Quantifying Risk in Complex Systems

Quantifying risk is a big topic in a number of fields. Whether it be the risk of terrorism, a stock market crash, or, as in my work, animal population extinction, we often want to know something about the probability, magnitude, and, ideally, the timing of negative outcomes. This is not a straightforward process in many social and biological systems. I’d be interested in hearing how people from other disciplines approach risk analysis. If anyone else is interested in this topic, let me know (and/or put your name below).

James james.battin (at) noaa.gov

Hi James,

I do some work in this area. Risk management is an enormous field. A basic tool in the policy world proposes the following metric:

• Risk = Threat * Vulnerability * Consequence
  • Threat is the frequency of a given adverse event.
  • Consequence is the maximum damage to the system you're studying from the event.
  • Vulnerability is the degree to which damage can be limited by preventive measures.

Theoretically, you can calculate the value of this in any terms you choose, including dollar terms. You might end up with some ugly integrals if you try to do this exhaustively, particularly if you couldn't be sure of the distribution of incident severity over time. (Think how different these would look for Gaussians, exponentials, power laws, and uniform probability distributions.)

Lots of other risk metrics exist. I only mention this one because it's commonly used in government, and because you can use it to evaluate the cost-effectiveness of particular policy interventions--not because it is precisely correct for a specific class of problems.

Benmazzotta 23:06, 20 June 2007 (MDT)