我国货币政策运行状态的量化分析与预测——基于非平稳离散选择模型和SVM方法比较

The Analysis and Forecasting on Monetary Policy Stance of China——Based on the Comparison between Estimations of Nonstationary Discrete Model and SVM

本文利用非平稳离散选择(NSD)模型,在两种时间刻度下,对我国央行货币政策操作中调整法定存款准备金率和定期利率的动态行为进行量化分析和预测,并与支持向量机(SVM)的预测结果进行比较.结果表明,核心经济变量及其纵向相对水平变化对央行货币政策调控决策具有显著且较优的解释能力.根据样本外模型预测结果,本文认为以月度为单位对央行制定执行货币政策行为进行分析预测比以季度为单位更为合适.虽然SVM模型整体样本外预测能力优于NSD模型,但NSD核心差分变量模型对央行上下调整政策行为具有较好的预测能力。本文结论对央行的货币政策调整决策行为具有一定的解释能力,有助于市场主体衡量经济运行状态,及时把握央行的货币政策操作动向.

Under two scales of time unit focus this paper on quantitatively studying and foreseeing the PBC（People＇s Bank of China）＇s dynamic monetary policy regulations on Required Reserve Ratio and One-year Deposit and Credit Interest Rate, through the Non-Stationary Discrete（hereafter NSD） model, whose estimated results are then used to compared with the results gained from Support Vector Machine method. As the results shows, key economic variables and their vertical changes present a significant contribution in explaining the PBC＇s monetary policy decisions. The Out-of-Samples fitting performance suggests that it would be better to analyze the central bank monetary policy making and implementing behaviors in month than that in season. Although SVM models have a higher forecasting accuracy over the NSD models in general, the NSD model with the key variables as well as their vertical changes show their strong ability in capturing the hiking and cutting decisions in monetary policy regulation correctly. To some extent, our conclusions can explain the PBC＇s dynamic behaviors of modifying the Required Reserve Ratio and One-year Deposit and Credit Interest Rate effectively and could be helpful in evaluating the economic statue in China and the PBC＇s monetary policy stance in time.