The credit risk premium moves in function of the initial assessment or rating and its possible changes over time. In addition, the rating of countries affects, in a straightforward way, the issues of public debt and, indirectly, those of companies based in the corresponding country. The credit rating assignment is not a totally transparent process made explicit by the rating agencies for commercial reasons, and therefore there is a divergence between the rating models they use.
The aim of this project is to build a simple and innovative framework based on latent variable methods for ordinal data, in particular with the focus on generalized linear mixed effects model using the rank likelihood. Statistical inference on the parameter of the proposed model is performed from a Bayesian perspective with the use of Markov Chain Monte Carlo methods. The main intent of this project is to demonstrate how the dynamic model we introduce, due to its simplicity and efficiency, outperforms the existing procedures used in this area, and provides multiple insights for future research.
One of the main novelties of our method consists in incorporating the rank likelihood approach into the well known framework of Generalized Mixed Effects Models applied to economic time series. A Bayesian modeling will be used which provides us with a richer information in comparison with the frequentist approach. The use of time series for each of the countries ensures the presence of a dynamic factor in the method.
Furthermore, in comparison to the existing models used so far to predict the government ratings, we are planning to make a forecasting for a longer period generating a large number of possible scenarios which would guarantee high accuracy of predictions.