论文部分内容阅读
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。
Credit rating is the possibility of evaluating a debt repayment or default loss rate. Because of the high risk, small amount, and unreal financial data of loans for industrial small businesses, commercial banks cannot accurately assess the credit risk of industrial small business loans. Therefore, the construction of a reasonable credit rating system has become an urgent problem to be solved. This paper constructs a logistic regression equation between a certain indicator and a default state. By comparing the log-likelihood of a zero model containing only a constant term with the log-likelihood of a complete model containing only one index, we construct it. For the ~2 statistic, the greater the two log-likelihood biases are, if there is no index, the greater the contribution of the index to the distinction between default and non-default, and the easier the indicator is to retain the indicator. It is ensured that the selected indicators all have a significant ability to distinguish the default status, and make up for the inadequacy of the existing research to distinguish the default status from the indicators. Second, by calculating the correlation coefficient of any two indicators within the same criterion layer, determine the degree of repetition of the information reflected by these two indicators. In the two indicators where the correlation coefficient is greater than a certain threshold, delete the χ2 statistic is small, that is, the state of violation. Differentiating indicators with a small degree not only avoids the information redundancy of the index system, but also avoids the accidental deletion of strong indicators for the judgment of the default state. To change the disadvantages of existing studies that remove one subjectively from a large correlation index. The third is to extract evidence from the loan data of a regional commercial bank distributed in 28 cities across the country, and establish a set of 26 indicators consisting of asset-liability ratio, cost-profit rate, and credit grants in the past three years. Index System for the Evaluation of Small Business Credit Risks. Fourth, through the credit rating of small industrial enterprises, not only the credit rating of each small enterprise, but also the corresponding default loss rate of each loan small enterprise, and the existing credit rating study only calculates the credit score of the loan customer and Insufficient ratings.