Prediction Markets

Introduction

My goal for this article is to explain prediction markets and what can be done with them.

Example

Suppose we want to know if Ron Paul will become the next president. The traditional way to gain this knowledge would be to conduct surveys. Since this is boring i want to explain another approach.
Suppose we have contracts that grant the holder of that contract one dollar if Ron Paul becomes president. Otherwise the contract is naught.
How much would you pay for such a contract? The fair price of the contract is clearly the expectation that Ron Paul becomes president. Suppose he becomes president with 20% probability. Then with 20% the contract is worth one dollar, otherwise zero. Thus the risk-neutral price is 0.2 dollars.
The problem is that we do not know the probability in advance so prediction markets work the other way around. We simply issue the contract and let the price be influenced by demand and supply. If you think the contract is too cheap and the fair price should be higher, you can buy some contracts and make money if you were correct. If the price then settles at 0.3 dollars, we can interpret this as the probability of Ron Paul becoming president.

Theory

Of course this not only works with presidency, but with any binary event that eventually “expires”. For example you could bet on the weather tomorrow. By issuing multiple contracts, we can also bet on discrete (non-binary) events. But where does the money come from? Well, insiders are better informed than the general public and thus they can make money with their information. But by providing this information through buying or selling such a contract, it becomes public in the price of the contract. So the money gained is simply the price which was included in that released information.

Problems

Why aren’t we all using prediction markets? One problem is that often we have a lack of liquidity. Then the spread becomes too large to make accurate predictions. For example, what does it tell you, if you can buy a contract on an event at 0.6$ and sell it at 0.3$? It only gives you an interval for the probability of that event occuring.

Another issue occurs when people can issue their own bets. For example they could bet on the date of the death of famous persons. This is similar to how assassination markets work. In an assassination market one bets on the date of the death and if it does not happen, the money spent is simply collected in a pot. However, if one is correct about the date of death, then one receives the pot. This provides an incentive to murder the person in question, after betting on the correct date of death which the murder knows, and thereby collect the pot. So, prediction markets in the worst case can degenerate to crowdfunded murder by providing financial incentives to change the outcome of the event underlying the bet.