中国外汇市场压力的动态评估与货币危机预警--基于动态Logit和神经网络模型的研究

Dynamic Assessment of China's Exchange Market Pressure and Currency Crisis Warning--Based on Dynamic Logit and Neural Network Model

我国汇率改革和人民币国际化导致汇率波动幅度显著增加,及时监测我国外汇市场波动并进行预警是抵御外部不确定性冲击的前提条件。本文首先构造了外汇市场压力(EMP)指数,以此反映外汇市场波动并及时识别我国出现货币危机的可能性;然后从宏观经济、银行体系和外部冲击三方面选取预警指标,建立四种不同形式的Logit模型,进行HL与预测-期望分析检验,结果发现动态Logit模型拟合效果最优,结果表明,外汇储备增长率和通货膨胀率与中国货币危机发生概率呈反向关系,M2/国内信贷与中国货币危机发生概率呈正向关系;最后基于动态模型进行的样本外预测表明,我国发生货币危机的概率非常低,通过BP神经网络方法、改变货币危机的识别方式进行稳健性检验,验证了我国外汇市场平稳波动、出现货币危机概率很小的结论。

China's exchange rate reform and RMB internationalization lead to significant increase in the fluctuation range of the exchange rate.Monitoring the fluctuation of China's foreign exchange market and early warning timely are the prerequisites for resisting the impact of external shocks.This paper aims to monitor and warn the fluctuation of China's foreign exchange market.First,we calculate the foreign exchange market pressure index(EMP)to reflect the fluctuations of the foreign exchange market and identify whether there is a currency crisis in China;then we select early warning indicators from the three aspects:macro economy,banking system and external shocks,and build four different forms of Logit models.Through HL and forecast-expectation analysis tests,we find that the dynamic Logit model has the best fitting effect.The model shows that the probability of currency crisis happening is negatively correlated with the foreign exchange reserve and inflation rate,but positively correlated with M2/domestic credit.Finally,this paper predicts the probability of the currency crisis happening in the following year based on dynamic model.By the BP neural network model and changing the recognition method,we conduct robust tests and conclude that China's foreign exchange market fluctuates steadily and the probability of currency crisis happening is very low.