On the Bayesian Risk Evaluation of Minimum Guarantees in Variable Annuities

Long-term equity return models have a substantial impact on calculating reserves and capital requirements for minimum guarantees in variable annuities (VAs). Under a Bayesian statistical framework, we reinvestigate the standard long-term equity return models (independent lognormal, regime-switching lognormal and stochastic log volatility models) with the Korean equity market data. Our empirical analysis shows that the long-term behaviors of the Korean market are best described by the regime-switching lognormal models, and there can be significant difference in the amount of the reserves and the capital requirements depending on the model selection. The long-term nature of the VA contracts requires that more caution be paid for modeling the mean part of the stochastic log volatility model.