Definition:Frequency-severity analysis
📊 Frequency-severity analysis is a foundational actuarial technique used in insurance to decompose loss experience into two distinct components: how often claims occur (frequency) and how large those claims are when they do occur (severity). By separating these two dimensions, actuaries and underwriters can build more granular models of expected losses, identify emerging trends in either component independently, and price policies with greater precision. The method is central to virtually every line of business, from motor insurance and workers' compensation to property catastrophe modeling and health cost projections.
⚙️ In practice, an insurer collects historical claims data and organizes it into frequency distributions (number of claims per exposure unit over a defined period) and severity distributions (the dollar or currency amount of each individual claim). These distributions are then analyzed separately — often fitted to statistical models such as Poisson or negative binomial distributions for frequency, and lognormal or Pareto distributions for severity. The product of expected frequency and expected severity yields the pure premium, which forms the starting point for ratemaking. Regulatory frameworks across jurisdictions expect this kind of rigor: Solvency II in Europe, the RBC framework in the United States, and C-ROSS in China all require insurers to demonstrate that their technical provisions and capital charges rest on defensible loss modeling, and frequency-severity decomposition is often the bedrock of that demonstration.
💡 The real power of the technique lies in its diagnostic value. A rising loss ratio could stem from more claims, costlier claims, or both — and the strategic response differs dramatically depending on which driver is at work. If frequency is climbing in a commercial auto portfolio, the insurer might tighten risk selection criteria or adjust deductibles; if severity is spiking due to social inflation or higher medical costs, the response might involve revising policy limits or pursuing reinsurance protection for large losses. Insurtech firms have further enhanced frequency-severity analysis by integrating telematics, IoT sensor data, and machine learning to update frequency and severity estimates in near real time, moving the discipline from retrospective analysis toward predictive and even prescriptive pricing.
Related concepts: