Race, Place, and Building a Base:Local Latino Threat and the Nascent Trump Campaign for President

ABSTRACT

A prominent feature of Donald Trump's campaign for President was the use of racially inflammatory rhetoric and fear over immigration—specifically from Mexico—to galvanize the electorate. Despite the commonly accepted assertion that hostility toward Mexican immigrants was an important attractor of core supporters to his base, analysts and academics alike have failed to explore the role that environmental indicators of perceived threat from immigration, such as residing in an area with a growing Latino population, played in generating support for Trump early in his campaign. We demonstrate that residing in a high Latino growth area is predictive of support for Trump following, but not before, his utterance of inflammatory and bellicose comments about Mexican immigrants. Our results suggest that, in addition to the importance of racial resentment and economic frustration, support for Trump in the early campaign period represented an adversarial reaction among Americans to Latino-led diversity.

Word Count: 2,750

On June 16, 2015, Donald Trump gave a rousing presidential announcement speech,

sparking an immediate reaction among the public, media, and political scientists. A prominent

feature of the Trump campaign was the galvanizing role that race played throughout his

campaign. Trump's campaign was marked by a focus on themes of fear and safety, with

recurrent attention given to the putative threat of rampant immigration from Mexico. Key

moments in his campaign were his reference to Mexican immigrants as "criminals" and "rapists"

and his promise to build a "great wall" along the border and make Mexico pay for the wall.

These statements, occurring during the summer of 2015, were met with significant controversy,

as many commentators decried them as explicitly racist. Trump’s blunt statements violated many

tenets of candidate behavior, including the norm of equality and racial tolerance (Mendelberg

2001). Nevertheless, Trump's poll numbers continued to rise in the lead-up to the 2016 Primary

Elections, causing pundits and scholars to search for explanations for Trump's political ascent.

Over the course of the campaign trail, many explanations were offered for the rise of

Trump. Importantly, nearly all of these explanations point to voters' characteristics, such as

Anglo heritage and lack of education1, economic frustration2, racial resentment3,

authoritarianism4, and ethnocentrism5. Left out of the discussion so far is the role that

environmental factors, such as local ethnic context, may have played in influencing citizens'

gravitation toward Trump. Indeed, while scholarship has identified fear of immigration and

1 http://www.nytimes.com/2016/07/26/upshot/the-one-demographic-that-is-hurting-hillary-clinton.html?rref=collection%2Fbyline%2Fnate-cohn&action=click&contentCollection=undefined&region=stream&module=stream_unit&version=latest&contentPlacement=21&pgtype=collection2 https://www.washingtonpost.com/news/monkey-cage/wp/2016/03/04/how-political-science-helps-explain-the-rise-of-trump-part-3-its-the-economy-stupid/3 https://www.washingtonpost.com/news/monkey-cage/wp/2016/03/03/how-political-science-helps-explain-the-rise-of-trump-the-role-of-white-identity-and-grievances/4 http://www.politico.com/magazine/story/2016/01/donald-trump-2016-authoritarian-2135335 https://www.washingtonpost.com/news/monkey-cage/wp/2016/02/28/what-differentiates-trump-supporters-from-other-republicans-ethnocentrism/?tid=a_inl

2

growing minority populations as an important predictor of support for Trump6, it is unclear

whether locally experienced increases in immigration-driven ethnic diversity drew citizens to

Trump's base. In the early campaign period, as Trump increasingly established himself as the

hard-line opponent of immigration, it is possible that rising Trump support came from areas of

the country experiencing large growths in immigrant populations.

One literature that provides a basis for the expectation of an effect of citizens' local ethnic

context on their evaluations of Trump is the public opinion research on immigration. A dominant

thread of work in this literature focuses on how citizens' racial context influences their perception

of threat from immigrants (Hainmueller and Hopkins 2014). This research indicates that

Americans are particularly sensitive to changes in levels of diversity (Newman and Velez 2014),

and that increases in local immigrant populations can trigger threat perceptions and activate

support for restrictive policies (Enos 2014; Hopkins 2010; Newman 2013). This work provides a

foundation for the expectation that local ethnic change was an important element in the genesis

of Trump's base of electoral support. In addition to helping explain where Trump support

emerged, this research provides insights about when the effects of ethnic context should

manifest. According to Hopkins (2010), many aspects of citizens' local environments—such as

increasing diversity—will capture their attention but remain non-political until prompted by

external agents, such as the media or political elites, to be connected with specific policy issues.

The timing of Trump's racially inflammatory statements represent cut-points in time, where,

following each controversial statement, increasing national attention was given to his campaign

and the issue of immigration. This leads to the expectation that residing in a context with

6 https://www.washingtonpost.com/news/monkey-cage/wp/2016/03/05/new-poll-shows-trump-supporters-more-likely-to-fear-a-majority-minority-u-s/

3

growing immigrant minority populations served as a contextual factor attracting residents to

Trump's campaign, but only after his controversial anti-immigrant statements.

Beyond informing expectations concerning where and when immigration-driven ethnic

change may have translated into support for Trump, the literature offers insights about what

immigrant group should be most influential. Prior research indicates that not all immigrant

minorities are equally threatening, as some immigrant groups, such as Asians, are constructed as

"model minorities" (Maddux et al. 2008), while others, such as Latinos, are constructed as threats

(Chavez 2013). Prior observational research finds that while residing near large Asian

populations is associated with less opposition to immigration, residing near large Latino

populations is associated with greater opposition to immigration (Ha 2010). Adding to this,

experimental research finds that citizens are more incensed by information about the costs of

immigration when the group in question is Latino (Brader et al. 2008), and express greater

disapproval of transgressive behaviors (e.g., being undocumented) when engaged in by a Latino

immigrant (Hartman et al. 2014).

This work leads us to the hypothesis that residing in an area where the Latino population

experienced considerable growth should serve as a predictor of Trump support, but only after

Trump's inflammatory and bellicose pronouncements concerning Mexican immigration. Prior to

these statements, there should be no systematic differences in Trump support between citizens

residing in low versus high Latino growth areas. In short, we seek to test whether a pivotal factor

in the building of Trump's base was his receipt of increasingly favorable evaluations among

Americans residing in contexts experiencing Latino-led ethnic diversification. Such a test offers

insight on whether an important component of the rise of Trump was a backlash among

Americans not simply to the abstract threat of a racially diverse and majority-minority America,

4

but rather to the lived-experience and palpable threat of Latino population growth challenging

prevailing racial hierarchies and community identities.

EVIDENCE USING NATIONAL SURVEY DATA

One challenge in testing our hypothesis is identifying national survey data conducted at

key moments in the nascent Trump campaign. These key moments are (1) the period of time

after his entry as a potential candidate in March 2015 (e.g., his formation of an exploratory

committee on March 18th) but before his infamous comments about Mexican immigrants, and (2)

the period of time following his campaign announcement speech on June 16th, 2015, where he

made his "rapists" and "criminals" comment about Mexican immigrants. These pockets of time

represent key moments in the nascent Trump campaign, serving as cut-points demarcating the

period before and after Trump established himself as a hard-line opponent of immigration and

racial provocateur. In addition to occurring within these points in time, any useable data would

require (1) items soliciting evaluations of Trump, (2) geocodes for respondents’ local area of

residence, and (3) a range of relevant control variables.

We located four surveys conducted by Monmouth University Polling Institute that met

these criteria: (1) the March 2015 Poll, conducted (March 30th - April 2nd) immediately after

Trump formed his exploratory committee, (2) the June 2015 Poll, conducted (June 11th-14th)

right before Trump's announcement speech and "rapists" comment, the July 2015 Poll, conducted

(July 9th - 12th) roughly a month after Trump's announcement speech, and the August 2015 Poll,

conducted (July 30th - August 2nd) weeks after Trump's July 11th speech proclaiming he would

build a wall along the U.S.-Mexico border. Each survey asked Republican respondents whether

they hold a favorable or unfavorable attitude toward Donald Trump7. Additionally, each survey 7 This includes Republicans and Republican-leaners. Respondents identifying with the Democratic Party were asked to evaluate Democratic candidates for President. While it would be of interest to know whether Trump's racial appeals resonated with Independent and Democratic voters residing in high Latino Growth contexts, the branched nature of the candidate favorability questions in the Monmouth surveys preclude this type of analysis, which we

5

includes geocodes for respondents' county of residence and a range of demographic control

variables and other relevant controls, such as disapproval of Obama.

We merged the Monmouth data into a single dataset, creating a unique Time variable for

each survey indicating survey-month8. We then merged this with data from the 2000 Decennial

Census and 2010-2014 American Community Surveys, which enabled us to construct our key

independent variable, Latino Growth, which is the percentage point increase in county Latino

populations between 2000 and 2010-20149. We conducted a moderated multiple regression

analysis10, where we assess the marginal effect of Latino Growth conditional upon Time—that is,

across the March 2015 - August 2015 time span of our combined Monmouth data. Our analysis

includes a range of individual and contextual controls, and we provide information about

question wording, variable measurement, and descriptive statistics in Appendix A.

We present the results from our analysis in Figure 1, Panel A, which plots the estimated

effects of county Latino Growth on the probability of reporting a "favorable" opinion of Trump

across each survey. Full results are presented in Table B1 in Appendix B. As can be seen, prior

to Trump's issuance of his two most infamous anti-Mexican statements, we do not observe a

statistically significant effect of Latino Growth on evaluations of Trump. However, immediately

following the "rapists" comment, the coefficient begins to take on a positive value (though

statistically insignificant), and after Trump had made both the "rapists" and the "wall"

believe represents an important direction for future research. 8 The variable Time is coded as follows: "1" for the March 2015 Poll, "2" for the June 2015 Poll, "3" for the July 2015 Poll, and "4" for the August 2015 Poll. 9 To provide complete coverage of all counties in the Monmouth data, we relied upon the 2010-2014 Five-Year American Community Survey (ACS) data file. Our use of a 10-14 year time frame for measuring Latino Growth is consistent with leading research on ethnic change in political science and sociology (Alexseev 2006; Citrin et al. 1990; Green et al. 1998; Hopkins 2010; Newman 2013), which measure change over time spans roughly comparable in length to the one selected in our analysis.10 Given the hierarchal structure of the data, where individual respondents are embedded within counties, we estimated a random intercepts multilevel logistic regression model. Our results entirely hold when estimating a single-level logistic regression model (see Table B3 in Appendix B). Our main results presented in Figure 1, Panel A (and Table B1) also hold when applying survey weights (see Table B4).

6

statements, we observe a positive and statistically significant effect of Latino Growth11. The

pattern of coefficients indicate a significant and positive interaction of Latino Growth with Time

(β= 2.57, s.e.= 1.27, p<.05). This effect is not an artifact of the interactive model specification,

as the pattern of estimates for Latino Growth derived from the moderated regression are

comparable to those obtained from by-survey regressions (Table B2 in Appendix B). These

results provide support for the hypothesis that residing in a context experiencing Latino

population growth created a latent source of support for Trump that was "activated" by his

racially inflammatory statements about Mexican immigrants.

ROBUSTNESS CHECKS

Our results hold when excluding Latinos from the analysis (Table B5) and when

measuring Latino Growth over a wider time period, such as 1990 to 2014 (Table B6).

Interestingly, we find that our results for Latino Growth weaken when focusing more specifically

on Latino Immigrant Growth (Table B7), suggesting that support for Trump following his anti-

Mexican comments is indicative of hostility toward Latinos in general12. An important

alternative model is one that includes dynamic versions of key contextual variables—

specifically, the economic variables that may serve as indicators of downward changes in

economic conditions. Indeed, Trump's campaign focused on the loss of manufacturing jobs,

unemployment, and the slipping "American dream"; thus, it is possible that the estimated effects

of Latino Growth are capturing the effects of changes in income levels, manufacturing jobs, and

11 The size of this effect is displayed at the bottom of Table B1 in Appendix B. The effect is substantively significant, as a 5th to 95th percentile increase in Latino Growth in August 2015 is associated with a .13 increase in the probability of favoring Trump.12 Less than half of the Latino population in the U.S. is foreign-born, and most of its recent growth has been due to birth rates instead of immigration (http://www.pewhispanic.org/2014/04/29/hispanic-nativity-shift/). Nevertheless, there persists a strong link between Latinos and immigration in the mass media (Valentino et al. 2013) and the minds of many Americans (Perez 2010). It is likely that the average American does not differentiate in their response to observing native-born vs. foreign-born Latino population growth. In fact, recent research by Rocha et al. (2011) demonstrates that Anglo American support for restrictive immigration policy increases the most in response to residing near large native-born, as compared to foreign-born, Latino populations.

7

employment. Table B8 demonstrates this is not the case, as our core results hold when ∆Median

Income, ∆Manufacturing Jobs, and ∆Unemployment, are interacted with Time.

An additional issue is the problem of residential self-selection and potentially non-

random assignment of the "treatment" variable in observational data. If residential self-selection

were at work, there is no reason to expect such an effect to be more or less operative across the

Monmouth surveys, and thus, to serve as an explanation for the varying effects of Latino Growth

in March 2015 versus September 2015. Despite this, we estimated an alternative model

controlling for county demographics in the year 2000 as possible pretreatment confounders.

These confounders are characteristics of counties in the year 2000 that may have influenced both

where Latino Growth occurred and where Trump support evolved. On top of this, our 2014

contextual measures may be influenced by Latino Growth, and thus, introduce post-treatment

bias. Table B9 reveals that when we replace our 2014 contextual controls with 2000 controls, our

results actually strengthen.

One additional concern is that the pattern of coefficients observed in Figure 1, Panel A,

represent some unidentified trend that led Republican identifiers in high Latino Growth contexts

to report increasingly favorable evaluations of in-party elites in general. Over time, candidates

were investing more resources in campaigning and citizens may have become increasingly

polarized, especially in diversifying contexts given the salience of immigration as a political

issue. In Figure 1, Panel B, we present the results from our model for Trump subbing in

evaluations for other prominent Republican candidates for President (full results reported in

Table B1). The pattern we observe for Trump that is indexed to the timing of his inflammatory

statements is unique to evaluations of Trump and not observed for evaluations of the other

candidates. Additionally, it may have been the case that Trump intensified his campaigning

8

efforts in states with early primaries, and such states also house counties experiencing high

Latino growth. To check this, we re-estimated our model including a dummy variable for

residence in an early primary state and its interaction with Time, and find that our core results

entirely hold, and that no significant effects are observed for residence in an early primary state

nor its interaction with time (Table B10).

One final concern is that the significant effect of Latino Growth observed in the August

2015 Monmouth survey is a fluke and not indicative of Trump's racial rhetoric activating support

in high Latino growth contexts. In response, we obtained the 2015 Governance Survey

conducted by the Pew Research Center13 in the months (August 27th and October 4th)

immediately following the latest Monmouth poll in our analysis. In addition to enabling us to

assess whether we observe similar effects of Latino Growth on favorability evaluations of Trump

roughly one month later, the Pew data contains a range of potentially relevant controls not

included in the Monmouth data, such as respondent income, nationalism, racial resentment,

political trust, and dissatisfaction with the "direction" of the nation. We present the results from

this analysis in Table 1. As can be seen, we continue to observe a positive relationship between

Latino Growth and evaluations of Trump14. Additionally, this relationship is unique to

evaluations of Trump, as we observe null effects for Latino Growth on evaluations of Bush,

Cruz, and Rubio in this Pew data (Table B11).

CONCLUSION

Our findings indicate that support for Trump early in the campaign was drawn from areas

where citizens had lived experience with Latino growth, suggesting that the political ascent of

13 Drawn from landline and cellular RDD (AAPOR3 response rate: 9.5% landline, 8.3% cellular). N=6,004, however, two phases of the survey and different versions of the questionnaire were randomly administered to half-samples. This left N=1,491 respondents in Phase B receiving questionnaire Form 1, which were the subset of respondents given items about Trump and key control variables. Of these N=1,491 respondents, N=636 identified as republican or republican-leaning. This analysis is restricted to this subset of respondents. 14 These results essentially hold in early 2016 as well (Table B12).

9

Trump represents an adversarial reaction among racially threatened Americans to the expansion

of Latino populations in their own communities. By demonstrating the interaction between racial

appeals and citizens' local racial context, this research note makes a broader contribution by

building a bridge between existing research on race-based campaign appeals and the effect of

racial context on public opinion and electoral behavior (Hutchings and Valentino 2004).

References

Alexseev, M. A. 2006. "Ballot-box vigilantism? Ethnic population shifts and xenophobic voting in post-Soviet Russia." Political Behavior 28(3): 211-240.

10

Brader, T., Valentino, N. A. and Suhay, E. 2008. What Triggers Public Opposition to Immigration? Anxiety, Group Cues, and Immigration Threat. American Journal of Political Science, 52: 959–978.

Chavez, L. 2013. The Latino threat: Constructing immigrants, citizens, and the nation. Palo Alto: Stanford University Press.

Citrin, J., Reingold, B., and Green, D.P. 1990. American Identity and the Politics of Ethnic Change. Journal of Politics, 52(4): 1124-1154.

Enos, R. 2014. Causal Effect of Intergroup Contact on Exclusionary Attitudes. Proceedings of the National Academy of Sciences, 111(10): 3699-3704.

Green, D.P., Strolovitch, D.Z., and Wong, J.S. 1998. Defended Neighborhoods, Integration, and Racially Motivated Crime. American Journal of Sociology, 104(2): 372-403.

Ha, Shang E. 2010. "The Consequences of Multiracial Contexts on Public Attitudes Toward Immigration." Political Research Quarterly 63(1): 29-42.

Hainmueller, J. and Hopkins, D.J. 2014. Public attitudes toward immigration. Annual Review of Political Science, 17: 225-249.

Hartman, T.K., Newman, B.J. & Scott Bell, C. 2014. Decoding prejudice toward Hispanics: Group cues and public reactions to threatening immigrant behavior. Politcal Behav-

ior, 36(1): 143-163. 

Hutchings, V., and Valentino, N. 2004. The Centrality of Race in American Politics. Annual Review of Political Science, 7: 383-408.

Hopkins, D. 2010. Politicized Places: Explaining Where and When Immigrants Provoke Local Opposition. The American Political Science Review, 104(1), 40-60. 

Maddux, W., Galinksy, A.D., Cuddy. A. J., and Polifroni. M. 2008. When being a model minority is good…and bad: Realistic threat explains negativity towards Asian Americans. Personality and Social Psychology Bulletin, 34(1):74-89.

Mendelberg, T. 2001. The race card: Campaign strategy, implicit messages, and the norm of equality. Princeton, NJ: Princeton University Press.

Newman. B. 2013. Acculturating contexts and Anglo opposition to immigration in the United States. American Journal of Political Science, 57(2): pp. 374-390.

Newman, B. and Velez, Y. 2014. Group size versus change? Assessing Americans’ perception of local immigration. Political Research Quarterly, 67(2), pp. 293-303.

11

Nteta, Tatishe M. and Brian Schaffner. 2016. "New Poll Shows Trump Supporters are More Likely to Fear A Majority-Minority U.S." The Washington Post. Available online at: https://www.washingtonpost.com/news/monkey-cage/wp/2016/03/05/new-poll-shows-trump-supporters-more-likely-to-fear-a-majority-minority-u-s/

Parker, C.S. and Barreto, M. 2014. Change they can’t believe in: the Tea party and reactionary politics in America. Princeton: Princeton University Press.

Pérez, Efrén O. 2010. "Explicit evidence on the import of implicit attitudes: The IAT and immigration policy judgments." Political Behavior 32(4): 517-545.

Rocha, R.R., Longoria, T., Wrinkle, R.D., Knoll, B.R., Polinard, J.L., and Wenzel, J. 2011. Ethnic Context and Immigration Policy Preferences among Latinos and Anglos. Social Science Quarterly, 92(1): 1-19.

Valentino, Nicholas A., Ted Brader, and Ashley E. Jardina. 2013. "Immigration opposition among US Whites: General ethnocentrism or media priming of attitudes about Latinos?." Political Psychology 34(2): 149-166.

Figure 1. Effect of Local Hispanic Growth on Favorability of Republican Candidates in Nascent Stage of Presidential Campaign (Republicans Only)

Panel A. Donald Trump

12

-.4-.2

0.2

.4.6

M.E

. of L

atin

o G

row

th

3/2015 6/2015 Rapists 7/2015 Wall 8/2015

Survey Date

Panel B. Other Candidates

-.4-.2

0.2

.4.6

M.E

. of L

atin

o G

row

th

3/2015 6/2015 7/2015 8/2015

Survey Date

Bush

Cruz

Kasich

Rubio

Table 1. Replication−Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period (Republicans Only)

B SELocal Ethnic Context

13

Latino Growth 1.81* (.892)Contextual Controls Median Income -.396 (.797) Unemployment Rate .634 (.860) Manufacturing Jobs .184 (.615) Percent Black -.169 (.763) Romney Vote .470 (.617) Population Density 1.33 (1.18)Individual Controls Education -1.18** (.390) Income .396 (.372) Age .004 (.005) Male .358* (.177) Black -.247 (.558) Latino -.806* (.363) Asian .605 (.617) Dissatisfied -.007 (.274) Nationalism .351 (.286) Racial Resentment .865** (.276) Obama Disapproval 1.11** (.407) Ideology -.097 (.444) Trust in Government -.342 (.486)Constant -2.75 (.946)# Level 1 units (Individuals) 632# Level 2 units (County) 415Source: 2015 Governance Survey (August 27th – October 4th), Pew Research Center Notes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

14

15

APPENDIX AQuestion Wording and Variable Measurement

I. Contextual Variables

Latino GrowthThis variable is the percentage point change in the estimated percent Latino in each county between the 2000 Decennial Census and the 2010-14 five year American Community Survey. Mean=3.81, standard deviation=2.68Manufacturing JobsThis variable is the percent of the total employed civilian population that is 16 years and over within a county working reporting working in a job classified by the Census Bureau as a "Manufacturing" job. Derived from the 2010-2014 ACS. Mean= 10.8%, standard deviation=5.5. Romney VoteThis is the percent of votes within a county in the 2012 Presidential Election cast for Republican candidate Mitt Romney. This data was obtained from David Leip's Atlas of U.S. Presidential Elections. Mean= 48%, standard deviation=15.2

