American Government
Iolpoboqo
13.pdf
Tyler T. Reny is a doctoral candidate in the Political Science Department, University of California, Los Angeles, CA, USA. Loren Collingwood is an assistant professor in the Department of Political Science at the University of California, Riverside, CA, USA. Ali A. Valenzuela is an assistant professor in the Department of Politics at Princeton University, Princeton, NJ, USA. The authors would like to thank Matt Barreto, Bryan Wilcox-Archuleta, and discussants and participants at the Western Political Science Association, the UCLA Mini-Conference on Race and Ethnic Politics, the Midwest Political Science Association and American Political Science Association annual conferences, and anonymous reviewers for their helpful feedback. Any errors are the authors’ alone. *Address correspondence to Tyler T. Reny, University of California, Los Angeles, Political Science Department, 4289 Bunche Hall, Los Angeles, CA 90095, USA; email: [email protected].
© The Author(s) 2019. Published by Oxford University Press on behalf of the American Association for Public Opinion Research. All rights reserved. For permissions, please e-mail: [email protected]
VOTE SWITCHING IN THE 2016 ELECTION HOW RACIAL AND IMMIGRATION ATTITUDES, NOT ECONOMICS, EXPLAIN SHIFTS IN WHITE VOTING
TYLER T. RENY* LOREN COLLINGWOOD ALI A. VALENZUELA
Abstract In the aftermath of Donald Trump’s 2016 electoral-college victory, journalists focused heavily on the White working class (WWC) and the relationship between economic anxiety, racial attitudes, immi- gration attitudes, and support for Trump. One hypothesized but untested proposition for Donald Trump’s success is that his unorthodox candi- dacy, particularly his rhetoric surrounding economic marginalization and immigration, shifted WWC voters who did not vote Republican in 2012 into his coalition. Using a large national survey, we examine: (1) whether racial and immigration attitudes or economic dislocation and marginality were the main correlates of vote switching; and (2) whether this phenomenon was isolated among the White working class. Findings indicate that a nontrivial number of White voters switched their votes in the 2016 election to Trump or Clinton, that this vote switching was more associated with racial and immigration attitudes than economic factors, and that the phenomenon occurred among both working-class and nonworking-class Whites, though many more working-class Whites switched than did nonworking-class Whites. Our findings suggest that racial and immigration attitudes may be continuing to sort White voters into new partisan camps and further polarize the parties.
Public Opinion Quarterly, Vol. 83, No. 1, Spring 2019, pp. 91–113
doi:10.1093/poq/nfz011
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
The realignment of party coalitions around issues of race and civil rights stands as one of the most consequential political developments of the twen- tieth century. By the 1990s, political elites were well sorted into racially lib- eral and racially conservative camps, and most politically informed voters had followed suit (Carmines and Stimson 1989; Schickler 2016; Kuziemko and Washington 2018), suggesting that there may be little further room for racial attitudes to influence White Americans’ partisan attachments.
More recently, however, the two-term presidency of the nation’s first Black president, partisan polarization over immigration policy, and visible and rapid Latino population growth may be further transforming mass partisan- ship. Existing partisan coalitions that have characterized the two parties in American politics for several decades are shifting and may be contributing to the further partisan realignment of White citizens. As Republicans have pushed right on race, Democrats are increasingly relying on minority voters to win elections, strengthening the link between the Democratic Party and racial and ethnic minorities (Frymer 2010; Abrajano and Hajnal 2015; Tesler 2016).
This paper tests whether Donald Trump and Hillary Clinton’s unique can- didacies may have facilitated vote switching in the 2016 election, a precursor to durable partisan change. First, did a sizable number of White voters switch their vote in 2016 and was this vote switching unique to the White working class? Second, are immigration and racial attitudes or economic dislocation and marginality more strongly associated with this vote switching?
We find evidence that a nontrivial number of both working-class and non- working-class White voters did switch their votes in the 2016 election and that this vote switching was associated more with racial and immigration attitudes than economic factors.
This paper contributes to a growing literature on White racial attitudes and White responses to demographic change and perceived immigrant threat in American politics. While others have shown that contextual demo- graphic threat contributed to Trump support during the 2016 primary election (Newman, Shah, and Collingwood 2018), and that racial and immigration at- titudes were associated with Trump support among voters (Mutz 2018; Sides, Tesler, and Vavreck 2018), this paper is the first to thoroughly examine the correlates of vote switching in the 2016 election. Our findings suggest that the United States may be in the midst of further electoral realignment as partisan voting continues to polarize around issues of race and immigration.
Racial Realignment: Conceptions of Partisanship and Partisan Change
Partisanship is one of the most widely studied phenomena in political science. Traditional views of partisan identification focus on the issues, ideologies, and groups connected to each party, what Huddy, Mason, and Aarøe (2015)
Reny, Collingwood, and Valenzuela92
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
and Huddy, Bankert, and Davies (2018) call the instrumental conception of partisanship (Berelson, Lazarsfeld, and McPhee 1954; Campbell et al. 1960; Abramowitz and Saunders 2006). More recently, scholars have conceived of partisanship as a social identity, comparable to race or religion (Green, Palmquist, and Schickler 2002; Huddy, Mason, and Aarøe 2015; Huddy and Bankert 2017; Mason and Wronski 2018; Rothschild et al. 2018). In this view, partisan affiliation is akin to a salient social group attachment (Tajfel 1981) and partisanship the result of “comparing a judgment about oneself with one’s conception of a social group. As people reflect on whether they are Democrats or Republicans (or neither), they call to mind a mental image, or stereotype, or what these sorts of people are like and square those images with their own conception” (Green, Palmquist, and Schickler 2002, 8).
While these two conceptions are often pitted against each another as mutually exclusive, they need not be. Studies of voter conceptions of partisanship find evidence for both (Rothschild et al. 2018), and both allow for partisan change, albeit via different mechanisms. Instrumental partisan change can occur if par- ties shift positions on issues that may be important to certain voters. Changes in partisan identity can occur if the primary social groups that make up parties change (i.e., which social groups “go with” each party). We argue that three parallel trends have opened the door for vote switching in 2016 among partisans via both channels: the election of Barack Obama, mass immigration from Latin American countries, and the slow collapse of American manufacturing.
While these three trends affect all voters, there is reason to believe that the White working class—increasingly alienated from the two-party system, threatened by demographic change, and unsure of its future economic pro- spects—was uniquely positioned to be cross-mobilized in the 2016 election. Indeed, politicians have not courted the White working class for some time. Democratic base-building strategies have focused on fostering a growing Latino and Asian American electorate (Barreto, Collingwood, and Manzano 2010; Wong et al. 2011; Abrajano and Hajnal 2015), rather than courting working-class Whites. The contemporary Republican Party similarly has struggled with White working-class mobilization. Though it has not been shy about using dog-whistle racial appeals to try to appeal to racially conserva- tive Whites (López 2015), it may be too strongly associated with the wealthy elite (Ahler and Sood 2018) for working-class Whites to feel like they belong (Green, Palmquist, and Schickler 2002). As a result, the White working class has felt increasingly alienated from both parties, neither of which look like their group or are perceived as representing their group’s interests (Gest 2016).
America’s First Black President
The election of a Black man to the White House served as a highly visible and symbolic political shock (Parker and Barreto 2013), shattering the long era
Vote Switching in the 2016 Election 93
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
of racial silence ushered in by Bill Clinton’s presidency (Tesler 2016). The Democratic Party no longer was associated just with civil rights and Black voters but had succeeded in electing an African American to the most powerful and visible position in the world. As a result of President Obama’s election, racial attitudes began to spill over into Americans’ evaluations of numerous political phenomena, including economic trends, public policies, and public figures (Tesler 2016; Yadon and Piston 2018; Enders and Scott 2018).
This increased racialization of American politics spilled over into partisan- ship as well (Tesler 2016), as low-information Whites with lower levels of attitude constraint (Converse 1964) increasingly linked their racial attitudes with their partisan identities, producing a racially polarized shift of White ra- cial liberals toward the Democrats and White racial conservatives toward the Republicans.
