Not quite any shape. De Servigny notes two weakenesses: can't be bi-modal (like a two humped camel) and doesn't handle zero (no recovery) or 1.0 (fully recovery) well. His point is the beta is "flexible" unlike most of the other distributions we look at. Many of the credit portfolio models (e.g., CreditMetrics) use beta distribution owing to its flexibility.
To my knowledge, however, unlike the GEV (mentioned in Wilmott's EVT) which is mathematically derived (it "converges" based on the EVT assumptions), there is nothing "magic" about beta against recoveries. Rather, it is favored for two reasons:
1. flexible (i.e., to fit various scenarios)
2. compact; only need bounds and two params
I am frankly not expert on beta distribution - it is related to binomial/gamma, and I *think* with certain params it may converge it to the others but i cannot recall exactly...
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.