Macro Stress Testing The Credit Risk Of Commercial Banks In China Based On The Worst Case Scenario Analysis

Since the internationa l financ ia l cris is, the world’s financ ia l regulators haveemphas ized the prevention of syste mic risk in the bank ing syste m. As an importanttechno logy of measuring the system’s vulnerability, stress testing has been extreme lywide ly applied by foreign financ ia l regulators. China’s financ ia l regulators andscholars also ha ve been doing researches on relative theories and practices. But inmost studies, our scho lars simp ly copy fore ign models. Especia lly in the stressscenario des ign, they usua lly take expert subjective assumptio ns or d irect selectio n ofhistorica l cris is scenarios as a shock. It could exist some defects, such as scenariodoes not have the economic prospective, the degree of impact is not extre me enoughor not cover all risk points. All of above would make the stress test result ins ignificant.Therefore, in order to scientific a nd rationa l use the techniq ue of stress testing; it isparticularly important to improve the optimization profiles link of scenario design.This thes is aims to do optimizatio n in aspects of credit risk stress testingscenarios design and introduce the Worst Case Scenario Ana lys is estimation method.By constructing a model o f the macroecono mic scenario, we select11macroeconomicexplanatory variab les with the time span of29quarters and screen two principa lcompone nt factors as the impact factor. At the same time, we choose the rate ofnon-performing loans of the bank ing sector as a pressure ind icator. Subsequently wederive the only long-term co-who le equation as a linear approximation of themaximum loss based on the VAR mode l. Essentia lly we assume that risk factorsfo llow a multivariate joint elliptica lly contoured distributio n and apply Ma hala nobisdistance as the measure of the probability of occurrence of the severe scenarios. Afterthat we will do stress test respective ly us ing scenario des ign of history scenariodesign, assumptio n scenario des ign and the worst case ana lys is scenario design.Fina lly we compare three methods from two aspects of extre me degree of impactscenarios and the probability of occurrence, the fo llowing conc lus ions can be drawnfro m the test results: First, assuming the risk factors fo llow a norma l distrib utioncompared to the worst-case scenario estimate of the Ellipso id high distrib ution cancover a wider range of risks and identify more loan loss vo latility than the former.Besides, the loss functio n with the Mahala nobis distance can be established; Second,the method to establis h the function o f the loss and the Maha lanob is distanceestimated by the worst-case scenario estimation, the probability of occurrence of the severe scenarios can be measured. Therefore it could eliminate the d imens ionlessimpact and dime ns ion dependence; Third, VAR model of long-term cointe grationequation can be used as a linear approximation o f loss function in the worst-casescenario estimation; Fourth, there are some limitatio ns on the traditio na l scenarioanalys is method to identify potentia l risk fac tors and the impact duration, but theworst-case scenario ana lys is estimation can re flect the multi-year dyna mically cha ngepath of shock factors so that they would be certainly forward-looking. This is a goodcomp le ment to the traditio na l stress test scenario ana lys is; F ifth, it is necessary toimprove China’s financ ia l statistics syste m.This paper analyzes the defects of traditio na l methods of credit risk stress testscenarios design and introduce the worst scenario ana lys is estimatio n method. Thenthe fo llowing scenario design methodology conc luded lessons from co mparisons,inc lud ing the stress test scenarios to grasp the economic e xtreme degree offorward-looking, to measure the probability of occurrence of stress test scenarios, toimprove China’s financ ia l statistics syste m and stress test scenarios design syste m.