Changing Demographics and Immigration Attitudes
The racial symbolism of electing the first non-White president has been cou- pled with rapid demographic change. It is likely that attitudes toward other non-White outgroups, like Latino immigrants, also spill over into White par- tisanship. “Latino threat” has been operationalized as both contextual and symbolic, with anti-immigrant attitudes being triggered by local demographic shifts (Newman 2013; Enos 2014) as well as national rhetoric and trends (Abrajano and Hajnal 2015). As a result, attitudes toward a variety of policy issues like welfare, health, and education now are associated with immigration attitudes and Latino affect (Fox 2004; Abrajano and Hajnal 2015).
More importantly, Latino affect and fear of demographic change has also been linked to individual-level ideology and partisanship (Valentino, Brader, and Jardina 2013; Craig and Richeson 2014; Abrajano and Hajnal 2015). Lab experiments have shown that exposure to news about shifting demographics moves White Americans in an ideologically conservative direction and to- ward the Republican Party (Craig and Richeson 2014), a shift also seen in observational data (Abrajano and Hajnal 2015). Ostfeld (2018) finds that when White voters learn about Democratic outreach to Latinos, they become less supportive of the Democratic Party. Indeed, Abrajano and Hajnal (2015) show that in the near term, Latino population growth will likely result in many White Americans shifting into the Republican Party as partisan elites continue to polarize on issues of immigration and race.
Partisan Groups, Issues, and Vote Switching
How do these visible changes translate into vote switching and partisan change? According to instrumental views of partisan change, the increased political
Reny, Collingwood, and Valenzuela94
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
attention to racialized issues (policing, immigration) during Obama’s tenure and the increased reliance on non-White voters is shifting the Democratic Party’s median position on issues away from the median White citizen’s pos- ition, resulting in White shifts toward the Republican Party as White voters update their partisanship to match their policy positions. According to iden- tity-based conceptions of partisan change, the increased perception of the Democratic Party as a coalition of non-White voters is changing perceptions of where many Whites feel they belong.
There is evidence that both processes are occurring, with perceptions of policy shifts following logically from perceptions of a diversifying Democratic Party. There is little doubt that Obama’s election increased the visibility of Black voters as a core Democratic constituency (Tesler 2016). For instance, Americans consistently overestimate the proportion of Democrats that are Black (41.9 percent compared to the true composition of 23.9 percent) (Ahler and Sood 2018). In addition, White voters, Republicans in particular, are very likely to implicitly associate the Democratic Party with African Americans and the Republican Party with Whites (Zhirkov and Valentino 2017).
The Democratic Party is also increasingly associated with Latinos (Abrajano and Hajnal 2015). The Democratic Party, particularly Democratic presidential candidates, frequently and openly courts Latino voters (Collingwood, Barreto, and Garcia-Rios 2014), which has been shown to turn off many White voters (Ostfeld 2018), and the majority of Latino elected officials are Democratic.1
This shift in real and perceived composition of parties is no doubt inter- twined with perceptions of the ideological orientation and issue priorities of the Democratic Party. As the party has diversified, White Americans have in- creasingly perceived the Democratic Party as being further from their own positions on issues (Zingher 2018), and increasingly aligned with issue pri- orities of African Americans (Tesler 2016) and immigrants (Abrajano and Hajnal 2015; Ostfeld 2018).
Regardless of the conceptualization of partisanship and partisan change, evi- dence suggests that White voters are increasingly perceiving the Democratic Party as the party of racial and ethnic minorities and racially liberal policy and the Republican Party as the party of White Americans and racially conservative policy. Together, these trends could lead to vote switching and eventual stable shifts in White partisanship. White voters who are racially conservative, who have more punitive immigration attitudes, or who live in communities under- going rapid demographic change may be particularly put off by the Democratic Party’s increasing diversity and shifting issue priorities and drawn to Trump for his clear and consistent anti-immigrant policy positions and rhetoric appealing specifically to White voters. At the same time, Donald Trump’s immigration
1. According to the National Association of Latino Elected and Appointed Officials, among the partisan offices held by Latinos in 2014, 88 percent were Democrats and Latino voters are increas- ingly voting Democratic (Lopez et al. 2016).
Vote Switching in the 2016 Election 95
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
policy proposals and rhetoric may have driven more traditional, business-ori- ented, and racially moderate White voters who are comfortable with diversity away from the Republican presidential candidate and toward Clinton, who em- braced a more accommodating position on racial and immigration issues.
We therefore put forth our first group of hypotheses related to vote switching:
H1a: Racial attitudes: White voters who express more conservative ra- cial attitudes will be more likely to switch their vote to Trump than similarly situated White voters with more liberal racial attitudes.
H1b: Anti-immigrant attitudes: White voters who express more punitive views on immigration will be more likely to switch their vote to Trump than similarly situated White voters with less punitive views on immigration.
H1c: Latino immigrant threat: White voters living in counties under- going rapid Latino growth will be more likely to switch their vote to Trump relative to similarly situated White voters who live in counties with lower levels of Latino growth.
H2a: Racial accommodation: White voters who express more liberal racial attitudes will be more likely to switch their vote to Clinton than simi- larly situated White voters with more conservative racial attitudes.
H2b: Pro-immigrant attitudes: White voters who express less punitive views on immigration will be more likely to switch to Clinton than similarly situated White voters with more punitive views on immigration.
H3: The relationship between racial attitudes, immigration attitudes, Latino threat, and vote switching to Trump will be stronger among working-class than nonworking-class Whites. The relationship be- tween racial attitudes, immigration attitudes, Latino threat, and vote switching to Clinton will be stronger among nonworking-class than working-class Whites.
Economic Marginality and Local Economic Dislocation
We have argued that White voters are a prime target for Trump’s racially conserv- ative rhetoric, particularly after Obama’s presidency and in an era of increased immigration. Recent economic changes and dislocation in an era of globaliza- tion and worker disaffection may have also driven White voters, particularly the White working class, to support the populist appeals of Donald Trump, whose rhetoric often dovetailed anti-immigrant with anti-globalization and anti–free trade themes. Indeed, the media were quick to declare economic dislocation as a key driver of White voting for Trump (Adams 2016; Sargent 2017).
The White working class has been hit particularly hard by structural economic changes (Gest 2016). Today there are three times as many white-collar workers as manual workers, and wages are stagnant for those without a college educa- tion (Teixeira and Abramowitz 2008). In this sense, manufacturing decline may
Reny, Collingwood, and Valenzuela96
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
be disproportionately felt among the White working class (Meyerson 2015). In addition, the upward mobility and union protections that defined the working class’s support for Democrats throughout the middle of the twentieth century is no longer a reality. The post-recession job recovery during President Obama’s tenure benefited almost exclusively college-educated workers, leaving out many middle-income earners (Carnevale, Jayasundera, and Gulish 2015). These eco- nomic dislocations have been compounded by the fraying of the community- based institutions that used to provide safety nets in times of need (Putnam 2001).
Moreover, a broad body of work in political science argues that eco- nomic conditions play an outsize role in determining the outcomes of elec- tions (Lewis-Beck and Stegmaier 2000). Political scientists regularly forecast elections using macroeconomic metrics such as second-quarter GDP growth (Abramowitz 2016) and change in unemployment (Jerome and Jerome- Speziari 2016). This body of work suggests that voters who switch from one party to another may do so for retrospective economic reasons—their personal and local economic conditions have deteriorated under the leadership of the party from which they switched (Fiorina 1981).
Thus, despite the large body of work showing that racial and immigration atti- tudes play a central role in recent voting trends, we cannot discount the possibility that White individuals who switched votes in 2016, particularly White working-class voters, did so because they were economically marginalized and, consistent with theories of retrospective voting, did not see Hillary Clinton’s Democratic Party as one that would address their economic concerns after eight years of Democratic control of the White House. Conversely, individuals who had not supported a Democratic president in the previous election but who have seen economic im- provements under a Democratic president, or who live in a thriving local economy, may have been drawn to switch allegiances to Clinton in the 2016 election.
