我国商业银行效率估计不一致检验与实证

The Test of Inconsistency of Efficiency Estimations of Chinese Commercial Banks and Empirical Analysis

为了检验我国商业银行效率估计的一致性,提出了一种类Hosmer-Lemeshow-致性检验方法(得到基于Kappa的U检验的验证),对5篇主流期刊文献进行检验,认为它们的效率估计相对排序基本不一致。不一致的主要原因是样本选择和投入产出结构的不同,而估计方法的影响有限。收集了1998～2004年14家主要银行的数据,采用同样的随机前沿方法估计出4种不同投入产出结构下银行的效率,结果表明效率估计对模型结构十分敏感。指出了这种现象的本质,并给出若干建议。

In order to test the consistency of efficiency estimations of commercial banks in China, the paper firstly proposes a Hosmer-Lemeshow-like inconsistency testing method and uses it to test 5 comparable literature papers that revealed significant inconsistency among the estimations. The result shows that the key reasons for such inconsistency lie in sampling and different input-output model structures, however, estimation method does not influence the results significantly. Using the data of 14 main banks in China during 1998 -2004, the authors estimates banks efficiency under 4 scenarios of different input-output structures with the same stochastic frontier method. Paired T and consistency test of the estimation results show that efficiency estimation is quite sensitive to model structure, which in fact implies different definitions of efficiency. Based on the empirical results, several conclusions and suggestions are reached in the end.