Big Data

Commentary: The Diminishing Returns of Big-Data Campaigning

What if we were all wrong about the Obama campaign’s vaunted technological advantage?

That’s what two Harvard researchers are suggesting. In a new head-to-head comparison of voter turnout in battleground states, what many believe gave President Obama an edge—his data-driven efforts at getting out the vote—might not have mattered much after all.

In fact, the researchers discovered, when it came to turning out his swing-state base, that GOP nominee Mitt Romney actually performed about as well as Obama did. The difference came down to little more than a percentage point. If those numbers are right, then either Obama’s ground game was less effective than everyone says, or Republicans aren’t quite at the disadvantage they think.

“You see a lot of talk in the media,” researcher Ryan Enos told me. “You know, ‘Obama has a vastly superior technological campaign’ and stuff. If it was vastly superior, it was 1.6 percentage points vastly superior.”

Here’s how the study went down. There are 38 media markets in the United States that each cover a portion of a swing state and a portion of a non-swing state. Cross-referencing that with a national voter file,  Enos and fellow political scientist Andrew Fowler were able to identify 42 million individual voters. All of them were seeing the same ads on TV but, depending on which side of the border they lived on, they were getting different getting-out-the-vote treatments, or personal outreach from campaigns encouraging them to vote, from either campaign.

Compared with their counterparts in non-battleground states, swing-state voters turned out at a significantly higher rate—4.5 percentage points higher, to be exact. And when you break it down by party, the results are even more pronounced. Team Obama managed to turn out more Democrats in swing states than it did in safe states, by a margin of 15.4 percentage points. Registered swing-state Republicans, meanwhile, were 13.8 percentage points more likely to visit the polls than those across state lines.

So Obama bested Romney on swing-state turnout by 1.6 percentage points. In a closer race, that might be enough to flip a state. (What actually happened: Only four swing states were decided by less than 5 points. The race just wasn't that close.) But instead of focusing on whether 2012 could’ve ended differently, it’s better simply to conclude that the figure contains a whole range of possible stories.

Does that match up with the postelection narrative? Not really. A lot of ink has been spilled analyzing what Republicans did wrong, and what Democrats did right. One of the major lessons both sides drew from the experience was that data and behavioral analysis were key factors in the outcome—an idea propelled by political journalists infatuated with the novelty of Big Data. (I’m just as guilty of this as anyone.) That conclusion has done a lot to shape parties' behavior in 2013. Republicans, for example, are busy building a new data warehouse to compete with Democrats.

None of this is to suggest that data and A-B testing and understanding what motivates people is a worthless enterprise. But it does imply that the marginal utility of doing all these things isn’t as great as we might assume. In light of that, perhaps the digital deficit that Republicans think they face isn’t so serious after all. But the bad news for them is that if technology doesn’t explain why they lost, the party’s demographic and policy shortcomings become more of a problem.

Maybe you think 1.6 percentage points is actually a lot. Or maybe you think it isn’t a lot, but that Obama’s massive investments in turning out the vote were worth it nonetheless (the “every bit counts” argument). Maybe you think technology made a difference in other ways (fundraising, signing up new voters or persuading people who may have been on the fence). That’s something people will have to decide for themselves.

Threatwatch Alert

Thousands of cyber attacks occur each day

See the latest threats

JOIN THE DISCUSSION

Close [ x ] More from Nextgov
// 6:18 PM ET
X CLOSE Don't show again

Like us on Facebook