Tipsheet

Let's Face It: American Political Polling is Broken

In my final pre-election analysis, I touched on a number of subjects and made a handful of predictions.  Some of it is holding up rather well; other aspects, not so much.  I'll revisit all of it once the dust settles, but I'd like to highlight this passage: "We've told you about some of the experimental polling methods that paint a far rosier picture of Trump's standing in the race. Given some track record of success on that front, I don't think they should be dismissed out of hand. If Trump pulls off another epic upset, the teams at Trafalgar and USC will instantly become highly sought-after gurus as the polling world attempts to pick up its shattered pieces." As of this writing, Trump has not pulled off said upset, but he's at least come rather close to doing so.  And it's very clear that there were a wide array of significant polling misfires all across the electoral map.  Again.  

It's true that polling is an imprecise science, and such guesstimates will never be perfect.  That's why margins of error exist.  But broadly speaking, over the last few cycles, how often have pollsters gotten big races really wrong by overestimating a Republican's standing?  I'm sure a small handful of examples must exist (selection bias and exhaustion may be taking a toll on me), but the biggest polling misses that immediately jump to mind involve Democrats' chances being repeatedly overestimated by the 'experts' before being contradicted by actual voters.  GOP chances were underestimated in 2014.  In 2016.  In 2018 (less so).  And now in 2020.  I don't say this lightly, but:

One counter-point is that "prove the fake news media suppression polls wrong!" could be a GOP motivational tool in a way that "voter suppression!" is used by Democrats.  That said, a small rant about this cycle's polling:

RealClearPolitics' Tom Bevan, who was savaged by the poll snobs for deigning to allow the Trafalgars of the world to influence the averages on his site, is making similar points:

The True Believers may not want to hear it, but there is a real problem in American political polling.  It may not be as acute or widespread as some critics suggest, but it's absolutely real.  Will the industry adjust, or stick with 'fake it til you make it' arrogance?  I'll leave you with this:

Hardly perfection, but might you say it's a tad better than the garbage put out by many competitors?  Just a little?  One more: