Stochastic models for asset and liability modelling in South Africa or elsewhere

Research in the area of stochastic models for actuarial use in South Africa is limited torelatively few publications. Until recently, there has been little focus on actuarialstochastic models that describe the empirical stochastic behaviour of South Africanfinancial and economic variables. A notable exception is Thomson’s (1996) proposedmethodology and model. This thesis presents a collection of five papers that werepresented at conferences or submitted for peer review in the South African ActuarialJournal between 1996 and 2006. References to subsequent publications in the field arealso provided. Such research has implications for medium and long-term financialsimulations, capital adequacy, resilience reserving and asset allocation benchmarks aswell as for the immunization of short-term interest rate risk, for investment policydetermination and the general quantification and management of risk pertaining to thoseassets and liabilities.This thesis reviews Thomson’s model and methodology from both a statistical andeconomic perspective, and identifies various problems and limitations in that approach.New stochastic models for actuarial use in South Africa are proposed that improve theasset and liability modelling process and risk quantification. In particular, a new MultipleMarkov-Switching (MMS) model framework is presented for modelling South Africanassets and liabilities, together with an optimal immunization framework for nominalliability cash flows. The MMS model is a descriptive model with structural features andparameter estimates based on historical data. However, it also incorporates theoreticalaspects in its design, thereby providing a balance between purely theoretical models and those based only on empirical considerations.