Can Markets Predict Elections?
Can markets predict the next presidential election? On one sports gambling website people are literally betting they can. Other websites are running virtual "prediction markets" in an attempt to harness the information that eludes pollsters. They solicit participation from the general public in much the same way real markets do in the real world. And yet there are important differences between the gambling markets and the virtual markets. It comes down to whether traders are using real resources or just playing games.
The two approaches represent the difference between Mises's theory of calculation and Hayek's theory of knowledge. I hope to demonstrate that those services in which people use property and not just knowledge will be better predictors.
The topic of prediction markets made headlines last year because of a public relations disaster called the Policy Analysis Market. PAM was a multi-million-dollar Pentagon research program that would have involved investors risking small amounts of money (around $100) on contingent contracts related to the social, economic, and political future of Middle Eastern countries. But when certain US Senators learned of the program, they labeled it a terrorism futures market, and publicly denounced the idea. They claimed the project proposed "that we trade in death," and expressed outrage that the Pentagon was developing "a federal betting parlor on atrocities".
As the political hysteria died down, some journalists started to discuss the actual merits and risks of using market mechanisms to predict geo-political events. Criticisms run from general skepticism about mixing government programs with market structures to specific concerns about how the incentives of would-be terrorists might be affected. The advocates cite econometric studies and point to recent success stories among less politically charged experiments. Whether or not terrorism in particular can be successfully forecast through such markets, there is compelling evidence that decentralized systems of market-like bidding can generate more accurate predictions than traditional forecasting techniques such as surveys or focus groups.
Twenty years ago economist Richard Roll demonstrated that the price of orange juice futures predicted temperature change better than the National Weather Service. But it's easier to believe that natural, impersonal forces, such as the weather, can be accurately forecast than it is to believe in predictive systems more generally. What about future human events, such as entertainment or politics? These too have been the subjects of contingent contracts sold in futures markets. The success stories include the Hollywood Stock Exchange, which accurately predicted 35 of the 40 Oscar nominees in the top eight categories, and went on to pick all six of the top winners at the Academy Awards.
The Iowa Electronics Market (IEM) — which trades stocks for the candidates in upcoming public elections — is another famous winner. Over the course of 14 elections, the Iowa exchange's prices were on average a half a percentage point closer to the final results than were the final polls before voting. In addition, the market generates better predictions sooner than the polls do. According to Joyce Berg, a University of Iowa professor who helped organize the IEM's political exchange, markets "tend to predict events really well when no one person knows the answer — when information is distributed among many people with different knowledge bases."
This is Hayek's knowledge thesis in action. (Here are the lastest prices.)
The Hayekian Thesis
"The capitalist economy is, for Hayek, a valuable means of disseminating knowledge from one individual to another through the pricing 'signals' of the free market." — Murray N. Rothbard
In his 1945 article, "The Use of Knowledge in Society," Hayek attacks the would-be experts and central planners of socialism for failing to see that the knowledge necessary for a rational allocation of resources "never exists in concentrated or integrated form," that "the 'data' from which the economic calculus starts are never for the whole society 'given' to a single mind."
Hayek goes on to speculate that the 20th Century's widespread faith in socialist central planning was the result, in part, of the many significant advances in the natural sciences, and a general trend toward an age of experts and engineers. But according to Hayek and the Austrian School tradition that produced him, it is a mistake to apply to economics the methods of physics: "This misconception … is due to an erroneous transfer to social phenomena of the habits of thought we have developed in dealing with the phenomena of nature."
After the collapse of many would-be socialist economies in the East and more limited experiments in social engineering in the West, there is a new faith, however hesitant, in the power of markets and prices, and a general distrust of both centralization in particular and of expert elites in general. The social and political worlds are better understood to be extremely complex and ever changing, making them a poor fit for single minds and central plans.
Hayek seems to have been right: knowledge that is implicit, dispersed, and inaccessible by traditional, conscious methods can be organized through markets to create more rational calculation than can elite experts.
