Tuesday, 6 November 2012

Updated: Nate Silver and the Election

Update - Nov 7: Results posted below.

I've been following Nate Silver's numbers on the 2012 US presidential election. Simply fascinating. If he's right, and I'm led to believe his predictions are scary-accurate, this is not a race at all despite how badly the news wants to portray it as one.

They have all the predictions posted on FiveThirtyEight.

Not being American but still wanting to play along, I compiled his state by state predictions into a spreadsheet I'll be using later tonight. I'm actually much more interested in how close Silver's estimates will be to results than I am in the actual election at this point. You can download my spreadsheet below if you want to play at home, too.

Just enter the results in the Obama and Romney "Actual" columns as they roll in and you'll be able to see the difference between results and Silver's numbers.


With 90%+ of results in now, I am both amazed, yet not surprised, at the results of Nate Silver's predictions. I've learned previously to take poker players proficient in statistic's predictions seriously, especially when they come out with a 91% confidence rate.

Florida still hasn't been given to Obama, but assuming he retains his slight, predicted lead there, Nate will have picked every single state correctly.

I changed my spreadsheet a bit to give me a Vote Share Accuracy (100% - absolute value of the % difference) column instead of just the % Difference from predicted. Not only did Nate pick every state correctly, his Accuracy with respect to vote share by state averages out to 96.8% with the biggest outlier being West Virginia where Obama received 14% less support than expected. He had 17 states where his predictions were 99%+ accurate.

You can download a completed version of this spreadsheet below.


  1. fascinating following mr. silver this year, and it's great to see a poker player go on and dominate another field so impressively.

    without any data to support it, i wonder if he missed in WV/HI in the presidential contest due to dearths in polling and erratic voter participation rates. HI had an open senate race, but this was not deemed very competitive, and WV had nothing interesting going on either.

    in addition, hawaii has, i believe, the lowest voter turnout.

    in the senate contests, he missed ND badly and MT slightly. i don't know much about the MT electorate, but native american populations make the dakotas tougher to poll. indeed, heitkamp not only crushed among these voters, but they provided a decent boost to her vote total.

    1. You are correct. Hawaii was the lowest at 43.6% turnout.

      From the reports I've read, such as this one:


      there's some people that consider being a state an 'occupation' and they'll never vote, along with people who know that the state will always be Democratic by a landslide, and yet others who don't see civic involvement as fitting their laid back lifestyle.

      But there are some potential voters that don't bother because the election has been called before they get to the polls as they're 6 hours behind EST. I think those are the ones that should be accommodated.

      There's a ton of suggestions out there about how to fix Hawaii's low turnout, but the obvious one that I don't really see anywhere is, by virtue of their significant time zone disadvantage, to simply let them vote the day before everyone else.

  2. Interesting! I hadn't seen anyone crunch Nate Silver's numbers quite like this.

    I wonder to what degree West Virginia just wasn't being polled led to the inaccuracy, versus how much Romney's last minute push in Ohio against Obama's supposed coal policies spilled over into neighboring WV -- and of course there was a lot of pressure by the Big Coal barons on their employees.

    I suspect mostly bad polling; after all, Obama nearly lost the perfunctory Democratic primary in WV to a convicted criminal way back in May.

    (Of course, what's really killing coal, the lifeblood of mountainous West Virginia, is all the fracking driving down the price of natural gas, not Obama's clean air initiatives.)

    Do they still have coal left in Newcastle? Ah, drat, Wikipedia suggests my idiom is a 100 years out of date.

    Which reminds me of the joke where little Timmy isn't paying attention in economics class, so the teacher demands he name to her any year, and the U.S. coal exports for that year. Timmy replies: "1492, none."

    1. Another interesting state to look at is Ohio.


      Over the past 40 years they've historically been a Republican state, and although they have shifted closer to the middle, the 4/10 Democratic victories there have been slim, while 4/10 Republican victories have been landslides.

      As Cohen points out in the article, polls showed white working-class voters, who Romney was targeting, supporting Obama at higher rates than the national average, viewing Obama as the saviour of the auto-maker industry and Romney as the type of businessman that caused the problem.