Economic Populism and Bandwagon Bigotry: Obama-to-Trump Voters and the Cross Pressures of the 2016 Election

Abstract

Through an analysis of validated voters in the 2016 American National Election Study, this article considers the voters who supported Obama in 2012 and Trump in 2016. More than 5.7 million in total, Obama-to-Trump voters were crucial to Trump’s victory in the Electoral College. They were more likely to be white, working class, and resident in the Midwest. They had lower levels of political interest, were centrist in both party affiliation and ideology, and were late deciders for the 2016 election. On economic interests, they were centrists, except for trade policy, which they viewed, on average, as a greater threat than other voters. They claimed to have more experience with economic vulnerability than Democratic loyalists of comparable social standing. On racial attitudes, including the racialized economic topic of immigration, they had a profile similar to Republican loyalists. While their support of Trump may be attributable to surging white nativism, this article argues for an alternative explanation. Voters who were attracted by Trump’s economic populism only joined his coalition if they could accept his racialized rhetoric. As a result, the Trump bandwagon predominantly attracted generically bigoted voters with racial attitudes similar to Republican loyalists.

Full text

https://doi.org/10.1177/2378023119871119
Socius: Sociological Research for
a Dynamic World
Volume 5: 1 –15
© The Author(s) 2019
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/2378023119871119
srd.sagepub.com
Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-
NonCommercial 4.0 License (http://www.creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction
and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages
(https://us.sagepub.com/en-us/nam/open-access-at-sage).
Original Article
Introduction
When investigating the outcome of the 2016 presidential
election, many quantitative social scientists have asked
causes-of-effects questions: What configuration of causes
mattered, and which component causes mattered the most?
Guided by this logic of analysis, researchers have debated
the relative importance of economic interests, racial preju-
dice, white nativism, and ethno-nationalism, often using
vote-choice regression models for the full electorate that esti-
mate net associations simultaneously for measures of each of
these causes.
While important for our public understanding of US poli-
tics, these studies are also important for the scholarly litera-
ture itself, which, in the main, did not suggest Trump would
win by defeating establishment Republicans in the primaries
and then by flipping battleground states in the Midwest in the
general election. Understanding the causes of an unantici-
pated event is therefore an important goal for scholarly
research, and a causes-of-effects strategy is one approach
that can be taken.
Consider the following prominent attempt, which, if
media coverage is an indication, is likely to become one of
the most widely cited analyses of this type. In their book
Identity Crisis: The 2016 Presidential Campaign and the
Battle for the Meaning of America, Sides, Tesler, and Vavreck
(2018) present regression results that the net association
between racial attitudes and voting Republican increased
between 2012 and 2016. In contrast, the net association
between economic interests and voting Republican did not
change much at all (see their Tables A8.4 and A8.5).
Incorporating a few more claims about religion and immi-
gration based on their models, they concluded that
the origins of Trump’s unique appeal in the general election
were similar to the origins of his appeal in the primary: in both
cases, his candidacy helped to make identity-inflected issues
central to voters’ choices. And it was these issues that largely
explained the most notable demographic divide in the electorate:
between voters with more or less formal education.
871119SRDXXX10.1177/2378023119871119SociusMorgan and Lee
research-article2019
1Johns Hopkins University, Baltimore, MD, USA
Corresponding Author:
Stephen L. Morgan, Department of Sociology, Johns Hopkins University,
3400 N. Charles St., Baltimore, MD 21218, USA.
Email: stephen.morgan@jhu.edu
Economic Populism and Bandwagon
Bigotry: Obama-to-Trump Voters and the
Cross Pressures of the 2016 Election
Stephen L. Morgan1 and Jiwon Lee1
Abstract
Through an analysis of validated voters in the 2016 American National Election Study, this article considers the voters
who supported Obama in 2012 and Trump in 2016. More than 5.7 million in total, Obama-to-Trump voters were
crucial to Trump’s victory in the Electoral College. They were more likely to be white, working class, and resident in
the Midwest. They had lower levels of political interest, were centrist in both party affiliation and ideology, and were
late deciders for the 2016 election. On economic interests, they were centrists, except for trade policy, which they
viewed, on average, as a greater threat than other voters. They claimed to have more experience with economic
vulnerability than Democratic loyalists of comparable social standing. On racial attitudes, including the racialized
economic topic of immigration, they had a profile similar to Republican loyalists. While their support of Trump may be
attributable to surging white nativism, this article argues for an alternative explanation. Voters who were attracted by
Trump’s economic populism only joined his coalition if they could accept his racialized rhetoric. As a result, the Trump
bandwagon predominantly attracted generically bigoted voters with racial attitudes similar to Republican loyalists.
Keywords
electoral studies, political sociology, public opinion, cross pressures, voting

2 Socius: Sociological Research for a Dynamic World
The activation of these issues helped Trump win because there
were so many Obama voters whose views on these issues were
arguably closer to Trump’s than to Obama’s or Clinton’s—and
these voters were especially prevalent in battleground states.
Their shift to Trump helped him prevail in the Electoral College,
even while losing the popular vote. (Sides et al. 2018:178–79)
According to most arguments like this one, the activation of
identity, especially racial identity, is the primary mechanism
that delivered the victory to Trump in the general election.
This activation attracted additional white voters, dispropor-
tionately from the working class, into Trump’s winning
coalition.
How convincing is identity activation as the key explana-
tion? Whether Trump spoke about identity-relevant issues
during his campaign is not under serious debate. He certainly
did, and he has doubled down on the same strategy of rhe-
torical provocation multiple times since taking office. The
core question that remains under debate is whether the shift
of many white, working-class voters from support for Obama
in 2012 to support for Trump in 2016 can be “largely
explained” by identity activation in voters themselves. As we
discuss in the following, the primary alternative position is
that Trump’s consistent campaign theme of renegotiating
trade deals with China and Mexico to “bring back jobs” was
especially appealing to voters in Midwestern battleground
states and that its appeal was reinforced by his separate com-
mitment to reduce competition for working-class jobs and
wages by restricting immigration.
To assess the evidence for these positions while comple-
menting the relative-importance regression models of others,
in this article we analyze shifting electoral coalitions directly.
We return to the earliest sociological tradition of voting anal-
ysis, analyzing cross pressures like those that Lazarsfeld and
his colleagues considered in their work on elections in the
1940s. Although we consider the full electorate at the begin-
ning of our analysis, we then narrow our focus to non-His-
panic whites who voted for Obama in 2012 and then Trump
in 2016, with secondary consideration of non-Hispanic
whites who did not vote in 2012, even though eligible, but
then voted for Trump in 2016. These are the two largest
groups to join the coalition of Trump voters, after declining
to vote for Romney in 2012, and they include individuals
with the racial self-identification that makes them most sus-
ceptible to identity motivations of the sort that Trump is said
to have activated.
In the next section, we detail our theoretical orientation by
taking positions on the relevance for our analysis of several
voting concepts from the sociological literature. We then
summarize the current state of the debate on the 2016 elec-
tion, beyond the prominent piece by Sides et al. (2018) dis-
cussed previously. In so doing, we also lay the groundwork
for the subsequent empirical analysis that will inform our
core research question: How did Trump build his winning
coalition of voters? We offer an answer to this question in our
discussion section, contrasting two narratives for why his
coalition of voters secured the Electoral College victory that
eluded Romney: (1) surging white nativism and (2) eco-
nomic populism with bandwagon bigotry. The evidence
more strongly supports the latter, as we will explain below.
Back to Basics: Lazarsfeld and Cross-Pressured
Voters
The concept of cross pressures is most commonly attributed
to the 1944 book, The People’s Choice by Lazarsfeld,
Berelson, and Gaudet (see also Berelson, Lazarsfeld, and
McPhee [1954] 1986). In their analysis of the 1940 election,
Lazarsfeld and his colleagues identified two classic cross-
pressured groups—working-class protestants and affluent
Catholics—whose apparent loyalties to class and religion did
not promote consistent and “natural” support for either
Republican or Democratic candidates. For their study, they
utilized a monthly panel survey—the first of its kind—to
examine how respondents’ intended votes changed as candi-
dates were nominated in contested party conventions, cam-
paigns were then conducted, and final votes were cast.
Lazarsfeld and his colleagues ([1944] 1948) introduced
cross pressures as a general interpretive tool for empirical
analysis, especially appropriate for understanding the late
decisions that can be crucial for election outcomes. They
introduced the concept in this passage:
those who made their choice in the late days of the campaign
were people subject to more cross-pressures. By “cross-
pressures” we mean the conflicts and inconsistencies among the
factors which influence vote decision. Some of these factors in
the environment of the voter may influence him towards the
Republicans while others may operate in favor of the Democrats.
In other words, cross-pressures upon the voter drive him in
opposite directions. (P. 53)
The guiding result for the 1940 election, which was further
bolstered by a subsequent analysis of the 1948 election, was
their empirical demonstration of how cross pressures created
uncertainty and indecision that had to be resolved in order to
decide how to vote. In their analyses, cross-pressured voters
often failed to pass the “acid test” of actually turning up to
vote. And when they did vote, they often decided on their
preferred candidate only shortly before the election.
With this conceptualization, cross pressures are not sim-
ply conflicting attitudes and identities that are relevant to
political matters in general. Cross pressures are the specific
positions in the stratification order, interactional patterns in
primary and secondary groups, and issue preferences that
generate uncertainty and indecision as individuals deliberate
on future vote choices. In the aggregate, it is the resolution of
these cross pressures that determines electoral outcomes in
close contests, with variation across elections shaped by
social change at the individual level and political change at

Morgan and Lee 3
the party and candidate levels. Lazarsfeld and his colleagues
saw the study of how cross pressures are resolved to be a
major topic of study in voting research, and they identified
the potential mechanistic importance of media sources, cam-
paign propaganda, and social influence in structuring final
voting decisions.
Political scientists continue to use the language of cross
pressures in electoral studies, although often only as inter-
pretive language for characterizing putative swing voters.1
In sociology, however, the concept of a “cleavage,” as ini-
tially elaborated at length by Lipset, has received much
more attention in contemporary political sociology (see
Lipset 1960; Lipset and Rokkan 1967: chapter four).2
Cleavages are features of social structure that shape both
party strategy and the political predispositions of voters. In
typical usage, the cleavage concept is deployed to demar-
cate persistent political conflict across recognized social
divisions that generate alternative political interests (see
Manza and Brooks 1999). When voters have interests that
arise from cross-cutting cleavages, opportunities emerge
for parties, as well as entrepreneurial candidates, to assem-
ble new voting coalitions. In this way, a dynamic relation-
ship can emerge between voter preferences and party
appeals, with election outcomes determining whether new
coalitions arise. Emergent coalitions that persist across
multiple elections can then alter the understood mapping of
cleavages to parties.
For the 2016 election, cross pressures were present—as
we will show in the following—and these could be consid-
ered to arise from cross-cutting cleavages in social structure.
From this perspective, the current debate over the party loy-
alty of the white working class can be interpreted as a debate
over the primacy of the race and class identities that are lega-
cies of twentieth-century cleavages. This is a reasonable
position, but we embrace the concept of cross pressures
instead. It directs attention to the uncertainty and indecision
that operate at the individual level, where vote decisions are
enacted, rather than the plate tectonics of Lipset’s macroso-
ciology. It is also too early, and the 2016 election too unusual,
to evaluate the strong narrative that a crystallized race cleav-
age has trumped an eroding class cleavage in US politics.
Finally, because our inspiration is Lazarsfeld’s original
conception of cross pressures, which must be factors that
influence a particular vote decision, we depart modestly
from one notable and interesting usage of cross pressures in
the recent sociological literature. Baldassarri and Goldberg
(2014) consider the relationships among 40 attitudes that
were measured at least three times between 1984 and 2004
among voters and eligible nonvoters in eight iterations of the
American National Election Studies (ANES). They offer a
cluster analysis that classifies the eligible electorate in each
year into three groups of individuals—ideologues, alterna-
tives, and agnostics—who, they argue, adhere to alternative
political belief systems. For their yearly model solutions, the
alternatives have aggregate attitudinal patterns that
Baldassarri and Goldberg conclude are consistent with the
existence of cross pressures in the electorate as a whole, with
the archetype being those who are morally conservative but
economically liberal.
In their analysis, Baldassarri and Goldberg (2014) make
no reference to the Lazarsfeldian tradition of modeling actual
vote choices by analyzing cross pressures. Instead, they
attempt to unravel patterns of political polarization between
1984 and 2004 by using covariation from their attitudinal
measures twice: first to partition the electorate by a measure
of relationality and then to examine between-individual pat-
terns within the identified clusters. In so doing, they argue
that to understand trends in US politics, “it is necessary to
consider citizens’ political preferences as making up an inter-
dependent gestalt rather than a collection of independent atti-
tudinal vectors” (Baldassarri and Goldberg 2014:53).3
Our goal for analysis is much different. We do not aim to
classify individuals based on relational configurations of
attitudes or identify component belief systems based on
whether a cluster of individuals can be identified that does
not easily map on to a political party. We take it for granted,
based on decades of distinguished work by others, that atti-
tude and interest conflict is pervasive and, furthermore, that
complex configurations of attitudes and interests are endemic
to diverse modern political cultures that are constrained by
electoral rules that reinforce two-party dominance.
Instead, we aim to participate in the construction of a plau-
sible explanation for a particularly vexing recent election, and
1For example, Sides, Tesler, and Vavreck (2018) use the language of
cross pressures in this way. For a more general discussion and other
examples, see Brader, Tucker, and Therriault (2014).
2Lipset studied with Lazarsfeld and drew on the concept of cross
pressures when developing early versions of cleavage theory.
Likewise, cleavages are present as a secondary analytic distinction
in Berelson, Lazarsfeld, and McPhee ([1954] 1986). Lazarsfeld
later indicated that he left voting research after the 1950s because
he was interested in how decisions were made, while most sociolo-
gists were not. As he recounted in Lazarsfeld and Stehr (1982:152),
his work on voting was marked by “a strong concentration on deci-
sion processes,” asking “how did [people] decide how to vote.” This
focus became a “homeless skill or art” because “it was too individual
for sociologists, and too introspective for academic psychologists”
(the latter of whom had become “strongly behaviorist-oriented”).
3Although focused on the political science concept of a political
belief system, their analysis builds most directly on sociological
studies of social change and polarization that do not consider elec-
toral behavior (e.g., DiMaggio, Evans, and Bryson 1996; Evans
2003). This is presumably why Baldassarri and Goldberg (2014)
do not consider particular elections and the votes that determined
them. They also do not use many years of the American National
Election Studies (ANES), including data available for 2008 and
2012, which makes it difficult to connect their interpretations to
voting patterns from the Obama-era onward.

4 Socius: Sociological Research for a Dynamic World
we maintain that modeling how voters appear to have resolved
their cross pressures is vital for building a specific explanation
that can achieve a scholarly consensus. The clusters of the
electorate for our analysis, like for Lazarsfeld, are the actual
voting patterns that suggest how cross pressures were resolved.
These are the patterns that are most relevant for generating an
explanation of a specific election outcome. In particular, we
will show that much can be learned about the 2016 election by
considering how Obama-to-Trump voters were cross pres-
sured for the 2016 election and why their switch from the 2012
Democratic nominee to the 2016 Republican nominee must be
part of any convincing explanation for the 2016 outcome.
Competing Explanations for the 2016 Election
Outcome
Although estimates vary, Obama-to-Trump voters repre-
sented between 9 percent and 14 percent of Trump’s 62.9 mil-
lion general election voters, and they were disproportionately
prevalent among members of the white working class (see
e.g., Morgan and Lee 2018, Table 1 and Table S2, as well as
the supplemental material to this article). Partisanship
remained strong in 2016, and most stalwart Republican voters
were insufficiently attracted by Clinton’s campaign to break
with their party’s nominated candidate. Some Republican
loyalists embraced Trump, and some merely accepted him,
but rather few defected based on either the substance or style
of Trump’s appeal. Had enough of them done so, Obama-to-
Trump voters would not have mattered much for the outcome,
even though they could still have been of analytic interest.
Instead, Obama-to-Trump voters proved crucial for the out-
come because they expanded the reliably Republican elector-
ate by a sufficient amount to more than make up for the
defection of Romney-to-Clinton voters (see Tables S1 and S2
in the supplemental material for a comparison). With the sup-
port of Obama-to-Trump voters, numbering at least 5.7 mil-
lion, Trump secured a narrow victory in the Electoral College.
As debates over the election outcome have unfolded,
many topics remain on the table for consideration. An impor-
tant one is how best to characterize the decisions made by
Obama-to-Trump voters. At this point, on this question, a full
distribution of conclusions can be found in the literature
based on the relative weighting of economic interests and
some version of race-based group identity.
At one end of the distribution are those who appear to
strongly favor explanations based on white nativism, racial
resentment, and notions of status threat (see Mutz 2018a,
2018b), even if the argument is sometimes indirect (see
Jardina 2019). If there is an opposite end of the spectrum, it
would have to be those who focus on economic populism,
inflected by place-based economic interests, while holding
aside whether racial attitudes might be a complementary nar-
rative (see McQuarrie 2017; see also Judis 2016 for the pri-
mary election; cf. Judis 2018). In this vein, Stiglitz (2018:515)
focuses on the core of Trump’s economic appeal:
Trump’s campaign for the presidency exploited discontent
amongst large swaths of the American population. Blaming
others—migrants and “unfair” trade practices of other
countries—for their economic plight was much easier and more
satisfying than blaming, say, changes in technology and other
changes in the economy and the way the country had managed
those changes.
Responding to this diagnosis, Trump promised to renegotiate
trade agreements, and to assert America’s market power. He
threatened to impose trade restrictions—in the case of Mexico
and China, tariffs of 20% and 45% respectively—if our trading
partners did not respond “appropriately.” He focused on bilateral
trade deficits, promising to reduce those, and by implication,
even more strongly, the multilateral trade deficit. The reasoning
was simple: with less imports (or more exports), employment
and wages of his constituents would increase.
The argument of those who focus on economic interests is
that Trump’s appeal was effective, even understandable, if
accepted by voters in US counties that have had lagging eco-
nomic growth for the past several decades.
Most other scholars grapple with how to fit together
mechanisms based on economic interests and racial identity.
As noted previously, Sides et al. (2017, 2018) give only a
small amount of explanatory power to economic interests,
and mostly when economic arguments are sufficiently
“racialized.” This position is similar to the perspective of
Abramowitz (2018:140) that “it really was (mostly) about
race.” Norris and Inglehart (2018) discuss the declining rel-
evance of traditional left-right voting that was driven by eco-
nomic interests, but they argue that the sources of the
“cultural backlash” that benefitted Trump had diverse ori-
gins, including resentment of lost economic standing. This
argument echoes the position implied by the work of Lamont,
Park, and Ayala-Hurtado (2017), which documents the seem-
ing success of Trump’s effort to praise the dignity of work-
ing-class labor and the prosperity it helped to deliver in the
twentieth century.
Closer to the economic interest position, Morgan and Lee
(2017b, 2018) emphasize the complementarity of explana-
tions based on economic interests and racial prejudice yet
still give more weight to economic interests than most oth-
ers. In Morgan (2018b), a prejudice-incorporating narrative
is offered, which casts defections from the Democratic party
as a response to appeals for renegotiated trade agreements
and immigration restrictions that could benefit working-class
voters. The voters who responded to the appeal were those
with baseline levels of racial prejudice that did not render
Trump’s racialized rhetoric sufficiently disqualifying.
Many other streams of research have developed in response
to Trump’s election, including some that place Trump’s vic-
tory in comparative perspective in order to evaluate whether it
should be considered a right populist response (e.g.,
Bonikowski 2017; Mudde and Kaltwasser 2018), some that
probe the structural conditions that promoted the appeal of
Trump as a “lying demagogue” (see Hahl, Kim, and Zuckerman

Morgan and Lee 5
Sivan 2018), some that investigate Trump’s appeal to evan-
gelical Christians (Whitehead, Perry, and Baker 2018), and
some that consider the extent to which Trump’s economic
arguments had resonance with voters because of their folk,
rather than professional, modality (see Swedberg 2018; see
also Stiglitz 2018). Finally, more work will surely develop that
is candidate-specific, assessing the role of bravado and celeb-
rity in activating base evaluations with little issue-specific
rationale, building on the type of analysis encouraged by
Lodge and Taber (2013). A full explanation of the 2016 elec-
tion will need to incorporate results and conclusions from a
broad range of studies such as these. Our contribution is to
develop conclusions about the observable cross pressures of
the 2016 election, as explained previously, to enrich the debate
on the relative importance of Trump’s appeals to the economic
interests and racial identities of his supporters.
Data
We analyze validated voters from the ANES 2016 Time-
Series Study (see American National Election Studies 2017).
Details on the selection of the analytic sample are offered in
the supplemental material to this article, including how we
utilized vote-validation information to code votes cast in
2012 and 2016 as well as votes cast in 2016 among those
who did not vote in 2012.
Methods of Analysis
The core of our analysis is a comparison of the attributes,
attitudes, and issue preferences of three groups of validated
voters: Obama-to-Trump voters and two groups of party loy-
alists, Obama-Clinton voters and Romney-Trump voters. For
group comparisons such as these, we offer raw unadjusted
differences as well as differences that are adjusted for age,
gender, region, education, and class (as well as race-ethnicity
when we do not narrow the analytic sample to non-Hispanic
whites at the outset). For the latter set of adjusted differences,
we use logistic regression to estimate a balancing score based
on the adjustment variables. We then use a ratio weight to
standardize the group differences to the target group (Obama-
to-Trump voters) while using the same variables in between-
group weighted regression equations to adjust for lingering
imbalance not eliminated by the balancing score.4
Readers of the causal inference literature are correct to
infer that this strategy is a direct analog to doubly robust
inverse-probability-of-treatment weighting in the service of
estimating the average treatment effect for the treated. In this
case, the “treatment” is simply membership in the focal
group of interest for the analysis, and no assumption is intro-
duced that this group is a causal treatment of any type. It is
simply group membership, and the methods employed are
therefore best thought of as traditional table standardization
techniques, often associated with Kish (1987) and the demo-
graphic research tradition that preceded the full development
of the potential outcomes model of causality.
Measures
Details of the coding of basic demographic variables, educa-
tion, and class are also available in the supplemental mate-
rial. We conform to the conventions of the literature. For our
modeling of differences across types of voters, we analyze
92 distinct outcome measures. The most complete set of
results, which models all 92 outcomes separately, is offered
only in the supplemental material. For the main text of the
article, we group 85 of these outcomes into 24 composite
variables, scaled to have means of zero and standard devia-
tions of one. Not only do scaled composites promote rough
comparability, we are able to effectively reduce the dimen-
sionality of the data by an appropriate amount to preserve
space and bring the core patterns into clear relief (by averag-
ing over redundant variation from single indicators).
All composites and scaling procedures are detailed in the
supplemental material (see especially Table S9 for specific
question wordings), and our decisions are generally in line
with prevailing practices. Most important, the supplemental
material presents the additional results that demonstrate that
our findings in the main text are not driven by decisions
about whether or how to construct composite variables.
Results
Who were the Obama-to-Trump voters, and who were the
new Trump voters who sat out the 2012 election for a reason
other than age? Table 1 presents demographic profiles of
these two groups along with two reference groups, Obama-
Clinton voters and Romney-Trump voters, among validated
voters in the ANES. Consistent with Morgan and Lee (2018),
Obama-to-Trump voters were more likely than Obama-
Clinton voters to be members of the white working class.
The last four columns restrict the sample to white-only, non-
Hispanic (WONH) voters, and here the class differences are
particularly pronounced.5 In addition, Obama-to-Trump vot-
ers were more prevalent in the Midwest region, especially
when considering WONH voters only.
In contrast, the 2016 Trump voters who chose not to vote
in 2012 for a reason other than age appear to align more
4Tables S3 and S4 in the supplemental material show that the bal-
ancing scores perform well, implying that the additional covariance
adjustments play a very minor role.
5The measure of social class is explained in Morgan and Lee (2017a,
2017b), based on a coding of the dominant social class schema in
sociology, updated for use with 2010 census occupation codes (see
Morgan 2017 for details). For implementation with 2016 ANES
occupation codes, see the online supplement of Morgan and Lee
(2018). For the current article, the coding across all classes is less
precise than desired because of the detail provided by the ANES
occupation codes. For this reason, we always use the class measure
alongside the education measure in this article.

6 Socius: Sociological Research for a Dynamic World
closely with Romney-Trump voters, and they were some-
what more likely to reside in the South. Comparisons by edu-
cation and class are complicated by the fact that these voters
were also somewhat younger, but overall, their distribution is
also closer to Romney-Trump voters. Partly for economy of
space, and partly because this second group of voters is more
similar to party-loyalist Republicans, we will offer additional
results for this group of voters only in the supplemental
material. The Obama-to-Trump voters will be our primary
focus for the remainder of this section.
Cross Pressures and Obama-to-Trump Voters
Table 2 presents a first analysis of cross pressures, measured
by political predispositions and social attachments. As
explained in the methods section, the table compares Obama-
to-Trump voters to party-loyalist voters with and without
adjustments for age, race, gender, education, and class. Given
the focus of the debate summarized in the introduction, we
present results in the main text only for WONH voters, offer-
ing analogous results for all voters only in the supplemental
material (see Tables S5–S8).
To understand the table, consider the first row, which
reports results for a conventional 7-point scale of party iden-
tification. Without any adjustments (except narrowing the
sample to WONH voters only), Obama-to-Trump voters were
2.195 points above Obama-Clinton voters and 2.046 points
below Romney-Clinton voters (with standard errors of .263
and .261, respectively) on a scale of party identification where
a 1 is a strong Democrat and 7 is a strong Republican. With
adjustments, the differences change little, to 2.296 above and
1.965 points below. The proper interpretation is that Obama-
to-Trump voters are mostly centrists on party identification,
which is consistent with the second row of the table that pres-
ents differences for level of conservative ideology. This latter
result, however, is based on a four-item composite variable,
Table 1. Demographic and Class Differences between Obama-to-Trump Voters, New 2016 Trump Voters, and Party Loyalists.
All Validated Voters WONH Validated Voters Only
Voters in 2016
Obama and
Clinton
Obama and
Trump
No 2012
Vote, Then
Trump
Romney
and Trump
Obama and
Clinton
Obama and
Trump
No 2012
Vote, Then
Trump
Romney
and Trump
Age in years 54.66 59.80 46.53 57.31 56.41 60.99 46.65 57.59
Female .58 .53 .50 .52 .55 .53 .51 .53
WONH .61 .79 .91 .93 — — — —
Region
Northeast .26 .24 .15 .14 .26 .26 .14 .15
Midwest .19 .32 .22 .28 .24 .38 .22 .29
South .31 .31 .47 .39 .22 .27 .48 .38
West .25 .13 .16 .18 .27 .09 .16 .18
Education
High school or less .26 .50 .28 .28 .20 .48 .30 .28
Some college .26 .30 .41 .33 .24 .29 .41 .34
BA or more .48 .20 .30 .39 .55 .22 .29 .38
Missing .00 .01 .01 .00 .00 .01 .01 .00
EGP class
I .12 .05 .06 .12 .15 .06 .06 .12
II .35 .19 .21 .28 .38 .19 .20 .28
IIIa .09 .12 .08 .13 .08 .13 .09 .13
IIIb .11 .12 .13 .10 .09 .14 .14 .11
IVab .05 .03 .07 .09 .06 .02 .07 .10
V .05 .08 .12 .08 .05 .09 .12 .08
VI .03 .16 .06 .04 .04 .10 .06 .03
VIIa .13 .15 .14 .08 .09 .16 .14 .08
VIIb .01 .01 .00 .00 .01 .01 .00 .00
Military .00 .00 .01 .01 .00 .00 .01 .01
Missing .05 .10 .11 .08 .04 .09 .11 .08
Raw N711 99 214 626 470 80 193 581
Source: American National Election Studies 2016 Time Series Study (December 18, 2018 release).
Note: The sample is restricted to those aged 23 or older in 2016 so that those who were ineligible to vote in 2012 because of age were excluded.
Working-class categories include IIlb, VI, VIIa, and VIIb. Classes IIIa and V are generally considered intermediate classes. Classes I and II are professional
and semiprofessional workers, while class IVab includes self-employed workers. WONH = white-only, non-Hispanic.

Morgan and Lee 7
the details of which are offered in the supplemental material.
As noted in the row label, this conservative ideology compos-
ite variable (like all others in this table and subsequent ones)
was scaled to a standardized metric, with mean of 0 and stan-
dard deviation of 1. For this reason, the differences for the
second row are in standard deviation units and can be
(roughly) compared to differences based on other composite
variables (see the following). And, for this outcome, the
adjustment matters to some extent. Without an adjustment,
Obama-to-Trump voters are slightly farther away from
Obama-Clinton voters. After the adjustment for age, gender,
education, and class (and a perfect match on race), they are
more nearly equidistant from both types of party loyalists.
The change is attributable to using the doubly robust weight-
ing estimator to focus the comparison of Obama-to-Trump
voters to weighted sets of party loyalists who have the same
marginal distributions on the adjustment variables.
The remaining results in the table show that Obama-to-
Trump voters were more likely than Romney-Trump voters
to have involvement with unions and have positive feelings
about them. They were the least interested in politics, in
comparison to all party loyalists, and they were more likely
to feel that the rich control politics than were Romney-Trump
voters. Obama-to-Trump voters did not discuss politics as
frequently with family and friends, and they were more
likely to have decided which candidate to support within one
month of the date of the election. Altogether, Obama-to-
Trump voters have many of the general predispositions of
cross-pressured voters noted in the literature, especially
lower levels of interest and more pre-election indecision,
matching the original profile constructed by Lazarsfeld et al.
([1944] 1948).
The final panel of Table 2 presents differences for three
composite variables specifically used to assess a type of
support for the “Make America Great Again” theme of the
Trump campaign, interpreted literally. In a piece of schol-
arship that has mostly fallen off the citation radar screen,
Lipset and Raab (1970) analyze the “Quondam complex,”
which they introduce as “the condition of those who have
more of a stake in the past than in the present” (p. 460),
Table 2. Differences in Political Predispositions and Social Attachments, WONH Voters Only.
Raw Difference Adjusted Difference
From Obama-
Clinton Voters
From Romney-
Trump Voters
From Obama-
Clinton Voters
From Romney-
Trump Voters
Attachment and political identity
Party identification (7-point scale from
1 for strong Democrat to 7 for strong
Republican)
2.195 −2.046 2.296 −1.965
(.263) (.261) (.212) (.218)
Conservative ideology and affect
(standardized composite of 4 variables)
.975 −.787 .786 −.747
(.061) (.069) (.070) (.068)
Union involvement and affect
(standardized composite of 2 variables)
−.047 .800 −.213 .640
(.206) (.204) (.144) (.137)
Political engagement
Political interest (standardized composite
of 9 variables)
−.574 −.356 −.512 −.240
(.103) (.117) (.115) (.115)
Rich control politics (standardized
composite of 2 variables)
.035 .408 −.187 .432
(.134) (.123) (.129) (.125)
Days per week talk about politics with
family or friends (scale from 0 to 7 days)
−2.066 −1.219 −1.611 −1.103
(.394) (.398) (.339) (.317)
Decided how to vote within one month
of the election (indicator variable)
.331 .251 .263 .286
(.074) (.065) (.069) (.064)
Quondam complex
Value tradition and the past (standardized
composite of 4 variables)
1.160 −.158 .864 −.216
(.115) (.103) (.095) (.083)
Accept changes in morality (standardized
composite of 2 variables)
−.555 .508 −.481 .462
(.129) (.124) (.125) (.130)
Country needs free thinkers who will
have the courage to defy traditional
ways (5-point agreement scale)
.063 .276 .247 .348
(.115) (.116) (.113) (.111)
Source: American National Election Studie 2016 Time Series Study (December 18, 2018 release).
Note: Standard errors of the differences are in parentheses. The sample is restricted to those aged 23 or older in 2016 so that those who were ineligible
to vote in 2012 because of age were excluded. The raw N varies slightly by outcome but is usually 550 for the first and third columns and 657 for the
second and fourth columns, with the same 80 Obama-to-Trump voters in both sets of differences. Full details are provided in the supplemental material
(see Tables S10–S13). WONH = white-only, non-Hispanic.

8 Socius: Sociological Research for a Dynamic World
then elaborate as “a preponderance of symbolic investment
in the past, related to some past group identity which has
declined in symbolic significance” (p. 462), before quali-
fying, after its usage, that it is better thought of as a
“descriptive category” than “another analytic category” (p.
504). Table 2 shows that Obama-to-Trump voters have
notable but not atypical Quondam tendencies. They are
more likely than Obama-Clinton voters to value tradition
and the past, but they are squarely in between both party-
loyalist groups in accepting changes in morality. And, per-
haps most surprising, they rate higher by a bit than both
groups on supporting free thinkers who have the courage
to defy traditional ways. This last finding could point to
support for Trump’s persona, his penchant for defying
“political correctness,” or perhaps a more general desire
for change that drew these voters to Trump. Overall, how-
ever, Obama-to-Trump voters appear to have considerably
weaker Quondam tendencies than Romney-Trump voters,
especially when accepting changes in morality. This last
set of findings is consistent with the profile of cross-pres-
sured voters and less so that of the preservatist, right-wing
voters who were of interest to Lipset and Raab in their
analysis.
The Economic Interests of Obama-to-Trump
Voters
Table 3 presents analogous differences for eight composite
variables that measure economic interests in several differ-
ent ways, again for WONH voters only. As with all such
point-in-time comparisons, it is impossible to determine
how much the views of voters have been shaped by the posi-
tions of the candidates, and a weakness of the ANES project
is that many of its attitudinal measures are collected only
after the election.6 Nonetheless, the results in Table 3 sug-
gest that Obama-to-Trump voters are somewhat distinctive.
They reported more economic vulnerability than Obama-
Clinton voters, and they were solidly in between both groups
on economic progressivism and economic libertarianism.
These differences are slightly smaller for the models that
adjust for gender, age, education, and class.
Obama-to-Trump voters were more likely than party-
loyalist voters of both types to see global trade as a threat,
Table 3. Differences in Material and Economic Interests, WONH Voters Only.
Raw Difference Adjusted Difference
From Obama-
Clinton Voters
From Romney-
Trump Voters
From Obama-
Clinton Voters
From Romney-
Trump Voters
Personal and family conditions
Economic vulnerability
(standardized composite of 4 variables)
.665 .021 .530 −.018
(.163) (.164) (.172) (.152)
Economic policy predisposition
Economic progressivism
(standardized composite of 8 variables)
−.548 .850 −.465 .788
(.132) (.129) (.120) (.109)
Economic libertarianism
(standardized composite of 6 variables)
.702 −.648 .508 −.643
(.113) (.106) (.127) (.099)
Effects of trade and environmental policy
Global trade is a threat
(standardized composite of 4 variables)
.841 .265 .644 .269
(.123) (.129) (.109) (.102)
Environmental protection harms jobs
(7-point scale)
1.233 −1.156 1.087 −1.178
(.285) (.290) (.235) (.237)
Views of the economy as a whole
Economy is healthy (standardized composite
of 3 variables)
−.931 .312 −.721 .336
(.106) (.104) (.141) (.097)
Increasing opportunity and equality
(standardized composite of 3 variables)
−.181 −.374 −.314 −.365
(.143) (.139) (.158) (.119)
Source: American National Election Studies 2016 Time Series Study (December 18, 2018 release).
Note: Standard errors of the differences are in parentheses. The sample is restricted to those aged 23 or older in 2016 so that those who were ineligible
to vote in 2012 because of age were excluded. The raw N varies slightly by outcome but is usually 550 for the first and third columns and 657 for the
second and fourth columns, with the same 80 Obama-to-Trump voters in both sets of differences. Full details are provided in the supplemental material
(see Tables S10–S13). WONH = white-only, non-Hispanic.
6And since the political environment is always shifting, measures
from prior years (which are not available for the 2016 ANES) have
their own taint and do not reflect a primordial baseline that can be
relied on while assuming no elite-oriented distortion.

Morgan and Lee 9
especially in comparison to Obama-Clinton voters. They
were more centrist on the tradeoffs between environmental
protections and job growth, and the large differences in both
directions suggest how much party loyalists consistently
differ from each other on this issue.7 In addition, Obama-to-
Trump voters viewed the health of the economy in 2016 as
much more compromised than did Obama-Clinton voters,
and they were less convinced than Romney-Trump voters
that economic opportunities in the United States were
increasing or that inequality is stable or trending lower.
After adjustments, they may also differ in the same way
from Obama-Clinton voters.
Altogether, Obama-to-Trump voters saw themselves as
more economically vulnerable and the US economy as less
healthy than Obama-Clinton voters did on average. While
they were between both parties on general policy positions,
they saw global trade as more of a threat than did Obama-
Clinton voters and a little bit more of a threat than did
Romney-Trump voters. Given Trump’s economic message
on global trade (see the Stiglitz quote in the introduction), it
would not be surprising if some Obama-to-Trump voters
helped to resolve their cross pressures by seizing on this
issue as a reason to defect from Obama to join the Trump
coalition.
Racial Affect and Views on Policies to Address
Racial Inequality
Table 4 presents 10 sets of differences that, overall, suggest
that Obama-to-Trump voters had views on matters of race
that, in the main, resembled those of Romney-Trump voters
rather than Obama-Clinton voters. Again, these results are
restricted to WONH voters only, but see the supplemental
material for analogous results for all voters.
The overall pattern is clear for racial affect toward blacks
and Hispanics, prevalence of discrimination against whites,
support for government spending to help blacks, resistance to
the government offering “favors” to help blacks overcome
obstacles, and preference for majoritarian dominance. Obama-
to-Trump voters differed sharply from the perspectives of
Obama-Clinton voters on these matters, but they present a typi-
cal profile of generic Republican voters on these attitudes.
Aside from this baseline similarity across measures, a
small amount of more subtle variation is present. Obama-to-
Trump voters professed a bit more support for specific affir-
mative action policies than did Romney-Trump voters while
remaining much less supportive than Obama-Clinton voters.
They saw blacks and Hispanics facing slightly more obstacles
but nowhere near the same degree that Obama-Clinton voters
did. Nonetheless, they departed a bit from Romney-Trump
voters, and this pattern is consistent with a more centrist posi-
tion on abstract economic policies and a willingness for gov-
ernment intervention, as shown in Table 4.
For specific measures that may tap white nativism, such
as the composite for white racial affect and the measure of
discrimination against whites, Obama-to-Trump voters were
perhaps very slightly higher than even Romney-Trump vot-
ers. But, these differences, which are about the sizes of their
standard errors in most cases, do not align with other mea-
sures, such as the final indicator for whether whites should
work together to change laws unfair to whites (or the differ-
ences already discussed on support for affirmative action).
A key point to keep in mind, and one that we will discuss
at length in the following, is that Obama-to-Trump voters did
indeed vote for Obama in 2012 before defecting to Trump in
2016. Thus, it is hard to argue that these cross-pressured vot-
ers have a baseline voting tendency that is predetermined by
fixed racial prejudice of some form. They flipped from one
election to the next, and this sets them apart from Romney-
Trump voters. In other words, while Table 4 shows a lot of
similarity between Obama-to-Trump and Romney-Trump
voters, their votes in 2012 suggest an important difference on
racial sentiment and how it matters for their votes.
Immigration and Racialized Economics
For the last set of differences, Table 5 presents results for
four measures that probe attitudes toward immigration as
well as views on the economic consequences of immigration
policy. Not too dissimilar from the racial attitudes presented
in Table 4, the differences in Table 5 suggest that Obama-to-
Trump voters were also similar to Romney-Trump voters and
very much at odds with Obama-Clinton voters. They had
more negative attitudes toward all immigrants in general and
for those who are illegal/unauthorized/undocumented. They
were much more likely than Obama-Clinton voters to believe
that immigrants take away jobs and hurt the economy. And as
a result, they were less likely to favor increases in immigra-
tion. It is possible that they viewed immigration as more of a
threat to jobs than Romney-Trump voters, but only slightly
so and at a level that could be attributed to sampling error.
The main difference is with Obama-Clinton voters, who, like
their favored candidates, saw immigration as much less of a
threat to economic security.
Given the connection that Trump drew between immigra-
tion flows and competition for working-class jobs and wages
as well as the position of Clinton that such connections are
only a minor concern, it is not surprising that this contingent
of Obama voters may have resolved their cross pressures by
defecting to Trump. The results presented do not permit a
characterization of this mechanism as more “racial” or more
7Nonetheless, this measure is particularly difficult for a respondent
to interpret because of the way the 7-point scale is worded, including
an unreferenced “it” that we interpret as intending to mean “regulate
business to protect the environment” but that a satisficing respon-
dent could interpret as “no regulation,” implying that “no regulation”
“will not work” and “will cost jobs.” It may be that the middle-range
responses of this group represent a tendency to be confused by the
question and select values in the middle of the scale.

10 Socius: Sociological Research for a Dynamic World
Table 4. Differences in Racial Affect and Policies to Address Racial Inequality, WONH Voters Only.
Raw Difference Adjusted Difference
From Obama-
Clinton Voters
From Romney-
Trump Voters
From Obama-
Clinton Voters
From Romney-
Trump Voters
Racial affect
Positive black racial affect
(standardized composite of 4 variables)
−.760 .018 −.665 −.032
(.149) (.150) (.120) (.110)
Positive Hispanic racial affect
(standardized composite of 3 variables)
−.414 −.150 −.320 −.053
(.171) (.154) (.131) (.112)
Positive white racial affect
(standardized composite of 3 variables)
.478 .152 .426 .199
(.173) (.172) (.144) (.151)
Views of racial obstacles and discrimination
Blacks and Hispanics confront racial obstacles
(standardized composite of 3 variables)
−.766 .237 −.633 .243
(.124) (.099) (.114) (.104)
Discrimination against whites (standardized
composite of 2 variables)
.924 .101 .775 .133
(.144) (.146) (.138) (.128)
Policies to address racial differences
Government should help blacks (standardized
composite of 3 variables)
−1.175 .055 −.957 .027
(.111) (.098) (.098) (.098)
Blacks should work harder and should not
receive favors (standardized composite
of 2 variables)
1.270 .066 .951 .007
(.109) (.082) (.089) (.079)
Opposition to affirmative action (standardized
composite of 2 variables)
.616 −.272 .352 −.262
(.132) (.119) (.122) (.085)
White domination and solidarity
Majoritarian racial dominance
(standardized composite of 2 variables)
.962 .032 .648 −.006
(.131) (.122) (.111) (.113)
Important that whites work together to
change laws unfair to whites (5-point scale)
.658 −.053 .702 −.198
(.203) (.194) (.171) (.160)
Source: American National Election Studies 2016 Time Series Study (December 18, 2018 release).
Note: Standard errors of the differences are in parentheses. The sample is restricted to those aged 23 or older in 2016 so that those who were ineligible
to vote in 2012 because of age were excluded. The raw N varies slightly by outcome but is usually 550 for the first and third columns and 657 for the
second and fourth columns, with the same 80 Obama-to-Trump voters in both sets of differences. Full details are provided in the supplemental material
(see Tables S10–S13). WONH = white-only, non-Hispanic.
Table 5. Differences in Attitudes toward Immigrants and the Economic Consequences of Immigration, WONH Voters Only.
Raw Difference Adjusted Difference
From Obama-
Clinton Voters
From Romney-
Trump Voters
From Obama-
Clinton Voters
From Romney-
Trump Voters
Attitudes toward immigrants
Positive immigration affect (standardized
composite of 6 variables)
−1.298 −.183 −.967 −.124
(.134) (.125) (.103) (.091)
Negative views of unauthorized
immigrants (standardized composite of
2 variables)
1.099 −.020 .944 .002
(.152) (.140) (.115) (.115)
Economic consequences of immigration
Immigrants take jobs and hurt economy
(standardized composite of 2 variables)
1.149 .132 .926 .162
(.148) (.155) (.137) (.136)
Number of immigrants should be
increased (5-point scale)
−1.161 .085 −.940 .128
(.173) (.155) (.144) (.134)
Source: American National Election Studies 2016 Time Series Study (December 18, 2018 release).
Notes: Standard errors of the differences are in parentheses. The sample is restricted to those aged 23 or older in 2016 so that those who were ineligible
to vote in 2012 because of age were excluded. The raw N varies slightly by outcome but is usually 550 for the first and third columns and 657 for the
second and fourth columns, with the same 80 Obama-to-Trump voters in both sets of differences. Full details are provided in the supplemental material
(see Tables S10–S13). WONH = white-only, non-Hispanic.

Morgan and Lee 11
“economic,” and for this reason we adopt the language of
Sides et al. (2018) that this is an instantiation of racialized
economics, much more so than for the types of differences
presented in Table 4.
How Much Additional Clarity Would Many-
Cause, Multiple-Regression Models Deliver?
Impatient readers may not appreciate the level of detail in the
foregoing tables, and if so, they will like the supplemental
material even less, where the same results are offered for
alternative analytic samples and in addition for each underly-
ing indicator (see Tables S10–S13). When the goal of analy-
sis is to consider groups like cross-pressured voters, who by
definition lie in the interior of the multidimensional cloud of
points that relates attitudes and interests to candidate prefer-
ences, adjusted group differences reveal the core patterns in
the data that need to be considered.
In holding this position, we would appear to be in the
minority. Most other quantitively oriented scholars in these
debates have centered their analyses on many-cause, multi-
ple-regression models that attempt to explain how each of
these factors predicts vote choices, net of each other, and
without focusing in any explicit way on the crucial Obama-
to-Trump voters. To allay concerns that we are withholding
the insight of such analysis strategies, we offer multiple
regression results in the supplemental material (see Tables
S14 and S15 and their associated interpretations in the
text). We conclude the supplement material by explaining
why we believe that many-cause, multiple-regression mod-
els do not provide the insight that others have claimed.
Conclusions
Our results suggest seven primary conclusions on Obama-to-
Trump voters based on an analysis of validated voters in the
ANES:
1. Likely more than 5.7 million in number, Obama-to-
Trump voters constituted a large enough share of
Trump’s voters that they were essential for his win.
2. They were more likely to self-identify as white-only,
non-Hispanic (or WONH, using our acronym) and
especially within this group, had lower levels of edu-
cation, higher rates of working-class membership,
and lived in the Midwest.
Focusing specifically on Obama-to-Trump voters who self-
identified as WONH, in the main text of this article, we have
offered results that allow us to conclude:
3. They had lower levels of political interest and were
centrist in both party affiliation and ideology.
4. They were more likely to be late deciders for the
2016 election.
5. On economic interests and issue positions, they were
centrists, except for trade policy, which they saw as a
greater threat than did both Democratic and
Republican party loyalists.
6. They claimed to have more experience with eco-
nomic vulnerability than Democratic party loyalists
of comparable social standing.
7. On a broad range of racial attitudes, including the
racialized economic topic of immigration, they had a
profile very similar to Republican loyalists, with
slightly more recognition of racial obstacles than
Republican loyalists and slightly more support for
affirmative action.
In the supplemental material (see Tables S12 and S13), we
offer a set of additional results that suggest three conclusions
worthy of mention here:
8. Voters who declined to vote for any candidate in
2012 but who turned out to vote for Trump in 2016
had profiles similar to Republican loyalists on both
demographic characteristics and attitudes toward
race.
9. They were more centrist on economic issues than
Republican loyalists but less centrist than Obama-to-
Trump voters.
10. They had less political interest than Democratic and
Republican loyalists, but they were not late deciders
for the 2016 election.
One contribution of this last group of Trump voters was to
boost the turnout rate in 2016 relative to 2012, disproportion-
ately for the white working class (see Morgan and Lee
2017b). Overall, these conclusions reinforce our view that it
is the Obama-to-Trump voters who deserve the most inter-
pretive work, but full results are available in the supplemen-
tal material for readers who wish to consider this second
group more comprehensively.
Discussion
With our first-order conclusions in hand, in this section we
tackle the harder question: Why did a substantial number of
Obama’s voters defect to Trump, delivering a victory in the
Electoral College? We provide two plausible answers: (1)
surging white nativism and (2) economic populism with
bandwagon bigotry. Before presenting the answers, we sum-
marize some additional empirical results that are vital for
understanding why we also argue that the second answer is
more convincing than the first.
Relevant Results from Recent Research
Before developing a more general explanation for the choices
of Obama-to-Trump voters, three sets of related findings

12 Socius: Sociological Research for a Dynamic World
from the recent literature should be foregrounded. First,
results from the 2004 through 2016 General Social Surveys
(GSS) show that the racial attitudes of whites, including
those from the working class, did not change much for the
2016 election cycle. If any meaningful change is present, it is
one of moderation in opposition to affirmative action and in
racial prejudice (see Morgan and Lee 2017b, Figure S7, for
results with 4-item and 11-item scales of each).
Second, results from the 2018 GSS further suggest that if
any change is present in the GSS results since the 2016 elec-
tion, it is clearly in the direction of a further moderation of
racial attitudes (see Holland 2019). In other words, a com-
parison of both the 2016 and 2018 GSS results to those from
previous years provides evidence that should allay the emi-
nently reasonable concern that Trump’s emphasis on identity
issues has pushed the racial prejudice and resentment of
whites higher since he entered national politics in 2015.
Third, work with the ANES is consistent with the GSS in
demonstrating little or no change in the racial attitudes of
whites between 2012 and 2016, and the ANES measures were
mostly collected in close proximity to the elections (usually
for the November-December postelection surveys). This lack
of change is not particularly clear in the work that has empha-
sized identity mechanisms. Sides et al. (2018), for example,
offer many claims about changing associations between vote
choices and racial attitudes between 2012 and 2016. But only
in their final chapter (see Figure 9.3) do they present the key
over-time results from the ANES and GSS on these racial atti-
tudes. And these trends are inconsistent with the claim that
identity activation increased white nativism among Trump
voters during the 2016 election campaign. Likewise, Jardina
(2019; Table 3.2), in a book that comprehensively examines
the specific activation of white identity in modern US poli-
tics, shows that a comparison of the 2012 and 2016 ANES
data reveals no increase (and possibly a decline) in “white
racial identity” between 2012 and 2016. However, the finding
is deemphasized, and the book concludes without a direct
analysis of the 2016 general election using the ANES data,
leaving the overall claim—that whiteness is a major emergent
issue in national politics—less than fully evaluated as a deter-
minant of the 2016 general election.8
The importance of these results deserves emphasis. The
ANES and GSS, which are widely regarded as the highest
quality surveys available on these matters, could have pro-
vided evidence for movement in racial attitudes consistent
with concerns about a surge in white nativism or more gen-
eral racial prejudice. They do not provide any evidence to
support the existence of such a surge, and these results must
be folded into a consideration of how best to interpret the
cross-pressured voters that broke for Trump at the end of the
campaign. We now turn to the two plausible explanations
for why many cross-pressured Obama voters defected to
Trump.
Possible Answer: Surging White Nativism
This explanation begins with the 2012 election. In that
election, Romney engaged in standard “racial politics,”
nudging voters to think about race indirectly by using
facially race-neutral arguments about government depen-
dency. The best example was his leaked talk to donors
where he stated on May 17, 2012: “There are 47 percent of
the people who will vote for the president [Obama] no
matter what . . . who are dependent upon government, who
believe that they are victims.” Campaigning in this way, he
reminded Republican voters that he agreed with the posi-
tions most strongly associated with Reagan’s campaign
against welfare dependency, with its clear racial overtones,
but without the explicitness that less urbane politicians
would embrace.
When then considering how to vote in 2012, a sufficient
number of cross-pressured voters decided to support Obama,
in part because their economic centrism left them unim-
pressed by Romney. He was characterized by the Obama
campaign as an out-of-touch elitist, whose personal wealth
was generated by advising firms to lower the wages of work-
ers. These cross-pressured voters supported Obama because
they wanted to give him another four years to see if he could
grow the economy in an equitable and balanced way, and the
support of labor unions, who reminded them of Obama’s
auto bailout and other stimulus funding, was a crucial factor
in their thinking. Altogether, according to this explanation,
Romney did not prime racial sentiments overtly enough to
excite white voters, and thus the cross-pressured voters from
the working class resolved to support Obama for economic
reasons.
When Trump campaigned in 2016, the election had an
entirely different feel, not just because of the absence of an
incumbent. Following from the “white nativism” explana-
tion advanced in the literature, it must have been the case that
white voters, especially those from the working class, saw
the election this way:
8Jardina (2019) analyzes ANES data from the 2016 primaries, using
what is known as the 2016 ANES Pilot Study. She also uses the
2016 general election ANES data (the same “Time-Series Study” we
analyze) to consider the distribution of white identity (see Table 3.2),
which did not change between 2012 and 2016, and also for other
types of analysis (see, e.g., her Tables 4.2–4.5, 8.1). In our reading,
we can find no results in the book that use the available ANES 2016
Time-Series Study data to directly analyze the relationship between
white identity and the 2016 general election outcome. For her analy-
sis of the 2016 general election, she opts instead to use briefly a two-
wave YouGov study that she fielded with some of her own measures
and appears to have 480 valid respondents (and, we assume, uses
YouGov’s weighted quota sampling strategy of opt-in empaneled
respondents, unlike the full probability sample of the ANES).

Morgan and Lee 13
While Clinton uses her campaign speeches to celebrate diversity
and global citizenship, Trump speaks about how powerful
America was in the past, and that this power had nothing to do
with diversity. Instead, it was white Americans who won wars
and built the modern economy.
Trump is courageous to point out that America’s security is
threatened by Muslims, and its government overwhelmed by
Mexicans who entered the country illegally. And he is obviously
correct in reminding everyone that America’s manufacturing
base has been undermined by globalists who want to embrace all
other nations, rather than support their own.
For these reasons, Trump is right that the country cannot return
to its greatness without looking to what worked in the past.
Clinton is wrong when arguing that the goal of eliminating racial
inequality is essential for moving the country forward. Such
inequalities are inevitable, for complex reasons, some of which
cannot be changed, and Trump is right that the government
should align itself with the people of the country who have given
the most in the past. It is patriotic to celebrate American
greatness, and one should not feel ashamed that it was mostly
white Americans who are responsible for that greatness.
Thinking in this way, many white voters, especially those
with low levels of education and limited experiences in
racially diverse environments, chose to shift loyalties from
Obama to Trump. Their loyalty to Obama was narrowly
instrumental in the first place and had little or no affective
foundation. In contrast, the way Trump spoke about the
country was something they could get excited about.
Is this explanation correct? It is hard to dismiss it as
entirely unreasonable, and surely many voters responded to
Trump’s campaign in a way that could be quite similar to
our characterization. The question that must be confronted,
however, is whether this way of thinking can be attributed
to those who actually voted for Obama in 2012. It is more
likely, we think, that voters prone to this type of efferves-
cent white nativism are precisely the sorts of voters who
have favored Republicans consistently. Obama-to-Trump
voters had not been convinced in 2012 by Tea Party activ-
ists or the widely reported 2011 birtherist claims of Trump
himself. They had instead decided to turn out to support
Obama despite their underlying prejudice. For this reason,
the aforementioned thinking seems much more applicable
to voters who either (1) supported Romney in 2012 because
the Democratic party does not welcome bigotry or (2) did
not vote in 2012 because neither party endorsed white
supremacy.
Still, it is possible that Trump’s identity rhetoric was con-
vincing in a new way in 2016, perhaps because of the appeal
of his bombast, and that some Obama-to-Trump voters sup-
ported him primarily because of it. The only thing we are
absolutely certain of is that we are not among those con-
vinced by arguments of the sort advanced by Mutz (2018a,
2018b), which suggest that many white Obama voters came
to regret their 2012 votes, fell into an affective trap in
defense, and then were transformed in 2016 because Trump
brought out their latent feelings of status threat.9
A More Convincing Answer: Economic Populism
with Bandwagon Bigotry
We see a larger role for economic interests in their own right,
directly considered and acted on.10 To understand our posi-
tion on the 2016 election, we recommend that readers first
search for the words trade and job in the many online sources
for the transcripts of the three presidential debates in 2016.
Trump was relentless and Clinton almost defenseless when
he argued that jobs had been lost to China and Mexico and
that he, by restricting immigration and renegotiating trade
deals, could bring them back. This is the argument summa-
rized by Stiglitz (2018), as presented in the quotation in the
introduction.
Here is how this explanation is structured in light of the
evidence that Obama-to-Trump voters differed from
Democratic loyalists in their racial attitudes: First, the expla-
nation for supporting Obama in 2012 is largely the same as
presented previously for the first stage of the “surging white
nativism” explanation. When voting for Obama in 2012,
Obama-to-Trump voters indicated a willingness to hold aside
their racial prejudice because Romney was unappealing,
especially on economic matters.
By 2016, these voters, who had generic Republican levels
of racial prejudice already, viewed Trump’s identity rhetoric
as unexceptional and thus largely ignorable. They believed
that supporting employment in manufacturing, as Obama did
in the Midwest for the automobile industry, was a reason to
have voted for Obama in 2012, and especially when the alter-
native was a candidate who built his own personal wealth by
recommending the overseas outsourcing of jobs. These vot-
ers also saw their sentiments toward recent immigrants not as
prejudice, or as a defense of national culture, but simply as a
reflection of their view that low-skilled immigrants are a
threat to employment security for working-class voters in
their states. They could not accept what they heard Democrats
claiming, which was that debates over immigration are not
9Mutz’s particular argument has no support because of methodolog-
ical errors in her first piece, even apart from measurement assump-
tions that are hard to accept (see Morgan 2018a, 2018b).
10We should note also that many Obama voters from the working
class, especially those who self-identify as nonwhite or Hispanic,
found Trump’s rhetoric on race disqualifying. These voters mostly
supported Clinton, but it is unknown whether some of the turnout
decline of nonwhites in the Midwest could have resulted from unre-
solved cross pressures that prevented some of these voters from
turning out to support Clinton. Some of these voters may have sup-
ported Trump’s trade and immigration proposals, but the ANES
sample is too small to support a direct analysis.

14 Socius: Sociological Research for a Dynamic World
economic at their core, and they were offended that media
commentators imputed racist motives to what they felt were
legitimate economic concerns.
Finally, if many WONH Obama-to-Trump voters had
felt genuinely race neutral, perhaps using the reasonable
justification that they had voted for Obama only four years
prior, then voting for Trump would not feel like a racist act.
If these same voters were cross pressured and associated
regularly with other voters who had palpable enthusiasm
for Trump’s overall message and style, then the path of
least resistance in navigating the cross pressures may have
been to vote for Trump. It could almost seem harmless to do
so if the media were correct in predicting that Clinton
would cruise to a victory anyway. And, if some of these
voters instead savored a change of pace in national politics,
even a transgressive disruption to the Washington estab-
lishment that Clinton had been a part of for 25 years, then
Trump would surely deliver and perhaps also strengthen the
economy at the same time.
We do not, of course, have direct evidence for many of
the components of this explanation, but it is consistent with
the results we have presented. And, we see a lot of inconsis-
tency between empirical results and explanations that focus
instead on some version of surging white nativism. If white
identity was activated so effectively that group identity is
the most important factor to consider, one would expect to
see measures of racial prejudice, opposition to affirmative
action, and white racial consciousness to have increased in
2016 for whites and then, probably, increasing further since
Trump took office. The evidence that we have seen all runs
counter to this implication. Racial attitudes and new mea-
sures of white identity show no direct evidence that Obama-
to-Trump voters or even Trump voters in general changed
in the direction suggested by the surging-white-nativism
explanation.
The biggest change that we have seen since 2015 is the
strong perception on the part of voters that Trump favored
unprecedented restrictions on immigration and renegotiated
trade agreements, and the evidence is consistent with the
idea that these positions won him centrist voters prone to
appeals grounded in economic populism. Because voters
who were repelled by Trump’s identity rhetoric did not shift
to support him, the bandwagon predominantly attracted
generically bigoted voters who differed little from Republican
loyalists on average levels of racial prejudice and white
nativism.
Acknowledgments
We thank Jesper Sørensen for his comments.
ORCID iD
Stephen L. Morgan https://orcid.org/0000-0003-2198-1381
Supplemental Material
Supplemental material for this article is available with the article
online.
References
Abramowitz, Alan I. 2018. The Great Alignment: Race, Party
Transformation, and the Rise of Donald Trump. New Haven,
CT: Yale University Press.
American National Election Studies. 2017. “User’s Guide and
Codebook for the ANES 2016 Time Series Study.” The
University of Michigan and Stanford University, Ann Arbor,
MI and Palo Alto, CA.
Baldassarri, Delia, and Amir Goldberg. 2014. “Neither Ideologues
nor Agnostics: Alternative Voters’ Belief System in an Age
of Partisan Politics.” American Journal of Sociology 120(1):
45–95.
Berelson, Bernard, Paul Felix Lazarsfeld, and William N. McPhee.
[1954] 1986. Voting: A Study of Opinion Formation in a
Presidential Campaign. Chicago: University of Chicago Press.
Bonikowski, Bart. 2017. “Ethno-Nationalist Populism and the
Mobilization of Collective Resentment.” The British Journal
of Sociology 68:S181–213.
Brader, Ted, Joshua A. Tucker, and Andrew Therriault. 2014.
“Cross Pressure Scores: An Individual-Level Measure of
Cumulative Partisan Pressures Arising from Social Group
Memberships.” Political Behavior 36:23–51.
DiMaggio, Paul, John Evans, and Bethany Bryson. 1996. “Have
American’s Social Attitudes Become More Polarized?”
American Journal of Sociology 102(3):690–755.
Evans, John H. 2003. “Have Americans’ Attitudes Become More
Polarized?—An Update.” Social Science Quarterly 84:71–90.
Hahl, Oliver, Minjae Kim, and Ezra W. Zuckerman Sivan. 2018.
“The Authentic Appeal of the Lying Demagogue: Proclaiming
the Deeper Truth about Political Illegitimacy.” American
Sociological Review 83(1):1–33.
Jardina, Ashley. 2019. White Identity Politics. New York:
Cambridge University Press.
Judis, John B. 2016. The Populist Explosion: How the Great
Recession Transformed American and European Politics. New
York: Columbia Global Reports.
Judis, John B. 2018. The Nationalist Revival. New York: Columbia
Global Reports.
Holland, Jesse J. 2019. “More American Say Gov’t Spends Too
Little Helping Blacks.” Associated Press, March 27. https://
apnews.com/69f6eccb65b242789a1e86e15d755567.
Kish, Leslie. 1987. Statistical Design for Research. New York:
Wiley.
Lamont, Michèle, Bo Yun Park, and Elena Ayala-Hurtado. 2017.
“Trump’s Electoral Speeches and His Appeal to the American
White Working Class.” The British Journal of Sociology
68:S153–80.
Lazarsfeld, Paul Felix, Bernard Berelson, and Hazel Gaudet. [1944]
1948. The People’s Choice: How the Voter Makes up His Mind
in a Presidential Campaign. New York: Columbia University
Press.
Lazarsfeld, Paul F., and Nico Stehr. 1982. “A Conversation with
Paul F. Lazarsfeld.” The American Sociologist 17:150–55.

Morgan and Lee 15
Lipset, Seymour Martin. 1960. Political Man: The Social Bases of
Politics. Garden City, NY: Doubleday.
Lipset, Seymour Martin, and Earl Raab. 1970. The Politics of
Unreason: Right-Wing Extremism in America, 1790–1977.
New York: Harper Collins.
Lipset, Seymour Martin, and Stein Rokkan, eds. 1967. Party
Systems and Voter Alignments: Cross-National Perspectives.
New York: Free Press.
Lodge, Milton, and Charles S. Taber. 2013. The Rationalizing
Voter. Cambridge: Cambridge University Press.
Manza, Jeff, and Clem Brooks. 1999. Social Cleavages and
Political Change: Voter Alignments and Us Party Coalitions.
New York: Oxford University Press.
McQuarrie, Michael. 2017. “The Revolt of the Rust Belt: Place and
Politics in the Age of Anger.” The British Journal of Sociology
68:S120–52.
Morgan, Stephen L. 2017. “A Coding of Social Class for the
General Social Survey.” GSS Methodological Report No. 125,
National Opinion Research Center.
Morgan, Stephen L. 2018a. “Correct Interpretations of Fixed-
Effects Models, Specification Decisions, and Self-
Reports of Intended Votes: A Response to Mutz.” Socius.
doi:10.1177/2378023118811502
Morgan, Stephen L. 2018b. “Status Threat, Material Interests,
and the 2016 Presidential Vote.” Socius. doi:10.1177/23780
23118788217
Morgan, Stephen L., and Jiwon Lee. 2017a. “Social Class and
Party Identification during the Clinton, Bush, and Obama
Presidencies.” Sociological Science 4:394–423.
Morgan, Stephen L., and Jiwon Lee. 2017b. “The White Working
Class and Voter Turnout in U.S. Presidential Elections, 2004 to
2016.” Sociological Science 4:656–85.
Morgan, Stephen L., and Jiwon Lee. 2018. “Trump Voters and the
White Working Class.” Sociological Science 5:234–45.
Mudde, Cas, and Cristóbal Rovira Kaltwasser. 2018. “Studying
Populism in Comparative Perspective: Reflections on the
Contemporary and Future Research Agenda.” Comparative
Political Studies 51:1667–93.
Mutz, Diana C. 2018a. “Response to Morgan: On the Role of Status
Threat and Material Interests in the 2016 Election.” Socius.
doi:10.1177/2378023118808619
Mutz, Diana C. 2018b. “Status Threat, Not Economic Hardship,
Explains the 2016 Presidential Vote.” Proceedings of the
National Academy of Sciences 115:E4330–39.
Norris, Pippa, and Ronald Inglehart. 2018. Cultural Backlash:
Trump, Brexit, and the Rise of Authoritarian-Populism. New
York: Cambridge University Press.
Sides, John, Michael Tesler, and Lynn Vavreck. 2017. “The
2016 U.S. Election: How Trump Lost and Won.” Journal of
Democracy 28:34–44.
Sides, John, Michael Tesler, and Lynn Vavreck. 2018. Identity
Crisis: The 2016 Presidential Campaign and the Battle for
the Meaning of America. Princeton: Princeton University
Press.
Stiglitz, Joseph E. 2018. “Trump and Globalization.” Journal of
Policy Modeling 40:515–28.
Swedberg, Richard. 2018. “Folk Economics and Its Role in Trump’s
Presidential Campaign: An Exploratory Study.” Theory and
Society 47:1–36.
Whitehead, Andrew L., Samuel L. Perry, and Joseph O. Baker.
2018. “Make America Christian Again: Christian Nationalism
and Voting for Donald Trump in the 2016 Presidential
Election.” Sociology of Religion 79:147–71.
Author Biographies
Stephen L. Morgan is a Bloomberg Distinguished Professor in the
Krieger School of Arts and Sciences and in the School of Education
at Johns Hopkins University. He is a co–principal investigator of
the General Social Survey, and his current areas of scholarly
research include stratification, public opinion, causal inference, and
survey methodology.
Jiwon Lee is a PhD candidate in Department of Sociology at
Johns Hopkins University. His current areas of scholarly interest
include social stratification, public opinion, and quantitative
methodology.