President Donald Trump has a theory about how to overcome America's racial divides — and no, it doesn't involve him clearly and forthrightly condemning the violent white supremacist rallies being carried out in his name by avowed racists and neo-Nazis. It involves jobs.
"I really think jobs are going to have a big impact," he told reporters on Tuesday. "If we continue to create jobs — over a million — substantially more than a million, and you see just the other day, the car companies come in with Foxconn, I think if we continue to create jobs at levels that I'm creating jobs, I think that's going to have a tremendous impact — positive impact — on race relations."
In the context of Trump's others remarks at that press conference — which saw him empathizing with white nationalist rioters in Charlottesville, Virginia, and defending monuments to the Confederacy — this might sound reasonable. It's not a totally implausible theory, that the country becomes more tolerant during economic booms and that white Americans become more racially prejudiced during recessions or stagnation.
More from Vox:
Why Trump's Charlottesville crisis is shocking but not surprising
This expert on political violence thinks Trump is making neo-Nazi attacks more likely
I'm a black Southerner. I had to go abroad to see a statue celebrating black liberation.
But the evidence for the theory is mixed at best. In many cases, it's hard to see much correlation between objective economic conditions and the status of race relations.
"There's no easy answer to 'overcoming racial divides,' and there probably is no 'overcoming' them," Michael Tesler, professor of political science at UC Irvine and a leading expert on race and American politics, says. "The easiest way to mitigate them, though, would be for politicians of all stripes to acknowledge the realities of the historical and ongoing white supremacy in America, and vocally condemn them."
What we know about the economy and racism
One particularly compelling study on this question is a 1998 paper by political scientists Donald Green (Columbia), Jack Glaser (Berkeley), and Andrew Rich (now at the Truman Scholarship Foundation). It's infamously hard to measure white people's views on race, especially now that there's a significant taboo against overt, old-fashioned racism.
So instead, Green, Glaser, and Rich looked at data on the incidence of lynchings. Lynchings were a disturbingly common way through which white communities in the South expressed their racism. It would stand to reason that if racist sentiment ebbs and flows with the economy, then the rate of lynchings should as well.
They found little or no correlation between changes in real disposable gross national product per capita (a measure of economic well-being that allows comparisons back to the 19th century) and the number of lynchings. From 1882 to 1920, high cotton prices — a good proxy for economic well-being in the American South at the time — actually had a slightly positive relationship with lynching rates. That is, when the cotton economy was booming, lynchings, if anything, increased.
The inverse appeared to happen as the Great Depression hit. "Between 1930 and 1931, real per capita GNP declined by 8.5 percent, and yet lynchings dropped from 20 to 12," Green, Glaser, and Rich write. "The following year, real per capita GNP dropped again, this time by an astonishing 15.4 percent, and lynchings fell to 6." As the economy cratered, the most brutal manifestation of Southern racial prejudice grew less common.
Of course, that evidence is describing a previous era in American history. So Green, Glaser, and Rich extended the analysis to look at racist and homophobic hate crimes. Looking at monthly data from New York City, from 1987 to 1995, they found that unemployment rates had no real relationship with hate crime rates, regardless of hate crime type.
Some other studies looking at public opinion data have turned up similar results. Looking at 1998 to 2002 data from the General Social Survey, and matching respondents' answers with characteristics of their metro area or county, Penn State sociologist Marylee Taylor found that while the education level of an area's white population can help predict levels of expressed racial resentment or prejudice, the white population's economic condition had no effect.
"Education level remains a significant predictor for most racial attitude measures and white economic status is never significant," Taylor writes. "White education level is not a proxy for material hardship in the community: Limited education among white residents has a pronounced net effect on white racial attitudes; economic hardship has none."
Even more recent evidence, looking at the Obama era, confirms these general findings. The Obama Effect, a 2014 book by UMass's Seth K. Goldman and Penn's Diana Mutz, uses a panel survey following 20,000 people, who were interviewed as many as five times a year. That's an unusually rich source of information on how voters' views are changing.
Mutz and Goldman conclude that whites' racial prejudice toward African Americans actually declined from the summer of 2008 to Obama's inauguration. This was in spite of the fact that the US economy was in recession for that entire period, which saw the collapse of Lehman Brothers and the downturn's explosion into a full-on financial crisis. The effects wore off between Obama's inauguration and 2010, but the data fits into a broaderliterature showing that racial prejudice did not increase during the economic crisis.
"There was little change in aggregate levels of racial resentment, anti-black stereotypes, or anti-black affect from before to after the recession," UC Irvine's Tesler summarizes.
Trump could be driving up racial polarization further
What did increase, Tesler finds in his book Post-Racial or Most-Racial?: Race and Politics in the Obama Era, is the degree of partisan polarization based on race. There used to be relatively modest differences between Republicans and Democrats' stated views on racial controversies; the same share of each party disapproved of Bernhard Goetz, the white man who shot four black youths he claimed were mugging him back in 1984, and while 50 percent of Democrats agreed in 1995 that OJ Simpson was innocent, so did 41 percent of Republicans.
By contrast, 68 percent of Democrats disapproved of George Zimmerman's acquittal in the killing of Trayvon Martin in 2013, while only 20 percent of Republicans did. The pattern extends even to events that aren't outright clashes over racism — 53 percent of Democrats supported 12 Years a Slave's Oscar win for Best Picture in 2014, while only 15 percent of Republicans did. While in 1992, Democrats were 16 percentage points more likely than Republicans to say that the Confederate flag was racist, by 2016 they were 44 points more likely.
Tesler has argued that this kind of polarization had a particularly important effect on less-educated white voters. He has found that less-educated whites indicating low levels of racial resentment did not, in fact, flee the Democratic Party while Obama was president; only those with high levels of racial resentment, or who expressed the belief that discrimination against black Americans is relatively rare, did so. This was a group that Trump was able to court extremely successfully.
And Trump's position as president has the potential to make the situation worse. Public opinion researchers have long known that voters typically take their views on political questions from elected officials and other people they trust. UC Berkeley political scientist Gabriel Lenz's book Follow the Leader makes this point in great detail, and it's been powerfully illustrated in recent experimental work by Stanford's David Broockman and Washington University in St. Louis's Daniel Butler, who got a number of state legislators to randomly send out letters to constituents and found that receiving letters often led constituents to adopt their representative's opinion, even when they had disagreed before.
This has been a particularly important mechanism for racial equality. "One thing that has led to greater racial tolerance in the past is elite consensuses from both Democrats and Republicans on things like biological equality of the races, interracial marriage, and desegregation," Tesler says.
Trump is not questioning the consensus on those three issues — not yet, anyway. But he's questioned elite consensus about the evils of the Confederacy, and about the necessity of condemning white supremacists and neo-Nazis in strong terms. Given what we know about the ease with which voters take cues from politicians of their party, this has deeply concerning implications for public opinion. Charlottesville could be just the beginning.
Commentary by Dylan Matthews, a writer at Vox. Follow him on Twitter @dylanmatt.
For more insight from CNBC contributors, follow @CNBCopinion on Twitter.