CreditRisk+: Poisson modelling
About this app
Move from Bernoulli defaults to a Poisson approximation of default counts and an analytical credit loss distribution.
The actuarial approach
CreditRisk+ is an actuarial credit portfolio model. The portfolio is split into bands of similar borrowers. If each of M borrowers defaults independently with probability PD, the number of defaults is binomial. CreditRisk+ replaces this with the Poisson approximation Po(λ), λ = M · PD, which makes the loss distribution analytically tractable. Correlation enters because the default rate itself is random: it fluctuates with a gamma-distributed sector factor that affects all borrowers in the sector together.
What the app shows
The left card compares the exact Bernoulli (binomial) distribution of the default count with its Poisson approximation, marks the largest deviation and shades the region k > M that Poisson allows but reality does not. The right card shows the conditional Poisson distribution for the chosen economic state and the loss distribution with a random rate. “Draw defaults” builds a histogram from random samples. The metrics show λ, P(≥ 1 default) and the tail risk with correlation.
Try it
Raise the PD from 0.5 % towards 50 % and watch the Poisson approximation get worse; for small PDs the two distributions almost coincide. Change the number of borrowers M. Then move the risk factor towards a bad economy: the conditional rate rises and the peak shifts right. With higher sector volatility, the right tail of the loss distribution becomes heavier than pure Poisson.
Frequently asked questions
- What is the CreditRisk+ model?
- CreditRisk+ is a credit portfolio model built on insurance mathematics. It describes the number of defaults with Poisson distributions whose rates vary through gamma-distributed sector factors. This lets the loss distribution be computed analytically, without Monte Carlo simulation.
- When is the Poisson distribution a good approximation to the binomial?
- A Poisson distribution with λ = n · p approximates the binomial well when n is large and p is small. Credit portfolios with low default probabilities usually meet this condition. For large PDs the error becomes visible, partly because Poisson allows more defaults than there are borrowers.
- How does CreditRisk+ capture default correlation?
- In CreditRisk+ the default rate of a sector is not fixed but a random variable. In bad years it is high for all borrowers in that sector at once. This widens the loss distribution and makes its right tail heavier than with independent defaults.
- What can I try in the CreditRisk+ app?
- You set the PD, the number of borrowers in the band, the sector volatility and an economic risk factor. The app compares Bernoulli and Poisson, shows the conditional distribution and the loss distribution with a random rate, and builds a sample histogram at the press of a button.
Subject: Credit risk
More from Credit risk
- Adverse selection
- Default correlation surface in 3D
- CreditPortfolioView: macro factors
- Exposure profile
- Integrated risk measurement
- Rating migration explorer
- PD master scale
- Risk-bearing capacity cockpit
A VisuApp by heyprof: interactive, free in your browser, no sign-up. What is a VisuApp? · All apps