The pitfalls of political punditry
Matt Yglesias puts his finger on why you can’t trust political pundits to tell you what’s happening in an election race:
…ordinary people ordinarily just don’t pay attention to politics at all. They don’t think it’s interesting. But … [t]he political press, by contrast, finds electoral contests to be a kind of fascinating game about which it’s amusing to do tons of stories regarding the ins-and-outs of tactical gambits…
The kinds of people who are political reporters are self-selected to enjoy commenting on the race as if it’s a game, simply because talking about politics day in-day out requires them to do just that. To think that they are, in their view of politics, remotely representative of the voting population is ridiculous. In fact, their analysis of political events is highly unlikely to represent the views of the average voter.
It is likely that the pundits are portraying the election contest as far closer than it really is. A landslide is hardly the most interesting thing to comment upon: there will be plenty of time to pick over the bones of the losing campaign when the election’s done. A close contest makes for far more exciting discussion and they probably have an incentive to depict it as such, regardless of the true situation.
I would far rather trust aggregations of information such as you find on prediction markets. However, it may similarly be argued that those who engage in share trading on such markets are not representative; so the aggregate price statistic you see is a biased indicator of the actual probabilities. Which way the distribution of share traders is skewed I don’t know.
Well, if you knew which way us traders were skewed, you could certainly go in and make a lot of money off of us.
It would be wonderful to have that sort of insight 🙂
it was interesting to look at the development of the Obama/McCain stocks on iPredict in that respect, though. They consistently favoured Obama far more than InTrade’s prices for the first month of trading. Perhaps this can, in part, be explained by the more politically liberal stance of NZers?
I don’t know anything about financial markets, so is this a common phenomenon — for people to be influenced in their predictions by their preferences?
Many of your assumptions are off. Commentators differ in what they offer – some are partisan, some not. Some are policy wonks, some are tabloid.
I got the 2005 election right and Centrebet did not. In an earlier series on the blog, predicting the election date, the market was right. Colin James boasts that he has correctly predicted every election result (bar one).
There is a stat doing the rounds lately that ~80% of the coverage of minor parties around the time of the last election was not to do with their policies, but the ‘game’ of possible coalition partners. Really quite a deplorable situation.
Tim:
I’m sorry, I was loose with my words there. I am thinking of non-partisan political reporters rather than partisan commentators. Those who make a living off telling the world what’s happening rather than those who are aiming to persuade others.
With regards to your second point, I find it unsurprising that many commentators are correct in their predictions. They’re predicting a binary variable with information that gives them a far better than 50% chance of being right. I’d expect them to get it right most of the time with those odds. However, whether they are right or not, I think they have an incentive to TALK about the election as though it’s a close race.