The media loves to say Trump's support is fueled by economic anxiety. But race, religion, and immigration are far more closely tied to his support.

You can ask just one simple question to find out whether someone likes Donald Trump more than Hillary Clinton: Is Barack Obama a Muslim? If they are white and the answer is yes, 89 percent of the time that person will have a higher opinion of Trump than Clinton.

That’s more accurate than asking people if it’s harder to move up the income ladder than it was for their parents (54 percent), whether they oppose trade deals (66 percent), or if they think the economy is worse now than last year (81 percent). It’s even more accurate than asking them if they are Republican (87 percent).

Those results come from the 2016 American National Election Study (ANES) pilot survey. My analysis indicates that economic status and attitudes do little to explain support for Donald Trump.

These results might be rather surprising since most political commentators have sought to root Trump’s appeal in the economic anxieties of working-class whites. As George Packer recently wrote in the New Yorker:

The base of the [Republican] Party, the middle-aged white working class, has suffered at least as much as any demographic group because of globalization, low-wage immigrant labor, and free trade. Trump sensed the rage that flared from this pain and made it the fuel of his campaign.

Other analysts, however, have found that support for Trump is rooted in animosity and resentment toward various minority groups, especially African Americans,immigrants, and Muslims.

So which is the key factor in explaining Trump support? I believe it is the latter, based on my analysis of the data from the 2016 ANES.

How I determined Trump support isn't about the economy

Here’s how I examined the 2016 ANES pilot survey, which includes a number of questions on economic attitudes as well as attitudes toward race, religion, and immigration.

I compared feeling thermometer evaluations for Donald Trump and those for Hillary Clinton. These evaluations run from 100 (most positive) to 0 (most negative). By looking at the difference in these evaluations, voters were ranked from 100 (most positive toward Clinton, most negative toward Trump) and -100 (most negative toward Clinton, most positive toward Trump). Those in the middle (a score of zero) were equally positive (or negative) toward the two candidates.

The first was a variable measuring how optimistic or pessimistic respondents were about economic opportunity in the US consisting of the combined results of these two questions:

Do you think people’s ability to improve their financial well-being is now better, worse, or the same as it was 20 years ago?

Compared with your parents, do you think it is easier, harder, or neither easier nor harder for you to move up the income ladder?

I also included a variable that asked whether respondents believed the nation's economy is now better, about the same, or worse than it was a year ago. Since Trump has made opposition to free trade agreements a key part of his campaign message, another variable measured whether respondents favor or oppose the US making free trade agreements with other countries.

To measure attitudes toward different races, religions, and immigrant groups, I included a variable measuring racial resentment, which combines the responses to the following statements:

Over the past few years, black people have gotten less than they deserve.

Irish, Italian, Jewish, and many other minorities overcame prejudice and worked their way up. Black people should do the same without any special favors.

It's really a matter of some people not trying hard enough; if black people would only try harder they could be just as well off as white people.

Generations of slavery and discrimination have created conditions that make it difficult for black people to work their way out of the lower class.

Other variables include whether respondents believe that President Obama is a Muslim, whether the number of legal immigrants should be increased or decreased, and how well respondents think the word "violent" describes Muslims.

Here's what I found

Next I ran a regression model to see how much impact the different variables had. The results of the model are shown in the graph below. The dots represent the impact that variable has on support for Clinton versus Trump.

Since all variables have been scaled from 0 to 1, the location of the dots shows the impact of going from the lowest to highest level of the variable. Dots on the negative or left side of the line indicate that respondents become more favorable toward Trump as they move from lower to higher levels for the variable. Dots on the right or positive side of the graph indicate that they become more favorable to Hillary Clinton.

For example, the result for the party identification variable is -57, indicating that Republicans (1) were, on average, 57 points more favorable to Trump than Clinton compared with strong Democrats (0). Similarly, the older the voter, the more likely he is to support Hillary Clinton. The lines intersecting the dots represent the 95 percent confidence intervals. When the lines touch zero, the result is not statistically significant.

Racial resentment is a more important predictor of Trump support than economic insecurity. (Josh Rosenblat/Vox)

The graph indicates that neither income nor economic pessimism has a statistically significant impact on evaluations of Clinton versus Trump. On the other hand, those who think the economy is worse now than a year ago and those opposed to free trade agreements are more likely to support Trump, and these results are statistically significant.

