Bayesian Modeling of Regime-Switching Models: An Application in Multistream Asset Pricing

• Activity Number: 416
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): David Engler*+ and Brian Hartman
• Companies: Brigham Young University and University of Connecticut
• Keywords: Bayesian ; Regime-switching ; Asset pricing ; GARCH

Abstract

Writers of insurance guarantees have increasingly sought pricing improvement through model complexity. However, complex options often have no closed-form pricing solution and guarantees must be priced through stochastic simulation of the underlying asset. Regime-switching models are an intuitive way to incorporate stochastic volatility into the simulated asset price and have been shown to accurately model single asset streams (e.g., stock index data). Many guarantees, however, are based on multiple assets. Ignoring between-asset correlation can expose the writer to significant pricing risk. Recently, examination of two multivariate model families, regime-switching and GARCH, has been conducted. Whereas the multivariate GARCH models were found to generally fit better throughout central elements of the distribution, models based on regime-switching provided better fit in the tail. Since most investment guarantees are written to protect against tail risk, regime-switching models seem an appropriate choice. Because maximum-likelihood estimation can be problematic in complicated models, we adopt a Bayesian approach for the modeling of regime-switching data.