Nassim Taleb and comments about risk managemenet
May 15, 2007
I would like to draw your attention to an interesting short magazine piece by Nassim Taleb and Avital Pilpel. Taleb is a well-known name among derivative traders and risk managers, and more recently, he began to publish texts of a more philosophical nature about risk. This piece relates to Taleb’s more recent intellectual trend and, I would assume, is related to his most recent book.
Some of the concepts discussed in the piece would not be new to ‘sociologists of risk’
(if we can put such a short label on this broad church). However, placing these concepts alongside a discussion about the limitations of the historical VaR concept (a very brief one, as the piece is very short) and with a reference to the self-referential nature of statistical distributions creates a nice opening for a sociological discussion about contemporary risk management. A couple of points:
First, in addition to the critical analysis regarding the nature of statistical distributions as a central tool in risk management, it is worth to note that such distributions do not take into account the cause-and-effect nature of risk management. That is, once an organization has encountered a risk event, it is very likely that it will change practices in such a way that will alter the probability of similar event happening again.
The second point is a development of the first one and refers more specifically to the constitutive power of risk events. Taleb and Pilpel suggest that the less frequent an event is, the more severe it will be. They imply that there is something about the nature of risk events that causes and maintains this inverse relation between frequency and severity:
[T]he severity of the event, will be in almost all cases inversely
proportional to its frequency: the ten-year flood will be more
frequent than the 100 year flood – and the 100 year flood will be more
I would like to suggest that instead of referring to some implicit external causality that would explain this inverse relation, it could be a good idea to look at risks through the social dynamics in which they unfold. In particular, maybe the severity of risks should be measured in relation to the organizational environment in which they occur. Events of various degrees of severity take place at various frequencies. However, when events happen frequently enough, organizations become ‘used to them’. That is, institutional reactions to risk events become more routine-based, procedural even, and as a result, a lower degree if severity is assigned to the events. This is not an ‘optical illusion’ – the more prepared an organization is to a specific risk event, the less impact that event would have on the organization. Hence, rare event are more severe not because of some mysterious ‘long-term’ cycle that makes them more devastating but because they are rare and organizations have less opportunities to prepare for them.