Joint-outcome prediction markets for climate risks

select scenarios for scenario analysis and “stress testing” exercises. These market-based likeli- hoods of different emissions pathways would also provide information about the credibility of policies to curb emissions. For example, if an international treaty to drastically reduce emis- sions was agreed but the market implied probabilities of lower emission scenarios did not increase, this could be interpreted as a sign of a lack of confidence in the effectiveness of the treaty. As well as GHG concentrations and global temperatures, a platform for prediction markets which attracts expert participants could be used to produce collective forecasts of many cli- mate-related variables. This could include those pertaining to physical risks, such as sea-level rise, and regional temperatures and rainfall, but also those that drive transition risks, such as the uptake of renewable technologies. Establishing a prediction market to aggregate forward-looking climate information would remove the need to select a single forecast provider, which is a challenge for users of climate forecasts because, unlike with weather forecasts, they cannot rely on the forecaster’s track record to evaluate their quality. Because the aim of the markets would be to obtain information from participants, rather than revenue, these participants would not have to “pay-to-play”, avoiding the regulatory issues that have hindered the development of prediction markets. Instead, they could be endowed with credits representing a share of funding provided by orga-

nizations seeking credible, collective forecasts of climate risks. Some of this funding would also be used to provide the subsidy required by the market maker to incentivize participants. The markets could provide a method for distributing funding to researchers in climate-related areas in a way that makes them accountable for the accuracy of forecasts and allows the provid- ers of funding to specify the outcomes for which they seek predictions. An objection to using prediction markets to inform climate change policies is that some participants may attempt to manipulate the markets to distort climate predictions to either exaggerate or downplay the risks of climate change, depending on the agenda of the partici- pant. This is more of a risk with open “pay-to-play” markets, where a participant willing to lose money can distort the forecast generated by the market. Such behaviour, however, will create profit opportunities for other participants, and these additional incentives can actually enhance accuracy [28]. Nevertheless, the suspicion by decision makers that manipulators are taking part in a prediction market can lead them to under-utilize the information generated by the market, even if there are no manipulators present [29]. It is therefore crucial that predic- tion markets intended for use by policy makers are monitored so forecast users are confident they are not being manipulated. Conclusion The experimental markets described in this paper prove that the problem of limited liquidity,

that has hindered previous prediction markets for climate-related variables, can be overcome with a subsidised market maker. They also demonstrate that using an automated market maker allows markets with large numbers of outcomes, including joint outcomes, to function. Although the markets converged well to the actual values, six independent forecasts are not sufficient to provide a statistically meaningful determination of the accuracy of the prediction market approach compared with other methods. Furthermore, none of the markets had a fore- cast horizon of more than seven months, so whether markets with significantly longer hori- zons would behave in a similar way is an open question. Running markets with much shorter horizons would produce enough data to perform a statistically rigorous verification study, but shorter horizons would accentuate the question of the validity of the results to multi-year pre- dictions. This validation problem is, of course, intrinsic to all methods for predicting long- PLOS ONE Joint-outcome prediction markets for climate risks PLOS ONE | https://doi.org/10.1371/journal.pone.0309164 August 30, 2024 12 / 14