It may be easy to talk a good game when it comes to race, but a new study from Pew Research Center reveals the hidden racial biases that people carry. To get at the truth, researchers used an Implicit Association Test (IAT), which tracks how quickly people associate good or bad words with specific groups. The study focused on black, Asian, white, biracial black-white, and biracial Asian-white participants, with an eye toward discovering if biracial people are less likely to be biased. It turns out that they are not.

The experiment was run in two parts. In the first part, black, white and black-white biracial people were shown images on a computer with photos of either a black or white man, and a word that means good or bad. They are asked to—as quickly as they can—press a button on the keyboard that either agrees or disagrees with using the word as an accurate descriptor. The same was repeated with a group of Asian, white and Asian-white biracial people.

The results showed that most of those polled had some racial preference. Among black-only respondents, just 26 percent exhibited no or negligible amounts of bias. For Asian-only people, that number was 20 percent. For whites in the first group, the number was 27 percent. Among whites in the first test, that number was 27 percent; it was 30 percent for the second group.

But most people exhibited a clear bias, even those who identify with more than one racial group. Forty-eight percent of people who identify as white-only favored whites over blacks; 50 percent of them preferred whites over Asians. For black-white biracial people, 42 percent preferred whites, while 35 percent preferred blacks. And when it comes to black people, 45 percent preferred blacks over whites, while 29 percent were more likely to associate positive things with white people. 

In the second group, 42 percent of Asians preferred Asians, while 38 percent preferred whites. Among biracial Asian-white people, 39 percent preferred whites, versus 38 percent who preferred Asians. 

In general, the numbers stayed the same even when divided by age (39 and younger, versus 40 and up), gender, education and political party. And when the same people were polled for explicit bias, they were less likely to reveal it.
 

Previous IATs have shown that the biases they uncover are associated with actions. For example, a 2007 study found that doctors’ IAT scores were reflected in the treatment plans they created for their patients.