Assessing the Political Distinctiveness of White Millennials: How Race and Generation Shape Racial and Political Attitudes in a Changing America

White Americans will soon lose their majority status—news that provokes group threat and a conservative response. Yet an alternative outcome focused on white millennials is also possible. This study examines whether young whites are distinct in their racial attitudes and how they react to demographic change. Using two nationally representative surveys from 2012 and2016 and a nationally representative experiment from 2016, we find that race affects attitudes more than generation, and in no case are white millennials as racially liberal as nonwhites. Exposure to information about changing demography makes white millennials more conservative on some questions, but what matters more is whether respondents are Republicans and identify as white. White millennials are hardly immune to the power of race to shape their attitudes.

Reading the news these days without seeing stories about race relations, immigration, or concerns about America's changing identity is difficult at best. The impending loss of majority status for whites in the United States often looms large in these stories, which feature headlines such as "It's Official: Minority Babies Are the Majority Among the Nation's Infants" (Cohn 2016) and "For First Time, Minority Students Expected to Be Majority in U.S. Public Schools this Fall" (Strauss 2014). In light of these changes, an unprecedented research question has emerged: how are white Americans reacting to the predicted loss of their majority status? Extant research concludes that this news drives a social and political wedge between whites and nonwhites. For example, exposure to information about these population changes can cause whites to become more conservative, less tolerant of diversity, and more likely to think that being white is important to their sense of self (Outten et al. 2012; Hutchings et al. 2012; Danbold and Huo 2014; Jardina 2014; Craig and Richeson 2014b). These changing demographics and the threat they pose to the status of whites as the majority have been cited as key drivers of the success of Donald Trump's candidacy for president (Tesler and Sides 2016; Major, Blodorn, and Blascovich 2016; Sides, Tesler, and Vavreck 2018).

Over the past decade, however, a more optimistic story line about population change has also emerged, one focused on the so-called millennial generation, which consists of those Americans born after 1980 (Taylor 2014). The millennial narrative is more hopeful, portraying a group of young whites who are much more comfortable with diversity than their predecessors (Madland and Teixeira 2009; Teixeira 2011; Winograd and Hais 2011; Marcotte 2013; Feldmann 2014; Cillizza 2014; Tierney 2014; Simmons-Duffin 2014). Their alleged comfort with diversity is not due solely to millennials' being young; it is also a characteristic attributed to the diversity of the generation itself. As David Madland and Ruy Teixiera write, "Because of their diversity, Millennials' attitudes about and experiences with race are dramatically different from earlier generations . . . For millennials, race is "no big deal," an attitude that will increasingly characterize society as a whole as the millennials age and our march toward a majority-minority nation continues" (2009, 11). Although not stated directly, what underlies this expectation is the idea that increased diversity will promote increased contact with ethnic outgroups, which in turn will improve intergroup attitudes in the foreseeable future. This optimistic set of expectations has a long history in social science and is known as contact theory (Allport 1954). In the world of electoral politics, this portrait of a young generation comfortable with diversity fuels a narrative that the Republican Party has a "young person" problem, and comes with warnings that the Republican Party will age itself out of existence if it does not do more to appeal to the nation's young adults and their more racially liberal policy preferences (Rampell 2016; Cillizza 2014; Brownstein 2016).

If white millennials are as progressive as popular portrayals indicate, particularly with respect to racial issues, then white youth might be less susceptible to well-known psychological tendencies associated with racial group membership than whites in other generations, who have been shown to feel threatened by the nation's changing demography (see, for example, Craig and Richeson 2014a, also 2018). The goal of this study was therefore to assess claims made about white millennials. In particular, we examined whether white millennials are in fact more politically and racially liberal than older whites. We examine this question two ways. First, we used the 2012 and 2016 Cooperative Congressional Election Study (CCES) to compare the political views of white millennials to those of nonwhite millennials, older whites, and older nonwhites. In this portion of the study, we paid particular attention to questions about race, immigration, ideology, and partisanship. We then report the results of an experiment designed to examine whether white millennials and nonmillennials reacted similarly to information about demographic trends in the United States. This experiment built on important investigations conducted by Craig and Richeson (2014a, 2014b). Dependent variables in this portion of the analysis include attitudes about immigration policy, affirmative action, and evaluations of ethnic outgroups.

We find that white millennials are slightly more liberal than older whites, but that their views are closer to the views of older whites than they are to those of nonwhite millennials. We also find that white millennials are not more "immune" to the ways in which information about demographic change can promote more conservative attitudes. When such information moves attitudes in a more conservative direction, it does so for the young and old alike. We conclude that, on the issues examined here, race is a more powerful determinant of political preferences than generation. Moreover, the 2016 presidential election does not appear to have altered this pattern. Donald Trump's leadership of the Republican Party and his racially antagonistic campaign did not have a unique effect on young whites; Trump does not appear to have driven them toward or away from conservative policy stances, despite claims that his issue positions and rhetoric would drive them to the Democrats.¹

WHO ARE MILLENNIALS?

