(Jabin Botsford/The Washington Post)

In state after state, a recurrent question of the 2016 campaign has been how much two powerful social forces, economic anxiety and racial anxiety, are driving support vote for Donald Trump. There is a good deal of evidence that both factors are at play in the Republican presidential primary races.

Trump’s supporters, for instance, disproportionately lack any college education — a demographic that has suffered badly over the past 20 years in the economy. Perhaps most striking is that support for Trump is highly correlated with counties where death rates for working-class whites have spiked. Polls also suggest that Trump does well among voters with greater racial and immigrant antipathy, and that he is popular in places where Google searches indicate that bias is prevalent.

But a crucial unanswered question concerns how these two forces interact. Are Trump’s supporters mainly motivated by a shared concern about white status and identity? Or are economic concerns the basis of Trump’s support?

To investigate the relationship between these factors, The Washington Post and ABC News asked several new questions in a national poll earlier this month.

Among many other topics, respondents were asked whether they were struggling economically — or whether they were comfortable and moving up. They were also asked whether they thought it was more of a problem that African Americans and Latinos are “losing out because of preferences for whites” or whether whites are “losing out because of preferences for blacks and Hispanics.” The survey asked this question in order to gauge the sentiment that racial and ethnic groups don’t just feel they are facing difficulties in general, but that those losses are being caused by other group’s gains.

The following chart summarizes results from Republicans and Republican-leaning independent registered voters. Trump had the support of 34 percent of Republican-leaning voters overall. But he did particularly well among people who said they are struggling economically, with 40 percent of their support, and even better – 43 percent –among people who said that whites are losing out.


All in all, the results suggest that both economic and racial anxieties are driving supporters to Trump. To learn more about the connection between these two kinds of concerns, The Post performed a series of statistical tests on the data.

The Post found (see how below) that both economic troubles and feelings that whites are losing out have a strong – and independent – impact on Trump’s supporters. Republicans who are worried about maintaining their economic situation are more likely to support Trump, regardless of whether they think that whites losing out to other groups is a big problem. Likewise, those who feel that whites are losing out are more likely to support Trump regardless of their level of economic anxiety. The analysis took both factors into account, and also controlled for demographics, income and political ideology.

Those who voiced concerns about white status appeared to be even more likely to support Trump than those who said they were struggling economically, but the results did not clearly show which concern was more important among Trump’s coalition. That’s akin to when a candidate is leading another candidate in a poll, but the lead is within the poll’s margin of error.

The results are represented in a chart below. The orange bars represent the relative importance of both sources of Trump’s support in statistical terms. Each bar shows an estimate of how holding one attitude versus another affects the chances that a person supports Trump, taking into account demographic differences. The odds that a person who feels strongly that whites are losing out supports Trump are more than three times higher than for a demographically and financially similar person who feels blacks or Hispanics are losing out or that neither group is losing out more. Likewise, the odds a person who says he or she is struggling financially supports Trump are about twice as high as someone who says he or she is comfortable or moving up economically. The width of bars indicate the possible range in each effect’s size with 95 percent confidence, since the results are only a statistical estimate based on a random sample of the population.


None of these measures is perfect, of course. Pollsters use many types of questions to measure economic anxiety and racial anxiety, and it’s possible more sensitive measures might have resulted in different effects. But the persistent impact of both measures on Trump’s support in this survey, even after statistical tests controlled for the effect of demographics and other attitudes, is evidence that both factors matter.

These are only two of many factors behind Trump’s support, according to analysis of the Post-ABC data. Controlling for multiple demographic factors and the above attitudes, men, for instance, were significantly more likely to support Trump. White respondents without college degrees were also especially supportive, as were Republicans who do not identify as white evangelical Protestants. Republicans who felt “strongly” that the political system is dysfunctional were more likely to support Trump even when controlling for other factors.

Methodology

The Post-ABC poll was conducted March 3 to 6 among a random national sample of 1,000 adults reached on land-line and cellular phones. The survey interviewed 433 Republicans and GOP-leaning independents and 400 Republican-leaning registered voters.

With the data from the group of 433 participants, The Post used a statistical technique known as a logistic regression to estimate the chances that Republican and GOP-leaning independents with various traits would support Trump. The regression took into account demographic and political differences between participants, such their education, whether they were evangelical Protestants, the incomes they reported in the poll, whether they lived in rural, suburban or urban areas, whether they identified as Republican or independent and whether they described their political ideology as conservative, very conservative, moderate or liberal. The model also included age, sex and race as predictors of Trump’s support.

If respondents felt “strongly” that it is a bigger problem that whites are losing out to minorities, their responses  were coded as 1; if respondents said they felt “somewhat” that whites are losing out, their answers were coded as 0.5. The responses of those who said blacks or Hispanics losing out to whites was a bigger problem, or that both groups or neither group were losing out, were coded as 0. For the question on personal finances, respondents who said they are “struggling” to stay in their current economic class were coded as 1, while those who said they are “comfortable” or “moving up beyond” their class were coded as 0. The odds ratios reported above are the differences in the estimated chances that a respondent supports Trump based on a response of 0 versus a response of 1.

The Post tested the direct connection between whether Republicans-leaning respondents said that they were struggling economically and whether they said that whites are losing out to other groups, and found the correlation was weak. The coefficient of correlation was 0.1, and the upper bound of the p-value was 0.068.