ABSTRACT
Until recently, forecasting health-related events has been uncommon compared to other elds, such as metrology, nance, and geology. One reason is that traditional forecasting approaches require reliable data for long periods of time, and these data generally need to be updated quickly. Unfortunately, in comparison to other elds, health data is almost always old by the time it becomes available. The absence of timely disease data has motivated several syndromic surveillance efforts using alternative information such as drug sales (Hogan et al. 2003, Welliver et al. 1979, Magruder et al. 2003, Davies
CONTENTS
8.1 Introduction ................................................................................................ 145 8.2 Futures Markets ......................................................................................... 146 8.3 Prediction Markets .................................................................................... 147
8.3.1 How Do Prediction Markets Work? ............................................ 147 8.3.2 Requirements for Prediction Markets ......................................... 148 8.3.3 Examples of Prediction Markets .................................................. 150 8.3.4 Prediction Markets for Public Health ......................................... 150
8.4 Novel Inuenza A (H1N1) Case Study ................................................... 151 8.5 Conclusions ................................................................................................. 158 References ............................................................................................................. 159
at al. 2003), emergency department visits (Irvin et al. 2003, Yaun et al. 2004, Suyama et al. 2003), absentee data (Lenaway et al. 1995) and Internet search query log data (Polgreen et al. 2008, Ginsberg et al. 2009). In the case of seasonal inuenza, these approaches have provided improved lead-time over traditional disease activity reports (Dailey et al. 2007). However, all of these approaches rely on quantitative data that can produce “false alarms” or miss abrupt changes. Often the addition of human interpretation can supplement such quantitative data streams, but it is difcult to aggregate subjective data. In this chapter we propose a relatively new method for gathering and aggregating disease information. This method involves operation of specialized futures markets called prediction markets and inviting health experts to trade in these markets. The prices generated in these markets can provide a consensus view regarding the likelihood of future disease-related events. After a brief discussion of futures markets and prediction markets, we present data from a pilot novel inuenza A (H1N1) prediction market.
