Boy, we certainly didn’t see that coming. I’m talking about Donald Trump’s (seemingly) unlikely win in the 2016 presidential election, of course. You may recall from last year’s UF/PC newsletter the essay by Michael Martinez on forecasting models which, at the time, pointed to a narrow win for Hillary Clinton – except for the econometric models (the most interesting to me because they predict outcomes months in advance), which reached exactly the opposite conclusion.
Bully for the forecasting models, right? Wrong! Almost all of these models are designed to predict the winner of the two-party popular vote. And guess who that was. Yep, Hillary Clinton. Perhaps the most famous presidential forecaster is Allan Lichtman, a history professor at American University whose so-called “13 keys” model is said to have accurately pre- or post-dicted accurately every election since 1860. Lichtman has received considerable media attention recently for having correctly anticipated Trump’s win in 2016 – except that like the other models, his is focused on the popular vote. Having claimed credit for being accurate with Al Gore’s popular vote win in 2000, I don’t see how he can now take credit for Trump’s electoral vote win in 2016. To put it simply, any model that forecast a Trump popular vote win was wrong. Period.
Which takes us to the various media polls in 2016, almost all of which (even as Election Day neared) anticipated a Clinton victory. While the pollsters have taken a lot of heat for this, especially in Trumpland, the fact is that they nailed it. At least in terms of the national vote, that is. Where the polls broke down was at the state level, where projections in places like Pennsylvania, Michigan, and Wisconsin either (a) did not catch what appears to have been a late pro-Republican surge or (b) employed sampling methods that missed a lot of Trump supporters in those states to begin with. At the national level, however, their numbers were pretty much on target.
I don’t bring any of this up with the intent of trying to rehabilitate the reputation of either presidential forecasters or (national) pollsters. They did their job in 2016, and for the most part did it well. But what about myself and my colleagues in academia, very few of whom believed that Donald Trump had a snowball’s chance of being elected president – a conclusion based in part on our almost universal conviction that, never mind what might happen in November, Trump’s chances of being the Republican nominee were basically nil. (Jeb Bush, anyone?)
I attended a panel at the 2016 American Political Science Association meetings in Philadelphia that focused on the presidential nominating process and posed the question, Was 2016 an outlier, or the new normal? For both nominations and the general election (and looking beyond just the presidency), that’s what public opinion and voting behavior scholars are going to be asking themselves for at least the next four years.