Friday, June 24, 2016

Brexit: When political science does and doesn't make a prediction

I'm going to start with a little story.  There will be a lot of quoted quotes.  I'll dispense with most of the quote marks.  You'll see why.

My grad school mentor, Nelson W. Polsby, was fond of referencing former Washington Post editor Bill Bradlee, who used to say that he wanted subscribers to pick up their copy of the paper every morning, look at the headline, and say, "holy shit!"  (Nelson's wording).  Nelson thought that journalists would need to call up professors and ask, "holy shit, right prof?  Holy shit?"  And our job, according to Nelson, was to tell them, "no, not holy shit.  This is exactly what we expect, and well within the scope of history, and here are the books you should read."

1)  You can see why I dispensed with most of the quote marks.

2)  If you are one of the very few regular readers, you can see either Nelson's influence on me, or why I latched onto him in grad school, and not somebody else.  I do get a kick out of telling people they need to read the books I've already read.

Anyway, let's check the Washington Post web page this morning and... OH HOLY FUCKING SHIT!

Wait a minute.  Let me put on my 'professor hat.'  I'm supposed to tell everyone that this isn't a holy shit moment.  Scope of history, nothing new under the sun, and... oh I can't do it.  Holy fucking shit!  Brexit is happening!

So, Brexit is happening, but Grexit didn't, even though Greece had to close their banks and impose currency controls for days, giving them all of the pain of Grexit with none of the benefits... Ummm...

Time for some admissions.  I made wrong calls in both cases.  Once Greece closed their banks that summer, I thought Grexit was a done deal.  Once the polls started to shift this week, I thought Brexit was moving off the table.  Yeah, maybe this is why I'm an Americanist.

But there is more to it, and the title of the post hints at it.  With Brexit, we were doing something that we call, in social science, trying to predict outside the bounds of our data set.  A nationwide referendum on leaving the EU just isn't something we had seen before.  Of course, we had seen the Scotland vote on leaving the UK, and they stayed, but not the same thing.  This is a unique event.  The problem with a unique event is that we have no models.

Contrast that with presidential general elections (as opposed to primaries, which, see my comments here).  Same thing, over and over again, so he have good models.  Brexit?  No history.  No history means no models.  No models means uncharted territory.

Yes, I was surprised.  The polls were close, but moving against Brexit.

Still, sorry Nelson, but this is a "holy shit" moment.

No comments:

Post a Comment