Monday, October 6, 2014

Beware of Outlier Polls

With the midterms less than a month away, some rather interesting polling is giving some Democrats and Republicans all kinds of goosebumps.

First the Democrats. In Kentucky, a poll by Survey USA shows Alison Lundergan Grimes ahead of Mitch McConnell by 2 points. It's the first poll in months that has shown Grimes ahead in the race and, predictably, it has garnered a lot of attention in progressive publications.

For Republicans, two polls out of Colorado, both by Quinnipiac, show the Republican candidates for governor and senator up by substantial margins over the Democratic incumbents. And, just like the Survey USA poll in Kentucky, both polls are polar opposites of virtually every other poll taken. Naturally, conservative publications are falling over themselves in anticipation of flipping the state.

Well, permit me to throw some cold water on both camps, but both Quinnipiac polls in Colorado and the Survey USA poll in Kentucky are nothing more than outliers. What is an outlier? Basically, it's anything that is so far removed from other measurable data that it could only have been the result of a statistical error. In 2012, you'll recall, Republicans were all pointing to a Gallup poll that showed Mitt Romney up by four points over Barack Obama nationally. It was the only national poll showing the former Massachusetts governor ahead in the race and, as it turns out, it was completely wrong. In fact, it was Obama who won the election by four points.

Outlier polls are fairly easy to spot, if you're paying attention. For instance, when you see one candidate ahead consistently in several polls and then, all of a sudden, one poll shows the other candidate ahead, you have to conclude that either all the other polls are wrong, which is highly unlikely, or that the lone poll is an outlier.

Kentucky and Colorado aren't the only states that have been the benefactors of outlier polling. New Hampshire and Georgia have been victimized by them, as well. In New Hampshire, Jeanne Shaheen has led pretty much from the beginning in her reelection bid over Scott Brown. Then a poll taken in early September showed the race tied. In Georgia, David Perdue, likewise, has been consistently ahead of Michelle Nunn when one lone poll showed Nunn ahead by four points. While these polls initially caused a stir in both races, it's worth noting that they have not altered the overall trajectory of either. As things stand now, both Shaheen and Purdue are favored to win. And, though it pains me to admit it, old Turtle face in Kentucky is going to win too.

The moral of this story is simple: if something is too good to be true, then it probably is.

No comments: