In the new issue of Bloomberg Businessweek, I have a short feature about how Google’s executive chairman, Eric Schmidt, is investing millions in a new company with Dan Wagner and two dozen other veterans of the Obama campaign’s data analytics team. Schmidt was, not surprisingly, full of praise for the engineers, statisticians, and scientists whose work informed the strategy of a campaign that won by 5 million votes. But given the collaborative effort and many moving parts of a modern presidential campaign, it can be difficult to isolate the data team’s contribution. One great example, though, is the intricate mathematical models of swing states that Wagner and colleagues built that were meant to offer an alternative glimpse of the state of the race to public polls and even the campaign’s internal polls.
As David Plouffe, then a senior White House adviser, explained in my story, the data team’s models proved to be much steadier and more accurate than even the traditional tracking polls the campaign was also conducting. A number of Obama vets repeated this claim to me, so I asked them to provide some evidence to back it up, and they did. Here, for the first time, is a chart based on internal data that shows how the Obama campaign’s swing state model performed against the much maligned Gallup poll over the last several months of the race. This was the campaign’s daily “horserace” projection of the outcome, based on a nightly survey of 10,000 people.