Date of Completion

8-5-2016

Embargo Period

2-1-2017

Keywords

Long-Term Care, Actuarial Assumptions, Generalized Linear Model, Incidence Rates

Major Advisor

Professor Jeyaraj Vadiveloo

Associate Advisor

Professor Emiliano Valdez

Associate Advisor

Professor Guojun Gan

Field of Study

Mathematics

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

Generalized Linear Model Approach to Adjusting Expected Assumptions of Long-Term Care Incidence Rates

Rozita Ramli, PhD

University of Connecticut, 2016

Actuarial assumptions are needed in most of actuarial works, for example pricing and reserving and setting capital standards. In long-term care (LTC) insurance, three main actuarial assumptions are incidence, termination and utilization rates. These assumptions can be seen as factors affecting the financial progress in LTC insurance business. Insurance companies must make sure the assumptions they use adequately reflect actual experience to avoid adverse financial consequences. Hence, a regular systematic methodology to monitor expected assumptions is essential. This work proposes a methodology to adjusting expected assumptions of LTC incidence rates. The methodology uses a Generalized Linear Model (GLM) approach in modeling incidence rates and the adjustment is parallel to an established industry technique of stochastic deferred acquisition cost (DAC) unlocking. The methodology uses confidence intervals as a decision tool to decide if any adjustment is needed. We also bring in a credibility component into the methodology to provide rationalization for choosing the appropriate significant level of the confidence interval when applying the methodology. Lastly we study the effect of size of historical experience used in the adjustment on the efficiency of the methodology. Similar to how the stochastic DAC unlocking technique is an acceptable technique by regulators and practitioners, we foresee our GLM methodology to be accepted as a basis for adjusting assumptions under Principle-Based Reserving in the future.

COinS