II. Monmouth Polls

Evaluation of Donald TrumpRespondents were asked: "I’m going to read you a few names of people who are running for presi-dent in 2016. Please tell me if your general impression of each is favorable or unfavorable, or if you don’t really have an opinion." They were asked this question in reference to a list of Republican nominees, including "Businessman Donald Trump". Response options were "Favorable" or "Unfa-vorable." From this question, we constructed an item coded "1" for respondents reporting "Favor-able" evaluations of Trump and "0" for respondents reporting "Unfavorable" evaluations of Trump or failing to provide an answer or reporting "no opinion." The percent of respondents in the data report-ing favorable evaluations of Trump by survey are as follows: (1) March 2015=25.07%, (2) June 2015= 21.4%, (3) July 2015=43.4%, and (4) August 2015=52%. EducationRespondents were asked: "What was the last grade in school you completed?", with the following re-sponse options: (1) 8th grade or less, (2) High School incomplete (Grades 9, 10 and 11), (3) High School Complete (Grade 12), (4) Vocational/Technical School, (5) Some College, (6) Junior College Graduate (2 year, Associate Degree), (7) 4 year College Graduate (Bachelor’s Degree), and (8) Grad-uate School (Masters, Law/Medical School, etc.). Mean= 5.51, standard deviation=1.98.AgeRespondents were asked to report their age on their last birthday. This item was open response. Mean= 53, standard deviation=17.8RaceRespondents were asked whether or not they are of Latino or Hispanic origin, as well as to report whether they are of White, Black, or Asian origin. Using responses to these items, we constructed dummy variables for non-Hispanic Whites, non-Hispanic Blacks, non-Hispanic Asians, and Latin@ respondents. ChildrenRespondents were asked if they have any children under the age of 18. Those who reported "Yes" were coded "1" and those who reported "No" were coded "0".

1

Obama DisapprovalRespondents were asked whether they approve or disapprove of the job Barack Obama is doing as President. Those who reported "disapprove" were coded "1" and those who reported "approve" were coded "0". Those reporting no opinion or not answering the question were coded "0". 49.4% of respondents across the combined Monmouth data disapproved of Obama.IdeologyRespondents were asked: "In general, would you describe your political views as liberal, moderate, or conservative?" Response options for this item ranged from (1)-"Very Liberal" to (5)-"Very Conservative." Mean= 3.2, standard deviation=1.1.

III. 2015 Governance Survey, Pew Research Center

NationalismRespondents were asked: "Which of these statements best describes your opinion about the Unites States?" The response options were: (1)-"The U.S. stands above all other countries in the world", (2)-"The U.S. is one of the greatest countries in the world, along with some others," and (3)-"There are other countries that are better than the U.S." From this item, we constructed a three category ordinal variable reversing the coding of the original variable, so that higher values indicate higher levels of nationalism (i.e., agreement with the notion that the U.S. stand above all other countries in the world). Racial ResentmentRespondents were asked: "I'm going to read you some pairs of statements that will help us understand how you feel about a number of things. As I read each pair, tell me whether the FIRST statement or the SECOND statement comes closer to your own views — even if neither is exactly right," and then presented with the following pair of statements: "Racial discrimination is the main reason why many black people can't get ahead these days" and [OR] "Blacks who can't get ahead in this country are mostly responsible for their own condition." Respondents who reported agreeing with the second statement were coded "1" and those agreeing with the former statement were coded "0". Trust in GovernmentRespondents were asked: " How much of the time do you think you can trust the government in Washington to do what is right? Just about always, most of the time, or only some of the time?" From this item we constructed a variable coded from (1)-"Never", to (4)-"Just About Always."DissatisfiedRespondents were asked: "All in all, are you satisfied or dissatisfied with the way things are going in this country today?" Using this item, we constructed a dichotomous variable coded "1" for those reporting being "dissatisfied" and "0" for those reporting being "satisfied" or reporting no opinion or failing to provide an answer. Trump EvaluationRespondents were asked: "Is your overall opinion of Donald J. Trump very favorable, mostly favorable, mostly unfavorable, or very unfavorable?" From this item, we constructed a dichotomous variable, coded "1" for respondents reporting "very favorable" or "favorable" and "0" otherwise.

IV. 2016 Religion & Politics Survey and 2016 March Political Survey, Pew Research Center

Trump Performance EvaluationRespondents were asked: "A year from now, a new president will be inaugurated...Regardless of who you currently support, I’d like to know what kind of president you think each of the following would

2

be if elected in November 2016? If Donald Trump were to become president do you think he would be a great, good, average, poor, or terrible president?" To simplify our analysis, we collapsed this item into a three category item coded "3" for respondents reporting "great" or "good," coded "2" for those reporting "average" and "1" for those reporting "poor" or "horrible."