H4a: Economic marginality: White citizens who are economically mar- ginal—whose perceived economic well-being has deteriorated or who are experiencing relative economic deprivation—will be more likely to switch their vote to Trump than similarly situated voters who are not economically marginalized.
H4b: Local economic dislocation: White citizens living in counties under- going economic decline—growth in unemployment or loss in manu- facturing—will be more likely to switch their votes to Trump, relative to similarly situated voters who do not live in such counties.
H5a: Economic integration: White citizens who are economically inte- grated—whose perceived economic well-being has improved or who are not experiencing relative economic deprivation—will be more likely to switch their vote to Clinton than similarly situated voters who are not so economically integrated.
H5b: Local economic expansion: White citizens living in counties under- going economic growth—declines in unemployment or increases in
Vote Switching in the 2016 Election 97
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
manufacturing—will be more likely to switch their votes to Clinton, relative to similarly situated voters who do not live in such counties.
H6: The relationship between economic indicators and vote switching for Trump will be stronger among working-class than nonworking- class Whites. The relationship between economic indicators and vote switching for Clinton will be stronger among nonworking- class than working-class Whites.
Methods
We use a large opt-in Internet panel survey, the 2016 Cooperative Congressional Election Studies (CCES) Survey, to test our hypotheses (Ansolabahere and Schaffner 2017). The CCES is administered by YouGov/Polimetrix and has an interview period of September to November. The CCES sample selection follows a two-stage sample-matching process. First, YouGov draws a stratified random sample from the 2012 American Community Survey (ACS) respond- ents. This sample is then matched to members of the YouGov/Polimetrix opt- in panel, such that the resulting panel looks the same on observables as the national population.2 The resulting survey includes 64,600 completed inter- views with a within-panel participation rate of 41.9 percent and an AAPOR re- sponse rate 1 of 13.9 percent. The final sample is weighted to be representative of the US adult population. Finally, the 2016 vote has been validated using the Catalist database of registered voters in the United States.3
For Trump-switching models, we restrict the data to only examine White 2016 voters who voted in 2012 for either the Democratic candidate, Barack Obama, or a third-party candidate, because these are the only voters who are eligible to switch (n = 19,296). For Clinton-switching models, we restrict the sample to White 2016 voters who voted in 2012 for either the Republican can- didate (Romney) or a third-party candidate (n = 17,493). Split-sample mod- els of the White working class further restrict our sample sizes to n = 10,341 for Trump models and n = 11,299 for Clinton models.4 We present results
2. While online nonprobability samples typically include more politically and civically engaged individuals, a Pew Research Center study finds that YouGov surveys show the smallest devia- tions from benchmarks compared to other well-known online opt-in survey panel competitors (Kennedy et al. 2016; Rivers 2016), producing a national sample that is deemed largely repre- sentative and accurate. 3. See https://cces.gov.harvard.edu/ for full details about the survey methodology, including full question wordings, sampling frame, sampling design, response rates, and voter list matching. 4. We define white working class as those without a four-year college degree. There are numerous ways to define working class. Educational levels, which we use for our models, serve as a proxy for skill and human capital, which is increasingly essential in our changing economy (Carnevale, Jayasundera, and Gulish 2015). Of course, those with college degrees can hold blue-collar jobs and those without college degrees can be (and frequently are) very successful financially.
Reny, Collingwood, and Valenzuela98
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
for working-class Whites, nonworking-class Whites, and all Whites in each analysis.
Our dependent variables are voting for Trump (1 = yes, 0 = no) and vot- ing for Clinton (1 = yes, 0 = no). Because of the model sample restriction, a Trump vote switcher can be defined as a White 2016 Trump voter who voted in 2012 for Barack Obama (the Democrat) or a third-party candidate. A Clinton vote switcher is a White 2016 Clinton voter who voted in 2012 for Mitt Romney (the Republican) or a third-party candidate.5 We outline most of the possible vote combinations for 2012 and 2016 voters in table 1 and display the proportion of nonworking-class and working-class Whites who fall into each strata. Not surprisingly, the vast majority of voters are congruent voters (Romney to Trump and Obama to Clinton). Among vote switchers, the focus of this paper, about 6 percent of White working-class and 2.4 percent of White nonworking-class voters switched to Trump and 2 percent of White working-class and 3.1 percent of White nonworking-class voters switched to Clinton. Given the over 50.5 million college-educated
Table 1. Vote switching combinations
2012 Vote 2016 Vote Non-WC Whites WWC
Congruent voting Romney Trump 35.2% 50.4% Obama Clinton 48.4% 31.5% Other Other 1.3% 1.0% Partisan vote switching Romney Clinton 3.1% 2.0% Other Clinton 1.1% 0.3% Obama Trump 2.4% 6.2% Other Trump 0.7% 1.4% Total N 9,129 13,842
Note.—Partisan vote-switching combinations and weighted percentage of all nonworking- class White and working-class White adult voters who voted in 2012 and 2016. Congruent voting figures are shown for comparison. Note that the columns do not sum to 100 percent because sev- eral vote combinations were omitted from the table, including demobilization (Romney, Obama, or Other in 2012 to not voting in 2016), third-party switching (Romney, Obama, or Other in 2012 to third party in 2016), and mobilization (not voting in 2012 to voting for Trump, Clinton, or Other in 2016).
Nevertheless, using income to determine working class can be arbitrary, depending on region and cut-points used, and is often poorly reported on surveys (Teixeira and Abramowitz 2008). We thus settle on a definition of working class as lacking a four-year college degree. We estimated similar models defining working class as those in the lower tercile of the income distribution and find very similar results, which are presented in Online Appendix A. 5. Given concerns of bias related to recalling a vote cast in 2012—due to poor memory or simply social desirability and lying—we undertake a number of additional analyses in Online Appendix B. They assess the extent that misreport could bias the results of these analyses. In line with Rivers and Lauderdale (2016), we conclude that very few respondents lie about which candidate they supported in the previous election, reducing concerns about significant bias in the measure.
Vote Switching in the 2016 Election 99
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
White voters and over 46.4 million working-class White voters in 2016 (as estimated by CNN 2016 exit polls), these percentages are not trivial and suggest that, in raw numbers, many more working-class Whites than non- working-class Whites switched their votes in 2016 from Obama to Trump and far fewer from Romney to Clinton.
Nonetheless, these descriptive statistics do not say anything about which factors were most strongly related to switching to either Trump or Clinton and whether those factors varied by partisan affiliation. To answer these questions, we use logistic regression to model vote switching as a function of racial, im- migration, and economic attitudes and contexts. Rather than pool across par- tisans, we conduct our analyses separately among voters who identify with the two major parties or as Independents.6
For racial and immigration attitudes, we relied on two batteries of ques- tions. We combine three questions about acknowledgment of race and racism into a scale of racial attitudes (α = 0.68; average r = 0.42) and recode it to range between 0 (racially liberal) and 1 (racially conservative). For individ- ual-level immigration attitudes, respondents chose which of four immigration policy proposals they supported. The four questions were combined into a single immigration attitude scale (α = 0.69, average r = 0.35) and recoded to fall between 0 (least punitive) and 1 (most punitive).7 Finally, to measure demographic change, we calculated Latino growth as the percentage change in the county Latino population from 2000 to 2014.
We measure economic marginality and local economic dislocation each in two ways. Economic marginality is operationalized as family-level retrospective economic evaluation and relative economic deprivation. Retrospective economic evaluations were measured with a question about whether over the previous four years the respondent’s household annual in- come increased or decreased. The responses were recoded to fall between (0) for increased a lot and (1) for decreased a lot. Relative deprivation is a combination of the respondent’s self-reported family income and their sur- rounding economic environment. We code the respondent as economically marginal if their family income is lower (1) or higher (0) than the median income in their county of residence.
