In this paper, we propose a new hazard model for default prediction. In the model, macroeconomic exposures are formed to be linear functions of observable rm characteristics. With this feature, the model allows not only time-varying but also rm-specic exposures on macroeconomic risk factors. Empirical tests are performed in Korean market using the default data from 1993 to 2005. Our model outperforms alternative models with regard to the power of forecasting default of rms. We also nd that IT, health care and consumer companies are more exposed to changes in USD/KRW exchange rate volatility. Also, high credit quality rms are found to be more sensitive to macroeconomic eects, which is consistent with previous researches.