本文基于措辞提取方法,以《中国货币政策执行报告》为文本基础,对中国央行沟通指数进行测度。本文对货币政策状态的判断、关键措辞的提取、措辞权重的确定以及指数的计算等环节进行了详细的阐述,以达到全面性、稳健性以及可复制性的测度目标。测度方案的敏感性分析表明:增加国外经济形势的措辞描述,由于其与国内信息的差异性使得指数更为平滑;改变措辞类型的权重对测算结果影响不大;改变措辞类型的数量虽然不会改变指数所显示的货币政策倾向,但会明显改变其强度;改变个别时期的货币政策状态判断,对有效措辞类型的选取有一定影响,但对最终沟通指数的构建影响甚微。进一步对货币政策工具的谱分析显示,相比数量型和价格型两类实际干预工具,央行沟通的谱密度在各个频域的波峰分布较为均衡,体现出既强调快速反应,又重视政策长期连续性的特点。
In this paper, we transform words from China Monetary Policy Report into quantitative indicators based on the decoding algorithm. We provide a detailed description of the judgment of the monetary policy phases, the identification of significant signal words, the determination of the weights of the signal words and the calculation of the indicator to make our algorithm comprehensive, robust and replicable. The sensitivity analysis shows that: (1) the indicator becomes smoother when we include the signal words describing the foreign economic situations; (2) changing the weights of the signal words doesn' t have much influence upon the final indicator; (3) changing the number of signal words will not change the direction of the monetary policy but will influence the intensity significantly; (4) changing the judgment of controversial monetary policy phases influences the chosen signal words but has little influence upon the final result. The further spectral analysis upon the monetary instruments shows that the spectral density of central bank communication is distributed more uniformly among different frequency domains compared with quantitative and pricing instruments, which reflects its characteristic of reacting fast and emphasizing the long-term continuity of policies.