Below, I’ll examine three narratives that became widely accepted about the 2016 election and see how they stack up against the ANES data.
The rich, the poor, and the in-between
The first narrative was about how income affected vote choice. Trump was said to be unusually appealing to low-income voters, especially in the Midwest, compared with recent Republican presidential nominees. True or false?
The ANES provides us data on income and presidential vote choice going back to 1948. To remove the effects of inflation and rising prosperity, I plot the percentage voting for the Republican presidential candidate relative to the overall sample, by where they rank in U.S. income, from the top to the bottom fifth. The dashed horizontal line shows the average likelihood of voting for the GOP presidential candidate that year; a point above that means an income cohort was more likely than the other groups to vote for the Republican. To most directly test the Donald Trump income hypothesis, I’ve restricted this analysis to white voters.
Authoritarians or not?
Many commentators and social scientists wrote about how much about authoritarianism influenced voters. Authoritarianism, as used by political scientists, isn’t the same as fascism; it’s a psychological disposition in which voters have an aversion to social change and threats to social order. Since respondents might not want to say they fear chaos or are drawn to strong leadership, this disposition is measured by asking voters about the right way to rear children.
The idea is that voters anxious about change and disorder will say it’s best to encourage children to follow rules. For instance, respondents are asked whether it’s better when children are considerate (likely more liberal) or well-behaved (likely more authoritarian), or whether they should be self-reliant (likely more liberal) or obedient (likely more authoritarian).
The next chart shows how white GOP presidential voters have answered these questions since 2000. As we can see, Trump’s voters appear a little less authoritarian than recent white Republican voters.
Did racism affect the voting?
Many observers debated how important Trump’s racial appeals were to his voters. During the campaign, Trump made overt racial comments, with seemingly little electoral penalty. Could the unusual 2016 race have further affected Americans’ racial attitudes?
To test this, I use what is called the “symbolic racism scale” to compare whites who voted for the Democratic presidential candidate with those who voted for the Republican. This scale measures racial attitudes among respondents who know that it’s socially unacceptable to say things perceived as racially prejudiced. Rather than asking overtly prejudiced questions — “do you believe blacks are lazy” — we ask whether racial inequalities today are a result of social bias or personal lack of effort and irresponsibility.
In the chart below, you can see the scores for white voters who supported the Democratic and Republican presidential candidates between 1988 and 2016. For clarity, the second and fourth items have been reversed so that the larger values always indicate higher animosity.
Since 1988, we’ve never seen such a clear correspondence between vote choice and racial perceptions. The biggest movement was among those who voted for the Democrat, who were far less likely to agree with attitudes coded as more racially biased.
So which of these had the biggest influence?
Finally, the statistical tool of regression can tease apart which had more influence on the 2016 vote: authoritarianism or symbolic racism, after controlling for education, race, ideology, and age. Moving from the 50th to the 75th percentile in the authoritarian scale made someone about 3 percent more likely to vote for Trump. The same jump on the SRS scale made someone 20 percent more likely to vote for Trump.
Racial attitudes made a bigger difference in electing Trump than authoritarianism.
Thomas Wood is an assistant professor of political science at Ohio State University. He studies public opinion and elections. Follow him on Twitter @thomasjwood.
Note: The first graph comparing the presidential vote by income levels has been replaced with a version using the correct survey weight for the 1948-2012 data. This does not change the substantive findings.