地方政府或有隐性债务风险预警系统构建与应用研究--基于BP神经网络分析法

Research on the Warning System of the Contingent and Implicit Risk of Local Goverment Debt——With the BP Neural Network Method

政府债务风险是影响宏观经济稳定和公众预期的重要因素,其防控的关键在于事前预警,提早防范。依据形成地方政府债务风险的内外部影响因素,应用AHP、熵值法与BP神经网络法,构建包含显性债务风险、或有隐性债务风险和财政经济运行风险在内的3类一级指标和11项二级指标的地方政府或有隐性债务风险预警指标体系。结果表明,2017年我国半数省市政府债务风险处于中警状态,需加强对风险集聚的中西部地区融资平台债务、PPP债务融资、地方国有企业债务、商业银行不良贷款等政府或有隐性债务来源的风险分类防控。设计规范化风险防控流程及分类应对策略--风险规避、风险转移、风险减轻、风险接受与损失控制等,为财政提早应对风险、安排偿债优先秩序、优化财政收支结构等提供信号导向和政策参考。

Government debt risk is an important factor affecting macroeconomic stability and expectations. The key to its prevention and control lies in pre-warning and early prevention. According to the internal and external factors influencing the formation of local government debt risk, a warning indicator system including explicit debt risk, contingent debt risk and fiscal and economic operation risk which was constructed by the comprehensive application of AHP, the entropy method and the BP neural network method. The results show that in 2017, more than half of China’s provincial and municipal government debt risks are in a state of moderate warning, and that it is necessary to strengthen the risk prevention and control of financing platform debt, PPP debt financing, local state-owned enterprise debt, and commercial bank non-performing loans in the central and western regions where risks are relatively high. Then the standardized risk prevention and control procedures and classified response strategies - risk avoidance, risk transfer, risk mitigation, risk acceptance and loss control are designed. The strategies can provide policy suggestions for the early fiscal response to risks, the debt repayment priority arrangement, and the optimization of fiscal revenue.