No one – not Donald Trump’s advisors and perhaps not even Trump himself – saw it coming. Within just a few weeks after announcing his candidacy for the Republican nomination in June 2015, Trump had seized an early lead in the polls that he would never relinquish. Rolling to victories in most of the early GOP primaries, by March his momentum seemed unstoppable and by early May 2016 he had essentially cinched the Party’s nomination for the presidency. In the process, Trump vanquished an experienced field of Republican contenders described by many commentators at the beginning of the campaign as the Party’s “most talented field ever” (e.g. Collinson 2015; Schneider 2015).
In many respects, the most surprising outcome of the 2016 election was Trump’s seizure of the Republican nomination rather than his general election victory over Hillary Clinton (in the electoral college). The general election outcome is less surprising that it fits well with either of two widely-held models of contemporary presidential elections (and American politics more generally). Forecasting models based on largely on macroeconomic fundamentals predicted a close election, with a majority finding a narrow Democratic victory (in the popular vote) the most likely [see Campbell (2016) for a summary]. The final outcome (with Clinton winning the popular vote by a little over 2%) was right in line with those forecasts, with the surprising twist lying only in the Electoral College verdict that handed Trump the presidency.
Polarization models asserting that partisan identities in the mass public have hardened in recent decades are also vindicated by the general election result (see e.g. Baldassarri and Gelman 2008; Abramowitz 2010; Hare et al. 2015). In spite of all the noise generated by the unusual Trump candidacy and the 2016 campaign, a remarkably strong partisan divide largely held to form. Polarization models suggest that the possibility of a significant number of Republican voters crossing over to vote for (an unpopular) Democratic candidate in the present environment would have been very unlikely. The final outcome in fact looks quite a bit like a replay of the 2012 general election, with Trump holding Republican identifiers and benefiting from very modest shifts among a few key electoral groups in key vote states to produce the final outcome.
If the general election outcome seems less surprising on closer inspection than many have suggested (Jacobs and Ceaser 2016), consider the primary campaign. Trump seized the Republican nomination against the wisdom predictions of the dominant social science theories of the nominating process. The Trump campaign largely disregarded – or failed to mobilize – any of the well-established factors long identified as critical: money, endorsements, field organization, and message discipline (e.g. Kamarck 2016). Trump unambiguously lost the “invisible primary,” whether we construe that as the process by which donors and party elites decide on a candidate (Reiter 1985; Aldrich 1995) or if we take a broader conception of the “party” as including leading activists who get involved in the nominating processes because of intense policy concerns (Cohen et al. 2008). By any definition, Trump’s primary success challenges what we thought we knew about the importance of party insiders and primary dynamics in the nominating process of American presidential elections. Trump did not build a grass-roots organization anywhere in the primary process (Geraghty 2016). He spent significantly less than the other leading challengers (Clark 2016). He received virtually no endorsements from leading elected officials until very late in the primary season (FiveThirtyEight 2016). Conservative media commentators by and large either opposed a Trump nomination or refused to endorse him, even after his nomination appeared to be fait accompli (see e.g. Slate magazine’s weekly survey of influential conservative commentators’ responses to the Trump campaign at www.slate.com).
Trump’s triumph in the GOP primaries occurred above all else for one simple reason: a decisive group of Republican primary voters disregarded the opposition to Trump from virtually all party elites to cast ballots for him. This was true in the early part of the campaign with the full slate of candidates, and later as the field narrowed. Various efforts to mount an “anyone but Trump” effort largely went nowhere. To be sure, Trump’s ability to dominate media coverage of the primary campaign, a widely noted and documented component of the race, helped him build and sustain support. But in contrast to other outsider candidates whose temporary rise in the polls typically generates a flurry of intense media coverage and voter evaluation, leading to discovery of warts and eventual decline – the “discovery, scrutiny, decline” model proposed by Sides and Vavreck (2013) – Trump won the Republican nomination against expectations, even as negative information about him flowed into media coverage of the campaign.
So who were the Trump voters who put him over the top? We develop a political sociological analysis of the social bases of support for Trump during the critical phase of his rise to the Republican nomination. This question is important for understanding not just how Trump prevailed, but also how the underlying mass base of the GOP (as reflected in its primary electorate) has evolved. In particular, we test the widely voiced hypothesis that a critical source of Trump’s support in the Republican primaries came from his appeal to working class and/or downwardly mobile and insecure middle class voters. Were the “Trumpenvolk” [as Oliver and Rahn (2016) snidely, but cleverly, describe them] drawn disproportionately from the ranks of the white working class or downwardly mobile middle class voters responding to a “populist” message? This is a question that particularly resonates among political sociologists, for, as we note in the next section, the “working class authoritarianism” thesis of Seymour Martin Lipset (and later variants) has been widely invoked to explain the Trump breakthrough. Its core proposition – that it is those citizens threatened by economic change who are most drawn to the parties and candidates of the far right – offers a distinctive model for not just the rise of Trumpism, but also for surging support for nationalist right-wing parties around the world. Our analysis extends beyond the question of class alignments to also consider the broader question of “where did the votes come from” that got Trump to the nomination. Here we employ the classical model of Axelrod (1972, 1986) to see how Trump fared across a range of electorally significant groups.
Our findings lead us to conclude that Trump’s primary support was not systematically derived from successful appeals to disadvantaged or downwardly mobile voters. Trump’s voters were, on the whole, significantly more affluent and better educated than the average voters in primary states. We find slightly higher levels of Trump support among those who think it is harder to climb the economic ladder in bivariate analyses, but when statistical controls are applied this finding is significantly moderated.
The paper is in three parts. We begin with a brief discussion of competing accounts – journalistic as well as scholarly – about Trump’s electoral base in the primary season. We then turn, in part two, to our analysis of Trump’s (and his leading competitors’) appeal to key GOP constituencies. We draw upon two sets of data, the American National Election Study pilot in January 2016 and exit poll interviews of GOP primary voters. We explore both individual-level and state-level factors. We conclude, in the paper’s final section, with some brief reflections about the significance of these findings and what they may signal for the future.
Theoretical Accounts of the Trump Coalition
The subfield of political sociology initially developed, from the 1930s onward, out of a perpetual puzzle, one that reappears in every election (and is arguably amplified in 2016): why is it that voters do not always appear to choose the candidates and parties who most closely represent their material interests (at least insofar as those interests can plausibly be discerned by social science analysts)? Or, as Shapiro (2002) pithily puts it, “why don’t the poor soak the rich?”
In the early postwar period, the work of Lipset and his colleagues probably did the most to define the subfield and focus attention on the complex role of class and other “cross-cutting” social divisions, or cleavages, in structuring political preferences and voting behavior [see especially Lipset 1963, 1981 ; Lipset and Rokkan 1967; see also Alford (1963)]. In the essays gathered together in his widely read 1960 book, Political Man, Lipset developed what he would later characterize – in the 1981 postscript to the reissue of the book – as an “apolitical Marxist” approach to explaining the social origins of democracy, fascism, communism, and the social bases of modern political parties (Lipset 1981 , p. 521). The values that sustain democratic societies were viewed by Lipset to be more prevalent in societies with a large and stable bloc of middle-class citizens, especially where education levels were relatively high. Authoritarian preferences, by contrast, could be traced to marginalized groups or classes, including workers (Lipset’s famous formulation of the thesis of “working class authoritarianism”), small business owners and other economically vulnerable groups.1 Feeling threatened, these citizens are prone to look for scapegoats – capitalism, immigrants, elites, Jews, bankers, racial and ethnic minorities – that political entrepreneurs of the right or left can offer as the source of their problems. When tough times arise, and economic marginalization increases, the potential for authoritarian movements and candidates to find popular support increases. Related versions can be found in many of the competing studies of “populism,” McCarthyism, and other expressions of mass politics (e.g. Kornhauser 1959; Bell 1960). Hofstadter’s (1965) famous “paranoid style” essay was an influential extension of the idea, developing a account of countersubversive movements in American history emphasizing the anti-elite and irrational elements of mass hysteria. The working class authoritarian model has been tested in many contexts with mixed results. Most later studies have concluded that it is education, not an occupation- or income-based measure of class, that is most closely associated with support for intolerance, racism, extreme nationalism, and right-wing parties (Lipsitz 1965; Grabb 1979; Dekker and Ester 1987). If the concept of class is respecified to include education, however, the model becomes more successful in accounting for empirical phenomena such as lower-class support for conservative or authoritarian ideas, parties or movements, as well as in accounting for the decline in class voting (e.g. Kohn 1977; Davis 1982; Houtman 2003). A related approach to understanding Trump’s breakthrough during the Republican primaries has focused on the disconnect between the party’s conservative policy platforms of recent years and the more moderate preferences of its many non-affluent partisans and voters. The national GOP mainstream in the last three decades has adopted an increasingly unified position on two critical policy components seemingly at odds with the thrust of Trumpism: (1) welfare state retrenchment and “entitlement reform” to reduce long-term budget deficits and limit the scope of government income subsidies; and (2) support for free markets and an international free trade regime with limited or no protections for the losers. The GOP leadership has also largely maintained a moderate approach to immigration reform and civil rights laws, albeit with more vigorous internal disagreement.2
Trump’s 2016 primary campaign was difficult to characterize ideologically and was extremely vague on policy specifics. It is tempting to read backward from what is emerging as his team has moved into the White House and sought greater alignment with the GOP mainstream. Nevertheless, a few grand and repeated themes from his primary campaign unambiguously stand out – anti-immigrant/“build the wall” rhetoric, opposition to multilateral free trade deals, tax cuts and corporate tax reform, and deregulation. On some issues Trump could appear significantly more sympathetic to the plight of American workers and their families than his primary opponents (Gest 2016). In particular, Trump’s promises to increase employment by improving trade deals and removing unauthorized immigrants could resonate in an era when employment levels for working age white men have declined, as might his calls for massive infrastructure spending and the jobs that would create. Trump also offered – during the primary campaign – little or no support for entitlement reform. His position on health policy was ambiguous; while later in the primary season and during the general election he criticized Obamacare as a “disaster”, during the early primary season he also made vague promises to make health insurance accessible to all and in early debates he even spoke favorably of a Canadian-style single-payer healthcare system. Overall, it is hardly stretching things to say that prior to 2016 it would have shocked most close observers of American politics to think that many of Trump’s policy positions would be endorsed by the Republican presidential nominee.
Numerous early analyses of Trump’s appeal have reintroduced versions of the thesis that his support is rooted in unhappy working class voters. For example, David Brooks (2016) declared “Trump voters are a coalition of the dispossessed [who] have suffered lost jobs, lost wages, lost dreams.” Veteran political commentator Schneider (2015) suggested “What we’re seeing right now in American politics is class warfare…[but] it’s not the working class versus the 1 percent. It’s the working class versus the educated elite. In fact, one of the richest men in the world is leading the revolt: Trump” (Schneider 2015). The New Yorker’s James Surowiecki (2016) declared “You could not ask for a better illustration of the complexity of ordinary Americans’ attitudes toward class, wealth, and social identity than the fact that a billionaire’s popularity among working-class voters has given him the lead in the race for the Republican Presidential nomination.” One can find many other examples of such interpretations (e.g. Weigel 2015; Bonvillian 2016; Irwin and Katz 2016; Sides and Farrell 2016).
It is entirely plausible to argue that on the campaign trail Trump indeed sounded like a populist candidate (at least as long as we disregard the fact that he never directly touches on issues of inequality and redistribution). Systematically examining candidate rhetoric and the use of populist phrases in their campaign speeches, Oliver and Rahn (2016) show that Trump used far more such phrases than any other Republican or Democratic candidates (including Bernie Sanders). Trump relentlessly attacks elites at every opportunity (most notably the media, who became the most visible “elite” during the campaign). Such attacks are also hallmarks of classical populist campaigns. Oliver and Rahn conclude that Trump supporters in the primary season were drawn to a classic populist campaign that is hostile to elites and makes grand but unspecific promises to improve lives, and defeat the conspiratorial forces standing in the way of enhanced well-being.
Sociologist Andrew Cherlin (2016) has advanced a variant of the populist/working class view. Cherlin argues that the key to understanding Trump’s support among non-college educated voters is their perception of being left behind and/or lacking opportunities to achieve the living standard of their parents. Survey respondents who report feeling downwardly mobile were much more likely to be anti-immigrant, and exhibit higher levels of distrust in mainstream institutions. However, Cherlin does not offer a direct analysis of Trump support, only inferring indirectly that “No other candidate in this race has addressed the concerns of downwardly mobile working-class whites so directly” as Trump, and that his working class voters are looking for renewed opportunities for upward mobility.
Against these accounts, there are plenty of reasons to suspect that the Trump-as-champion-of-the-working-class thesis is less than convincing. For example, Trump did not, so far as we can tell, ever once mention rising income and wealth inequality in the course of his primary campaign, and definitely not in the Republican primary debates. Given the avalanche of recent discussion of rising inequality, and the implications it poses for working families, Trump’s systematic neglect of these questions marks his populism as distinctly non-redistributive. His numerous suggestions, by contrast, of tax-cutting in favor of wealthy households and intention to pursue business-friendly deregulation and corporate tax rate cuts drift far away from a populist policy agenda.
To be sure, as observers of “authoritarian” populism have long noted, vague and contradictory messaging is common in populist campaigns. Trump’s promises to create jobs by renegotiating trade deals and stemming immigration could be a source of appeal in an era when employment levels for working age white men have declined (e.g. Gest 2016). Further, his anti-elite message, targeting the media in particular but government as well, had clear populist elements. Certainly Trump routinely declared the bankruptcy of the political system and that the political class as utterly incompetent, while he (as an accomplished business executive) knows how to “get things done,” and would “make deals” that will end governmental incompetence and perceived national decline.
We turn now to our empirical investigation. After briefly describing the data, we present our findings in three stages. First, in the style of Axelrod’s (e.g. 1972, 1986) classic examination of “where the votes come from,” we look at support Trump’s compared to other GOP candidates in various sociodemographic categories at the beginning of the nomination process.3 Here, we ask how many voters in each category reported preferring Trump over other GOP candidates, and what percentage of those preferring Trump fall into a given sociodemographic category. In this way, we can distinguish, a la Axelrod, between a group’s “loyalty” and its “contribution,” respectively, to the Trump campaign. This examination provides a first cut at the broader questions of the paper. Next, we turn to exit poll data to evaluate Trump’s performance among actual GOP primary voters. These data have very limited measures, but they do allow us to examine three vitally important sociodemographic characteristics: income, education, and state economic context. Of special note, the exit poll data, combined with state-level Census data, allow us to estimate the “average” Trump voter and compare that voter to both the average GOP primary voter (as reflected in the exit polls, as well as the average resident in each state (as reflected in state income and education means)). Finally, we develop a multivariate analysis which considers the possibility that while Trump’s level of support may be relatively consistent across socioeconomic status, the reasons for this support may vary. In other words, we consider the possibility that respondents’ subjective perceptions of their social mobility chances influence their support for Trump net of other factors (such as income, education, and partisanship).
Our analysis draws from two sources of data: the American National Election Survey’s (ANES) 2016 pilot survey, fielded in January, and exit poll results from 25 states with primaries before May 4th (when Trump’s victories on May 3 in the contested Indiana primary led most observers to declare him the presumed GOP nominee, with his two remaining competitors dropping out shortly thereafter).4 These data are complementary in several respects. For one, they bookend the primary season, with the ANES data collected only a few weeks before the Iowa caucus while the exit polls ran though the primary season. Second, whereas the exit polls reflect actual voting patterns, the ANES includes a wider array of attitudinal items and information about the sociodemographic profiles of survey respondents, permitting multivariate investigation. Finally, although the data were collected in quite different ways and at different moments, the general patterns in our findings are consistent between the ANES and exit polls.
The ANES pilot survey was fielded from January 22nd to 28th, 2016, and contains 1200 observations (we use an analysis sample of the 1000 observations with no missing data on key variables). We examine two dependent variables in this analysis. The first ANES item measuring Trump support asks respondents which Republican candidate among the ten current candidates they prefer. The second measure is a feeling thermometer which asks respondents how warmly they feel toward Trump (and other GOP candidates). The feeling thermometer is a scale ranging between 0 and 100, where a value of 100 represents very positive feelings, 0 very negative feelings, and 50 represents neutral feelings. The mean score on the feeling thermometer for Trump is fairly low: 38.33 (S.D. 36.53). However, this masks an extreme partisan difference; Republican identifiers report a mean feeling thermometer score of 64.37 towards Trump, while Democrats a very low 18.17. Summary statistics for the Trump feeling thermometer across other socio-demographic variables are provided in the supplementary materials (Table S1).
Exit polls were collected by Edison Research for the National Election Pool group of news organizations. Participants are recruited and interviewed as they exit randomly selected polling locations. Not every state’s exit polls include the same question items; we report results from the 26 states with complete data that held primaries on or before May 3rd, 2016. The total number of respondents and poll dates for each state are reported in supplementary Table S2.
Who Supported Trump?
We first explore the Axelrod question. Axelrod made an important distinction between “loyalty” and “contribution,” noting that a highly loyal but small group may provide a less significant contribution to an electoral coalition than a larger but less loyal group. To see this, let’s consider the general election context. It is more impactful for example, to receive 50% of the votes of whites than 90% of African Americans, as there are far more whites in the electorate as a whole. Loyalty scores highlight the degree to which a group is disproportionately attracted to a candidate, while contribution score reflects what share of the candidate’s total vote comes from that group, and thus provide an overall picture of the electoral coalition of candidates.
The ANES pilot asked respondents who among the Republican candidates they most preferred (and who was their second choice). In Table 1, we report these preferences across several sociodemographic categories. We compare those who prefer Trump to those who prefer Ted Cruz, and then to all respondents who reported preferring any GOP candidate other than Trump.5 Again, we report two measures: the percentage of each group who prefer a given candidate (“loyalty”) and the percentage of all those who prefer a candidate from a given group (“contribution”).
At a first glance, Trump’s overall popularity relative to his competitors is apparent. Nearly 27% of all whites, 22% of women, 27% of born again Christians, and 27% of southerners say they prefer Trump over other GOP candidates (Column A). These rates of support are four or five times the average rates of support for other GOP candidates, and more than twice what Cruz earns across these categories of respondent (surprisingly, even among evangelical Christians, although Cruz receives a significantly higher share of his total support from this group). Trump wins similarly high levels of popularity across all income and education categories, although loyalty declines among those with college degrees and those earning more than $110,000 per year (the top quintile of the income distribution). However, even as highly educated and high income voters prefer Trump less than their lower income and education counterparts, they are nevertheless more likely to prefer Trump than Cruz and more likely to prefer Trump than the average GOP primary candidate. For example, although only 18% of respondents with an income above $110,000 report Trump as their preferred candidate, the average GOP candidate (excluding Trump) wins only 7% of these respondents.
Although Trump bests any other GOP candidate among the wealthy and highly educated, these groups represent a smaller share of his supporters than they do of other candidates’ supporters. This distinction becomes clearer when we turn to the columns reporting the aggregate contributions of each group to candidates’ total votes (Column B). For example, 11% of those who prefer Trump earn over $110,000 per year. By contrast, on average, nearly 19% of those who prefer the other GOP candidates earn this income. In other words, the other GOP candidates rely more on high income respondents for support, whereas a larger part of the share of Trump supporters are middle or low income. This finding partially supports the view that the Trump coalition was distinctive in class terms, yet it is also important to note that Trump still outperforms his competitors among the high income respondents. Since the NES pilot was conducted before any votes were actually cast, however, and is a nationally representative sample rather than the much narrower group who vote in GOP primaries, we can only indirectly infer that Trump’s support was weighted downward. In the next section, we examine who actually voted for him as reflected in the exit polls.
Trump’s Share of Primary Voters Among Income and Education Groups
What about the actual voters? Figures 1 and 2 compare the median income and mean years of education, respectively, for exit poll respondents overall versus those who voted for Trump.6 Looking first at Figure 1, we notice that the median income for Trump voters nationally is about $9000 less than the median income for all GOP primary voters. However, the median Trump voter makes some $17,000 more than the median American. This same pattern repeats in virtually every state exit poll: Trump voters are less affluent than GOP voters in the state, but more affluent than the state median income (see Appendix Table A1 and A2).
Similar results obtain with respect to education. Whereas Trump voters have on average 0.28 fewer years of schooling than Republican primary voters in general, they have 0.84 more years of education than their state’s average resident. Taken together, these results show that while Trump’s voters may be somewhat less affluent and educated than the average GOP voter, they represent considerably higher socioeconomic status than others in their states.
In short: with a median income of $73,000 a year and nearly 15 years of education on average, the Trump electorate is relatively privileged. However, these voters are also slightly less affluent and educated than supporters of other GOP candidates, reflecting the patterns of participation in GOP primaries, in which voters are significantly more affluent and better educated than the average citizen of primary states. From this kind of evidence, it is very difficult to view Trump’s rise as reflecting a successful mobilization of downwardly mobile voters from below. Can these findings be reconciled with the rich ethnographies of scholars like Hochschild (2016) and Gest (2016), who have provided searching accounts of the attraction of white working class citizens to Trump? These scholars have argued, on the basis of ethnographic fieldwork in southwestern Louisiana and northeastern Ohio respectively, that white working class anger at the loss of economic stability and opportunity drive a significant percentage towards Trump (or to a Trump-like agenda). We think this work provides some powerful insights for understanding why so many non-college educated white voters looked favorably upon Trump, and voted for him in the general election. Nontheless, these groups were not the main source of Trump’s primary voters.
Much of the commentary about Trump’s surprising victories in electorally-rich Midwestern states in the general election, where the impact of global economic trends have been felt most sharply, is that Americans (and the working class in particular) disproportionately turned to Trump where local economic contexts were poor. What about during the primary season? Did Trump beat his competitors in places that are facing economic decline?
To consider this possibility, we look at the variation in Trump’s vote share across states in relation to three state-level economic variables: the 2015 unemployment rate, year-over change in the unemployment rate since 2014, and manufacturing as a share of total employment (all of these data come from the BLS). While we lack more fine-grained data in the exit polls, state unemployment rates and changes in unemployment rate reflect the labor market opportunities for state residents and provide significant variation across the country. Manufacturing as a share of total employment proxies both the “blue collar” composition of a state and the threat of import competition to local industry. Figures 3 through 5 depict the bivariate relationship between state-level economic conditions and Trump’s primary vote share for all voters, low income (<$50,000 per year) voters, and voters without a college degree, respectively. The story that emerges from each figure is the same: neither unemployment nor manufacturing as a share of the workforce predict Trump’s performance in a state primary, whether among all voters, the low income, or the less educated.
We cannot evaluate economic conditions at a more fine-grained level than the state as a whole. This is a clear limitation. However, taking advantage of Gallup’s Daily Tracking Survey microdata, Rothwell and Diego-Rosell (2016) are able to match survey respondents to their zip code, county, and commuting zone. Their results parallel our results here; they find that measures of local economic conditions, including intergenerational income mobility and the size of import-competing industries, fail to predict respondents’ support for Trump (Rothwell and Diego-Rosell 2016, p. 14).
Perceived Income Mobility and Trump Support by Income and Education
So far, we have evaluated how Trump support varies across concrete measures of socioeconomic status. Based on the January 2016 ANES and the exit polls, we maintain that Trump’s popularity was relatively broad and more affluent and better educated than most previous analysts have suggested. Further, it is not concentrated in regions facing economic decline. However, we can still interrogate one final important question. It is possible that Trump’s voters had different reasons to support him, depending on their personal economic situation or subjective view of the national economy, than non-Trump GOP voters. In other words, did those Trump supporters who were economically vulnerable (or perceived themselves to be so) more likely to support Trump? To assess this, we now consider a subjective measure of economic wellbeing included in the ANES pilot. This item asks respondents whether they think it was “easier to climb the income ladder 20 years ago than today.” This question captures well one of the prominent explanations for Trump’s appeal (e.g. Cherlin 2016), that is, that his populism speaks to Americans concerned about their chances for who perceive threats to upward mobility. We test whether this sense of stalled income mobility predicts positive attitudes toward Trump (using the ANES feeling thermometer), and whether this relationship varies across levels of income and education. The importance of objective or subjective income mobility has long been thought to be a classical source of working class authoritarianism (Lipset 1981 ), and, generally speaking, sociologists have devoted a great deal of attention to the centrality of perceived mobility chances for political attitudes (see e.g. De Graaf, Nieuwbeerta, and Heath 1995).
The bivariate relationship between subjective income mobility and respondents’ feelings toward Trump (and other GOP candidates) is depicted in Figure 6.7 Average positive feelings toward Trump are consistent across levels of respondents’ sense of income mobility, ranging between scores of 58 and 69 on the feeling thermometer. However, when we compare feelings toward Trump to the average feelings toward his GOP primary contenders, two noteworthy variations appear. Respondents who think that it is equally possible to climb the income ladder today as it was 20 years ago report an average of 10 more points in positive feelings toward other Republicans than toward Trump. In contrast, respondents who think it is a lot harder to climb the income ladder today report 20 more points in positive feelings toward Trump than toward other GOP candidates. In other words, although positive feelings toward Trump are consistent irrespective of respondents’ perceived income mobility, respondents who feel stalled upward mobility report feeling more positively toward Trump than they do toward other Republican candidates (and vice versa among respondents who feel their upward mobility has not changed in 20 years).
The multivariate analysis summarized in Figures 7 and 8 digs deeper into the null association between Trump support and subjective mobility to test for heterogeneity at different levels of socio-economic status provide nuance (Table A3 in the appendix presents the full results from these models). We regress the Trump feeling thermometer item on the measure of subjective income mobility, then test whether income or education moderate this relationship.8 Figures 7 and 8 below plot the predicted value of the Trump feeling thermometer as a function of subjective mobility in interaction with income and education, respectively, after controlling for basic demographic covariates. Generally speaking, the relationship between subjective income mobility and Trump support does not vary across levels of income and education.9
While these results are not definitive, they strongly suggest that at the beginning of the GOP primary season, subjective income mobility was not a strong predictor of Trump support. This holds whether we are considering the question in general terms, or among the working and middle class. Trump’s rise to the GOP nomination did not rest on mobilizing or bringing into the fold the most threatened segments of these classes.
We conclude that the view that Trump’s rise was fostered by his appeals to economically insecure voters is misplaced. Back-of-the-envelope assessments that mistook or misread the exit poll data showing Trump voters as lower-income or less-educated than other GOP contenders seemed to support such conclusions. But such arguments generally neglected to adjust for the demography of the GOP primary electorate, and thereby misread what was happening. But this conclusion holds even when we turn to nationally representative surveys including non-voters. Our regional analysis raises doubts about the role of regional economic distress as predictors of Trump support.
We do not want to claim more than we have demonstrated. Factors not considered here – most notably, the puzzle of how Trump continually dominated media coverage of the candidates throughout the primary season – are unquestionably important for Trump’s rise, and would require a separate analysis. Trump’s unique use of social media also deserves attention. His ability to communicate directly to the public is perhaps a harbinger of a new style of campaigning that challenges assumptions about political communication and the possibilities of media manipulation.
Other analysts have focused on the role of Trump’s appeal as an outsider in an era when traditional government institutions and the politicians are at a low point. There is ample and longstanding evidence that declining trust in government and public institutions is an important factor in citizens’ evaluation of national political institutions (e.g. Smith 2012; Hetherington and Rudolf 2015). And with regard to the more specific question about willingness to endorse government action, it is one of the few areas where the partisan divide in the mass public has increased; Republican identifiers are more likely to be opposed to a government role in the economy or trust the government to do what is best than Democratic identifiers, and this gap has grown over the past two decades (Baldassarri and Park 2016). However, we lack good measures of trust in government and candidate support from the primary season to rigorously investigate this possibility.
Similarly, our analysis does not consider how Trump’s appeals found ideological traction among a significant bloc of GOP primary voters. A line of commentary focusing on the growing racial and anti-immigrant partisan divide in the Obama era (e.g. Lopez 2014; Abrajano and Hajnal 2015; Tesler 2016) raises the possibility of a racial partisan realignment that may have played into Trump’s hands (as the dominant voice of anti-immigrant sentiment and, arguably, racial resentment). This possibility constitutes a potentially powerful model of emerging divisions within the GOP electorate that Trump’s campaign could exploit. Elsewhere we consider the role of broader cultural and ideological forces, including the role of racial resentment and anti-immigrant sentiment alongside the role of a broader countersubversive demonology in driving enthusiasm for Trump’s candidacy (Manza and Crowley 2017).
We conclude with a reminder that assertions about “authoritarian” tendencies among less affluent or educated citizens is often overblown. In his classical examination of the social bases of McCarthyism, Rogin (1967) noted that political sociological accounts of right-wing movements have sometimes been overly willing (given mixed evidence) to place the onus for “irrational” beliefs on the working class, even as authoritarian populism generally finds support across a wide-range of the public in those critical junctures where it explodes on the scene. The case of Trumpism may well reflect this larger pattern. Disbelief on the part of media commentators in the possibility that Trump’s voters are not just the downtrodden and poorly educated motivated the dominant line of commentary about the 2016 primary campaign. It is time to put those views to rest.
We thank Maria Soto for superb research assistance as we developed this paper, and Alex Maresca for help at the end. We received valuable feedback on an earlier draft from the Culture Workshop at New York University, and thank both the participants and co-conveners Paul DiMaggio and Iddo Tavory
Abrajano, Marisa, and Zoltan Hajnal. 2015. White Backlash: Immigration, Race, and American Politics. Princeton: Princeton University. Google Scholar
Abramowitz, Alan I. 2010. The Disappearing Center. New Haven: Yale University Press. Google Scholar
Aldrich, John H. 1995. Why Parties? The Origins and Transformation of Political Parties in America. Chicago: University of Chicago Press. Google Scholar
Alford, Robert. 1963. Party and Society: The Anglo-American Democracies. Chicago: Rand-McNally. Google Scholar
Axelrod, Robert. 1972. “Where Do the Votes Come From? An Analysis of Electoral Coalitions.” American Political Science Review 66: 11–20. Google Scholar
Baldassarri, Delia, and Andrew Gellman. 2008. “Partisans without Constraint: Political Polarization and Trends in American Public Opinion.” American Journal of Sociology 114: 408–446. CrossrefGoogle Scholar
Baldassarri, Delia, and Barum Park. 2016. “Was there a Culture War? Polarization and Secular Trends In U.S. Public Opinion.” Paper presented at the 2016 American Political Science Association Annual Meeting, Philadelphia, PA, September 3. Google Scholar
Bell, Daniel. 1960. The End of Ideology. On the Exhaustion of Political Ideas in the Fifties. New York: Free Press.Google Scholar
Bonvillian, William. 2016. “Donald Trump’s Voters and the Decline of American Manufacturing.” Issues in Science and Technology. Summer. Available at http://issues.org/32-4/donald-trumps-voters-and-the-decline-of-american-manufacturing/.
Brooks, David. 2016. “Not Trump, Not Ever.” New York Times March 18. http://www.nytimes.com/2016/03/18/opinion/no-not-trump-not-ever.html.
Campbell, James. 2016. “Seeing the Forest for the Trees: Presidential Election Forecasts and the Fundamentals.” PSNow September 22. Accessed September 29, 2016. Available at http://www.politicalsciencenow.com/seeing-the-forest-for-the-trees-presidential-election-forecasts-the-fundamentals/.
Cherlin, Andrew. 2016. “The Downwardly Mobile for Trump.” New York Times August 25. Accessed September 5, 2016. Available at http://www.nytimes.com/2016/08/25/opinion/campaign-stops/the-downwardly-mobile-for-trump.html?_r=0.
Collinson, Stephen. 2015. “Donald Trump Sinks GOP’s Dream Team.” CNN Politics August 18. Accessed September 15, 2016. Available at http://www.cnn.com/2015/08/18/politics/donald-trump-republican-2016-field/.
Clark, Dan. 2016. “Trump Was Outspent by His Closest Primary Opponents.” Politifact July 1. Available at http://www.politifact.com/new-york/statements/2016/jul/01/michael-caputo/trump-was-outspent-his-closest-primary-opponents/.
Cohen, Marty, David Karol, Hans Noel, and John Zaller. 2008. The Party Decides: Presidential Nominats Before and After Reform. Chicago: University of Chicago Press. Google Scholar
Davis, James. 1982. “Changeable Weather in a Cooling Climate Atop the Liberal Plateau: Conversion and Replacement in Forty-Two General Social Survey Items, 1972–1989.” Public Opinion Quarterly 56: 261–306. Google Scholar
De Graaf, Nan Dirk, Paul Nieuwbeerta, and Anthony Heath. 1995. “Class Mobility and Political Preferences: Individual and Contextual Effects.”American Journal of Sociology 100: 997–1027. CrossrefGoogle Scholar
Dionne, E. J. 2016. Why the Right Went Wrong: Conservatism From Goldwater to Trump and Beyond. New York: Simon and Shuster. Google Scholar
FiveThirtyEight.com. 2016. “The Endorsement Primary.” Available at https://projects.fivethirtyeight.com/2016-endorsement-primary/.
Geraghty, Jim. 2016. “Trump Has No One to Blame But Himself for Weak Ground Game.” National Review April 11. Available at http://www.nationalreview.com/article/433915/donald-trump-lost-delegates-weak-ground-game.
Gest, Justin. 2016. The New Minority: White Working Class Politics in an Age of Immigration and Inequality. New York: Oxford University Press. Google Scholar
Hacker, Jacob, and Paul Pierson. 2010. Winner-Take-All Politics. New York: Simon and Shuster. Google Scholar
Hare, Christopher, David A. Armstrong, Ryan Bakker, Royce Carroll, and Keith T. Poole. 2015. “Using Bayesian Aldrich-McKelvey Scaling to Study Citizens’ Ideological Preferences and Perceptions.” American Journal of Political Science 59: 759–774. CrossrefGoogle Scholar
Hochschild, Arlie. 2016. Strangers in Their Own Land. New York: The New Press. Google Scholar
Hofstadter, Richard. 1965. The Paranoid Style in American Politics and Other Essays. New York: Vintage Books. Google Scholar
Houtman, Dick. 2003. “Lipset and ‘Working-Class’ Authoritarianism.” American Sociologist 34: 86–105. Google Scholar
Irwin, Neil, and Josh Katz. 2016. “The Geography of Trumpism” The New York Times March 12. Available at http://www.nytimes.com/2016/03/13/upshot/the-geography-of-trumpism.html.
