By Mark Mellman - 01/23/08 12:01 AM EST
It was not Iraq, but for pollsters involved in the New Hampshire debacle (fortunately not us), it seemed like the professional equivalent.
With almost as many explanations circulating as there are voters in New Hampshire, sorting through them would take more like 60,000 words than 600, but I’ll take a shot at some of the prominent theories, while adding to the jumble with a couple of my own.
An important caveat, though: With too many theories chasing too little data, no one can prove anything.
Nonetheless, we can rule out some frequently proffered possibilities. The composition of the electorate was not dramatically different from that anticipated by the polls, nor did pollsters simply stop their interviewing too early.
Despite some misleading attempts to save face, most of those who were wrong continued to be wrong in their late polling.
Perhaps the most widely discussed explanation is the “(Tom) Bradley effect” — the supposed tendency of white voters to exaggerate their support for black candidates. Historically this theory occasionally appears to be true, but mostly is not, which hardly inspires confidence in its explanatory power.
Moreover, in this instance, the theory runs head-on into three problems. First, could it really be that just days earlier, Iowans exhibited no symptoms of a malady that so afflicted New Hampshire?
Second, how is it that New Hampshire voters were perfectly willing to express a preference for Sen. Hillary Rodham Clinton (D-N.Y.) over Sen. Barack Obama (D-Ill.) for the year preceding the primary, but suddenly developed this racial consciousness about poll questions during the five days after Iowa?
Perhaps most significantly, the “Bradley effect” does not fit the facts. Only our rather foolish preoccupation with poll margins creates the appearance that New Hampshire conforms to the theory. Yes, Obama was “way ahead in the polls,” but lost on Election Day. However, Obama’s 36 percent of the vote was almost exactly the 37 percent predicted by the polls.
The candidate whose final tally was substantially under-predicted by the polls was Hillary Clinton. On average, they predicted she would garner 30 percent, but her actual share was 9 points higher. What cries out for explanation is not the failure of polls to mirror Obama’s vote, but rather the inaccurate measure of Clinton’s support.
Clinton may have been a victim of something similar in theory to the “Bradley effect,” which I might label the “embarrassment effect.”
After her Iowa loss, much of the press, along with the public discourse, turned harshly negative toward Clinton. Our own post-Iowa study of word-of-mouth in New Hampshire revealed a poisonous milieu for Sen. Clinton. Perhaps, in the face of all that negativity, her supporters were reluctant, even embarrassed, to admit they were still planning to cast ballots for her.
While less likely to account for the New Hampshire fiasco, the rather bizarre way pollsters treat undecideds deserves mention. Many labor to reduce the level of indecision at all cost. Insider publications regularly lampoon polls registering large numbers of undecided voters.
Yet what does it mean that in the final CNN/MUR/UNH New Hampshire poll, for example, just 6 percent were undecided, but, in the very next question, only half claimed to have “definitely decided’ for whom they would vote? At the very least, it suggests the horse-race numbers may have been something of an illusion.
Whatever one’s favorite explanation of the New Hampshire fiasco, it reinforces the fact that the most important errors in polling are not the statistical margins of error on which we lavish so much attention.
Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients include the majority leaders of both the House and Senate.