基于混频数据的互联网金融风险测度及预测研究

Research on Internet Financial Risk Measurement and Prediction Based on the Mixed Frequency Data

从宏观经济、传统金融行业、互联网行业和互联网金融行业等维度,采集2014年1月~2022年3月的9组混频风险变量,结合频率转换法和动态因子模型对互联网金融风险进行实时、系统性测度;基于马尔可夫区制转换模型识别风险变化特征;根据该特征分别进行样本外滚动预测和固定系数预测。研究表明:混频变量对风险变化有较高的解释能力;风险的高低区制转换特征明显;互联网金融风险整体呈下降趋势,未来虽会经历温和上升,但仍处于低风险区制;随着互联网金融向其“金融”本质回归,风险的顺周期性及与传统金融风险的共振效应日趋显现。

Nine groups of mixed frequency risk variables during the period of January 2014 to March 2022 from macroeconomy, traditional finance industry, internet industry, and internet finance industry are adopted to timely and systematically measure the systemic risk of internet finance using the frequency-converting method and dynamic factor model. This research also use the Markov-switching model to identify the risk movement property and conduct out-of-sample forecasting using fixed parameter and rolling window forecasting methods. The results show that the mixed frequency risk variables have strong explanatory power in explaining the movements of internet financial risk;Internet finance risk has a significant regime-switching property with the risk level switching from high to low;Internet financial risk level exhibits a downward trend and will experience a moderate increase in the near future. However, it still resides in the low-risk regime in general. As internet finance return to its “finance” nature, the pro-cyclicality of risks and the resonance effect with traditional financial risks are increasingly emerging.