Justin Wolfers, assistant professor at Wharton University in Pennsylvania, conducted an interesting study on what he calls “prediction markets*,” i.e. markets in which the participants put a price on the probability that a particular event will occur. The odds that bettors place on horse race results and on the probability that a certain horse will win, and on the outcomes of boxing matches and of baseball, football and hockey games, are examples of markets where, by observing the opinions of participants expressed as the price they are willing to pay to bet on the probability of a result, a general consensus is reached on the probable result of an event. An interesting aspect of this study is that derivatives are also part of prediction markets: options put a price on the probability that the price of an underlying asset will reach a certain level within a certain period of time. And futures contracts illustrate the consensus that investors have reached on the price that an underlying asset will reach at a future date. For example, Federal Funds futures represent the probable Fed Funds rate at the expiration of the contract, which rate was fixed according to a consensus established by market participants.
Wolfer’s study is based on the efficient market hypothesis (EMH), according to which stock prices fully reflect available information and react almost immediately and objectively to any new information. In the strong form of the hypothesis, prices reflect all available information; in its semi-strong form, prices reflect all publicly available information; and in its weak form, prices reflect only past data. Whatever the form, prices can reflect, summarize and aggregate a huge amount of information. In fact, it is on this premise that technical analysis is essentially based: as the change in the price of a stock is the result of a change in the general consensus, it reflects, summarizes and aggregates a host of information. But what is the predictive value of such a consensus?
To answer this question, Justin Wolfers compiled the outcomes and odds of 21,885 Major League Baseball games played between 1991 and 2000 and 3,791 National Football League games from 1984 to 2000.
The results are telling: he found that when the market gave a team a 33% chance of winning, the team won 31% of the time; when the market established the odds at 67%, the team won 65% of the time; and when the market set the odds at 90%, the team won 88% of the time. So the consensus has a real predictive value. To profit from these predictions, investors have to exploit the error. Predicting an outcome doesn’t pay off, because of the odds they have to accept or the price they have to pay to place a bet.
Take, for example, an at-the-money option (at par) which expires in one month and which is offered for one dollar. What the market is telling us is that according to highly complex statistical formulas, it is very unlikely that at the end of the month, the price of the stock will be more than one dollar above the current price. In addition, this probability gradually decreases with time. That is why very few investors are able to profit from long positions in the case of options held to expiration. To make a profit, the persons who established the consensus must be wrong and, although this is often the case, the error must be sufficiently large and occur in the right direction. Consensus errors are random and unpredictable. That is why it may be preferable to buy options when the implied volatility is lower than the historical volatility, in the hope that the former will quickly realign with the latter. You then have to hope that the market underestimated the potential change in the value of the options.* Justin Wolfers, Prediction Markets: The Collective Knowledge of Market Participants, CFA Institute Conference Proceedings Quarterly, June 2009.
** Investors should read the document describing the risks inherent in trading options. Options are not suitable for all types of investors.