Risk Markets And Politics

Monday, January 08, 2007

HP and corporate prediction markets mechanisms

Leslie Fine's announcement that HP had signed with Pfizer to deploy an "information gathering tool" could mark a turn for corporate prediction market consultants, regardless of nomenclature. Pfizer had enough internal resources to build their own prediction markets, and ran them for about a year. As Leslie tells it, Pfizer encountered some issues that HP had already solved. First, Pfizer wasn't interested in "crowd" knowledge as much as aggregating the opinions of small groups — of experts, one might say. (A scoring rule was helpful to solve the liquidity problem.) Second, time spent following a market is a cost, so Pfizer wasn't thrilled with a standing market, nor the CDA format. Finally, HP's mechanism doesn't automatically relay information back to everyone, which has an appeal to managers and also works against information cascades.

Two of Pfizer's infelicities can arguably be expressed as costs: how to minimize the number of participants and the time they spend on the mechanism?

Keeping the aggregate price private is even more desireable in situations where participants are external to the "patron", who after all has paid for that valuable information.

As Robin Hanson noted, it's not as though prediction markets are strictly opposed to expert opinion. Larger samples are more robust, but the marginal costs and advantages of adding participants must be weighed, and will vary across companies and questions. HP also cluster-samples participants to reduce costs.


Post a Comment

<< Home