Commonly Asked Questions about the Influenza Prediction Markets

Why was the influenza market started?
How can information from the influenza market be used?
How does the influenza market work?
Could the influenza market be applied nationally?
Who has access to data from the Influenza Market?
Specifically, how can the influenza market be used to help prevent influenza?
Have prediction markets been used for other purposes?
What are the main reasons that prediction markets work?
What are the future plans for the influenza market?
Can prediction markets be applied to other infectious diseases?
Why do the traders participate?
Do they actually earn money based on the success of their trades?
What is the difference between gambling and futures or prediction markets?
How are the predictions translated into timing and are the predictions accurate?
Are there any currently available forecasts for influenza activity?
What active surveillance for influenza is currently being done?
What are the limitations of Influenza Surveillance?

Why was the influenza market started?

Currently there are no methods to accurately forecast influenza activity.
If local health officials and health care workers could predict an outbreak, even 1-2 weeks in advance, they could initiate measures to modify the severity of the outbreak.
The Centers for Disease Control and Prevention devotes significant resources to influenza surveillance, but by the time this information is published, it is 1-2 weeks old.

How can information from the influenza market be used?

An accurate forecast of influenza activity would be helpful even 1-2 weeks in advance: measures could be initiated to modify the severity of the outbreak. For example, public health officials could allocate resources to increase vaccination rates among high-risk individuals and health care workers. Prophylactic medication, such as neuroamidiase inhibitors, could be administered to persons in high-risk groups (e.g. nursing home residents and staff). In addition, if increased influenza activity could be predicted, hospital administrators could plan ahead and ensure that adequate numbers of staff are available to care for an increased number of patients admitted for influenza complications.

How does the influenza market work?

At the beginning of the influenza season, we recruit a diverse group of health care workers with information about influenza activity. Each of the traders is given an educational grant of $100 with which to trade. This grant grows or shrinks over the course of the influenza season depending upon the accuracy of the individual trader's predictions. Potential traders include physicians, nurses, pharmacists, clinical microbiologists and epidemiologists. Each of these traders has access to unique information regarding current and future influenza activity. For example, if a clinical microbiologist suddenly sees an increase in the number of respiratory cultures positive for influenza, influenza will likely increase in her community during the next few weeks. Such a trader could log-on to our website and purchase shares for "widespread influenza activity" over the next few weeks. She would also likely try to sell her shares of "little-or-no influenza activity". On the other hand, if a nurse had seen several patients with influenza in the emergency room for the last few weeks and then saw no more patients, he would be tempted to sell all his "widespread influenza activity" shares and try to buy as many "little-or-no activity" shares as he could.

Each influenza "share" or "contract" is associated with a specified level of influenza activity during a specific week and is worth $1.00 when trading ends if that turns out to be the observed level of activity. Thus, the price at which trades take place represents the consensus belief about the likelihood of that level of activity. For example, if the price of widespread activity for a given week was $0.90, the market expects influenza to be widespread with 90% probability.

Each influenza contract is based on the Centers for Disease Control's weekly color-coded system: RED=Widespread, BLUE=Regional, PURPLE=Local, GREEN=Sporadic, YELLOW=No Activity. Individual trader's predictions are judged accurate or not based upon the levels of activity reported by the CDC.

Could the influenza market be applied nationally?

Yes, our software runs over the internet so any health care worker with information about influenza and with internet access could trade. However, we would need to recruit traders from around the country. Our software is capable of handling thousands of traders. Thus, we only need to recruit and train a group of traders from diverse geographic regions and diverse health care fields. Next year we plan to start recruiting traders from around the country from existing infectious disease surveillance networks.

Who has access to data from the Influenza Market?

The information from prediction markets would be available to anyone who wanted to log-on to our web site. The probabilities for different levels of future activity would be displayed in an easy-to-interpret graphical format. This information is updated with each new trade. Thus, the data is available in real time.

Specifically, how can the influenza market be used to help prevent influenza?

Although influenza epidemics occur each year, the peak activity can differ by up to 2-3 months, making it difficult to identify the best time for vaccination campaigns. The influenza season can start early (as it did in 2003-2004) before vaccination campaigns are completed. Alternatively, activity might peak three months after employees have been vaccinated against influenza (as it did in 2004-2005). Because antibody titers can drop significantly in the months immediately following vaccination, giving the vaccine too early may also be problematic.

It takes 1-2 weeks for the vaccine to generate a protective immune response. Thus, 2-3 weeks of advance warning before an increase in influenza activity could provide enough time for medical directors in long-term care facilities to increase vaccination rates among vulnerable populations and also to increase their staff vaccination rates.

A few weeks of advance notice regarding a surge in influenza activity would also help health care providers and pharmacists to anticipate increased demand for prophylactic antiviral medications (e.g., neuraminidase inhibitors). These medications, if given prior to the development of illness are effective at preventing influenza. Also, if given within the first 24-30 hours of symptoms they ameliorate the major symptoms associated with influenza.

Finally, knowing that influenza is circulating in the community or will soon circulate can dramatically increase the ability of physicians to correctly diagnose influenza and thus treat the illness.

Have prediction markets been used for other purposes?

Experimental prediction markets have been used successfully to forecast future events in several settings. For example, Eli Lilly has used internal markets to predict which developmental drugs have the best chance of advancing though clinical trials. Hewlett-Packard (HP) has used experimental markets to forecast sales of its printers; these markets have outperformed standard statistical sales forecasts. Similarly, The Hollywood Stock Exchange has accurately predicted Oscar nominees and repeatedly forecasts opening-weekend returns more accurately than does the movie industry. The Foresight Exchange operates markets to predict a variety of events such as whether specific mathematical conjectures will be proven.

What are the main reasons that prediction markets work?

Prediction markets, with as few as 20 or 30 traders, can yield good results, but the accuracy tends to improve as the number of traders increases. These markets work because they: (1) aggregate information from all participants, each of whom has different information about the issue in question; (2) provide incentives that encourage knowledgeable participants to reveal their knowledge in their trades; (3) provide feedback to participants-through market prices, traders learn about the beliefs of other traders and are motivated to collect more information; (4) allow traders to share their knowledge anonymously, thereby, encouraging traders to signal information through the market that they might not state publicly.

What are the future plans for the influenza market?

The Robert Wood Johnson Foundation just awarded our project a five-year, $1.1 million dollar grant to expand our influenza prediction markets. In addition to expanding into a national influenza market, we will open a market to help predict which influenza strains should be included in the subsequent year's vaccine. This can be accomplished by opening a market that aggregates information from influenza experts in the fields of virology, epidemiology, and clinical microbiology.

Our future plans also include opening an avian influenza market.

Can prediction markets be applied to other infectious diseases?

Prediction markets for infectious diseases also have applications beyond influenza. New prediction markets could be created almost overnight to address new emerging infectious diseases even before a specific etiologic agent has been identified. For example, information about SARS existed in the spring of 2003, and a prediction market for SARS could have quickly, accurately, and inexpensively aggregated expert opinion about this new disease. Such a prediction market would have been especially useful, given the absence of any preexisting SARS surveillance system.

Why do the traders participate?

Traders trade for three reasons:

They are interested in influenza and have information about its activity. Traders are interested in volunteering information in the same way that they are when filling out any other medical survey.
Unlike surveys, prediction markets allow participants to view what other participants think in real-time. This encourages traders to keep participating.
Surveys often provide a small financial incentive for participants to respond, but with prediction markets the amount of the incentive depends upon the accuracy of the participants' predictions. If participants do well, their account grows; if their predictions are inaccurate, their account shrinks. This performance feedback tends to spur further participation.

Do they actually earn money based on the success of their trades?

Traders in our market actually do get a check in the form of a financial grant to use for education purposes (e.g., buy books, or to help pay for a conference). The grant is awarded at the end of the season and the size depends on how successful participants were in predicting influenza activity.

The financial amount involved (traders start with $100) is not enough to really encourage traders to trade up to several times a week for an entire influenza season. Most participants reported that they truly enjoyed the process of trading itself and seeing how accurate their forecasts are compared to those of other traders.

What is the difference between gambling and futures or prediction markets?

In futures markets investors make investments based upon their beliefs about the future, and they may profit or lose depending upon the outcomes of their investments. Unlike gambling, however, futures markets provide some "social good," the most common of which is to allow investors to "hedge risk" or to protect themselves against future price increases.

For example, agricultural futures markets have caused a net stabilization of commodity prices, which greatly benefited the industry by reducing uncertainty about future prices.

The markets run by the University of Iowa are run for educational and academic research purposes, which is their "social good".

Since 1988, the Iowa Electronic Markets (IEM), run by University of Iowa's Tippie College of Business, has run markets to teach students at the University of Iowa and over 100 other colleges and universities throughout the world how markets work. Also, several peer-reviewed research articles have been written based on research employing IEM data. Vernon Smith, in his acceptance speech for winning the 2002 Nobel Prize in Economics in 2002, cited the IEM as one of the best demonstrations of how efficiently markets aggregate information about uncertain future events.

