Creating and validating risk models [29]

Go to: Summary | Previous | Next   
Bullet points include: Sobehart, Keenan and Stein article (in reading list) notes that default risk models can err in two different ways, and reducing one type comes at the expense of increasing the other (c.f. ‘power’ of statistical tests, false positives vs. false negatives): Indicate low risk when risk is in fact high (Type 1 error) Indicate high risk when risk is in fact low (Type 2 error) Model Actual Low credit qualityHigh credit quality Low credit quality Correct assessment Opportunity costs. Lost potential profits. Lost interest income and origination fees. Premature selling High credit quality Lost interest and principal through defaults. Recovery costs. Loss in market value Correct assessment

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile