Rather than parse survey data, he tried to measure white racism indirectly, by finding out where people had Googled various permutations of the N-word in the three years prior to 2008. The more such searches in a given area, the higher its "racially charged search" score.
After factoring in other variables, such as the vote share John Kerry received in 2004, Stephens-Davidowitz found that Obama's 2008 share of the vote falls as a given area's racially charged search score rises. Overall, he concluded, Obama suffered a net loss of between 2 and 4 percentage points in the national popular vote, compared to what he would have gotten if the whole country were as tolerant as the most racially tolerant areas.
I suppose you could quibble endlessly with Stephens-Davidowitz's methodology. But let's assume he's right. And let's further assume that Obama still carries a 2 to 4 point handicap.
I would still argue this is a glass-half-full story about race and racism. It's shameful that there are so many people in this country who are still racist, and who still judged Obama through that lens. It is — or should be — a source of consolation, however, that they were unable to change the course of events in 2008. The bad news: There's a lot of white rage out there. The good news: It's impotent rage.
Why do I say this? Presidential elections are not contests to win the national popular vote. They are contests for a majority of electoral votes, in 50 state elections, not all of which are equally competitive.
And the racially-charged searches were not evenly distributed across the country. The most racist pockets of opinion that Stephens-Davidowitz found were in West Virginia, upstate New York, rural Illinois, eastern Ohio, southern Mississippi, western Pennsylvania, and southern Oklahoma.
Except Ohio, which had 20 electoral votes in 2008, and which Obama carried with 51.5 percent of the vote, none of these were swing states. Obama rolled up huge victory margins in NY, IL and PA, just as previous Democratic candidates had. And he got clobbered in MS, OK and WV, just like Kerry and Al Gore did.
Among 2008's swing states, only Missouri presents a case in which the final result — a 3,900-vote margin for John McCain — was both unfavorable to Obama and within the 2 to 4 percentage points range of lost votes suggested by Stephens-Davidowitz. At most, then, racism as measured by Stephens-Davidowitz cost Obama Missouri's 10 electoral votes in 2008.
To be sure, if Obama is still suffering from a 2 to 4 percentage point racial handicap in 2012, and if racial hostility remains concentrated in the same states, then racism puts him at risk of losing the states he carried by fewer than four points in 2008: OH (18 electoral votes this time), NC (15), FL (29), IN (11), VA (13), CO (9). That would cost him 95 electoral votes of the 365 he won in 2008, but he would still squeak through to re-election with 270 electoral votes.
This particular scenario does not seem especially likely, though. Stephens-Davidowitz does not provide a state-by-state estimate for Obama's net popular vote loss due to racism. But he does rank the states in order of highest racially charged search scores to lowest. On that list, CO ranks 49th — suggesting that if Obama does lose the Rocky Mountain State, it won't be due to racial animosity.
Of the top 25 (or, if you prefer, worst 25) states, only four, OH, NV, NC, and FL, are considered "in play" in 2008. You could say that MI is in play as well, and that race might imperil Obama there; MI ranks 6th on Stephens-Davidowitz's list, though Obama carried it by 16 points in 2008. But Mitt Romney's home-state attachments may be a confounding variable.
Economic conditions do not favor Obama as they did in 2008. The wind was at his back in 2008; it will be in his face this time. Much else has changed in the last four years as well: the nature of the Republican coalition, the persona of Obama's opponent, the international situation. And Obama is running on his record as an incumbent, not as a challenger trying to make history. If unemployment were even one percentage point lower than it is today, we might not be having this discussion.
The next study of the relationship between racially-charged Google searches and Obama's vote share will have to find ways to measure those variables, and to factor them into any analysis of the connection between his race, white racism and his performance in 2012 relative to 2008.