A2007
Title: Change-point prediction/detection as alternative to insurer traditional hedging for equity benefit guarantees
Authors: Mark Zanecki - IHA Consultants (United States) [presenting]
Abstract: An AI-enabled daily change point prediction and detection process is introduced, including a downloadable Excel summary model for hands-on analysis. Prospective change point identification is demonstrated for selected USA indexes, ETFs, and stocks, alongside retrospective change point detection with near 100\% accuracy. Theoretical foundations using discrete Fokker-Planck approximation are reviewed. Statistical support is established through t-tests, ANOVA, and linear algebra eigen-analysis via principal component methods. Prospective change points within the subsequent 3 trade days are predicted with 90\% or greater accuracy and confirmed retrospectively with near 100\% accuracy given sufficient run-out time. The Henriksson and Mertons sufficient statistic for investment return value, defined as the sum of condition probabilities, exceeds 1.0 and approaches 2.0. An accurate daily prospective and retrospective change point framework with sufficient accuracy provides on-demand, timely equity benefit guarantee and risk mitigation at significantly lower cost compared to traditional hedging. Time permitting, a synopsis of other applications is discussed.