According to Paul Taylor of the pewresearch Center, the millennial generation begins with those Americans born after 1980 (2014). Members of this group are called millennials because they came of age at the dawn of the new millennium (pewresearch Center 2015b).² This generation is notably large, and is now the largest generation in the American labor force (Fry 2015).

Taylor's analysis concludes that millennials are more liberal, Democratic, and racially diverse than other generations. Importantly, he notes that millennials are "at ease with racial, ethnic, and sexual diversity" (2014, 33). Taylor has a lot of company in depicting the millennial generation as unique in its comfort with diversity. Morley Winograd and Michael Hais, for instance, write that growing up in a more diverse society than its predecessors has led this generation to have "relatively colorblind attitudes on race relations" and "far more positive attitudes toward immigrants and their impact on society than older generations" (2008, 95, 96). Others echo this view (Madland and Teixeira 2009; Teixeira 2011; Winograd and Hais 2011; Marcotte 2013; Feldmann 2014; Cillizza 2014; Tierney 2014; Simmons-Duffin 2014). An important factor driving optimism about the racial harmony found among this generation is its own racial composition. According to recent estimates, this generation is only 57 percent white, non-Hispanic, relative to 72 percent of baby boomers, the next largest generation (pewresearch Center 2015a).³

What is missing from these optimistic characterizations of millennials, however, is an examination of their attitudes by race. Perhaps it is the relatively large proportion of nonwhites in this generation that makes the group seem so much more comfortable with diversity compared to older, but also whiter, generations. Whether millennial members of the country's racial majority share that comfort is an important question, particularly as whites edge ever closer to losing their majority status. The few extant examinations of the political attitudes of white millennials present a mixed picture. A 2014 report from the pewresearch Center uncovers gaps in attitudes among millennials by race, including presidential approval ratings and views on the scope of government, and a report by the Public Religion Research Institute describes a wide range of differences among millennials by race (Jones, Cox, and Banchoff 2012). A 2014 analysis by the Institute of Politics at Harvard also points to racial divisions among millennials on a range of issues (Harvard Public Opinion Project 2014). These reports, however, do not go beyond cross tabulations and do not include systematic comparisons of millennials to older Americans. They also fail to report the sample size for their nonwhite millennial respondents; given that most nationally representative surveys have few nonwhite respondents of any age, the subset of nonwhites who are born after 1980 in these surveys is likely to be quite small.

Scholarly analyses of the attitudes of millennials are also hard to find, though research on white attitudes by age exist. For example, Tatishe Nteta and Jill Greenlee find that white Americans who came of political age during Obama's first election are less likely to exhibit racial resentment than older white Americans (2013). Moreover, the ability of contact with blacks to mitigate racial resentment was enhanced for younger whites, whereas such contact exacerbated racial resentment among the oldest whites. Nteta and Greenlee conclude that coming of political age during Obama's presidency could diminish racial resentment among whites in the long term.

A study by Gary Jacobson suggests that this kind of "generational imprinting" is not confined to those young whites who might be predisposed to support Obama on account of their partisan identification (2016). Across many measures, he finds that young Republicans differ considerably from their older counterparts.⁴ For example, young Republicans were much less likely to call themselves conservative and more progressive in their views on same sex marriage, immigration, and racial resentment. Notably, Democrats showed greater similarity in their views across age groups on these issues than Republicans did. Younger Republicans were also less likely to watch Fox News and to believe that Obama is foreign born or a Muslim than older Republicans. Jacobson concludes that younger Americans are less polarized along ideological lines than their older counterparts.

A contrasting perspective on racial attitudes among young adults comes from a study by Tyrone Forman and Amanda Lewis, who examined the attitudes of white high school seniors from 1976 to 2000 (2015). Their analysis finds that whites in the later period were more likely to agree that it is not their business if some minority groups get unfair treatment. Instead of finding an embrace of diversity, Forman and Lewis argue that their findings indicate "a growing sense of cognitive distance and disengagement from racial matters in general" (2015, 1418). In thinking that racial issues do not concern them, the so-called comfort with diversity that young whites seem to feel might actually be an indifference that ignores or denies ongoing racial injustice.

Overall, however, research suggests that younger whites today may in fact be more racially liberal than older whites. Additional research shows that people do not necessarily become more conservative with age, as previously thought, and that generational attitudes are formed by political events of the period in which people enter adulthood, as well as by the aggregate characteristics of the generation itself, such as whether the generation as a whole has more educational attainment than previous generations (Davis 2013; Schwadel and Garneau 2014; Danigelis, Culter, and Hardy 2007; Abrajano and Lundgren 2015; Cook 2014; Osborne, Sears, and Valentino 2011; Alwin and Krosnick 1991; Ghitza and Gelman 2014). Taken together, these studies suggest that the attitudes among the large millennial generation will be stable and will drive the socialization process among their children in the years to come. Put another way, there is reason to believe that the diversity of this generation, along with the political climate during its coming of political age (including the election of the first black president and the liberalization of several policies, such as gay rights and health care), will have a lasting cohort effect that will render this generation more politically and socially liberal than prior generations.

