This Election Is Probably Over
Nate Cohn says that Hillary Clinton has a 74% chance of winning:
The Upshot’s elections model suggests that Hillary Clinton is favored to win the presidency, based on the latest state and national polls. A victory by Mr. Trump remains quite possible: Mrs. Clinton’s chance of losing is about the same as the probability that an N.F.L. kicker misses a 45-yard field goal.
I’m not a fan of this analogy. A field goal is a one-off event. Any random thing can change it entirely. The kicker gets distracted for a split second. There may be a gust of wind. He can be off by a split second. Elections, on the other hand, are far less likely to change due to some fractional thing. They are considerably more fundamental. It depends not on the performance of one or a handful of people, but the collective decisions of millions.
To clarify, I don’t take issue with the 74% figure in and of itself. I’m not arguing that it’s actually 50% or 95%, but with the framing. We know the likelihood of a field goal by accumulated data with a very large sample set. However, any given field goal is subject to an infinite number of factors. A field goal can be missed because the kicker got distracted for a second. When a kicker misses a field goal, you can be relatively confident that if given another opportunity, the outcome could be entirely different. It is rarely the case that there’s just nothing a kicker can do about a field goal outcome. Well, there is also the offensive line and the defense, but we’re still dealing with the performance of a very limited number of individuals. That makes results extremely variable. A field goal can be missed because the kicker got distracted or a defender jumped just a quarter of an inch higher than usual or an offensive lineman slipped or the center didn’t snap the ball correctly.
In 2012, there is no single thing that Mitt Romney could have done to win the election. There is likely nothing the GOP could have done. Obama could have completely tanked it with a rousing speech about how he hates America I suppose, but that’s about it. The same applies to 2008, except moreso. I’m not saying that Nothing Matters, as clearly the economic collapse affected the outcome. But short of that catastrophic event, the impact of any particular thing would have been marginal in an election where the margins weren’t very marginal at all. The election was over not just when the Lehman Brothers went under, but it was over before then. We just didn’t know it yet.
If we had run the 2008 election over and over again, though, the outcome would have been the same. The game-changing event would have occurred and changed the game in almost every run of the election. So while Nate Silver may have given McCain a 10% chance at winning the election before Lehman happened, if we were to run the election 100 times, McCain wouldn’t have won 10 times unless Lehman had only a 90% chance of happening. But Lehman was unknown at that point. If there was a 10% chance, it was based on an uncertainty of things that overwhelmingly were already in play but just not known about. None of it was as spontaneous as a distracted kicker, a tripping center, or a defender high jump.
Further, when it comes to elections, we’re dealing with a much smaller sample set. The last election was almost four years ago, and both of the primary variables (the candidates) have changed. Whereas when we analyze the likelihood of a kicker getting it through the uprights we’re dealing with a sample set of actual instances, with elections we’re basing it on models that are projections and predictions. At this stage of the game, that’s where most of the uncertainty lies. So if we talk about a 75% chance of Clinton winning, we’re talking much less about another Lehman incident and much more about the models being wrong. That’s the probability we’re assessing.
This wasn’t always the case. I do not believe that elections are decided before they begin. Things matter. I have been relatively confident of a Clinton victory since the nominees were decided, but there were a lot of things that could have shaken my confidence, including things the campaigns did. However, with each passing week, the band of probabilities has contracted. Hillary’s GOTV is fundamentally in place, as is Trump’s lack of one. More and more voters have made up their minds. The amount of time left for a major terrorist attack on US soil has gone down. The likelihood of another Lehman has gone down. The chances of some new revelation about Clinton or Trump have gone down. There is simply less time for it to happen. And if it does happen, it’s likely something planned and therefore as inevitable as Lehman. If Hillary Clinton is sitting on something especially damning of Trump, or vice-versa, then that October Surprise occurs in at least 95 of 100 potential runs of the election.
None of this is to say that any of us know how it’s going to end. While the width of the band of possibilities has been contracting, we still don’t know the baseline. Hillary Clinton could do better than the polls suggest, as well as they do, or worse. If that 26% comes to fruition, the most likely explanation will be that the polls failed to capture the public sentiment. Any mind-changes that occur from here to election day are likely to be marginal and pre-ordained. I don’t believe there are enough of them to change the outcome of the election from a Clinton victory. So if Trump wins, it’s likely to be because the baseline was wrong. For example, that the LA Times had the best model all along.
As it happens, I believe the polls will be off by a couple of points. Except I believe it’s more likely to be in the other direction, for a variety of reasons. By election day, I expect FiveThirtyEight to have Clinton ahead by three, RCP to have her ahead by four, Huffington Post to have her ahead by five, and for her to actually win by six. That’s a guess, but we’ll use that as my personal baseline. Six point margin, plus or minus a couple points.
Now, a margin of 4-8 points is hardly a bold prediction, so I expect to be awarded no points if that’s how things shake out (unless it’s really close to six). Which is the problem with my theory as presented. It’s hard to confirm or disprove. But not impossible. If Donald Trump keeps screwing around and his support erodes in the polls and those polls are confirmed on election day, then I will have been wrong. Likewise, if Trump goes on his best behavior, the gap closes and it’s a very competitive election, I may have been wrong. If neither of these things happen, though, it won’t necessarily confirm my view. A lot of it is simply going to depend on the particulars of the polls and Trump’s (and Clinton’s) behavior. So basically, “I’ll know it when I see it,” which is admittedly unsatisfying.
While Tod declared the election over in June, I didn’t. And I’ve been frustrated by the creaks and trips of the Clinton campaign. I’ve never been worried about a loss, but I still believed that things mattered. For the margin, if nothing else. I don’t want Clinton to get cocky because you never know and there’s no reason to tempt fate, but the election of “LOL Nothing Matters” has finally reached, or after the first debate has reached, the point where I believe it doesn’t matter. Everything likely to have an effect already has, or is at least already coming down the pike. Stuff mattering is predicated on a closer election than the one that exists.
Assuming, of course, no strokes, heart attacks, or assassination attempts.
Feature Image by DonkeyHotey