James Surowiecki, business columnist for The New Yorker magazine and author of The Wisdom of Crowds, has studied the developing technologies for accessing collective knowledge and is an advocate of prediction markets. But he suggests a certain irony: businesses, based in a culture of hierarchy and status-based leadership, might be slow to embrace the power of the more decentralized, more "democratic" mechanisms of the market!
I would like to suggest, however, that even if Surowiecki is correct about a certain reluctance from the traditional private sector, prediction markets will not achieve their full potential until they incorporate the power of profit-seeking self-interest, which can only be found in the private world of risk and reward. Government-developed "planned markets" such as the Pentagon's PAM, "virtual markets" such as the Hollywood Stock Exchange, and low-risk hampered markets such as the IEM (which only allows $500 trading accounts) will not be able to operate as efficiently or accurately as would true capitalist markets, which allow for real profit and real loss.
The Misesian Thesis
"For Mises, in short, the key to the capitalist market economy and its successful functioning is the entrepreneurial forecasting and decisionmaking of private owners and investors. " — Murray N. Rothbard
A quarter-century before Hayek published "The Use of Knowledge in Society," his mentor, Ludwig von Mises, shook the world of socialist economists and their fellow travelers with a monograph entitled, "Economic Calculation in the Socialist Commonwealth."
Mises's thesis was that economic calculation — the ability to allocate resources most efficiently toward the end products that are most needed or wanted — depends entirely on a price system, which reflects the balance of supply and demand.
The Misesian argument was devastating. Prior to his paper, the universally acknowledged challenge to socialism was the incentive problem, namely, Where is the incentive to work harder, to increase efficiency, to innovate — or even to work at all — when there is no pay-off for such behavior? Or, as it was often summarized, Under socialism, who will take out the garbage?
The socialist reply was that collectivist society would create collectivist psychology, a new Socialist Man who works for collective interest rather than old-fashioned self-interest.
Mises granted the socialists their absurd Socialist Man. For the sake of argument, he said, let's assume that everyone is well-intentioned, altruistic, and aligned in their collective goals. Even so, said Mises, without a free price system, scarce resources can never be allocated efficiently.
The calculation debate went on for years. In the end, many socialists were persuaded of the need for markets and for prices, but failed to see why markets and prices had to be based in private property, especially in the private ownership of capital goods — the famous means of production in Marx's language. Private ownership of capital goods is, after all, the formal definition of capitalism, and it was exactly such private ownership that socialism sought to abolish.
For the socialists, it was impossible to believe that central planning was less rational than the chaos of competing selfish interests. Their response to Mises — other than denial — came from Oskar Lange, who eventually became chief economic planner of Poland's Politburo. Lange said, "socialists have certainly good reason to be grateful" to Mises for forcing them to "recognize the importance of an adequate system of economic accounting to guide the allocation of resources in a socialist economy." He even suggested that a "statue of Professor Mises ought to occupy an honorable place in the great hall of the … Central Planning Board of a socialist state" in "recognition of the great service rendered by him" to the theory and practice of socialism.
Lange's "Misesian" revision to traditional Marxism came to be known as Market Socialism, wherein socialist planners could order managers to make up prices based on the costs of production. Given this starting point for prices, different bureaus in the central planning agency could behave as if they were bidding against each other for capital resources. Trial and error, it was claimed, could bring supply and demand together in rational socialist prices.
The later and more influential response was a claim from mathematical economists that given sufficient knowledge of economic conditions, there existed various mathematical equations that could solve the problem of efficient allocation.
It was in response to these mathematical socialists, that Hayek wrote his "Use of Knowledge" essay, emphasizing that their given-sufficient-knowledge condition was so impractical as to yield the entire socialist mission nearly impossible.
Unfortunately, Hayek's knowledge thesis came to overshadow Mises's original point that property-based prices are indispensable to rational calculation. For years, economists talked as if there were a single, homogenous Mises-Hayek thesis from the Austrian School — as if Hayek had incorporated and expanded on Mises's original argument. Toward the end of the Twentieth Century, however, Misesians began to "de-homogenize" the thesis. They insisted Mises's original claim — that there is no true cost-benefit analysis without actual human beings bearing the costs and reaping the benefits — is distinct from, and stronger than the later Hayekian emphasis on knowledge coordination.
