Hewlett-Packard printer-sales prediction-market experiment

Hewlett-Packard’s printer-sales prediction-market experiment tested whether employees drawn from different parts of the company could forecast printer sales more accurately by trading on their judgments than the company could through its usual forecasting process. Run in the late 1990s, the small, short-lived markets outperformed Hewlett-Packard’s forecasts 75 percent of the time over three years, offering a practical example of collective judgment at work inside a corporation.

Design and Participation

Economists Charles Plott and Kay-Yut Chen established artificial markets in which Hewlett-Packard employees bought and sold shares according to what they expected printer sales to be in the coming month or quarter. Participants came from different parts of the company, a choice intended to bring varied information and perspectives into the market rather than confining the forecast to a single department or specialist group.

The experiment was modest in scale and duration. Each market included between twenty and thirty employees and ran for only a week. Trading took place during lunch and in the evening. Its significance did not depend on a large trading population or a long-running exchange: the markets converted the participants’ separate expectations into a collective forecast that could be compared with Hewlett-Packard’s own predictions.

Results and Organizational Significance

Over the three years of the experiment, the market’s forecasts outperformed the company’s forecasts 75 percent of the time. The result suggested that employees distributed across a company could collectively know more about near-term sales than the organization’s established forecasting process captured. Their different positions within Hewlett-Packard supplied the diversity that the market design sought to harness; trading provided a mechanism for aggregating their judgments.

The experiment served as one of the book’s concrete examples of how decision markets might help companies make forecasts despite their relative reluctance to adopt internal markets. It demonstrated a specific, limited use: aggregating employee expectations about printer sales, not guaranteeing better corporate decisions generally. Its force lies in the contrast between a small group trading briefly and a company’s formal forecast—and in the market’s repeated advantage across the three-year trial.