The Wisdom of Crowds Explained

The wisdom of crowds is the finding that a large group of independent guesses, averaged together, is usually more accurate than almost any single person in the group, including the experts.

It sounds like it should not work. Most people in a crowd are not experts, plenty are simply wrong, and any one guess can be wildly off. Yet when you collect enough independent estimates and combine them, the errors tend to cancel out and what remains is a signal that tracks the truth remarkably well. Understanding why is the key to understanding a lot of modern forecasting, from polls to prediction markets.

The ox that started it all

The most famous demonstration is also the simplest. At a country fair, ordinary fairgoers were invited to guess the weight of an ox, each writing down a number on a ticket for a small prize. No one conferred, and most had no special expertise in livestock. On its own, any single ticket was close to worthless as a forecast.

In 1907 the statistician Galton gathered those tickets and studied them, publishing the result in his paper “Vox Populi” in Nature. Across 787 valid entries, the median guess was 1,207 lb against an actual dressed weight of 1,198 lb, an error of under 1%. The crowd, taken as a whole, was almost exactly right.

That is the whole idea in miniature. No single guesser was reliably accurate, and Galton had rather expected the crowd to look foolish. Instead, the middle of all those independent guesses landed within a fraction of a percent of the truth. The aggregate knew something that no individual reliably did.

Why the aggregate beats the individual

The mechanism is not magic, it is arithmetic. Think of each person’s guess as the truth plus an error. Some people guess too high, some too low, and the size of their mistakes varies. When you average many independent guesses, the positive and negative errors tend to offset each other. What survives the averaging is the part they have in common, which is the underlying truth they are all circling.

The catch is the word independent. For the errors to cancel, they have to be scattered in different directions rather than all leaning the same way. When guesses are genuinely independent and drawn from varied perspectives, adding more people keeps sharpening the aggregate. When everyone makes the same mistake for the same reason, averaging just preserves that shared mistake, and the crowd is no wiser than one biased individual.

The idea was popularized decades later by the journalist Surowiecki, whose 2004 book The Wisdom of Crowds gathered dozens of cases, from guessing jellybeans in a jar to locating a lost submarine, where an aggregated judgment outperformed the individual experts inside it.

The three conditions for a wise crowd

A crowd is not automatically wise. Put people in the wrong conditions and the group can be dumber than any of its members, prone to panics, bubbles, and herd behavior. The wisdom shows up only when three conditions hold together.

Diversity of opinion

Each person should bring something different to the table, some private information, a different model of the world, or a distinct interpretation of the same facts. Diversity is what makes the errors point in different directions in the first place. A room full of identical thinkers, however smart, only ever produces one guess repeated many times.

Independence

People should form their estimates on their own, before being swayed by what everyone else thinks. The moment guesses start copying one another, independence collapses and errors stop cancelling. This is why showing the running average as people vote, or letting the loudest voice go first, quietly destroys the wisdom you were trying to capture.

A way to aggregate

Finally, the scattered private judgments need to be combined into a single collective answer. A simple average or median can do it, as Galton’s ox showed. A vote can do it. And a market can do it, letting each person express both their view and how strongly they hold it through the size of their bet.

The mathematics behind this is old. As far back as 1785, Condorcet’s jury theorem showed that if each voter is more likely than not to be right and votes independently, the probability that the majority is right climbs toward certainty as the group grows. It is the wisdom of crowds stated as a theorem.

How prediction markets operationalize it

A prediction market is arguably the most refined machine for harvesting crowd wisdom. Instead of asking everyone for a number and averaging, it lets people buy and sell shares in an outcome. The price at which those shares trade becomes the crowd’s aggregate estimate of how likely the outcome is, updating live as new people trade.

Crucially, a market improves on a plain average in two ways. It weights people by conviction, because a confident trader stakes more and moves the price more than a hesitant one. And it rewards being right with money, which pulls in the informed, filters out lazy guesses, and gives everyone a reason to correct a price they believe is wrong.

If the mechanic itself is new to you, start with prediction markets explained, then come back for the crowd-wisdom lens on why the price is trustworthy.

This is exactly why deep, liquid markets forecast so well: they satisfy all three conditions at once. Traders are diverse, they act on their own private reasons, and the price aggregates every trade into one continuously updated probability.

For the full accuracy argument and the evidence behind it, see why prediction markets are accurate.

When crowds are not wise

The same conditions that make a crowd wise, when broken, make it foolish. It is worth knowing the failure modes so you can spot a crowd you should not trust.

• Herding: when people copy each other instead of thinking independently, errors stop cancelling and the group can stampede in one direction.
• Information cascades: if early, visible guesses anchor everyone who follows, the crowd amplifies a single early mistake instead of correcting it.
• Shared bias: when everyone leans the same way for the same reason, averaging preserves the bias rather than removing it.
• Too little diversity: a crowd of near-identical experts has little error to cancel, so adding more of them barely helps.
• No real aggregation: a shouting match or a show of hands swayed by the loudest voice is not the same as a clean average of independent views.

Many of these traps are individual mental shortcuts scaled up to a group. To see the underlying errors, read cognitive biases in forecasting.

See it for yourself

The wisdom of crowds is easiest to believe once you have watched a live price move toward the truth as informed people trade against it. You do not need real money to feel the effect, only a crowd and a clear question.

Clutch lets you predict real news and sports with in-app credits and shows you the crowd’s price on every question, so you can watch collective wisdom form in real time. Get the app and put a crowd to the test.