The Dissident
Donate | Contact | Masthead | Discuss



Everything You Know About Politics

The Thoughtless Orthodoxy

Wal-Mart: The Big Friendly Giant


Islamic Democracy; An Exercise in Futility?

The Wisdom of Markets Isn't the Wisdom of Crowds

Hunters, Liars, and Philosopher Kings

Modern Economics as a Flight from Reality




books + ideas + provocations
The Wisdom of Crowds or of Markets?

discuss this article

ON THE NIGHT BEFORE THE 2004 PRESIDENTIAL ELECTION, Jon Zogby, a leading pollster, confidently predicted a John Kerry landslide. At the same time, two electronic futures markets, and the Iowa Electronic Markets (IEM), indicated that George Bush’s chances of winning reelection stood at 51 percent, approximately the percentage of the popular vote that he in fact captured. While it may be surprising that groups of gamblers outperformed a well-informed expert, it actually happens all the time. The IEM has been more accurate than traditional polls at predicting the outcome of presidential elections since 1988, when it was created.

In The Wisdom of Crowds, James Surowiecki argues that remarkable yet commonplace events like these prove that under the right circumstances, a “group’s decisions will, over time, be intellectually superior to the isolated individual, no matter how smart or well-informed he is.” Displaying a dizzying ability to jump between economics, popular culture, computer science, biology, politics, psychology, and history, Surowiecki presents abundant evidence for his claim. To him, a group, and thus a case study, can be anything from a crowd at a livestock festival to a management team at NASA, a corporation, or a mass of investors. All of these groups, he finds, are “wiser” than individual experts.

Surowiecki is not proposing that some kind of magic powers of intelligence animate groups. To make wiser-than-individual decisions, he concludes, a group must have an adequate supply of three qualities: diversity, independence, and decentralization.

A diverse group is one whose members have different opinions or perspectives. Independence is the degree to which a group’s individual members develop their own opinions, without allowing the opinions of others to affect them. And a group is decentralized when its decision-making power is not located in one place or person. These three conditions allow a group’s members to draw on their own local and specific knowledge — even though (in fact, precisely because) nobody is in charge.

Surowiecki acknowledges that each group member’s opinions will contain both information and error. His key move is to claim that, given enough diversity, independence, and decentralization, the random errors of group members will cancel each other out. When the three conditions are satisfied, then, all that will remain is as much local information as possible, without any systematic bias that would distort it.

Thus, Surowiecki believes that we should “ask the crowd” rather than search for the smartest person or the most respected expert to solve a complicated problem.

Surowiecki impressively documents the pitfalls of expertise worship. While experts certainly do exist, their areas of competence tend to be so narrow that they rarely have enough information to make expert decisions. “There is no real evidence,” Surowiecki writes, “that one can become an expert in something as broad as ‘decision-making’ or ‘policy’ or ‘strategy,’” because decisions involve judgments of the future — about which nobody can possess very good information.

Furthermore, as Surowiecki shows, on almost every topic there is no single, definitive expert, but instead a large collection of them; and they are likely to disagree with each other. These disagreements resist resolution because experts organize their knowledge by leaning on theories that serve, for all practical purposes, as ideologies. Since The Wisdom of Crowds appeared, Princeton University Press has published Philip Tetlock’s Expert Political Judgment: How Good Is It? How Can We Know?, which impressively documents much the same point.

Surowiecki recognizes that his case for group wisdom may be interpreted as an endorsement of democracy, but he rebuts this interpretation on disappointingly narrow grounds. He does not believe that the electorate’s decisions should be presumed wise, he writes, because the political decisions entrusted to the people “involve values, tradeoffs, and choices about what kind of society people should live in,” and those decisions do not have objective answers. Therefore, they cannot be considered wise or unwise.

