混频数据信息下的时变货币政策传导行为研究——基于混频TVP-FAVAR模型

现实经济环境中,货币政策操作受到金融市场及宏观经济信息的共同影响.如何基于混频大数据信息分析货币政策行为机制是需解决的现实问题.为此,本文提出一种混频时变参数因子增广向量自回归(MF-TVP-FAVAR)模型.基于宏观经济及金融市场等多维度混频数据信息的实证结果表明:首先,MF-TVP-FAVAR模型在宏观金融混频数据中提取的金融形势指数(FCI)能较好地表征宏观经济先行趋势,为货币政策的制定提供前瞻性信息.其次,混频TVP-FAVAR模型可以捕捉价格型和数量型货币政策传导的高频时变特征.与货币供应量相比,利率传导对产出的影响具有滞后性.利率传导随着利率市场化改革愈发畅通,而信贷传导机制因财政政策搭配等问题日渐受阻.再次,货币政策传导效果受到经济周期影响,无论产出效应还是价格效应,经济上行时期,货币政策传导机制都比经济衰退时期更加通畅.价格型和数量型传导机制在经济下行时的作用效果均会减弱,但数量型货币政策更易受到经济周期的影响.最后,货币政策对FCI的冲击响应具有时变性,说明金融市场信息冲击对我国货币政策调控具有结构性的动态影响.货币当局制定尤其是微调货币政策时应及时评估金融市场信息冲击的影响.

The Time-varying Transmission Mechanism of Monetary Policy with Mixed Frequency Data: Evidence from MF-TVP-FAVAR Model

The global economy is facing increasing uncertainty, and the financial market is becoming more fragile. China's macro-economy is also facing problems such as economic structural adjustment and financial risk agglomeration, which make the relationship between monetary policy and macro-economy more challenging. The gradual reform of Chinese interest rate marketization and the rapid development of Fintech are also leading to a complex financial big data environment in terms of monetary policy. To understand the dynamic behavior of the monetary policy transmission mechanism, the available macroeconomic and financial market big data information must be utilized. Effectively analyzing the monetary policy mechanism through big data is thus a critical problem.
The transmission mechanism for monetary policies attracts extensive research attention. Some believe that the credit (quantitative) transmission mechanism is the main factor, while others suggest that the interest rate mechanism is more effective. Monetary policies typically exhibit time-varying features due to the business cycle, and thus a time-varying parameter vector autoregression (TVP-VAR) model is proposed to capture the behavior of monetary policies. The factor-augmented VAR (FAVAR) model is also used to analyze monetary policy, as it effectively utilizes real economic data.
The traditional TVP-FAVAR model uses only the same frequency data. However, the frequency of macroeconomic data is completely different from that of financial market data. Mixed frequency data are therefore widespread in actual economic activities. Effectively using such data to construct a TVP-FAVAR model and then analyze the monetary policy behavioral mechanism is therefore the challenge we face, and the aim of this study.
We propose a mixed frequency TVP-FAVAR (MF-TVP-FAVAR) model. We collect Chinese mixed frequency big data for our empirical study. The main advantage of the MF-TVP-FAVAR model is that it maximizes the integration of high-frequency financial market information and low-frequency macroeconomic information, and effectively extracts unobservable potential factors from a large amount of information. These advantages help us to more accurately analyze the time-varying relationships of monetary policy, macro indicators, and financial market indicators.
The mixed frequency data are mainly derived from China's quarterly and monthly macro data, and monthly financial data are also included. The sample period is from January 1997 to December 2017. The data sources are the National Bureau of Statistics and the WIND database.
The main conclusions of this paper are as follows. First, based on the MF-TVP-FAVAR model, the Financial Condition Index (FCI) extracted from financial market big data can better establish the dynamics of China's financial situation. This index is a leading indicator that can be used to measure economic performance, and an auxiliary indicator of the intermediary target of monetary policy. The FCI has a significant positive impact on interest rates and money supply, and this impact shows time-varying features.
Second, based on the time-varying response function of monetary policy shocks, the MF-TVP-FAVAR model captures the time-varying features of the impact of monetary policy at a macroeconomic level. This impact can be identified through the monthly observation frequency, which significantly improves the timeliness of the monetary policy transmission mechanism. Unlike the money supply, the impact of interest rate transmission on output shows a lag effect. Interest rate transmission has become smoother with interest rate marketization, while the credit transmission mechanism is increasingly blocked by fiscal policy coordination.
Finally, the business cycle has a significant impact on the transmission mechanism of monetary policy. We find that both the output effect and the price effect of this mechanism are more fluent during an economic boom than in a recession. Thus, monetary policy transmission is obviously cyclical. However, compared with price-based monetary policies, quantitative monetary policies are more susceptible to the impact of the business cycle.