Back when I first started obsessing online about polls and trends and probabilities (2004), there weren’t a whole lot of resources. There was 270towin.com (still with the coolest maps around), there was electoral-vote.com and there was Real Clear Politics. Not much else. I’ve always liked RCP because they were great data collectors. They were pretty right wing in their editorial stuff but nobody had the range of data they did. So I’ve stuck with them. Nate Silver came along later, and he’s fun, but these days he’s more interested in looking good than in being right, so I’m a little down on him too.
Today, after my several rants (and a rare double rickroll), I decided I was done with RCP, and set out to look at some new sites. Two of them caught my eye, I’ve heard of both but never really dug in to either one too much.
Today I did. Let’s see what I found.
The first is the Princeton Election Consortium. Oh my good lord, so many nerds in one place. Honey, I’m home! Founded by Princeton neuroscience professor Sam Wang (this year’s Nate Silver), the PEC is a treasure trove of fun for data geeks. Just yesterday, there was a fun interview between Wang and fellow PEC-er and Princeton history prof Julian Zelizer.
The PEC model sets the probability of Hillary Clinton winning the election at 98-99%. Let’s be clear, that doesn’t mean it will be a rout, but that she will win. The PEC snapshot electoral vote outcome is 312-226. (I did not look at this site before making my map this afternoon. Honest.)
The HuffPost presidential forecast gives Democratic presidential nominee Hillary Clinton a 98 percent chance of winning the general election on Tuesday. That means we’re pretty darn certain that ― barring some major catastrophe, scandal or nearly every single poll being wrong ― Clinton will be elected.
But that doesn’t necessarily mean Clinton will win in a landslide. It’s still a close race in several states; Clinton could win with as few as 273 electoral votes. Or she could blow the race out and win 341 or more. The high win probability doesn’t choose between those scenarios ― it just means that the model shows Clinton below 270 is very unlikely.
* * *
So the 98 percent chance of a Clinton win is not a 98 percent chance that she wins by a certain number of electoral votes ― it’s the percent chance that she wins more than 270 electoral votes. It’s less important that Clinton wins each individual state than it is that she wins enough states in this type of model.
That’s why the HuffPost model is so certain of a Clinton win, even though several states are still close. Ohio, for one, is right on a razor’s edge ― it’s changing directions practically every time the model runs. Clinton is neck-and-neck with Republican presidential nominee Donald Trump as they vie for the state’s 18 electoral votes. Trump needs those to win. Clinton does not.
In fact, Clinton has 302 electoral votes in just the states in which she’s 90 percent or more likely to win. That includes Michigan, Virginia, Pennsylvania, Wisconsin, Colorado, Florida and New Hampshire. That’s because polls in those states have shown consistent trends toward Clinton ― and nothing in the last week has been definitive enough to change those trends.
I also found, linked to the PEC site, a series of maps from various professional prognosticators made with the 270towin.com charts. Here’s a bunch of maps and even better, here’s a chart that has predictions by a whole bunch of people who get paid to do this stuff. Nobody predicts Clinton with fewer than 272 electoral votes, nobody predicts Trump getting more than 214. Tossups range from 12-103 electoral votes.
So take RCP and CNN and anyone else who things “poll of polls” is a reasonable methodology and dump ’em in the trash. There’s a a wide range of available data out there, and virtually all of it will dissolve the knot in the pit of your stomach. Is Trump doing better? Yes. Does that mean he’s going to win? Almost assuredly not.
And if you don’t want to trust the guy who suckered you with a fake LA Times headline earlier tonight, then read one of the other nerds with vastly better credentials than me anyway. I’m an amateur – these guys are pros. I’m good at spotting bullshit, but I have no idea what “Bayesian” even means. You should listen to the pros. You’ll feel better. Really.