Economic dislocation is operationalized as change in county-level manu- facturing and change in county-level unemployment. Manufacturing loss is calculated as the percentage change in county manufacturing employment
6. Pooled models, presented in Online Appendix C, return substantively similar results. 7. Full question wording, distributions for key covariates, and scale statistics can be found in Online Appendix D. Readers might be concerned that horse race models pitting the regression coefficient of single items, which are more prone to measurement error, against scales, which are less prone to measurement error (Ansolabehere, Rodden, and Snyder 2008), is setting us up for an unfair comparison. Online Appendix E presents additional models where the scales have been disaggregated, with no differences emerging.
Reny, Collingwood, and Valenzuela100
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
T ab
le 2
. T
ru m
p v
ot e
sh if
t
P re
di ct
or s
of s
hi ft
in g
to T
ru m
p in
2 01
6
D
ep en
de nt
v ar
ia bl
e:
T ru
m p
sw it
ch D
em (
al l W
hi te
s) In
d (a
ll W
hi te
s) G
O P
a ll
W hi
te s)
D em
( W
W C
) In
d (W
W C
) G
O P
( W
W C
) D
em (
no n-
W W
C )
In d
(n on
-W W
C )
G O
P (
no n-
W W
C )
(1 )
(2 )
(3 )
(4 )
(5 )
(6 )
(7 )
(8 )
(9 )
R ac
ia l
at ti
tu de
s 3.
23 9*
** (
0. 33
8) 2.
55 9*
** (
0. 23
8) 1.
54 8*
** (
0. 39
8) 2.
89 5*
** (
0. 37
9) 2.
21 6*
** (
0. 27
9) 1.
54 4*
* (0
.4 80
) 4.
59 5*
** (
0. 75
6) 3.
39 5*
** (
0. 45
9) 1.
54 8*
( 0.
73 6)
Im m
ig ra
ti on
a tt
it ud
es 2.
02 4*
** (
0. 21
1) 1.
95 2*
** (
0. 16
1) 1.
15 4*
** (
0. 24
4) 2.
01 7*
** (
0. 23
6) 1.
80 2*
** (
0. 18
9) 1.
06 4*
** (
0. 28
7) 1.
95 1*
** (
0. 47
6) 2.
33 7*
** (
0. 30
8) 1.
47 5*
* (0
.4 77
) P
ct . L
at in
o gr
ow th
( 00
–1 4)
0. 00
3* (
0. 00
1) 0.
00 1
(0 .0
01 )
0. 00
1 (0
.0 01
) 0.
00 3*
( 0.
00 1)
0. 00
04 (
0. 00
1) 0.
00 01
( 0.
00 2)
–0 .0
01 (
0. 00
5) 0.
00 1
(0 .0
02 )
0. 00
3 (0
.0 03
) F
am il
y ec
on s
it ua
ti on
w or
se 2.
01 5*
** (
0. 27
9) 0.
53 2*
* (0
.2 00
) 0.
82 2*
* (0
.3 08
) 2.
08 8*
** (
0. 31
6) 0.
70 0*
* (0
.2 34
) 0.
78 8*
( 0.
36 3)
1. 51
0* (
0. 61
0) 0.
08 1
(0 .3
89 )
0. 72
6 (0
.6 04
) R
el at
iv e
de pr
iv at
io n
–0 .1
76 (
0. 21
7) –0
.3 02
( 0.
16 3)
–0 .2
72 (
0. 25
9) –0
.1 20
( 0.
24 5)
–0 .3
47 (
0. 19
0) –0
.3 12
( 0.
30 1)
–0 .3
09 (
0. 48
1) –0
.1 74
( 0.
32 3)
–0 .1
21 (
0. 52
5) P
ct . m
an uf
ac tu
ri ng
l os
s (0
0– 14
) –0
.0 03
( 0.
00 6)
0. 00
5 (0
.0 04
) 0.
00 7
(0 .0
06 )
–0 .0
10 (
0. 00
6) 0.
00 5
(0 .0
05 )
0. 00
8 (0
.0 07
) 0.
02 3*
* (0
.0 08
) 0.
00 4
(0 .0
08 )
0. 00
2 (0
.0 13
) P
ct . u
ne m
pl oy
m en
t di
ff
(0 0–
14 )
–0 .0
03 (
0. 00
1) 0.
00 2*
( 0.
00 1)
0. 00
4* (
0. 00
2) –0
.0 04
* (0
.0 02
) 0.
00 2
(0 .0
01 )
0. 00
3 (0
.0 02
) 0.
00 4
(0 .0
03 )
0. 00
1 (0
.0 02
) 0.
00 6
(0 .0
04 )
C on
tr ol
s? √
√ √
√ √
√ √
√ √
O bs
er va
ti on
s 9,
38 9
5, 35
7 91
5 4,
88 7
2, 93
6 59
9 4,
50 2
2, 42
1 31
6 L
og l
ik el
ih oo
d –9
66 .9
69 –1
,4 87
.1 20
–5 32
.0 05
–7 46
.2 79
–1 ,0
57 .2
67 –3
81 .1
49 –2
09 .0
87 –4
20 .5
17 –1
47 .8
58 A
ka ik
e in
f. c
ri t.
1, 96
5. 93
8 3,
00 6.
24 0
1, 09
6. 01
1 1,
52 2.
55 8
2, 14
4. 53
4 79
2. 29
7 44
8. 17
4 87
1. 03
4 32
5. 71
6
N o
te .—
U ns
ta nd
ar di
ze d
lo gi
st ic
r eg
re ss
io n
co ef
fi ci
en ts
. S ta
nd ar
d er
ro rs
i n
pa re
nt he
se s.
C on
tr ol
v ar
ia bl
es a
re o
m it
te d
fr om
t ab
le f
or p
re se
nt at
io n.
F ul
l re
gr es
si on
t ab
le s
av ai
la bl
e in
O nl
in e
A pp
en di
x H
.
* p <
0 .0
5; *
* p <
0 .0
1; *
** p <
0 .0
01 (
tw o-
ta il
ed )
Vote Switching in the 2016 Election 101
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
T ab
le 3
. C
li n
to n
v ot
e sh
if t
P re
di ct
or s
of s
hi ft
in g
to C
li nt
on i
n 20
16
D
ep en
de nt
v ar
ia bl
e:
C li
nt on
s w
it ch
D em
( al
l W hi
te s)
In d
(a ll
W hi
te s)
G O
P (
al l W
hi te
s) D
em (
W W
C )
In d
(W W
C )
G O
P (
W W
C )
D em
( no
n- W
W C
) In
d (n
on -W
W C
) G
O P
( no
n- W
W C
)
(1 )
(2 )
(3 )
(4 )
(5 )
(6 )
(7 )
(8 )
(9 )
R ac
ia l
at ti
tu de
s –1
.2 53
* (0
.5 98
) 4.
20 8
** *
(0 .4
20 )
–4 .1
53 **
* (0
.5 46
) –0
.7 58
( 0.
67 1)
3. 79
5* **
( 0.
59 4)
4. 28
8* **
( 0.
77 3)
–3 .2
62 *
(1 .5
04 )
–4 .8
46 **
* (0
.6 11
) –3
.9 48
** *
(0 .7
85 )
Im m
ig ra
ti on
a tt
it ud
es –1
.2 40
** *
(0 .3
57 )
1. 49
7* **
( 0.
24 2)
–2 .0
80 **
* (0
.2 96
) 1.
29 9*
* (0
.4 08
) 1.
31 6*
** (
0. 33
7) 1.
99 3*
** (
0. 41
7) –0
.8 38
( 0.
85 8)
–1 .7
19 **
* (0
.3 56
) –2
.2 07
** *
(0 .4
27 )
P ct
. L at
in o
gr ow
th
(0 0-
14 )
0. 00
2 (0
.0 02
) –0
.0 04
* (0
.0 01
) –0
.0 01
( 0.
