President Obama was re-elected yesterday. It came as a surprise to some, but I’ve known for some time that there was a higher likelihood that Obama would win than Romney. I want to share how I came to that conclusion and why this is an illustrative example of how we can predict better results by choosing more reliable data sources in everything we do.
The winner of this election is Nate Silver
Nate Silver is a statistician that runs the New York Times FiveThirtyEight blog. I could go into his eclectic background making a living for a time playing online poker, as co-author of the Baseball Almanac, but his current claim to fame is for his statistical models of elections.
His model takes a meta-analysis of many polls and weighs them based upon their past performance and expected biases. Of course, I don’t understand the specific mechanisms nor do I have to. In 2008, Silver correctly estimated 49 of the 50 states for the Presidential election (Indiana was the anomaly which went to Obama by a 1% margin). Sufficed to say that in his first iteration he proved quite successful.
But in 2012, many pundits criticized or marginalized Silver’s model as partisan (he stated that he personally supports Obama). So I followed his blog a lot, especially after the first debate when Romney gave a commanding performance on a national stage. Silver’s model showed an impact but still favored Obama’s re-election. The night before the election, pundits were calling the election a toss-up. Silver projected Obama’s chance for re-election higher than 90%.
So, who was right? Once Florida is called, Silver will have accurately predicted 100% of the states. Between 2008 and 2012, he will have a 99% success rate, which I consider pretty good. Forbes also ran an outstanding article lauding Silver but also Drew Linzer of Votamatic and Sam Wang of Princeton, who ran similarly accurate models of the electoral results.
What this means and the five whys
A strange thing happened last night when I was watching Fox News. Karl Rove challenged the network for calling Ohio for Obama so early in the evening, and in his remarks he said something about Romney getting an “inside straight.” As it happens, “inside straight” was the poker analogy that Nate Silver used to describe Romney’s odds of winning the electoral college. I think this and the myriad examples of punditry before and after the election show that talking points are less about accuracy and more about substantiating a particular point of view. I’m pretty sure that Rove knew what was going to happen.
There was an infographic floating around in earlier in the year showing that men were the predominant users of Pinterest in the U.K. The information was gleaned from Google DoubleClick AdPlanner. I used the tool to look up every other social network I could think of, and it showed each dominated by men. It was clearly wrong.
When I shared my findings people came out of the woodwork to insist that there was a plausible explanation for this, and that British men love pinning like they love Mr. Bean: inexplicably. The point being that oftentimes we’ll adopt a viewpoint and hold on to in spite of contrary evidence. Maybe that’s why the world needs pundits?
But the qualitative difference between social media / marketing bloggers like Christopher Penn, Jeremiah Owyang, and Charlene Li and much of the blogosphere (my own site included) is vast. There are a lot of big personalities that don’t share the astute, measured insights that analytical bloggers do. And I think it’s important to understand the difference.
There’s a concept in lean manufacturing called the “Five Whys.” It’s the same mechanism that my three year old tries to use to get candy at the store. You take a statement like: “UK Men love Pinterest” and ask why five times.
UK Men love Pinterest. Why?
Because an infographic says so. Why?
Because Google DoubleClick Ad Planner said so. Why?
Because all social media is dominated by men in the UK, but nowhere else. Why?
Because the data’s likely inaccurate.
The point is that anything that people tell you about what you need to be doing with your social media or marketing or whatever should be substantiated to your satisfaction. If you’re happy targeting men on Pinterest or believing that an election that’s not very close actually is – then your threshold for substantiation would be lower. But if you’re mitigating business risk, ot are the type of person who hates surprises – you may want to dig a little deeper.
It’s one thing if you woke up this morning and were disappointed that Mitt Romney isn’t President: you’ll get over that. It hits home a little more when you’ve applied a bunch of time, money and best practices to your Facebook Page and can’t discern whether it’s doing anything for you. That sucks.