宏观压力测试下商业银行零售信贷产品PD 模型预测研究

通过将宏观经济指标与商业银行零售信贷产品住房按揭PD构建宏观变量预测模型,得到预测显著的GDP、CPI、HPI等三个宏观经济指标,再观察其不同滞后阶数组合VAR模型的AICC值,最终选取宏观经济因子高阶项构建回归方程和进行压力测试。研究结果发现:从施压时点开始,不同压力情景下PD均开始缓慢增长趋势,其中重度情景下PD增幅最大。说明使用宏观经济因子的阶乘能更好捕捉上述特征,PD预测模型能准确描述风险传导过程,此举可有效帮助商业银行加强零售信贷领域风险管理。

Under the Macroscopic Stress Test Commercial Bank Retail Credit Products PD Model Prediction Research

Stress testing is designated by the Basel Committee as an important tool for identifying, measuring and controlling liquidity risk and is an important part of the enterprise risk management framework. The China Banking Regulatory Commission also requires banks to establish a stress testing framework to effectively manage capital so that they hold sufficient capital to withstand risks at all stages of the economic cycle and assess potential losses in relatively extreme scenarios. However, the domestic retail banking stress test method has not yet fully unified the standard. The applicability of various measurement models under different scenarios needs further testing, and the study of the credit risk of retail banks is also discussed in combination with domestic macroeconomic variables. Based on Chinese macroeconomic operation rules, a bank's retail credit product policy and business operation mode are combined, as well as the availability of relevant historical data, to design a stress test plan for the retail credit portfolio. The stress test program has good operability and good decision-making reference value, which is extended to other similar retail banks to help modern retail banks strengthen risk management.The data of the housing mortgage default probability index are from the data of Bank A from 2012 to 2016, and the macro factor indicator data are from the data of China National Bureau of Statistics website from 2006 to 2016. By using house mortgage loan data collected from a commercial bank, a prediction model for probability of default is constructed based on macroeconomic indicators. Economic indicators that proved to be significant predictors to default probability are trained by value at risk model, including gross domestic product (GDP), consumer price index(CPI) and Herrick Payoff Index(HPI). Through observing the AICC value of the VAR model of different lag order combinations, the high order items of those indicators are used to build the regression equation and to do the stress test. The results show that the probability of default begin to increase slowly from the stressing point, and the largest increase occurs under the condition of severe stress. It is indicated that the factorial of macroeconomic factors can better capture the above characteristics, and the PD prediction model can accurately describe the risk conduction process, which can be a strong support to commercial bank for the risk management in retail businesses.