Although this study is framed as an examination of the millennial generation, people born in 1979 are likely to have similar experiences to those born in 1981. Yet a long line of research on political socialization focuses on younger members of the electorate, and in such studies it is routine to examine respondents by cohort. The "young adult" cohort varies in definition, but is often conceptualized as adults who are eighteen through twenty-nine, eighteen through thirty, or eighteen through twenty-five (Cook 2014; Nteta and Greenlee 2013; Alwin and Krosnick 1991; Dinas 2013; Abrajano and Lundgren 2015). This time frame is theoretically interesting and important because it is the time in which political attitudes appear most likely to be affected by contemporary political events and set the stage for a person's long-term political identity (Stoker 2014; Sears and Brown 2013; Ghitza and Gelman 2014; Kinder 2006; Hopkins 2014). Focusing on millennials, therefore, allows this analysis to engage with both conventional wisdom in popular culture, where the term millennial is used frequently, as well as with scholarly examinations of how political socialization affects cohorts, which focus on "young adults" more generally.⁵

EXPAND THE CIRCLE OF WE OR CIRCLE THE WAGONS?

Population change and perceived threats to the group's status can lead people to become more or less inclusive. Contact theory and generational imprinting would predict greater inclu-sivity, whereas social identity theory would predict greater defense of one's in-group and perhaps greater denigration of out-groups.

One theoretical process underlying the notion that today's young whites will be more comfortable with diversity than previous generations is contact theory. Contact theory, in its most basic form, posits that intergroup contact can enhance intergroup harmony and reduce prejudice (Allport 1954; Forbes 1997). Meta-analyses of contact theory studies conclude that the totality of evidence supports the theory's main tenets, particularly when the contact in question involves interpersonal friendships (Pettigrew and Tropp 2006; Pettigrew et al. 2011). For millennials, their diversity leads observers to believe that they will be more racially liberal and less threatened by demographic change than older Americans. Among a generation this diverse, increased interracial contact is an assumption. Recent polling supports this assumption, finding that Americans under thirty are much more likely to date and marry outside their race than older Americans (Dunsmuir 2013). Given the findings from research on contact theory, the diversity of this generation, and the rise of interracial and in-terethnic relationships, white millennials are therefore expected by many to react to demographic change more favorably than older whites.

White millennials may be more politically and racially liberal than older whites for other reasons as well. Chief among them is their pre-adult and early-adult socialization experiences across domains (home, school, politics, popular culture, and so on), that could reinforce racial liberalism. Growing up in a post-civil rights environment with new norms about acceptable racial discourse, witnessing the election of a black president during their politically formative years, seeing more diversity in entertainment, and more, could result in racially liberal generational imprinting, even for those young whites who maintain racially homogeneous interpersonal networks (Mutz and Goldman 2010; Dovidio et al. 2011; Mendelberg 2001; Nteta and Greenlee 2013; Goldman 2012).

Despite the potential power of interracial contact and the increased presence of more racially tolerant environments, there are reasons to believe that whites in the United States might not expand the circle of we so readily in the face of increasing ethnic diversity (Hollinger 1995). Studies show that nontrivial segments of the white population feel that whites are discriminated against, identify as white, and exhibit a sense of linked fate (Norton and Sommers 2011; Jardina 2014; Schildkraut 2017). Whites are more likely to think of themselves in terms of their racial identity when they are primed to consider threats to the group's status, and identification as white can promote group-interested preferences (Branscombe, Schmitt, and Schiffhauer 2007; Goren and Plaut 2012; Outten et al. 2012; Hutchings et al. 2012; Lowery et al. 2006). Maureen Craig and Jennifer Richeson find that informing white Americans of demographic projections leads them to adopt a more conservative political outlook on a range of issues and promotes greater levels of both implicit and explicit racial bias (on outlook, 2014b; on bias, 2014a). Felix Danbold and Yuen Huo find that such projections lead whites to perceive that their status as the "prototypical" American is threatened which, in turn, decreases support for diversity (2014). Ashley Jardina finds that the impact of identifying as white on policy attitudes has increased over time. Together, this research suggests that as the nation's population is becoming ever more diverse, whites are reacting by closing ranks around the group, identifying more strongly as white, and having that identification become a more prominent influence over their political beliefs (2014).

These reactions are in line with the predictions of social identity theory, which maintains that salient group identities can be powerful forces shaping political attitudes and behaviors (Tajfel 1982; Tajfel and Turner 1986). Perceptions of threat to the group enhance the group's salience and lead group members to approach their environment in a group-interested manner (Schmitt and Branscombe 2002). Related to social identity theory is group position theory, the idea that perceived threats to the position of one's group in the social hierarchy, such as those whites feel in response to efforts to redress racial inequality, can generate group-interested attitudes and behaviors (Blumer 1958; Bobo and Hutchings 1996; Bobo 1999).

What we do not know, however, is whether white Americans of different generations respond similarly to information about the changing ethnic composition of the country. The studies described do not focus on age or generation. Several relied on small MTurk samples or college-age samples, which preclude meaningful analysis by generation. Of those that used larger nationally representative samples, analyses that control for age are rarely reported. In the exceptions where the results do control for age, it is used as a continuous variable: specific attention to millennials is lacking. Moreover, in studies that control for age, its effect is erratic: older Americans are more likely to identify as white than younger Americans in one study (Jardina 2014), but in another age is insignificant (Hutchings et al. 2012). In short, despite the value of the studies cited thus far, they do not address the specific question that motivates this analysis.

CROSS-SECTION ANALYSIS

Our first examination of whether white millen-nials are more racially and politically liberal than older whites relied on cross-sectional survey analysis, which allowed us to compare the attitudes of white millennials with those of other groups. The results show the relative effects of race and generation on attitudes about party, ideology, racial resentment, affirmative action, and immigration policy. This portion of the analysis does not include a measure of interracial contact, an important theoretical mechanism that might promote differences between white millennials and older whites; it is therefore not a direct test of contact theory. Nor do we have direct measures that capture the socialization experiences of our respondents. What we can do, however, is examine whether our results are more consistent with the expectations of social identity theory or more consistent with the expectations of theories that would predict greater racial and political liberalism. The goal here is to establish patterns of political preferences among whites of different generations and among millennials of different races. Establishing these patterns is an important step in assessing claims about the promise and expectations cast upon the nation's white young adults.

The data for this analysis come from the 2012 and 2016 Cooperative Congressional Election Studies. The CCES is a national survey conducted online by YouGov/Polimetrix that includes before and after election waves (Ansolabehere 2013). The 2012 survey included 54,535 respondents. Of those, 74 percent were white and 16.5 percent were millennials. Of the millennials, 5,121 were white and 3,881 were nonwhite. The 2016 CCES included 12,465 white millennials and 6,959 nonwhite millennials (Ansolabehere and Schaffner 2017). The large CCES sample sizes thus present a unique opportunity to examine the questions under investigation here.⁶

Despite the advantage of having a large number of nonwhite millennials, it is significant that the CCES is conducted only in English. Language use is an important marker of acculturation, and acculturation is a powerful predictor of political attitudes among Latinos and Asians in the United States.⁷ Specifically, respondents with lower levels of acculturation tend to exhibit more liberal policy preferences and greater concern for racial issues, such as descriptive representation (Branton 2007; Barreto and Pedraza 2009; Manzano and Sanchez 2010; Sanchez and Masuoka 2010; Fraga et al. 2012; Schildkraut 2013). The lack of a non-English questionnaire makes it harder to detect differences between white and nonwhite millennials; any differences that do emerge are likely to be even stronger in the broader population.⁸

Across all dependent variables, we tested the extent to which race and generation shaped responses. In particular, we looked for whether the views of white millennials more closely resembled the views of nonwhite millennials or the views of older whites. If the former, then perhaps the optimistic portrait of how millen-nials feel about racial matters is true. If the latter, then we can conclude that being white still matters a great deal in shaping how white mil-lennials form their political views. The 2012 analysis focused on five variables. First, we examined partisanship and ideology. These broad political outlooks are arguably more associated with race in our current era than in the past, and prior research argues that information about the nation's changing demographics leads whites to adopt more conservative preferences across a range of issues, even if those issues are not racial in nature (Abrajano and Hajnal 2015; Craig and Richeson 2014b; pewresearch Center 2012). Then we examined racial resentment, support for affirmative action, and attitudes about immigration policy. In 2016, a different set of questions was fielded to examine racial attitudes, and those variables were analyzed as well, in addition to partisanship, ideology, and attitudes about immigration policy.

For each dependent variable, we ran three statistical models. We first examined whether millennial respondents held more liberal preferences than older respondents. Then we controlled for race to determine whether being in the millennial generation still affects attitudes. The last model examined how being white and a millennial interact to shape attitudes. For models that include this interaction term, the coefficient on generation can be interpreted as the effect of being millennial on the political attitudes of nonwhites, and the coefficient on race can be interpreted as the effect of being white on the political attitudes of nonmillenni-als. The coefficient on the interaction term indicates whether being millennial has a unique effect on the attitudes of whites. Given the conventional wisdom about the political outlook of white millennials, the hypothesis examined was whether this coefficient is negative, which would indicate a more liberal attitude relative to older whites. All models also controlled for the respondent's level of education and whether the respondent resides in the South.⁹ For ease of presentation, we display the results in terms of predicted outcomes derived from the full model.¹⁰ Tables of results are included in the appendix (see tables A1 through A5).

Party and Ideology

Partisanship is measured on a 7-point scale that runs from strong Democrat to strong Republican,

Figure 1.

Predicted Partisan Identification, 2012

Source: Authors' calculations based on 2012 Cooperative Congressional Election Survey (Ansolabehere 2013).

Figure 2.

Predicted Partisan Identification, 2016

Source: Authors' calculations based on 2016 Cooperative Congressional Election Survey (Ansolabehere and Schaffner 2017).

and ideology on a 7-point scale that runs from very liberal to very conservative. The results, which can be found in figures 1 and 2 (party) and 3 and 4 (ideology), indicate that being white is associated with greater identification as Republican and conservative. For partisanship, being in the millennial generation lessens one's attachment to the GOP, but only for whites (that is, the interaction was negative). It is clear from figures 1 and 2, however, that despite the statistical significance of being a white millennial, white millennials were actually closer to older whites in their partisan identification than to nonwhite millennials.

Figure 3.

Predicted Ideology, 2012

Source: Authors' calculations based on 2012 Cooperative Congressional Election Survey (Ansolabehere 2013).

Figure 4.

Predicted Ideology, 2016

Source: calculations based on 2016 Cooperative Congressional Election Survey (Ansolabehere and Schaffner 2017).

For ideology, whites were more conservative than nonwhites, and millennials were more liberal than nonmillennials. The interaction between race and generation was significant in 2016 but not in 2012. The predicted outcomes displayed in figures 3 and 4 indicate that in 2012, the ideological self-categorization of white millennials was squarely in between that of older whites on the one hand and nonwhite millennials on the other. In 2016, white millennials were closer to their nonwhite counterparts than to older whites. Ideology in 2016 is the only analysis in this study where we find this to be true.

Figure 5.

Predicted Level of Racial Resentment, 2012

Source: Authors' calculations based on 2012 Cooperative Congressional Election Survey (Ansolabehere 2013).

Little evidence indicates that the 2016 presidential campaign drove young people from the Republican Party. The mean predicted partisan affiliation and ideological categorization of white millennials is nearly identical in 2012 and 2016. In other words, the partisan and ideological makeup of white millennials was stable and—especially with respect to party—very close to the pattern exhibited among older whites.

Racial Resentment

No set of issues draws as much optimism among observers of millennials as racial issues. Observers claim that young people today are growing up amid so much diversity that we are witnessing a generation that is uniquely at ease with racial issues. As Winograd and Hais put it, "For Millennials of all backgrounds, racial and ethnic equality and inclusivity is a message they have been hearing all their lives and one in which they firmly believe" (2011, 32). To examine such optimism, we turn now to attitudes related to race, starting with racial resentment. The 2012 CCES measured racial resentment with two questions. Respondents were asked the extent to which they agreed with the following statements: "Generations of slavery and discrimination have created conditions that make it difficult for Blacks to work their way out of the lower class"; and "The Irish, Italians, Jews and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors" (for a discussion of these and other racial resentment measures, see Kinder and Sanders 1996; Tesler and Sears 2010). Both items were combined into a 5-point scale coded such that a higher score indicates more resentment (a=0.76).

The results, presented in figure 5, are similar to the results for partisanship. For white millennials, being white led them to have more resentment than nonwhite millennials, and being a millennial led them have less resentment than older whites. Again, the interaction between race and generation is negative and significant, indicating that young whites are more racially liberal than older whites. Nonetheless, as is true of partisanship, the liberalizing effect of generation is modest; the predicted level of resentment for white millennials (3.45) is closer to that for older whites (3.72) than to the predicted level for nonwhite millennials (2.93).

Figure 6.

Predicted Probability of Affirmative Action Preference, 2012

Source: Authors' calculations based on 2012 Cooperative Congressional Election Survey (Ansolabehere 2013).

Affirmative Action

Turning to affirmative action, respondents in 2012 were asked, "Affirmative action programs give preference to racial minorities in employment and college admissions in order to correct for past discrimination. Do you support or oppose affirmative action?" Four response options were given, ranging from strongly support to strongly oppose. Consistent with our other results, white millennials resembled older whites more than nonwhite millennials, despite a significant negative interaction between race and generation. White millennials were slightly less likely to strongly oppose affirmative action than older whites (35 percent versus 42 percent), but strong opposition is the plurality response for both groups, as illustrated in figure 6.

Attitudes About Discrimination

The 2016 CCES did not ask about racial resentment or affirmative action but instead about respondents' awareness of white privilege and racial discrimination. They were asked whether they agree or disagree with the following statements: "White people in the U.S. have certain advantages because of the color of their skin";

and "Racial problems in the U.S. are rare, isolated situations." The results for these questions, depicted in figures 7 and 8, conform to the pattern uncovered thus far. The interaction between being white and millennial was negative and significant in both cases. In one of them, whether having white skin confers advantages (figure 7), the liberalizing effect of being young was severely offset by being white. In sum, across 2012 and 2016, using a variety of measures about race, the general pattern remains: the racial attitudes of white millennials were remarkably similar to those of older whites.

Immigration Policy

Turning to immigration, the pattern is repeated yet again. Here, the dependent variable for 2012 was a 7-point scale, coded 0 to 1, on whether respondents support or oppose six immigration policies (a = 0.80). The policies are to provide legal status for immigrants in the country illegally, increase border patrol, allow police to question anyone they think might be in the country illegally, fine U.S. businesses that hire immigrants who are in the country illegally, prohibit social services for people in the country

Figure 7.

