RESEARCHING POLARIZATION AND CULTURE

The language of a culture war was introduced into social science discourse in 1991 in James Davison Hunter’s eponymous book Culture Wars: The Struggle to Define America, one of several decrying American social fragmentation (e.g., Schlesinger 1991). These and similar works argued that multiculturalism as a critical idea and a social fact magnified social groups’ differences, potentially rendering the country ungovernable.

In political science, these concerns have been explored primarily in two diverging bodies of research, one focused on social groups and group-related issues such as multiculturalism, social values, and immigration, and the other emphasizing divisions along the lines of party, initially among elites (especially Congress), and more recently at the mass level. Debate within the second of these research streams was galvanized in 2005 by Culture War? The Myth of a Polarized America (Fiorina , Abrams, and Pope 2005). This book debunked the idea that the mass public was polarized and that polarization was increasing, arguing that apparent mass polarization reflected elite divisions that presented voters with polarized choices, masking the moderate attitudes of the public. Since then, research on mass polarization has largely followed Fiorina et al., defining polarization as partisanship in identification and voting behavior, and exploring how issue positions and deeper values and beliefs widen partisan differences (Abramowitz Reference Abramowitz2011; Abramowitz and Saunders Reference Abramowitz and Saunders2008; Coffey Reference Coffey2011; DiMaggio, Evans, and Bryson Reference DiMaggio, Evans and Bryson1996; Evans Reference Evans2003; Fiorina and Abrams 2008; Hetherington Reference Hetherington2009; Layman, Carsey, and Horowitz Reference Layman, Carsey and Horowitz2006; Masket , Winburn, and Wright 2012; Pew 2012).

Although the term culture wars is used as shorthand for research into mass and elite partisan polarization, analysis of polarization in terms of the first body of research (on major social groups and opinions about them) has been lacking, despite the central role of social groups in the concept of culture (e.g., Dalton Reference Dalton and Almond2000; Layman, Carsey, and Horowitz Reference Layman, Carsey and Horowitz2006). Recent research has investigated individuals’ basic beliefs and values as components of political culture, but most research on these and other attitudinal differences between social groups continues to focus on their effects on partisan divisions (Baker Reference Baker2005; Brewer and Stonecash Reference Brewer and Stonecash2007; Carsey and Layman Reference Carsey and Layman2006). The rare research defining polarization among social groups in terms of fundamental inter-group attitudes, such as identity and antipathy toward groups, typically is limited to a few social cleavages such as race, religion, and ethnocentrism (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012; Kinder and Kam 2009).

Here, I build on these streams of research into polarization and culture by assessing individuals’ evaluations of social groups as measured by their favorability toward social groups. To measure both the degree of polarization among groups and trends in polarization, I use public opinion data collected between 1964 and 2012 by the American National Election Studies (ANES). The ANES assesses group favorability using “feeling thermometers” asking respondents to rate how favorable and “warm” or unfavorable and “cool” they feel toward social groups on five important group cleavages: race, class, age, gender, and religion. I use these favorability or thermometer measures to assess two types of inter-group polarization, building on Hetherington’s (Reference Hetherington2009) distinction between polarization defined as differences of opinion between groups and polarization defined as the effects of group opinion differences. (In discussing the data analysis I use “feelings,” “warmth,” and “favorability” to refer to the feeling thermometer ratings).

My approach is two-fold. First, I analyze direct inter-group antipathy as respondents’ favorability toward social groups, measured separately for members of the groups on different sides of a cleavage, with the size and trends in the differences between group members indicating inter-group polarization. For example, I compare blacks’ and whites’ feelings of favorability toward blacks and toward whites in terms of how positively members of these groups rate their own group (in-group) positively and the other group (out-group) negatively. The size of this gap indexes polarization and trends in the gaps indicate changes in the degree of inter-group polarization. This use of group favorability or “thermometer” questions is consistent with an extensive body of prior attitudinal and electoral research on fundamental inter-group feelings, including research on polarization in issue opinions and partisanship (Brewer and Stonecash Reference Brewer and Stonecash2007; Evans Reference Evans2003; Hetherington Reference Hetherington2009; Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012; Kinder and Kam 2009). For example, Iyengar et al. (Reference Iyengar, Sood and Lelkes2012) found that thermometer ratings of the “out-party” declined from 1978 to 1988 and that party activists were consistently more polarized than nonactivists.

Second, I use the group favorability measures to assess a second type of polarization: the extent to which beliefs about social groups are linked to group members’ partisanship. This analysis calculates the correlation between polarization on the group favorability measures with party identification separately for the different groups on a given social cleavage. The higher the absolute value of the correlation, the stronger the relationship between people’s feelings about groups. Thus, while Hetherington’s measure of impact is the strength of the relationship between ideological and partisan identifications, I operationalize polarization’s impact as the strength of the relationship between group favorability and partisan identification and carry out separate analyses of these relationships for members of the different social groups.

This dual analytic strategy brings culture, in the form of individuals’ favorability or antipathy toward both in-groups and out-groups, back into political analysis of the culture war. It captures both the polarization dimension of opinion divergence between groups used by Evans (Reference Evans2003) and his colleagues (DiMaggio, Evans, and Bryson Reference DiMaggio, Evans and Bryson1996), Abramowitz and Saunders (Reference Abramowitz and Saunders2008), and Fiorina and Abrams (2008), as well as the dimension of “sorting” or “conflict extension” used by Layman et al. (Reference Layman, Carsey and Horowitz2006) and Hetherington (Reference Hetherington2009). The first measure indicates direct antipathy between social groups, and the latter indicates the degree to which group affiliations are linked to partisan divisions. In the data analysis, I assess historical trends and current levels of polarization using both measures.

Because social group polarization can have grave consequences, the direct intergroup favorability measure is important as an index of the degree of social conflict in a society. When social group divisions widen, political parties and other institutions must manage increasing conflict with the potential for violence. If the cleavages in social groups are tightly linked to partisan differences, this robs the party system in particular and the political system in general of its flexibility in forming coalitions and forging compromises to ease conflict (Dalton Reference Dalton2008). If my analysis indicates that intergroup polarization is intensifying and increasingly implicated in partisan divisions, it would support those scholars who claim there is increasing polarization with a potential for system breakdown (e.g., Abramowitz and Saunders Reference Abramowitz and Saunders2008), in contrast to those who argue there is no significant or growing polarization (e.g., Fiorina and Abrams Reference Fiorina and Abrams2008).

DATA ANALYSIS AND DISCUSSION

This analysis employs the ANES time series, which incorporated favorability, or thermometer ratings beginning in 1964. I focus on the period beginning 1972 when more social groups were incorporated into the ANES. Inconsistency from year to year in the inclusion of group favorability measures limits the analysis of the group cleavages of class, age, and gender to favorability toward only one of the two groups on the cleavage. For the racial group cleavage, consistent data on feelings toward both blacks and whites is available, so I construct a difference measure by subtracting the thermometer score for each respondent’s in-group from the rating for their out-group. On the religion cleavage, I separately analyze favorability ratings for the groups asked about (Protestants, Catholics, Jews, and Christian Fundamentalists), broken down by the religious group members identifiable within the limits of the ANES data.^{Footnote 1}

Race

Given the prominent role of racial issues in American politics, it would be surprising not to find sharp polarization. Several writers have argued that racial issues and black-white opinion differences are large, growing, and prime forces in shaping party cleavages (e.g., Carmines and Stimson Reference Carmines and Stimson1989; Gilens Reference Gilens1999; Hutchings and Valentino 2004; Schlesinger 1991).

Figure 1a plots the trends in feelings toward blacks and whites, calculated as the difference in thermometer scores for the two groups. The two lines represent the mean scores on this polarization measure, which range from -100 to 100, with positive scores indicating more positive feelings toward one’s in-group, and negative scores indicating more positive feelings toward one’s out-group. The trend clearly shows decreasing, not increasing polarization, with both blacks and whites becoming more evenly balanced over time in their assessments of the two groups. This net measure of polarization declined from 25 points for both groups in 1964 to about 10–15 points in the 1980s; since 1996 it has been historically low at 3–14 points.

Figure 1a Black-White Thermometers Differenced (1964–2012)

(Color online)

Figure 1b presents trends in the correlation (Pearson’s r) between the thermometer scores presented in figure 1a and party identification. Although people’s feelings toward blacks and whites have become less polarized, they might still contribute to party polarization if these feelings increasingly affect party identification. However, the trend in figure 1b shows little change since 1964 and no evidence of high current polarization. The relationship between group thermometers and party is largely steady among blacks, declining slightly in 1976 and again in 2004, then increasing slightly. Among whites, feeling more positively toward whites was initially weakly associated with Democratic identification, but by 1972 there was no relationship, and beginning in 2000 favoring whites was weakly associated with Republican identification, the same association made previously by blacks. The magnitude of the correlations since 1976 has generally ranged from roughly 0 to 0.15. As a point of comparison, the correlation between an ideological polarization thermometer of liberals and conservatives and party identification was .33 in 1972 and .59 in 2004, and Hetherington (Reference Hetherington2009) reports that the correlation between ideological identification and party identification doubled from 0.28 in 1972 to 0.57 in 2004.^{Footnote 2}

In sum, among blacks and whites, neither the differences in affect toward racial groups, nor the linkage between group affect and party identification has grown substantially. Despite expectations about increasing racial tensions, racial polarization has not occurred.

