Prediction Markets Accurately Predict


As I write this, the chance of Congress passing a law by December, 2006 that allows U.S. taxpayers to divert Social Security taxes to private accounts is about 1 in 100. The chance of Hamas recognizing Israel by year end is about 8 in 100, and that of a Palestinian state by year end is 2 in 100. Will the U.S. and/or Israel execute an overt air strike against Iran by the end of December? There is about a 20 percent chance it could happen (one chance in five.)

All probabilities are taken from contracts traded on prediction markets run by Tradesports. I have no monetary interest in this company and no independent opinions on the odds of these events. Those whose opinions differ and wish to bet against the market can do so at Tradesports. My interest in this article is (1) in the uncanny ability of prediction markets to predict events more accurately than individuals can, (2) what this fact means for such things as a terror prediction market and claims that markets are irrational, and (3) what this means for speculation.

We can dispose of item #3 quickly. It is easy to find articles urging speculation in one asset or another predicated on an air strike (perhaps nuclear) against Iran. The chances of this happening are actually somewhat lower today than last January. They were quite a bit lower, but they have been rising as Israeli bombing expands and nears Syria. The behavior of the odds over time can help someone decide what or what not to speculate in if he believes that they help measure the risk that the Lebanon War will spread and worsen. They do.

How do we get probabilities out of these contracts? Their prices give us that. Right now, the two main contenders in the National League Pennant race are the New York Mets and the St. Louis Cardinals. A contract buyer for a Mets win pays about 34 cents, and if the Mets win he collects $1. Disregarding interest on money and risk aversion, this is a fair bet if the chance of the Mets winning is a 34 percent chance. Then one is paying 34 cents for something worth 0.34 x $1 = $0.34. This is why the price of the contract is, to a good approximation, the market’s probability of the Mets winning the pennant. The Cardinals have a 29 percent chance of winning.

Prediction markets are a form of speculative market in which the contract prices directly measure the probability that an event will occur.

Political predictions

Elections are due in 2006 and 2008. The market expects Republicans to end up controlling the Senate after the 2006 elections. The probability of Republicans maintaining Senate control is about 75 percent. (If you bet 75 cents on the Republicans and they win, you’ll get back $1. If you bet 25 cents on the Democrats and they win, you’ll get back $1.) The House race is quite different. There the probability of the Republicans holding the House is 46 percent. That makes the Democrats slight favorites to win control.

Should the Democrats, who currently favor Hillary Clinton as their nominee, run her for office in 2008? Probably not, according to the prediction markets. At present, frontrunner Clinton has a 42 percent probability of winning the Democratic nomination, while John McCain leads the Republican field with a 38 percent chance. There are significant chances that others may win each of these nominations. However, at present the market gives Clinton a 20 percent chance of winning the presidential election and McCain a 22 percent chance. According to these Tradesports markets, the Democrats would therefore be wise to look elsewhere for a candidate. Are politicians sneaking a look at these prediction markets? They should.

Prediction market accuracy

In an article in support of rational markets, Mark Rubinstein relates this story:

"At 3:15 p.m. on May 27, 1968, the submarine USS Scorpion was officially declared missing with all 99 men aboard. She was somewhere within a 20-mile-wide circle in the Atlantic, far below implosion depth. Five months later, after extensive search efforts, her location within that circle was still undetermined. John Craven, the Navy’s top deep-water scientist, had all but given up. As a last gasp, he asked a group of submarine and salvage experts to bet on the probabilities of different scenarios that could have occurred. Averaging their responses, he pinpointed the exact location (within 220 yards) where the missing sub was found."

James Surowiecki in his book The Wisdom of Crowds tells the story of the game show "Who Wants to be a Millionaire" in which a contestant could ask an expert for help with a question or ask the audience. The experts were right 65 percent of the time, and the audience was right 91 percent of the time.

Jude Wanniski related a story told to him by Jack Treynor, a finance guru. Treynor had his class guess the number of jelly beans in a jar holding 850 beans. The average guess was within 3 percent of the total. Wanniski, by the way, correctly realized that this supported the efficiency of financial markets. He also, in my opinion incorrectly, construed this as proof of the efficiency of political markets, an opinion he expanded upon in The Way the World Works.

Prediction markets in general perform exceedingly well compared to individual forecasts. In his article on prediction markets, Philip O’Connor writes: "In fact, studies of prediction markets have found that the market price does a better job of predicting future events than all but a tiny percentage of individual guesses. The analysis below of the Virtual Super 12 shows the average selection, an average or constructed market price, to be better than 99% of participants’ selections."

He continues: "A short list of other evidence includes the following:

  • Markets that predict elections have been shown to outperform the predictions of opinion polls.
  • Prediction markets on movie box-office receipts and more obscure events have been shown to correspond closely with actual outcomes.
  • Sports gambling markets are excellent predictors of actual outcomes.
  • Laboratory experiments demonstrate that markets do the best job of aggregating information across participants in a controlled setting."

The bits of information possessed by independently thinking individuals are aggregated into market price, just as they are averaged into consensus judgments about jelly beans or correct multiple choice answers. The resulting outcomes tend to be more accurate than those of the individuals in the group and often more accurate than experts.

