Get Herald RSSThroughout the 2008 primary election cycle, there have been polls from nearly every state, and many have been completely inaccurate. Just think back to the polls showing Obama up by double digits prior to the New Hampshire primary, or the exit polls leaked on the Drudge Report on the evening of Feb. 5. Each campaign has internal pollsters, and the media relies on polls to determine whether there is movement in the electorate and to gauge the effect of certain stories, advertisements, endorsements or gaffes on potential voters.
Currently, most news about polling relates to general election matchups between McCain and Obama or McCain and Clinton, rather than upcoming primary elections. Many polls show both Democrats ahead of McCain, and state-by-state general election breakdowns suggest different optimal geographic strategies. On the whole, most polling indicates that either Democrat has the capacity to win.
This is all well and good. If polls are being conducted, the American public certainly has a right to know about them. But the fact remains that, oftentimes, people put too much weight in polls without subjecting them to scrutiny.
Any given poll could have flaws in sampling or methodology. Polls are often conducted over the phone, but many voters primarily use cell phones, not landlines. Polls often have to oversample minority groups to achieve reliable samples, and polls rarely (if ever) address issues of turnout accurately. Yet these structural issues pale in comparison with several intractable polling dilemmas.
A poll relies on individuals - in this case voters - voluntarily and accurately reporting their private choices. Unfortunately, the incentive to lie in exit polls, opinion polls or any other kind of poll is pretty strong. One example of this is the "Bradley effect," in which non-white candidates do well in polls but then do not achieve the same results at the ballot box.
One theory suggests that people lie to pollsters about their support for a candidate of color in order to avoid appearing prejudiced. Regardless of the empirical validity of this theory, which is difficult to verify, race could be one of many factors that give individuals an incentive to lie to pollsters.
This incentive to lie is a much more serious problem than methodological flaws. As the Democratic race becomes increasingly bitter and divisive, we need to reevaluate the degree to which polls, at the state and national level, can accurately forecast the result of the general election.
Recent polls have shown a potentially disturbing level of Democratic defection to McCain. A recent Gallup poll reported that 28 percent of Clinton supporters would vote for McCain over Obama, and that 19 percent of Obama supporters would choose McCain over Clinton. Many Clinton supporters try to brandish this as evidence that Obama is unelectable. Meanwhile, Obama supporters try to use this as evidence that Clinton should bow out before the party is irreparably divided.
Well, no offense to Gallup or anything, but that poll result seems like one of a number of bogus results in this election cycle. A rational supporter of either Hillary or Obama has a vested interest in making her disfavored Democratic candidate appear unelectable. If she believed it would sway public opinion, she would be more than willing to lie to pollsters. The most surprising aspect of this poll is not that so many said that they would defect, but that so few said so. A similar trend was reported during the 2000 Republican primary. Right after McCain lost the nomination to Bush, about half of McCain supporters said that they would vote for Gore. Fortunately for Bush, that didn't happen and he was elected.
Polls are flawed, imprecise instruments for measuring trends in the larger population. And while they aren't going anywhere, everybody should take a deep breath.
The general election is still six months away. While John McCain will be a formidable competitor for either Democratic candidate, Democratic voters should stop worrying about general election polls. Whether these polls show the effects of a divisive primary battle or claim that McCain will win California by 20 points, they all have issues of reliability and deserve serious scrutiny. And Democrats should still have time to rally around their candidate and defeat McCain in the fall.
Max Chaiken '09, a coordinator of Brown Students for Obama, is actually a covert pollster, lurking around campaign offices
