Time for the bloated, flailing political polling industry to crash and start over
What got me going on this was the latest ridiculous high-profile political polling debacles, namely, the California recall and the New Jersey gubernatorial election. But it’s been a long time coming. I’m going to assume that as someone intelligent, knowledgeable, cool, and righteous enough to be visiting this website, you recall the various examples to which I’ll be referring. So I don’t have to pack this full of hyperlinks.
For a while, up to maybe six or seven years ago, I could have been fairly described as a polls junkie during election seasons. I didn’t draw away from that because of any serious loss of confidence in polling integrity or methodology. Rather, I just started to feel like obsessing over every 1-2 point swing probably wasn’t the best use of my time and effort.
That said, I just don’t remember seeing nearly as much truly crap polling, back then. Of course there were all kinds of individual outliers – to further abuse an exceedingly tired phrase, polling is indeed an “inexact science.” And I know the explanation for so much error these days is supposed to be that with contemporary communications, and contemporary attitudes to responding to, say, robo-polling, it’s pretty much impossible to get good samples. But the reality remains that in practically every profession, this kind of repeated blundering would put you out of work, in a hurry. Hence, the title of this.
I’m not sure that everyone knows that political horse-race polls are actually, at least for the most part, “loss leaders” for the people doing them. They’re about marketing. But that doesn’t excuse sheer, repeated incompetence, individual or collective, either.
I stopped accepting Star Tribune polling as legitimate after Wardlow + 8, in 2018. Ditto for KSTP’s, after Tina + 1, just before the election, in 2020. Not just because of those individual polls. Those were culminations, in both cases, of plenty of suspiciously-timed misses, always to the right, over the years. But I’m not here to throw around accusations I can’t prove.
And, after all, in 2020 the big national miss was to the left in the presidential race. While in 2018, the polling was generally so good in congressional races that as I recall at least a couple of the the higher-profile aggregators/predictors were able to nail, right on the head, how many U.S. House seats Democrats would pick up.
So I’m not trying to claim large-scale, consistent bias. Rather, it’s that the numbers are all over the place, as likely to be way wrong as spot-on. A lot of pollsters do seem to be “herding,” presumably so that if they’re way wrong it won’t be more so than many others were. And the really unfortunate side of that, right now, is that when election results are well off of what the polls said, it just provides more ammo for the “rigged elections” crowd. Nobody should be taking that shit seriously, but a lot of people are.
As the linked article notes the priority for most pollsters isn’t trying to always crush it with their accuracy, whatever that might take, which I’m not claiming is easy to do in any case. So they “herd.” And it seems like there are more polling outfits out there, all the time. Thus, bloated and flailing, and in big need of big change that probably won’t come soon.
Thanks for your feedback. If we like what you have to say, it may appear in a future post of reader reactions.