5 Questions with Seth I. Stephens-Davidowitz

FM sits down with someone who uses Google searches to predict election results.

Fifteen Minutes: You’re a PhD student in the economics department. When and how did you start thinking about race and politics?

Seth I. Stephens-Davidowitz: Really, I started thinking about Google data. And I just found Google data and was like, “Wow, you can measure pretty much anything with Google data: what people search, where they search it, when they search it.” I started thinking about Google data and then said, “What question can I ask that might be hard to answer otherwise?” And I’m kind of a political junkie, so it was in some sense natural that I’d focus on politics.

FM: Your research about the effects of racial prejudice on voting uses Google Insights, which is a relatively new tool. What does this mean as a research tool? Does it make things more accessible for political science research at Harvard and elsewhere?

SSD: It’s obvious that within a few years it’s going to be really leading to new insights on all kinds of questions, particularly the sensitive topics. Because people don’t like to admit certain things to surveys—you know, racism, drugs, sex, etc. And everybody admits everything to Google.

FM: So has your research changed the way you use Google for non-research purposes?

SSD: Yeah, I Google myself a lot more now that I’m in newspapers and stuff. Other than that, no. [Google searches are] all private and anonymous so [people who analyze them] have no clue. Nobody knows—so I’ve never been careful. I just type in whatever.

FM: You concluded that race actually does matter in electoral politics. How might racial prejudice influence the outcome of the 2012 election?

SSD: You always have to include the caveats. The future doesn’t resemble the past always. There’s no question, I think, when you look at even the early polling data. [Obama is] still struggling in many of the parts of the country where he was struggling before, certain parts of the South, West Virginia, etc., where racial animus is a big issue. He may be doing better in the Midwest than I would have expected based on the data, so that’s kind of interesting. We’ll see. I think [race] is there, it’s an issue, it’s a drag on [Obama’s] vote total in many parts of the country including some swing states. We’ll see exactly how many—if it’s as big of an effect, if it’s a little smaller of an effect. We have to wait and see the actual voting data.

FM: What will you be looking for before the election?

SSD: I tend to look at the Google data. You tend to get insights before polling becomes available. I knew right after Romney’s convention that it was a bust, because nobody was looking up “vote” or “voting” or “donate Romney,” whereas after every other convention these searches spike up. So in some sense you didn’t have to wait for the polling to know that that was a pretty lousy convention. Things like that. I might have some other projects in the line as well along similar veins.