Predicting Mortality Impacts by Underwriting Class, Distribution Channel, and Geography

Mortality prediction is one of the most critical processes a life insurance company performs, as it directly impacts the pricing of its products, driving the company’s profitability. This client project investigates potential mortality impacts through three different lenses: accelerated underwriting, underwriting class assignments (i.e., risk classification), and geography. It addresses the accelerated underwriting objective of achieving at least 30% acceleration using LASSO logistic regression, decision trees, and artificial neural networks to predict if a policy will be in the underwriting class preferred or better, a proxy for mortality rates. The client’s application data from 2019 to 2023, containing demographic information, prescription history, health disclosures, etc., formed the basis for these models. Then, using the same dataset, it addresses the risk classification objective of building new underwriting guidelines based on applicant characteristics using K-Means to evaluate the company’s current underwriting guidelines. Third, it addresses the geographic mortality impact objective of increasing profits by at least 5% using CDC geographic mortality data by determining if different distribution channels exhibit different mortality profiles. Finally, it addresses both ethical and regulatory concerns with respect to the data, accelerated underwriting, and risk classification to determine the feasibility of the first three objectives being implemented.