FiveThirtyEight, to their credit, went and did an analysis of what was easily the most embarrassing mistake of Tuesday night: to wit, why Bernie Sanders won the Michigan Democratic primary when everybody was confidently saying Hillary Clinton by a lot*. It's a very interesting read (short version: bad assumptions, coupled with a poor polling environment, sprinkled with a bunch of idiots who apparently decided that strategic voting for a Republican was safe enough because hey, Hillary has this**), but I was struck by this passage: "“Polls on the Republican race, including ours, were generally OK,” said Barbara Carvalho of the Marist Poll." Ms. Carvalho is referring to the Republican Michigan primary, and: no. No, the polls were not generally OK.
The above is, of course, the March polling from Real Clear Politics survey of the Michigan Republican primary, and let's be honest: as our own streiff keeps telling us, you could get better results from winos. The RCP average only manages to get close to Donald Trump's actual result because of that God-awful ARG poll that showed John Kasich winning, and a Monmouth poll that, uniquely, turned out to be very close. Otherwise: the pollsters persistently kept over-estimating Trump's lead (this is a common problem) and underestimating Ted Cruz's numbers. They likewise mostly all got the eventual order wrong, which is actually pretty significant in this case: Kasich was aiming for a strong second, and instead ended up in third. And I can't help but notice that the only Marist poll listed there overestimated Trump and had Marco Rubio in third place, not Kasich.
We are getting perilously close to a situation where there's no point in looking at state-level polling at all. It was bad enough with the dog's breakfast over here on the Republican side; but if they can't poll accurately on the Democratic one - a race with only two candidates and an almost palpable sense of tedium about the whole thing - then why bother? At this point, the results are only right by accident***. And, given that we apparently may be fighting this election out all the way to the convention; 'only right by accident' is not going to cut it.
Which is why I'm harping on this. The pollsters themselves don't really seem to be overly concerned that they keep missing the results, if that FiveThirtyEight article is any indication. They probably should be, given that this is the second election cycle in a row where they are consistently getting results wrong...
*Including me, if anybody had asked.
**I freely admit that this is merely one possible interpretation of why 7% of Republican voters in the primary were actually Democrats. By the way: state Republican parties need to close their primaries and caucuses, and get rid of same-day registration, too. This nonsense has gone on long enough.
***We care about stuff like the order and final percentages because it's a way to check that the pollster is doing something more consistently accurate and descriptive than somebody who is simply guessing who'll win the race. Because guessing works great... right up to the point where you have something like the Democratic Michigan primary. Then you get blindsided. And even then at least the pollsters have a theoretical ability to analyze what they did, and what they did wrong. But you can't cross-check a gut feeling.