Summary:
In this article researchers took the methodology of prediction markets also known as information markets, or future markets, and applied it to the medical field. The purposed study purposed that a group of medical experts from various field who were privy to different bodies of information could aggregate and accurately forecast what FLU strains would be prevalent in the upcoming year.
Prediction markets work by having "traders" invest money in outcomes that they believe will come true. As time progresses and more traders invest in differing options, the buy in prices for probable outcomes shift to become more expensive and less expensive to invest in low probability outcomes. These market prices reflect the traders collective confidence that an outcome will happen. Due to the potential to earn money for a correct selection the participant is encouraged to invest in something that will yield a payoff.
As the more people invest into the market that as time progresses that market will come closer and closer to mirroring what will actually occur in the future. Prediction markets have a very high forecasting accuracy. For example if eighty-five percent of investors suggest this will happen, then that has an eighty-five percent chance of happen. This mimics the principle that the more one flips a coin over time the closer it will be to fifty percent heads and fifty percent tails.
Three things are highlighted that are key to making a prediction market valid as a methodology. The first is that a prediction market needs a diverse group of traders, to pull information from a large amount of sources. Sufficient amount of traders is also required as there must be enough to encourage market value changes but not so many that there is a large amount of error. The correct number of traders is still up for debate as research has yielded a wide array of results with different amount of traders and there has been no consensus of the issue. The third criteria a prediction market must have is that there must be incentives to trade within the market or it would essentially be a betting pool and not a market.
When in application 62 medical professionals from various subfields in medicine participated in the market. The market introduced new contracts, which are the equivalent to shares, into the market every two weeks. As the market went on the predictions where more accurate throughout, both with correctly forecasting the strain color, but also within one color variation (See Figure 1). The researchers concluded that even though there was some wiggle room within the results that given the volatility of the FLU market it was an overall success.
Critique:
Prediction markets have inherent strengths of having a malleable method throughout the course of an experiment to accurately represent what is likely to happen in the future. Prediction markets are versatile, as they can be applied to many different fields. They also trump surveys as they allow for informed decisions throughout the entire process by being allowing trade through the process, as opposed to a survey which is filled out once. This allows for confidence in an event, to rise and fall appropriately leading to an overall better forecasting agent. The downfall to prediction markets are they require a great deal of time and effort to set up as well as run. They also require people who care about an outcome to invest into a market with money. This investment raises an ethical question in some fields were it may seem inappropriate to be investing physical money such as medicine, or college sport outcomes. The last pitfall prediction markets have is that they can be swayed by manipulators who are able to purchase large portions of a market if said big buyer wanted to suppress an unpopular opinion to them.Sources: Use of Prediction Markets to Forecast Infectious Disease Activity
Posts
At first Ian, I did not understand why the researchers organized their research the way they did until I look at figure 2 within the source. With the color-coded system that they used from the CDC (red, widespread; blue, regional; purple, local; green, sporadic; yellow, no activity) they were able to accurately forecast the onset, duration, and intensity of influenza activity in Iowa week by week. The ability to forecast the extent of disease activity and the timing and geographical location of influenza outbreaks would be helpful for the medical field to prepare for such events. However, the authors did mention that prediction markets will never replace traditional infectious diseases surveillance, though they view it as a potential supplement. In my opinion, prediction market could be used to design more sensitive and efficient traditional surveillance systems within the future.
ReplyDeleteMichael I agree with your point that fine-tuned prediction markets could help better identify influenza strains in the medical field. However I believe that traditional infectious disease surveillance systems were what the researchers were trying to aggregate data from. In essence the prediction market would be an average of all the possibilities and probabilities of certain strains becoming a problem. It would boil down to looking for outliers in nominal group technique, based upon maybe one field such as pediatrics found a very contagious strain that other fields would not have seen or detected.
ReplyDeleteDid the article highlight any keys to making prediction markets valid that are applicable to other areas?
ReplyDeleteYou mention that prediction markets require people who care about an outcome to invest into a market with money. With that being said, is there any evidence whether prediction markets using real money with traders (as a reward) produce better accuracy than prediction markets with fake money? I believe creating a prediction market design using real money would create an incentive for traders to become more fully invested in predicting outcomes.
ReplyDeleteI find it interesting how many of these studies used prediction markets on medical research. I would never think to do that. I like how versatile this methodology is.
ReplyDelete