Across the United States, white Americans show subtle or "implicit" biases against blacks -- biases they mostly don't even realize they have. This has been established through lots of research, but it's not perfectly uniform across the land. Whites in some states show more bias, overall, than in others.
That's the takeaway from the map above, courtesy of Project Implicit, which is based on the scores of 1.5 million voluntary takers of the Implicit Association Test or IAT (which detects subtle or unconscious racial preferences), and which we examined earlier this month. But we didn't know why the map looked this way -- with levels of uncontrolled bias higher in the U.S. Southeast and East Coast (but not so much New England).
Dominic Packer, a psychologist at Lehigh University, has a surprising (and troubling) answer: Unconscious racial bias, he shows in a new analysis, is higher in U.S. states in which there is a higher ratio of black citizens to white citizens -- or in other words, in which there are relatively more black people for every white person.
“These findings are inconsistent with any simple hypothesis that contact between members of different racial groups will lead to reduced bias," wrote Packer in his analysis. However, he added, they are quite consistent with the idea that in states with more black citizens, whites may perceive "greater competition for political, cultural, and economic resources" or "greater risk for cross-race crime."
Packer provided this visualization, first recoloring the IAT scores map above to a new color scheme, and then showing a map using the same colors for the ratio of black to white residents in the state:
In a 2014 presentation, Project Implicit researchers had already suggested that this relationship might exist. Packer simply went further, seeking to uncover correlations between state level differences in IAT scores (for whites) and a number of sociological and demographic factors, which are known to vary by U.S. state. Factors that he considered included levels of income and income inequality, history as a slave-holding state, political ideology, and the ratio of white to black residents in the state.
Interestingly, the research found that how a state voted in 2008 -- for Obama versus McCain -- did not explain anything about the state's average IAT score. There was simply no relationship to politics. However, other factors, like levels of income inequality and whether the state was once a slave-holding state, did correlate with implicit bias scores.
Since many of the factors listed above actually correlate with each other, raising questions as to which factor might be truly primary, Packer then took another step. That was to perform a statistical analysis -- called a regression -- that would seek to even more precisely define which variable is linked to the pattern of implicit racial bias.
The result was striking: The ratio of white to black residents in a given state explained over 50 percent of the variability in various states' white participants’ Implicit Association Test scores. "That’s pretty big, for anything social scienc-y," said Packer of the result.
Or to state the result in a different way: “States where Whites outnumber Blacks substantially in the population have lower average IAT scores,” wrote Packer. “In contrast, states where Blacks make up proportionally more of the population have higher average IAT scores.”
So do these relatively small differences in average implicit bias scores, across U.S. states, actually matter? Packer thinks that they do, suggesting that they may predict actual behaviors. To show as much, he turned to a much-discussed 2012 state-by-state map of racist tweets, and analyzed that map in light of the distribution of implicit bias by whites.
Here, fascinatingly, Packer found that states with higher implicit bias scores had more racist tweets. Furthermore, he found an interaction between whether or not a state voted for Romney (versus Obama) and whether it produced a lot of racist tweets -- meaning that in pro-Romney states, a high state IAT score predicted lots of racist tweets, whereas in states that went for Obama, it didn't really predict anything:
Because Packer’s analysis was not peer reviewed or submitted to and published in a journal, we sought comments from other researchers on it. Generally they offered praise, but also noted a number of caveats.
David Amodio, a neuroscientist at New York University who studies prejudice, said that Packer's analyses "appear to be done well." But he also cautioned that correlation is still not causation -- so we still don’t necessarily know the fundamental reason for the quite strong correlation that Packer’s analysis revealed.
“There could be a million reasons for the observed correlation between IAT scores and the ratio of Blacks and Whites in a state,” said Amodio.
The University of Washington's Anthony Greenwald, who created the Implicit Association test in 1995, had another point of note. While observing that Packer’s analysis "went interestingly beyond what we already knew," he warned that the result might look different if the analysis was at the level of U.S. counties, rather than U.S. states. An analysis at the level of counties, he suggested, might reveal a rural-urban divide in whites’ implicit racial bias levels that a coarser state level analysis can’t tease out.
In particular, Greenwald singled out Washington, D.C., from Packer’s analysis -- an entirely urban “state” where blacks make up just under 50 percent of the population, but where white people's IAT scores were relatively low, in comparison with those in other states. "The finding that Washington, D.C., is a huge outlier is by itself an indicator that population ratio provides no simple explanation,” said Greenwald. “And it seems plausible that cities with relatively high Black:White ratios may have lower mean IAT scores for Whites than do surrounding rural areas with lower Black:White ratios.”
Packer agrees that doing the analysis at the level of counties, rather than states, would lead to interesting new insights -- but he doesn't think it would invalidate his overall conclusion. But such a county-level analysis, he said, is "really important" as well.
So there is still more research to do in order to further home in and refine explanations for the striking pattern shown in the map at the start of this article. Let’s hope that happens forthwith: Knowing our biases, and why they exist, is the first step toward correcting them.