Forecasting and analyzing insurance companies' ratings

Insurance companies sell protection to policy holders against many types of risks: property damage or loss, health and casualties, financial losses, etc. In return for this risk protection, insurance companies receive a premium from the policy holder, which is used to cover expenses and the expected risk. For longer-term risk protections, part of the premiums are invested to get higher yields. Although the protection buyer mitigates the individual risk to the large and better diversified portfolio of the insurer, it does not mean that the risk is completely reduced since the insurer may default his obligations. Insurers need to have sufficient equity or buffer capital to meet their obligations in adverse conditions when their losses on the diversified portfolio exceed the expected losses. Ratings provide an assessment of the ability of the insurer to meet its obligations to policy holders and debt holders. In this paper, the relationship between financial ratios and the rating is analyzed for different types of insurance companies using advanced statistical techniques that are able to detect non-linear relationship. The resulting rating model approach is similar to the approach for a low default portfolio, which uses a common set of explanatory variables (such as capitalization, profitability, leverage and size) which is generally applicable for all insurance types, and is complemented with insurance type specific ratios. The resulting model is found to yield a good accuracy, with 75% of the model ratings differing at most one notch from the external rating.