The Wisdom of Crowds

When the Many are Smarter than the Few and How Collective Wisdom Shapes Business, Economics, Societies, and Nations

1/1/2005

By James Surowiecki

Doubleday 2004 Paperback Buy here

Book Review
Reviewed by Ilya Somin

Is it possible for the seemingly ignorant many to make better decisions than the expert few? To what extent can ordinary people be trusted to make their own decisions in economic and political markets? Political philosophers and social scientists have debated these questions for centuries. James Surowiecki’s The Wisdom of Crowds is a readable and engaging summary of the evidence showing that “crowds” of ordinary people often can and do make better decisions than small groups with a greater average level of knowledge and expertise. While many of Surowiecki’s points are convincing, he fails to take adequate account of the impact of incentives and selection effects on the ability of ordinary people to acquire knowledge and use it effectively. As a result, his argument, while very persuasive for many private sector settings, is too optimistic about the ability of voters to make good decisions in the political process.  Nonetheless, Surowiecki’s book  is a well-written contribution to a crucial debate, one that might help counter the elitist tendencies of much political and intellectual discourse.

How Crowds Can be Wise

 Intuitively, it may seem unlikely that a large crowd of laypeople might make better decisions than a small group composed of experts on the issue in question. Yet, as Surowiecki  recognizes, this counterintuitive conclusion might nonetheless hold true because the large crowd might have a greater total amount of knowledge even if the small group of experts has a greater average amount. For example, if each of 1 million laypeople has 1 unit of relevant information, their total knowledge of the subject turns out to be greater than that of 100 experts each of whom has 1000 units. Although each individual expert knows 1000 times more than each individual layperson, the total knowledge of all the laypeople is ten times greater than that of the experts  (1 million total information units versus 100,000). This core insight is not a new one; Aristotle (whose contribution Surowiecki unfortunately fails to acknowledge) made a similar argument 2,400 years ago in his reply to Plato’s critique of democracy.  Yet, Surowiecki makes a useful contribution in outlining situations in which the greater superior total knowledge of a crowd can be effectively harnessed in ways that may lead it to outperform smaller, more expert, groups. 

Surowiecki argues that “four conditions characterize wise crowds: diversity of opinion (each person should have some private information, even if it’s just an eccentric interpretation of known facts), independence (people’s opinions are not determined by the opinions of those around them), decentralization (people are able to specialize and draw on local knowledge), and aggregation (some mechanism exists for turning private judgments into a collective decision” (10). Diversity is useful because if each of the members of the crowd not only has just a small amount of knowledge but is limited to the same knowledge as everyone else does, there is little to be gained from aggregating it. If each of the million laypeople in my example above had the same one unit of information, their combined total knowledge will also be one, not one million. Independence facilitates good decisionmaking, Surowiecki argues, because it prevents people from simply going along with an opinion because others hold the same view, thereby depriving the group of the benefits of their particular bit of knowledge. Decentralized specialization, in turn, is useful because it increases the likelihood that people will have knowledge that is both nonduplicative and helpful. Finally, aggregation is essential because a group’s knowledge has little value unless it can be harnessed to improve decision-making. 

Surowiecki provides some striking examples of cases where “crowds” of laypeople outperform smaller groups of experts in situations where his four criteria are met. For example, betting lines on professional sports contests created by bookies on the basis of wagers placed by millions of nonspecialist bettors are extremely accurate at predicting the outcomes of games (11-14), usually more so than the picks of experts. Similarly, even the most sophisticated professional investors and mutual fund managers fail to consistently outperform the stock market as a whole, which reflects the judgments of millions of nonexpert investors with an enormous amount of cumulative diverse knowledge (233-35).  Perhaps most interesting of all, the decentralized knowledge of the stock market accurately determined which of several NASA contractors was responsible for the defects that caused the 1986 explosion of the Challenger space shuttle months before aa expert Presidential Commission reached the same conclusion (7-10). Within hours of the Challenger’s demise, the guilty firm, Morton Thiokol, saw its stock fall precipitously, while other space shuttle contractors suffered much smaller drops and soon recovered most of their losses even as Thiokol stock continued to fall (7-8). Six months later, the presidential commission of inquiry confirmed that a defective Martin Thiokol part was indeed the cause of the tragedy (8). 

Surowiecki also takes note of situations where the failure to meet one or more of his preconditions leads to dysfunctional collective decisions. For example, highly homogenous groups, especially small ones, might be prone to “groupthink” in which the absence of diversity leads the group  members to reinforce each other’s preexisting biases rather than make objective decisions (ch. 2). Groupthink among NASA officials may have been responsible for the most recent space shuttle catastrophe in 2003 (173-76). Similarly, although the United States had large number of diverse intelligence agencies prior to the September 11 attacks, they failed to properly assess the threat posed by Al Qaeda in part because there was no effective mechanism for aggregating their diverse knowledge (66-69). 

Many of Surowiecki’s examples are both interesting and compelling. He effectively demonstrates that a large crowd can often be  “smarter” than a small group of experts.

Incentives and Selection Effects

Unfortunately, Surowiecki’s account is not as persuasive as it might have been because he largely ignores the importance of incentives and selection effects in increasing the odds that large groups will make the right decision. It is no accident that many of his best examples of good decisionmaking by crowds come from the operation of markets, notably sports betting and stocks. Market participants have strong incentives to try to acquire accurate information and act on it rationally because - quite simply - doing so enables them to make a bigger profit. It is particularly important that individual actors in markets can profit from increasing their knowledge even if all or most other market participants remain ignorant. Indeed, investors who correctly predict an increase or decrease in the value of a stock that others fail to foresee stand to make particularly high profits. Sports bettors who correctly foresee the victory of an underdog stand to make a killing.  While Surowiecki is clearly aware that incentives matter, he fails to systematically integrate this fact into his analysis. 

What economists call “selection effects” also have an important impact in determining whether a large group reaches correct decisions or not. Even if investors in the stock market choose to ignore relevant information or otherwise act foolishly despite strong incentives not to do so, the market can still function effectively in the long run. An investor who fails to respond to incentives and as a result makes bad decisions will tend to lose money over time   Conversely, those who make good decisions will tend to increase their wealth. As a result, competitive markets will, over time, transfer wealth from bad decisionmakers to good ones, thereby increasing the overall quality of decisionmaking in the system. Moreover, the existence of selection effects ensures that many market participants who are unwilling or unable to make good decisions will choose not to enter that particular market in the first place. Certainly, people who are completely ignorant of sports are far less likely to become sports bettors than are avid fans. Thus, poor decisionmakers are “selected out” of market processes, both because they tend to lose money over time and because  the prospect of failure may deter them from even entering the market in the first place.

Markets vs. Voting

 Surowiecki’s neglect of the importance of incentives and selection effects is particularly evident in his discussion of democracy, which he praises as an effective example of large-group decisionmaking (260-70). Surowiecki recognizes that most voters are often ignorant of even basic facts about politics and public policy (265-66). For example, he notes that  most do not understand the basics of how Social Security works, and a 2003 poll found that half the population did not realize that there had been a tax cut during the previous two years (266). In fact, survey research provides examples of ignorance even more glaring than these. For example, in a 2004 survey, 70 percent did not know that Congress had recently passed a prescription drug bill, despite the fact that this was arguably the most important domestic legislation passed during the Bush administration, and the most expensive new domestic program in decades. Nonetheless, Surowiecki concludes that it is “more than plausible” that  voters will “pick the candidate who will make the right decision” (id.).

Unlike markets, however, elections often lack the incentives and selection effects that increase the odds of good decisions. Even if a voter makes a tremendous effort to become informed about the candidates, parties, and issues involved, the chance that his or her vote will actually affect the outcome is infinitesimally small - less than 1 in 100 million in a presidential election, for instance. Only if other voters are equally well informed will that knowledge make a difference; but if others are well informed, any given individual still has every reason to remain ignorant because the election will likely come out the “right” way in any event.

Thus, there is very little incentive for individual voters to acquire knowledge and use it effectively, at least so long as their only reason for doing so is to increase the chance of choosing the “right” candidate. For this reason, scholars have long recognized that even highly intelligent voters might reasonably choose to be “rationally ignorant” of politics; quite simply, the knowledge acquired by any one voter will almost surely make no difference. Obviously, some voters acquire political knowledge for other reasons, such as simple interest in public policy, but as surveys show that, most do not. Surowiecki himself partially recognizes this, when he notes that widespread political ignorance on the part of voters “is no sign of their lack of intelligence” but is rather “a sign of their lack of information which itself is an indication of their lack of interest in political details” (id.).

Similarly, unlike markets, elections are not influenced by selection effects.  For understandable reasons, voters who make bad decisions are not eliminated for the rolls prior to the next election nor do they even have their influence diminished relative to those voters who made good ones. Thus, there is very little tendency for the electorate to become more knowledgeable and sophisticated over time, as market participants do. Despite increasing education levels and a quantum leap in the availability of information thanks to cable television news and the internet, studies show political knowledge has stayed at roughly the same low level for at least fifty years. This is in stark contrast with Surowiecki’s account of stock market and corporate decisionmaking, where a great deal of new knowledge has been successfully incorporated into the system over the same time period (chs. 10-11).

Democratic elections also often fail Surowiecki’s own criteria of diversity, independence, and decentralized specialization. The independence criterion is especially likely to be violated because people routinely acquire political biases from those around them rather than attempting to analyze the issues for themselves. For this reason, political biases are usually nonrandomly distributed and therefore often do not “cancel themselves out, rather than reinforcing each other,” as Surowiecki’s criteria require (227). Unlike in economic markets, voters have relatively little incentive to challenge the views of their friends and associates, since casting a better-informed ballot is unlikely to have any impact on electoral outcomes.  Decentralized specialization is also rarely present in political markets because few voters have specialized knowledge of any policy issues; surveys show that the small minority with unusually high knowledge levels tend to be more knowledgeable across the board rather than experts in one particular field of policy.

Democratic electorates do much better on Surowiecki’s diversity criterion. But even here, there is often less diversity than meets the eye.  Relatively few voters have truly unique knowledge about politics, and that uniqueness is not easily translated into disproportionate electoral influence.

The absence of good incentives and selection effects among voters enable mistaken biases caused by lack of independence to persist for much longer than they do in economic markets. For example, economists have known for centuries that free trade is the best possible policy for the nation as a whole; on this point, there is virtually unanimous agreement in the economics profession across the political spectrum. Yet surveys have consistently shown that most voters fail to recognize this and continue to harbor wrongheaded protectionist sentiments, a state of affairs that has persisted for decades. Because the case for free trade is complex and counterintuitive, most voters lack the incentive to acquire the necessary knowledge to understand it. Unfortunately, similar misperceptions abound on a wide range of other issues.

Conclusion

Surowiecki’s book is a highly readable and interesting defense of popular knowledge against the elitism of rule by experts. Unfortunately, his defense of voting is not nearly as effective as his arguments for the effectiveness of crowd knowledge in market settings. Surowiecki sees “rule… by a technocratic elite” (267) as the only alternative to democracy. Certainly, he is right to conclude that democratic government is preferable to “trusting an insulated, unelected elite to make the right decisions” (id.).

But there is another and more appealing alternative to what he calls “democracy’s failings” (267), one that he - perhaps unwittingly - has himself defended. By strictly limiting the powers of government, we can give more scope to decisionmaking by ordinary people in market and civil society settings - precisely the circumstances in which Surowiecki himself has demonstrated the superior wisdom of crowds. Obviously, information problems are not the only factors that must be taken into account in determining the proper role of government in society. Yet Surowiecki’s analysis, combined with that of other writers, suggests that it should have a greater role in the debate than has so far been the case. A smaller and less complicated government might even be easier for rationally ignorant voters to understand, monitor, and control.  Sometimes, the wisest crowd is the one that governs least.