As we hit the Eve of Election (great, “Eve of Destruction” is now in my head...), thought it might be worthwhile to mention a couple of things to be aware of as we’re inundated by numbers in these final hours.
I’ve worked in stats for sports for many years, so while I don’t think I can call myself a “statistician” in the professional sense of the word, I do a lot with numbers and probably have a better grasp than most. In particular, I’d like to make some points about percentages, of which we’ll be seeing a ton starting in just over 24 hours. I have two general points:
1 — Don’t get freaked by changes in vote percentages in individual races.
2 — Nobody will be winning 88% of an election.
For point number one, I have to say this is a major pet peeve, something I even see our front pagers do regularly on Election Nights.
Just because a percentage gap is narrowing, the race is not necessarily getting “closer.”
I’ll explain this as best I can. Ten votes are counted in a local election; Candidate A receives seven, and Candidate B receives three. Thus:
Candidate A = 70%
Candidate B = 30%
With me so far? Okay, so 10 more votes are counted, with A getting six and B getting four. Total now is A 13, B 7.
Candidate A = 65%
Candidate B = 35%
“OH MY GOD, IT’S GETTING CLOSER!!!” No, it’s not. Candidate A has actually increased her (sic) lead from four votes to six. It can be more difficult, but look at the actual vote totals. I’ve seen proclamations about races getting closer when the leader has actually added thousands of votes to an advantage. Obviously, the more votes are counted, the less the percentages will be subject to movement.
=============
My footing is a little less sure on point two, so anyone who can either clarify or refute it is welcome.
Unlike polling, the final result will be either Hillary winning (100%) or not (0%).
Polling aggregators, much like the polls themselves, are meant to give an idea of where things stand at any given point in time. As many have pointed out, in an Electoral College system, national polling figures are basically useless (see: Gore, Al).
State-by-state figures and aggregators seem to exist to gather more of those numbers to give a “big picture” of the information. But here’s the thing; as should be obvious, the vote percentages are not equal to the probability of victory. For example, if someone is polling at 70%, there is not a 70% chance of that person winning — even allowing for margin of error, it is virtually 100% certain.
I guess my point here is, in an election like this, where every single aggregator still has Hillary with a significant chance of winning, I would expect the victory probability numbers to start approaching 100% (like Sam Wang, or the comments Markos has made) as the election nears. An aggregate on November 7 should not be the same as one on September 7.
Especially in the context of Nate Silver’s numbers, it’s worth noting that, if Hillary wins, he will be as “correct” as Wang. I’m not sure exactly how one will say that his percentage was better or worse, since they use different methodologies and can both stick to them and claim to be right.
I’m trying to quantify why I have such a high confidence rate based on the numbers I’m seeing — I can be a hand-wringer myself, but like many it just strikes me that the margin of error would need to go against Hillary in the vast majority of circumstances, which just seems highly improbable.
Stat geek session over. I’ll be watching returns throughout the night as you will, and will be lurking about The Great Orange Satan throughout. (Wait… considering the Republican nominee, do we need a new nickname now?!?!)