"The Investment Forecast," by Reiner Schnur. Nature, 31 January 2002.
There are currently many models of the world's climate, but each has various uncertainties. To minimize the effects of these uncertainties, the Intergovernmental Panel on Climate Change uses a consensus approach, in which they average projections from several models. This has proven to be effective in predicting, say, worldwide average temperature. However, a new study shows that the consensus method chronically underestimates the probablility of extreme weather like flooding. The averaging tends to smooth out the predictions of each model, and thus may well yield a zero probability for flooding even when each component model predicts at least one flood. The study's authors argue that it's better to average the probability of extreme weather given by each model.
The curious title of the article seems to come from the fact that this study could be used by investors interested in the probability of extreme market fluctuations. However, nearly the entire article deals exclusively with climate change. There is also another study mentioned, which shows that flood frequency has increased during the 20th century, with a very strong probability that the increase is a result of human activity.
--- Rafe Jones