Social Class and Income

Figures 2a and 2b analyze feelings toward the middle class from 1972 to 1984 and from 2004 to 2012 (the thermometer for “working class” was asked only from 2004 to 2012, and results of data analyses of both measures were similar). Both self-identified working-class and middle-class respondents expressed highly positive feelings, averaging 75 on the 100-point thermometer. These feelings toward the middle class are unrelated to partisanship (figure 2b), with correlations ranging from -.1 to +.1. The results are identical when they are broken down by respondents in the top- and bottom-third of income instead of class.

Figure 1b White-Black Thermometer and Party Identification

(Color online)

Feelings toward poor people were assessed from 1972 to 2012 except 1996 (not shown). Middle-class respondents’ ratings ranged from 67 to 73 and were consistently lower by about five points compared to working-class respondents’ ratings of 72 to 77. Feelings toward poor people were only weakly linked to Democratic Party identification regardless of the social class or income of the respondents, with correlations ranging from about -.2 in the 1980s to about -.1 in the 2000s.

As with race, none of these measures shows high or increasing levels of polarization in either feelings toward social classes or in the linkage between class feelings and partisanship.

Age

Figures 3a and 3b present data on feelings about older or elderly people from 1976 to 2004 (feelings toward young people were asked only in 1972–1980 and 2004). Here again, no evidence of polarization is seen. Feelings are strongly positive among both younger respondents (18–34) and those 65 or older, who are slightly warmer by 3–7 points over the period. The relationship between warmth toward the elderly and partisanship is weak, around -.1 (positive feelings toward the elderly are associated with Democratic identification) for both younger and older respondents. Neither measure indicates growing polarization over time. Analysis of the limited years in which the ANES asked the young people thermometer question indicates similar small gaps in warmth and little effect on partisanship (not shown).

Figure 2a Middle Class Thermometer (1972–1984, 2004–2012)

(Color online)

Gender

Feelings about women and men have seldom been assessed by the ANES. Warmth toward women was asked only in 1976, 1984, 1988, and 2004. As figures 4a and 4b indicate, there is little polarization on this measure, with ratings consistently between 74 and 83, with women slightly warmer. The relationship between thermometer ratings and partisanship is weak and virtually the same for both groups, exceeding .07 only in 1980 for men as shown in figure 4b. As with age, class, and race, polarization as indexed by people’s feelings toward their own social groups or their out-groups, and the impact of those feelings on partisanship is weak and stable.

Figure 2b Correlations (Pearson's r) (Middle Class Thermometer by Party Identification)

(Color online)

Religion

Religion presents the analytic challenge of assessing multiple religious groups that can be (and have been) defined in several ways with inconsistent measures. The ANES has asked feeling thermometer questions about Catholics (1964—2012 except 1980 and 1996), Protestants (1964–76 and 2000), Jews (1964–76, 1988–92, and 2000–08), and Christian Fundamentalists (1988–2012). The ANES has used generally consistent measures for respondents’ affiliations as Catholics and Protestants. However, there have been serious problems and inconsistencies over time in the measures distinguishing Mainline from Fundamentalist Protestants, a key cleavage in recent accounts of values-based issue polarization (Brewer and Stonecash Reference Brewer and Stonecash2007; Evans Reference Evans2003; Layman, Carsey, and Horowitz 2006— see note 1). In this section, I analyze the four religious thermometers in turn, first discussing Protestants’ and Catholics’ responses, then Mainline and Evangelical Protestants when they diverge, with the caveat that the measurement of these two categories mandates caution in interpreting the data, especially after 1996.^{Footnote 3}

Feelings toward Catholics are presented in figure 5a. Catholic respondents’ warmth has ranged from 74 to 80 since 1972, while Protestants’ warmth has grown from the lower to upper 60s, slightly reducing the gap between the groups. As shown in figure 5b, feelings toward Catholics are at best weakly related to partisanship. For Catholics, warmth toward their in-group is slightly associated with Democratic identification after 1976, except in 2004 and 2012. This may reflect the public criticism of 2004 Democratic presidential nominee Senator Kerry and President Obama in 2012 by some Catholic leaders and secular pundits. The opinions of Mainline and Evangelical Protestants (not shown) are nearly identical except in 2000, when warmth toward Catholics is associated with Democratic identification among Mainline Protestants (r+-.15) and Republican identification among Evangelicals (r = .18) Overall, none of these religious groups have feelings toward Catholics that are negative or strongly partisan, nor are these increasing.

