}

Tuesday, November 22, 2016

Professor predicts impeachment


The CBS News video above features Allan Lichtman, a professor at American University, who has correctly predicted very presidential election since 1984, including this year. Now he says Don will be impeached. Is that valid?

My first thought when I watched the video was that this is an example of why most academics aren’t pundits on television: What makes a person good for one field does not necessarily make them good for the other (a few notable exceptions notwithstanding). Even so, he does manage to stay on point, which is more than most pundits can manage.

His “prediction” about Don being impeached is actually more of a hunch, as he admits, but one that’s at least reasonable—though more likely to be probable. It seems to me the main motivation for Republicans to do this would be if Don breaks with them and prevents them from enacting their agenda, or especially if his antics threaten to cost them seats in the midterm elections. The Republican Party, after all, cares about nothing other than holding power.

One thing I did think was hilarious was both of them dissing polling, as if the professor has some unique magical way of always being right. The polls were wrong this year—obviously—but they have been very accurate in other years. Sooner or later, the professor will get it wrong, too: He’s a human predicting the behaviour of other humans, and that means his (admittedly) long streak of correctly calling the winner will inevitably end. He’s been very lucky, but he’s not infallible any more than pollsters are always fallible. We’re talking rationality, after all, not superstition or magic.

So, I think his viewpoint is interesting. Some of what he had to say seemed pretty self-evidently true, and some of it was debatable—and that’s just his talking about why he was right. It’ll be even more interesting to see if his luck holds out for another election or not.

5 comments:

Arthur Schenck (AmeriNZ) said...

Maybe. But it's also possible—likely, I'd argue—that polling companies will return to normal. I say that first because the companies will be analysing what went wrong in order to fix their models and methods—their income depends on being right, so they have a strong incentive to get it right every time (or as close as possible to that goal). Second, the only reason people lie to pollsters is when they're embarrassed to admit who they're voting for and don't want to reveal their actual choice "publicly", a kind of more generalised version of the “Bradley Effect.

rogerogreen said...

I'll tell you that the Census Bureau has been getting a lot of pushback on people answering their surveys. Some are undocumented workers, some are otherwise marginalized, but they don't trust the govt

rogerogreen said...

But I think the polls will be less correct, as people avoid landlines, avoid pollsters, lie to pollsters, et al

Arthur Schenck (AmeriNZ) said...

Maybe. But it's also possible—likely, I'd argue—that polling companies will return to normal. I say that first because the companies will be analysing what went wrong in order to fix their models and methods—their income depends on being right, so they have a strong incentive to get it right every time (or as close as possible to that goal). Second, the only reason people lie to pollsters is when they're embarrassed to admit who they're voting for and don't want to reveal their actual choice "publicly", a kind of more generalised version of the “Bradley Effect".

https://en.wikipedia.org/wiki/Bradley_effect

rogerogreen said...

And this: http://billmoyers.com/story/tv-news-long-dark-night-soul/ underestimating the Trump vote. Why? The best guess is that so many in their samples are refusing to play along that the pollsters, receiving fewer data to deal with, have to tinker with their models, hoping to compensate for the refusals.