Five Thirty Eight

Andrew Sullivan pimps Nate Silver and Co. of fivethirtyeight.com, and rightly so. Silver was an invaluable resource for polling junkies, and I hope he finds a way to repurpose his website now that the election’s over. Sullivan sees Five Thirty Eight’s success as the triumph of new media, and I suppose there’s some truth to this. I have a hard time imagining a newspaper or television network combining polling statistics, state-by-state vignettes, and razor-sharp commentary into one awesome political website. That said, it’s also worth remembering that Silver’s statistical model was entirely dependent on other people’s polling data. This doesn’t detract from his accomplishment – it’s just a healthy reminder that even new media gurus depend on traditional reporting more than we’d like to admit. It’s a fairly banal observation, but the future of internet-based media probably involves some sort of symbiotic relationship with traditional outlets. Which is basically a long way of saying that you should still subscribe to newspapers.

Advertisements

10 Comments

Filed under The Media

10 responses to “Five Thirty Eight

  1. Hawerchuk

    By doing so well on his first try, Nate has essentially defused a new specious argument: Nate’s just a guy with a computer; his opinion isn’t as valuable as the pundits who use their guts to analyze politics. As statistical analysis took over baseball, people made this argument a lot – they freaked out when Nate’s system predicted the 2007 White Sox would go 72-90. Then they went 72-90. The key point: there’s always room for new analysis, and you always want as much information as possible – stats and gut scouting.

  2. As far as punditry goes, I totally agree – I really appreciated Silver’s innovative approach to electoral statistics. I was just pointing out that his site relies on other people’s data, which says something important about the continued relevance of “old media” outlets.

  3. Hawerchuk

    The same is true of baseball – Nate took stats collected by other people and analyzed them. His approach really represents a change in the way people look at politics – every outlier poll used to get splashed across the headlines. You can’t do that with 538 evaluating your reliability.

    The other really important innovation (which I think proved incomprehensible to most of the people I referred to the site) is the distribution of outcomes. Monte Carlo analysis (Nate’s way of generating the spiky chart on the right-hand side) can be a great way to analyze a problem that’s too hard to otherwise understand. That technique absolutely destroyed the MSM’s election analysis – there isn’t a reporter out there with enough math background to have even attempted it.

  4. Shannon

    There was a story on dailykos (http://www.dailykos.com/storyonly/2008/11/6/11859/8702/311/655656) that has now been taken down that showed that Nate wasn’t so hot in predicting the election. Pollster and RCP did a better job.

    Also, Nate’s electoral vote prediction was off in favor of McCain. And electors cannot vote in fractions.

    Nate also predicted the Cubs would win the World Series.

  5. I’m not sure if evaluating 538’s accuracy based on a single election result is fair. Also, RCP and Pollster.com are institutions – 538 is run by a couple of amateurs. If anything, it’s amazing they were even in the same league.

  6. That really depends on your definition of success, Will. The real value that Silver and Quinn added was not in the number crunching, but in the analysis of all the other polls.

    They provided persuasive and learned reasons behind why one set of polls were more reliable than other and then, lest we disbelieved them, proved themselves right with the final prediction.

  7. JOHNNY

    I think Nate’s particular talent was in taking all the polls, and not just averaging them, but deciding which were better polls and weighting them accordingly.

    I know Nate isn’t doing the polling, but i’m pretty sure that if he were to, he would never come up with a result that showed McCain was winning 18-29’s by 50%.

  8. “Pollster and RCP did a better job.

    Also, Nate’s electoral vote prediction was off in favor of McCain. And electors cannot vote in fractions.”

    Umm, electors cannot vote in fractions? Who knew??

    Anyway, I was surprised by your assertion that Pollster and RCP did a better job, so I did the analysis myself.

    For those interested, here is the full result:

    http://musefree.wordpress.com/2008/11/07/the-real-winner-contd/

  9. Howard

    Nate Silver introduced several new wrinkles into polling analysis: the weighting of polls according to their past performance, the combining of state-level and national polls to show trends, and most importantly, the use of demographical data to link trends in one state to other states with similar demographics.

    His use of demographic data in the primaries was how he made a name for himself, and that was the biggest innovation he came up with. I think the true test of how well he did would be to see how much his use of demographics improved his projections — or alternatively, whether they actually made them worse. But that was his biggest contribution — the ability to take a poll of Nevada and show how shifts in preferences there might affect the way New Mexico would vote.

    There were plenty of polling aggregators already, from the dumbed-down RCP averages to Pollster.com’s more intelligent smoothing. Silver took what they had done and (presumably) improved on it using data they had never thought to look at.

  10. Pingback: Robert Sharp » Blog Archive » Linklog for 4th November to 15th November

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s