« Do Scientists Already Know This Stuff? | Main | Bounty Slander »

May 17, 2008


I worry whether markets will actually make predictions that differ substantially from the consensus achieved by other means. I'd be curious if that has happened much, within the limited existing track record (like IEM, Intrade, FX) - cases where the "favored estimate" is contradicted by the markets, and the markets turn out to be right. If markets turn out mostly to track the already-existing consensus, they will not be too useful.

I think the fear is misplaced, because liquid wealth is liquid power, and to a significant degree buys and trumps all other status indicators (this will only increase as all status enhancing/signalling traits become increasingly malleable). People who seek to signal status will rationally be drawn to markets where they have an information advantage, if for no other reason than to acquire wealth to pay for services and products to increase their ability to signal status.

However, on another hand our crude society is deformed from its rational dragon-slaying potential by our various primate social aesthetics. So I think your parable does explain a large part of the tepid reception of information/prediction markets thus far.

Actually I thought of an example of what I asked about, where the markets were making a substantially different prediction than the socially or officially favored estimate. In the contest for the U.S. Democratic Presidential nomination, Obama and Clinton have generally been presented until recently as both being roughly evenly matched. Only within the last two weeks, since the Indiana and North Carolina elections on May 6, has the consensus narrative changed to be that Obama is the presumptive nominee and now the only question is when and how Clinton will get out of the race.

However, the betting markets have anticipated this for far longer. In fact, Obama has had more than a 70% chance of winning the nomination since mid-Februrary on the Intrade market. For months, there was this disconnect between the media narrative of a closely-fought and tight race, with much excitement and attention to all the ups and downs, and the markets' consistent view that Obama was the overwhelming favorite. Only in these past two weeks have the media abandoned their pretense of it being an even race, and now the narrative story has changed to match what the markets have been saying all this time.

I don't know if the markets will beat the consensus of experts any time soon, but I think they're likely to beat media reporting for reasons Hal laid out: the media has a vesting in a horse race narrative separate from reality. I do note that the markets all strongly predicted Hillary Clinton to win the nomination at the outset of such markets (at least at the first reporting of them in publications like slate). This should be recognized so we don't succumb to massaging the data by cherry-picking the moments when the markets favored Obama.

Hopefully, news media and viewers like to pretend they care about accuracy, but if in fact they don't care very much relative to other ends, then prediction market popularity will depend on how well they acheive those other ends.

Hal, yes, there are many cases where markets appear to lead media consensus. Would be nice to get clearer data on this though.

I was thinking of something like the Gene Sweepstakes where scientists bet on the number of genes in the human genome. The mean was 61,710; the minimum was 27,462. (See also the FX betting.) But the actual number turned out to be about 21,000, lower than the minimum. Here, it seemed that the betting market merely confirmed the same consensus which one might have acquired among scientists through other means, and was not particularly close to the eventual result. The question is whether Idea Futures markets will succeed in discovering cases where the "official" scientific (as opposed to media, as in my example) consensus misrepresents the actual views of working scientists.

The comments to this entry are closed.

Less Wrong (sister site)

May 2009

Sun Mon Tue Wed Thu Fri Sat
          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30