"White People in the U.S. Have Certain Advantages Because of the Color of Their Skin," 2016

Source: Authors' calculations based on 2016 Cooperative Congressional Election Survey (Ansolabehere and Schaffner 2017).

Figure 8.

"Racial Problems in the U.S. Are Rare, Isolated Situations," 2016

Source: Authors' calculations based on 2016 Cooperative Congressional Election Survey (Ansolabehere and Schaffner 2017).

illegally, and deny birthright citizenship. In 2016, it was a 5-point scale made from four questions (a = 0.69): to provide legal status for immigrants in the country illegally, increase border patrol, grant legal status to children who were brought to the United States illegally, and deport "illegal immigrants." In both years, a higher score indicates a greater preference for restrictive policies. Here again, the interaction between race and generation is negative and

Figure 9.

Predicted Immigration Restriction Score, 2012

Source: Authors' calculations based on 2012 Cooperative Congressional Election Survey (Ansolabehere 2013).

Figure 10.

Predicted Immigration Restriction Score, 2016

Source: Authors' calculations based on 2016 Cooperative Congressional Election Survey (Ansolabehere and Schaffner 2017).

significant, but the positive and significant effect for race, which promotes a more conservative stance, offsets the liberalizing push that being a millennial brings. As the predicted probabilities presented in figures 9 and 10 indicate, young whites had more restrictive immigration preferences than nonwhites, and their views more closely resembled those of older whites than of their nonwhite co-millennials.

In all models and with all dependent variables, residing in the South is associated with more conservative attitudes. Ideally, we would also have been able to control for whether respondents live in rural or urban settings. Unfortunately, such a measure was not available. Although white millennials appear in the aggregate to have views similar to those of older whites, heterogeneous preferences among whites need to be acknowledged and investigated. It is plausible that whites in urban settings are more racially liberal than whites in rural settings given the greater diversity found in urban areas. More relevant for our purposes, it is important to consider whether white millennials are uniquely affected by living in a rural or urban area relative to older whites. This question is one that future research should pursue.

To sum up, in four of the five tests in 2012 (all but ideology) and in all tests in 2016, the interaction between race and generation was negative and significant, indicating that white millennials showed a more liberal pattern of political preferences than their white elders did. Yet in all models, race maintains a significant and strong effect that offsets much of the racial and partisan liberalism promoted by being in the millennial generation. Across almost all tests, we see that white millennials tended to have views on party, ideology, race, and immigration closer to those of older whites than to those of nonwhite millennials.¹¹ In short, contrary to the claim Madland and Teixiera assert, race appears to remain a "big deal" for white millennials (2009).¹²

EXPERIMENTAL ANALYSIS

To probe this question further, we examined whether millennial whites and older whites responded similarly when presented with the specter of losing their status as the nation's racial majority. As noted earlier, social identity theory lays out reasons to expect whites to close ranks around their group in response to perceived threats to their status, and existing studies indicate that when shown information about population projections, whites do become more conservative on a wide range of issues. But are white millennials less prone to this tendency? Does the liberalizing effect of their generation offset their perception of threat? Exploring this question is the goal of the analysis presented in this section.

Here we show results of an experiment involving a nationally representative sample of white Americans (N = 955, 227 of whom are white millennials). The experiment was conducted online in 2016 by GfK Custom Research as part of a partnership with Time Sharing Experiments for the Social Sciences (Druckman and Freese 2016). It built on the important experiment conducted by Craig and Richeson, in which some respondents were shown a press release indicating that the nation will soon be majority-minority and a control group read about internal migration within the country. Their analysis reveals that whites in the treatment group were more likely than whites in the control group to have negative evaluations of ethnic groups (2014a) and to have more conservative preferences on a range of race-related and race-neutral public policies (2014b). Yet their study had only 415 white respondents, only eighty-five of whom were millennials. It also lacked measures of potential mechanisms that could drive the supposed differences between white millennials and older whites on matters of race.

Our experiment replicated features of the original Craig and Richeson study, using their press releases in our treatment and control conditions (see appendix). We included a larger sample of white millennials (N = 227), and to examine one potential mechanism, a measure of interracial contact. We asked respondents whether they personally knew anyone black, Hispanic, or Asian. If they said yes, we asked whether they had any close friends who are black, Hispanic, or Asian. Their answer to this question is our measure of interracial contact.¹³ We concentrated on this form of close interpersonal contact because research indicates that contact through friendships is more likely to have a beneficial effect on intergroup attitudes than casual contact that results from diverse schools, workplaces, and neighborhoods (for a thorough example, see Ellison, Shin, and Leal 2011). We also asked respondents (before the manipulation) how important being white was to their identity.¹⁴ The main hypothesis examined here is that white millennials are less likely than older whites to be affected by the treatment. In other words, white millennials are less likely than older whites to be moved in a conservative direction and are less likely to exhibit lower evaluations of ethnic outgroups after reading about demographic projections. After testing this hypothesis, we looked at interracial contact and examined whether millennial respondents were uniquely affected by having nonwhite friends.

