James Annan, a climate scientist working with a Japanese climate research center, recently posted an interesting analysis to the discussion list for the Foresight Exchange. (FX)
There has been an ongoing discussion on the FX lists about observed biases in those (play money) markets. Some people believe that there’s a bias toward the midpoint in all prices, so that few claims actually carry prices that are near the extremes, even though there are many claims that are believed to be nearly certain or nearly
impossible (AKA, “crackpot claims”). This is justified because play money allows (perhaps even encourages) participants to bet on long shots. In addition, betting for the long shot is cheap and would have a large payback if it ever came in.
Others think they see a bias toward the negative–more claims trade significantly lower than even odds than trade significantly higher than even. The usual justification for this bias is that FX prefers claims that are phrased as positives (even X will happen) so people don’t have to reason about double negatives in order to undestand the claim. The most interesting claims are those that will be expected to be controversial, so it’s easy to get approval for claims on unlikely occurrences. The result is a preponderance of crackpot claims, so low prices shouldn’t have been unexpected.
Given that context, James did an analysis of all claims that closed before the year 2000, sorting them according to their prices as of a year before their due dates. He then divided them into 4 groups based on price, and graphed the result.
Two quotes to notice from James’ email messages:
“It looks like the market has some credibility, but a substantial bias in favour of lower-priced claims which are clearly over-priced. Nothing with a price of 0.36 or lower with one year to go actually came true!”
“Note that 75% of all these claims eventually paid $0, and almost half of them were in my bottom bin.”
This looks to me like further fodder for the discussion about the effects of real money on prediction markets. I don’t think anyone has found a similar bias in any real-money market.