The Wisdom of Crowds

The Wisdom of Crowds names the surprising capacity of a properly organized group to reach judgments that are often more accurate than those of its individual members, including its smartest ones. The principle does not depend on a crowd being made up of experts. It depends on preserving different pieces of information and perspectives, allowing people to judge independently, drawing on local knowledge, and combining individual estimates into a collective result. Under those conditions, ordinary, imperfect judgments can add up to an unusually good picture of the world.

From individual limits to collective intelligence

The idea begins from a modest view of human decision makers. People have less information than they would like, limited foresight, and little ability or appetite for elaborate cost-benefit calculations. They often settle for what seems good enough, and emotion can affect their conclusions. Herbert Simon described this condition as bounded rationality. The book’s claim is not that individuals cease to be fallible, but that their errors can be offset when judgments are aggregated in the right way. A group can therefore make a good decision without being governed by exceptional intelligence or composed mainly of exceptionally informed people.

This collective capacity appears across settings that at first seem unrelated. Search engines such as Google draw on the organization of a vast body of webpages to find a relevant result; betting markets make it difficult for even informed gamblers to beat the collective forecast; and the Iowa Electronic Markets have predicted election outcomes more accurately than major polls in many comparisons. The same general idea helps explain how markets coordinate the supply of everyday goods, why people pay taxes and volunteer to coach Little League, and why collective judgment matters to science and business. The examples differ in their mechanisms, but each raises the same question: how can dispersed knowledge become a usable answer?

The phrase also challenges the instinct to search for a single authoritative mind. People commonly assume that valuable knowledge is concentrated in a few heads and attribute a crowd’s success to its hidden experts rather than to the group’s combined judgment. Jack B. Soll and Richard Larrick call this impulse chasing the expert. The book treats it as a costly mistake: a crowd includes its geniuses, but it can also contribute information and perspectives that no one expert possesses. Its success is easy to overlook precisely because the visible result—a price, a forecast, or an estimate—can appear to have come from nowhere.

The conditions that make a crowd wise

The book identifies four conditions behind a wise crowd: diversity of opinion, independence, decentralization, and aggregation. Diversity means that participants bring some private information, even if it takes the form of an unusual interpretation of known facts. Independence means that their views are not simply determined by what others around them think. Decentralization gives people room to specialize and use local knowledge. Aggregation provides a mechanism for turning individual judgments into a collective decision. Together, these conditions allow a group to pool what its members know without allowing one voice or a shared error to dictate the answer.

The logic can be expressed as a distinction between information and error. Each estimate contains both. If a sufficiently large group of diverse, independent people makes a prediction or estimates a probability, averaging their answers can cancel individual errors while preserving information. But cancellation is not magic: if the group has no relevant information to begin with, the average will not become wise merely because many people contributed. The group’s collective verdict can be powerful because it often contains far more information than observers expect, not because any crowd is automatically right.

The difference between averaging physical performance and averaging judgment makes the claim vivid. Average the times of a hundred people running a 100-meter race, and the result will be slower than the fastest runners’ times. Average their answers to a question or their attempts to solve a problem, however, and the group’s answer may equal or outperform that of its smartest member. For many physical tasks, the average means mediocrity; in decision making, it can mean excellence. The contrast captures the book’s central surprise: aggregation can improve judgment in a way it cannot improve every kind of performance.

Markets as mechanisms for combining judgments

Decision markets offer one practical method for aggregation, but they are not the only one. Stock prices, votes, point spreads, pari-mutuel odds, computer algorithms, and futures contracts all turn multiple judgments into a shared signal. A market price can be understood as a collective estimate: participants act on different information, and the price changes as their decisions meet. Its usefulness depends less on the special superiority of any one market design than on whether the process preserves diversity, independence, and decentralization and then aggregates judgments effectively.

The Iowa Electronic Markets provide a detailed example. A study of forty-nine elections from 1988 to 2000 found that election-eve prices missed vote shares, on average, by 1.37 percentage points in presidential elections, 3.43 points in other United States elections, and 2.12 points in foreign elections. The market generally outperformed major national polls and was more accurate even months before an election. Across 596 polls released during presidential elections between 1988 and 2000, the market price on the day each poll appeared was more accurate three-quarters of the time. Unlike the polls, its forecasts were comparatively steady, changing sharply mainly in response to new information.

