Our paper "Incorporating idiosyncratic risk factors into CAT bond pricing: A machine learning approach" won the best paper award at the ECOS conference in Brno!
⛈️ Catastrophe bonds are a type of insurance-linked securities (ILS) that have been gaining increasing importance as an instrument for addressing climate change related risks and losses. They provide a way to transfer some of the risk insurers and reinsurers face, to the market and interested investors. For investors in turn, these instruments are a nice diversifier as, by construction, they are mostly uncorrelated to financial market risk. They are (or should be even purely) linked to the occurence of a catastrophic event and its probability of occurence.
💲 However, idiosyncratic risk factors play a significant role for the primary market pricing. Moreover, this primary market pricing is heavily dependent on external ratings of expected loss measures. We provide an analysis of an extensive set of idiosyncratic risk factors and show how these can be used to improve pricing forecasts in the primary market. We also compare a set of ML models and their predictive performance using different variable sets.
💻 We find that several models are significantly capable of incorporating idiosyncratic factors, while others are not. The importance of idiosyncratic factors is a potential hint at inefficencies as, by construction, same perils should be priced identically. Nevertheless, it seems, investors trust some insurers more than others.