基于似然比检验的工业小企业债信评级研究

Facility Rating of Small Industrial Enterprises Based on Likelihood Ratio Test

债信评级是评价一笔债务偿还的可能性或违约损失率。由于工业小企业贷款存在风险高、额度小、财务数据不真实等特点,使商业银行无法准确对工业小企业贷款的信用风险进行科学评估。因此,构建一套合理的债信评级体系成为亟待解决的问题。本文一是构造某一个指标与违约状态之间的逻辑回归方程,通过对比仅含常数项的零模型的对数似然值与仅含有某一个指标的完整模型的对数似然值,构造χ~2统计量,若有、无某指标时的两个对数似然值偏差越大,则该指标对区分违约与非违约状态的贡献越大,该指标越易保留的思路对指标进行遴选,保证遴选出的指标都对违约状态具有显著的区分能力,弥补现有研究不以能否区分违约状态对指标进行筛选的不足。二是通过计算同一准则层内任意两个指标的相关系数,确定这两个指标反映信息的重复程度,在相关系数大于某一阈值的两个指标中,删除χ2统计量小、即对违约状态区分程度小的指标,既避免指标体系的信息冗余、又避免误删对违约状态判别能力强指标。改变现有研究在相关系数大的指标中人为主观删除一个的弊端。三是通过提取中国某区域性商业银行分布在全国28个城市分支行的贷款数据进行实证,建立了由资产负债率、成本利润率、近三年企业授信情况等26个指标构成的适用于工业小企业信用风险评价的指标体系。四是通过对工业小企业进行债信评级,不仅得到每个小企业的信用等级,还得到每个贷款小企业对应的违约损失率,改变现有的信用评级研究仅仅计算贷款客户的信用得分和进行评分排序的不足。

Facility rating aims to evaluate the possibility of repayment or the loss given default（LGD）of one debt.It is hard for commercial banks to accurately evaluate the credit risk of small business loans,because of mall enterprises＇ high risk,small amount,and untrue financial data,etc.This paper established facility rating index system for small enterprises,which selected indicators based on the ability of distinguishing default state for the first time,while the second time was to avoid the information redundancy of selected indicators in the same guidelines layer.First,establishing logistic regression equation about default state and one indicator,then constructed χ^2-statistic by comparing the log-likelihood values between the zero-model which has no any indicator and the full-model which has an indicator.The greater the deviation of those two log-likelihood values,the easier the indicator can distinguish the default state,that is to say,the indicator should be retained.It compensates the disadvantage that the existing research has nothing to do with the default state when screening indicators.Second,the paper avoided information redundant by calculating the correlation coefficient between any two indicators in the same criteria layer,if the correlation coefficient between these two indicators is greater than a threshold,then remove the indicator which has small χ^2-value.This method can avoid redundant information and mistakenly deleting the indicator which has more significant impact on default state.What＇s more,it changes the disadvantage that the existing researches subjectively delete one indicator when the correlation coefficient is greater than the threshold.The results shows that the established debt rating system of small industrial businesses,including 26 indicators,such as asset-liability ratio,cost margins,the corporate credit situation nearly 3years,by extracting the related data of 28 regional commercial bank branches of China.the paper can not only get the credit rating of the loan enterprises,but also get the loss given default of per credit rating.It changes the disadvantages that the existing research can only calculate the credit score and give credit rating.