Predicting credit ratings and transition probabilities: a simple cumulative link model with firm-specific frailty
扩展了累积链接模型，通过引入企业特定的不可观测脆弱性变量来预测信用评级分布和转移概率，实证表明该方法比传统模型更准确稳健。
There has been a relatively large body of literature addressing the question of predicting credit ratings and transition probabilities. Using frailties to model and predict credit events has generally been shown to provide better prediction outcomes than models without frailties. The paper takes this approach and uses it to extend the general class of cumulative link models (CLM). In particular we impose a positive correlation structure on CLM between repeated ratings from the same firm by assigning an unobservable frailty variable to each firm. We first apply the resulting model to predict credit rating distributions for individual firms and then transform the results to make our target predictions of credit ratings and transition probabilities. Our predictions enjoy using firm-specific and macroeconomic covariate information and having simple computation and interpretation. As an empirical illustration, S&P long-term issuer credit rating (LTR) examples are provided. Using an expanding rolling window approach, our empirical results confirm that the extended model provides better and more robust out-of-time performance than its alternatives because the former yields more accurate predictions of S&P LTRs and transition probabilities.