In terms of credit scoring, an important issue is the quality of a rating model, namely the discriminatory power. In this thesis, we aim to improve the discriminatory power of rating models by using a forecast of the factors. Therefore, three types of rating models are analyzed: a standard scorecard rating model based on logistic regression, a rating model with macroeconomic forecast where the forecast of macroeconomic factors is based on a vector autoregressive model, and a rating model with financial forecast where the forecast is based on linear regression from the macroeconomic to the financial factors. Their performances on short-term and long-term prediction data sets are measured by Somers' D. The asymptotic standard deviation of Somers' D is calculated as the indicator of stability. After comparison between the three model types, the one with highest Somers' D is identified as an optimal choice. At the end we suggest an applicable model of the optimal type with reasonable factor selection and coefficients calibration.     «

In terms of credit scoring, an important issue is the quality of a rating model, namely the discriminatory power. In this thesis, we aim to improve the discriminatory power of rating models by using a forecast of the factors. Therefore, three types of rating models are analyzed: a standard scorecard rating model based on logistic regression, a rating model with macroeconomic forecast where the forecast of macroeconomic factors is based on a vector autoregressive model, and a rating model with fi...     »