To paraphrase from an interesting article, the jelly bean in a jar experiment and predictive markets prove that the power of the crowd will beat even the experts, hence the markets generally are efficient. The exception is when the experts are given too much air time biasing the crowds opinion distorting the cancellation of error in divergent opinions.

Two things this teaches:

1. We should value all opinions
2. The more we know the humbler we should be to this

Adam Robinson contends: 

With respect to the esteemed Mr. Sears, guessing stock prices is nothing like guessing jelly beans.

In the first place, there's a risk associated with markets that does not exist with jelly beans. But more important, in markets the "guesses" of participants influence each other, with all sorts of feedback loops that do not exist at state fairs guessing jelly beans.

It would be interesting to see whether this second factor can be tested in the field: run the counting contest but post each of the guesses, in ranked order, along with high guess, low guess, median guess, average guess for easy overview, and see whether that information influences participants' guesses (for comparison, two groups could be tested, one with access to the information and one without).

A prize and a penalty could also be attached to guesses, and for a further dimension, see whether and how much participants would be willing to have access to the "market" information. 

Russell Sears notes: 

The predictive markets (the markets that predict things like election results, winners of Oscars, etc) do run very close to the proposed modification of the jelly bean test, with the added bonus that the "bettor" must be willing to lose and hence think he has an edge, like the stock markets.

One example is of a multiple choice of one out of four listing of the one not being in the old music group The Monkees. If 7% know all, 5% know two, and 5% know one, and rest don’t have a clue. With all votes random, besides the choices eliminated due to knowledge, it takes a fairly small number of players to get it right. This would work even if the "expert" did not have perfect knowledge, such as the Oscars, or the stock market. 





Speak your mind

2 Comments so far

  1. Gary Rogan on August 21, 2007 6:50 pm

    The crowd experiments are usually conducted when the crowd is calm. It would be interesting to see the results of the following experiments, obviously with fake danger/rewards, on how well the crowd does against the experts under the following simulated conditions (1) each member of the crowd is expected to receive a substantial monetary reward/incur a substantial monetary loss (a) if they are correct/incorrect (b) the crowd’s guess is correct/incorrect (2) each member of the crowd expects something terrible or something woderful to happen regardless of their guess upon completion of the experiment (3) each member of the crowd is aware of the direction of the decision making (a) as it’s really happening (b) faked to simulate an incorrect guess. The results of these experiments would be more indicative of how financial markets behave, and I would expect them to diverge from the “calm crowd” results.

  2. Steve Scoles on August 23, 2007 7:25 pm

    Further to Adam’s comment ‘in markets the “guesses” of participants influence each other’ … I completely agree. The jelly bean experiment is a very simple way of showing how collectives *can* be very wise (and as Russell points out, we should be very humble much of the time to the crowd), but it is nowhere near as dynamic and interactive as a financial market.

    The article referenced in Russell’s post continues on and does a very brief overview of Complex Adaptive Systems. The feedback loops noted by Adam (an inherent characteristic of an adaptive system) are very critical to understanding markets.

    This is my first time on this site and it is a very fascinating collection of thinking. I’m honoured that someone thought my article worthy of such a cool site.


Resources & Links