What Google Knows About the Presidential Race

Political insight: people lie to pollsters, and probably on Facebook. But not to Google’s search bar

David Talbotarchive page

October 22, 2012

Lately I haven’t been telling the full truth to people calling my house asking about my political views. Why should I tell them anything? Of course, since I keep telling Massachusetts senator Scott Brown’s people that I haven’t made a final decision, they keep calling. This tends to tie them down. I feel bad about that.

People fib to poll takers and survey makers. They put their best face on Facebook. But by contrast, they tend to reveal their true colors in Google searches—and this can provide some insight into what will happen in the presidential race, as we were reminded in a New York Times piece yesterday by a Harvard PhD student, Seth Stephens-Davidowitz.

Four years ago, in October of 2008, search rates tended to predict high black turnout. Searches for voting information overall were slightly lower than they’d been in October 2004 but were higher in states with high black populations—North Carolina, Georgia, and Mississippi. While this isn’t surprising, it does show that Google searches can predict behavior.

The piece is worth reading for insights including this one: “Turnout might be expected to be higher in Ohio in 2012 than it was in 2004 or 2008.”

And: “Areas with the largest black populations are, on average, Googling for voting information at rates similar to those of 2008, rather than 2004, levels. By this metric, it does seem that pollsters should assume a black share of the electorate similar to that of 2008, when African-Americans made up an estimated 12 percent of the electorate, rather than 2004, when it was 11 percent—a good sign for Mr. Obama.

And: “There is nice news for Mitt Romney in the Google data, too: voting searches are higher in Idaho Falls and Salt Lake City, the two media markets with the largest Mormon populations. While neither Idaho nor Utah is a swing state, increased Mormon turnout might help Mr. Romney somewhat in two important swing states: Nevada (7 percent Mormon) and Colorado (3 percent Mormon).”

His overall conclusion: “Mr. Obama’s opponents hope that the 2012 electorate will be less favorable to Democrats, more like the 2004 electorate. My early analysis of Google search data says: don’t count on it.”