Jacobs, Nicholas, and James Ceaser. 2016. “The 2016 Presidential Election by the Numbers and in Historical Perspective.” The Forum 14: 361–383. Google Scholar
Kamarck, Elaine. 2016. Primary Politics: Everything You Need to Know About How America Nominates Its President. 2nd ed. Washington: Brookings Institute Press. Google Scholar
Kohn, Melvin L. 1977 . Class and Conformity: A Study in Values. 2nd ed. Chicago: University of Chicago Press. Google Scholar
Hetherington, Marc, and Thomas Rudolf. 2015. Why Washington Won’t Work: Polarization, Political Trust, and the Governing Crisis. Chicago: University of Chicago Press. Google Scholar
Kornhauser, William. 1959. The Politics of Mass Society. Glencoe, IL: The Free Press. Google Scholar
Lipset, Seymour Martin. 1963. The First New Nation. New York: Anchor Books. Google Scholar
Lipset, Seymour Martin. 1981 . Political Man. Expanded Edition. Baltimore: Johns Hopkins University Press. Google Scholar
Lipset, Seymour Martin, and Stein Rokkan. 1967. Party Systems and Voter Alignments. Glencoe, IL: The Free Press. Google Scholar
Lopez, Ian. 2014. Dog Whistle Politics: How Coded Racial Appeals Have Reinvented Racism and Wrecked the Middle Class. New York: Oxford University Press. Google Scholar
Manza, Jeff, and Ned Crowley. 2017. “Countersubversive Entrepreneur: Demonology and the Rise of Donald Trump.” Unpublished paper, Department of Sociology, New York University. Google Scholar
Oliver, J. Eric, and Wendy M. Rahn. 2016. “Rise of the Trumpenvolk: Populism in the 2016 Election.” The ANNALS of the American Academy of Political and Social Science 667: 189–206. CrossrefGoogle Scholar
Reiter, Howard L. 1985. Selecting the President: The Nominating Process in Transition. Philadelphia: The University of Pennsylvania Press. Google Scholar
Rogin, Michael. 1967. The Intellectuals and McCarthy. Cambridge: MIT Press. Google Scholar
Rothwell, Jonathan T., and Pablo Diego-Rosell. 2016. “Explaining Nationalist Political Views: The Case of Donald Trump” Working paper. Available at https://ssrn.com/abstract=2822059.
Schneider, Bill. 2015. “Obama’s Terror Speech Perfectly Highlights Reasons for Trump’s Rise.” Reuters December 7. Accessed September 15, 2016. Available at http://blogs.reuters.com/great-debate/2015/12/07/obamas-terror-speech-perfectly-highlights-reasons-for-trumps-rise/.
Shapiro, Ian. 2002. “Why the Poor Don’t Soak the Rich: Notes on Democracy and Distribution.” Daedalus 130: 118–128. Google Scholar
Sides, John, and Lynn Vavreck. 2013. The Gamble: Choice and Chance in the 2012 Presidential Election. Princeton: Princeton University Press. Google Scholar
Sides, John, and Henry Farrell. 2016. “The Science of Trump: Explaining the Rise of an Unlikely Candidate.” The Monkey Cage. Available via Amazon Digital Services at https://www.amazon.com/Science-Trump-Explaining-Unlikely-Candidate-ebook/dp/B01IL9DTG8.
Silver, Nate. 2016. “The Myth of Trump’s Working Class Supporters.” FiveThirtyEight May 3. Available at https://fivethirtyeight.com/features/the-mythology-of-trumps-working-class-support/.
Smith, Tom. 2012. “Trends in Confidence in Institutions, 1973–2006.” In Social Trends in American Life: Findings from the General Social Survey, ed. Peter V. Marsden, pp. 177–211. Princeton: Princeton University Press. Google Scholar
Stouffer, Samuel. 1955. Communism, Conformity, and Civil Rights. New York: Doubleday. Google Scholar
Surowiecki, James. 2016. “Donald Trump’s Sales Pitch.” The New Yorker August 10/17. Available at http://www.newyorker.com/magazine/2015/08/10/donald-trumps-sales-pitch.
Tesler, Michael. 2016. Post-Racial or Most-Racial? Race and Politics in the Obama Era. Chicago: University of Chicago Press. Google Scholar
Weigel, David. 2015. “Why Donald Trump makes Sense to many Voters – Even some Democrats.” The Washington Post. August 15. Available at https://www.washingtonpost.com/politics/why-donald-trump-makes-sense-to-a-lot-of-voters--even-some-democrats/2015/08/15/cee648f0-42bf-11e5-8ab4-c73967a143d3_story.html.
Zelizer, Julian. 2007. “Seizing Power: Conservatives and Congress Since the 1970s.” In The Transformation of American Politics: Activist Government and the Rise of Conservatism, edited by Paul Pierson, and Theda Skocpol, 105–134. Princeton: Princeton University Press. Google Scholar
The online version of this article offers supplementary material (DOI: https://doi.org/10.1515/for-2017-0002) offers supplementary material, available to authorized users.
In the US context, Lipset’s thesis drew from research on the timely subject of social and political tolerance, especially concerning the civil liberties of free speech and association (but also racial prejudice), which was found to be greater among those with higher education and occupational status and in positions of community leadership (e.g. Stouffer 1955).
The exact question wording is “Regardless of whether you plan on voting in the Republican primary, which Republican candidate do you prefer?” Three-hundred and fifty five respondents indicated “none.”
Data for individual states are reported in the appendix in Tables A1 and A2. The method for interpolating median income and mean years education from exit poll data is included in the supplementary materials. Our analysis parallels and extends that of Silver (2016).
With one exception: As shown on the right hand side of Figure 7, respondents who feel stalled income mobility and earn over $110,000 per year report much less positive feelings toward Trump. However, high income respondents reporting very low subjective income mobility represent a predictably sparse cell; they number only 17, so this estimate comes with high error.