Figure 3a Elderly / Older People Feeling Thermometer (1976–1988, 1996–2004)

(Color online)

Figures 6a and 6b present the data on feelings toward Jews. Among both Protestants and Catholics, feelings are positive and nearly identical, ranging from about 60 to 70 (figure 6a). Warmth toward Jews is not related to partisanship among either group in most years, but in 2004 Protestants who feel warmly toward Jews are slightly more Republican. A separate analysis (not shown) indicates this shift occurs among both Mainline and Evangelical Protestants.

Figure 3b Correlations (Pearson's r) Elderly Thermometer by Party Identification

(Color online)

The feeling thermometer toward Protestants was only asked in 1964—1976, and 2000. Protestant respondents are more positive than Catholics by about five points, with ratings above 60 for both groups (figure 7a). For neither group are positive feelings toward Protestants related to partisanship, with correlations hovering near 0 in all three years (figure 7b). Among Mainline and Fundamentalist groups, positive feelings were nearly identical, although in 2000 positive feelings toward Protestants were weakly associated with Democratic identification among Mainline (r=-.10) and Republican identification among Evangelical (r=.07) Protestants.

Figure 4a Women Feeling Thermometer (1976, 1984–1988, and 2004)

(Color online)

Feelings toward Christian Fundamentalists could be expected to evoke stronger and more politically potent feelings, given the attention paid to the political activism of this group and the strong moral traditionalism in political views associated with them. Figure 8a indicates that Protestants feel more warmly (positively) toward Christian Fundamentalists than do Catholics, but the gap has hovered around 10 points, and both groups’ feelings toward Christian Fundamentalists warmed slightly in 2004 and 2008. Figure 8b shows that Catholics’ feelings toward Christian Fundamentalists have no impact on their party identification. Among Protestants warmth was moderately associated with Republican identification in 2004 (r = .19; among white Protestants r = .35) and 2008 (r = .15; among white Protestants r = .27).

This analysis indicates that there is currently a disconnect between partisan and social cleavages; social group animosity and partisan differences do not reinforce or exacerbate one another, rendering prospects for a real “culture war” on the scale of direct conflicts such as the former Yugoslavia or Northern Ireland reassuringly remote.

Figure 4b Correlations (Pearson's r) (Women Thermometer by Party Identification)

(Color online)

For Mainline Protestants (not shown), warmth was related to Republican identification both in 2004 (.28) and 2008 (.10); for Evangelicals the relationship ranged from .12 to .16 from 2004 through 2012. The correlations among white Mainline Protestants ranged from .26 to .35; among White Evangelicals from .32 in 2004 and .27 in 2008, dropping to .12 in 2012. Although these are small effects compared to the impact of ideological thermometers on identification, they are larger than for any other group members on all other group thermometers, and may indicate a slight trend toward greater consistency between feelings toward specific religious groups and partisan identifications. The ANES data for future years may indicate whether this linkage strengthens and extends to other religious groups, stabilizes, or weakens.

Overall, as with the social cleavages of race, class, age, and sex, there is little evidence of sharp polarization across religious groups in terms of either group members’ feelings toward their in-and out-groups or the impact of their feelings about groups on their partisan attachments.

CONCLUSION AND IMPLICATIONS: WHITHER POLARIZATION AND CULTURE WARS?

Is polarization a problem, and has it grown into a culture war? The answer partly depends on how we conceptualize and measure polarization and where we look for its effects. This analysis extended the definition of polarization to incorporate two important, but overlooked, dimensions of political polarization: in-group favoritism and out-group hostility among social groups and their impact on partisanship. By either measure, there is not a high level of polarization on any major social group cleavage in the United States, and polarization has not increased.

This is positive news. Disagreements among the mass public about issues and values, no matter how contested, have not (yet) extended into direct inter-group antipathy. Moreover, basic emotions about social groups are largely unrelated to people’s partisan identification. This analysis indicates that there is currently a disconnect between partisan and social cleavages; social group animosity and partisan differences do not reinforce or exacerbate one another, rendering prospects for a real “culture war” on the scale of direct conflicts such as the former Yugoslavia or Northern Ireland reassuringly remote.

There are minor exceptions, primarily in recent feelings toward Christian Fundamentalists and in the association between feelings toward Fundamentalists and partisanship among Protestants. This sorting could foreshadow future polarization, but as of 2012 the trend was recent, inconsistent, and small, with correlations above .20 only among white Protestants and no attendant change in their mean thermometer ratings.

