Over the next year and (almost) and half, you're going to hear a lot of historical "rules" from 538, the Upshot, etc. Here's a counter example to keep in mind when they start to sound persuasive.

 The pattern was clearer than almost any of the rules that were dredged up by data journalists in the past couple of elections to support this or that prediction. This is a forty year run with plenty of examples of candidates with and without the trait and 100% accuracy.

Then it just stopped. The lesson here is that even with the most convincing historical precedent based argument, you shouldn't assume the future should look like the past.

Monday, November 18, 2019

Fun with Political Trivia

This picks up on a recent thread (telling which one might be too much of a clue). The ones and zeros represent a trait of Democratic candidates from 1964 to 2004. Take a look and think about it for a moment. Here's a hint, the trait is something associated with each man well before he ran for president.

Johnson           1
Humphrey       0
McGovern       0      
Carter              1           
Mondale          0           
Dukakis           0           
Clinton            1                 
Gore                1                      
Kerry               0


As you might have guessed, the relationship between this trait and the popular vote didn't hold in the previous or following elections. The trait is not at all obscure. It was well known at the time and figured prominently into their political personas, This is not a trick question.

Put bluntly, for decades, every time the Democrats ran a Southerner, they won the popular vote (and possibly the electoral college, depending on your thoughts about Bush v. Gore). Every time they ran a Northerner they lost.

Currently, those candidates who are being held up as electable (Biden, Mayor Pete, etc) are all white men from the North. We already pointed out that a white male from either party has not won the popular vote since 2004. Add Northern to the mix and you have to go back to 1960 to find a successful Democratic candidate.                 

I can already hear some of you out there trying to reduce this to the ideological spectrum, but you're likely to run into problems. For starters, though there's plenty of JFK fan fiction to the contrary, it's difficult to argue that LBJ was a conservative Democrat, and even putting him aside, the relationship isn't that strong. There are a few liberal Southern Democrats and more than a few conservative Northern Democrats. Even in the primaries in question, it's not clear that the Southerners were the most conservative candidates on at least some issues. Jerry Brown was arguably more fiscally conservative than either Carter or Clinton while you didn't get further to the right on communism than Scoop Jackson.

You could make more headway chipping away at special cases. LBJ was already in office and he had the martyrdom of his predecessor. Humphrey probably would have won if not for Vietnam. Carter might have owed his victory to Watergate. Reagan was unstoppable in 1984. We could keep going which is kind of the point.

This kind of historical pattern can be useful as an existence proof (___ can happen if ____ has happened before) and conversation starter. It can be great for suggesting hypotheses (a paper on the role of regional tensions and biases in presidential elections comes to mind), but with n=9, even with  r-squared=1, I'm not about to go in the prediction business. I have no trouble imagining Buttigieg or Biden becoming president (my concerns with Joe are actuarial more than electoral).

Almost by definition, each presidential election is unique and you shouldn't assume that even strong patterns are going to hold in the future. Unfortunately, much of what passes for early campaign political analysis in supposedly data driven sites like the Upshot is often little more than a string of n=1, not all that apt anecdotes. Remember how Trump couldn't get the nomination because Herman Cain rose then faded and Trump was supposedly just like Cain? It was, for a while, one of the most popular arguments among political data journalists, appearing multiple times in the NYT alone.

For the most part, the same people who made those ridiculous points insisting that Trump couldn't get the nomination (let alone the presidency) are still around, still using the same flawed reasoning, and still having far too many people take them seriously.