This is the third in a series of posts on what has gone wrong in Political Science this year. The first defined "Type A Trump denialism" as the persistent belief that, despite all evidence to the country, Trump could never win. No, it's not over, but we all clearly underestimated The Donald. In the second post, I began describing Type B denialism, which is refusing to admit that the models that led us to underestimate Trump were the problem. I began with The Party Decides, which I never believed. Now it is time to address a model that I did believe, coming from Sides & Vavreck's The Gamble. That one, I assign.
Sides & Vavreck performed a detailed and careful analysis of what happened in 2012, both the primary and general elections. Obviously, I will focus on the Republican nomination contest, which was, at the time, the most entertaining ever. Obviously, it has now been upstaged.
Their basic argument about the 2012 Republican nomination contest was that the field consisted of Mitt Romney, and a group of people whom few could imagine as president, like my personal favorite, Michele Bachman, Herman Cain, Newt Gingrich, Rick don't-Google-me Santorum, and others. Each of the non-Romneys had a boomlet in the polls, driven by media attention. However, once the media turned to scrutiny, each of their campaigns collapsed due to the intrinsic weaknesses of the candidates, leaving Romney the eventual winner almost by default. It fit the data, made substantive sense, and fit within my expectations that year. Once Tim Pawlenty ran out of money and Rick Perry did his Rick Perry thing on the debate state, there were no other credible candidates, and I discounted each of the non-Romney boomlets as ephemeral. The Gamble confirmed my expectations, so it must have been right, right?
Fast forward to 2015. Donald Trump enters the race, looking very much like the kind of candidate who would get a boomlet from media attention, whither under media scrutiny, and then whine. Trump takes an early lead in the polls, and it looks very much like the start of the discovery-scrutiny-decline model from 2012. John Sides begins writing the obvious things at The Monkey Cage. See, for example, here, here, and here.
The problem, of course, is that media attention shifted to scrutiny time and again. There were Trump's comments about Mexican rapists, his attacks on John McCain for having been held captive in Vietnam, his comments about Megyn Kelly's menstrual cycles... His numbers never went down. How many times has Trump's candidacy been declared dead? It has happened so many times that it is a joke at The Onion.
I freely admit here that I wrote off Trump's early lead based on The Gamble. I expected his campaign to collapse based on The Gamble. Several times. There must come a point, though, at which we acknowledge that Trump's campaign has withstood more scrutiny than any of the 2012 candidates subject to the discovery-scrutiny-decline model.
What do we learn? Not precisely that the cycle is wrong. After all, it explained Ben Carson this year. Rather, Trump can withstand scrutiny like an established, serious politician. Trump's polling lead was never simply about disinterested survey respondents casually stating support for the candidate whose name they heard most frequently. Instead, Trump's support was more substantive, and hence less susceptible to scrutiny. What, precisely, keeps Trump's supporters in his camp? A topic for another post. What distinguished Trump from Herman Cain, though, is that everyone already knew who Trump is, and had formed opinions about him. For Herman Cain, not so much. Thus, when media attention turned negative for Cain, he crumbled since his support was basically illusory. Trump has a base of support among Republican primary voters.
Where we went wrong, then, was in thinking that Trump was Herman Cain. He isn't. He is The Donald. To whom can we compare his candidacy? Umm... We can't. We are outside the bounds of our data set. The fact that we can go so far from historical precedent, though, is exactly my point about why we need to reassess our models. Maybe Trump is just the black swan that cannot be explained, and years from now, he will be remembered as an anomaly. Right now, though, we need to acknowledge that none of our models are holding up very well.
Once again, the standard disclaimer. The race isn't over yet, and while the betting odds keep holding Trump at or below 50% at PredictWise, the fact that he is still standing, and still leading means we need to reassess the models that predicted his doom so many times. Even if he loses, he outperformed by so much that our models require scrutiny.
I'll still probably assign The Gamble, but this is a problem...