One of the problems with following election news here in the US, is that the various media outlets often try to convey a sense of drama in the outcome, even when it’s not really there. For this reason, I try not too pay much attention to most of the prognosticators out there. Their track record been pretty dismal so far this year.
What has been fairly predictive however, are straight polling numbers. The polls don’t always get it right (every poll itself requires considerable judgment from the pollsters), but the raw results have done much better in this election cycle than the pundits. Many pundits ignored Trump’s early but consistent strength in the polls, while the polls themselves largely told the story that would unfold.
With that in mind, I have a few number oriented sites that I follow to keep my finger on the pulse of the election.
The first is FiveThirtyEight. These guys are pretty well known. Their name is obviously a reference to the number of presidential electors in the electoral college. Nate Silver, the founding statistician of the site, famously predicted the results of the 2012 election, much to the consternation of conservatives at the time, who strongly felt that he had to be wrong. He wasn’t. That said, while I’m a fan of FiveThirtyEight, it sometimes feels to me like too much in their analysis is outside of the actual polling data, and that doesn’t necessarily increase the accuracy of their projections, although they’re usually good about providing the raw polling numbers if you’re willing to scroll a bit.
A site that I just recently discovered is Princeton Election Consortium. I’m not sure how long they’ve been around. It was founded by Sam Wang, a neuroscientist, and has a periodically updated model which gives a probability distribution of who will win under current conditions. I’ve seen a lot of recommendations for this site, and Wang himself reportedly has a good track record going back to 2004.
But as much attention as these sites get, they’re not my favorites. There are two other sites that I visit more often as the election gets closer. They are actually run by a couple of amateur poll aggregators, although they’ve been doing it longer and more successfully than many of the other sites out there.
The first is Electoral-Vote, run by Andrew Tanenbaum. If you’re an old computer nerd like me, that name might sound familiar. Tanenbaum wrote a number of well regarded computer science books in the 80s and 90s, one of which was about writing an operating system called Minix. Minix was a project which inspired a certain computer science student at the time named Linus Torvalds to take a shot at building his own operating system, which just happens to be the kernel at the center of what we now call Linux, the OS that powers a sizable chunk of the internet, not to mention Android phones.
Tanenbaum is an expatriate progressive (living in the Netherlands), which concerned me when I first discovered his site in 2004. I was a little concerned that maybe I was just reading disguised propaganda. (I’m now a liberal progressive myself, but wasn’t quite there yet in 2004, and in any case I was looking for the most accurate numbers.) But Tanenbaum himself pointed out that his numbers were similar to another polling aggregation site run by a born-again Christian conservative named Scott Elliott. Elliott’s site is called ElectionProjection.
Since 2004, both of these guys have been collecting polling data to make a prediction on how the electoral college vote will come out. And they’ve both been remarkably accurate.
Tanenbaum has been pretty open about his straightforward methodology. (It may be that Elliott has been equally open but I haven’t scoured his site as closely.) Tanenbaum simply takes, for each state, the average of all public polls conducted within the last five days. If there hasn’t been any polling in the last five days, he uses the last one(s) published.
Notably, neither of these guys wimp out when the polls in a particular state are close (as many news organizations tend to do). They identify which states are close, but still count that state as a win for whichever candidate is slightly ahead, counting on any inaccuracies in the margins of error to cancel each other out.
The result are projections of the electoral college that, while a snapshot, usually are remarkably consistent once we’re well past the conventions. Both sites also make projections for the Senate, and Elliott makes projections for the House. (Tanenbaum used to do House projections but found it to be too labor intensive.)
A quick glance at the sites will show the current state of the race. Even with Trump’s current post-convention bounce and lead in the national polls, Clinton is still (barely) leading in the electoral college, the only vote tally that matters in Presidential elections. (As anyone old enough to remember the 2000 election knows all too well.) But the current confidence that the Democrats will take the Senate, or are within striking distance of the House, appears to be misplaced.
Still, it’s early days yet, and the shape of the race will almost certainly change, although the history of the last few elections is that we’ll get a good idea of what’s going to happen in the next few weeks.
Finally, I can’t do a post on political web sites without mentioning my favorite overall political sites. Taegan Goddard’s Political Wire is a blog aggregating intelligent political commentary from several sources, admittedly with a progressive tilt. I also follow Vox and FiveThirtyEight’s article feeds, both of which I can highly recommend.
Assuming you’re not disgusted by the whole mess and eager for it to just be over, what sites do you use to keep track of the election? Any major resources that I’m missing?