Predicting bond ratings using publicly available information

This paper developed a model to predict bond ratings using financial and non-financial variables although the rating agencies believe that bond ratings could not be replicated quantitatively by using a computer model. The model used an artificial intelligence (AI) technique that is non-parametric and designed to capture a dynamic relationship between input and output variables. The results showed that bond rating could be assessed quite accurately and critical variables were successfully identified. In addition the investment grade bonds were successfully distinguished from the speculative bonds.