【摘要】 利用格兰杰非因果关系检验确定贝叶斯网络节点,反映经济变量之间的影响方式。采用误判率最小原则,确定预警指标临界值。利用贝叶斯网络学习,确定贝叶斯网络节点的后验概率。利用贝叶斯网络推理,测算地方政府债务风险,计算预警指标变化对省级政府债务违约概率的影响。研究结果表明:财政收入/财政支出与GDP增速/债务增速是预警省级政府债务风险的最重要指标,保持债务依存度、GDP增速/债务增速和民间投资增速/政府债务增速在适度区间内,能够有效降低省级政府债务风险。更多还原

【Abstract】 In this paper,we apply Granger Causality test to determine the Bayesian network nodes.By the principle of minimizing false alarm rate,we determine the optimal threshold of stress index.The MLS method was used to estimate Bayesian network parameters,thus the government debt stress model was established.By the Bayesian network inference methods,we determined the province government debt risk.Through finding the changes of debt default probabilities resulting from the change of stress index,we determined the impact of stress index on the default risk.Results show that the revenue/expenditure ratio and GDP growth/debt growth are the most important indicator.It can reduce the government debt risk significantly to keep debt dependence,GDP growth/debt growth,and civil investment growth/government debt growth in the moderate range.更多还原