The residual correlation assumption of the CPV model was adjusted, and then, the stress scenarios generation model and the risk conduction model can be handled separately. Therefore, the partial least square method can be used to estimate the parameters of credit risk conduction model , avoiding the problem that macroeconomic factors could not be contained in the stress testing system for multicollinearity. The parameters of stress scenarios generation model were estimated through seemly unrelated regression, and Monte Carlo simulation method was used for stress scenarios generation when there exist lagged terms of macroeconomic factor in credit risk conduction model . Case analysis results show that the proposed stress testing method can be effectively applied to banking reverse cycle management.