A poll may not vote, but those who answer it do
That’s what Vice President Walter Mondale used to say at every stop in the closing days of the campaign. “Polls don’t vote. People vote.”
The only problem, of course, is that polls do measure people’s preferences, typically with greater accuracy in general elections than in primaries and caucuses, where it is more difficult to predict what the universe of voters will look like. So when you’re down in the polls as the election is approaching, it generally means that you’re down.
In recent days, there has been a spate of stories about so-called “biased” polls — biased against Mitt Romney.
The national polls show a close race within the margin of error. State by state polls show President Obama with a somewhat larger margin in key states, but with Romney still in striking distance. The race, in short, is far from over, and anyone who thinks otherwise has never been around races that really were over (Mondale/Reagan) or felt the bottom fall out of a campaign because of a bad debate (Reagan/Carter; Dukakis/Bush).
So what’s with this business about biased polls?
Advocacy groups often generate biased polls to support their position, aid fundraising, influence policy and the like. On abortion, for instance, I used to say — and I think it is still true — that if you let me write the poll, it will show the country to be overwhelmingly pro-choice.
“Push polls” involve the same sort of game, although no one even pretends that they’re really polls. “Mitt Romney thinks people who receive Medicare are ‘victims’ who don’t take personal responsibility. Knowing that, Mrs. Senior Citizen, are you more or less likely to support Mr. Romney?” Less. “Barack Obama’s former minister Rev. Jeremiah Wright said in 2009 that he hadn’t spoken to the president because ‘them Jews aren’t going to let him talk to me.’ Knowing that, Mr. Jewish Voter, are you more or less likely to support the president?” Less.
But that isn’t how horse race polls are conducted. The big issue in horse race polls is weighting the poll so it reflects the demographics of what the voter pool will look like. You don’t want to have too many young people, too few seniors or whatever, because if you do, your results will be wrong. Similar national polls with differing results occur due to slight variations in how turnout is predicted.
Even the best poll is a snapshot, not a prediction. It tells you where voters are today, not Nov. 6.
That’s why virtually every horse race question is followed by one about just how sure you are. Are you certain you are going to vote for X? In this race, more than 90 percent of each candidate’s supporters say they won’t change their minds, which is a very high number and a reflection of a very polarized electorate.
It’s also why people pay so much attention to the non-horse race measures. If you’re the incumbent, you want to see people feeling confident (or more confident than they were last month) about their economic situation. Do you think the country is on the right track or the wrong track? The right track is where the incumbent obviously wants people to be. Do you think the candidate cares about or understands the problems of people like you? Too many nos means trouble.
At the end of the day, the most important issue for any pollster worth paying attention to is not whether their poll helps or hurts the candidate they (secretly or not so secretly) favor. It’s whether it’s right.
Always believe your pollster when he gives you bad news, the late pollster Tubby Harrison used to say. For my part, I always believe the worst poll. The easiest way to lose is to believe you’ve won when the fat lady hasn’t even started humming.
