Joint-Outcome Expert Prediction Markets for Climate Risks: Conditional Forecasting with the LMSR

Abstract

CRUCIAL’s prediction market platform implements the Logarithmic Market Scoring Rule (LMSR) automated market maker (Hanson 2003), which mitigates the principal challenges facing the user-pays Climate Service Provider (CSP) sector: incentives, abundance of methodological & procedural choice, need for multi-disciplinarity, and ‘circularity’. The platform has been used successfully for hosting 24 climate-related markets with maturities ranging from 2 to 12 months. This sample of markets perform well in terms of probabilistic calibration and information-add over historical climatological benchmarks. Moreover, a subset of these markets implemented a ‘joint-outcome market’ structure, which embeds a spectrum of conditional forecasts. This subsample demonstrates the viability of running joint-outcome markets with CRUCIAL’s LMSR platform despite the large outcome space (5,207 distinct partitions) and the complexity that this and the two-dimensional structure poses for participants. Both the hard data generated by these markets (price and volume series) as well as qualitative feedback from participants indicate that the markets were well-behaved and the participants were able to adapt quickly to the joint-outcome market structure as a mechanism for expressing their expertise.