Front page

Blarchive

Blinkered

15 Dec 12 | Re: Changing of the media geek guard | Link-U-Post

Remember the US election? The main thing I noticed from my UK perspective was that it spawned a thousand moderately interesting articles about Nate Silver, the statistician who correctly predicted the result in all 50 states. This achievement crowned him as a Top Media Geek: Nate Silver to become the next Malcolm Gladwell, says the Gawker website. He will write books, he will be applauded on websites, he will be asked his opinion on every possible issue over the next few years that involves polling or statistics or predictions in some way.

But what about the old Malcolm Gladwell? He’s Canadian, I know, so probably he wasn’t interested, but theoretically, could he have predicted the US election results as well as Nate Silver did?

Silver’s method wasn’t particularly complicated in principle. It took advantage of the fact that most election pundits’ approach was laughably weak: they would speak to a few people, watch the news, mix that in with their own prejudices and basically pluck predictions out of the air. Silver, by contrast, studied the polling data, appraised it for reliability and slant, and painstakingly put together his best guess at how people would vote based on how as many people as possible said they would vote when asked.

In his book Blink, Malcolm Gladwell gives us an idea of what his approach might have been. Blink is all about the notion that a snap decision based on an initial, albeit limited impression can often be the best one (padded out to just over 300 pages, of course). In essence, then, Gladwell suggests behaving like a particularly stupid version of the hapless traditional pundits who Silver has just beaten to an electoral pulp. Careful consideration of all the available data? Pshaaaaw to that: just blink! You wonder why Silver bothered.

Silver bothered, of course, because Gladwell is talking bunk, and his ascendency would appear to leave Gladwell in a tricky position. In their media careers, they are going for the exact same audience. Silver has just provided inarguable proof that his approaches work, using one of the biggest world news events as a platform. His approaches are built from solid planks of hard data and held together with sturdy mathematical bolts, whereas Gladwell spins candy floss out of cherry-picked examples and obviously made-up numbers – “ten thousand hours of practice”, anyone?

I pointed out in my review of Gladwell’s Outliers (sorry if it comes off as peevish - I had just read all of Outliers) that Gladwell’s “discoveries” were only useful for “predicting” events that had already happened. So as much as the question of who the media wheel out to talk about this sort of thing seems pretty unimportant, it must be a good thing if the general public’s go-to stats guy is a guy who knows how to make proper and honest use of stats.

Posted by JUANITO SANTA CRUZ at 22:26

[Back to main blog]

[Or dive into the blarchive...]

---

Take me home