00 2)
0. 00
1 (0
.0 02
) –0
.0 04
( 0.
00 2)
–0 .0
03 (
0. 00
3) 0.
00 9
(0 .0
07 )
–0 .0
04 *
(0 .0
02 )
0. 00
1 (0
.0 03
) F
am il
y ec
on s
it ua
ti on
w
or se
–0 .4
95 (
0. 48
5) –0
.7 90
** (
0. 28
2) –1
.4 00
** *
(0 .3
85 )
–0 .1
96 (
0. 57
0) –0
.4 94
( 0.
42 0)
–1 .0
57 (
0. 54
2) –1
.3 79
( 1.
04 8)
–1 .0
36 **
( 0.
38 4)
–1 .8
38 **
* (0
.5 56
) R
el at
iv e
de pr
iv at
io n
–0 .0
03 (
0. 34
7) 0.
07 7
(0 .2
22 )
0. 01
5 (0
.2 90
) 0.
11 7
(0 .4
14 )
0. 19
3 (0
.3 30
) 0.
44 2
(0 .4
01 )
–0 .5
75 (
0. 73
0) –0
.0 38
( 0.
30 9)
–0 .4
54 (
0. 44
6) P
ct . m
an uf
ac tu
ri ng
lo
ss (
00 -1
4) –0
.0 05
( 0.
00 9)
–0 .0
01 (
0. 00
5) 0.
00 5
(0 .0
07 )
–0 .0
06 (
0. 01
0) –0
.0 05
( 0.
00 8)
0. 00
9 (0
.0 09
) 0.
00 5
(0 .0
19 )
0. 00
2 (0
.0 07
) –0
.0 03
( 0.
01 1)
P ct
. u ne
m pl
oy m
en t
di ff
( 00
-1 4)
0. 00
3 (0
.0 02
) 0.
00 3
(0 .0
02 )
0. 00
5* *
(0 .0
02 )
0. 00
5 (0
.0 03
) 0.
00 2
(0 .0
02 )
0. 00
6* (
0. 00
3) –0
.0 05
( 0.
00 6)
0. 00
4 (0
.0 02
) 0.
00 4
(0 .0
03 )
C on
tr ol
s? √
√ √
√ √
√ √
√ √
O bs
er va
ti on
s 58
4 5,
52 6
7, 92
5 43
5 3,
42 6
5, 23
8 14
9 2,
10 0
2, 68
7 L
og l
ik el
ih oo
d –2
87 .7
15 –8
32 .6
46 –5
54 .3
41 –2
07 .0
86 –4
13 .0
15 –2
98 .9
33 –7
3. 03
0 –4
08 .0
75 –2
51 .3
12 A
ka ik
e in
f. c
ri t.
60 7.
43 0
1, 69
7. 29
2 1,
14 0.
68 2
44 4.
17 2
85 6.
03 1
62 7.
86 5
17 6.
06 1
84 6.
15 1
53 2.
62 5
N o
te .—
U ns
ta nd
ar di
ze d
lo gi
st ic
r eg
re ss
io n
co ef
fi ci
en ts
. S ta
nd ar
d er
ro rs
i n
pa re
nt he
se s.
C on
tr ol
v ar
ia bl
es a
re o
m it
te d
fr om
t ab
le f
or p
re se
nt at
io n.
F ul
l re
gr es
si on
t ab
le s
av ai
la bl
e in
O nl
in e
A pp
en di
x H
. *
p <
0 .0
5; *
* p <
0 .0
1; *
** p <
0 .0
01 (
tw o-
ta il
ed )
Reny, Collingwood, and Valenzuela102
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
between 2000 and 2014 and change in unemployment as the percentage change in unemployment rate at the county level between 2000 and 2014.8
Beyond these key independent variables, our analyses included several con- trol variables that may be related to vote switching, including change in county foreign-born population, personal income, employment status, self-reported ideology, union membership, gender, geographic region, and in pooled all- White-respondent models, education.9
Results
Tables 2 and 3 present results from the main logistic regression model that include racial and immigration attitudes and contexts (H1, H2) and eco- nomic factors (H4, H5) for all Whites (columns 1 through 3), the White working class (columns 4 through 6), and nonworking-class Whites (col- umns 7 through 9; H3 and H6). Control variables are omitted from the table for space concerns, but full regression tables are presented in Online Appendix H. Because logistic regression coefficients are difficult to inter- pret, we simulate counterfactuals and plot the results for each variable of interest.10
We begin by looking at the role of racial and immigration factors on vote switching for Trump. Figure 1 displays the effect of moving racial attitudes (min to max), immigration attitudes (min to max), and county-level Latino population (mean ± 2 s.d.) on the probability of vote switching for all White (circles), White working-class (triangles), and White nonworking-class (squares) Democrats, Independents, and Republicans.11
First, the associations between each variable and switching for Trump for working-class and nonworking-class Whites are generally not statistically dis- tinguishable, with the exception of immigration attitudes among Democrats
8. Online Appendix F presents results using the same contextual economic measures but change as measured between shorter time spans. These results are robust to these alternative specifications. 9. The full model is: Vote ~ β0 + β1 RacialAttitudes + β2 ImmigrationAttitudes + β3 HispanicGrowth (00–14) + β4 RetrospectiveEconomics + β5 RelativeDeprivation + β6 ManufacturingLoss (00–14) + β7 CountyUnemploymentChange (00–14) + β8 Income + β9 Unemployed + β10 ForeignBornChange(00–14) + β11 Union + β12 Female + β13 Ideology + β14 South + β15 College. For each variable, DK and “Refuse” responses are recoded as missing, with the exception of ideol- ogy where DK respondents were recoded as moderates (Treier and Hillygus 2009). We have also run the core analyses using imputed values via the MICE package in R. The results, which are substantively identical, are reported in Online Appendix G. 10. Model fit statistics are presented in Online Appendix I. 11. All point estimates in figures are estimated from coefficients in tables 2 and 3. Respondents are split by party because baseline propensity to switch should vary by partisanship. For instance, it will be easier for a self-identified Republican who voted for Obama in 2012 to “come home” to his or her party in 2016 than it will be to get a Democrat who voted for Obama in 2012 to vote for Trump in 2016.
Vote Switching in the 2016 Election 103
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
and Independents. While working-class Whites were more likely to switch their vote to Trump in 2016 than nonworking-class Whites, both working-class and nonworking-class Whites with strong racially conservative or punitive im- migration views were more likely to switch than those with racially liberal or pro-immigration views. These relationships are similar across subgroups for all models.
Second, the association between racial and immigration attitudes and switch- ing to Trump is stronger among Independents and Republicans than among Democrats. It is easier for Trump’s campaign to “bring home” Republicans or sway Independents than to persuade Democrats to vote across party lines. Nevertheless, moving White Democratic racial conservatism and punitive im- migration attitudes from their minimum to maximum values, holding all other variables at their means, is associated with a 12.6 (95 percent CI: [7.4,20.4]) and 3.7 (95 percent CI: [2.5,5.2]) percentage-point increase in the likelihood of switching to Trump in 2016, a relationship that only strengthens in the WWC sample.12
Third, the findings show little support that county-level demographic change is associated with vote switching. While the point estimates are posi- tive, they are substantively small and generally statistically indistinguishable from zero. If we simulate the probability of Trump vote switching for the full range of Latino population change (–100 percent to 1,409 percent), the point estimates increase substantially for working-class White Democrats (66 per- centage points), Independents (18 percentage points), and Republicans (6 percentage points). However, because we are extrapolating to extreme outli- ers, these estimates are highly imprecise. These results could be due to the fact that politics is increasingly becoming nationalized, fueled by declining local media (Prior 2007; Martin and McCrain 2018) and decreasing knowl- edge of and interest in local political events (Hopkins 2018). These find- ings echo the sociotropic literature on immigration attitudes that suggests that immigration attitudes are driven more by national than local concerns of the cultural and economic threat posed by immigrants (Hainmueller and Hopkins 2014).