Dependent Variables

In one paper, Craig and Richeson use their experiment to show that whites in the treatment group had more conservative preferences on several race-related and race-neutral policies than the control group (2014b). We used the same questions they did in their study. The first race-related policy item asked respondents whether they thought the time it takes for immigrants to become eligible for U.S. citizenship should be increased or decreased (5-point scale). The second asked whether the overall number of immigrants allowed into the United States should be increased or decreased (5-point scale). The third asked whether respondents agreed or disagreed with preferential hiring and promotion of racial minorities (7-point scale). In the interest of space, this analysis focuses on these three policy questions; analysis of race-neutral policies is not included.

In a second study, Craig and Richeson find that the treatment led whites to have more negative attitudes toward blacks, Latinos, and Asian Americans, but not toward whites (2014a). We assessed these group attitudes just as Craig and Richeson did, using traditional feeling thermometers in which respondents were asked to rate how warm (favorable) or cold (unfavorable) they felt toward particular groups in society, 100 being the warmest and 0 the coldest. In both of their papers, Craig and Richeson note that the treatment also promotes feelings of threat to group status. In keeping with their study, we asked participants the extent to which they agreed or disagreed with the following statement: "If they increase in status, racial minorities are likely to reduce the influence of white Americans in society" (7-point scale, higher feelings of threat coded as 7). In the Craig and Richeson study, whites in the treatment group were more likely to agree with this statement than whites in the control group.

Next, we tested whether exposure to information about population change affected white millennials and older whites similarly. We examined the three racial policy questions, the feeling thermometers, and the measure of status threat described here.

Are Millennials "Immune"?

To determine whether the treatment condition influenced the attitudes of millennials differently than older whites, we conducted two-way ANOVAs on all of the dependent variables described with the treatment condition (Racial Shift versus Control) and generation (Millennial, Nonmillennial) as fixed factors. The findings regarding the main effects for being in the Racial Shift condition and for being a millennial appear in the first two blocks of results in table 1. The findings regarding the interaction of condition and generation are in the last block of results. Overall, the results are in line with the findings from the CCES analysis: white millennials were often more racially liberal than older whites, but their views were still powerfully shaped by their race.

Looking first at the direct effect of being in the Racial Shift condition as opposed to the control, we find a significant and direct effect moving people in a racially conservative direction on three of the eight measures (period of time to naturalize, feeling thermometer for blacks, and racial threat); a fourth measure is just shy of significance at the 95 percent confidence level (feeling thermometer for Asians).

Looking at the direct effect of being in the millennial generation, the results indicate more racially liberal views among millennials on the overall level of immigration and feelings of racial threat; the feeling thermometers for blacks and whites are just shy of significance. In sum, when looking at direct effects, the Racial Shift treatment results in more racially conservative views on four of the eight dependent variables and being in the millennial generation results in more racially liberal views on four of the eight dependent variables.

However, the last block of table 1 notably shows that for all eight dependent variables, the interaction between generation and condition was not statistically significant. In other words, in no case did we find that white millennials were less likely than older whites to be pushed in a conservative direction when they read about census population projections. When we saw significant results for the treatment, they applied to millennials and nonmil-lennials similarly.

What About Contact?

Despite finding that white millennials are not immune to the perceived threats that population changes promote relative to older whites, it is still possible that younger whites have more meaningful or significant interracial friendships than older whites, a possibility we began to explore in this final set of analyses. Using the same dependent variables analyzed in table 1, we ran a series of ordered probit and regression analyses that controlled for whether the respondent was in the treatment condition, whether the respondent was a millennial, how important the respondent said being white is to his or her identity, whether the respondent had any close friends who were black, Hispanic, or Asian, and his or her partisan identification. We also included a term to capture the interaction between one's generation and having any close nonwhite friends. The coefficient on this interaction term indicates whether having nonwhite friends affected millennials differently than older whites. In our sample, 67.7 percent of millen-nials reported having a close nonwhite friend, but so did 67.4 percent of older whites. In the aggregate, millennials were not more likely than older whites to report interracial friendships, a curious finding in and of itself that merits further scrutiny and additional data collection. But were those friendships a more powerful influence over racial attitudes?

The results for our analyses appear in table 2 (racial policy and racial threat) and table 3 (feeling thermometers). On racial policy and racial threat, we find that having nonwhite friends does not affect attitudes. If anything, white millennials with nonwhite friends were slightly more likely than others to say that immigration levels should be decreased. Rather, the most consistent factors leading to conservative racial policy views are whether respondents said that being white is important to them and whether they were Republican. Being in the Racial Shift condition also moved respondents in a conservative direction in two of the four measures (increase citizenship time and status threat).

Racial identity, partisanship, and being in the Racial Shift condition also mattered a great deal for the feeling thermometers (table 3). Here, however, we also find that nonwhite friendships led to higher thermometer scores for blacks, Hispanics, and Asians. Only on the feeling thermometers scores for blacks did we find that the millennials were affected more than older whites. In sum, interracial friendships mattered more for outgroup evaluations

Table 1.

ANOVA Results for Condition and Generation

Table 2.

