A claims frequency distribution is the probability distribution of the number of claims an insurance policy, risk group or portfolio produces over a period, usually a year. Combined with a claims severity distribution for the size of each claim, it forms the frequency-severity model at the heart of non-life pricing and capital modelling.
The Poisson distribution is the standard starting point. It has a single parameter , the expected number of claims, and its variance equals its mean; with , the probability of no claim in a year is . Real portfolios usually show more variation than the Poisson allows, because risks differ and shared conditions such as weather affect many policies at once, so the negative binomial distribution, whose variance exceeds its mean, is widely used. The binomial distribution suits cases with a fixed maximum number of claims. Frequency models are often fitted with generalised linear models to support pricing segmentation and are simulated in Monte Carlo capital models.