It is possible that other group attitudes—for example, attitudes toward other racial groups or recent immigrants, or attitudes among smaller social sub-groups—are more polarized and more consequential for party identification. But in this article, I have explored all of the social cleavages for which there is adequate data.^{Footnote 4} Moreover, prior research on issue, ideological, and partisan polarization would lead us to expect the black-white cleavage and the cleavages of class, age, sex, and religion to be the locus of animosity toward social groups and deep-seated beliefs about equality, the role of government, and justice (e.g., Carmines and Stimson Reference Carmines and Stimson1989; Fiorina and Abrams Reference Fiorina and Abrams2008; Hetherington Reference Hetherington2009; Hutchings and Valentino 2004). If a real culture war were raging in the American polity, feelings toward social groups would be tightly linked with both policy and partisan differences.

The results of my analysis complement recent findings that ideological identifications and perhaps issue positions are becoming “sorted” or part of a “conflict extension” that brings them more in line with partisanship and voting and reinforces existing partisan cleavages (Bartels Reference Bartels2006; Fiorina and Abrams 2008; Hetherington Reference Hetherington2009; Layman, Carsey, and Horowitz Reference Layman, Carsey and Horowitz2006; Levendusky Reference Levendusky2009). At the same time, my analysis does not contradict the argument that opinion on issues engaging fundamental values and beliefs have diverged along partisan lines (Abramowitz and Saunders Reference Abramowitz and Saunders2008). Specifically, my findings suggest that sorting and conflict extension may be confined to purely “political” attitudes about policies and issues, and have not become linked to deeper “social” beliefs about groups, which would exacerbate inter-group hostility. Thus, opinions about social groups may be compartmentalized from partisan and issue beliefs during the 1964–2012 period. This is consistent with Iyengar et al.’s (Reference Iyengar, Sood and Lelkes2012) finding that partisan group polarization has increased since 1976 while racial polarization has declined. In the future, if polarization of partisan political beliefs intensifies or becomes closely linked to issues that engage social group cleavages, the current relative peace across group cleavages may break down and become a widening gulf. Yet in my analysis the only hint of such a trend is the recent widening of some religious differences.

Words matter, and the term “culture war” serves us badly on many fronts. Metaphors of war are common currency in descriptions of political election “campaigns” and the “war rooms” from which campaigns “deploy” resources to “battleground” states. The language both debases the seriousness of real physical violence and armed conflict and inflates the sense of gravity about differences in partisan, issue, and social group attitudes among the mass public. The term is also grossly inaccurate, as even the growing issue and ideological differences identified in other research are not particularly deep or intense (cf. Fiorina and Abrams 2008; Hetherington Reference Hetherington2009), and as my results show, do not extend to inter-group hostility.

The martial metaphor may be somewhat useful in distinguishing the responses of political elites from those of the mass public. While partisan leaders in Congress, campaigns, interest groups, and media may issue calls to arms to “culture warriors,” the mass public has not adopted the role of troops taking up arms against one another. The recent sorting among the mass public into more distinctly ideological camps has not degenerated into cultural or civil war. Instead, the intergroup attitudes examined here have been and remain largely civil. This is in keeping with the norms of tolerance and the promise of democratic political regimes to provide an arena in which conflicts originating in civil society can be articulated and resolved without endangering either individuals or society. Perhaps the gap in civility among elites, and between some elites and the mass public, is where we should search for metaphors to characterize cultural differences.

Figure 5a Catholics Feeling Thermometer (1964–1976, 1984–1992, 2000–2012)

Note: See footnote 1 and text regarding measurement issues in 2008 and 2012. (Color online)

Figure 5b Catholics Feeling Thermometer (1964–1976, 1984–1992, 2000–2012)

Note: See footnote 1 and text regarding measurement issues in 2008 and 2012.(Color online)

Figure 6a Jews Feeling Thermometer (1964–1976, 1988–1992, 2000–2008)

Note: See footnote 1 and text regarding measurement issues in 2008. (Color online)

Figure 6b Correlations (Pearon's r) Jews Thermometer and Party Identification

Note: See footnote 1 and text regarding measurement issues in 2008. (Color online)

Figure 7a Protestants Feeling Thermometer (1964–1976, 2000)

(Color online)

Figure 7b Correlations (Pearson's r)(Protestant Thermometer and Party Identification)

(Color online)

Figure 8a Christian Fundamentalists Feeling Thermometer (1988–2012)

Note: See footnote 1 and text regarding measurement issues in 2008 and 2012. (Color online)

Figure 8b Correlation (Pearson's r)(Christian Fundamentalists Thermometer and Party Identification)

Note: See footnote 1 and text regarding measurement issues in 2008 and 2012. (Color online)