3

APPENDIX BResults Tables and Auxiliary Results

[Tables Appear on Following Page]

4

5

Table B1. Effect of Local Latino Growth on Evaluations of Republican Candidates in Early Campaign Period (Republicans Only)

Placebo Tests

Trump Bush Cruz Kasich RubioLocal Ethnic Context Latino Growth -1.32 (.913) -.204 (.808) 1.41 (.865) -1.04 (.928) 1.17 (.839)Moderator Time .604 (.433) .093 (.406) .404 (.429) .295 (.456) .388 (.419)Interaction Growth × Time 2.57* (1.27) .882 (1.18) -1.10 (1.24) .108 (1.35) -.235 (1.22)Contextual Controls Median Income -.561 (.518) -.693 (.509) -.101 (.539) .384 (.542) .709 (.525) Unemployment Rate .280 (.552) -.835 (.549) -.675 (.580) .633 (.587) -.081 (.559) Manufacturing Jobs -.726 (.422) -.073 (.410) .002 (.432) .786^ (.438) -.122 (.419) Percent Black .757 (.423) .152 (.424) -.065 (.454) .129 (.456) .309 (.435) Romney Vote .964* (.412) .483 (.411) .244 (.435) -.234 (.439) .353 (.413) Population Density .499 (1.04) 1.88 (1.30) -1.70 (1.60) -1.25 (1.38) .126 (1.06)Individual Controls Education -.327 (.224) .549* (.217) .180 (.229) .820*** (.245) 1.05*** (.223) Age .004 (.004) .020*** (.004) -.001 (.004) .023*** (.005) .017*** (.004) Male .239* (.117) .028 (.112) .475*** (.119) .462*** (.125) .353** (.116) Black -.159 (.423) -.817* (.420) -.504 (.459) -.101 (.515) -.192 (.416) Latino .024 (.302) .413 (.285) .398 (.298) -.100 (.343) .567^ (.301) Asian -1.00^ (.565) .191 (.440) 1.22* (.496) .164 (.471) .751 (.472) Children .277^ (.150) .031 (.146) -.028 (.153) -.045 (.170) -.029 (.148) Obama Disapproval .136 (.187) -.218 (.178) .881*** (.197) .352 (.215) .533** (.180) Ideology .753** (.260) -.936*** (.250) 2.69*** (.283) .389 (.278) 1.69*** (.263)Constant -2.22 (.647) -.507 (.614) -3.39 (.667) -4.05 (.708) -4.28 (.655)# Level 1 units (Individuals) 1,457 1,457 1,457 1,457 1,457# Level 2 units (County) 743 743 743 743 743∆Pr(y) / ∆ Latino Growth: @ 3/2015 -.07 @ 8/2015 .13

Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from random intercepts logistic regression models estimated using xtlogit in Stata®, standard errors in parentheses. ∆Pr(y) / ∆ Latino Growth is the change in the Pr (Y=Favorable) associated with a 5th to 95th percentile change in Latino Growth in the March and August 2015 surveys.^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

6

7

Table B2. Effect of Local Latino Growth on Evaluations of Trump in Early Campaign Period, By Survey Analysis (Republicans Only)

3/2015 6/2015 7/2015 8/2015Local Ethnic Context Latino Growth -.378 (1.14) -1.06 (1.56) .409 (.966) 1.61^ (.941)Contextual Controls Median Income -1.06 (1.22) 1.30 (1.39) -1.46 (1.14) -.226 (.907) Unemployment Rate .431 (1.18) -.331 (1.49) -.348 (1.13) 1.39 (1.04) Manufacturing Jobs -.269 (.821) .350 (1.07) -1.02 (.935) -1.81* (.797) Percent Black -.495 (.988) 1.92^ (1.09) .949 (.991) 2.02* (.818) Romney Vote 1.18 (.976) 1.20 (1.08) -.920 (.954) 1.89* (.783) Population Density -2.43 (7.70) -9.96 (11.81) -16.63^ (9.34) 2.70^ (1.46)Individual Controls Education -1.05* (.470) .457 (.610) .248 (.458) -.329 (.436) Age -.009 (.009) .002 (.011) .023* (.009) .001 (.007) Male .082 (.262) -.130 (.307) .067 (.242) .988*** (.221) Black .079 (.921) -1.22 (1.24) .119 (.847) -.355 (.802) Latino .544 (.542) -1.66 (1.13) .613 (.866) -.066 (.555) Asian -23.50 (163729.20) -17.62 (4820.04) -.940 (1.21) .049 (.818) Children .106 (.330) .402 (.403) .471 (.304) .213 (.286) Obama Disapproval -.709^ (.395) -.462 (.427) .479 (.390) 1.10** (.397) Ideology 1.12^ (.585) -.111 (.659) .929^ (.546) 1.08* (.473)Constant -.546 (1.24) -2.19 (1.57) -1.76 (1.40) -4.04 (1.13)# Level 1 units (Individuals) 352 348 335 422# Level 2 units (County) 282 269 261 309Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from random intercepts logistic regression models estimated using xtlogit in Stata®, standard errors in parentheses.^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

8

Table B3. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Using a Single-Level Regression Model (Republicans Only)

B SELocal Ethnic Context Latino Growth -1.31 (.906)Moderator Time .600 (.431)Interaction Latino Growth × Time 2.56* (1.27)Contextual Controls Median Income -.561 (.515) Unemployment Rate .276 (.547) Manufacturing Jobs -.723^ (.419) Percent Black .753^ (.420) Romney Vote .960* (.408) Population Density .501 (1.04)Individual Controls Education -.327 (.224) Age .004 (.004) Male .238* (.116) Black -.158 (.422) Latino .028 (.300) Asian -.998^ (.563) Children .276^ (.149) Obama Disapproval .137 (.186) Ideology .750** (.258)Constant -2.21 (.642)N 1,457Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a logistic regression model estimated using logit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

9

Table B4. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Applying Survey Weights (Republicans Only)

