Economics-based prediction

Research opportunities

Accurate forecasting is a problem that frustrates the corporate world. In particular, it is difficult to extract accurate information about future events (such as predicted quarterly revenues or expected product demand) from knowledgeable workers and managers.

In the wider world, when large numbers of people invest in company stocks, for example, prices reflect everything currently known about how those companies might perform in the future.

But when small teams are involved, markets don’t work. Even though these groups are often knowledgeable, predictions can be influenced by personal biases and relationships, or by team members trying to play the market to their own advantage.

Research focus

Current work

Our research shows that a pool of about nine people can consistently make better predictions than even its most perceptive individual member.

But how do you tap that knowledge while filtering out the biases? Research in experimental economics at HP Labs suggests you can do it in three ways:

• make the prediction process anonymous
• establish the risk attitudes of the participants
• ask people to back up their predictions with real money

Our most advanced project in this area is called BRAIN (Behaviorally Robust Aggregation of Information in Networks).

When using BRAIN to make better predictions, teams play a game that reveals both their appetite for risk and their predictive power. Because predictions are anonymous and players stand a chance to make some real money if they predict well, members of the group tend to offer their very best guesses. By factoring in the players' known appetite for risk and known skill at predicting, researchers running BRAIN can then accurately compile these predictions into a prediction that has the form of a probability distribution for a number of future events.

Because the whole exercise is a game, takes little time, and is run anonymously, teams don’t mind using the tool regularly. What’s more, it’s proving to be highly accurate.

In early trials BRAIN has proven to better predict future outcomes – expected monthly revenue, for example, or future product demand -- than traditional methods.

HP Labs is exploring other tools for future prediction, including a design for ‘decision insurance’ that lets risk-averse business managers be risk-neutral when deciding the future directions they want to pursue.

Technical contributions

This work is being applied outside the lab as part of a broader effort by HP to offer businesses the chance to better predict the future.

Related research

Learn more