Monday, October 17, 2016

The Arizona throw down

As I have mentioned before, there are a lot of election forecasting models this election. 538 seems to be the most popular, partly because it's run by the gay wizard of the 2012 election but also because the site is just better (more interactive and more accessible) than the others. I probably do check 538 the most, but I also try to check in with the others on a regular basis. Luckily, the Upshot has a handy comparison chart of the nine most prominent models.

The models have mostly all pointed in the same direction throughout. Currently, they all point to a Clinton win (as they have almost every day since they each launched). So they are not that far off from one another. But there are a few differences. While the ones that express their prediction in terms of a percentage all put Clinton's victory at between 87 and 98% likely, the ones that don't all say the likely presidential outcome is "lean Dem." "Lean Dem"?!?! how can that be consistent with the others that are giving chances in the 90s or close to it?

My current test case is Arizona. AZ is a weird outlier for the 538 model. As I write this, that model is giving AZ a 53.6% chance of being won by Clinton and the model has shown the state to be leaning Clinton for the past week or so. The other models all have Trump favored to take the state. The Upshot gives Trump an 84% chance, Daily Kos 64% Trump, Huffington Post 83% Trump, Predictwise 63% Trump, Princeton Election Consortium 96% Trump, and the Cook Report, Rothenberg & Gonzales, and Larry Sabato all have AZ as leaning Republican. It is interesting that the biggest contrast for AZ is between Nate Silver (aka 538) and Sam Wang (Princeton Election Consortium) because those two are longstanding rivals.

Maybe the models will start projecting AZ closer as election day approaches. And I realize that even Nate Silver is not predicting a sure thing in AZ--he is still giving Trump just short of a 50-50 chance of carrying the state. But who cares. I have the right to be arbitrary and I am arbitrarily picking what happens to AZ on November 8 to be my test case for which model is correct. Assuming they all pick the overall outcome correctly, if Clinton wins AZ, Nate Silver is the official wizard of this election. If Trump wins AZ, then Wang will get the crown (fuck the other 7). I'm sticking with this even if 538 shifts and starts predicting AZ will go for Trump down the line.