There are all sorts of reasons a voter may pick a candidate for president. Maybe the voter prefers the candidate’s policies, likes how the candidate speaks, or thinks that the candidate has more applicable experience or qualifications for the job. But a new study shows there’s another reason a voter might pick one candidate over another: race.
To measure this, Evan Soltas of the University of Oxford and David Broockman of Stanford University cleverly took advantage of a real-world experiment.
Every four years in Illinois’s Republican presidential primary, GOP voters select delegates who represent a specific candidate — and these delegates then go on to vote for the candidate they represent in the Republican convention, where the official candidate for president is named by the party. The delegate and candidate’s names are clearly printed on the ballot as voters make their pick. (All delegate systems generally work in a similar way.)
Soltas and Broockman asked a simple question: If delegates had names that implied they’re white, were they more likely to get votes than delegates who had names that implied they’re nonwhite? To answer this, they looked at the Republican presidential primary elections from 2000 to 2016, covering nearly 2,400 delegates and 22.3 million votes, with data taken from the Illinois State Board of Elections.
The result: Yes, delegates with stereotypically white names got more votes than those with stereotypically nonwhite names. The researchers found that nonwhite delegates got 10.5 percent fewer votes. To put that in context, Soltas and Broockman estimated the number of nonwhite candidates in Illinois would increase by 20 percent if this kind of discrimination wasn’t taking place.
“Indeed, we show that in several cases it appears very likely that voter taste-based discrimination against a presidential candidates’ nonwhite delegates reduced their vote totals sufficiently that delegates for other presidential candidates won and served instead,” the researchers concluded.
In other words, a delegate’s race was a substantial factor in whether a candidate got enough votes — so much so that it may have swung some of the delegate elections.
The researchers took advantage of a real-world experiment
The researchers went through several methodological checks to make sure the results were reflecting racial discrimination, not some other factor. But generally, the findings are credible because voters have essentially no other plausible reason beyond race to vote against a delegate who represents their candidate.
Here’s how it works: Let’s say there are six names on the ballot, with two candidates — Donald Trump and Jeb Bush — competing against one another. The delegates picked to vote for Trump are Tom, Bob, and Ahmed. The delegates picked to vote for Bush are John, Jane, and Harry.
These delegates, to be clear, are totally powerless. They don’t make any policy decisions. All they do is vote for the candidate they’re representing at the convention, and that vote is tallied up with the hundreds of others to pick the Republican nominee for president. They are required under convention rules to vote for the candidate they represent, or their vote is simply rescinded and the candidate they represent gets an extra vote to make up for the loss.
So really, voters are only picking their preferred presidential candidate. The delegates are just stand-ins for the formal process of picking that presidential candidate.
Yet the delegates’ names — to the extent they signaled a delegate’s race — seemed to affect the outcomes. So John, Jane, and Harry get the full backing of Bush supporters, and Tom and Bob get the full backing of Trump supporters. But Ahmed, one of three delegates for Trump, ends up lagging behind, getting fewer votes even from Trump’s backers.
Delegates’ race influenced the election results, even after controlling for other variables
Soltas and Broockman explained: “Examining the results by delegate race, we find clear evidence of discrimination against delegates whose names signal they are Hispanic, East Asian, Middle Eastern, or Indian. Our point estimates for discrimination against black delegates are similar, although fewer names clearly signal that the delegates are black, making our estimates of anti-black taste-based discrimination less precise.”
Notably, there was no apparent discrimination based on a delegate’s gender.
The researchers used three measures to gauge the potential racial signals that a delegate’s name sends: data from the US Census to match last names with their racial backgrounds, anthropological data for the etymology of full names, and a survey of workers on Amazon Mechanical Turk, a pay-for-work service, about what delegates’ names say about their race.
The results were not seemingly affected by order of appearance on a ballot. They also weren’t limited to certain election years, the actual candidates behind the delegates (although nonwhite delegates for nonwhite candidates, like Marco Rubio, appeared to face lower levels of discrimination), or geographic location. And the findings held up when the researchers controlled for other attributes related to the delegates, such as whether a delegate previously held office and therefore might be known to some of the public.
The major caveat to this study is it was only for Republican primary voters in Illinois. Although the researchers tracked several measures — number of hate crimes, racially charged Google search, and scores on implicit bias tests — that indicate Illinois isn’t more or less racist than other states, it’s possible that something about this single pool of voters influenced the results. Maybe other states would be more or less likely to discriminate. Or maybe the results would be different for other types of elections, like general elections in which minority candidates may be able to overcome racial barriers with better policy ideas or political messaging than their opponents. We just don’t know.
The bottom line: Racism influences elections
Whether the estimate of 10.5 percent fewer votes for nonwhite delegates applies across the board or not, the general finding is clear: The race of the person on the ballot plays a significant role in elections.
So even if minority politicians overcomes structural barriers — a higher likelihood of poverty, for example — to run for office, they will still face additional barriers at the ballot box.
Soltas and Broockman don’t propose specific policy solutions to close the gap. But the research generally suggests that to reduce people’s racial biases, personal, empathetic tactics are most effective — those that attempt to root out racial biases by getting people to think of how they would feel if they were in a minority group’s position. (For more on this research, read my in-depth explainer.)
One study co-authored by Broockman, for example, found that canvassing people’s homes and having a 10-minute, nonconfrontational conversation about transgender rights — in which people’s lived experiences were relayed so they could understand how prejudice feels personally — managed to reduce voters’ anti-trans attitudes for at least three months. Perhaps a similar model could be adapted to reach out to people with racist, sexist, or other deplorable views, although this idea needs more study.
To give minority candidates a better chance, Soltas and Broockman’s Illinois study suggests we’ll need more of this kind of empathetic, racism-breaking outreach in the future.