Mass–Elite Divides in Aversion to Social Change and Support for Donald Trump

Mass–Elite Divides in Aversion to Social Change and Support for Donald Trump. Matt Grossmann, Daniel Thaler. American Politics Research,  https://doi.org/10.1177/1532673X18772280

Abstract: Donald Trump won the American presidency in 2016 by overperforming expectations in upper Midwest states, surprising even Republican political elites. We argue that attitudes toward social change were an underappreciated dividing line between supporters of Trump and Hillary Clinton as well as between Republicans at the mass and elite levels. We introduce a concept and measure of aversion to (or acceptance of) social diversification and value change, assess the prevalence of these attitudes in the mass public and among political elites, and demonstrate its effects on support for Trump. Our research uses paired surveys of Michigan’s adult population and community of political elites in the Fall of 2016. Aversion to social change is strongly predictive of support for Trump at the mass level, even among racial minorities. But attitudes are far more accepting of social change among elites than the public and aversion to social change is not a factor explaining elite Trump support. If elites were as averse to social change as the electorate—and if that attitude mattered to their vote choice—they might have been as supportive of Trump. Views of social change were not as strongly related to congressional voting choices.

Keywords: political parties, vote choice, political elites, racial resentment, diversity

---
We sought to assess the relationship between attitudes toward social change and vote preference. Our measure of aversion to change is an additive scale made up of two components—respondents’ level of agreement with a pair of statements about changing cultural values:

1. “Our country is changing too fast, undermining traditional American values.”
2. “By accepting diverse cultures and lifestyles, our country is steadily improving.”

[...]

Our major dependent variable of interest, vote preference, is a three-category ordinal variable created from a survey item asking respondents which of the two major candidates they most support for the presidency in 2016. Each of these variables takes on a value of 0 if the respondent preferred Clinton, a value of 1 if the respondent preferred Trump, and a value of 0.5 if the respondent preferred another candidate or could not decide. A similar variable records the respondent’s preference between the major party candidates in their local congressional election.

Our measure of authoritarian attitudes is based on a measure used by Feldman and Stenner (1997). We constructed a 3-point scale from 0 to 1 from two binary items that asked respondents to choose which of a given pair of personal qualities is more important for a child to have: obedience versus self-reliance, and independence versus respect for elders. Preference for obedience and respect for elders were considered the more authoritarian choices. Our measure of racial resentment is a 9-point scale from 0 to 1 constructed from respondents’ reported level of agreement or disagreement with two statements about race—one positing that African Americans should overcome prejudice and work their way up without any special favors like some other minority groups did, and one (coded in the opposite direction) positing that generations of slavery and discrimination make it difficult for African Americans to work their way up financially. Higher values indicate higher levels of resentment.

Ethnocentrism is measured using a set of “feeling thermometer” questions for particular racial and religious groups, comparing the respondent’s rating of Whites to their rating of Blacks, Hispanics and Latinos, and Muslims. In particular, the variable is coded as the average difference between the score given by the respondent to “Whites” and the score the respondent gave to each of the three minority groups (rescaled from 0 to 1). Minority respondents are coded as having values of 0 on this ethnocentrism scale. Calculating the ethnocentrism of non-White respondents the same way does not change our conclusions in any significant way.