BACKGROUND

Elite polarization is a defining feature of contemporary American party politics. Indeed, examinations of congressional roll-call voting,Footnote ¹ campaign donations,Footnote ² party platforms,Footnote ³ candidate positioningFootnote ⁴ and party activists’ issue attitudesFootnote ⁵ all indicate that Democratic and Republican elites are more ideologically divergent from each other now than they were in previous decades. Elite ideological conflict alters citizens’ opinion formationFootnote ⁶ and perceptions of party differences,Footnote ⁷ motivates increasingly negative affect toward the outpartyFootnote ⁸ and lowers voter turnout.Footnote ⁹

A considerable body of evidence also suggests that prolonged and increasing elite polarization has strengthened the relationship between citizens’ ideological and partisan identifications, although scholars debate the precise contribution of partisan ‘sorting’ versus polarization to the observed increase among the mass public.Footnote ¹⁰ Understanding the heightened connection between ideology and partisanship is especially pressing due to these predispositions’ influence on subsequent attitudes and behaviors. In this article, we investigate a theoretically important but often overlooked factor linking individuals’ ideological and partisan identifications: core political values. Core political values influence both ideology and partisanship,Footnote ¹¹ and recent evidence suggests that self-identified liberals and conservatives hold distinct value orientations that contribute to broader divisions in American politics.Footnote ¹² Core values’ central role in the formation of individuals’ ideological identities, as well as the increasing integration of partisanship and ideology among the mass public, leads us to hypothesize that observed ideological and partisan sorting has largely been driven by values.

Examining Center for Political Studies American National Election Studies (ANES) surveys from 1988 to 2012, we show that the increased correlation between citizens’ ideological and partisan identifications observed in recent decades is asymmetric across general value orientations. Specifically, we find that the relationship has strengthened significantly only for individuals who maintain conservative value orientations. Although the association between ideology and partisanship is statistically and substantively significant in every year we analyze for the full mass public sample, the relationship has not changed over time for individuals who maintain liberal value orientations. The findings offer two significant contributions to our understanding of public opinion in the age of polarization.

First, the results demonstrate that the increasingly close connection between ideological and partisan identifications observed in the past two decades is in fact rooted in value polarization among the mass public, and that those holding conservative values are driving this polarization. Secondly, and relatedly, this asymmetry further highlights the impact of the symbolic content of Republican elites’ rhetoric on self-identified conservatives.Footnote ¹³ Namely, this rhetoric has enabled those in the mass public who maintain conservative values to connect their ideological and partisan attachments to a greater extent than their liberal counterparts. Values are the ‘glue’ of the political translation process that binds together citizens’ political attachments, and Republican elites’ communication strategy has strengthened these bonds for conservative individuals.

THEORETICAL FRAMEWORK

The starting point for our inquiry is the well-documented, increasingly close connection between individuals’ self-identified ideology and partisanship.Footnote ¹⁴ Although the changing empirical relationship is uncontested, its causes and meaning are debated. The familiar disagreement situates scholars who argue that the heightened correlation reflects deep and enduring ideological differences in the electorateFootnote ¹⁵ against those who believe that the increase is owed to partisan ‘sorting’, or the result of an elite realignment that now better matches liberals and conservatives to their natural partisan ‘homes’.Footnote ¹⁶ In this article, we endeavor to unpack the relationship between citizens’ ideological and partisan identifications and thus heed Levendusky’s call to ‘think about how citizens change and why this matters’.Footnote ¹⁷

We argue that if the resurgent association between ideology and partisanship reflects meaningful differences in citizens’ worldviews, then we should observe a conditional relationship among values, ideology and partisanship over time, as values represent fundamental political principles that shape individuals’ subsequent political attitudes, attachments and behavior.Footnote ¹⁸ If values play this hypothesized moderating role, then the sorting of individuals into properly matched ideological and partisan camps should not be considered benign because the sorting process involves dividing the electorate along durable and principled fault lines.Footnote ¹⁹

According to Rokeach, a value is ‘an enduring belief that a specific mode of conduct or end-state of existence is personally or socially preferable to an opposite or converse mode of conduct or end-state of existence’.Footnote ²⁰ Furthermore, Rokeach asserts that values are stable, have their origins in culture and society and animate and constrain attitudes across virtually all realms of human thought and behavior.Footnote ²¹ Subsequent work in the American context shows that values relate to symbolic political predispositionsFootnote ²² and influence citizens’ attitudes toward an array of social welfare issues,Footnote ²³ foreign policy alternatives,Footnote ²⁴ racial issuesFootnote ²⁵ and cultural questions such as gay rightsFootnote ²⁶ and abortion.Footnote ²⁷ Values further shape individuals’ candidate evaluationsFootnote ²⁸ and tolerance judgments.Footnote ²⁹ Footnote ³⁰

Most importantly for this article, value orientations also influence citizens’ ideological and partisan identities.Footnote ³¹ Core political values, especially the principles of resistance toward change and acceptance of social inequality, are widely understood to underlie liberal–conservative ideology.Footnote ³² These orientations further relate to liberals’ and conservatives’ disparate tolerance of uncertainty, threat and ambiguity,Footnote ³³ testifying to the social psychological bases of ideology.Footnote ³⁴ Values, then, in the language of Converse, are the ‘crowning postures’ of belief systems that ‘serve as sort of a glue to bind together many more specific attitudes’.Footnote ³⁵ Therefore, precisely because values are central to individuals’ belief systems and provide a basis for goal-oriented thought and behavior, to the degree that the increasingly close connection between ideological self-identifications and partisanship reflects mass public polarization, the relationship should have its origins in core value differences.

Our hypothesis regarding the role of values in the relationship between ideology and partisanship is further based on the dissemination and reception of elite cues in a polarized era. Elite party cues are critical to the formation and expression of mass public attitudes and evaluations of political stimuli.Footnote ³⁶ As a result of polarization, American political elites now provide citizens with ideologically clearer and more consistent cuesFootnote ³⁷ that are increasingly imbued with the language of values.Footnote ³⁸ On account of this polarized rhetoric, in addition to influencing individuals’ social welfare, cultural and foreign policy positions, political values now also divide public attitudes toward contemporary controversies such as voter identification lawsFootnote ³⁹ and health care reform.Footnote ⁴⁰ Evidence also suggests that political values contribute to attitudinal constraintFootnote ⁴¹ and ‘reinforce ideological distinctions’ in the electorate.Footnote ⁴²

Given the role of elite cues in mass public opinion formation and the importance of values for an array of citizens’ attitudes and identities, we argue that the existence of increasingly polarized elite position taking across all issue domains should help foster connections between individuals’ most fundamental political predispositions – core values – and their ideological and partisan attachments.Footnote ⁴³ As a result, we expect that core political value orientations moderate the observed heightened correlation between symbolic ideology and partisanship. More specifically, given the asymmetry of elite polarization and the distinctiveness of Republican and Democratic messaging, we argue that individuals with conservative value orientations are most likely to receive elite cues linking values to subsequent predispositions.

This hypothesized heterogeneity in the reception and activation of cues linking core values to subsequent orientations builds upon evidence suggesting that Republican elites communicate in symbolic terms more often than their Democratic counterparts.Footnote ⁴⁴ Republican elite rhetoric evokes abstract principles in favor of protecting personal freedom and traditional family arrangements, as well as reducing the role of government in the economy, whereas Democratic arguments tout concrete policy proposals and endorse particular programs.Footnote ⁴⁵ Republicans’ focus on ideological principles over specific policies to garner political support extends to claimed electoral mandates,Footnote ⁴⁶ campaign speeches, party platformsFootnote ⁴⁷ and State of the Union speeches.Footnote ⁴⁸ As a result of this consistently abstract messaging strategy and the empirical observation that Republicans in Congress are more ideologically extreme than Democrats,Footnote ⁴⁹ Republican partisans in the mass public are more ideologically orientated and aware of party differences than their Democratic counterparts.Footnote ⁵⁰ Republicans are also more likely than Democrats to prefer ideological purity to moderation and upholding policy principles to political compromise.Footnote ⁵¹ Taken together, the importance of values and elite cues for mass public opinion formation, the more frequent use of abstract rhetoric among Republican elites and the greater reliance on values and ideological symbolism among Republican identifiers in the mass public lead us to hypothesize that individuals with conservative value orientations drive the observed over-time increase in the correlation between ideological and partisan identifications.