B SELocal Ethnic Context Latino Growth -1.72 (1.21)Moderator Time .147 (.531)Interaction Latino Growth × Time 3.55* (1.62)Contextual Controls Median Income -.598 (.605) Unemployment Rate .054 (.623) Manufacturing Jobs -.480 (.471) Percent Black .700 (.486) Romney Vote 1.09* (.461) Population Density .334 (1.36)Individual Controls Education -.326 (.245) Age .004 (.004) Male .229 (.134) Black -.230 (.431) Latino .068 (.348) Asian -1.06^ (.627) Children .249 (.165) Obama Disapproval -.138 (.218) Ideology .581^ (.311)Constant -1.68 (.717)N 1,457Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

10

Table B5. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Excluding Latino Respondents (Republicans Only)

B SELocal Ethnic Context Latino Growth -1.26 (.949)Moderator Time .720 (.449)Interaction Latino Growth × Time 2.28^ (1.34)Contextual Controls Median Income -.468 (.525) Unemployment Rate .319 (.562) Manufacturing Jobs -.895* (.432) Percent Black .817^ (.429) Romney Vote .997* (.422) Population Density .530 (1.04)Individual Controls Education -.337 (.229) Age .004 (.004) Male .297* (.119) Black -.161 (.423) Asian -1.00^ (.563) Children .305* (.153) Obama Disapproval .124 (.193) Ideology .724** (.265)Constant -2.28 (.660)# Level 1 units (Individuals) 1,395# Level 2 units (County) 726Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

Table B6. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Using Alternative Latino Growth Measure (Republicans Only)

B SELocal Ethnic Context Latino Growth (1990-2014) -1.18 (.822)Moderator Time .982*** (.271)Interaction Latino Growth × Time 2.32* (1.14)Contextual Controls Median Income -.537 (.517) Unemployment Rate .275 (.557) Manufacturing Jobs -.705^ (.419) Percent Black .786^ (.427) Romney Vote .986* (.414) Population Density .502 (1.04)Individual Controls Education -.322 (.225) Age .004 (.004) Male .239* (.118) Black -.250 (.438) Latino .014 (.311) Asian -.993^ (.566) Children .276^ (.151) Obama Disapproval .170 (.189) Ideology .777** (.262)Constant -2.50 (.617)# Level 1 units (Individuals) 1451# Level 2 units (County) 742Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

11

Table B7. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Using Latino Immigrant Growth (Republicans Only)

B SELocal Ethnic Context Latino Immigrant Growth -1.61 (1.12)Moderator Time .712 (.630)Interaction Latino Imm. Growth × Time 1.80 (1.52)Contextual Controls Median Income -.425 (.520) Unemployment Rate .377 (.532) Manufacturing Jobs -.799^ (.416) Percent Black .833^ (.436) Romney Vote 1.00* (.414) Population Density .277 (1.04)Individual Controls Education -.331 (.224) Age .004 (.004) Male .230* (.117) Black -.111 (.421) Latino .017 (.298) Asian -1.01^ (.563) Children .267^ (.150) Obama Disapproval .128 (.187) Ideology .747** (.260)Constant -2.08 (.716)# Level 1 units (Individuals) 1457# Level 2 units (County) 743Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

12

Table B8. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Including Dynamic Economic Contextual Controls and Controlled Interactions (Republicans Only)

B SELocal Ethnic Context Latino Growth -.999 (.940)Moderator Time .291 (1.68)Interaction Latino Growth × Time 2.42^ (1.35)Contextual Controls ∆Median Income -1.34 (1.13) ∆Unemployment Rate -.071 (1.27) ∆Manufacturing Jobs .050 (.991) Percent Black .958* (.419) Romney Vote .935* (.388) Population Density .612 (1.06)Individual Controls Education -.334 (.223) Age .005 (.004) Male .239* (.117) Black -.135 (.422) Latino .040 (.302) Asian -.976^ (.568) Children .277^ (.150) Obama Disapproval .126 (.187) Ideology .759** (.261)Controlled Interactions ∆Median Income× Time .000 (.000) ∆Unemployment Rate × Time .039 (1.80) ∆Manufacturing Jobs ×Time -.051 (1.45)Constant -2.14 (1.31)# Level 1 units (Individuals) 1,457# Level 2 units (County) 743Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

13

Table B9. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Including Potential Pretreatment Confounders (Republicans Only)

B SELocal Ethnic Context Latino Growth -1.74^ (.996)Moderator Time .572 (.432)Interaction Latino Growth × Time 2.68* (1.28)

Contextual Controls Median Income 2000 -.692 (.489) Unemployment Rate 2000 -.332 (.712) Manufacturing Jobs 2000 -.412 (.397) Percent Hispanic 2000 .665 (.622) Percent Black 2000 .712^ (.397) Bush Vote 2000 .770^ (.433) Population Density 2000 .186 (1.07)

Individual Controls Education -.343 (.224) Age .004 (.004) Male .234* (.116) Black -.135 (.422) Latino -.053 (.307) Asian -1.09^ (.567) Children .271^ (.149) Obama Disapproval .141 (.186) Ideology .744** (.258)Constant -1.84 (.642)# Level 1 units (Individuals) 1,457# Level 2 units (County) 743Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

14

Table B10. Effect of Local Latino Growth on Evaluations of Donald Trump in Early Campaign Period, Controlling for Early Primary States (Republicans Only)

B SELocal Ethnic Context Latino Growth -1.31 (.914)Moderator Time .593 (.434)Interaction Latino Growth × Time 2.50* (1.28)

Contextual Controls Median Income -.539 (.520) Unemployment Rate .319 (.554) Manufacturing Jobs -.721^ (.425) Percent Black .750^ (.425) Romney Vote .983* (.414) Population Density .563 (1.04)