The market was small—never more than roughly eight hundred traders—and did not resemble the electorate. Most traders were men, and a disproportionate, though shrinking, share came from Iowa. They were not simply predicting their own votes; nevertheless, their market forecast of what voters across the country would do was better than asking voters directly what they intended to do. The example shows how a group can forecast behavior beyond its own membership, while also making clear that the relevant test is performance against outcomes, not demographic resemblance alone.

Sports betting supplies another test because forecasts can be compared with game results. At the Mirage, bookmaker Robert Walker adjusted the point spread in response to betting. Before the 2001 Super Bowl, the Baltimore Ravens opened as 2.5-point favorites; after a couple of early $3,000 bets on Baltimore, Walker moved the spread to three points. If bettors overwhelmingly favored one side, the line might be wrong, so it shifted. The bookmaker set the opening number, but once betting began, the spread increasingly reflected the bettors’ combined judgment. The final line was not a crystal ball: it only weakly predicted the final scores of most NFL games. Yet favorites covered the spread about as often as underdogs did, and no simple, persistent market mistake stood out. Roughly three-quarters of the time, the Mirage’s final line was the most reliable forecast available for NFL outcomes. In other sports, too, market odds generally captured available information; horse-racing odds, for example, reliably ranked horses by their chances and offered reasonably good estimates of winning probability.

These examples do not make every market or prediction problem suitable for decision markets. Some markets may fail to attract enough participants to produce an intelligent forecast; others may try to predict outcomes that cannot be predicted. The case for the method is conditional: where the problem is appropriate and the group meets the necessary conditions, a market can provide a quick way to transform diverse opinions into a single judgment. The broader lesson is that the mechanism matters, but the social and informational conditions it creates matter more.

Diversity in groups and organizations

In large markets, diversity often arises because people enter with different information, risk tolerances, time horizons, and styles of investment. Small teams and formal organizations cannot assume the same variety will appear by itself. They must actively select for cognitive diversity, because a few biased individuals can exert disproportionate influence in a small group. Here, diversity is not simply a virtue of representation; it is an input to problem solving, adding approaches and skills that would otherwise be missing.

Political scientist Scott Page demonstrated the point with computer-simulated agents. He formed groups of ten or twenty agents, each equipped with a different set of skills, and asked them to solve a relatively sophisticated problem. Some agents were highly effective and others less so. Yet groups containing both stronger and weaker agents almost always outperformed groups made up only of the strongest. Choosing a group randomly and letting it work could do as well as, or better than, spending time identifying the best agents and placing them together.

Page’s results do not imply that intelligence is irrelevant or that ignorance helps. The agents were not ignorant, and successful groups still included high performers. The finding is that intelligence alone does not guarantee a range of perspectives. If people selected as smart tend to have similar abilities, a group of them may possess a narrower collective toolbox than a more varied team. Adding people with different skills can improve the whole group even when those people are less effective on their own. The value lies in what the members know in combination, not in ranking them one by one.

Scope and limits of the principle

The book presents the principle both as a description of how the world already works and as a possibility for institutions that have yet to make full use of it. Collective judgment can be easy to miss, and even when its results are visible people may credit a few exceptional individuals instead. Voters, investors, consumers, and managers often look for the one person who has the answer. That habit helps explain why companies have shown little interest in using decision markets or tapping the collective knowledge of employees, despite the close fit between market forecasts and the uncertain choices corporate strategy must make.

Nor does aggregation guarantee truth. A crowd needs information in its estimates; without it, cancellation of errors cannot rescue the result. It also needs the conditions that prevent shared opinion from overwhelming independent judgment and that let local knowledge enter the process. The book’s argument is therefore not a blanket endorsement of majority opinion or of markets in every circumstance. It is a claim about what groups can accomplish when their judgments are diverse, independent, decentralized, and effectively combined—and a warning that an institution can possess collective intelligence without knowing how to use it.

That distinction gives the concept its practical force. No single technique, whether a futures market, a search engine, or a betting pool, is inherently wise. Each is a way of attempting to collect what many people know. The hard work is arranging the process so that knowledge survives the act of combination. When those conditions hold, the average can be more than a compromise among individual views: it can be a remarkably accurate judgment that no individual had to possess in full.