Mon 27 Apr 2026 - The loss amount for each insurance business line is a critical metric for risk analysis and fair pricing. Standard methods often model losses using annual data, assuming expense and distribution stability. This approach, however, can fail in high-inflation environments where insurer costs and loss dynamics shift significantly within a policy period.
This webinar introduces a time-dependent loss distribution analysis to address such volatility. Instead of an annual view, we index data (for example, by month) and fit a probability distribution at each interval, creating a time series of distributions. We then apply singular spectrum analysis (SSA), a non-parametric time series method, to this series. SSA extracts key patterns, models the evolution of the loss distribution, and enables forecasting of its future state.
This framework explicitly models the uncertainty in loss distributions by repeating the modelling process across sub-periods. It not only captures underlying trends and seasonal changes but also provides a forward-looking view, enhancing pricing and reserve setting in volatile economic climates.
Rahim Mahmoudvand, Bu-Ali Sina University/ University of Cagliari