A loss distribution is a probability distribution describing how likely each possible level of loss on a portfolio is over a given period, usually one year. In credit risk it is built from each exposure's probability of default, loss given default and exposure at default, together with assumptions about how defaults are correlated. Insurers build similar distributions from claim frequency and severity. Because large losses are rare but possible, credit loss distributions are typically skewed, with a long right-hand tail.
Two measures are read from it. The mean is the expected loss, which for a single loan is : a 1,000,000 exposure with a 2% default probability and 45% loss given default has an expected loss of 9,000. The unexpected loss is the gap between a high percentile, such as 99.9%, and the mean; banks hold economic capital against it, while expected loss is covered by pricing and provisions. Distributions are often estimated by Monte Carlo simulation.
Ref 1020000118