By Chris Wilson
In the last few days before the election, most pollsters reported that somewhere around 4 percent of voters remained undecided between President Barack Obama and Gov. Mitt Romney. Campaigns and outside groups spent roughly a billion dollars identifying and converting these equivocating voters in the home stretch of the campaign, both through relentless microtargeting and relentless macrotargeting--that is, suffocating the airwaves with a final cannonade of ads.
The best available profiles of undecided voters suggest that they were not so much the civically challenged simians on Saturday Night Live as a hyper-disenchanted sliver of the electorate unsure whether their vote was even worth deciding at all.
Research from Yahoo Labs now has identified a clue as to what was on the minds of these holdout swing voters. In the past few weeks, scientists in the Barcelona office have analyzed millions of anonymous queries entered in the Yahoo search engine during the final day of the campaign, focusing on those that mention both Obama and Romney. This turns up a curious subspecies of the American electorate: Those who felt it was worth deciding whom to vote for, but not until the last 24 of the 5,040 hours that elapsed between the time Romney shored up the nomination and the time polls opened.
As we close the books on the year of the undecided voter--my pick for Person of the Year--here is a portrait of the procrastivoters.
The picture that emerges is a cohort of voters who have generally sophisticated concerns for the country but who have absorbed almost no information during the campaign beyond the names of the candidates.
Try this one on for size: “Analysis of healthcare between Romney and Obama.”
Here we have people who care about the crisis in medical coverage in America, understand that the candidates and parties have divergent ideas of how to fix it, but how no clear idea what those differences are.
Here’s another: “obama vs romney educational student loans.” Let’s hope those loans were not for degrees in political science.
The Yahoo Labs team in Barcelona produced the following interactive in which you can see the four top last-minute queries for each of five major topics in the election.
If you click on the queries, you can see the contemporary results for each query. While there’s no guarantee that these are the same pages users saw on the day before the election, we can be reasonably confident that they’re similar. The demand for articles and advertisements on the differences between the candidates dropped to zero the day after the election, and most of the results you see here predate the opening of the polls.
And here we are reminded of the critical importance of search engine optimization, the evil practice by which companies and campaigns game the search engine algorithms to return their content higher than the competition. (Everyone, including Yahoo News, participates in this arms race.) When people searched for “will middle class taxes go up under Romney or Obama,” the first result is to a page on the Obama campaign’s website. You could argue that this question is intrinsically slanted in Obama’s favor, given that the health of the middle class was the most sledgehammered subject of the president’s reelection effort. Still, the campaign page appears above a story from the umpire site Politifact and an article on the conservative RedState.com. You can guess which one people click on most.
The same is true of the query “Mitt Romney and Barack Obama views on women.” The top result is a campaign page, but this was another central tenet of the Obama bid.
More often than not, to be fair, most of the top results are to publications. Make whatever jokes you like about the media; voters are much more likely to get a fair assessment of the difference between the candidates from a journalism shop than from anywhere else.
One of the trickiest dilemmas in political polling is how to predict who will turn out to vote. In fact, we now know from detailed reporting by CNN, The New Republic, and many others that the Romney campaign made several wrong guesses about who would eventually decide to turn out and vote, leading to its false sense of confidence on Election Day. The data here suggests that, contrary to what you might expect, these last-minute voters are not purely self-interested. While you do see some of this, with queries like “who is better for federal employees,” others seek the aggregate opinion of others: for example, “military general officer's endorsements” and another wonders “who are Christians voting for.”
These are not single-issue voters; they are single-proxy voters. It suggests that campaigns might better spend some of that advertising bounty running ads featuring endorsements by leaders of core constituencies, and spare us all another 30-second ad featuring the candidate himself. Or axe some of those ads and divert a few more dollars to the search-engine optimization team--a move that, while terrible for democracy given the shadiness of gaming search algorithms, would at least be a small palliative to our ad-addled nerves during the baseball playoffs.