Figure 2 displays the effect of moving racial attitudes (max to min) and immigration attitudes (max to min) on the probability of switching a vote to Clinton for all White (circles), White working-class (triangles), and White nonworking-class (squares) Democrats, Independents, and Republicans. Note that the direction of the counterfactual scenario is inverted to be consistent
12. Readers might be concerned that these relationships are endogenous and that respondents are simply learning and adopting the racial or immigration views of their candidate of choice. We are skeptical that this is the case, given that group antagonisms are generally crystallized attitudes (Tesler 2015). Nevertheless, in leveraging a panel dataset to examine how wave one (2011) racial and immigration attitudes are correlated with vote switching in 2016, we find similar trends, pre- sented in Online Appendix J, suggesting that racial and immigration attitudes preceded Trump’s rise.
Reny, Collingwood, and Valenzuela104
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
with hypotheses tested. In other words, these plots can be interpreted as the in- crease in the predicted probability of switching for Clinton given a shift from the most racially conservative to the most racially liberal and from the most punitive to the least punitive views on immigration.
Racial Attitudes Immigration Attitudes
Democrat Independent Republican Democrat Independent Republican
0.00
0.25
0.50
C h
a n
g e
P ro
b N
e w
C lin
to n
V o
te r
| P
a rt
y
Sample
All Whites
WWC
Non−WC Whites
Figure 2. Race, immigration, and switching to Clinton. Points indicate effect of moving each variable from its maximum to minimum value (except Latino growth, which was moved from 2 s.d. above its mean to 2 s.d. below) while holding all others at their means. Circles indicate model for all White respondents, triangles for just White working-class respondents, and squares for nonworking-class White respondents. Lines indicate simulated 95 percent confidence intervals.
Racial Attitudes Immigration Attitudes Latino Growth
Democrat Independent Republican Democrat Independent Republican Democrat Independent Republican
−0.5
0.0
0.5
1.0
C h a n g e P
ro b N
e w
T ru
m p V
o te
r | P
a rt
y
Sample
All Whites
WWC
Non−WC Whites
Figure 1. Race, immigration, and switching to Trump. Points indicate effect of moving each variable from its minimum to maximum value (except Latino growth, which was moved from 2 s.d. below to 2 s.d. above its mean so we aren’t extrapolating to extreme outliers) while holding all others at their means. Lines indicate simulated 95 percent confidence intervals.
Vote Switching in the 2016 Election 105
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
Similar trends emerge in the Clinton models as in the Trump models. The most racially liberal Democrats, Independents, and Republicans were more likely to switch to Clinton in the 2016 election than the most racially con- servative. This relationship is stronger among nonworking-class Whites than among working-class Whites. The same goes for Democrats, Independents, and Republicans who held the least punitive immigration views.
In sum, we find support for part of hypotheses H1 and H2. Symbolic racial and immigration attitudes were strongly associated with vote switching in the 2016 election. White voters who held punitive immigration or racially conser- vative views were more likely to switch to Trump in the 2016 election than those with pro-immigration or racially liberal views, who were more likely to switch to Clinton. This suggests that symbolic racial and immigration attitudes played an important role in shuffling some White voters in the 2016 election. We did not uncover strong evidence for the hypothesis that living in counties with the most rapidly changing Latino population was associated with vote switching to Trump. Most of the coefficients were positive and significant,
Manufacturing Loss Change Unemployment
Family Econ Relative Econ Dep
Democrat Independent Republican Democrat Independent Republican
−0.2
0.0
0.2
0.4
−0.2
0.0
0.2
0.4
C h a n g e P
ro b N
e w
T ru
m p V
o te
r | P
a rt
y
Sample
All Whites
WWC
Non−WC Whites
Figure 3. Economic marginality and switching to Trump. Points indicate effect of moving from minimum to maximum values (retrospective economic evaluations and economic deprivation) or from 2 s.d. below to above the mean (manufacturing loss and change in unemployment). Circles indicate model for all White respondents, triangles for just White working-class respondents, and squares for nonworking-class White respondents. Lines indicate simulated 95 percent confidence intervals.
Reny, Collingwood, and Valenzuela106
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
but the effects were too small to be substantively meaningful. With respect to H3, the effects of these attitudinal dispositions were slightly more asso- ciated with switching to Trump among working-class Whites and to Clinton among nonworking-class Whites, though the differences were small and often statistically indistinguishable. While more working-class Whites switched to Trump and more nonworking-class Whites to Clinton, the association between their symbolic racial and immigration attitudes and vote switching were not substantively different.13
Turning to economic indicators, figure 3 presents a similar plot with four panels for family economic marginality (min to max), relative economic
Manufacturing Loss Change Unemployment
Family Econ Relative Econ Dep
Democrat Independent Republican Democrat Independent Republican
−0.50
−0.25
0.00
0.25
0.50
0.75
−0.50
−0.25
0.00
0.25
0.50
0.75
C h a n g e P
ro b N
e w
C lin
to n V
o te
r | P
a rt
y
Sample
All Whites
WWC
Non−WC Whites
Figure 4. Economic integration and switching to Clinton. Points indicate effect of moving from minimum to maximum values (retrospective economic evaluations and economic deprivation) or from 2 s.d. below to above the mean (manufacturing loss and change in unemployment). Circles indicate model for all White respondents, triangles for just White working-class respondents, and squares for nonworking-class White respondents. Lines indicate simulated 95 percent confidence intervals.
13. Two explanations may speak to why racially conservative White voters were supporting Obama in 2012 in the first place. First, the 2016 election was far more racialized than the 2008 or 2012 elections, sending a clearer signal of racial positions between the two candidates, which might filter down to even the least politically aware citizens. Second, the 2016 election followed a longer trend of racially white conservative Democrats sorting into the Republican Party, a process that was far from complete in 2012 and will likely continue past 2016. We expand on these argu- ments in Online Appendix K.
Vote Switching in the 2016 Election 107
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
deprivation (min to max), county-level manufacturing loss (μ + / −2 s.d.), and change in county-level unemployment (μ + / −2 s.d.) for the same subgroups.
Across the board, weaker relationships exist between economic indica- tors and vote switching to Trump than our race and immigration measures. Hypothesis 4a predicted that White voters experiencing economic margin- ality—negative economic retrospective evaluations or relative economic de- privation—will be more likely to switch to Trump than those who do not. Weak support exists for this argument. The first panel of figure 3 shows that those with the strongest decline in family income over the previous year were slightly more likely to switch to Trump than those with improving family in- comes. White working-class Democrats and Independents who reported the steepest declines in family income were only about 5.4 (95 percent CI: [3.5,8]) and 6.9 (95 percent CI: [2.3,11.7]) percentage points more likely to switch to Trump. That jumps to an imprecisely estimated 19.4 points (95 percent CI: [3,35]) for Republicans. We find no relationship between relative economic deprivation and switching to Trump for any subgroup.14
While individual-level measures of economic marginality are only weakly associated with switching to Trump in 2016, perhaps contextual-level indica- tors are more robust predictors of vote switching given the Trump campaign’s focus on widespread job losses and manufacturing decline in the United States. Hypothesis 5a posited that White citizens who lived in economically declining counties were more likely to switch to Trump than similarly situated voters whose communities were not undergoing economic decline. As shown in figure 3, there is no relationship between county-level economic decline and vote switching in 2016.