DATA AND METHOD

In order to unpack over-time changes in the relationship between ideology and partisanship, we analyze pooled ANES surveys from 1988 to 2012 using the ANES Cumulative File. We begin with 1988 because the values batteries needed to test our key hypothesis regarding the role of core political values in linking ideology to partisanship were introduced in that year.Footnote ⁵² Fortunately, this starting point allows us to examine the relationship before the most vociferous elite polarization began.Footnote ⁵³ The two core political values that will be combined to serve as our primary independent variable are egalitarianism and moral traditionalism. Although agreement regarding precisely what values to analyze in studies of public opinion has been elusive,Footnote ⁵⁴ these two core political values represent long-standing and fundamental sources of political cleavage in American politics.

Egalitarianism captures the American cultural tension between the need for the government to ameliorate social and economic inequality, on the one hand, and an aversion to state interference in economic affairs and the belief that individuals’ lot in life is a product of their own choices, on the other.Footnote ⁵⁵ Moral traditionalism measures individuals’ preference for traditional social and family arrangements and aversion to change.Footnote ⁵⁶ Egalitarians are more likely to support social welfare spendingFootnote ⁵⁷ and policies to promote racial equality,Footnote ⁵⁸ and to identify as liberals and Democrats,Footnote ⁵⁹ than their less egalitarian counterparts. Moral traditionalists are more likely than other citizens to oppose abortionFootnote ⁶⁰ and gay marriage,Footnote ⁶¹ and to identify as conservatives and Republicans.Footnote ⁶² Recent scholarship thus demonstrates that equality and morality not only provide a basis for citizens’ orientations toward the political world, but also that the two values are opposed to one another.Footnote ⁶³

Given that both values theoretically and empirically exist at opposite ends of the same continuum and relate significantly to ideology and partisanship, we combine individuals’ responses to the standard six-item egalitarianism and four-item moral traditionalism batteries in order to create a core values scale in which lower scores represent more liberal value orientations and higher scores represent more conservative value orientations.Footnote ⁶⁴ The resulting scale provides a reliable measure (α=0.744) that we will use to test the potential over-time impact of these two core political values on the association between citizens’ ideological and partisan attachments.Footnote ⁶⁵

Examining first the bivariate correlation between ideological and partisan identifications over time is useful for understanding trends in the relationship during the period under investigation. Moreover, the potential endogenous nature of these predispositions’ relationship complicates a causal estimate of one’s effect on the other. Note that partisanship is measured using the standard seven-point ANES scale ranging from ‘strong Democrat’ to ‘strong Republican’. Ideology is measured similarly from ‘extremely liberal’ to ‘extremely conservative’. Figure 1 shows their correlation for each of the seven presidential election years that we examine. The evidence presented in the figure confirms what several previous studies have demonstrated:Footnote ⁶⁶ The correlation between citizens’ liberal–conservative identifications and partisanship has increased markedly over the course of the past generation, growing from approximately 0.40 in 1988 to over 0.60 in 2012. However, we argue that this correlation tells an incomplete story of the relationship between individuals’ ideological and partisan attachments. If our theory is correct, then we should observe a conditional relationship that is moderated by the core political values of egalitarianism and moral traditionalism. Additionally, given Republican elites’ greater emphasis on symbolic and principled rhetoric compared to Democratic elites, we expect that the correlation between ideology and partisanship will have increased more for individuals with conservative value orientations. Below, we test these hypotheses empirically.

Fig 1 Correlation between ideology and partisanship, 1988–2012 Note: vertical lines represent 95 per cent confidence intervals.

