Title: In-sample forecasting: Continuous chain ladder and extensions
Authors: Munir Hiabu - Cass Business School (United Kingdom) [presenting]
Abstract: The standard method to forecast outstanding liabilities in non-life insurance is the Chain Ladder Method. Data is hereby aggregated into a so-called loss triangle. The actuary handling the claims can usually choose from a drop-down menu wether the data should be aggregated in years, quarters, or months. We will argue that a) aggregation should be understood as a smoothing technique balancing bias and variance. b) If individual data is available, forecasts can be improved by replacing aggregation with kernel smoothers. c) In the continuous version with kernel smoothers, calendar time effects and operation time effects can be incorporated - effects classical Chain Ladder cannot handle. Asymptotic results, a data example, and a simulation study will be provided.