Predictions on the Money, Australian Financial Review, 19 May 2009
by Andrew Leigh

With economic forecasters now about as well-respected as astrologers and tarot card readers, it is little surprise that the federal government’s official projections for economic growth have been greeted with some scepticism. Yet the fact that predicting the future is difficult does not make it any less important. In tumultuous times, policymakers cannot afford to drive by watching only the rear-vision mirror.

One way to get a better sense of the future is to turn to prediction markets, which have proven surprisingly accurate in forecasting election results. Such markets aggregate information from a large number of individuals, weighting it in proportion to each person’s degree of certainty about the outcome. Ask experts what they think will happen, and you’ll discover that talk can be pretty cheap. But ask those experts how much money they’d wager that the predicted event will come to pass, and you’ll find out how confident they really are.

Over the past decade, a wealth of evidence has accumulated on the power of prediction markets. In the words of George Mason University’s Robin Hanson: ‘in every known head to head field comparison between speculative markets and other social institutions that forecast, the markets have been no less, and usually more, accurate. Orange Juice futures improve National Weather Service forecasts, horse race markets beat horse race experts, Oscar markets beat columnist forecasts, gas demand markets beat gas demand experts’. And many firms – including General Electric, Google, Hewlett Packard, Nokia and Pfizer – have established internal prediction markets to shape their decision-making.

For policymakers, two of the greatest unknowns are the state of the economy next year, and the likely trajectory of the H1N1 flu virus. On both counts, there is strong evidence that prediction markets can do a good job of forecasting the future. In an analysis of futures markets in economic indicators such as employment and retail sales, Refet Gurkaynak (Bilkent University) and Justin Wolfers (University of Pennsylvania) found that these ‘macro derivatives’ performed at least as well as the average from a panel of professional forecasters. Right now, online bookmaker Intrade estimates a 70 percent chance that US unemployment will top 10 percent by December 2009.

Influenza prediction markets have been the subject of a careful study by University of Iowa researchers Philip Polgreen, Forrest Nelson and George Neumann. Under the auspices of a new ‘Iowa Electronic Health Market’, they found that a market on future flu outbreaks (in which the traders are health care professionals) was able to accurately forecast flu activity up to one month in advance. These markets now provide estimates of the mortality rate for H1N1 swine influenza in the US by the end of July (a 93 percent chance that it will be below 1 in 100), the number of countries with at least one confirmed case by the end of July (the shortest odds are on “101 or more”), and when the US will be able to vaccinate 50 million or more people (most probably not until December 2009).

What do prediction markets say about unemployment and swine flu in Australia? Unfortunately, we don’t know, because such markets do not currently exist. Part of the problem is regulatory – gambling laws treat a bet on an important event the same way they treat a bet on two flies crawling up a wall. This ignores the fact that there is a public benefit to improving our knowledge about future macroeconomic and disease events. Governments should lightly regulate – if not subsidise – prediction markets.

Prediction markets are not perfect, but many of the early critiques turn out to have been overblown. Trading volumes are generally thick enough to provide good estimates, and arbitrage opportunities are rare. Market manipulation is difficult, and prices generally revert to their true value as smart money happily soaks up ideological wagers. Structural details – such as whether the markets are run as betting markets or futures exchanges – do not seem to matter much.

At risk of undermining my own status, the strongest argument for prediction markets is the abysmal record of expert commentators in forecasting the future. One study that followed up the forecasts of several hundred political and economic pundits concluded that their forecasts were about as accurate as would have been produced by a team of dart-throwing monkeys. Armed with grand theories, experts are often too slow to adapt their ideas to fit the facts.

Allowing betting on future events will create greater competition in the forecasting market, and increase the precision with which we can anticipate coming events. In a volatile world, a little more certainty about the future has substantial value. Time for policymakers to take a bet on prediction markets?

Andrew Leigh is an economist in the Research School of Social Sciences at the Australian National University.