“The more voter contacts you make in the field, the better. Then you can actually test your models and improve them.”
After the 2004 election, Democrats got serious about building a tech and online infrastructure. They had lost two presidential elections in a row, to George W. Bush, who won with the help of microtargeting pioneered by Karl Rove and Ken Mehlman, who went on to become the chairman of the Republican National Committee. Heading into 2008, the Democratic National Committee was committed to maintaining and updating its lists of voters and donors, as well as finding new ways to target them through email and canvassing.
On this week’s episode of our podcast What’s The Point, we conclude our history of political data with professor Daniel Kreiss of the University of North Carolina. In 2008, Barack Obama’s campaign was able not only to microtarget by demographic group, but also to track individuals within households and craft a message that helps drive them to the polls. Kreiss describes the Obama campaign’s efforts and how the GOP tried to push back in 2012.
Plus, we discuss how data is being deployed in this year’s campaign. A FiveThirtyEight team recently spent a week in Iowa reporting on the run-up to the Feb 1. caucuses. We’ll hear reports from the Hillary Clinton and Ted Cruz field offices.
To listen, stream or download the full episode above. Below are some photos from Iowa and a transcript of interview highlights. You can listen to Part 1 of the podcast here.
Volunteers make phone calls at Hillary Clinton headquarters in Davenport, Iowa.
Jody Avirgan
A call sheet used by volunteers at Hillary Clinton headquarters in Davenport, Iowa. Answers are logged on paper and then entered manually into the VoteBuilder database.
Jody Avirgan
Highlights from the recent history of political data
2008: Obama tracks individuals
Daniel Kreiss: In 2008, the Obama campaign really started to do predictive level modeling and individual level modeling. The older method for targeting voters was to generate static microtargeting categories of voters based on polling at a particular moment in time.
The insight that Obama staffers such as Dan Wagner had was that larger conceptual demographic or psychographic or behavioral groups didn’t really exist. Within every category of those voters, there was significant diversity. So what the Democratic Party did was move to a model that would be much more accurate if it was premised on individuals, not necessarily groups.
2008-12: Testing public data against the model
Kreiss: The foundation of all this is public data. Vote history. Party registration. Gender, age, geography. Things like marital status, children in the home. Race, in particular states. Any modeling is really built on these basic categories of public information.
Jody Avirgan: But there were all these reports that the Obama campaign was using car dealership information and cable box information and all this sophisticated information. … And you’re saying it’s just, like, race, gender, age, marital status?
Kreiss: Those basic categories are the most predictive. And campaigners will tell you this. But then additional data points add marginal predictive utility. So you get better. The more data that you can add, the system gets better. The more voter contacts you make in the field, the better. Then you can actually test your models and improve them. And that’s really part of the story of the Obama 2012 campaign. They put all these canvassers out in the field and started to generate field level data — that then makes its way back and tests the model.
Volunteers make phone calls at a Ted Cruz field office outside of Des Moines, Iowa.
Jody Avirgan
A call script used by the Ted Cruz campaign in Iowa. The phones are equipped with an interface that lets volunteers log voter responses directly into a database.
Jody Avirgan
2012: The GOP loses an election it thought it should have won
Kreiss: What I think happens is that in 2012, the Republicans finally lost an election they believed they should’ve won. And it’s at that moment when you see a reorientation of the party in this direction of technology, digital, data and analytics. The investments start to be made. I think Republicans had such the advantage in 2004 that they basically got to sit on that for a little while.
After 2012, you see the GOP start to hire much more in tech areas. You see a cluster of new firms, firms like Deep Root Analytics. There’s a race to catch up on the Republican side of the aisle.
2016 and beyond: Does any of this work?
Avirgan: If we think about the sweep of history from 1891 through today, does any of this work? Does it make a difference?
Kreiss: I love this question. I think the overriding consensus in political science and among practitioners is that it works on the margins. Generally, the metaphor I often hear from practitioners is that it’s worth a field goal. Data may help you turn out more voters than the next guy. You will be more efficient about how you spend your campaign resources. It enables you to raise money more efficiently.
But there are a lot of things that matter in electoral politics. The electoral context matters. Which party is in the White House matters. The issues that people run on matter. The state of the economy matters. These are deep structural factors.
But when you think about what campaigns can do, technology, digital data and analytics are things they actually have control over. And it’s valuable, but on the margins.
If you’re a fan of What’s The Point, subscribe on Apple Podcasts, and please leave a rating/review — that helps spread the word to other listeners. And be sure to check out our sports show Hot Takedown as well. Have something to say about this episode, or have an idea for a future show? Get in touch by email, on Twitter, or in the comments.
What’s The Point’s music was composed by Hrishikesh Hirway, host of the “Song Exploder” podcast. Download our theme music.