For more information about the IEM See: http://www.biz.uiowa.edu/iem/

The goal and "social good" of the influenza market is to produce influenza forecasts to help health care workers, public health officials and hospital administrators better prepare for increases in influenza activity.

How are the predictions translated into timing and are the predictions accurate?

For the 2004-2005 influenza season, we had traders make predictions from two week to eight weeks in advance. Every time a 2 week period closed another 2 week period opened.

Our predictions are very accurate from 2-3 weeks out, and even farther out traders are able to make predictions which would be helpful to the health care community. Because there are prices for each color contract, the traders give us the predicted probability that influenza activity is at one of the five influenza levels.

For the 2005-2006 influenza season, we are having traders make predictions from one to six weeks in advance, and every week during the influenza season is covered by a new set of contracts. Trading ends for a target week on Wednesday of that week. The following Saturday a new set of contracts for flu activity six weeks in the future is opened for trading.

Are there any currently available forecasts for influenza activity?

Currently, there are no widely available influenza forecasts. One could envision a very useful survey which would ask health care workers every week to predict how much influenza will be in the community during the next few weeks, based upon what health care workers are currently seeing. However, no such survey exists.

What active surveillance for influenza is currently being done?

Preventing influenza is a major public health priority and an important part of the effort to control the disease is surveillance. The seven components of the US influenza surveillance system include the following:

U.S. Influenza Sentinel Providers Surveillance Network (Figure A)
World Health Organization (WHO) and National Respiratory and Enteric Virus Surveillance System (NREVSS) Collaborating Laboratories (Figure B)
122 Cities Mortality Reporting System (Figure C)
New Vaccine Surveillance Network (NVSN) (Figure D)
State and Territorial Epidemiologists' Reports (Figures E and F)
Influenza-associated pediatric mortality
Emerging Infections Program (EIP)

Together, these networks are designed to provide a national overview of influenza activity.

What are the limitations of Influenza Surveillance?

Surveillance data is, inherently, a view of past activity, not a forecast of what to expect in the future. By the time most influenza surveillance data is collected and published, it is 1-2 weeks old.

The state and Territorial Epidemiologists' reports of influenza activity are the only state level information reported.

Thus, lots of data are being collected, but there are no currently used methods to aggregate this data in time to make useful predictions about future influenza activity.



Error processing SSI file	Commonly Asked Questions about the Influenza Prediction Markets Why was the influenza market started? How can information from the influenza market be used? How does the influenza market work? Could the influenza market be applied nationally? Who has access to data from the Influenza Market? Specifically, how can the influenza market be used to help prevent influenza? Have prediction markets been used for other purposes? What are the main reasons that prediction markets work? What are the future plans for the influenza market? Can prediction markets be applied to other infectious diseases? Why do the traders participate? Do they actually earn money based on the success of their trades? What is the difference between gambling and futures or prediction markets? How are the predictions translated into timing and are the predictions accurate? Are there any currently available forecasts for influenza activity? What active surveillance for influenza is currently being done? What are the limitations of Influenza Surveillance? Why was the influenza market started? Currently there are no methods to accurately forecast influenza activity. If local health officials and health care workers could predict an outbreak, even 1-2 weeks in advance, they could initiate measures to modify the severity of the outbreak. The Centers for Disease Control and Prevention devotes significant resources to influenza surveillance, but by the time this information is published, it is 1-2 weeks old. How can information from the influenza market be used? An accurate forecast of influenza activity would be helpful even 1-2 weeks in advance: measures could be initiated to modify the severity of the outbreak. For example, public health officials could allocate resources to increase vaccination rates among high-risk individuals and health care workers. Prophylactic medication, such as neuroamidiase inhibitors, could be administered to persons in high-risk groups (e.g. nursing home residents and staff). In addition, if increased influenza activity could be predicted, hospital administrators could plan ahead and ensure that adequate numbers of staff are available to care for an increased number of patients admitted for influenza complications. How does the influenza market work? At the beginning of the influenza season, we recruit a diverse group of health care workers with information about influenza activity. Each of the traders is given an educational grant of $100 with which to trade. This grant grows or shrinks over the course of the influenza season depending upon the accuracy of the individual trader's predictions. Potential traders include physicians, nurses, pharmacists, clinical microbiologists and epidemiologists. Each of these traders has access to unique information regarding current and future influenza activity. For example, if a clinical microbiologist suddenly sees an increase in the number of respiratory cultures positive for influenza, influenza will likely increase in her community during the next few weeks. Such a trader could log-on to our website and purchase shares for "widespread influenza activity" over the next few weeks. She would also likely try to sell her shares of "little-or-no influenza activity". On the other hand, if a nurse had seen several patients with influenza in the emergency room for the last few weeks and then saw no more patients, he would be tempted to sell all his "widespread influenza activity" shares and try to buy as many "little-or-no activity" shares as he could. Each influenza "share" or "contract" is associated with a specified level of influenza activity during a specific week and is worth $1.00 when trading ends if that turns out to be the observed level of activity. Thus, the price at which trades take place represents the consensus belief about the likelihood of that level of activity. For example, if the price of widespread activity for a given week was $0.90, the market expects influenza to be widespread with 90% probability. Each influenza contract is based on the Centers for Disease Control's weekly color-coded system: RED=Widespread, BLUE=Regional, PURPLE=Local, GREEN=Sporadic, YELLOW=No Activity. Individual trader's predictions are judged accurate or not based upon the levels of activity reported by the CDC. Could the influenza market be applied nationally? Yes, our software runs over the internet so any health care worker with information about influenza and with internet access could trade. However, we would need to recruit traders from around the country. Our software is capable of handling thousands of traders. Thus, we only need to recruit and train a group of traders from diverse geographic regions and diverse health care fields. Next year we plan to start recruiting traders from around the country from existing infectious disease surveillance networks. Who has access to data from the Influenza Market? The information from prediction markets would be available to anyone who wanted to log-on to our web site. The probabilities for different levels of future activity would be displayed in an easy-to-interpret graphical format. This information is updated with each new trade. Thus, the data is available in real time. Specifically, how can the influenza market be used to help prevent influenza? Although influenza epidemics occur each year, the peak activity can differ by up to 2-3 months, making it difficult to identify the best time for vaccination campaigns. The influenza season can start early (as it did in 2003-2004) before vaccination campaigns are completed. Alternatively, activity might peak three months after employees have been vaccinated against influenza (as it did in 2004-2005). Because antibody titers can drop significantly in the months immediately following vaccination, giving the vaccine too early may also be problematic. It takes 1-2 weeks for the vaccine to generate a protective immune response. Thus, 2-3 weeks of advance warning before an increase in influenza activity could provide enough time for medical directors in long-term care facilities to increase vaccination rates among vulnerable populations and also to increase their staff vaccination rates. A few weeks of advance notice regarding a surge in influenza activity would also help health care providers and pharmacists to anticipate increased demand for prophylactic antiviral medications (e.g., neuraminidase inhibitors). These medications, if given prior to the development of illness are effective at preventing influenza. Also, if given within the first 24-30 hours of symptoms they ameliorate the major symptoms associated with influenza. Finally, knowing that influenza is circulating in the community or will soon circulate can dramatically increase the ability of physicians to correctly diagnose influenza and thus treat the illness. Have prediction markets been used for other purposes? Experimental prediction markets have been used successfully to forecast future events in several settings. For example, Eli Lilly has used internal markets to predict which developmental drugs have the best chance of advancing though clinical trials. Hewlett-Packard (HP) has used experimental markets to forecast sales of its printers; these markets have outperformed standard statistical sales forecasts. Similarly, The Hollywood Stock Exchange has accurately predicted Oscar nominees and repeatedly forecasts opening-weekend returns more accurately than does the movie industry. The Foresight Exchange operates markets to predict a variety of events such as whether specific mathematical conjectures will be proven. What are the main reasons that prediction markets work? Prediction