The Data-Driven Campaign

Most of the decisions campaign strategists make are based on folk theories, by which I mean theories that have been developed over time by practitioners about what works and what doesn't, without much empirical evidence to demonstrate whether they're true or not. At the end of a campaign, you have a sense about what worked and what didn't, but the evidence is largely circumstantial, and the strategist's inclination toward self-justification -- even though we lost, my decisions were good ones, and we would have lost anyway, or alternatively, we won because everything I did was right -- makes clear-eyed assessment difficult.

But in the last few years, campaigns have had more quantitative tools available to them to make precise evaluations of the efficacy of various moves. Did canvassing that neighborhood with a lot of independents win over voters? Was it worth our time to send the candidate to that far-flung corner of the district? Did that radio ad buy move any votes? At the same time, some political scientists (most notably Donald Green of Columbia University) have been doing careful experiments to evaluate campaign techniques. What they've discovered is that some of those folk theories are valid, others less so.

Sasha Issenberg, a journalist best known until now for this glorious evisceration of David Brooks, is coming out with a new book soon about all this called "The Victory Lab." In an interview with the New York Times' David Leonhardt, he reveals something interesting:

Q: What makes Rick Perry's approach to politics different from that of other candidates?

Mr. Issenberg: No candidate has ever presided over a political operation so skeptical about the effectiveness of basic campaign tools and so committed to using social-science methods to rigorously test them.

As the 2006 election season approached, the governor's top strategist, Dave Carney, invited four political scientists into Perry's war room and asked them to impose experimental controls on any aspect of the campaign budget that they could randomize and measure. Over the course of that year, the eggheads, as they were known within the campaign, ran experiments testing the effectiveness of all the things that political consultants do reflexively and we take for granted: candidate appearances, TV ads, robocalls, direct mail. These were basically the political world's version of randomized drug trials, which had been used by academics but never from within a large-scale partisan campaign.

The findings from those 2006 tests dramatically changed how Carney prioritized the candidate's time and the campaign's money when Perry sought re-election again in 2010 and will inform the way he runs for president now.

As this shows, it's perfectly possible to be an anti-science know-nothing, as Perry apparently is, and also have very smart people working for you. But a lot of this is essentially campaigns catching up with what social scientists have known for some time. It's pretty remarkable that the Perry campaign allowed a bunch of scholars to conduct a randomized experiment with Perry's time, but what it demonstrated in the case Issenberg cites wasn't new: when you send a candidate to a town in person, the result is a flood of positive local news coverage (The candidate himself, right here in our town! Walking down Main Street! Having lunch at Merle's!), and that can have lasting electoral effects.

What's also interesting, as Issenberg points out near the end, is that if Perry is the nominee (or equally so, I'd bet, if Mitt Romney is), the 2012 general election will feature two campaigns that are both intensely data-driven, constantly monitoring their performance and making decisions based as much on the evidence they've gathered as on the gut feelings of a "senior strategist."

UPDATE: Forgot to paste in the first two paragraphs of this post. Complete now. Thank you for your patience.