CreditRisk+ and Actuarial Credit Models

Most portfolio credit models (like Merton-style structural models) try to explain why a firm defaults — its asset value falling below its debt. CreditRisk+, developed at Credit Suisse in 1997, sidesteps that entirely and borrows a much older idea from insurance: treat each obligor's default as a random event with a small, known probability, and model the number of defaults in the portfolio the way an actuary models the number of insurance claims in a year, using a Poisson-type distribution.

This actuarial framing has a real practical payoff: because it relies on a small number of inputs (default probabilities, exposures, and a way of correlating obligors through shared background factors like sector or region), it produces a portfolio loss distribution and credit-VaR figure through closed-form or near-closed-form mathematics, rather than requiring the heavy Monte Carlo simulation that structural models typically need.

Worked example. A portfolio of 500 loans each has a 1% annual default probability and is roughly independent. Treating defaults as Poisson-distributed with mean $\lambda = 500 \times 0.01 = 5$, the model gives a full distribution over how many defaults occur in a year — a roughly 84% chance of 7 or fewer defaults, and a tail probability of more than 12 defaults of under 2%. That distribution, combined with loss-given-default and exposure sizes, feeds directly into a portfolio loss curve without simulating each loan's individual asset path.

The trade-off is that CreditRisk+ treats default probabilities as fixed rather than driven by an evolving asset value, so it is faster to compute but less able to explain how correlated defaults arise from a shared economic shock — a gap that risk teams typically patch by layering sector-level correlation factors on top of the base Poisson structure.