I Asked Google And Google Said....

Charlottesville, Virginia

July 25, 2020

What areas of the country are the "most racist?" The answer is suggested in a new study just published in PLOS ONE.

Before anyone gets all bothered by this study, clearly there are flaws in this methodology. (I won't get into the fact that it is scary that Big Brother collects data that allows this type of analysis.). I'll note that the article doesn't provide any detail about the categories - how much is "much more" compared to "more than average." And there is an obvious assumption in the study that there is racism. But it is interesting that the results compare well with a similar analysis of racist tweets. (I agree, what is the definition a "racist tweet?")

The paper introduces a novel but makes-tons-of-sense-when-you-think-about-it method for measuring the incidence of racist attitudes: Google search data. The methodology comes from data scientist Seth Stephens-Davidowitz. He's used it before to measure the effect of racist attitudes on Barack Obama's electoral prospects.

"Google data, evidence suggests, are unlikely to suffer from major social censoring," Stephens-Davidowitz wrote in a previous paper. "Google searchers are online and likely alone, both of which make it easier to express socially taboo thoughts. Individuals, indeed, note that they are unusually forthcoming with Google." He also notes that the Google measure correlates strongly with other standard measures social science.

This is important, because racism is a notoriously tricky thing to measure. Traditional survey methods don't really work -- if you flat-out ask someone if they're racist, they will simply tell you no. That's partly because most racism in society today operates at the subconscious level, or gets vented anonymously online.

For the PLOS ONE paper, researchers looked at searches containing the N-word. People search frequently for it, roughly as often as searches for "migraine(s)," "economist," "sweater," "Daily Show," and "Lakers." (The authors attempted to control for variants of the N-word not necessarily intended as pejoratives, excluding the "a" version of the word that analysis revealed was often used "in different contexts compared to searches of the term ending in '-er'.")

It's also important to note that not all people searching for the N-word are motivated by racism, and that not all racists search for that word, either. But aggregated over several years and several million searches, the data give a pretty good approximation of where a particular type of racist attitude is the strongest.

Interestingly, on the map above the most concentrated cluster of racist searches happened not in the South, but rather along the spine of the Appalachians running from Georgia all the way up to New York and southern Vermont.

Other hotbeds of racist searches appear in areas of the Gulf Coast, Michigan's Upper Peninsula, and a large portion of Ohio. But the searches get rarer the further West you go. West of Texas, no region falls into the "much more than average" category. This map follows the general contours of a map of racist Tweets made by researchers at Humboldt State University.