Guessing Populations in a Market Setting

As Sunstein pointed out in his book Infotopia and as we discussed in class, one advantage of markets is that they allow people to portray their confidence by how much money they invest. If someone is confident that that they will be able to predict a certain outcome, they will typically invest more money than someone who is unsure of the outcome. This is a strength of prediction markets because the people with confidence have more of an impact than others with less confidence because they will invest more money, which will affect the market more. It is usually the case that the people who are confident have access to more relevant information that allows them to be able to predict an outcome more accurately. This helps to ensure that prediction markets reflect the relevant information that is available on the subject. Sunstein points this out in his book and provides several of examples that show the accuracy of prediction markets. He also points out situations where prediction markets have failed and highlights that they are especially ineffective where there is little information available and there exists a wide variety of possibilities for the outcome. One thing that came to my mind was the class exercise where we tried to estimate the population of Irvine, California.

In class last week we were discussing information cascades and tried to conduct an activity to see if an information cascade would occur. At first every member of the class took a turn guessing the population of Columbus, Ohio. Many people had no idea what the population would be and as a result the guesses were widely varied from student to student and no information cascade occurred. The class agreed that an information cascade would occur if there was one person who was very confident in their answer. The students who guessed after this person would adjust their guess towards the person who had confidently given their answer. To try to simulate this, one person in the class was given the correct population of Irvine, California but nobody knew who that person was. As a result, no information cascade developed because nobody could identify who had the correct answer.

After reading about and discussing markets I thought back to the class activity where we had tried to incite an information cascade. It became clear to me that if the activity had been approached as a prediction market where students wagered virtual money on the actual population, an information cascade would have almost certainly occurred. The students who were unsure of their guesses at first would wager very little money, which would allow the rest of the students in the class to see that they had no confidence in their answer. However, when it got to the person who had been given the correct population they would give their guess and wager a large sum of virtual money. This would signal to the rest of the class that the guess was made based on good information and the person wagering was confident in their answer. As a result, every student who guessed after that would predict a population very close to the guess given by the student who had wagered a lot of money. An information cascade would occur and the class would accurately predict the population of Irvine, California.

This example shows the power of a prediction market. Without the factor of betting virtual money there was no way for the students in the class to measure how confident someone was in their answer. This led to a wide variety of inaccurate answers and the class was unable to accurately predict the population. In the market setting though, the confident students with good information wagered high sums of money which outweighed the inaccurate guesses and the population was correctly predicted.