The gallery at Jezebel prompted some geographers to create a map of all the racist tweets.
The enterprising folks at Floating Sheep used software they created called DOLLY to collect geocoded tweets for the week beginning November 1. In other words, it’s possible to search Twitter by both location and keyword (some other examples here). If I understand it correctly, the DOLLY software allows this search process to be further refined to get data at a more granular level.
What they came up with is a map that allows us to understand, at a glance, how these everyday acts of overt racism are spatially distributed in the U.S.
This is valuable work and just the kind of thing that I’d think sociologists would be interested in doing (but I digress, slightly). The methodology here, buried in the footnotes on the original post, is a worth exploring a little further.
The research questions they pose are: “Are racist tweets relatively evenly distributed? Or, do some states have higher specializations in racist tweets?”
To answer these questions, they sampled the universe of tweets. Specifically, they: “collected tweets that contained the text ‘monkey’ or ‘nigger’ AND also contain the text ‘Obama’ OR ‘reelected’ OR ‘won.’ A quick, and very unsettling, examination of the search results revealed that this indeed was a good match for our target of election-related hate speech. We end up with a total of 395 of some of the nastiest tweets you might possibly imagine. And given that we’re talking about the Internet, that is really saying something.”
Following that, they took the number of “hate tweets” by state and divided by the total number in the U.S., that became the numerator. Then, they got their denominator by doing the same for all the tweets in the state, divided by all the tweets in the U.S., which is easier to understand expressed as a formula:
(# of Hate Tweets in State / # of Hate Tweets in USA)
(# of ALL Tweets in State / # of ALL Tweets in USA)
Based on this, they assign a number, or a Location Quotient (LQ), for “Post Election Racist Tweets.” They then rank order states based on their LQ’s.
The results they end up with (Alabama, Mississippi, Georgia end up with 3 highest LQ scores) are less interesting than their map and clever methodology. In their analysis, the writers do well to note that the “prevalence of post-election racist tweets is not strictly a southern phenomenon,” but the ranking of the LQ scores by states makes the opposite case.
I want to suggest here that the problem is twofold:
the way the research question is posed and
the state-level of analysis.
The researchers here frame their question in terms of state boundaries and posit something of a false dichotomy between “even distribution” of racist tweets on the one hand, and, “states that specialize” in racist tweets on the other hand. As anyone who has taken an undergraduate methods class can tell you, this research question shapes the kind of data collection you do, and the analysis you come up with at the end.
The state-level of analysis here is something of a distraction. I understand that since we just went through a presidential election, people are thinking in terms of “states” – swing states, blue states, red states, who carried the state – but here, it makes less sense.
What I see when I look at this map are population centers. Take my home state, of Texas. The red dots there are clustered around places where there’s population density – Houston, Dallas, and more along the I-35 corridor. And, compare that to where I live now, on the East Coast. There are red dots all along the Northeast corridor of I-95. At a glance, it looks like racist tweets are not evenly distributed across the U.S. but are concentrated where white people live.
Again, let me say, I appreciate this work immensely, but I think that the state-level questions are the least interesting, and ultimately least revealing set of questions for mapping racism through digital media. Instead, I’d be interested in seeing some other basic demographic info about percentage of white people in the population and the proportion of racist tweets.
My guess is that the LQ is highest where there is the highest proportion of white people, but that, as academics are so fond of saying, is an empirical question worth investigating.
More in posts to come on calling out racism in digital media, and the growing backlash against it.