First published in Business Life Magazine, March 2008
One of my favourite Dilbert cartoons begins with the pointy-haired boss asking for a revenue forecast on a project; it finishes with his subordinate screaming “Just tell me which lie to use!”
That might give a hint as to why internal corporate forecasts are often no good. It’s always hard to foretell the future, but harder still when managers are promoted on the basis of self-serving projections, or because they have successfully sat on bad news.
There is an alternative: something called a “prediction market”. Such markets enable a company’s employees to bet on, for example, whether the new flagship product will beat sales projections in the third quarter. The odds in the market should be a reality check for the official forecasts.
Consider a prediction market “ticket” that pays ten dollars if the flagship does indeed beat expectations in the third quarter, and nothing otherwise. The sales team, knowing that retailers are enthusiastic, might want to buy the ticket. The engineers, knowing the product is full of bugs, might want to sell. If the ticket trades at, say, three dollars, that means that the market’s opinion is that the sales target is only 30 per cent likely to be achieved.
A corporate prediction market allows employees to trade on the basis of their own unique information. They can do so anonymously, without fear of jeopardising their careers. And if they get it right, there is money to be made.
You might think that no company would dare to use a betting market to make forecasts, but the economists Justin Wolfers and Eric Zitzewitz have pointed to Arcelor Mittal, Chrysler, Eli Lilly, General Electric, Microsoft and others as having been publicly identified as experimenting with the idea. The first company to do so seems to have been Hewlett-Packard, during the first technology boom, and their tentative experiment did indeed outperform internal forecasts. The largest and most liquid prediction markets seem to be at Google, which has embraced the idea with gusto. Analysis by Wolfers and Zitzewitz, along with Google researcher Bo Cowgill, suggest that the markets have worked well – albeit with a bias towards overly sunny forecasts.
Google being Google, one trader wrote a software robot to do his trading for him and made money. But the traders who did best were those who had the longest tenure at the company. That might be a crumb of comfort for top managers. They may be usurped by prediction markets, but at least they can make a profit.