2000 United States presidential election

The 2000 United States presidential election serves as a test case for the Iowa Electronic Markets (IEM), where the price of a contract tied to a candidate’s share of the popular vote translated traders’ collective expectations into a measurable forecast. The book’s central point is that the market’s small, unrepresentative group of traders could predict election outcomes more accurately than polls, illustrating how a crowd’s structure and its method of combining information can matter more than its resemblance to the population being forecast.

The market’s vote-share contracts

One kind of IEM contract predicted the winner; another predicted a candidate’s percentage of the popular vote. In the vote-share market, the eventual percentage determined the contract’s payoff: a George W. Bush contract purchased for the 2000 election would have paid 48 cents when the election was over, because Bush received 48 percent of the vote. The contract price therefore offered a running estimate of the market’s forecast, while the final vote supplied a concrete result against which to assess it.

The comparison was not simply between the market and an individual poll. Across forty-nine elections held between 1988 and 2000, election-eve IEM prices missed the result by an average of 1.37 percentage points in presidential elections, 3.43 points in other U.S. elections, and 2.12 points in foreign elections. The book reports that the IEM generally outperformed major national polls; during the presidential elections from 1988 through 2000, its price was more accurate than the poll released on the same day three-quarters of the time.

A small and unrepresentative crowd

The IEM had never had more than about eight hundred traders. Most were men, and a disproportionate—though declining—share came from Iowa. They were not a miniature version of the national electorate, nor were they forecasting only how they themselves would vote. Yet their market forecasts of the country’s voters proved better than asking voters what they intended to do.

That contrast helps distinguish representativeness from predictive accuracy. The market’s prices changed over time, but were less volatile than polls, which could swing sharply; they tended to shift dramatically only when new information appeared. The book presents that relative stability as one reason the market provided a more reliable forecast, not as a claim that its prices were infallible.

Why aggregation can work

The election example is set within the book’s account of the conditions that can make a crowd wise: diversity of opinion, independent judgment, decentralization that permits specialization and local knowledge, and aggregation through a mechanism that turns separate judgments into a collective result. When sufficiently numerous, diverse, and independent people make estimates, their individual errors may cancel in the average, leaving information that no single participant necessarily possesses in full.

The market’s result is therefore significant not because traders collectively knew the electorate from the inside, but because their distinct judgments could be brought together and measured against the eventual vote. The election forecast illustrates the book’s broader distinction between a crowd’s composition and its capacity to combine information: a small group can produce a useful estimate without mirroring the people whose behavior it predicts.