In order to test our two hypotheses, we code respondents who score one or more standard deviations below and above the mean of our core political values scale as maintaining liberal and conservative value orientations, respectively. Figure 2 presents the correlation between ideology and partisanship separately for these two groups for the years 1988 to 2012. The correlations for individuals with conservative value orientations – that is, anti-egalitarian, morally traditional citizens – are shown in closed circles for each year, whereas individuals who maintain liberal value orientations are shown in open circles. The differences between the groups are immediately apparent. The correlation for each group is nearly identical and close to 0.30 in 1988, and the correlation increases modestly for both groups in 1992. However, as elite polarization further escalates in 1996, the correlation increases precipitously for those with conservative values, and it remains close to, or above, 0.60 through 2012. The story is strikingly different for those with liberal values. Instead of increasing apace through time, the correlation remains flat before fluctuating in 2008 and 2012 and never reaches beyond 0.40. The results provide support for our hypothesis that the over-time strengthening of the relationship between ideological and partisan identities is conditional on core values and driven by individuals with conservative value orientations. Next, we turn toward a regression analysis in order to ensure that the growing and asymmetric over-time influence of values on the association between ideology and partisanship observed in Figure 2 is a product value differences, rather than other individual-level social or political attitudes, beliefs or characteristics.

Fig. 2 Correlation between ideology and partisanship conditioned on core values, 1988–2012 Note: vertical lines represent 95 per cent confidence intervals.

The dependent variable in this model is Partisanship. However, we certainly understand and appreciate the primacy of partisanship in models of political attitudes, evaluations and behavior.Footnote ⁶⁷ Our modeling choice is driven by two considerations. First, we wish to reiterate that our interest in this article is to examine the conditional and controlled correlation between ideological and partisan predispositions. As a much larger literature investigates the determinants of partisanship relative to ideology, specifying partisanship as the dependent variable offers us better direction regarding the appropriate variables to include in our statistical model. The regression model then attempts to control for these other variables in order to ensure that the conditional correlations presented in Figure 2 are not spurious. Secondly, recent evidence suggests that symbolic ideology influences partisanship more than the reverse.Footnote ⁶⁸

We specify Ideology and Core Political Values as independent variables, and include retrospective economic evaluations in the model given their potential influence on partisanship.Footnote ⁶⁹ We also include an interaction between retrospective economic evaluations and a dummy variable coded 1 if a Democrat occupies the White House at the time of the election, as party control should condition the effect of economic retrospections.Footnote ⁷⁰ We also include in the model a two-item scale combining biblical literalism and church attendance because greater doctrinal orthodoxy and religious commitment are associated with Republican identification.Footnote ⁷¹ We also control for a host of demographic predictors of partisanship. We include gender, coded 1 for females, as women are more likely than men to identity as Democrats;Footnote ⁷² race, coded into four dummy variables representing whites, blacks, Latinos and individuals of mixed and other races, respectively, with whites representing the omitted reference category because they are expected to be more Republican than individuals of other races.Footnote ⁷³ We also include age, measured in years; education, a four-point, ordinal variable ranging from less than high school to an advanced degree; income, because individuals with higher incomes are more likely to identify as Republicans than their less wealthy counterparts;Footnote ⁷⁴ marital status, coded 1 for married, as married people are more likely to vote Republican than single individuals;Footnote ⁷⁵ union status, coded 1 if the respondent or anyone in his or her household is a member of a labor union, because union families are more likely to be Democrats;Footnote ⁷⁶ and a South dummy variable, coded 1 for respondents living in that region because the remnants of the Solid South might result in Southerners identifying more as Democrats relative to citizens elsewhere.Footnote ⁷⁷

The key relationship in our analysis is a three-way interaction involving ideology, values and time. Although three-way interactions can be difficult to interpret, we believe the task is made easier here due to the fact that one of the variables in this critical interaction is time. We model time using a variable that is coded 0 to 6 for each of the seven survey years included in our analysis. Thus the variable takes a value of 0 for 1988 and a value of 6 for 2012. The coefficient for the three-way interaction term will show the potential over-time increase in the effect of core political values on the relationship between ideology and partisanship. That is, core values are hypothesized to moderate the association between ideology and partisanship, and this moderating effect is expected to heighten over time, consistent with the polarization hypothesis evidenced in much prior research and confirmed in Figures 1 and 2. Table 1 shows the results of our model predicting partisanship, where cell entries are ordinary least squares regression coefficients, standard errors and two-sided p-values, respectively.Footnote ⁷⁸