THOSE WHO EXPRESS MORE RESENTMENT TOWARD AFRICAN AMERICANS, THOSE WHO THINK THE WORD "VIOLENT" DESCRIBES MUSLIMS WELL, AND THOSE WHO BELIEVE PRESIDENT OBAMA IS A MUSLIM HAVE MUCH MORE POSITIVE ATTITUDES OF TRUMP COMPARED WITH CLINTON

The graph also indicates that several of the racial and religious views have a significant impact on evaluations of the candidates. Those who express more resentment toward African Americans, those who think the word "violent" describes Muslims well, and those who believe President Obama is a Muslim have much more positive views of Trump compared with Clinton.

Attitudes toward increasing or decreasing legal immigration, however, do not have a statistically significant impact, a rather surprising finding given Trump’s attacks on undocumented immigrants and his promise to build a border wall.

Furthermore, the impact of these variables is generally greater than that of the economic variables. For, example, moving from the least to the most resentful view of African Americans increases support for Trump by 44 points, those who think Obama is a Muslim (54 percent of all Republicans) are 24 points more favorable to Trump, and those who think the word "violent" describes Muslims extremely well are about 13 points more pro-Trump than those who think it doesn’t describe them well at all.

This compares with an 11-point difference between those who are most opposed to free trade deals and those who are most in favor, and a 23-point gap between those who think the economy had gotten much better and those who think it had gotten much worse in the previous year.

At best, one can argue that Trump supporters are much more negative about the direction of the economy over the past year, but even this opinion may not represent actual economic dissatisfaction. In the year before the survey was administered (January 2015 to January 2016), unemployment fell from 5.7 to 4.9 percent, real GDP grew by 2.4 percent, inflation was almost nonexistent (0.1 percent), andpersonal income increased by 4.4 percent. Perhaps the economy hasn’t been booming, but there seems to be little evidence to support the contention shared by more than half of all Republicans that the economy got worse during 2015.

More likely, as political scientist John Sides has pointed out, evaluations of the economy have become a proxy for partisan and political preferences. And to the extent that opinions about President Obama influence evaluations of the economy, opinions about him are themselves strongly tied to opinions about race, immigration, and religion.

Indeed, Obama embodies of each of these concerns for a large segment of Republicans. He is obviously African-American; as mentioned previously, more than half of all Republicans believe he is a Muslim; and finally, only 29 percent of Republicans think he was born in the US. Thus, it’s no surprise that, as the graph below shows, opinions about Obama are heavily influenced by levels of racial resentment.

(Josh Rosenblat/Vox)

Rerunning the original regression model to control for whether respondents approve or disapprove of the job President Obama is doing produces the following results. Not surprisingly, the more strongly someone approves of President Obama, the more positive their evaluation of Hillary Clinton over Donald Trump.

RESENTMENT TOWARD AFRICAN AMERICANS, THE BELIEF THAT MUSLIMS ARE VIOLENT, AND THE PERCEPTION THAT PRESIDENT OBAMA IS A MUSLIM ALL REMAIN SIGNIFICANT EVEN WHEN CONTROLLING FOR ATTITUDES TOWARD OBAMA’S JOB APPROVAL

More interestingly, with this variable included, attitudes toward free trade and evaluations of the economy are no longer statistically significant. This suggests that Trump supporters’ opinions about the direction of the economy are less an objective evaluation of actual conditions than a chance to register disapproval of a president they strongly dislike.

At the same time, resentment toward African Americans, the belief that Muslims are violent, and the perception that President Obama is a Muslim all remain significant even when controlling for attitudes toward Obama’s job approval.

While there’s little evidence of a direct relationship between economic attitudes and support for Trump, there may be an indirect one in which economic pessimism combines with resentment against various out-groups, so that those who believe the economy is stacked against them are more likely to scapegoat black people, or immigrants, or Muslims for their problems. As political scientist Robert Putnam put it recently, perhaps economic deprivation has provided "dry tinder" for Trump’s inflammatory appeals to race and xenophobia.

To test this possibility, I reran the model to include an interaction between racial resentment and economic pessimism. The graph below shows no support for the "dry tinder" thesis. Racial resentment had no extra effect among those who were more pessimistic about the economy. If anything, racial resentment mattered more among those who were less pessimistic, but the difference is trivial.

Graphic by Vox's Josh Rosenblat

Bottom line: It's not the economy, stupid

Given these results, political analysts need to understand that Trump’s appeal has little to do with economics.

To paraphrase what Marco Rubio once famously said (over and over and over), "let’s dispel with this fiction" that Donald Trump is appealing to the economic anxieties of Americans. Instead, attitudes about race, religion, and immigration trump (pun intended) economics.

Philip Klinkner is the James S. Sherman Professor of Government at Hamilton College in Clinton, NY. He is the author (with Rogers Smith) of The Unsteady March: The Rise and Decline of Racial Equality in America.

Men shouldn't fear women. Trump should.