Common Risk Factors in the Cross Section of Catastrophe Bond Returns

Abstract

Historically, cat bonds have provided high single-digit average annual returns, paired with a low volatility and little correlation to other asset classes. While there is an extensive literature that explains (ex-ante) cat bonds spreads, there is no factor model in the academic literature that explains this (ex-post) realized return puzzle. Based on monthly quoted prices for the complete cat bond market from 2001 to 2020, we provide insights into relevant risk factors in the cross-section of cat bond returns. After investigating a battery of possible cat bond return factors in bivariate and multivariate portfolio sorts as well as Fama-MacBeth regressions, we propose a four-factor cat bond model. Its factors are the seasonality adjusted probability of first loss, a separate seasonality adjustment factor and the two corporate bond factors TERM and DEF from Fama & French. This novel four-factor model predicts 60% of the time series variation of the historical cat bond market returns - as opposed to 4% for the Fama & French three- or five-factor model - and substantially reduces the observable alpha of the cat bond market.