Finally, figure 4 displays the same results for the Clinton models. Hypothesis 5a posited that positive retrospective evaluations and positive relative family income would be associated with switching to Clinton. Once again, in flipping the direction of the counterfactual simulation to be consistent with our hypoth- eses, we find no substantively and statistically significant relationship between economic marginality or local economic dislocation and vote switching for Clinton. Similarly null results emerge for tests of local economic dislocation and vote switching for Clinton.15
14. While the retrospective measure is positively related to vote switching, we also note that evaluations of finances are influenced by a respondent’s partisanship and the party that happens to be in power (Healy, Persson, and Snowberg 2017), though far less so than evaluations of the national economy (Bartels 2002), suggesting that part of the effect found here could be simply reflecting partisanship. 15. Online Appendixes L and M present two additional analyses assessing mobilization/demo- bilization between 2012 and 2016 and assessing whether effects are amplified in swing states versus non-swing states. The relationships between the key IVs and outcomes look similar for mobilization (examining those who did not vote in 2012) but much weaker for demobilization. No strong differences emerged between swing and non-swing state residents.
Reny, Collingwood, and Valenzuela108
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
In sum, our analyses yield two core findings that both run counter to dom- inant media narrative on the 2016 election. First, we find a much stronger association between symbolic racial and immigration attitudes and switching for Trump and Clinton than between economic marginality or local economic dislocation and vote switching. In fact, we find marginally small or no asso- ciations between any of our economic indicators and vote switching in either direction. Second, while significantly more working-class Whites switched votes to Trump in 2016 than nonworking-class Whites, lending some credence to election reporting, little evidence exists that working-class Whites were sig- nificantly more motivated by racial and immigration attitudes to switch than nonworking-class Whites.
Discussion and Conclusion
The 2016 election was unique both for the unorthodox candidacy of Donald Trump and for featuring the first female nominee of the two major parties. Trump surprised the world by pulling off an upset victory, with unexpected wins in a number of “blue firewall” states. Subsequent media analyses of the election highlighted the role of both economic anxiety and racial and ethnic attitudes among the White working class in driving this outcome. In this inves- tigation, we sought to understand whether immigration or economics played a bigger role in this process, whether this vote switching was isolated among the working class, and whether voters were switching away from the Republican Party and toward Clinton as well.
Throughout this paper, we presented evidence that Trump’s and Clinton’s candidacies and campaign messages did likely have an effect on voting trends. White voters with racially conservative or anti-immigrant attitudes switched votes to Trump at a higher rate than those with more liberal views on these issues. At the same time, White voters who had liberal views on race and immi- gration moved toward Clinton. Congruent with media coverage, vote switch- ing to Trump was, in raw numbers, far more prevalent among the working class than the nonworking class, though the relationship between attitudes and switching did not vary significantly by class. The inverse was true for Clinton. We find little evidence that economic dislocation and marginality were signifi- cantly related to vote switching in 2016.
While this, by itself, is not evidence of partisan realignment, history sug- gests that significant changes in voting across party lines, particularly for the presidency, precede changes in party identities, the basis for realign- ments. This sequence of events played out during the Southern realignment (i.e., Democrats voting for GOP presidential candidates but maintaining their party attachment), and here we provide evidence that it may be happening again after two terms with a Black president and during an era of mass demo- graphic change due to immigration. Racial conservatives and those with the
Vote Switching in the 2016 Election 109
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
most punitive immigration views are moving right and were the most likely to switch to Trump in 2016. Our data suggest the same is happening in the op- posite direction as those with racially liberal or pro-immigration views may be sorting into the Democratic Party.
Our findings also speak to how elites are responding to changing demo- graphics and racial realities. As communities around the country diversify, immigration and race are increasingly dominating campaign messaging. Many White voters feel left behind as the Democratic Party becomes the party of highly educated Whites and a consortium of minority groups. The Republican Party, historically the party of the wealthy and of business interests, has not offered many of these White voters a home either. But after eight years of the nation’s first Black president, Trump, the candidate who spurned the GOP establishment and played so well to a sense of resentment over a changing country, reached out and signaled that he would, in so many words, make the country White and working class again.
Supplementary Data
Supplementary data are freely available at Public Opinion Quarterly online.
References
Abrajano, Marisa, and Zoltan Hajnal. 2015. White Backlash: Immigration, Race, and American Politics. Princeton, NJ: Princeton University Press.
Abramowitz, Alan I. 2016. “Will Time for Change Mean Time for Trump?” PS: Political Science & Politics 49:659–60.
Abramowitz, Alan I., and Kyle L. Saunders. 2006. “Exploring the Base of Partisanship in the American Electoral: Social Identity vs. Ideology.” Political Research Quarterly 59:175–97.
Adams, Kimberly. 2016. “Key for Trump Voters? Economic Anxiety.” Marketplace.org. Available at https://www.marketplace.org/2016/11/10/elections/economic-anxiety-key-some-trump-voters.
Ahler, Douglas, and Guarav Sood. 2018. “The Parties in Our Heads: Misperceptions About Party Composition and Their Consequences.” Journal of Politics 80:964–81.
Ansolabehere, Stephen, Jonathan Rodden, and James M. Snyder. 2008. “The Strength of Issues: Using Multiple Measures to Gauge Preference Stability, Ideological Constraint, and Issue Voting.” American Political Science Review 102:215–32.
Ansolabehere, Stephen and Brian F. Schaffner. 2017. “CCES Common Content, 2016.” Harvard Dataverse, V4, UNF:6:WhtR8dNtMzReHC295hA4cg== [fileUNF]. doi:10.7910/DVN/ GDF6Z0
Barreto, Matt A., Loren Collingwood, and Sylvia Manzano. 2010. “A New Measure of Group Influence in Presidential Elections: Assessing Latino Influence in 2008.” Political Research Quarterly 63:908–21.
Bartels, Larry. 2002. “Beyond the Running Tally.” Political Behavior 24:117–50. Berelson, Bernard R., Paul F. Lazarsfeld, and William N. McPhee. 1954. Voting: A Study of
Opinion Formation in a Presidential Campaign. Chicago: University of Chicago Press. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The
American Voter. New York: Wiley & Sons.
Reny, Collingwood, and Valenzuela110
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
Carmines, Edward G., and James A. Stimson. 1989. Issue Evolution: Race and the Transformation of American Politics. Princeton, NJ: Princeton University Press.
Carnevale, Anthony P., Tamara Jayasundera, and Artem Gulish. 2015. “Good Jobs Are Back: College Graduates Are First in Line.” Washington, DC: Georgetown University Center on Education and the Workforce. Available at https://cew.georgetown.edu/wp-content/uploads/ Good-Jobs_Full_Final.pdf.
Collingwood, Loren, Matt A. Barreto, and Sergio I. Garcia-Rios. 2014. “Revisiting Latino Voting: Cross-Racial Mobilization in the 2012 Election.” Political Research Quarterly 67:632–45.
Converse, Philip E. 1964. “The Nature of Belief Systems in Mass Publics.” Critical Review 18:1–74.
Craig, Maureen A., and Jennifer A. Richeson. 2014. “On the Precipice of a Majority-Minority America: Perceived Status Threat from the Racial Demographic Shift Affects White Americans’ Political Ideology.” Psychological Science 25:1189–97.
Enders, Adam M., and Jamil S. Scott. 2018. “The Increasing Racialization of American Electoral Politics.” American Politics Research. doi:10.1177/1532673X18755654.
Enos, Ryan D. 2014. “Causal Effect of Intergroup Contact on Exclusionary Attitudes.” Proceedings of the National Academy of Sciences 111:3699–3704.
Fiorina, Morris P. 1981. Retrospective Voting in American National Elections. New Haven, CT: Yale University Press.
Fox, Cybelle. 2004. “The Changing Color of Welfare? How Whites? Attitudes Toward Latinos Influence Support for Welfare.” American Journal of Sociology 110:580–625.
Frymer, Paul. 2010. Uneasy Alliances: Race and Party Competition in America. Princeton, NJ: Princeton University Press.
Gest, Justin. 2016. The New Minority: White Working Class Politics in an Age of Immigration and Inequality. New York: Oxford University Press.
Green, Donald P., Bradley Palmquist, and Eric Schickler. 2002. Partisan Hearts and Minds: Political Parties and the Social Identities of Voters. New Haven, CT: Yale University Press.
