“WHETHER YOU VOTE IS A MATTER OF PUBLIC RECORD!” the letter proclaimed, followed by a report card comparing the recipient’s voting record to his/her neighbors’. The letters sparked outrage … and higher turnout. (More)

The Victory Lab, Part II: Who Votes How, Why?

This week Morning Feature discusses Sasha Issenberg’s The Victory Lab: The Secret Science of Winning Campaigns. Yesterday we looked at the familiar, precinct-based analysis and planning. Today we see the rise of data-mining and campaign plans that target individual voters. Saturday we’ll ask whether micro-targeted voter contacts provided the Obama campaign’s decisive edge.

Sasha Issenberg writes the Victory Lab column for Slate and is the Washington correspondent for Monocle, where he covers politics, business, diplomacy, and culture. He covered the 2008 presidential election for the Boston Globe, and has also written for New York, The New York Times Magazine, The Washington Monthly, The Atlantic, and other magazines.

Yes, of course, I voted….

The idea for that voter turnout study began with a statistical anomaly. The Census Bureau conducts the Current Population Survey to gather employment data for the Bureau of Labor Statistics, and each November in a federal election year they also ask respondents if they’re registered to vote and, if so, whether voted in the election earlier that month. They report the data, which offer an early snapshot of overall voter turnout. And each November, the number of people who say they voted is higher than the state- and nationwide totals of actual votes cast, as released by elections officials.

Political scientists thought there must be a methodological flaw. Maybe the BLS random sample was skewed toward voters and overestimated turnout. Maybe more people voted than elections officials’ numbers showed, because their ballots not counted for one reason or another.

To resolve the anomaly, researchers checked voter rolls and followed up with specific survey respondents, and they uncovered a different explanation. Simply: many nonvoters lied. They told the BLS survey they voted, when voter rolls showed they had not.

A Lying Shame

That explained why BLS turnout estimates were higher than the actual turnout, but it raised deeper question: why would people lie about voting … and what might that lie reveal about why voters voted?

Traditionally, political gurus looked at voting as an economic act that was, or at least should be, grounded in rational analysis. Voters should look at the candidates and their policies, weighing pluses and minuses as applied to the voter’s own wants and needs. Of course no voter really does that, so the gurus shifted their attention to why voters use heuristics, information shortcuts such as party affiliation or a candidate’s gender. The gurus decided voters weighed the time cost of gathering enough information for a fully rational decision, balanced that against the tiny chance their vote would affect the election’s outcome, and decided the shortcuts were good enough.

To boost turnout, the gurus told candidates, parties, and other groups to appeal to voters’ sense of civic duty, with aspirational messages like “Your vote may decide this election!” or “Your vote is your voice!”

Yet researchers looking at the BLS data saw something else. Voters were well aware of their civic duty, as many lied about having voted. That suggested voting was more than an individual, calculated act. It was a social act. One researcher decided to test that theory during a Michigan primary election, sending out five mailers ranging from the aspirational (“Your vote is your voice!”), to statistical comparisons (“Over 70% of people don’t vote in primaries!” and “Only 30% of people vote in primaries!”), up to outright shame (“Here are the voters in your household and their voting records!” and the neighborhood mailer described above). A control group received no mailer about the upcoming primary.

The results were … astonishing.

The researchers couldn’t count the number of outraged calls, as their voice mailbox filled within minutes after they cleared it each day. People wrote angry letters. A local columnist, who often chided readers for not voting, penned a furious screed because his wife and adult children now knew he hadn’t voted in recent primaries.

Yet the data were clear. About 28% of the control group voted in that primary. Those who received aspirational and statistical messages turned out in the low 30s. Family shame letters boosted turnout to 35%, and those faced with neighborhood shame turned out at the rate of 37%.

Geeks vs. Gurus

The gurus didn’t like the geeks’ results in that and several other studies, because the geeks’ data showed the gurus were little more than articulate storytellers who believed their own stories. The gurus would sagely tell candidates – or cable TV audiences – who would vote, who wouldn’t, and why. This ad or that event would be a game-changer, they would argue, often citing a past election they had managed where a similar ad or event had preceded some twitch in the polls.

Their arguments felt convincing, and their advice was expensive. Indeed a candidate’s hiring a top guru was – the gurus said and many in the media agreed – a litmus test of political viability. Into that flattering, profitable bubble waded the geeks, with data they said proved that much of the gurus’ advice was folklore and hokum.

What’s more, the geeks claimed they could predict which voters were persuadable, and by what messages, including many voters the gurus would never have even considered. Their regression analyses showed that a 26-year-old single woman with a college degree who mentors high school students … has more in common with a 33-year-old college-educated single woman who donates to the ACLU and lives in a reliably Democratic precinct … than she does with the other, mostly older women in the Republican base precinct where she lives with her parents.

Have someone call that 26-year-old single woman with a pitch about your candidate’s passionate commitment to education, the geeks said, and she’s likely to vote. You’re more likely to get her vote if the call comes from a local volunteer rather than a paid call center, more likely still if the caller asks questions rather than reading from a ‘robo-script,’ more likely still if the candidate sends her a mailer. And she’s most likely to vote if a local volunteer knocks on her door and speaks with her face-to-face.

That combination of data-mining, sophisticated mathematical analysis, and personalized messaging was the key to winning elections, the geeks insisted. And they had studies to prove it.

But as we’ll see tomorrow, the gurus weren’t convinced, and they haven’t given up.


Happy Friday!