Under the current financial and economic situation, risk control of financial institutes is becoming one of the most significant factors to stabilize the economic condition. One of the most important methods for risk controlling is the evaluation of rating for each company, and now a generalized linear regression model such as the ordered logit model has been widely used in the preceding studies of the rating forecasting model.
In statistical modeling, the multi-nominal logit model as well as many other models have been developed from the multi-nominal logit model and are expected to decrease the I.I.A. (Independence from Irrelevant Alternatives) assumption. However, these models have not usually been used in the study of rating forecasting model so far.
In this study, we will construct a rating forecasting model based on the normal ordered logit model, multi-nominal logit model and nested logit model, and compare the model performance among these models. Finally, we will confirm the validity of these models by applying them to actual publicly disclosed rating data.