By the fall of the Berlin Wall and the collapse of Soviet communism in 1989–90, Western economists were forced to acknowledge what East-Block economists had already learned the hard way: Mises was right!
But is this concession stronger than the claim that Hayek was right?
Prediction markets offer us an opportunity to compare — on Hayek's territory, so to speak — the strength of Mises's original emphasis.
The Hollywood Stock Exchange, for instance, where both stocks and dollars are only virtual, can be seen as exactly the sort of "play market" that Mises dismissed the Market Socialists for proposing: "They want people to play market as children play war, railroad, or school. They do not comprehend how such childish play differs from the real thing it tries to imitate."
But as we've already seen, the Hollywood Stock Exchange, however "unreal" it is as a market, is nevertheless a prediction system superior to even rigorous implementations of more traditional methods. Aren't we guilty of the perfection fallacy — contrasting the very best available reality with some hypothetically perfect alternative — if we say that a virtual Hayekian market is inferior to an all-together non-existent Misesian competitor?
Profit, Research & Arbitrage
"It is at least conceivable that one can order a manager to play market and act as if he were enjoying the profits and suffering losses; but it is clearly ludicrous to ask investors and capital speculators to act as if their fortunes were at stake." — Murray N. Rothbard
The Misesian competitor is not, of course, all-together non-existent, since actual, property-based, profit-driven commodity futures markets operate so efficiently that, as noted above, they beat the predictive ability of the National Weather Service.
Another empirical comparison could be conducted between the accuracy of the odds generated in traditional horse race gambling to any of the virtual Hayekian systems. The fact that a horse running at 2-to-1 odds will tend to win a third of the time tells us a lot about the ability of the gambling market to generate rational predictions. How accurate would those odds be in a system where gamblers couldn't collect their winnings?
Without conducting an empirical study on sports gambling, a couple of simpler theoretical demonstrations are available. The first is the cost of research and the need for profits to justify such costs; the second is the absence from virtual markets of the purely profit-based phenomenon of arbitrage.
The research point is so straight-forward that it can be addressed in two simple questions:
1. Is it possible that research undertaken by market players might improve the accuracy of their bids?
2. Will such research take place in a "market" that won't pay for it?
The arbitrage issue requires a bit more explanation. The first time I looked at TradeSports.com — an Irish sports gambling site that operates through an international exchange of contingent contracts — they were putting the chances of President Bush's re-election at 60% (i.e., a $100 BUSH WINS contract was priced at $60). The clients of TradeSports.com spend real money on real contracts, taking real risks in pursuit of real reward. The virtual prediction market site, NewsFutures.com also offers contingent contracts on President Bush's re-election chances, but both purchases and pay-offs are done with play money. Their "market" gave Bush only a 49% chance of re-election. This was in the spring of 2004. One obvious empirical test would be to compare how close the different sites come to predicting the actual election results — will the local knowledge of the local American website outperform the profit-driven predictions of the international gamblers?
As I write this, in the summer of 2004, I can't yet know. What I do know is that an 11-point discrepancy in contract prices couldn't last if both markets were based on the risk of real money. If both markets really paid a hundred dollars for their $100 contracts — and if they were both active enough to allow a large number of such contracts to be traded — then arbitrageurs could guarantee a certain return in November. By buying the cheaper contract in each market — the $40 BUSH LOSES in one market and the $49 BUSH WINS contract in the other — they could guarantee a $100 pay-off for an $89 expenditure (ignoring transaction costs). Note that the arbitrageur doesn't need to have any opinion of Bush's re-election chances: he is, after all, betting both ways. All that is necessary for the arbitrage profit opportunity is a disagreement between the two markets: a discrepancy in their prices.
The very activity of arbitrage would, of course, drive the two markets toward a common price: by raising the demand for the cheaper contracts, arbitrage would drive up the price of each, until the discrepancy is eliminated. As long as the discrepancy remains, the arbitrage opportunity will remain.
