cincinnatus_c (
cincinnatus_c) wrote2017-06-16 10:51 pm
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
Like a shoal of fish in a tank going nowhere
Did you know: in the 2017 general election, the British Conservatives got their highest share of the popular vote since at least 1983 and possibly 1979, had their greatest increase in share of the popular vote over the previous election since 1979, and [ETA] had the biggest increase in vote share of a sitting Conservative government since 1924. What a triumph for Theresa May, eh! No wonder she looks tired. (Gah.)
Anyway, I would never have noticed any of that if not for this article on fivethirtyeight.com, about why the election result shouldn't have been as surprising as lots of people took it to be, and why the polls generally missed in the direction they did--i.e., they generally had a bigger gap between the Conservatives and Labor than there was in the actual vote. The upshot is they missed the way they did because the pollsters overcorrected from last time, when they underestimated the Conservative vote share by underestimating Conservative voter turnout. Which again goes to show how opaque and misleading the publicly reported poll results are, because they are not actually poll results but inferences from the poll results and some secret sauce to what the results of an election would be. If polls were just giving you poll results it would be nonsense to say that they "missed"--it would be like mistaking the current weather conditions report for the weather forecast.
Which brings me back to something I've been meaning to note about what I was saying about the flukiness of the last American election. I said that Nate Silver gave Trump about a 30% of winning, so, you run the election twice more and there's a bit better than even chance that he wins one of those, too. But I don't think that's Silver's view. I think (e.g. from his framing of the issue in his series about why Trump outperformed everyone's expectations) Silver's view is that the 0.3 probability was an epistemic probability and not an ontological probability (or however statisticians would put that)--i.e., we could say with 70% confidence that the actual probability of Trump winning was 0. To put the point another way, I think Silver thinks the result was basically determinate, and perfect polls could not have failed to predict the result perfectly (or, you know, the chances of their failing would have been practically insignificant). So if you run the election twice more--I mean the very same election on the very same day, not a Holy Shit We Elected Donald Trump do-over--the results would be more or less exactly the same. Of course you're dealing with such large numbers that it seems like there's very little wiggle room for anything to go differently--I mean for people to change their minds between the last polls and the election in a way that actually swings the result one way or another. Then again, when you look at the actual swing states, you might start to think there is enough wiggle room to make a difference. Anyway, I'm not sure I know enough to know how to think about this, but I have the feeling that there's something like a basic and more or less unresolveable problem about human freedom here. (And once again I'm reminded of Hume saying that you're better off trying to persuade the stone walls than you are trying to persuade the guard to let you out of prison.)
And in other loose ends ... it probably makes sense that mythology comes after (fantastical) history, eh? I mean that history (fantastical or otherwise) gets mythologized. This is a big part of The Diviners--the supersession of (Scottish(-Canadian) and Metis) history by mythology: Morag comes to realize that there's no point visiting the ancestral places because it's the mythology that matters and not the actual history. And so that question (it suddenly just occurs to me now) that recurs through the book, how far back it all starts, how far back anything goes, turns out to be the wrong question, or at least to have a different kind of answer. This wasn't set in motion, let alone set in stone, by all that back there (or at least, insofar as it was, that's not the important thing), but it is what it is because of how the story of all that is told....
Currently at Havelock: 19.4. High today: 29.6.
Anyway, I would never have noticed any of that if not for this article on fivethirtyeight.com, about why the election result shouldn't have been as surprising as lots of people took it to be, and why the polls generally missed in the direction they did--i.e., they generally had a bigger gap between the Conservatives and Labor than there was in the actual vote. The upshot is they missed the way they did because the pollsters overcorrected from last time, when they underestimated the Conservative vote share by underestimating Conservative voter turnout. Which again goes to show how opaque and misleading the publicly reported poll results are, because they are not actually poll results but inferences from the poll results and some secret sauce to what the results of an election would be. If polls were just giving you poll results it would be nonsense to say that they "missed"--it would be like mistaking the current weather conditions report for the weather forecast.
Which brings me back to something I've been meaning to note about what I was saying about the flukiness of the last American election. I said that Nate Silver gave Trump about a 30% of winning, so, you run the election twice more and there's a bit better than even chance that he wins one of those, too. But I don't think that's Silver's view. I think (e.g. from his framing of the issue in his series about why Trump outperformed everyone's expectations) Silver's view is that the 0.3 probability was an epistemic probability and not an ontological probability (or however statisticians would put that)--i.e., we could say with 70% confidence that the actual probability of Trump winning was 0. To put the point another way, I think Silver thinks the result was basically determinate, and perfect polls could not have failed to predict the result perfectly (or, you know, the chances of their failing would have been practically insignificant). So if you run the election twice more--I mean the very same election on the very same day, not a Holy Shit We Elected Donald Trump do-over--the results would be more or less exactly the same. Of course you're dealing with such large numbers that it seems like there's very little wiggle room for anything to go differently--I mean for people to change their minds between the last polls and the election in a way that actually swings the result one way or another. Then again, when you look at the actual swing states, you might start to think there is enough wiggle room to make a difference. Anyway, I'm not sure I know enough to know how to think about this, but I have the feeling that there's something like a basic and more or less unresolveable problem about human freedom here. (And once again I'm reminded of Hume saying that you're better off trying to persuade the stone walls than you are trying to persuade the guard to let you out of prison.)
And in other loose ends ... it probably makes sense that mythology comes after (fantastical) history, eh? I mean that history (fantastical or otherwise) gets mythologized. This is a big part of The Diviners--the supersession of (Scottish(-Canadian) and Metis) history by mythology: Morag comes to realize that there's no point visiting the ancestral places because it's the mythology that matters and not the actual history. And so that question (it suddenly just occurs to me now) that recurs through the book, how far back it all starts, how far back anything goes, turns out to be the wrong question, or at least to have a different kind of answer. This wasn't set in motion, let alone set in stone, by all that back there (or at least, insofar as it was, that's not the important thing), but it is what it is because of how the story of all that is told....
Currently at Havelock: 19.4. High today: 29.6.