The H1N1 Flu Prediction Market

As kids across the country are starting to go back to school this month, parents, health officials and school administrators are wondering whether or not we will see a resurgence of H1N1. While no one can forecast the future, we asked Phil Polgreen and Forrest Nelson, the directors of the Iowa Electronic Health Markets, to explain what they’ve learned so far from their H1N1 market, a pilot prediction tool, and discuss what their expert traders believe will happen in the fall. The Iowa team has been running prediction markets for avian influenza (bird flu) and for seasonal influenza for several years.

On Friday, April 24, some of the first news reports about a frightening new influenza virus surfaced here in the United States. That news prompted us to re-tool our prediction market for the bird flu in order to develop a market for this new strain. We immediately started working on the questions for the new market and recruiting traders.

We had already assembled a group of microbiologists, doctors, public health officials, epidemiologists and others that had inside information about influenza, including trends in seasonal and bird flu. We gave those experts $100 in virtual money (makeshift dollars that can be used to buy and sell shares on the market) and asked them to start trading. By April 28, the H1N1 market was up and running.

Right off the bat, most of our traders predicted that by the end of May, there would be more than 1,101 cases across the U.S., with a low the mortality rate. As it turned out, they were right: As of June 1, the Centers for Disease Control and Prevention had reported 10,053 cases of swine flu in all 50 states and the District of Columbia with 17 deaths at that time.

What’s remarkable about the forecast is that in late April, news reports played up the danger factor, reporting that the virus was killing young, relatively healthy people in Mexico. The fear was that H1N1 would spread rapidly in the U.S. too and be a formidable killer. Our expert traders didn’t buy into the hype, betting instead on a future that resembles the course actually played out by H1N1. In fact, the market predicted that H1N1 would spread widely and quickly but the mortality rate would be less than 1 percent during the first few weeks of the outbreak.

This market, like the one for seasonal flu, is a spin-off of a project that began with a Web-based tool we developed to forecast the outcome of presidential elections. Such markets now are used in a wide range of settings in place of or as supplement to traditional tracking methods like opinion polls and statistical forecasts, and they have proven to be remarkably accurate. Here, at the University of Iowa, we thought that the power of these markets, which assemble traders with lots of expert knowledge, could be used to predict the course of an emerging public health threat—like the H1N1 virus.

The H1N1 prediction market is a pilot project; we did not expect public health officials to base decisions on the information it provides for management of the current pandemic. But if these markets do remain as accurate as early indicators are showing, public health officials might one day use such tools to stay one step ahead of a quickly spreading outbreak.

Many of today’s public health officials rely on slow surveillance methods such as lab tests that – while undeniably critical – provide snapshots of the past. Using this new form of intelligence, public health officials could plan ahead and ramp up production of a vaccine, or perhaps launch an education campaign to curb the spread of the virus from one person to the next. We see great potential for prediction markets to harness the wisdom of the crowd – to quickly aggregate the beliefs and educated guesses, as well as hard evidence, held by a crowd of health experts and to present that information in an easily interpreted and accurate guide to the future.