No matter what your politics, you have to admit that Barack Obama gave an electric speech in Denver at the Democratic Convention. He tried to set the terms for the debate for the Presidential election to be about global issues rather than a particular candidate's stance on hot-button issues like abortion, gun-control, and gay
marriage. Immediately, though, the pundits all began to ask the same question. What will this do to the polling numbers?
The way we measure the success of a candidate during the election is through political polling. Pollsters ask "likely voters" who they would vote for in the election. The polls then give a margin of error (usually in the neighborhood of 2-3 points). This use of polls raises a more general question, which is what factors are likely to make a poll accurate.
First off, it is important to determine what you mean by accuracy. After all, what we are really interested in is how voters will act in the voting booth in November. So, the first goal of pollsters is simply to make sure they are getting an accurate idea of what people are thinking right now.
They start by polling people in the actual population of interest-likely voters. There are plenty of people who live in the United States who will not vote, either because they are not allowed to or because they choose not to vote.
Then, it is important to ask the question in a way that people are likely to answer truthfully (as they feel right now). For example, many years ago, I was reading a newspaper article that had a survey in it saying that 90% of people said they would help the homeless. A representative from the director of a city group that provided aid to the homeless was quoted as saying, "Those results indicate that 90% of people are unwilling to say that they are not interested in helping the homeless." In this election, questions about the degree that race will play a role in the election are probably less likely to be answered truthfully than questions about the importance of the economy in the election.
But the most important factor in the ability of a poll to accurately predict the election comes from the concept of attitude-behavior consistency. Essentially, 30-years-worth of research in Psychology makes clear that a person's behavior will match the attitude that they state when the situation they are in when they express the attitude closely matches the situation they are in when they carry out the behavior.
Let's take some obvious examples for starters. If you ask most people at random whether they would ever drive while drunk, they will say that they won't. Even leaving aside people's unwillingness to say that they will drive drunk, they probably still underestimate how likely they are to drive drunk. That is because they are (probably) not drunk when you asked them the question, and they are probably not at a party or a bar a few miles from home late at night. When people are actually in that situation, they end up expressing behavior that is different from what they believe they will do when you just ask the question.
Election polls are the same way. If you ask someone in August how they are going to vote, they give you the best answer they can. But the world is a different place in August than it will be in November. Candidates will say things between now and then that can affect people's beliefs. The world can change. And in the end, factors that do not seem so important when you are asked a question now may become important when you are actually standing in the voting booth ready to cast your vote.
And that means that you need to take polling number with a grain of salt. At best, they reflect what people are willing to tell a pollster about what they are feeling right now. And as you look at those poll numbers, you should ask yourself whether your own behavior ought to be influenced by what a few people tell you now that they might do in November.
And finally, if you are eligible to vote, then vote. Please.
