Let the data speak: a new approach to sovereign risk assessment
- Insurance companies and asset managers need to rely on proprietary internal rating for sovereign bonds in order not to be exclusively dependent on rating agencies.
- We introduce a new internal rating model that provides a purely quantitative assessment of sovereign credit risk, with-out resorting to potentially opaque subjective adjustments made by rating agencies. It uses a framework based on em-pirical evidence on the determinants of sovereign stress. The model relies on a limited set of parameters which are empirically validated and supported by economic theory. It can easily be applied to a large set of countries.
- As no manual, qualitative adjustments are made, our model results – opposite to those of the rating agencies – cannot be criticised for subjective bias.
- The model uses available macro data (including medium-term projections) in an efficient and transparent way, blending two approaches: a) a panel regression model, to perform out-of-sample projections, and b) a machine learning algorithm (k-means) to cluster countries according to credit risk indicators.
- We currently cover 72 countries, with a focus on those most relevant for a liability-driven manager like Generali Insurance Asset Management (GIAM). The structure of the model is flexible enough to allow for a quick extension of the country coverage. In total, the model is based on a manageable set of roughly 20 economic time series variables.
Download the full publication below