At a time of increasing tension over President Donald Trump’s rhetoric on undocumented immigrants and people of color, several Democrats have charged that Trump’s language is encouraging hate crimes. To make their case, Democrats cite a startling statistic.
"In 2016, counties hosting a Trump rally saw a 226% spike in hate crimes," reads a post on Facebook from the presidential campaign of Bernie Sanders.
A tweet from Rep. Ilhan Omar, D-Minn., makes the same claim.
A reader saw the Sanders post on social media and sent it to us, asking whether it was accurate.
We found that the statistic comes from an academic paper. But the Sanders post makes the increase seem dramatic and causal. The findings are actually much more nuanced.
The post cites the Washington Post as its source, and we tracked that down to a post on one of the newspaper’s blogs, the Monkey Cage.
On March 22, the blog published a post titled, "Counties that hosted a 2016 Trump rally saw a 226% increase in hate crimes."
The post summarized a paper written by three academics: Ayal Feinberg, an assistant professor of political science at Texas A&M University-Commerce; Regina Branton, a professor of political science at the University of North Texas; and Valerie Martinez-Ebers, a professor of political science at the University of North Texas. (Sanders’ campaign did not respond to an inquiry.)
The authors told PolitiFact that the paper is currently under peer review for publication in an academic publication. The authors presented it in 2019 at meetings of both the Southern Political Science Association and the Western Political Science Association.
Here’s how the authors summarized their work in the Post:
Using the Anti-Defamation League’s Hate, Extremism, Anti-Semitism, Terrorism map data (HEAT map), we examined whether there was a correlation between the counties that hosted one of Trump’s 275 presidential campaign rallies in 2016 and increased incidents of hate crimes in subsequent months.
To test this, we aggregated hate-crime incident data and Trump rally data to the county level and then used statistical tools to estimate a rally’s impact. We included controls for factors such as the county’s crime rates, its number of active hate groups, its minority populations, its percentage with college educations, its location in the country and the month when the rallies occurred.
We found that counties that had hosted a 2016 Trump campaign rally saw a 226% increase in reported hate crimes over comparable counties that did not host such a rally.
In their posts, Sanders (and Omar) referenced the headline in the Monkey Cage blog post almost verbatim. But there’s a discrepancy between the headline’s phrasing and what the researchers’ academic paper looked at.
The phrasing used by Sanders implies that hate crimes spiked in one jurisdiction, jumping from one level before the rally to a rate three times higher afterward. That’s not the case. As the study summary says, the 226% difference compares Trump-rally counties to non-Trump rally counties with similar demographics.
So a more accurate phrasing might be: "Counties that hosted a 2016 Trump rally had 226% more hate crimes than counties that did not."
There are other nuances that aren’t captured by the statistic Sanders posted, experts told PolitiFact.
In an interview, Feinberg, one of the coauthors of the study cited in the Post article, noted a few general challenges facing the field of hate-crime studies.
First, he said, there are limitations in "how extremist incidents, events motivated by hate, and hate crimes lack a cohesive definition" under the law. What constitutes a hate crime in one state may not in another. Also, some of the data is self-reported, making it hard to tell how many hate crimes in a given jurisdiction are unreported.
In addition, Feinberg said, "not all police departments investigate hate crimes or perceive incidents motivated by prejudice" with the same degree of aggressiveness.
Feinberg said that these caveats are always top of mind for hate-crimes researchers.
He added that a 2018 research project he collaborated on did not find that the inclusion of far-right parties in governing coalitions in Europe had an effect on reported anti-Semitic incidents. So when starting work on the Trump paper, he came in to the research with an open mind.
Before undertaking the study, he said, "I was not sure that Trump rhetoric would have any impact on reported extremist events, hate incidents, anti-Semitism, or hate crimes."
Meanwhile, Brian Levin, Kevin Grisham, and John Reitzel, three of the contributors to a California State University-San Bernardino study of hate crime trends in recent years, told PolitiFact that it’s hard to know whether other factors could have played a role in boosting hate crimes in the counties with a Trump rally.
"It is hard to untangle all these relationships without some interviews and in-depth qualitative data," Grisham said. "I think the data is fascinating and may be the first step in a larger and more in-depth project. But, we cannot say with absolute certainty that hate follows the rallies."
For instance, Grisham said, it could be that the rallies may be a contributing cause rather than a dominant one. "Anyone using that data should be aware of the complexity of extremist actions and their connections to rhetoric. Like anything with human beings, nothing is black and white."
Another potential problem: A 226% difference sounds massive, but in many locales, that may mean only a small numerical difference, especially the types of low-population counties that tended to host Trump rallies. The difference between one hate crime and three is 200%, but with numbers like that, a small difference in the reporting of crimes or police investigations could cloud the question of whether Trump’s rallies had an effect.
There is "statistical noise" when raw numbers are small, Reitzel said. Focusing on percentage changes, "even where the numbers are true, can still distort interpretation," he said.
Overall, the study is "certainly interesting and suggestive," Levin said. "But I would have preferred to see a comparison that’s apples to apples."
Sanders’ post said, "In 2016, counties hosting a Trump rally saw a 226% spike in hate crimes."
This is from an academic study. However, it’s worth noting some caveats -- that the data can be subject to "statistical noise" and jurisdictional differences in hate crime definitions and police aggressiveness, and that cause and effect are hard to pinpoint.
Perhaps most important is that Sanders’ wording implies that the 226% jump stems from comparing hate crimes before and after a Trump rally within the same county, when in fact it’s a comparison from Trump rally counties to similar counties that did not host a Trump rally.
The statement is partially accurate but leaves out important details, so we rate it Half True.