Stress testing incorporating causal dependency [6]

Go to: Summary | Previous | Next   
Bullet points include: Frequentist approaches, when taken to the limit, are not obviously appropriate for forward-looking risk measurement Because we always need to assume something about time stationarity Because they assume all information of relevance is ‘in the data’ But are implicit in many common implementations of VaR and/or extrapolation into the tails of distributions How might we incorporate to our advantage subjective views (aka ‘expert opinion’) in our assessments?

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile