Watching the analysis of Tuesday’s stunner on the Dem side of the New Hampshire primary, what I most wanted was to hear one of the bigfoot pundits say: “We were all wrong. We didn’t see this coming. The polls misled us. We don’t know why it happened and we should stop pretending that we do.” And then shut up for a while and let some more actual votes be counted.
The stunner last night was not that Sen. Hillary Rodham Clinton did better in New Hampshire than she did in Iowa. Iowa and New Hampshire often produce different winners. The surprise last night was all about the results not matching the pre-election polls, which showed consistently that Sen. Barack Obama would win New Hampshire big.
I’m looking at pollster.com’s compilation of the 22 most recent polls of likely New Hampshire Democratic primary voters taken after the Iowa results were in. (Note, that’s 22 polls of the same question over a five-day span. Any chance our political culture could benefit from a little less polling?)
One of them shows Clinton ahead by two points, and it is one of the older polls with one of the smaller sample sizes. One of them shows Clinton and Obama tied (also one of the older polls). Twenty of them show Obama ahead. In nine of the polls, Obama’s lead was 10 or more percentage points. In just three of the polls was Obama’s lead less than five points.
We are often cautioned against taking polls as predictions, They are only snapshots of a moment, and the moment is never the same as the election. And then there’s the fact that margins of error are always published (although many of these polls, if compared against the final result, were outside of the margin). I often pass along this caution myself, even while acknowledging that, like most political junkies, I am irresistibly drawn to the latest polling data.
But in the modern era of constant polling, using different methodologies and question wording, I can’t think of another instance in which such a strong consensus among poll results so close to an election turned out to be this far off from the final result.
A small thought experiment
Imagine that there had been no polls taken in New Hampshire. Before Iowa, when it appeared that Clinton might not win there, it was often reported that her campaign was attempting to set up a windbreak in New Hampshire to stop Obama from building irresistible momentum. In the no-polls fantasy, this strategy would have remained somewhat credible, and been mentioned alongside the possibility that it would fail and Obama would win New Hampshire with his post-Iowa momentum.
If Clinton had won narrowly and we hadn’t just seen 20 polls with big Obama leads, we would not have been shocked. But, because of the polling consensus, that idea was swept away.
Over the course of the late evening, many pundits pondered the impact of the footage showing Clinton choking up while talking about her motives for soldiering on. I normally eschew predictions, but I predict this moment will enter political legend, along with the greatly overrated Howard Dean scream of 2004 (people keep trying to forget that the scream followed Dean’s defeat in Iowa and therefore could not have caused it), and a similar misty moment by Sen. Ed Muskie in 1972 (although in Muskie’s case, the teary voice supposedly ruined his standing, while Clinton’s mistiness saved hers).
It may very well be that Clinton’s misty moment explains her rally. We don’t know and we never will. I would appreciate any pundit who embraces this fairly mushy analysis point who also acknowledges that it is simply a guess that can neither be confirmed nor falsified.
Even the elderly women who (according to the hypothesis and according to exit polls that showed Clinton with a huge margin among older women) shifted to Clinton after seeing her tear up, probably can’t say that that was the reason, unless the media onslaught convinces them that it was so.