Why We Like Polls
November 15, 2013 - by
For all our disagreements, Carter and I agree on one thing: the value of polls. But we constantly have to explain why – and defend ourselves.
Most recent was a delightful dinner discussion with a group of learned and accomplished gentlemen from Chapel Hill. For the first half hour, we were peppered by questions about why campaigns do polls. Most of the questions had a distinct whiff of disapproval. One gentleman in particular took great umbrage that we would call him and ask for his opinion.
Similarly, Thomas Mills blogged this week about candidates who resist polls and demand: “Why do I need a poll to tell me what I think?”
You don’t (or shouldn’t) do a poll to tell you what to think. You do a poll to tell you what the voters think. As well as why they think that way, what they know and don’t know, how they react when they get new information and hear new arguments, where they get their information, etc., etc.
Polls – and all forms of public opinion research – are simply ways of listening to voters. Not listening just to anecdotes, or your friends, or your donors, or the activists who come to your events, but to real people who care enough to vote in an election.
What is so bad about listening to voters?
The other night, we got a lot of questions about how reliable polls are. That’s understandable. There are good polls and bad polls, on-target polls and off-by-a-mile polls.
Methodology is important. Did you get a representative sample of voters? Did you get the turnout model right? (See: Mitt Romney campaign.) How do you reach people who don’t have land lines? How do you word questions fairly? In what order do you ask questions?
Then there’s this perennial: “How can a poll of 300 people possibly predict how millions of people will vote?” Carter pointed out that it is purely a matter of statistical probability. If you poll a representative sample of 300 people, the results will be accurate within 4-5 percentage points 95 percent of the time.
In the end, Carter and I both fall back on experience. Over and over, we have seen good polls beat hunches and guesswork.
Here’s my best example. In 1991, Jim Hunt started thinking about running for governor a third time. Nearly everybody who had been with him in the 1984 Helms campaign was against it. “There’s no way,” they said. Hunt was damaged beyond repair, he could never come back, he could never overcome the harm that Jesse Helms’ negative ads had done.
Then we asked the voters. They had a different opinion: “He was a good governor.” “He proved he can do the job.” “Things have gotten worse since he left office.” “I’d like to see him run again.”
Hunt’s insiders were too inside. They were too close. They hadn’t forgotten 1984. The voters had.
Fortunately, Hunt listened to the voters.
Posted in General
Why We Like Polls
November 15, 2013/
For all our disagreements, Carter and I agree on one thing: the value of polls. But we constantly have to explain why – and defend ourselves.
Most recent was a delightful dinner discussion with a group of learned and accomplished gentlemen from Chapel Hill. For the first half hour, we were peppered by questions about why campaigns do polls. Most of the questions had a distinct whiff of disapproval. One gentleman in particular took great umbrage that we would call him and ask for his opinion.
Similarly, Thomas Mills blogged this week about candidates who resist polls and demand: “Why do I need a poll to tell me what I think?”
You don’t (or shouldn’t) do a poll to tell you what to think. You do a poll to tell you what the voters think. As well as why they think that way, what they know and don’t know, how they react when they get new information and hear new arguments, where they get their information, etc., etc.
Polls – and all forms of public opinion research – are simply ways of listening to voters. Not listening just to anecdotes, or your friends, or your donors, or the activists who come to your events, but to real people who care enough to vote in an election.
What is so bad about listening to voters?
The other night, we got a lot of questions about how reliable polls are. That’s understandable. There are good polls and bad polls, on-target polls and off-by-a-mile polls.
Methodology is important. Did you get a representative sample of voters? Did you get the turnout model right? (See: Mitt Romney campaign.) How do you reach people who don’t have land lines? How do you word questions fairly? In what order do you ask questions?
Then there’s this perennial: “How can a poll of 300 people possibly predict how millions of people will vote?” Carter pointed out that it is purely a matter of statistical probability. If you poll a representative sample of 300 people, the results will be accurate within 4-5 percentage points 95 percent of the time.
In the end, Carter and I both fall back on experience. Over and over, we have seen good polls beat hunches and guesswork.
Here’s my best example. In 1991, Jim Hunt started thinking about running for governor a third time. Nearly everybody who had been with him in the 1984 Helms campaign was against it. “There’s no way,” they said. Hunt was damaged beyond repair, he could never come back, he could never overcome the harm that Jesse Helms’ negative ads had done.
Then we asked the voters. They had a different opinion: “He was a good governor.” “He proved he can do the job.” “Things have gotten worse since he left office.” “I’d like to see him run again.”
Hunt’s insiders were too inside. They were too close. They hadn’t forgotten 1984. The voters had.
Fortunately, Hunt listened to the voters.
Posted in General