Policy makers, the press, and others are increasingly turning to forecasting to predict things like gas and housing prices. And when choosing which forecasts to trust, it would seem perfectly logical to turn to individuals who have been accurate in the past. But based on research I helped conduct, this turns out not to be the case.
Instead, we find that forecasts are more accurate when they are drawn from large and diverse groups, even if some individual members of that group have been wrong before.
That’s because the larger the group, the greater the inputs as well as the likelihood that wrong pieces of information will cancel each other out.
This law of large numbers logic doesn’t hold when forecasts are drawn from small groups, making such groups less reliable, even when their individual members have been correct in the past.
But even if your forecasts are drawn from larger and diverse groups, caution is still needed. While technology enhances the ability to gather the wisdom of the crowd, it also creates opportunities for manipulation. Our research paper, which will appear in a forthcoming issue of Management Science, discusses what to consider when assessing forecasts and which features are likely to make them more accurate.
But here’s one safe prediction: The use of forecasting will continue to grow.
P.J. Lamberson is a senior lecturer in Systems Dynamics at MIT Sloan.