AGORA use case: Predicting hurricanes

Mark Roulston, Senior Data Scientist •

IMG Hurricane

Problem

The Atlantic hurricane season runs from June to November. On average, 6 hurricanes form each season but the actual number is variable with 2005 seeing a record 15 hurricanes. Hurricanes, particularly hurricanes that make landfall, are of interest to reinsurers who essentially provide insurance for the insurance companies that sell policies covering the wind and flood damage associated with storms.

Approach

As part of its participation in the Lloyd’s Lab InsurTech accelerator Hivemind used AGORA to produce expert consensus forecasts of how active the 2020 hurricane season would be. Experts in seasonal weather forecasting and hurricane prediction were invited to participate in a prediction market on AGORA to predict the total number of hurricanes during the 2020 season as well as the number that would make landfall in the United States. With on-platform credits, participants were able to buy and sell contracts corresponding to different numbers of hurricanes.

Outcome

Fifteen experts participated in the market. They made over 1,000 trades during the duration of the market. From their trading activity, AGORA generated implied probability distributions for the numbers of hurricanes and landfalling hurricanes. These distributions evolved during the season as participants updated their positions in response to new information, particularly the formation of new depressions and storms. At the end of season, after the market had been settled, the participants who had accumulated the most on-platform credits received cash rewards.