Hainmueller, Jens, and Daniel Hopkins. 2014. “Public Attitudes Toward Immigration.” Annual Review of Political Science 17:225–49.
Healy, Andrew J., Mikael Persson, and Erik Snowberg. 2017. “Digging into the Pocketbook: Evidence on Economic Voting from Income Registry Data Matched to a Voter Survey.” American Political Science Review 111:771–85.
Hopkins, Daniel. 2018. The Increasingly United States: How and Why American Political Behavior Nationalized. Chicago, IL: University of Chicago Press.
Huddy, Leonie, and Alexa Bankert. 2017. “Political Partisanship as a Social Identity.” Oxford Research Encyclopedia of Politics 1:1–30.
Huddy, Leonie, Alexa Bankert, and Caitlin Davies. 2018. “Expressive Versus Instrumental Partisanship in Multiparty European Systems.” Political Psychology 39:173–99.
Huddy, Leonie, Lilliana Mason, and Lene Aarøe. 2015. “Expressive Partisanship: Campaign Involvement, Political Emotion, and Partisan Identity.” American Political Science Review 109:1–17.
Jerome, Bruno, and Veronique Jerome-Speziari. 2016. “State-Level Forecasts for the 2016 US Presidential Election: Political Economy Model Predicts Hillary Clinton Victory.” PS: Political Science and Politics 49:680–86.
Kennedy, Courtney, Andrew Mercer, Scott Keeter, Nick Hatley, Kyley McGeeney, and Alejandra Gimenez. 2016. “Evaluating Online Nonprobability Surveys.” Pew Research Center. Available at http://www.pewresearch.org/2016/05/02/ evaluating-online-nonprobability-surveys/.
Kuziemko, Ilyana, and Ebonya Washington. 2018. “Why Did the Democrats Lose the South? Bringing New Data to an Old Debate.” American Economic Review 108: 2830-67.
Lewis-Beck, Michael S., and Mary Stegmaier. 2000. “Economic Determinants of Electoral Outcomes.” Annual Review of Political Science 3:183–219.
Vote Switching in the 2016 Election 111
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
López, Ian Haney. 2015. Dog Whistle Politics: How Coded Racial Appeals Have Reinvented Racism and Wrecked the Middle Class. New York: Oxford University Press.
Lopez, Mark Hugo, Ana Gonzalez-Barrera, Jens Manuel Krogstad, and Gustavo Lopez. 2016. “Latinos and the Political Parties.” Pew Research Center. Available at http://www.pewhispanic. org/2016/10/11/latinos-and-the-political-parties/.
Martin, Gregory J., and Josh McCrain. 2018. “Local News and National Politics.” American Political Science Review. Available at http://joshuamccrain.com/localnews.pdf.
Mason, Lilliana, and Julie Wronski. 2018. “One Tribe to Bind Them All: How Our Social Group Attachments Strengthen Partisanship.” Political Psychology 39:257–77.
Meyerson, Harold. 2015. “America’s White Working Class Is a Dying Breed.” Washington Post. Available at https://www.washingtonpost.com/opinions/the-white-working- class-is-a-dying-breed/2015/11/04/f2220170-8323-11e5-a7ca-6ab6ec20f839_story. html?utm_term=.6c68115a1433.
Mutz, Diana. 2018. “Status Threat, Not Economic Hardship, Explains the 2016 Presidential Vote.” Proceedings of the National Academies of Sciences 115: E4330-E4339.
Newman, Benjamin J. 2013. “Acculturating Contexts and Anglo Opposition to Immigration in the United States.” American Journal of Political Science 57:374–90.
Newman, Benjamin J., Sono Shah, and Loren Collingwood. 2018. “Race, Place, and Building a Base: Latino Population Growth and the Nascent Trump Campaign for President.” Public Opinion Quarterly 82:122–34.
Ostfeld, Mara Cecilia. 2018. “The New White Flight? The Effects of Political Appeals to Latinos on White Democrats.” Political Behavior: 1–22.
Parker, Christopher S. and Matthew A. Barreto. 2013. Change They Can’t Believe In: The Tea Party and Reactionary Politics in America. Princeton, NJ: Princeton University Press.
Prior, Markus. 2007. Post-Broadcast Democracy: How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. New York: Cambridge University Press.
Putnam, Robert D. 2001. Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster.
Rivers, Doug. 2016. “Pew Research: YouGov Consistently Outperforms Competitors on Accuracy.” Available at https://today.yougov.com/topics/finance/articles-reports/2016/05/13/ pew-research-yougov.
Rivers, Doug, and Benjamin Lauderdale. 2016. “Beware the phantom swings: why dramatic bounces in the polls aren’t always what they seem.” YouGov.com. Available at https://today.yougov.com/topics/ politics/articles-reports/2016/11/01/beware-phantom-swings-why-dramatic-swings-in-the-p.
Rothschild, Jacob E., Adam J. Howat, Richard M. Shafranek, and Ethan C. Busby. 2018. “Pigeonholing Partisans: Stereotypes of Party Supporters and Partisan Polarization.” Political Behavior.
Sargent, Greg. 2017. “Why Did Trump Win? New Research by Democrats Offers a Worrisome Answer.” Washington Post. Available at https://www.washingtonpost.com/blogs/plum- line/wp/2017/05/01/why-did-trump-win-new-research-by-democrats-offers-a-worrisome- answer/?utm_term=.79246d50a861.
Schickler, Eric. 2016. Racial Realignment: The Transformation of American Liberalism, 1932– 1965. Princeton, NJ: Princeton University Press.
Sides, John, Michael Tesler, and Lynn Vavreck. 2018. Identity Crisis: The 2016 Presidential Campaign and the Battle for the Meaning of America. Princeton, NJ: Princeton University Press.
Tajfel, Henri. 1981. Human Groups and Social Categories. Cambridge: Cambridge University Press.
Teixeira, Ruy, and Alan Abramowitz. 2008. “The Decline of the White Working Class and the Rise of a Mass Upper Middle Class.” Brookings Institution. Available at https://www.brook- ings.edu/wp-content/uploads/2016/06/04_demographics_teixeira.pdf.
Reny, Collingwood, and Valenzuela112
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
Tesler, Michael. 2015. “Priming Predispositions and Changing Policy Positions: An Account of When Mass Opinion Is Primed or Changed.” American Journal of Political Science 59:806–24.
———. 2016. Post-Racial or Most-Racial? Race and Politics in the Obama Era. Chicago: University of Chicago Press.
Treier, Shawn, and D. Sunshine Hillygus. 2009. “The Nature of Political Ideology in the Contemporary Electorate.” Public Opinion Quarterly 73:679–703.
Valentino, Nicholas, Ted Brader, and Ashley Jardina. 2013. “Immigration Opposition Among U.S. Whites: General Ethnocentrism or Media Priming of Attitudes About Latinos?” Political Psychology 34:149–66.
Wong, Janelle, Karthick Ramakrishnan, Taeku Lee, and Jane Junn. 2011. Asian American Political Participation: Emerging Constituents and Their Political Identities. New York: Russell Sage Foundation.
Yadon, Nicole, and Spencer Piston. 2018. “Examining Whites’ Anti-Black Attitudes After Obama’s Presidency.” Politics, Groups, and Identities. doi:10.1080/21565503.2018.1438953
Zhirkov, Kirill, and Nicholas A. Valentino. 2017. “Blue Is Black and Red Is White? Affective Polarization and the Racialized Schemas of U.S. Party Coalitions.” Paper presented at the Midwest Political Science Association, Chicago, IL, 2017.
Zingher, Joshua N. 2018. “Polarization, Demographic Change, and White Flight from the Democratic Party.” Journal of Politics 80:860–72.
Vote Switching in the 2016 Election 113
D ow
nloaded from https://academ
ic.oup.com /poq/article-abstract/83/1/91/5494625 by U
niversity of C alifornia, Los A
ngeles user on 22 M ay 2019
