Exclusive: How Democrats Won The Data War In 2008

Part one in a three-part series: the 2008 Data Wars, Re-examined. Later today: Catalist's after action report, posted in full. Tomorrow: Inside the Service Employees International Union's sophisticated targeting.

Get-out-the-vote operations mounted by the Obama campaign, the Democratic Party and progressive organizations mobilized more than one million dedicated volunteers on Election Day. But it was buttressed by a year-long, psychographic voter targeting and contact operation, the likes of which Democrats had never before participated in. In 2008, the principal repository of Democratic data was Catalist, a for-profit company that acted as the conductor for a data-driven symphony of more than 90 liberal groups, like the Service Employees Union -- and the DNC -- and the Obama campaign. The Catalist data was crunched by the Analyst Institute, a DC-based organization that was set up to perform rigorous experiments like these on progressive voter contact methods.

According to the analysis, those registered voters contacted by Catalist member groups turned out at a rate of 74.6%; the voters who weren't turned out in proportions roughly equivalent to the national average -- about 60.4%. In four states, the number of new votes cast by liberals exceeded Obama's victory margin: in Ohio, Florida, Indiana in North Carolina. If you assume that only 60% of these voters chose Obama, the margin was still greater than Obama's in North Carolina and Indiana, both essential to his victory. With the caveat that correlation does not equal causation, the report provides convincing, if not absolute, evidence that the progressive/Democratic data-mining and targeting operation measurably helped elect Barack Obama.

The Atlantic has obtained Catalist's official after-action report, marked "proprietary and confidential."

Assessing this part of the election has been difficult until now, because the people involved, including the Democratic National Committee, the Obama campaign and interest groups, treat the data as a state secret. They don't want to let the other side know how good they've got it. The Republican microtargeting effort helped re-elect George W. Bush in 2004 and probably saved Republicans a dozen or more House seats in 2006. Democrats were slow to realize the benefits of the technology and, until recent election cycles, lacked the organizational acuity to figure out how to catch up.

Catalist's member groups contacted a universe of 49 million adults more than 127 million times; about half of them managed to vote, representing more than 20% of all the votes cast. With a major caveat attached -- it's hard to know why individual voters are motivated to cast ballots on Election Day -- the report concludes that progressive data-driven targeting was "essential to progressive victories." In the 16 states targeted by liberal groups, more than one in three voters was contacted at least once before Election Day.

The targeting wasn't limited to Election Day. Catalist member groups generated more than 7 million new voter registrations, according to the data -- more than half of whom turned out to vote. Catalist members claim to have accumulated more than 60 million new data points on 35 million Americans as the result of its canvassing activity.

So did it work?

"The data tells us that the volume of contact was extraordinarily large; it was orders of magnitude larger than in 2004," said Laura Quinn, Catalist's chief executive officer. But it all might have been noise had the Obama campaign not embraced a methodology that used the data effectively, she said. "They really religiously built relationships and kept up person to person contact. This was an extraordinary candidate, and it was an extraordinary election, and that opened the door to places that progressives hadn't been before. It really looms large in the data."

David Plouffe, Obama's campaign manager, said the data and modeling served as "great radar about the electorate" and helped the campaign develop smart tactics and metrics (and strategy) about delivering message. "It served a once in a generation candidate and his perfect for the times message well," he said.

To the extent that battle-hardened field generals of political campaigns engage in philosophical debates about their trade, the efficacy of microtargeting is bound to come up. Steve Hildebrand, who set up Obama's voter contact and turnout operation for the primary and general elections, once called the microtargeters' claims "magical." Though he worked well with Ken Strasma, who modeled voter universes for the campaign, he is among those campaign operatives who consider microtargeting a tool that has, at best, marginal effects on elections. Indeed, as the Catalist report admits, "The data in-hand, for the most part, is still insufficient to do more than indicate probable 'causality' in some cases, but not certain causality across the board." What that means, in essence, is that to determine whether the activity generated the outcomes -- the modeling and targeting made individuals more likely vote -- techniques must be tested alongside control groups of voters who don't get the special, individualized attention.

The more people were contacted, the more likely they were to vote -- particularly if the contact was made by a person, speaking to them on their stoops. (This tracks with the research findings of Donald Green, a Yale University political scientist who has studied voter contact efforts.)

In a memorandum attached to the findings, Catalist officials take pains to note that correlation does not imply causation, and that it is next to impossible to prove that microtargeting moved votes.

But by controlling for demographic variables, the evidence is considerable that the tailored messaging, contacts and targeting played a significant role.

Later today: read the Catalist after-action report for yourself.

Until the 2008 cycle, it was hard to get Democratic groups, and even Democratic state parties, to share technology and data. Democrats relied on static and easily obtainable information, like voter files and party registration data, to find other Democrats. Pollsters were often able to target particular messages to particular groups of people, but it was not until 2008 that targeting, for Democrats, increased its magnification to the level of the individual. Perhaps even more importantly, the type of targeting and voter contact activities was less effective. When Democrats identified a likely voter or a potential voter based on telephone contact, they lacked the resources and know-how to follow up in a way that increased the likelihood that the voter in question would actually turn out to vote. Republicans, by contrasted, had pretty much mastered this process.