Table 1 Predicting Partisanship, 1988–2012 ANES

The first crucial point to note in Table 1 is that the constitutive terms comprising the three-way interaction – ideology, value orientations and time – are interpretable only when the other variables included in the interaction are equal to zero.Footnote ⁷⁹ Therefore, Table 1 shows that ideology’s relationship to partisanship is positive and statistically significant for individuals with the most liberal value orientations in the year 1988. As expected, even before polarization accelerated sharply, more conservative respondents were more likely to identify as Republicans. Similarly, the influence of values on party identification is significant for the most liberal respondents in that year (that is, respondents who maintained conservative value orientations were more likely than other self-identified liberals to identity as Republicans in 1988).

Two of the two-way interactions are also notable. First, the non-significant ideology and value orientations interaction indicates that core political values did not moderate the relationship between ideology and partisanship in 1988, a result confirming the evidence that we presented in Figure 2, which shows no difference among individuals with liberal and conservative value orientations in that year. Secondly, the non-significant interaction between the ideology and time variables indicates that ideology has not become a stronger predictor of partisanship over time for individuals who maintain the most liberal value orientations.Footnote ⁸⁰

All of the control variables are significant and in the expected direction, save for the non-significant effects of religiosity and being married. The coefficient for the retrospective economic evaluations constitutive term is expectedly positive and significant, meaning that individuals who believe the economy has improved over the past four years are more likely to identify as Republicans when the incumbent president represents the Republican Party, controlling for ideology, values and other factors. Conversely, but also as expected, the coefficient for the interaction term between retrospective economic evaluations and the Democratic president variable is negative and significant, indicating that individuals are more likely to identify as Democrats if they perceive that the economy has improved during a Democratic chief executive’s tenure. Lastly, the positive and significant impact of the Democratic president constitutive term indicates that individuals who perceive the economy as having worsened during a Democratic presidency are more likely to identify as Republicans. All of these results indicate that favorable economic conditions improve the partisan balance in the electorate for the president’s party.Footnote ⁸¹ This result is noteworthy for our purposes because it highlights the independent influence of ideology, values and issues on citizens’ partisan attachments. If these orientations were identical or otherwise interchangeable, then we likely would not observe the distinct effects for each that we do here.

We now turn to the key relationship in the model, the three-way interaction involving ideology, core values and time. Although exceedingly difficult to interpret from Table 1 alone, the statistically significant coefficient indicates that a significant three-way interaction exists over some range of the moderator variable.Footnote ⁸² Of course, understanding the nature and extent of the three-way interaction is made much easier through a graphical presentation of the relationship. We show this relationship in Figure 3.

Fig. 3 Estimated marginal effect of ideology on partisanship conditioned on core values, 1988–2012 Note: shaded areas represent 95 per cent confidence intervals

Figure 3 shows the marginal effect of ideological self-identifications on partisanship for individuals with liberal (dashed line) and conservative (solid line) value orientations separately over time. The first aspect to note about the figure is that ideology is significantly related to partisanship for both groups in every year, as we would expect given the correlations presented in Figure 2. Secondly, in the early years of our analysis, beginning in 1988, there are no significant differences among individuals with liberal versus conservative value orientations. Core value differences did not connect citizens’ ideological and partisan identifications differently for citizens with liberal versus conservative value orientations before elite polarization became fully entrenched in American politics. However, the situation has changed substantially in the age of polarization. As elites have diverged over time, the role of conservative value orientations in the relationship between ideology and partisanship has increased dramatically, consistent with arguments that the substance and style of Republican elite rhetoric has helped conservative members of the mass public better connect their core values to subsequent predispositions in a polarized era. The result also importantly shows that the relationship between ideology and partisanship has remained nearly unchanged over time for individuals with liberal value orientations. Statistically significant differences in the marginal effect of ideology on partisanship across the two groups emerged in 1996 and have persisted thereafter.