It is indeed true that some political disputes, such as the abortion controversy, depend on values and are not really disagreements over facts. But despite the hullabaloo over “values voters,” most political decisions involve choices about the means to achieve agreed upon ends, like economic prosperity, a functional education system, and military security. It’s possible, then, that Surowiecki’s thesis is even more ambitious than he admits. It may be that his argument is not merely neutral about mass democracy, but subversive of it. For when factual issues are at stake, the democratic “crowd” may be just as clueless as a so-called expert.

Unlike most writers, Surowiecki acknowledges the public’s overwhelming ignorance of facts that might be germane to deciding the best policy to achieve given objectives. Then again, according to his theory, as long as the three conditions hold, the members of a group don’t need to be well informed for them to make wise collective decisions; so maybe the electorate’s ignorance shouldn’t be taken to undermine democracy even when it comes to policy decisions that demand knowledge of facts. Why shouldn’t an electorate that meets Surowiecki’s three conditions reach wise decisions?

Unfortunately, almost all of Surowiecki’s examples of group wisdom involve crowds of relatively informed people solving simple problems, such as determining the number of jellybeans in a jar or the weight of a slaughtered ox. This type of evidence doesn’t suggest that groups of uninformed people will necessarily do well at solving complicated problems, such as those that democracy poses when it asks “the crowd” to decide on fiscal, military, or education policy. The evidence that comes closest to suggesting that groups of ordinary people can outperform experts is the impressive performance of markets. In fact, the success of markets is what motivated Surowiecki, a New Yorker business journalist, to study the concept of collective wisdom in the first place. But while markets do allow imperfectly informed and fallible people to produce collectively wise decisions, it doesn’t follow that the same people, as voters, will also make wise decisions. The reason is that Surowiecki’s three conditions of diversity, decentralization, and independence aren’t enough to explain the wisdom of market crowds.

To use one of Surowiecki’s examples, none of the consumers who contributed to the triumph of the gasoline-powered car in the early 1900s needed to know how the car or the alternatives to it worked; they could have been uniform in their opinions about that, and their opinions might have been uniformly wrong, but collective wisdom would have still emerged. All that’s necessary to correct a mistaken group opinion in markets is the ability of members of the group to try different alternatives. The wisdom of the crowd in Surowiecki’s example consisted of people’s ability to draw on the local knowledge gained from experimenting with different types of car, if they didn’t like the first one they tried. But if you’d asked any of the individual car buyers why their experience had shown that gasoline engines worked better than the alternatives, very few of them would have been able to supply a correct answer. If you put such a complex question to the whole group — as elections do — their answer probably would have been wrong.

Even a diverse, independent, and decentralized group wouldn’t have been likely to guess right about which type of engine was best unless they had already been able to experiment with different answers. Prior to their experience with different engines, posing that question to the crowd would have been like asking members of today’s public to guess at which theory of the origins of the universe is true: the probability that the crowd would guess correctly is minuscule, because their opinions would be uninformed by experimentation.

The advantage of markets, then, is not that by virtue of their diversity, independence, and decentralization, they make people more knowledgeable about complicated topics. The advantage is that experimentation — or what market ideologues call “freedom of choice” — makes each consumer’s evolving local knowledge of his or her own “preferences” suffice, regardless of how poorly informed he may be about the more complex question of how to satisfy those preferences. Different entrepreneurs provide different answers to that question, and without knowing anything about these different answers except how well they perform in the consumer’s experience, the crowd of consumers mimics wisdom without actually becoming knowledgeable about those answers. (It’s the entrepreneurs — not the crowd of consumers — who have to be diverse, independent, and decentralized, so that the consumers have different choices among which to experiment.)

Surowiecki claims that errors are random, and cancel each other out, so that the truth tends to prevail in the form of mass opinion. But the weakness of his book is in never examining the mechanism that ensures the canceling out of errors, and thus the emergence of truth — even under the conditions he specifies. It can’t be that the randomness of error alone ensures that “truth will out,” since no given opinion is as likely to be true as to be false. If there is only one truth but there are many opinions, most of them will be wrong, by logical necessity. The chance of somebody arriving at a given error may be random, but all those randomly generated errors will dwarf the truth, since the chance that someone will arrive at it is equal to the chance that they will arrive at any one among many possible errors.

Of course, there are degrees of truth and degrees of error. But what is the mechanism to pick out the most true opinion from the crowd of more-mistaken ones?

The scientist’s mechanism is laboratory experimentation. The market’s mechanism is a less controlled, but still effective, version of the same thing: the purchases of consumers, who can decline to repeat purchases if they’re dissatisfied with the first one. Consumers’ initially mistaken opinions about a very limited, simple topic — which product they will like — are corrected by their ability to “change their minds” and try alternatives if they don’t like the initial results. Best of all, their opinions about the question of what they prefer don’t need to be informed by anything but direct experience. Thus, their opinions about, say, automotive design can be completely wrong — as they surely will be, in most cases — but they can still unwittingly choose cars that have the best engines.

Contrast the situation in markets with the situation in politics. Imagine that instead of billions of consumer experiments happening every day, all Americans got one bundle of products every two or four years. That’s what mass democracy is like, when the “products” in question are public policies. The citizens of a democracy have no individual ability to experiment with alternative policies, and this reduces to the vanishing point their ability to gather evidence that would let them recognize, let alone reverse, their mistakes. In a democracy, the crowd acts as one. Everyone must follow the majority, and if things seem to be going badly, it’s just a matter of opinion whether the minority’s policies would have done any better. Everything now does boil down to how well informed people are about arcane matters — such as the cause of recessions or of wars going badly.

Political decision makers have to decide issues of social causation, not through direct experience akin to buying a lemon from Chevrolet and trying Toyota next time — social causation is too complicated for that — but by assessing abstract economic, sociological, and historical theories of the sort that divide even scholars in those fields. Lacking experimental evidence on these questions, the electorate must turn to its opinions, and every study has found these to be woefully uninformed.

Here is where Surowiecki is more subversive than he recognizes. His evidence of expert disagreement should give pause not just to Platonists, but to democrats. If rule by experts is a non-starter because there are few experimental correctives to ideologically biased expert judgments, the same is true of the political judgments of the crowd. As long as the best opinion is outnumbered by worse opinions — again, a logical necessity if, indeed, opinions are diverse — then there needs to be some mechanism to weed out all those bad opinions if democracy is to produce wise policy decisions. But if experimentation serves that function in markets and in (natural) science, then such a mechanism is sorely missing in politics. Whether we’re ruled by the few or the many, in politics we’re ruled by somebody’s opinion, and if that opinion is uninformed by experimentation, then it isn’t likely to be correct. Surowiecki’s book can therefore be read as an argument not about experts against crowds, but about rule by opinion — the expert’s or the crowd’s — against rule by experience, grounded in experimentation.

All political decisions, for lack of experimentation, will tend to be made by unchecked opinion. But not all decisions need to be political. Military strategy, perhaps. But school pedagogy, road construction, cable TV licensing, employment practices, consumer protection? In each area that’s governed politically, a legal monopoly imposes one opinion about the best public policy on all residents of the jurisdiction. This means that erroneous opinions are likely to prevail, and that errors are unlikely to be weeded out, because erroneous opinions are commonplace, and experimentation is banned by law. If individuals were freer to choose among employers, schools, and so on, they’d be able to discover their mistakes by generating “local knowledge” about what works and what doesn’t — without having to become pedagogical experts or traffic engineers.

Perhaps this means we should research trading political decision-making mechanisms for market mechanisms, or even transforming the former into the latter. By bringing attention to the problem with relying on (expert) opinion, and by pointing out the relative wisdom of markets, Surowiecki’s book is a commendable and fascinating first step on this path.

discuss this article
Matt Weinshall graduated from Harvard in 2002.






the articles

on your campus