Alongside these basic approaches to sharing risk emerged the notion of risk assessment, as people realized three things. First, the ‘value’ of goods or persons exposed to dangers was not identical; second, some perils were more dangerous than others; and third, some hazards were more likely to happen than others. Not surprisingly, the evolution of risk assessment closely paralleled the evolution of an insurance sector. Historical examples include the establishment of such insurance organizations as Lloyds in the seventeenth century to insure cargo during sea transportation, or Nicolas Barbon’s insurance company to insure buildings against fi re in the aftermath of the Great Fire of London in 1666 (Hanson 2002). Risk is now generally defi ned as being a function of both the expected losses that can be caused by an event and of the probability of this event. The harsher the loss and the more likely the event, the greater the overall risk. These two factors, loss and the probability of the loss, are typically treated as sequential, independent events so that, in simplifi ed terms, quantifi cation of risk is expressed as:

Risk = (probability of an accident) × (losses per accident)

Along with risk assessment goes the notion of risk management (Louisot and Gaultier-Gaillard 2007; Anderson et al. 2009). Risk research has developed

as society has become increasingly aware of its capacity-or incapacity-to manage risk, particularly when new potential sources of risk are presented. One characteristic of risk management is that risk is always acknowledged and assessed a posteriori. Scientifi c proof can only be put forward when harm has already been done. This is the only way to obtain an estimation of an unobserved quantity on the basis of empirical data. It was only after the London fi re destroyed 13,200 houses and left some 70,000 Londoners homeless that Nicolas Barbon’s idea to insure buildings against fi re could fi rst germinate and then become successful (Wainwright 1953). It was only in 2007, when Hurricane Dean approached oil-production facilities off the coast of Mexico, that the oil market began focusing on the potential for weather-related supply disruptions (Tan and Evans 2007). The paradox is that risk can only be envisaged in the light of past events and yet bears on the probability of the occurrence of future events. This paradox leads to three types of risk management.