In Defense of Nate Silver -- and Math

Bernie supporters are lining up to deny math due to Michigan's shocking and historic primary results...
Avatar:
Author:
Publish date:
Social count:
602
Bernie supporters are lining up to deny math due to Michigan's shocking and historic primary results...

More than a few Bernie Sanders supporters have popped into my various social media mentions to inform me that Nate Silver and FiveThirtyEight.com were brutally wrong on the Michigan primary. 

Obvious analysis is obvious. 

The polling was historically wrong -- perhaps the most inaccurate statistical forecast since New Hampshire 1984, or, in recent memory, New Hampshire 2008 when Hillary Clinton shockingly overtook Barack Obama's robust polling lead to win the primary there, portending a long and contentious haul to securing a presumptive nominee.

I can't possibly disagree with the conclusion that the polling was wrong, and therefore Nate's polling average and forecast were subsequently wrong, too. I'll also eat a fair amount of crow for reporting on the math. The truth is, however, FiveThirtyEight.com isn't a polling outfit and, as far as I know, has never put an actual poll of its own in the field. Thus, Nate and his crew rely upon outside pollsters, each of which are ranked and weighted based on accuracy. This time, they were all wrong. Every poll was glaringly wrong. Big time. As a result, Nate was wrong, too.

Indeed, the most inaccurate poll in the last several days, conducted by Mitchell Research, also happened to carry the most weight. Mitchell showed Hillary with a 37 point margin going into the primary. Wildly wrong.

Michigan polls

The most recent Michigan polls, per FiveThirtyEight.

Some of the reasons for the unnervingly inaccurate polling being discussed today include the following:

1) The open primary in Michigan allowed the allegedly complacent Hillary supporters to vote in the GOP primary as a means of thwarting Trump. It also flummoxed pollsters due to an influx of unaffiliated voters siding with Bernie. According to exit polls, Clinton easily won among registered Democrats alone, 60 to 40 percent, give or take. Bernie won among independent voters.

2) Turnout far exceeded 2008, and therefore the model was universally wrong. The Washington Post's Philip Bump reminded us that 2008's Michigan primary was an outlier due to the state moving the primary to January and subsequently being punished by the party. Both Barack Obama and John Edwards, in solidarity with the DNC, chose to not compete in the state and therefore Hillary was the only Democrat on the ballot. Turnout was dampened as a result. This year, however, turnout was much more energized. If they hadn't already, pollsters should have factored for this disparity.

3) More African-Americans voted for Bernie than in previous contests. Bernie's margin with black voters jumped from around 17 or 20 percent elsewhere to 30 percent in Michigan. We can assume this is the manifestation of Bernie's strength and momentum.

Bernie won this one fair and square, and his supporters should be rightfully thrilled with the results. 

What's pissing me off this morning are Democrats who are proudly burying Nate Silver's math, suggesting his reign as the gold standard is over and so they're not going to pay attention to him any more. You know what this reminds me of? Republicans who deny the existence of the climate crisis because three percent of scientists disagree that it's happening. How quickly liberals, too, deny math when an outlier contradicts nearly unimpeachable accuracy.

In the 2012 election, Nate Silver was 100 percent accurate on the presidential race. In 2008, too. Actually, strike that. Nate missed Indiana, which went to Obama by a margin of 0.1 percent. Couple these numbers with eerily spot-on accuracy in 2010 (34 out of 36 Senate races predicted accurately), 2014 (roughly the same results at 2010) and, yes, this 2016 primary season, and any rational, realistic analysis of Nate's track record shows that he's missed very few election results out of literally hundreds or more. So, then, to write off Nate Silver because he was wrong about Michigan is roughly as foolish as Mitt Romney's 2012 campaign relying too heavily on its own polling, while scoffing at Nate's alleged liberal bias. Worse, a sizable faction of Democrats are now veering toward the Dean Chambers "unskewed polls" nonsense. Remember him? No? There's a good reason for that. He made a fool of himself and no one takes him seriously any more.

Yes, Bernie won Michigan. Yes, he deserves a round of applause for it. Yes, Nate Silver's forecast was wrong. But, no, that doesn't negate what's repeatedly proven to be mind-boggling predictive accuracy. (And by the way, yes, Hillary ended up winning the day with a net gain of 18 pledged delegates.)

Does this toss into chaos the polls for upcoming open primary states like Ohio and Illinois? I'd be lying if I said I'm as confident in the polling averages as I was yesterday, but math is still math. Everything else is guesswork based on considerably less accurate metrics such as crowd sizes, cable news punditry or whatever constitutes momentum. 

Returning to our climate crisis denial metaphor, it's snowing in Boston therefore the climate crisis is a hoax -- likewise, Michigan was wrong, therefore should we look at Nate Silver's math with an incredulous eye? 

Uh, no.