A Combinatorial Prediction Market for the US Elections

  • Sebastien Lahaie | MSR NYC

We present a new market maker that automatically provides liquidity across multiple logically interrelated securities. Our approach lies somewhere between the industry standard–treating related securities as independent and thus not transmitting any information from one security to another–and a full combinatorial market maker for which pricing is computationally intractable. Our market maker, based on convex optimization and constraint generation, is tractable like independent securities yet propagates some information among related securities like a combinatorial market maker, resulting in more complete information aggregation. We prove several favorable properties of our scheme and report on its information aggregation performance in a live implementation of the market for the 2012 US presidential elections, which drew hundreds of users.

Joint work with Miro Dudík, David Pennock, and David Rothschild.

Watch Next