Wednesday, May 31, 2017

Poll Arising Nation, Con't

The after-action report on 2016 polling was simple: national polls were correct (Clinton by 2%) but individual state polls were wrong, in some cases badly so.  Nate Cohn at the NY Times Upshot crew takes a look and theorizes that the big problems in state polling are fixable to some extent.

Nearly seven months after the presidential election, pollsters are still trying to answer a question that has rattled trust in their profession: Why did pre-election polls show Hillary Clinton leading Donald J. Trump in the battleground states that decided the presidency? Is political polling fundamentally broken? Or were the errors understandable and correctable? 
At their annual conference in New Orleans this month, polling experts were inching toward the latter, more optimistic explanation. And there is mounting evidence to support their view. 
At least three key types of error have emerged as likely contributors to the pro-Clinton bias in pre-election surveys. Undecided voters broke for Mr. Trump in the final days of the race, or in the voting booth. Turnout among Mr. Trump’s supporters was somewhat higher than expected. And state polls, in particular, understated Mr. Trump’s support in the decisive Rust Belt region, in part because those surveys did not adjust for the educational composition of the electorate — a key to the 2016 race. 
Some of these errors will be easier to fix than others. But all of them are good news for pollsters and others who depend on political surveys. 
It might seem strange to argue that the polls could miss the result of an election and could still be trusted in the future. But there are some kinds of polling errors that pollsters can accept, even if the public never will. 
There’s nothing pollsters can do, for example, if undecided voters break for one candidate in the final hours. Even an error that puts more blame on the pollsters might be acceptable, provided it can be fixed. 
Errors have happened enough in past elections to know that an upset was well within the realm of possibility in 2016. The Upshot model estimated that a polling misfire was about as likely as a baseball strikeout or a missed midrange field goal in football. It’s not pretty, but it happens and will happen again, and a team wouldn’t release a batter or a kicker because of a strikeout or a missed kick.

The reality is that confidence in polling is quite frankly nowhere near 100%, and is really more like 75 or 80% at most, even in a best case scenario.   Nate Silver famously gave Trump about a one in 3 chance of winning, and Trump did just that.

Of course Russian interference that fueled James Comey's political bombshells had a great effect on this "last-minute break" to Trump too, and of course the pollsters missed that.

I'm much less worried about pollsters than I am having free and fair elections in 2018 or 2020 with the demonstrably broken election system we have now, and Trump running that system.

No comments:

Post a Comment