Traditional incurred but not reported (IBNR) methods (the Loss Ratio, Chain-Ladder, and Bornheutter-Ferguson methods) rely to varying degrees on the assumption that the past provides a reasonable expectation of the future. However, the pandemic has caused an unprecedented change to processes across the insurance space. Some impacts, such as an increase in claim incidence, are easy to grasp, whereas understanding how a change in reporting speeds might impact a given IBNR method is less obvious.
Take a hypothetical look at business under various stress scenarios and discover the strengths and weaknesses of common IBNR reserving techniques under each scenario. Learn the extent to which predictive analytics can provide insight into the adequacy and accuracy of IBNR reserves. Examine both mortality and morbidity claims and explore various machine learning approaches, including polynomial regression and ensemble techniques such as bagging and boosting. Assess random forest and other boosting models and evaluate their performance in fitting and predicting stop-loss case level reserves.
By attending the session, you will be able to:
• Evaluate the impact of external changes on common actuarial techniques.
• Understand the appropriateness of assumptions underlying IBNR estimation approaches in different environments.
• Integrate advanced analytics to supplement traditional methods.
• Derive the proper predictors to construct an optimal predictive model.
• Evaluate performance of various types of predictive models.