Interesting article by Alan Greenspan in the FT today (http://www.ft.com/cms/s/0/edbdbcf6-f360-11dc-b6bc-0000779fd2ac.html). One of a number of recent articles pointing at weaknesses in the modelling of risk re: the recent credit crisis – in particular taking putting forward that the human sentiment (e.g. greed or fear of loosing market share in bull market) may well be irrational, but it is systemic and observable and should be considered as a "factor" within risk and economic models.
Greenspan is critical of statistical modelling for risk, saying that the seemingly stable correlation relationships between different assets collapse under times of extreme market stress. Seems like he and a number of other commentators are talking about Paul Wilmott’s crashmetrics ideas on correlations – see http://paul.wilmott.com/crashmetrics.cfm for some background.