A central component of our argument is that elite polarization drives both the increasingly strong empirical relationship observed between ideology and partisanship, and serves as the moderating effect of core political value orientations on this relationship. Given the existing evidence of growing elite polarization over time, we have used presidential election year as a proxy for elite polarization for ease of exposition and presentation. However, in order to ensure that election year represents a valid proxy for elite polarization, we also specify a regression analysis in which we substituted the time variable for the difference in the estimated first dimension DW-NOMINATE House chamber median between Republicans and Democrats for each Congress in the period that we examine.Footnote ⁸³ Thus, the estimated chamber medians are for the 100th, 102nd, 104th, 106th, 108th and 112th Congresses.Footnote ⁸⁴

Figure 4 shows that the three-way interaction involving ideology, core political values and elite polarization is substantively identical to that presented in Table 1 and Figure 3. Namely, the result both supports our hypotheses concerning the nature of the moderating effect of core values on the relationship between ideology and partisanship and illustrates that time is an excellent proxy for elite polarization in contemporary American politics.

Fig. 4 Estimated marginal effect of ideology on partisanship conditioned on core values using a direct measure of elite polarization Note: DW-NOMINATE scores, 1988–2012 ANES.

Overall, the results presented in Table 1 and Figures 2, 3 and 4 highlight the key moderating influence of core political values on the observed over-time strengthening of the relationship between ideology and partisanship. Although Americans’ ideological self-identifications correlate more tightly with their party identification today than in the past, our results show that this increase is confined to citizens who maintain conservative value orientations – that is, those who oppose efforts to reduce social and economic inequality and who favor traditional social arrangements. The association between ideology and partisanship has not changed over the past generation for individuals with liberal values. We conclude with a discussion of the implications of this conditional relationship for American public opinion and polarization.

Despite the above evidence, some readers might be concerned that our regression model results presented in Table 1 and Figures 3 and 4 are biased due to endogeneity because the dependent variable in the model, Partisanship, might cause individuals’ symbolic identification.Footnote ⁸⁵ Thus, Figure 5 presents a partial correlation plot between ideological self-identifications and partisanship conditioned on core values. Importantly, the partial correlations estimate the relationship between ideological and partisan attachments for individuals maintaining liberal and conservative value orientations, respectively, controlling for all of the other covariates featured in Table 1. The estimated partial correlations are significant for both groups in each election year that we examine, but the differences across groups are apparent and consistent with the evidence presented Table 1 and Figures 2–4. Namely, the estimated correlation between individuals’ ideological and partisan attachments has increased over time more among individuals maintaining conservative value orientations relative to those maintaining liberal value orientations. Thus, Figure 5 confirms the key finding reported in this section that the association between ideological and partisan predispositions is conditioned by core values and is asymmetric across liberal and conservative value holders.

Fig. 5 Partial correlation between ideology and partisanship conditioned on core values, controlling for other covariates Note: 1988–2012 ANES

DISCUSSION AND CONCLUSION

In his 1988 acceptance speech at the Republican National Convention, George H. W. Bush told the American people, ‘An election that is about ideas and values is also about philosophy. And I have one.’ Bush’s philosophy was one of community and moral traditionalism, and his message was that Republican presidents sought to be judged by the American people according to candidates’ core values. In Bill Clinton’s acceptance speech at the Democratic National Convention in 1992, he argued emphatically, ‘The thing that makes me angriest about what has gone wrong in the last twelve years is that our government has lost touch with our values, while our politicians continue to shout about them. I’m tired of it!’ In 2000, George W. Bush stated in his speech accepting his party’s nomination that Americans ‘know we must renew our values to restore our country’. Finally, in 2008, Barack Obama noted that Republican claims regarding the ‘abandonment of traditional values’ were nothing but ‘stale tactics to scare voters’. The message is clear: Conservative Republicans are instructed to ‘vote values’, whereas liberal Democrats should not be duped by appeals to moral traditionalism and the ‘loss of community’.Footnote ⁸⁶

Core political values are central to American politics and political identity, and they have only become more important in the age of polarization. Elite partisan rhetoric in recent decades consistently demands that values be the standard by which the mass public should understand and evaluate the political world. Considerable political science evidence shows that individuals better align their ideological and partisan identities today than in the past, but existing research does not sufficiently establish the role of core values in strengthening the relationship between ideological and partisan identifications.

The motivation for this article was to explain the demonstrable over-time change observed in the relationship between citizens’ ideological and partisan attachments. We necessarily approached the question against the backdrop of the intense and ongoing elite polarization that has profoundly altered the American political system, and we wondered specifically whether we might be able to uncover the extent to which the heightened relationship under study reflects value divisiveness. Our results hold important implications for our understanding of polarization in the mass pubic and, more generally, the role of core values in shaping citizens’ approach to politics.

Our first reported finding confirmed what many scholars before us have evidenced: The correlation between symbolic ideology and party identification among the mass public has increased significantly since the late 1980s. Owing to the improved clarity of elite cues, individuals who self-identify as ideologically conservative or liberal now better match these orientations to their Republican and Democratic party affiliations, respectively.Footnote ⁸⁷ However, this result ignores the moderating role of core values in the basic relationship between ideology and partisanship.

We hypothesized in this article that if polarizing elite cues foster closer connections among citizens’ political predispositions, then citizens should also use their fundamental beliefs about what constitutes a good and bad society to form ideological and partisan attachments, as we argued that values are the glue that holds together political thinking. We found empirical support for precisely this hypothesis. Core political value orientations – specifically, egalitarianism and moral traditionalism – condition the relationship between ideology and partisanship. Furthermore, the expected asymmetry of this relationship is consistent with a wealth of evidence showing that Republican elites communicate more abstractly and symbolically than their Democratic counterparts. Our findings show that this critical difference in rhetorical style – extending throughout the major party organizations from the campaign trail to the halls of Congress to activists and donor networks – is important for understanding mass public opinion heterogeneity.

Conservative messaging enables citizens who share conservative values to bring their ideological and partisan attachments into closer alignment. These individuals are significantly better today than a generation ago at linking their bedrock political values to subsequent orientations, and thus determining ‘what goes with what’ in the political realm. Individuals with liberal value orientations, however, are met more often with messages emphasizing concrete policy proposals and appeals to specific group interests rather than with broader defenses of abstract principles.Footnote ⁸⁸ Consequently, these individuals do not bring their core values to bear in forming ideological and partisan attachments any more relative to the pre-polarization period.

We believe that our results provide support for the mass polarization thesis, but with an important caveat. Our findings suggest that the over-time increase in the strength of the relationship between ideology and partisanship is rooted in genuine value differences. Thus even if elite rhetoric serves to sort individuals by their ideology into proper partisan homes, then we show that this sorting is driven by core political values, the fundamental principles that shape humans’ conceptions of good and bad in the political sphere. This evidence comports with recent work highlighting the value cleavages that exist among Democrats and Republicans. As Jacoby notes, ‘It seems reasonable to characterize such sharp differences in feelings about fundamental values as the existence of a culture war.’Footnote ⁸⁹ However – and here is the important caveat – our results also reveal that asymmetric elite polarization has confined this polarized sorting to individuals who maintain conservative orientations. The relationship between ideological and partisan identities does not appear to have changed among other citizens in the twenty-four-year period that we examine.

We have also raised more questions that necessarily must be left to future research to answer. For example, what is the impact of specific rhetoric, or the precise timing of elite communication, on observed changes among the mass public? A content analysis might be conducted to connect modifications in elite language use during this period to the changing relationships that we investigate in this article. Additionally, an experiment could likely be designed to tease out more carefully the effects of the asymmetric nature and reception of elite cues on the role of values in linking ideology and partisanship. Similarly, an experiment could be conducted to estimate the influence of other contextual influences, such as the presence or absence of specific political figures or issue mentions, on the strength of the relationship between individuals’ ideological and partisan attachments. Despite these and numerous other unresolved questions, we nonetheless argue that our results contribute fruitfully to an ongoing agenda assessing the effects of heightened elite conflict on American citizens’ political attitudes, beliefs and core political values.