The Corner

A Flashback to Primary Polls — Undercounting the Republican Base

In all the back and forth over the “skewed polls,” it’s worth remembering a very recent example of a candidate repeatedly out-performing his polling: Rick Santorum in the primaries. As a Romney supporter, I recall sweating out each and every primary where Mitt led in the polling by less than five points because I perceived Santorum would do better than polls would suggest (including better than exit polls suggested). Turns out, he did — sometimes much better.

The trend became so pronounced that by March New York Times poll guru Nate Silver gave the issue an extended treatment on his blog.  An example:

[T]he Mississippi polls implied that Mr. Santorum had only about a 2 percent chance of winning the state, according to the model. He trailed both Mr. Romney and Mr. Gingrich by seven or eight points in the polling average there. Against one candidate, such a deficit is sometimes surmountable, but against two it has very rarely produced a victory.

The phenomenon applied (but to a lesser extent) even when polling had been quite extensive — and was enough to swing three states in Santorum’s favor:

So far, 11 states have had robust polling, which I define as multiple polls in the field in the days immediately before a primary or caucus. These states provide the clearest measure of how a candidate did against his polls on primary day — cases in which the polling was sparse, or there was a late swing in the momentum that the polls were not current enough to capture, are excluded.

In these states, Mr. Santorum has outperformed his forecasts by an average of 2.3 percentage points thus far. That isn’t a huge difference, but it was enough to swing three states for him that the model predicted he would lose — Alabama, Iowa and Mississippi.

At the same time, Mitt was not underperforming at the polls (he did, on average, 0.2 percent better than polls indicated) but Santorum was significantly overperforming. While primary polls were obviously measuring only likely Republican-primary voters (and thus didn’t have nearly the weighting challenges presented by general-election polls), there are still a few takeaways from the serial polling inaccuracies in the Republican primary.

First, pollsters consistently have trouble polling the Republican base.  Again, speaking from the perspective of an Evangelical for Mitt, I was keenly aware of the energy for Santorum among many tea partiers and Evangelicals, and that energy was reflected in the electoral outcomes. Even when examining primary voters only, the pollsters companies still misread base turnout. Just as general-election polls tend to undercount Republicans, primary polls tended to undercount or misread the base.

Second (and consequently) the base must remain energized. Here’s where the Ryan pick not only is vital (sharpening the contrast between Obama and a truly conservative ticket), but his importance won’t necessarily show up in the polls because he appeals to the very population the pollsters so consistently undercount.  

Third, while the polls will be wrong, they likely won’t be wrong by much. Yes, the pollsters made mistakes. Yes, the mistakes were in one direction. But they were still mostly right; an average 2.3 point variation is not that large. In other words, we’re likely not in the grips of a heinous anti-conservative pollster conspiracy (by most pollsters, anyway) but rather pollsters are struggling to predict turnout from the population most removed from them culturally and most suspicious of their methods and intentions.

Again, I realize these polls were different, but a consistent pattern of recent mistakes can still be instructive. The media has trouble understanding the Republican base, and if the base remains energized it can potentially shock the world on November 6.


The Latest