Fine, you might say, but all that shows is that arbitrage will force the markets to reflect each other's predictions. It won't necessarily produce a better prediction in the process.
This objection assumes, however, that the original discrepancy was arbitrary, and that closing the gap between the two prices will only produce a new-but-still-arbitrary consensus price. If that were true, a simple averaging of the two prices could eliminate the discrepancy. But arbitrage isn't likely to produce a price that merely splits the difference.
Let's say that Market 1 sells $100 Event X contracts for $40 (and therefore sells Not-X contracts for $60). Market 2 has those prices reversed: it gives Event X a 60% likelihood. The arbitrageur will buy Market-1 X contracts for $40 each, paired with Market-2 Not-X contracts for $40 each, guaranteeing a $100 payoff for each $80 contract pair. As more arbitrage takes place, the gap between the Market-1 and Market-2 prices will disappear, but arbitrage would only result in an average price of $50 if the two markets were equally informed, equally confident, and equally liquid. What is far more likely to be the case is that one market will have "stickier" prices than the other. The price of an Event X contract in Market 1 might well rise above $40 as a result of arbitrage between the two markets, but it might move slowly and by smaller increments than the falling price of Event X contracts in Market 2. If the post-arbitrage price turns out to be $45 instead of an even split of 50/50, then the 45% odds given to Event X are likely to reflect a greater informedness, greater expertise, or greater confidence in the original Market-1 prediction against the chances of Event X coming to pass.
The arbitrageurs need know nothing about Event X. They don't need to have any opinion on the likelihood of any future event related to Event X. All they need are investment funds, a grasp of mathematics, and a self-interested pursuit of profit. But the result of their activity is a consensus price that better reflects the beliefs of the more confident and informed market players. The blind pursuit of profit squeezes the untapped potential from the prediction markets.
The prediction market advocates would certainly agree with Hayek's elaboration, but they have lost track of the original argument that Hayek was trying to elaborate. The more limits that are placed on the profit potential of prediction markets, the more limits are being put on their predictive power. The magic of market mechanisms goes well beyond the creation of consensus from a decentralized distribution of local, implicit, sometimes unconscious knowledge — the power of the market lies in its ability to harness the power of self-interest. By focusing on the question of knowledge, Hayekians lose track of the critical role of the arbitrageur, who contributes exactly as Adam Smith would have described:
… he intends only his own gain, and he is in this, as in many other cases, led by an invisible hand to promote an end which was no part of his intention … By pursuing his own interest he frequently promotes that of the society more effectually than when he really intends to promote it.
What the advocates of prediction markets need to rediscover is the importance of Mises's original insight — or perhaps just Adam Smith's.
 "Amid furor, Pentagon kills terrorism futures market." CNN.com/Inside Politics. Wednesday, July 30, 2003 Posted: 1:37 PM EDT (1737 GMT)
 R. Roll, "Orange juice and weather," American Economic Review, 74, 1984, 861–880.
 Murray N. Rothbard, "The End of Socialism and the Calculation Debate Revisited," Review of Austrian Economics, 1991, 5(2), pp. 51–76.
 F.A. Hayek, "The Use of Knowledge in Society," American Economic Review, XXXV, No. 4; September, 1945, pp. 519–30.
 "Die Wirtschaftsrechnung im sozialistischen Gemeinwesen." Archiv für Sozialwissenschaften, vol. 47 (1920).
 Oskar Lange, "On the Economic Theory of Socialism," Review of Economic Studies 4:123–142. 1936.
 See the writings in particular of Joseph Salerno and Hans-Hermann Hoppe in The Review of Austrian Economics. See also Salerno's postscript to the 1990 edition of Economic Calculation in the Socialist Commonwealth issued by the Mises Institute:
 Robert Heilbroner, "Reflections After Communism," The New Yorker, September 10, 1990, and "The Triumph of Capitalism," The New Yorker, January 23, 1989.
 Adam Smith, An Inquiry into the Nature and Causes of the Wealth of Nations. 1776.