边界Logistic违约率模型Bayes分析及实证研究

Bounded Logistic Default Model and Empirical Evidences from China

Cramer[1]指出了一般Logistic违约率模型容易出现的问题并提出了边界Logistic违约率模型。本文采用了不同于Cramer(2004)的Bayes分析方法对边界Logistic模型的后验分布的性质进行了分析,从理论上说明了边界Logistic违约率模型更优越的原因。然后利用中国公司数据展开实证研究,不仅找到了Cramer问题的中国证据,同时还发现Bayes边界Logistic违约率模型不仅能够克服Cramer问题,而且对临界值不敏感,同时预测效率也相对较高。

Cramer（2004） pointed out shortcoming of plain Logistic default model and put forward bounded Logistic model. This paper does some further research about bounded Logistic default model. We first demonstrate why bounded Logistic default model is superior to plain Logistic model theoretically through Bayes analysis. Then we give empirical evidences based on China companies＇ data. We not only find evidences about Cramer＇ s problem, but also find that bounded Logistic model can solve the Cramer＇s problem, which is not sensitive to critical value and has higher prediction efficiency.