Individual Controls Education -.315 (.225) Age .004 (.004) Male .244* (.117) Black -.148 (.423) Latino .031 (.302) Asian -1.02^ (.567) Children .264^ (.150) Obama Disapproval .136 (.187) Ideology .761** (.261) Early Primary State -.213 (.615) Early Primary×Time 1.059 (.936)Constant -2.26 (.650)# Level 1 units (Individuals) 1,457# Level 2 units (County) 743Source: Monmouth University PollsNotes: Entries are unstandardized regression coefficients from a random intercepts logistic regression model estimated using xtlogit in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

Replication Analysis Using Data From the Early Primary Election Season of 2016

While our research note manuscript is primarily interested in exploring the determinants

of Trump support in the early campaign period, it is nonetheless of interest to assess whether our

main finding extends into the 2016 Primary Election season. One limitation is that the Pew

Research Center did not solicit favorability evaluations of Trump again until after Trump was

elected president (the favorability toward Trump question was not asked again until the

December 2016 Political Survey). This said, Pew did conduct surveys in January and March of

15

Table B11. Effect of Local Latino Growth on Evaluations of Other Republican Candidates in Early Campaign Period (Republicans Only)

Jeb Bush Ted Cruz Marco RubioLocal Ethnic Context Latino Growth 1.32 (.873) .797 (.895) -.255 (.902)Contextual Controls Median Income -.065 (.793) -.779 (.823) -.202 (.834) Unemployment Rate -1.09 (.854) -.959 (.883) -.216 (.885) Manufacturing Jobs -.642 (.612) -1.38* (.629) -1.63* (.633) Percent Black 1.61* (.766) .365 (.784) 1.25 (.823) Romney Vote 1.35* (.616) 1.49* (.632) .935 (.642) Population Density .506 (1.17) 4.04** (1.56) 2.56^ (1.54)Individual Controls Education .342 (.385) -.819 (.398) .336 (.400) Income .181 (.368) .723^ (.383) 1.25*** (.386) Age .016** (.005) .000 (.005) .012* (.005) Male .063 (.176) .125 (.181) .115 (.183) Black 1.15* (.588) -.534 (.574) -.341 (.560) Latino .453 (.342) .281 (.353) .722* (.365) Asian -.618 (.640) -1.74^ (.892) -.293 (.652) Dissatisfied .034 (.271) .094 (.281) .581* (.279) Nationalism -.324 (.284) .630* (.291) 1.11*** (.297) Racial Resentment -.287 (.263) -.031 (.274) -.160 (.275) Obama Disapproval -.622 (.397) .515 (.415) .286 (.410) Ideology .551 (.440) 2.93*** (.483) 1.30** (.464) Trust in Government 2.11*** (.513) .123 (.500) 1.20* (.512)Constant -2.28 (.918) -3.27 (.951) -3.84 (.963)# Level 1 units (Individuals) 632 632 632# Level 2 units (County) 415 415 415Source: 2015 Governance Survey, Pew Research CenterNotes: Entries are unstandardized regression coefficients from random intercepts logistic regression models estimated using xtlogit in Stata®, standard errors in parentheses.^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.

2016 asking respondents to report if they believe Trump—if elected—would be a good, average,

or poor president. The January 2016 Religion and Politics Survey was conducted between

January 7-14, resulted in a sample of N=2,000, and achieved a response rate (AAPOR3) of 9.7%.

The March 2016 Political Survey was conducted March 17-27, resulted in a sample of N=2,250,

and achieved a response rate (AAPOR3) of 7.9%. Combining these data, we find that Latino

Growth exerted a positive effect (B= .965, s.e.=.602) that is significant using a one-tailed test (p=

.054) and nearly marginally significant using a two-tailed test (p= .108). When restricting the

analysis to non-Latino respondents, the effect of Latino Growth increases in size and significance

(B= 1.27, s.e.=.646, p<.05, two-tailed). The results are displayed in Table B12 below. While

slightly weaker than our previous results, the consistency of this result with our previous results

is impressive given that it is based upon a different dependent variable and on data collected well

into the primary election season.

16

17

Table B12. Effect of Local Latino Growth on Evaluations of the Performance of Donald Trump if Elected President (Republicans Only)

Full Sample Excluding Latino Respondents

B SE B SELocal Ethnic Context Latino Growth .965 (.603) 1.27* (.646)

Contextual Controls Median Income .159 (.590) .064 (.624) Unemployment Rate .637 (.747) .538 (.785) Manufacturing Jobs .436 (.435) .462 (.457) Percent Black .182 (.494) .180 (.524) Romney Vote .184 (.440) .075 (.471) Population Density .170 (1.05) .411 (1.09)

Individual Controls Education -.212 (.258) -.333 (.272) Income -.339 (.214) -.436^ (.225) Age .008* (.004) .009* (.004) Male .412*** (.121) .470*** (.127) Black -.157 (.448) -.247 (.453) Latino -.552* (.239) - - - - - - Asian .455 (.510) .492 (.533) Dissatisfied .089 (.215) .049 (.230) Obama Disapproval 1.35*** (.217) 1.34*** (.230) Ideology .560^ (.316) .435 (.341) Church Attendance -.839*** (.207) -.751*** (.214)Survey Fixed Effect 3/2016 -.393** (.127) -.415** (.134)Thresholds Cut 1 1.12 (.707) .983 (.746) Cut 2 2.02 (.710) 1.85 (.749)# Level 1 units (Individuals) 1,172 1,095# Level 2 units (County) 654 630Source: Pew Research Center, January 2016 Religion and Politics Survey and March 2016 Political SurveyNotes: Entries are unstandardized regression coefficients from a random intercepts ordered logistic regression model estimated using gllamm in Stata®, standard errors in parentheses. ^p<.10, *p<.05, **p<.01, ***p<.001, based upon two-tailed hypothesis tests.