On the right, we have seen a blatant alliance of the Republican Party and right-wing media in an attempt to force out a popular but embarrassing candidate. On the center/left, we have seen newspapers like the New York Times loudly point out that the emperor has no clothes while carefully avoiding the fact that he is standing in the middle of a nudist colony. (The bizarre alliance between Fox News and the New York Times on derailing the Donald Trump candidacy is a fascinating topic that will have to wait for another post.)
On the analytic side, where we are supposed to be above this sort of thing, more and more of the coverage is sliding into drunkard's light post territory: using data for support but not illumination.
I'm going to pick on Five Thirty-eight now, not because they are particularly bad, but because they are particularly good. In many ways, Nate Silver and company have set the standard for analytic political journalism, so if they have started making groupthink mistakes, you can be pretty sure that the pods have now absorbed everybody.
One of Nate Silver's big innovations in political reporting was that he understood the innate complexity of the problem and had the appropriate analytic tools to deal with it. That tradition makes the simplistic approach of this piece by Harry Enten all the more troubling.
Candidates In Donald Trump’s Position Have A Terrible Track RecordI have a lots of issues with this, but to keep things moving, I'll limit this post to just one. If you read this argument carefully (or at least, not carelessly) you will notice a substantial disconnect between thesis and argument. The shift is a fairly standard bit of statistical Three-card Monte. We start out with Trump ahead in Iowa, New Hampshire and nationally. Enten then shows us a long list of candidates going back to the eighties who were ahead in the summer by comparable margins before the primary but lost the nomination anyway, but if you keep your eyes on the queen, you'll notice something funny – none of his “examples” were actually ahead in Iowa, New Hampshire AND nationally.
Polls show Donald Trump leading in the Republican presidential primary. He’s leading nationally. He’s leading in Iowa. He’s leading in New Hampshire. That’s right — Donald Trump may end up winning … “Polling Leader for the Summer of 2015.”
Using Enten's own data, since 1980 it appears that no candidate has ever gone three for three and not gotten the nomination and only one candidate has gone two for three. That would be Hilary Clinton in 2008, and if you'll remember, she had a pretty good run, hardly the stuff of terrible track records.
Enten opens his piece with “Polls show Donald Trump leading in the Republican presidential primary. He’s leading nationally. He’s leading in Iowa. He’s leading in New Hampshire.” According to Enten's own data, for the past thirty-six years, every candidate in that group has gone on to win their party's nomination. In other words, his data point in the opposite direction of his conclusion.
So, am I saying that Trump is likely to be the Republican choice in 2016? Of course not. With the threat of a third party run, he did make things interesting by bringing a gun to a knife fight, but the primaries are a long way off and the election farther still. Between now and then, I strongly suspect that the feedback loop will peter out, the joke will grow old and the negatives will catch up with him.
We don't need polling data to tell us this and, more to the point, we don't have it.
[minor update: just corrected a small typo -- from 'inmate' to 'innate' -- but otherwise everything's the same.]