Know these first
- L3.7.07: Measuring frequency and severity
Math Atoms
By the end you will be able to
- 1Explain the Poisson approximation to defaults and say when it is safe.
- 2Build the CreditRisk+ loss distribution: gamma-mixed intensities by sector, exposure bands and Panjer's recursion, checked by FFT.
- 3Calibrate sector variances and compare the tail with the Gaussian one-factor model on the same book.
- 4Explain how independent sectors and the gamma's shape change the whole-book tail.
Why it matters
An insurer does not ask why a policyholder claims; it counts claims and sizes them. In 1997 Credit Suisse First Boston published CreditRisk+, which treats loan defaults the same way. No simulation, no asset values: one formula and a recursion give the whole loss distribution in milliseconds. On the bank's corporates it lands, to the nearest £10k, where L6.4.01: Portfolio loss distributions's exact model does at 99.9%, £56.54m. On the whole book it says £112.27m where the Gaussian model says £135.24m. This lesson shows why both are right about their own assumptions.