its passage of the ACA legislation. Finally, post-2012, the ACA has colloquially come to be known as “Obamacare”, further entrenching the close connection between Obama—and the source cue of Obama's race—with health care as an issue.

Obama also explicitly mentioned “Blacks” as potential beneficiaries of health care reform, which could also not only activate White racial resentment but also activate in-group identification among Black voters. For example, on March 18, 2008 in Philadelphia, PA, Obama gave a speech largely focused on race.Footnote ² The speech followed an ABC News report on March 13, 2008 about Obama's then-pastor, Reverend Jeremiah Wright, who delivered remarks that denounced America, in part, due to the country's treatment of Black Americans (Ross and El-Buri Reference Ross and El-Buri2008). Obama's message in the March 18^th speech articulated the need to foster unity across racial lines to help everyone in the country succeed. Throughout the speech, Obama made several explicit mentions of specific racial groups as potential beneficiaries of his policy agenda, including health care. This means that as early as February and March 2008, explicit and implicit cues could have primed racial attitudes on health care among both Black and White adults.

Survey data reveal moments in this timeframe where a racial divide on health reform emerges during the 2008 presidential campaign. Using rolling cross-sectional data from the National Annenberg Election Study (APPC 2010),Footnote ³ Figure 1 displays the difference in the 7-day moving average for support for government involvement in health reform for Black Democrats and White Democrats over time, by days away from March 18, 2008, the date that marked President Obama's speech on race. (For example, Day 0 corresponds to the average from March 15–21, 2008.) Positive values mean that Black Democrats were more supportive of government involvement in health care than White Democrats, on average.

Note: 2008 NAES. Points are the difference in 7-day averages between Black Democratic and White Democratic respondents. Figure displays a loess curve of the averages over time using data from all dates where the survey question is available. Dashed vertical line at “0” represents the day of Barack Obama's speech on race.

Figure 1. Racial divide in support for Government involvement in health care over time

The survey asks respondents whether they would prefer “having one health insurance program covering all Americans that would be administered by the government and paid for by taxpayers” (=1), “keeping the current system where many people get their insurance from private employers and some have no insurance” (=0), or neither system (=.5).Footnote ⁴ At Day 0 in Figure 1, a divergence emerges in the opinion of Black and White Democrats. During the timeframe immediately after March 18th, there is an increase in preference for government health care among Black Democrats and a moderate decrease among White Democrats.Footnote ⁵ These data represent a moment very early in Obama's campaign where public opinion on health care behaves in ways that correspond to Obama's salient association with the issue.

This early racial divide aligns with the findings from Lundberg et al. (Reference Lundberg, Payne, Pasek and Krosnick2017) that racially charged evaluations of Obama emerged prior to the 2008 presidential election and continued to have a strong relationship with attitudes on Obama's performance and policies throughout the president's tenure. Henderson and Hillygus (Reference Henderson and Hillygus2011) similarly find that racial attitudes among White adults predict health care policy attitudes as early as September 2008 and became more impactful by 2010. In a set of experiments, Knowles et al. (Reference Knowles, Lowery and Schaumberg2010) and Maxwell and Shields (Reference Maxwell and Shields2014) also show that the presence of Obama as the source or leader associated with health care reform is what activates the relationship between racial attitudes and health care opposition among White adults.

Thus, a considerable body of work supports the claims that health care is a policy area that experienced a growing racial divide between White and Black adults, that this racial gap in opinion appears to be at least partially tied to negative racial attitudes among White adults (and not reduced to other factors that may be correlated with race), and that Obama's close association with health reform and the ACA appears to correspond with and may partially explain the elevated relationship between racial resentment and attitudes toward health reform among White adults during Obama's tenure in office.

What this study seeks to contribute to the literature is twofold. First, the study evaluates public opinion on health care to assess if health care also became “racialized” among Black adults, in the sense that in-group identification led Black adults to become more supportive of health reform. Just as Obama's persistent, close association with health reform provided cues that activated negative racial attitudes among White individuals, Obama's association with the law and references to how Black Americans may benefit from the law, might have also activated in-group identification among Black adults. Second, I assess if health care opinion appears to become less racialized once a set of conditions that are “ripe for spillover” (Tesler Reference Tesler2012) are less salient in contemporary health reform. In particular, I assess if a new health care policy in the post-Obama era, which is, arguably, less tied to Obama as an actor, appears to have less of a racial divide than the ACA over the same time period.

DATA AND METHODS

The analysis contains two components. In the first component, I assess the relationship between opinion on health reform and race and racial attitudes using two nationally representative American National Election Study (ANES) surveys from 2012 and 2016. The 2012 ANES is a two-wave panel survey containing a pre- and post-election wave in the fall of 2012. The survey includes both face-to-face and online interviews. The 2016 ANES is a similar two-wave survey conducted before and after the 2016 presidential election. The analysis will include only respondents who identify as non-Hispanic White or identify as non-Hispanic Black. In addition, the primary outcome variable was asked in the pre-election wave, but because multiple covariates of interest are asked in the post-election waves, the sample will include only respondents who completed both waves. The resulting sample from 2012 includes 3275 White and 959 Black respondents, and the 2016 sample includes 2631 White and 343 Black respondents. The 2012 ANES had an oversample of Black/African American respondents. Because of this oversample, analyses of Black respondents in 2012 have greater statistical power than analyses from 2016. Analyses of both surveys apply ANES-provided sampling weights.

The primary dependent variable in these analyses is respondents’ favorability toward the ACA. This question was asked during the pre-election waves, only, in 2012 and 2016. The question wording was “Do you favor, oppose, or neither favor nor oppose the health care reform law passed in 2010? This law requires all Americans to buy health insurance and requires health insurance companies to accept everyone.” Respondents were coded on a seven-point scale (scaled from 0 to 1) from 0 = “Oppose, a great deal” to 1 = “Favor, a great deal” with “Neither favor nor oppose”/Don't know as the midpoint of the scale.

The primary independent variables in the analysis are partisanship, race, and racial attitudes. The race measure Black versus White is a dichotomous indicator for whether a respondent identifies as Black (=1) or White (=0). The partisanship measure will rely on the traditional ANES seven-point branching question (scaled from 0 to 1), which classifies respondents from “Strong Democrats” to “Strong Republicans” with “pure” independents and unaffiliated respondents at the midpoint.

To assess racial attitudes among White respondents, the analysis will use a four-question racial resentment measure, scaled from 0 to 1 (see Supplementary Materials for question wording). To assess the strength of identity among Black respondents, I use a measure of linked fate and a measure of identity importance. The linked fate question is on a four-point scale (scaled from 0 to 1) based on the question, “Do you think that what happens generally to Black people in this country will have something to do with what happens in your life? Will it affect you a lot, some, or not very much?” The importance of being Black to identity wording is: “How important is being Black or African-American to your identity?” Responses are on a five-point scale (scaled from 0 to 1 from “Not at all important” to “Extremely important”). In addition to the primary independent variables, the multivariate regression analyses will adjust for several other political, socioeconomic, and health-related covariates described in the Results section.

The second component of the analysis compares public opinion on the ACA to public opinion on the AHCA, a short-lived Republican proposal on health reform in the summer of 2017. To do so, I evaluate three surveys from the Kaiser Family Foundation monthly tracking polls from May, June, and July 2017.Footnote ⁶ These are nationally representative telephone-interview surveys that track public opinion on health reform. The analyses in the results examine these three surveys separately and in a pooled analysis. Both the separate and pooled analyses use sampling weights provided by the Kaiser Family Foundation (KFF) (2017) to weight the data to be nationally representative. These weights were developed to make a single, cross-sectional survey representative, but due to the similarities in the samples over the short timeframe in 2017, I apply the same weights in the pooled analysis.Footnote ⁷

In each of the three KFF polls, respondents were asked questions about both the ACA and the AHCA.Footnote ⁸ Favorability toward both the ACA and AHCA is asked on a five-point scale from “very unfavorable” to “very favorable.” Those who indicate “don't know” are coded at the midpoint of the scale. The variables are rescaled from 0 to 1. For the analysis, the AHCA variable is reverse coded so that the stereotypical most Democratic or liberal position for both measures is “1” (“very favorable” for the ACA, and “very unfavorable” for the AHCA).

Like the analysis of the ANES surveys, the primary explanatory variables in this analysis are race (Black versus White, coded as 1 = Black, 0 = White) and partisanship. The KFF polls do not use the same seven-point branching question for partisanship. Instead, in these analyses, partisanship is on a more limited five-point scale with “pure” independents and unaffiliated respondents as the midpoint. The analyses will also adjust for several political and demographic covariates. However, these surveys do not include measures of racial attitudes, which means the KFF analyses are limited to assessing the racial gap in opinion on the ACA and AHCA and not differences within racial groups. Thus, the analyses using the ANES data from 2012 and 2016 provide more direct evidence on whether racial attitudes and strength of racial identity underlie health care attitudes.

PUBLIC OPINION ON HEALTH REFORM POST-OBAMA

Will the racial divide on health care continue in a post-Obama policy era? To answer this question, the final analysis compares public opinion of the ACA to the AHCA in the summer of 2017. The first set of analyses combine the three KFF surveys in regressions using survey fixed effects. To directly assess if the racial divide is significantly greater for the ACA than the AHCA, I append answers to both questions as the outcomes in a single regression where race (Black = 1 versus White = 0) is interacted with an indicator for whether the respondent's answer is to the ACA (=0) or AHCA (=1) question. (Respondents were asked both questions.)

Table 7 presents the regression results. Model 1 includes a limited set of covariates, while Model 2 adds covariates for health insurance, gender, income, age, education, employment status, ideology, and health. (See full results for Table 7 in Supplementary Materials Table S9.) First, examining the “main effects” of the Black versus White coefficient, consistent with the ANES analyses, the results show that Black survey respondents hold significantly more positive views toward the ACA than White respondents, controlling on partisanship and other factors.

Table 7. Racial gap in attitudes toward AHCA versus ACA

Note: KFF data from May, June, and July 2017. Weighted. ***p < .001, *p < .05.

Moving to the comparison of the ACA and AHCA, the interaction between race and health policy measure is negative and significant. Recall, AHCA favorability is reverse coded, such that higher values for both the ACA and AHCA indicate support for the stereotypical liberal, Democratic position on the issue. To help interpret the interaction results, Figure 2 displays the estimated attitudes toward health reform from Model 1 for four types of respondents: where the covariates are set to a Democrat in the analysis who is Black or White and answering the question about the ACA or AHCA. Black survey respondents are significantly more supportive of the ACA than White respondents. In contrast, White survey respondents hold more unfavorable views of the AHCA than Black respondents. Thus, while Black respondents hold views more stereotypical of liberals and Democrats on the ACA, this does not hold for the AHCA, where the racial gap is smaller overall and in the opposite ideological direction.

Note: KFF data from May, June, and July 2017. Estimates and 95% confidence intervals based on regressions in Model 1 of Table 7. Month of survey fixed to June 2017 for estimated values.

Figure 2. Favorability on the ACA and unfavorability on the AHCA by race

One reason the racial gap may be smaller for the AHCA is that it was a new policy proposal, that was generally less popular, and about which the public had little information and few elite cues. Thus, individuals might not have known how to map predispositions onto their preferences on the AHCA in the same way that they do for the ACA. To assess this possibility, I evaluate the association of another more stable consideration—partisanship—with the ACA versus the AHCA by interacting partisanship with the indicator for whether the response is to the ACA or AHCA in regressions presented in Table 7, columns 3 and 4.

The results show that partisanship is consistently a strong, and significant predictor of support on the ACA (interpreting the main effect of partisanship). Going from a Democrat to a Republican is associated with about a 60-percentage-point shift in ACA support in Model 3 and nearly a 50-percentage-point shift in support in Model 4 (which includes additional covariates). The results do reveal a significantly weaker partisan gap on the AHCA relative to the ACA—5 percentage points less of a partisan gap on the AHCA in Models 3 and 4. Thus, another predisposition shows a diminished association with the AHCA compared to the ACA. However, this represents a 10% reduction of the relationship between partisanship and health care attitudes. In contrast, the racial gap on the ACA is generally negated on the AHCA.Footnote ¹²

The last results disaggregate the data by survey month and health policy measure. Figure 3 displays the coefficient for race (Black versus White) and 95% confidence intervals from regressions of health care views on race and the set of covariates from Table 7, column 2. The separate regression analyses reduce the statistical power to detect significant differences in the relationship between race and health care views on the ACA and AHCA, but will help reveal if the patterns vary greatly across surveys. As Figure 3 shows, the differences between the coefficients on race for the ACA and AHCA are largest in July 2017, but the overall pattern is consistent across surveys. The results show that the racial divide for the ACA is positive (suggesting greater support among Black respondents) for each survey and significantly different from zero in June and July 2017 at the p < .05 level. In contrast, the coefficient on race for the AHCA is generally smaller and is not distinguishable from zero in each survey.

Note: KFF data from May, June, and July 2017. Figure displays coefficient estimates for Black versus White and 95% confidence intervals from separate regressions of ACA favorability and AHCA unfavorability on race (Black versus White) and covariates.

Figure 3. Racial differences in ACA and AHCA opinion by survey