Media Briefings

USING MARKETS TO FORECAST CLIMATE CHANGE: New study shows their potential for prediction

  • Published Date: May 2013


Markets could be used to predict climate change and other long-term events, according to research by Dr Lionel Page and Professor Robert Clemen, published in the May 2013 issue of the Economic Journal.

Economists have long praised market mechanisms for their ability to aggregate dispersed information about what should be produced and in what quantities. This study shows how markets could be used to aggregate the dispersed knowledge of experts to create forecasts about long-term events. Co-author Dr Page says:

‘Our study shows that while such markets do not exist at the moment, it would be reasonably simple to create markets forecasting very long-term events, such as the future rise of the sea level or the level of depletion of natural resources.

‘On such a market, experts with very different perspectives – for example, geophysicists, biologists, economists and political scientists – could pool their different views, and market prices would give us a unique possibility to aggregate their diverse knowledge into one forecast.’

Prediction markets are a new kind of market, which have attracted the interest of economists and other social scientists in recent years. On these markets, people trade assets whose value depends on the realisation of a possible future event.

One form of prediction markets is well known to the public in the form of betting exchanges where punters can bet against each other without bookmakers as intermediaries. People willing to stake money on a future event can trade bets with other market participants.

While betting exchanges have kept the format and vocabulary of gambling, they are really markets where supply and demand determines the price of bets. As a consequence, the price of bets can be seen as an estimation of the probability of the event as estimated by a large number of informed individuals.

Over the last ten years, prediction markets have raised a lot of interest for their possible forecasting ability. Some prediction markets are now routinely used in US presidential elections to predict the winner alongside standard polls.

But the new study shows that in their current form, prediction markets are unable to predict long-term events (events more than a few months in the future). Due to the time needed to wait for the uncertainty on long-term events to be resolved one way or the other, traders are deterred by such markets. Dr Page says:

‘Prediction market traders do not want to buy a risky asset that could give them a 20% benefit in ten years when they will get a safe 60% return in a deposit at the bank.’

The authors go further and show that when a price can be observed on a market for an event taking place a substantial time into the future, it will be biased towards 50% (which is another consequence of traders’ reluctance to wait a long time for market resolution). Using a large dataset of prediction market prices, they find that this bias is indeed present for prediction markets about events taking place in several months. Dr Page says:

‘Unfortunately, our results show that the further away in time the event to be predicted, the less informative are the prices in prediction markets.’

But this negative result actually helps to point to new ways to overcome prediction markets’ present limitations. Having found the causes of the observed biases, Dr Page and Professor Clemen suggest simple modifications that could make it possible for prediction markets to give credible forecasts of future events.

ENDS


Notes for editors: ‘Do Prediction Markets Produce Well-calibrated Probability Forecasts?’ by Lionel Page and Robert Clemen is published in the May 2013 issue of the Economic Journal.

Lionel Page is at the Queensland University of Technology. Robert Clemen is at Duke University.

For further information: contact Lionel Page via email: lionel.page@qut.edu.au; or Romesh Vaitilingam on +44-7768-661095 (email: romesh@vaitilingam.com; Twitter: @econromesh).