Predicting Racial Policy Attitudes and Status Threat

than for policy preferences, though little evidence indicates that today's young whites are uniquely affected by such friendships. What mattered more was whether respondents were Republican and whether they said that being white was important to them. On this last measure, bivariate analysis shows that generation might matter: only 8.8 percent of millennials said that being white was extremely important, whereas 14.6 percent of nonmillennials did (p = .01). Exploring the relationship between age, race, and white racial identity—and how levels of white identity might change over time—is an important next step in this research agenda.

DISCUSSION AND CONCLUSION

The results of this analysis indicate that white millennials occupy a political space between their nonwhite counterparts and older whites. Notably, however, the findings fail to support the optimism about millennials commonly found in popular discourse with respect to racial attitudes. It is true that white millennials were somewhat more ideologically and racially liberal than older whites on some measures studied here, but far more often race was a stronger factor shaping their outlook. The first portion of our study showed that their partisan and racial policy views were closer to the views of older whites than to those of other millen-nials. The one exception was for ideological orientation in 2016, in which white millennials were closer to nonwhite millennials than they were to older whites. The second portion showed that they became more racially conservative on some measures after reading about population projections, just as older whites did. They did not become more conservative on every measure, but neither did nonmillennials. Together, these results offer a corrective to popular narratives that tell us that young Americans will usher in a more racially harmonious era and that the Republican Party needs to change if it hopes to count young Americans among its ranks. The Republican Party is mostly white (89 percent of GOP identifiers in our CCES samples were white), and white millennials are not very distinct from older whites on the issues examined here, including partisan identification.

Table 3.

Predicting Feeling Thermometer Scores

Although our study focused on comparisons between older and younger whites on the one hand and white and nonwhite millennials on the other, our findings also reveal an intriguing pattern among older and younger nonwhite respondents in the CCES surveys. In nearly all cases, the predicted outcomes for nonwhite respondents differed very little by generation. Our study leads us to conclude that race affects attitudes more than generation among whites on many (but not all) of the issues under examination here, and this seems to be especially true for nonwhites. This degree of racial solidarity is perhaps not surprising, yet it is important that it provides an additional corrective to narratives claiming that millennials will be more comfortable with diversity than older Americans: among nonwhites, partisan and racial liberalism appears to be the norm among the young and old alike.

The analysis also raised many questions for future research to consider. First, if white millennials are only slightly more racially liberal than their predecessors and close ranks around the group in the face of group threat despite being a vastly more diverse generation, should we instead place our hopes for greater racial unity on the subsequent generation, which is going to be more diverse still? Or will white members of Generation Z also be significantly affected by being white? Similarly, to what degree is there heterogeneity on racial and ethnic matters within the millennial generation? Perhaps younger white millennials, having arguably come of political age in a more politically and racially polarized context than older white millennials, are more racially and ideologically conservative than older members of their generation. A related question is what happens to the views of millennials as they age. Will their slight racial liberalism (relative to older whites) remain, or will they become more conservative over time?

Second, what exactly is the nature of interracial contact among white millennials? We know that rates of interracial dating and marriage are rising, but to what extent do white millennials truly spend their time in racially diverse settings? Moreover, it is possible that their interracial relationships do matter, but in ways not captured by the blunt measure of interracial friendship used here. In our experiment, younger and older whites did not differ in their likelihood of reporting interracial friendships. That could be an anomaly, or it could be a product of the blunt measure we use. It could also mean that younger whites do not actually have more meaningful interethnic contact than older whites, despite conventional wisdom that says otherwise.

Third, the meaning and dynamics of white racial identity are poorly understood. We know that white millennials seem less likely to say that being white is important than older whites and that its importance moves whites in a racially conservative direction. But how will the presence and power of white racial identity change as whites lose their majority status, and how much will this dynamic vary across whites of different ages and contexts?

Fourth, the intersection of race, age, and partisanship needs to be analyzed further. Among whites, partisanship has become a major dividing line on matters pertaining to race, yet younger Republicans showed signs of having some policy preferences and ideological identities distinct from those of older Republicans (Jacobson 2016). How that divergence develops alongside demographic changes in the years to come will be important to observe.

Finally, the extent to which living in an urban or rural context shapes the dynamics under investigation here should be examined. One important narrative that emerged after the 2016 presidential election is that rural whites felt particularly ignored by political elites, which made them especially amenable to Donald Trump's campaign rhetoric (Cramer 2016; Hochschild 2016). The degree to which this sentiment exists and shapes how people respond to information about demographic change—and whether that effect varies across generations—is an important question for future research to address.

As the nation edges ever closer to having a majority-minority population, messages about racial divisions seem like they are becoming more, rather than less, common. These divisions are exacerbated by political parties that are becoming more racially distinct and as the standard bearer of the Republican Party takes positions that are explicitly racially antagonistic. It is increasingly important to examine how people feel about the implications of such divisions, whether preferences differ across generations, and the mechanisms that affect how people react to information about our changing nation. Uncovering practical ways to elicit racial harmony will be one of the most pressing questions of our time, our research suggesting that relying on the allegedly more tolerant attitudes of millennials may not be enough.