The British Nate Silver

 by Martin Belam, 8 November 2012

Yesterday I glibly tweeted that I had trademarked the term “The British Nate Silver”.

Not for myself, you understand, but in recognition of the fact that his data analysis has played such a pivotal role in debate around the US election that it is inevitable that someone will get dubbed with that title in the run-up to the next General Election in the UK.

It will be very tempting for media organisations over here to attempt to set up something similar. The Telegraph and The Guardian, with their existing talented sets of data journalists, resources, and political ideologies, look the best bets.

Of course, as a few people have pointed out, when Nate Silver himself tried his hand at the UK elections, his first attempt at a model was way off. He himself said “it is experimental and should be regarded as such”, and I’m not sure if that shows that a similar approach wouldn’t work in the UK, or just indicates that Nate wasn’t steeped enough in our political culture to refine the model accordingly.

Paul Bradshaw has written the best blog post I’ve seen on the topic of what it means for journalists. Essentially, once you’ve got someone in the market applying statistical rigour, it makes everybody else relying on their old-fashioned approach, “hunches” and off the record briefings about how the campaign is going look data illiterate.

It isn’t just politics that could benefit from a bit more maths though. I’ve blogged before on here how about how Michael Blastland and Conrad Quilty-Harper have spoken passionately about the media’s failure to grasp numbers, or to present crime data in meaningful ways.

Let’s hope that there isn’t one British Nate Silver.

Let’s hope that this is a wake-up call for our media to apply more statistical thinking to their reporting.

Maths and statistics aren’t super-powers. You just need to care enough to use them properly when reporting.

If you want to get started, the Royal Statistical Society has resources for journalists, the media and public relations professionals.


Or here is a curve ball - why report polls during an election at all - it is banned in Italy.

I think it makes the discourse around the election a lot more interesting if there aren't polls

I also think that the 'could go either way' narrative was a much more interesting one to report than 'it looks like Obama's won'.

But Chris, would you rather the media reported stories in an interesting way, or an accurate way? much the same way that an "alien invasion" narrative is a more interesting one to report than "jobless figures rise".

Interesting, but utterly false, and destroying the credibility of news orgs that are less informed than any interested reader with access to a web browser.

If you want an interesting narrative that is also true, surely should have been (as it nearly always is in US elections) "the popular vote could go either way, but because of the way the electoral college works Obama is much more likely to emerge victorious". There's a big, important story in there about how the system is broken and fails to represent a huge swathe of the country at any given time -- which is, in part, responsible for the culture war between left and right.

Also, don't forget that it's not just the horse race that informs election editorial: there's also the fact that they don't really want to report a clear victory so they don't go totally the wrong way if the outcome is a surprise.

For example, under Silver's model there was, at various points, a 10-30% chance that Romney would win. As it's a binary outcome, any reporter who wrote that "Obama probably has it in the bag" probably would have got a pasting by Republicans if that had happened. That's because it's not just journalists who can't get the stats right, it's readers too.

Keep up to date on my new blog