资本计量方法改革、商业银行风险偏好与信贷配置

为提高银行业风险管理水平与信贷配置效率,监管部门于2014年开展了资本管理高级方法的试点工作。本文基于上市银行2010至2016年的微观数据,与银监会公布的行业信贷风险进行匹配,采用双重差分和三重差分法,实证分析前述改革如何影响试点银行的风险偏好和信贷调配。研究发现,在资本管理高级方法实施后:(1)试点银行显著降低了风险加权资产的规模;(2)试点银行风险偏好的变化存在非线性的特征,在调减高风险行业贷款的同时,并未显著增加最安全行业的贷款,而是增加了风险略高行业的贷款,体现出试点银行对风险与收益的权衡;(3)进一步将行业划分为“虚”与“实”,研究发现试点银行减少了房地产业(“虚”)、制造业(“实”)和建筑业(“实”)贷款,显著增加了金融业(“虚”)贷款。本文研究不仅丰富了资本监管方面的文献,也对金融支持供给侧结构性改革具有启示意义。

Reform of Capital Measurement Methods,Banks' Risk Preference and Credit Allocation

The 2008 global financial crisis is attributed to excessive risk-taking by commercial banks and the failure of financial regulation. In December 2010, the Basel committee officially released the Basel III standards, which revised multiple defects identified in the Basel II standards. Based on the Basel Accords, the China Banking Regulatory Commission (CBRC) introduced the New Capital Accord and planned to implement a series of reforms labeled Advanced Methods of Capital Management (AMCM). The AMCM consists of an internal ratings-based (IRB) approach to credit risk measurement, an internal model approach to market risk measurement, and an advanced measurement approach to operating risk measurement. In April 2014, the CBRC approved six banks to act as pilot adopters of AMCM by replacing the previous credit risk standard, under which risk weights were determined uniformly by the regulatory agency, with the IRB approach. This raises several issues in need of further study. Will AMCM implementation influence the effect of capital regulations, change commercial banks' risk preferences, or cause structural adjustments to credit allocation？
The partial implementation of AMCM enables us to use the difference-in-differences (DD) and triple differences (DDD) methods for empirical analysis. Using semi-annual and annual reports of listed commercial banks from 2010 to 2016, we collect data on 16 banks (6 banks as the treatment group and 10 banks as the control group). We use risk-weighted assets and the proportion of individual industry loans to total loans as the explanatory variables and other characteristics as the control variables. We further rely on the industry-specific non-performing loan ratio published by the CBRC as the measure of industry-level credit risk and then match this measure to our banking data to settle on 13 industries.
Empirical results using the DD method show that the pilot banks significantly reduce risk-weighted assets after AMCM implementation, which suggests that this policy successfully reduces banks' risk appetite. More importantly, this change in risk appetite may influence how credit is allocated to industries with different credit risk levels. Using the DDD method, we find that AMCM implementation prevents pilot banks from making loans to high-risk industries and has a nonlinear effect on loans to less risky industries; i.e., loans increase not for the least risky industries, but rather for ones with slightly higher risk. This illustrates banks' tradeoff between risks and returns. Due to the pro-cyclical nature of commercial bank credit, the influence of AMCM on credit allocation may have implications for economic activities. By dividing industries into virtual and real sectors, we find that pilot banks reduce loans to the real estate (virtual), manufacturing (real), and construction (real) industries, and they significantly increase loans to the financial industry (virtual). Therefore, pro-cyclical effects and banks' loan adjustment behavior may hinder the reallocation of credit from virtual to real sectors, which is not conducive to structural economic change.
Based on the above empirical results, this paper has several policy implications. For one thing, regulators should pay attention to the impacts of AMCM and encourage commercial banks to improve their customer data collection and risk analysis abilities to ensure the effectiveness of IRB systems. For another, regulators should implement macroeconomic regulations, such as counter-cyclical capital buffers, to mitigate the pro-cyclical effects arising from the IRB approach.
This paper contributes to the literature in several ways. First, there is presently disagreement over the effect of IRB approaches, and our paper enriches these discussions. Second, unlike previous domestic studies that explore bank credit behavior from the perspective of capital requirements, this paper analyzes the impact of changes in capital measurement methods on bank credit behavior. Finally, this paper offers insights on whether IRB methods can aid in the credit allocation flight from virtual to real sectors, sheds further light of the issue of financial resource flow from virtual to real sectors, and has value as a reference for regulatory authorities in setting policy.