markets, with as few as 20 or 30 traders, can yield good results, but the accuracy tends to improve as the number of traders increases. These markets work because they: (1) aggregate information from all participants, each of whom has different information about the issue in question; (2) provide incentives that encourage knowledgeable participants to reveal their knowledge in their trades; (3) provide feedback to participants-through market prices, traders learn about the beliefs of other traders and are motivated to collect more information; (4) allow traders to share their knowledge anonymously, thereby, encouraging traders to signal information through the market that they might not state publicly. What are the future plans for the influenza market? The Robert Wood Johnson Foundation just awarded our project a five-year, $1.1 million dollar grant to expand our influenza prediction markets. In addition to expanding into a national influenza market, we will open a market to help predict which influenza strains should be included in the subsequent year's vaccine. This can be accomplished by opening a market that aggregates information from influenza experts in the fields of virology, epidemiology, and clinical microbiology. Our future plans also include opening an avian influenza market. Can prediction markets be applied to other infectious diseases? Prediction markets for infectious diseases also have applications beyond influenza. New prediction markets could be created almost overnight to address new emerging infectious diseases even before a specific etiologic agent has been identified. For example, information about SARS existed in the spring of 2003, and a prediction market for SARS could have quickly, accurately, and inexpensively aggregated expert opinion about this new disease. Such a prediction market would have been especially useful, given the absence of any preexisting SARS surveillance system. Why do the traders participate? Traders trade for three reasons: They are interested in influenza and have information about its activity. Traders are interested in volunteering information in the same way that they are when filling out any other medical survey. Unlike surveys, prediction markets allow participants to view what other participants think in real-time. This encourages traders to keep participating. Surveys often provide a small financial incentive for participants to respond, but with prediction markets the amount of the incentive depends upon the accuracy of the participants' predictions. If participants do well, their account grows; if their predictions are inaccurate, their account shrinks. This performance feedback tends to spur further participation. Do they actually earn money based on the success of their trades? Traders in our market actually do get a check in the form of a financial grant to use for education purposes (e.g., buy books, or to help pay for a conference). The grant is awarded at the end of the season and the size depends on how successful participants were in predicting influenza activity. The financial amount involved (traders start with $100) is not enough to really encourage traders to trade up to several times a week for an entire influenza season. Most participants reported that they truly enjoyed the process of trading itself and seeing how accurate their forecasts are compared to those of other traders. What is the difference between gambling and futures or prediction markets? In futures markets investors make investments based upon their beliefs about the future, and they may profit or lose depending upon the outcomes of their investments. Unlike gambling, however, futures markets provide some "social good," the most common of which is to allow investors to "hedge risk" or to protect themselves against future price increases. For example, agricultural futures markets have caused a net stabilization of commodity prices, which greatly benefited the industry by reducing uncertainty about future prices. The markets run by the University of Iowa are run for educational and academic research purposes, which is their "social good". Since 1988, the Iowa Electronic Markets (IEM), run by University of Iowa's Tippie College of Business, has run markets to teach students at the University of Iowa and over 100 other colleges and universities throughout the world how markets work. Also, several peer-reviewed research articles have been written based on research employing IEM data. Vernon Smith, in his acceptance speech for winning the 2002 Nobel Prize in Economics in 2002, cited the IEM as one of the best demonstrations of how efficiently markets aggregate information about uncertain future events. For more information about the IEM See: http://www.biz.uiowa.edu/iem/ The goal and "social good" of the influenza market is to produce influenza forecasts to help health care workers, public health officials and hospital administrators better prepare for increases in influenza activity. How are the predictions translated into timing and are the predictions accurate? For the 2004-2005 influenza season, we had traders make predictions from two week to eight weeks in advance. Every time a 2 week period closed another 2 week period opened. Our predictions are very accurate from 2-3 weeks out, and even farther out traders are able to make predictions which would be helpful to the health care community. Because there are prices for each color contract, the traders give us the predicted probability that influenza activity is at one of the five influenza levels. For the 2005-2006 influenza season, we are having traders make predictions from one to six weeks in advance, and every week during the influenza season is covered by a new set of contracts. Trading ends for a target week on Wednesday of that week. The following Saturday a new set of contracts for flu activity six weeks in the future is opened for trading. Are there any currently available forecasts for influenza activity? Currently, there are no widely available influenza forecasts. One could envision a very useful survey which would ask health care workers every week to predict how much influenza will be in the community during the next few weeks, based upon what health care workers are currently seeing. However, no such survey exists. What active surveillance for influenza is currently being done? Preventing influenza is a major public health priority and an important part of the effort to control the disease is surveillance. The seven components of the US influenza surveillance system include the following: U.S. Influenza Sentinel Providers Surveillance Network (Figure A) World Health Organization (WHO) and National Respiratory and Enteric Virus Surveillance System (NREVSS) Collaborating Laboratories (Figure B) 122 Cities Mortality Reporting System (Figure C) New Vaccine Surveillance Network (NVSN) (Figure D) State and Territorial Epidemiologists' Reports (Figures E and F) Influenza-associated pediatric mortality Emerging Infections Program (EIP) Together, these networks are designed to provide a national overview of influenza activity. What are the limitations of Influenza Surveillance? Surveillance data is, inherently, a view of past activity, not a forecast of what to expect in the future. By the time most influenza surveillance data is collected and published, it is 1-2 weeks old. The state and Territorial Epidemiologists' reports of influenza activity are the only state level information reported. Thus, lots of data are being collected, but there are no currently used methods to aggregate this data in time to make useful predictions about future influenza activity.
Error processing SSI file