4. A Bayesian Pricing Model for CAT Bonds

This paper examines the impact of the 2005 hurricane season, particularly Hurricane Katrina, on the pricing of CAT bonds. We examine whether highly rated CAT bonds demonstrate a different relationship than subinvestment bonds between objective risk measures and the spread. The theoretical framework for this relationship is based on the Lance Financial (LFC) model, introduced by Lane (Rationale and results with the LFC cat bond pricing model, Discussion paper, Lane Financial LLC, Wilmette, 2003). The empirical results of treed Bayesian estimation confirm that the severity component of the spread has an increased impact, indicating a shift in investor perception during the pricing process. The impact of the conditional expected loss also significantly increases, but it contributes through its interaction with the attachment probability rather than through its variance. Finally, we show that the influence of conditional expected loss is also increased by investment-grade ratings, because investors who demand highly rated bonds may be more concerned about possible losses than junk bond investors.