In Defense of Public Opinion Polls
People lie. People say they'll do one thing and then do another. People have free will. Predicting what people will do is harder than predicting the weather. Those who religiously track political polling found this out the hard way last night.
The great thing about the atmosphere is that it lays its cards out for us to read. Everything is there in plain sight. The atmosphere has few tricks up its sleeve because it, unlike people, does not have free will. It follows the laws of physics. When the atmosphere catches us off-guard with a surprise snowstorm or a tornado, it's not the atmosphere's fault—that's on us for not fully understanding how the atmosphere works.
Public opinion polling firms have to predict how people think, whom people support, which issues are most important, what outside factors will influence how people will act, and ultimately if people will walk the talk and get out the vote. They use previous elections and previous electorates to predict for whom people will vote. Sometimes they're right, and sometimes they're wrong. In the 2014 midterms, they were really wrong.
It became clear early last night that the public opinion polls missed the mark when compared to the results. For Democrats, it was a shocking bloodbath on a night that was supposed to be a readjustment among red states. For Republicans, it was a vindication of something they knew all along.
Let's take a look at the preliminary final tallies in some of the most hotly-contested top-ticket races around the country.
Here are the results from the "close" Senate seats. The surprise of the night was Mark Warner's squeaker against Ed Gillespie. Warner's reelection was expected to be a cakewalk—polls showed him winning by nearly ten points—but barring the results of a potential recount, he won by 40,000 votes, or by six-tenths of one percent. The other shock was the race in Kansas, where polls showed incumbent Republican Pat Roberts losing to independent Greg Orman, only to win the election by nearly 11%. Thom Tillis won Kay Hagan's seat in North Carolina by two points in what was supposed to be a race destined for recounts.
Gubernatorial races across the country saw similar inaccuracies, especially in Maryland, where the Republican candidate pulled off one of the greatest upsets in modern political history.
I'll leave the reasoning behind deep biases in the polling to political scientists with more experience than I'll ever have, but one thing I've noticed after every election where the results take a sharp turn away from polling is that people blame the polls. They blame the polls. Think about that.
People blaming the polls for being wrong is like people blaming meteorologists when it snows and they called for rain. It was always going to snow, it's just that we didn't know it would snow until it snowed. When you have shocking but decisive victories like Pat Roberts' in Kansas or Larry Hogan's in Maryland, it's not because the electorate changed its mind at the last second—it was always going to vote for Pat Roberts or Larry Hogan. Last night was always going to be the not-so-close Republican wave it turned out to be. The polls got it wrong because they missed something, so it took everyone by surprise. Weather forecasters get their predictions wrong because they missed something or there was a factor involved that our mortal weather models couldn't comprehend or process.
Public opinion polling is much like weather forecasting in that both polling and weather forecasts provide us a snapshot in time of an ever-changing variable. That snapshot is usually a good indicator of what will really happen; keep in mind that our three-day forecasts today are as accurate as one-day forecasts were back in the 1980s, and public opinion polls are usually accurate to within a couple of points.
Sometimes they get it wrong. Sometimes it snows when we didn't think it would. Sometimes politicians win when we didn't think they would and close races turn out not to be that close at all. When forecasters get it wrong, they analyze what happened and learn from their errors so they don't make the same mistake again. Polling firms will have to do the same thing.
There is one big difference, though: you can't change the weather, but you can change the outcome of elections. Whether it was this year or 2012 or back in 1994, if you don't like that the polls misled you on the outcome of an election, do something about it next time and actually vote.
[Image: Maryland Governor-Elect Larry Hogan via AP]
