Any single poll question, or even a set of questions, is inevitably limited in its ability to divine the complexity of public opinion. Trying to determine why people behave the way they do is an especially difficult task, partly because the factors that influence their behavior may be elements of a holistic framework that can make it impossible to separate one factor from another. No matter how much we "crunch" the numbers, some questions just can't be answered.
This caveat about the realistic limits of polls came to mind when I was reading an article last week by William Saleten, the chief political correspondent for the online magazine, Slate. Titled, Kerried Away: The myth and math of Kerry's electability, the article asserts that the very polls that suggest Massachusetts Sen. John Kerry is the most electable of the Democratic candidates also provide evidence that he may not be.
Among his various arguments was one implying that voters interested in issues were not strong Kerry supporters. Saleten asked why Kerry had won 12 of the 14 primaries and caucuses held thus far. His answer:
"Not because voters agree with him on the issues. The reason, according to the exit polls, is that voters think he's the candidate most likely to beat Bush."
Saleten arrives at this dubious conclusion because in several exit polls, a question asked whether people voted for their candidate "more because you think he can defeat George W. Bush in November [or] he agrees with you on the major issues." Voters who chose the first option gave overwhelming support to Kerry, but voters who chose the "issues" option gave proportionately less support to Kerry and more support to his opponents.
Saleten's conclusion, that voters are not choosing Kerry because of the issues, assumes that voters who supported Kerry because they perceived him as the most electable did not also vote for him because they agreed with him on the issues. And that's where the limit of a single polling question comes into play.
The exit poll question did not allow voters to say they were voting for Kerry both because he was most electable and because he agreed with them on the issues. Voters had to make a choice. So, the more passionate anti-Bush voters -- and not surprisingly (given that these were Democratic primaries), there were many such voters -- were looking for a candidate who would have the best chance of beating Bush.
But did that mean these voters didn't also agree with Kerry's issue positions? There is nothing in the data to justify that conclusion. Moreover, given the slight differences that exist among the issue positions of the candidates, one might expect most Democrats to agree with the candidates on most issues.
Saleten's other principal argument is that, apart from the issues vs. electability comparison, candidates such as retired Gen. Wesley Clark and North Carolina Sen. John Edwards appealed proportionately more than Kerry did to independents and even Republicans who voted in the Democratic primaries. This suggests that in a general election, either of the two candidates would fare better against Bush than Kerry.
There are many problems with this argument, most notably that independents and Republicans voting in a Democratic primary are not likely to be representative of independents and Republicans nationally. Did they vote because they were out to sabotage the Democratic electorate's ability to choose the best candidate? If so, they can hardly be relied on to show that Kerry would not be the best candidate. That may be the reason they voted against him.
Another problem with his argument is one that Saleten immediately acknowledges -- that in the hypothetical head-to-head matches between each of the Democratic candidates and Bush, polls show Kerry doing the best. But Saleten says he wonders whether that advantage would hold in a general election, when Kerry is better known than he is now.
Here comes another caveat about polls. They are notoriously bad about predicting what people will do, especially when the predictions are for nine months into the future. And the reason why polls are so bad at predicting people’s behavior is that people themselves are wildly inaccurate about predicting their own behavior. Indeed, the whole electability question is at best a guess. No one really knows how much better or worse a candidate Kerry will make in comparison with other Democratic challengers. And no matter how many numbers we "crunch," polls can't answer that question either.