Creating a Climate Prediction Market

In partnership with Winton Capital, we are developing the concept of climate markets (see references below) into a workable demonstration for the purpose of gathering best information about the future climate.

We hosted an evening meeting at the Royal Society on 30 January 2017 to gather community feedback and consider the next steps for the project.

David Stainforth of CATS introduced the concept by talking about the sources of uncertainty in climate prediction, existing models of "expert elicitation".

Mark Roulston of Winton Capital then went into detail on the proposed structure of the platform, including past experiments with similar concepts, the types of bet that can be made, the intended players, the effect of the regulatory environment, and potential concerns to be addressed during development.


Questions raised by the audience on the day included:

Q. How do you distinguish expert opinion from someone making a guess?

A. We don't.  But with enough well-informed players, the market will converge on a consensus opinion.  Less well-informed players can choose to make less specific bets.

Q. UK betting companies refuse better informed professionals. Those professionals say that real sums with informed players move to Asian betting markets.Your view?

A. According to our preliminary investigations, the regulatory regime in the UK appears to be acceptable for our purposes.  We do not anticipate allowing high value bets.

Q. You are defining a scoring rule based on assumptions of rational bettors? What if bettors not "rational" in the sense of maximising expected return, e.g. having a logarithmic utility function?

A. Behavioural issues will be explored during development of the platform.

Q. How much could capping the max loss of the market skew the information generated?

A. As the aim of the project is to generate information, this is an important question which will be investigated during development of the platform.

Q. You may get risk neutral probabilities Q, mixing objective probabilities with investor stochastic discount factors dQ/dP. Which recovery theorem to disentangle?

A. This will be considered during development of the platform.

Q. What if your single co2 measurement is not made, e.g. if there's an eruption on Mauna Loa?

A. Contracts will be defined to account for such eventualities, for example by return of the initial investment.

Q. Will results be skewed due to the only larger long-term bets being made by those who can afford to live without the money in the meantime? Developing countries?

A. This will be considered in future development.

Event flyer