Measuring and managing market, credit and Op risk [34]

Go to: Summary | Previous | Next   
Bullet points include: Regulators interested in VaR model performance have studied VaR techniques empirically Risk system vendors do likewise (internally and in external publications) May involve ‘horse races’ between different models on actual or simulated return data Although then exposed to lookback bias Most obvious: in-sample versus out-of-sample But if we know the entire dataset in advance (e.g. it is the past) then can also arise because of how we select our underlying model families

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile