The best thing about the November vote for president is that we are finally returning to a “normal” ballot scenario. Most voters will turn out, partisans will vote for their party’s nominee and independents will decide the result. The worst thing about November’s balloting is that pre-election polls will provide an imperfect and erratic prediction of the result because of the pitfalls associated with polling those pivotal independents. Tighten up those seatbelts, poll-watchers. It’s going to be a wild ride.
There are already signs that this is occurring. Various public and private polls that I have reviewed over the past two months suggest widely divergent outcomes for independent voting.
{mosads}Sometimes there are oddly large month-to-month swings within a single polling organization’s numbers for independents. And even when you focus on specialized polls, like those of swing states being conducted by Gallup for USA Today or Purple Strategies, the independent voting is inconsistent. Gallup’s latest polling had President Obama up handily among independents in swing states, while the Purple Poll had Mitt Romney winning independents. What’s up?
The first and most important explanation for this situation is that independents make up such a small share of each poll’s sample that the margin of error for independents is very high. So while you might see a poll touting a 3 percent margin of error for its total results, the margin of error for the decisive independent subsample may be 6 percent or higher. If you see Obama leading Romney among independents, say by 5 points, 49 percent to 44, the president’s real percentage could be as low as 43 percent and Romney’s as high as 50, reversing the outcome, and all completely acceptable to a statistician.
But it’s not purely sampling error that’s to blame here. Some part of the problem lies with different polling methodologies. And independents themselves — with all their wishy-washiness — bear some responsibility. Pollsters’ methodologies come into play in several ways. First, some pollsters use actual voter registration to determine partisanship or independence. Others ask about partisanship, as an attitude. There are justifications for both approaches, but comparing results of the two methodologies can account for differences. This is a tough election to use attitudinal partisanship, because we know that a large percentage of registered partisans have since 2010 started saying they are “independents,” muddling that category. My own analysis of this phenomenon suggests that more Democrats are guilty of this misclassification. So polls that use partisanship (rather than registration) might be plumping their independent category with closeted Democrats.
The way that pollsters approach identifying likely voters can also affect the results for independents. Being indecisive, some independents are not so resolute when it comes to declaring their intent to vote. My analysis suggests that less-educated and lower-income independents are more likely to strongly profess their intent to turn out than well-off, well-educated independents. The latter might hedge and not pass the likelihood-of-voting screen, and because they are actually just as likely to vote, and more likely to vote Republican, their non-inclusion in some samples reduces the apparent support for a candidate like Romney.
Independents themselves are to blame, too. They really do have troubling making up their minds. And when they do make up their minds, they change their minds, again. And the poll numbers fluctuate. A week of good news for one candidate has them headed his way. The next week a new set of troublesome headlines causes a reversal of course. Each poll until November is going to reflect that old cliché of the results being “a snapshot of one point in time.” It might sound like a pollster’s cop-out, but it’s really going to be true this time because of independents.
David Hill is a pollster that has worked for Republican candidates and causes since 1984.