Wednesday, October 27, 2004

Economics - why political betting is not a sensible indicator

One of the fashionable fields in recent economic thought has been trying to apply rational expectations theory to political events. Famously, the Pentagon had to pull its "policy analysis market" due to outcry. Now, I see via Brad DeLong that a couple of academics have tried something similar, using a US sports betting website to trade futures in various political events. You can get the paper here.

Now, the theoretical underpinning for this is the rational-expectations hypothesis familiar to economics students. It states simply that market participants make their decisions through a rational calculation of costs and benefits based on all the information available. Although the possibility of their being mistaken is accepted, it is assumed that all errors are equally probable - that errors are randomly distributed. With enough data, then, there should be just as many errors on each side and they should be equally extreme. Therefore, they cancel out and can be ignored. This has important repercussions - for example, the price of a share is assumed to reflect a precisely accurate estimate of the firm's future profitability at any moment. Any attempt by government to reflate the economy is futile, in the extreme view, because everyone will simply adjust their behaviour instantly so that only inflation results.

In a betting market, then, the odds on any event should represent an accurate prediction of its likelihood. Now, a lot of us would tend to be sceptical about this. We might be even more so, when we see some of their results. Apparently, according to the market, the capture of Osama bin Laden becomes more likely as the election approaches. This can hardly reflect an accurate assessment of available information, as there is practically no information available. Clearly, this is a shadow effect of the belief that the capture of bin Laden would influence the presidential election. The nearer the election, the greater the impact - this seems intuitively sound. That doesn't mean, though, that the event is more likely. In fact, the hypothesis that the capture of bin Laden might occur in time to save the President's bacon is driving the market for the likelihood of Bin Laden's capture - an independent variable. There is no known connection between the two. There's been a lot of rumours, but nothing solid. One kind of rumour might be that a political betting exchange was seeing a lot of money going on. Yep, it's Soros's idea of reflexivity in town.

There are some serious questions involved with this. For a start, gambling isn't a rational activity. A majority of gamblers lose. Why should their decisions be treated as economically rational? Secondly, rational expectations theory assumes perfect information. In this case, the participants are betting on something they almost certainly have no information on. The only guide they have is the behaviour of others. Thirdly, there is no reason to believe that errors are random. In fact, crucially important instances of error tend to show not randomness but very strong patterns. Stock market bubbles would not occur if error was random - does anybody really believe that Bookham Technology's share price in early 2000 was a cosmically correct prediction of its profitability? Well, it was a very bad prediction. Politics provides an endless supply of these examples. The German leadership in the Second World War showed a strong tendency to make the same mistake again, and then to reinforce it. At the battle of Avranches the Panzer Group Eberbach was sent again and again further into the attack although its rear was increasingly threatened. The year before, something similar had happened at Stalingrad. The year afterwards it would happen at the battle of the Bulge. In the field of aviation safety, the human ability to make a mistake and keep on making it, convinced of the rightness of your cause, is tragically frequent. Men have been known to convince themselves that, even though a physical stall warning was wildly clattering the control column, the aircraft was not about to stall, and to yell down their colleagues until the crash.

Gamblers, of course, are the ultimate example of this trait. Although the house has the edge, they can be willing to keep betting in an effort to recoup losses, even running bigger risks. When you think of qualitative examples - where a particular structure of thought constrains reason - you could keep going forever on these. To get away from this psychological digression, think of the problem of representation. Is it possible that people who bet on political events in a manner favourable to George Bush share certain views or preferences - a particular political indifference curve - that affects their results?

No comments:

kostenloser Counter