To make this happen, the individuals should make independent assessments. If they all get together as a committee, talk things out, and reach a consensus opinion, we probably will not find this result. People on committees influence each other in many ways, as anyone who attends such meetings knows. Anonymity is absent, and information and independent opinion often are suppressed.

Hayek’s 1945 paper on knowledge and prices begins to explain why prediction markets predict accurately. Hayek pointed out that knowledge is diffused among many individuals. It is hidden throughout society and changes according to particular circumstances of time and place. Prices aggregate this information. Hayek observed: "We must look at the price system as such a mechanism for communicating information if we want to understand its real function…"

Each person guesses at the outcome of a future event with error or noise. Some guesses are too low, and some are too high. In a sample of such independent guesses, the errors tend to cancel out when an average is struck. And if the outcome depends on many variables that no one person can assess but which many people might know a little about, the average will incorporate more variables than any single person might be aware of. In markets, if some people have better information than others, they are more willing to bet and bet more because they are more sure of the outcome. The bottom line is that prices tend to aggregate and therefore communicate information, although of course not perfectly.

Markets get it right, usually better than individuals do. Noise cancels out. Markets are on target, as much as anyone can be. Submarines get found. Winning horses get picked. Candidates who win get picked ahead of time. Predictions markets ignore a good deal of noise.

Speculative markets

From prediction markets to speculative markets like bond, stock, commodity, and foreign exchange markets is not a big jump, even if the speculative markets do not forecast specific events the way that prediction markets do. In both cases, diverse people with diverse information interact to predict the future. The outcome is a price.

Anyone who thinks that the prices in speculative markets are systematically or routinely off-base had better seriously consider the evidence that prediction markets provide superior forecasts of the future. Anyone who bets against the prices in speculative markets had better have a good reason why.

Speculation is very important to make prices reflect information. Some people win and others lose, but in the speculative process the information that bettors have about the future is reflected in the prices. Smart speculators know they are always speculating against an informed market consensus that reflects all sorts of information, private and public.

There is overwhelming evidence, and most mutual fund investors can confirm it, that mutual funds rarely beat the market for any length of time. They are simply unable to make a guess about future market prices that is better than the guesses built into today’s market price. Despite all their research and focus on speculation, they do not make up an elite group of speculators who consistently better the markets. This suggests that speculative markets are like prediction markets. In neither case can any but a small fraction of people guess (or judge) better than the market’s consensus guess (or forecast.)

In both prediction and speculative markets, there are plenty of people making bad guesses based on little more than hunches. But they tend to lose to the better speculators and the speculators who have new information. The latter are more willing to place larger bets and move prices from day to day if they deviate too far from the values they perceive; that is, if they see profit opportunities. Markets make mistakes, for sure. They can’t forecast anywhere near perfectly. However, they are also very hard to beat because their prices aggregate diverse information.


Markets prices are usually hard to understand, simply because they do impound more information than any single person is aware of or can dig up, even after the fact. Researchers often don’t know what to make of them. A substantial contingent of "behavioral" economics and finance researchers has discovered hard-to-explain prices (called anomalies) in speculative markets. After a good deal of psychological research, important parts of which I believe are greatly flawed (see here), they have leaped on a new bandwagon, that investors act irrationally. Much of the behavioral research asks its questions in ways that elicit seemingly biased or poor answers from human beings. (They are asked in terms of probabilities.) When the same issues are put to people in ways that use frequencies, the usual means by which human beings through the ages have had to approach probability questions in order to survive, the apparent irrationalities and/or biases of human judgment vanish.

Behavioral devotees believe that investors are irrational and that market prices are irrational. They think that humans have built-in judgment biases that can’t cope with certain types of problems involving risk. They ignore the fact that mankind has survived and that its ways of coping with risk fill a long book. Now law review articles are appearing that call for more government regulation of financial markets. The only irrationality in all of this is a warped political system that supports academics who thrive on publishing novel drivel.

The fact that the stock markets peaked out in 1999—2000 is cited in support of their case. The "irrationalists," the behavioral economics and finance types, pay no attention to the Austrian business cycle theory and no attention to Federal Reserve policies. They pay no attention to the fact that the markets dropped most severely after President Bush made it clear that Iraq was of paramount importance to him and that he intended war. They pay no attention to the costs imposed on corporations by Sarbanes-Oxley legislation. They pay no attention to the barriers to proper market operations introduced by the government itself, one of the largest being the capital gains tax and the capital gains tax holding period, another being the discriminatory treatment of short-selling and the gains from short-selling. The government’s taxes discourage the realization of capital gains and they discourage short-sales, which are taxed at ordinary rates.

But they are also paying no attention to the glaring fact that prediction markets accurately predict. This fact flatly contradicts the erroneous notion that market prices are set by irrational investors.

Terror prediction market

An agency of the Department of Defense called DARPA wanted to start a terrorism prediction market, but Congressional uproar quickly quashed it. Senator Dorgan called the plan "unbelievably stupid," and Senator Wyden called it "a federal betting parlor on atrocities and terrorism." Hillary Clinton said it was "a futures market in death." Daschle said "This is just wrong." He thought it provided "an incentive to commit acts of terrorism." Senator Boxer thought "There is something very sick about it," and thought the originators of the plan should be fired. Paul Wolfowitz said "I share your shock at this kind of program," and Senator John Warner was quick to obtain assurances that the program would be killed after obtaining agreement from Senators Pat Roberts and Ted Stevens. Both Republicans and Democrats were against this innovation. Whether or not these politicians wished to maintain appearances or cultivate votes or genuinely were shocked and indignant, which is the least likely hypothesis, it is obvious that many such markets would diminish the prerogatives of the powerful and restore more openness to government and more power to people at large. Naturally, officials viscerally object to such markets that even hint of a loss of control. They might even demonstrate that markets are good for something other than regulating, taxing, and extracting campaign contributions and favors.

Both Senators Daschle and Boxer argued that the terror prediction market gave terrorists an incentive to commit terror acts because they could make money on them. This is a serious objection to consider. In the same vein, betting on sports and other contests gives the players and jockeys an incentive to cheat while participating, to shave points, throw the fight, etc., but in these cases the racetracks and owners have incentives to keep the sport clean so as not to damage their franchises. Sports fans do not want to watch fixed contests, unless they are wrestling matches. Corrupted television game shows quickly lose audience. Although terrorists are not interested in reputation and these arguments do not apply to them, they too have an incentive. It’s to conceal their plans.

The Senators overlooked a number of factors that more than meet their objection. The first is that the more the terrorist tries to profit by his deeds in the prediction market, the more he raises the price. This makes public that the terror act has a higher probability. This raises the chances of discovery and alerts the potential victims, which in turn lower the expected damage of a terror act. (An increase in the contract price may call forth preventative measures that subsequently lower the contract price.) By entering the market, the terrorist may make money but he also lowers the chance of success of his act. This restrains his use of the terror market. In fact, if he wants secrecy, and he must want this in order to succeed, his primary incentive is not to participate in the prediction market so as not to reveal his plans.

The second factor is that the prediction market gives an incentive to anyone, including those close to terrorists who care more about money than terrorist success, to ferret out information about possible terrorist acts. This is analogous to offering a reward to turn in a terrorist.

The third factor is that terrorists already can profit by their acts using existing markets in gold, currencies, particular stocks, etc. It is easier for them to hide their intentions in these larger markets that reflect more information about other matters than terror. They already have an incentive to make money from their terrorist acts, and they already can do this more secretly elsewhere.

Fourth, the Senators implicitly assumed that only terrorists have information about prospective terror acts. However, there are all sorts of people in the FBI and CIA, others doing police work, and ordinary citizens who actually have information or might have information that they have an incentive to profit from. If they can trade in a terror prediction market, their information gets reflected in price. This is far better than a bureaucratic system of five colors that no one pays any attention to.

On the last point, O’Connor observes: "However, there is information that is difficult to communicate and may be found by fortuitous methods, or serendipitous information. Take an example of a possible terrorist attack where the information is extremely valuable to society. The claim that only terrorists would have information about terrorist attacks is clearly wrong. There were many stories after 9/11 of honest American citizens, from FBI agents to flight school instructors, being aware of suspicious behaviour indicating some type of terrorist attack. For example, the actor James Woods on a flight before the 9/11 World Trade Center bombings notified authorities concerning the behaviour of some Middle Eastern men in first class who he thought were going to hijack the plane. Nothing was done with this information. Later it turned out that these were actual 9/11 terrorists on a dry run. A prediction market provides a way for serendipitous information to be communicated via a market price, instead of through bureaucratic organisations that often fail miserably to make use of the many signs that people know about."

Despite the horror expressed by our noble, educated, and dedicated leaders who are busy taking this country to greater and greater heights, a terror prediction market is an idea that entrepreneurs no doubt have considered and will again consider. Having the government run such a market is of course the kiss of death. The Tradesports contract on an air strike on Iran shows that contracts on manmade destructive events are feasible and valuable. Following this contract enables the public to better understand and interpret the statements made on both sides and the various diplomatic and other actions. This contract allows businesses and others who might be affected by such an air strike to better plan for the future. No one is saying that this market provides an incentive for U.S. and Israeli authorities to initiate an attack in order to make money. If serious planning made such an air strike imminent, information leakage among the many persons involved would probably drive the contract price up. This would provide some warning to the Iranians and others who might change their diplomatic tune to prevent it. It would inform the public that such a step was on its way, which would provide a chance to stop it. For these reasons, neoconservatives may have an incentive to close down the market. Any such move would chill prediction markets for a time, but then they’d probably move further offshore.

Prediction markets have a bright future, as long as states leave them alone. They can be useful for everything from the introduction of new products to the success of research and development ventures and even to predicting terror events or a war’s success. The men and women who run enterprises and rely on information that filters through hierarchies can make better decisions with such markets. The benefits in terms of improved planning for future contingencies are clear. We can safely predict that we will see more prediction markets.

Political Theatre

LRC Blog

LRC Podcasts