Abstract: Macroeconomic stability has long been an important goal of China's macroeconomic regulation, and it has been more widely concerned during the “New Normal” period. As a necessary condition for macroeconomic stability, financial stability is predicated on the understanding of the financial cycle. This requires careful study of the impact of financial cycle variables on macroeconomic operation in order to support policy precision. This paper uses the relevant quarterly data of financial cycle and macroeconomic variables, applies the filtering method to find the fluctuant and stationary period of macroeconomy, utilizes the machine learning algorithm to verify the matching degree between the selected input variables and output variables and derives the importance of the fluctuation of each input variable to the fluctuation of the output variables. This paper finds that although the macroeconomic performances in the period of stability and volatility are different, the sum of the importance of the asset price index and the credit level on each of macroeconomic variables is more than 50%, regardless of the period of stability or volatility, while the importance of other financial cycle variables such as leverage and risk exposures of financial institutions are limited, which indicates that it is necessary to shoot the arrow at the target. Compared with releasing liquidity with a big scale, it can focus on maintaining asset price stability and optimizing credit policy rationing to achieve optimal results and escort macroeconomic stability.

Key words: macroeconomic stability, financial cycle variables, heterogeneity, refinement policy

摘要： 宏观经济稳定长期以来作为中国宏观调控的重要目标，在“新常态”时期被更加广泛关注。金融稳定作为宏观经济稳定的必要条件，以洞悉金融周期为前提，要求细究金融周期变量对宏观经济运行的影响，为政策精准性提供支持。本文利用相关金融周期和宏观经济变量季度数据，使用滤波方法找出宏观经济的波动和平稳时期，运用机器学习算法验证所选输入变量与输出变量的匹配度，并测度各输入变量对输出变量的重要度。本文发现波动期和平稳时期宏观经济表现虽有不同，但资产价格指数和信贷水平对各宏观经济变量的重要度之和都达到50%以上，而其他金融周期变量如利差、杠杆率和金融机构风险暴露的重要度则有限。这表明相关政策需有的放矢，相较于无差别的大规模释放流动性，可将重心置于维护资产价格稳定和优化信贷政策配给，为宏观经济稳定保驾护航。

关键词: 宏观经济稳定, 金融周期变量, 异质性, 精细政策