Big Data Lessons From A Political Pollster
Why didn’t the pollsters get it right on predicting the outcome of a certain ballot issue or candidate election? People always ask that question after learning of some unexpected election result. They probably will do so again after catching up on the November election results.
The same question often surfaces during insurance meetings. If strategic decisions based on data don’t pan out as expected, someone is always asking “Why? What went wrong?”
Nationally known political pollster David Plouffe shed some light on this during LIMRA’s recent annual meeting in New York.
Although data and polling can help predict outcomes, he said, “data can’t capture emotion, momentum and irrationality.” To get the full picture, assessing original feedback is important, too.
That may not be an answer most people expect to hear from a data analyst, because the pollster did not push the data envelope to the exclusion of all other information. His message was that taking the emotional temperature is important, too.
Plouffe speaks from experience. He was senior advisor to President Barack Obama from 2011 to 2013, and also campaign manager of 2008 Obama for President. Currently, he is senior vice president of policy and strategy for Uber, a company that connects riders to drivers through apps in many cities.
Before the 2008 primaries in Iowa, Plouffe said, the Obama campaign’s traditional polling, as well as public polling, showed that the candidate would lose. That was in view of Hillary Clinton’s 30 point lead. As it turned out, however, Clinton came in third and Obama won by 8 points
What happened? Plouffe said the Obama team looked into the undecided voters and realized that the Clinton campaign would have a hard time winning them over.
The Obama pollsters “used internal data about who we thought might caucus,” he said. “Modeling that out suggested that if we reached our turnout goal (200,000), we were going to do something that no traditional polling would capture.”
The method the campaign used was connecting with voters directly. “We went out and talked to people,” he said. The campaign used phone and social media contacting, too. “We got original feedback, which is the most valuable data you could have.”
Lesson learned: “Traditional means of measuring politics were deeply flawed,” he said.
The team learned another lesson five days later. This was about the power of emotion to alter results.
On Jan. 7, the day before the 2008 New Hampshire primary, the public polls and the Obama polls showed that Obama was in the lead. Meanwhile, following her defeat in Iowa, Clinton had dropped from 50 points to about 32, Plouffe said. People still liked her, he said, but some had had changed over to Obama or were undecided.
But come Election Day, Clinton won the primary. This was a “massive shift,” Plouffe said. He attributed it to emotion.
The day before the election, he explained, Clinton went to a diner. While there, she talked about how much the election meant to her, why she was in it, and who she was fighting for, Plouffe said. “She became very emotional and teared up, which was very unusual for her.” In 12 hours, that “became a viral sensation,” he said, with just about every voter in New Hampshire seeing it. The voters who had left Clinton came back and she won, he said.
Lesson learned: “What data can’t capture is emotion on a ‘change event,’” he said. A change event is something that just happens that affects how people feel about an issue or candidate.
Tips for business leaders
In the end, 240,000 voters showed up for the Iowa primary, a lot more than 180,000, which is the most that Plouffe said had ever shown up.
Later in his speech, Plouffe offered a few pointers for business leaders about using data analytics in decision-making.
- “Data can’t live in a room,” he said. “It has to flow through the organization.” He alluded to one disagreement that came up during the Obama campaign in Ohio. Obama decided to go with the analytics and the data, he said.
- Data analytics get more accurate as time goes on. Even so, he said, “no model can anticipate emotion that is based on events.” In the campaigns, this means that some decisions go on instinct.
- “The more we talk to human beings, the more effective we become.” The most effective contacts in the Obama campaign were door-to-door, Plouffe said. The campaign found that people have more trust if they get information for someone they already know and trust. “Social dynamics is so very powerful,” he said.
- Stories and videos showing people who are just like the consumer have impact. That also has application to people buying life insurance, he said. So “understand as much as you can. Have conversations. Talk about the pros and cons. Go out and talk to human beings.”
- When using qualitative data, “be careful not to put too much statistical value on it,” he cautioned. But do listen to the “rich information.” That leads to the best insights that capture the moment. Those conversations also help uncover how people talk about things, in their own language.
- “Remember — you can lose people in a minute if they think you are being inauthentic.”