Andersen–Sidenius–Basu loss recursion
Builds a conditional portfolio loss distribution for credit tranches.
How it works
Conditional on the common factor, compute the loss distribution by adding obligors one at a time. Each new obligor leaves existing mass at the same loss with its survival probability and shifts mass upward by its default loss with its default probability. Integrating over the systematic factor restores the portfolio dependence.
Worked example
For two unit-loss loans with conditional default probabilities 0.1 and 0.2, the distribution at that factor value assigns 0.9 × 0.8 = 0.72 to no loss, 0.1 × 0.8 + 0.9 × 0.2 = 0.26 to a loss of one and 0.02 to a loss of two. A tranche attached at one counts only the last state.
Validation and limitations
Check that probability mass is preserved at each step and compare the first two moments with direct calculations. For heterogeneous loans, test discretisation and the integration over factor states near the tranche tail. Change the grid and compare with independently weighted simulation where possible.