Image via WikipediaI was going through my notebook, and I just cannot find the source for one of my notes.
I am pretty sure it is from The Economist, Feb 13, 2010, special report on financial risk, but I just cannot pin it down.
The idea was that to use probabilistic tools not just to work out expected outcomes and their distributions, but also to take the Unthinkable Horrific eventualities, and work backwards to puzzle out the most likely ways that they could come into being. And then use that to mitigate.
So, if you were considering building a nuclear power plant, you would not simply work forward to calculate expected outcomes and their distributions, but you would assemble a list of Unthinkable Horrific eventualities that relate to nuclear power plant. You would work backwards, to assemble a list of the ways those horrific eventualities could take place, along with a likelihood (or distribution over variables). Then you would have the basis for decisions about mitigation under economic constraints.
This cannot be done without complete transparency. Everyone needs access to your published assumptions, because you need to be called out when your view is too narrow. Ego and procedural niceties be damned.
Random notes: trees of scenarios, working backward under assumption of failure mode, black swans, alternatives to VAR (value-at-risk): conditional VAR, mean excess loss, mean shortfall, tail VAR.