Who will win the electoral vote on Tuesday, November 8th?
It is not what you say, but how you say it. For several days now, I’ve been told by some how totally wrong I am in my various analyses of the electoral map. Half the naysayers say “But but FiveThirtyEight says this, so you are wrong” and the other half say “No, no, Sam Wang at Princeton says that, so you are wrong!” But all along, we’ve all three been saying something very similar. The difference in how we say it is, Sam Wang says something like “I’ll eat my shorts if Clinton doesn’t win” and I say “I think Clinton will win, but Trump has a small chance.” But really, we have very similar estimates as to what the situation is. And, that is:
1) Hillary Clinton is more likely to win this election than is Donald Trump.
2) Regardless of the initial probability distribution one might have been imagining, this has changed over time so that the chance of a Trump win has been increasing a bit.
3) A number of states are in play, and broadly speaking, the list of states can not be robustly assigned to either candidate is similar.
I myself have been avoiding making specific probability statements because I think that the necessary assumptions to talk about behavior of the electorate out at the margins are unknown or unreliable.
As you know I developed a model that I used during the primaries, that I’m applying to the electoral college vote, with modifications. In short, the model, as used here, reflects whatever polling data are used to seed it, but modifies the outcome to reflect general patterns of behavior. This, I suspect, removes strange results that the polls sometimes give. But it may also miss strange thing the electorate sometimes does. Which is happening in a particular case, for a particular state? Nobody knows. If we knew that, we wouldn’t need to do the actual voting.
So, here, I’m giving you two separate sets of results, initially. First, as in my previous post, a distillation of what the polls themselves are actually saying, using this approach.
First from the polls only:
As noted in the figure, the polls give Secretary Clinton enough electoral votes to win, barely, with Nevada being exactly split between the two candidates. We’ll look at swing states more closely below, but for now, this is my suggestion for the best guess based on the polls. So, if Clinton takes Nevada, she’ll win by 8 electoral votes.
As I had noted earlier, my model should converge on the polls by this point in time, but since there are so many states within a percentage point either way of the 50%-50% line, my model and the polls tend to differ a bit. Overall, my model is more favorable to Clinton because it give her Florida and Nevada.
At this time, this is my best prediction of what I think will happen on Tuesday, unless there are secret unmeasurable forces having to do with unspoken voting behavior or get out the vote efforts.
This result, my model, is very similar to Sam Wang’s result.
One scary possibility is that Trump is gaining ground on Clinton. Looking at just the polls, there was a gaining of ground going on for a while, but it seemed to stop a few days ago. FiveThirtyEight agrees with that. But, what if all the polls end up being one percent off from what they say now, by the time Tuesday comes around? Can Trump then win?
The following moves all the states over by one point, from my modeled results (which I regard as more reliable than the polls) which, oddly, puts Pennsylvania right in the middle. Trump could win. Or Clinton could win.
It has been said that the Democrats may have a ground game, a GOTV plan, that is much superior to that of the Republicans. A good estimate of how that would change things is to add 2.5% to the Democrat’s votes, effectively for the swing states. In this case, Clinton is shown here to do about as well as anyone expected her to do. Don’t expect this, it will never happen, but this is more or less the maximum limit on where Clinton can go. Notice that Trump still takes Texas and Georgia, but may be a bit weak in Georgia.
Finally, by way of summary, here is a map that shows which states are either recognized by one analysis or another as a tossup, or that move back and forth across analyses or over short times scales or, as in the case of Georgia and Colorado, don’t change their color under those conditions but remain very close in percent distribution to those that do. (Note, for Maine, we are only talking about one electoral vote moving back and forth.) Regardless of which column these states actually end up in, they are states you want to watch to measure the strength of each candidate. Obviously, the eastern time zone states will be the most helpful in this regard early in the evening.
