Racial Bias, Even When We Have Good Intentions
By Sendhil Mullainathan
The deaths of African-Americans at the hands of the police in Ferguson, Mo., in Cleveland and on Staten Island have reignited a debate about race. Some argue that these events are isolated and that racism is a thing of the past. Others contend that they are merely the tip of the iceberg, highlighting that skin color still has a huge effect on how people are treated.
Arguments about race are often heated and anecdotal. As a social scientist, I naturally turn to empirical research for answers. As it turns out, an impressive body of research spanning decades addresses just these issues — and leads to some uncomfortable conclusions and makes us look at this debate from a different angle.
The central challenge of such research is isolating the effect of race from other factors. For example, we know African-Americans earn less income, on average, than whites. Maybe that is evidence that employers discriminate against them. But maybe not. We also know African-Americans tend to be stuck in neighborhoods with worse schools, and perhaps that — and not race directly — explains the wage gap. If so, perhaps policy should focus on place rather than race, as some argue.
But we can isolate the effect of race to some degree. A study I conducted in 2003 with Marianne Bertrand, an economist at the University of Chicago, illustrates how. We mailed thousands of résumés to employers with job openings and measured which ones were selected for callbacks for interviews. But before sending them, we randomly used stereotypically African-American names (such as “Jamal”) on some and stereotypically white names (like “Brendan”) on others.
The same résumé was roughly 50 percent more likely to result in callback for an interview if it had a “white” name. Because the résumés were statistically identical, any differences in outcomes could be attributed only to the factor we manipulated: the names.
Other studies have also examined race and employment. In a 2009 study, Devah Pager, Bruce Western and Bart Bonikowski, all now sociologists at Harvard, sent actual people to apply for low-wage jobs. They were given identical résumés and similar interview training. Their sobering finding was that African-American applicants with no criminal record were offered jobs at a rate as low as white applicants who had criminal records.
These kinds of methods have been used in a variety of research, especially in the last 20 years. Here are just some of the general findings:
• When doctors were shown patient histories and asked to make judgments about heart disease, they were much less likely to recommend cardiac catheterization (a helpful procedure) to black patients — even when their medical files were statistically identical to those of white patients.
• When whites and blacks were sent to bargain for a used car, blacks were offered initial prices roughly $700 higher, and they received far smaller concessions.
• Several studies found that sending emails with stereotypically black names in response to apartment-rental ads on Craigslist elicited fewer responses than sending ones with white names. A regularly repeated study by the federal Department of Housing and Urban Development sent African-Americans and whites to look at apartments and found that African-Americans were shown fewer apartments to rent and houses for sale.
• White state legislators were found to be less likely to respond to constituents with African-American names. This was true of legislators in both political parties.
• Emails sent to faculty members at universities, asking to talk about research opportunities, were more likely to get a reply if a stereotypically white name was used.
• Even eBay auctions were not immune. When iPods were auctioned on eBay, researchers randomly varied the skin color on the hand holding the iPod. A white hand holding the iPod received 21 percent more offers than a black hand.
The criminal justice system — the focus of current debates — is harder to examine this way. One study, though, found a clever method. The pools of people from which jurors are chosen are effectively random. Analyzing this natural experiment revealed that an all-white jury was 16 percentage points more likely to convict a black defendant than a white one, but when a jury had one black member, it convicted both at the same rate.
I could go on, but hopefully the sheer breadth of these findings impresses you, as it did me.
There are some counterexamples: Data show that some places, like elite colleges, most likely do favor minority applicants. But this evidence underlies that a helping hand in one area does not preclude harmful shoves in many other areas, including ignored résumés, unhelpful faculty members and reluctant landlords.
But this widespread discrimination is not necessarily a sign of widespread conscious prejudice.
When our own résumé study came out, many human-resources managers told us they were stunned. They prized creating diversity in their companies, yet here was evidence that they were doing anything but. How was that possible?
To use the language of the psychologist Daniel Kahneman, we think both fast and slow. When deciding what iPod to buy or which résumé to pursue, we weigh a few factors deliberately (“slow”). But for hundreds of other factors, we must rely on intuitive judgment — and we weigh these unconsciously (“fast”).
Even if, in our slow thinking, we work to avoid discrimination, it can easily creep into our fast thinking. Our snap judgments rely on all the associations we have — from fictional television shows to news reports. They use stereotypes, both the accurate and the inaccurate, both those we would want to use and ones we find repulsive.
We can’t articulate why one seller’s iPod photograph looks better; dozens of factors shape this snap judgment — and we might often be distraught to realize some of them. If we could make a slower, deliberate judgment we would use some of these factors (such as the quality of the photo), but ignore others (such as the color of the hand holding the iPod). But many factors escape our consciousness.
This kind of discrimination — crisply articulated in a 1995 article by the psychologists Mahzarin Banaji of Harvard and Anthony Greenwald of the University of Washington — has been studied by dozens of researchers who have documented implicit bias outside of our awareness.
The key to “fast thinking” discrimination is that we all share it. Good intentions do not guarantee immunity. One study published in 2007 asked subjects in a video-game simulation to shoot at people who were holding a gun. (Some were criminals; some were innocent bystanders.) African-Americans were shot at a higher rate, even those who were not holding guns.
Ugly pockets of conscious bigotry remain in this country, but most discrimination is more insidious. The urge to find and call out the bigot is powerful, and doing so is satisfying. But it is also a way to let ourselves off the hook. Rather than point fingers outward, we should look inward — and examine how, despite best intentions, we discriminate in ways big and small.
Sendhil Mullainathan is a professor of economics at Harvard. Follow him on Twitter at @m_sendhil.