Joint-outcome prediction markets for climate risks

prediction market exchanges, such as Intrade and Smarkets. In the Intrade market, the global temperature anomaly for 2012 was partitioned into intervals of 0.05˚C and contingent con- tracts defined for each interval [21]. The Intrade platform used a continuous double auction which directly matches buyers and sellers of contracts. Unfortunately, due to the relatively large number of contract types, and the esoteric nature of the topic, some of the markets suf- fered from low levels of trading, making it difficult to infer reliable probabilities from prices. Such thinly traded markets are described as lacking liquidity. These markets are also “zero sum”: the rewards of informed (or lucky) participants must come from the losses of unin- formed (or unlucky) ones. If the market does not attract enough uninformed participants, the potential rewards are not great enough to attract informed ones. Even if the liquidity problem can be solved, with longer-horizon temperature-only markets the problem of circularity makes their interpretation ambiguous. The circularity problem can be addressed by defining a two-dimensional space of possible outcomes. For example, GHG concentration (the policy variable) is partitioned into intervals on one axis and temperature anomaly is partitioned on the other. Therefore, each outcome is defined by simultaneous val- ues of GHG concentration and temperature. This structure allows participants to take contin- gent positions that the temperature will lie within a certain range if the concentration of GHGs

falls within a defined range. Each grid cell has a price (between 0.00 and 1.00) associated with it. Marginal price distributions for GHG concentration and temperature can be obtained by summing over each dimension separately, but the two-dimensional structure means that the market also implies a joint distribution of GHG concentration and temperature; essentially providing an estimate of transient climate sensitivity. A problem with a joint-outcome space, however, is that it greatly increases the number of outcomes, and hence the number of contract types to be traded. This exacerbates the liquidity problem. It might be unlikely that any two participants will want to trade identical collections of contracts, complicating the task of directly matching buyers and sellers; this is known as the “double coincidence of wants”. The problems of matching buyers with sellers and low liquidity can be solved by using an automated market maker, an algorithm that will always buy or sell baskets of contracts. All trades are with the market maker that sets prices based on its relative exposure to different out- comes. Market makers exist in many financial markets, and bookmakers in traditional betting can also be thought of as market makers. The aim of traditional market makers, however, is to make a profit from the “spread”, by selling contracts for slightly more than they buy them. But a market maker doesn’t need to be motivated by profit. Instead, a market maker can be willing to lose money in return for good information. Such a subsidised market maker changes a

zero-sum market into a positive-sum market, essentially taking the place of uninformed partic- ipants so that informed participants always have an incentive to take part. When subsidising a market through a market maker it must be carefully designed so that subsidies incentivise the provision of accurate information and cannot be “gamed” by participants. The Logarithmic Market Scoring Rule (LMSR) market maker [22] rewards participants according to the loga- rithmic scoring rule—a well-established proper scoring rule for probability forecasts [23]— based on their marginal contributions to the accuracy of the collective forecast. While in principle joint-outcome markets for GHG concentration and temperatures solves the circularity problem, it is not obvious that they will work in practice. To provide useful information, concentrations and temperatures should be finely partitioned (e.g., 10ppm inter- vals for carbon dioxide, 0.1˚C for temperature). The resulting outcome space is an order of magnitude larger than those featured in previous successful trials [24]. It also cannot be assumed that market participants will be able to engage with a complex market structure in a way that allows their views to be effectively expressed and aggregated. PLOS ONE Joint-outcome prediction markets for climate risks PLOS ONE | https://doi.org/10.1371/journal.pone.0309164 August 30, 2024 4 / 14