中国经济波动来自趋势还是周期——基于多种UC模型的比较分析

传统趋势周期分解方法存在理论基础不符实际、缺少数据生成过程等问题，相较而言，UC模型具有一定优势。论文采用贝叶斯方法和中国宏观季度数据，估计了多个UC模型以分解产出的趋势和周期。研究发现：(1) 断点期在2008Q1的无约束UC模型为最优模型；(2) 趋势新息波动大于周期新息波动，两者高度负相关；(3) 趋势增长率发生了结构性下降；(4) 经济下行源于趋势下行而非周期下行。论文的基本结论在双变量模型、其它数据、其它经典单变量方法下依然成立。论文为宏观调控和“供给侧”改革提供实证依据。

Trend Fluctuations or Cycle Fluctuations for China: Comparison Analysis Based on multiple UC Models

The traditional trend-cycle decomposition methods have some problems that the theoretical basis is not realistic and the data generation process is lack. In comparison, the UC model has some advantages. Using Bayesian technique and Chang et al. (2016)’s quarterly data, this paper decomposes the output into trend and cycle by several UC models. The results are as follows: (1) the unconstrained UC model with breakpoint in 2008Q1 is the best model; (2) the volatility of trend innovation is greater than that of cycle innovation, and they are highly negative correlated; (3) the trend growth rate has a structural decline; (4) the economic downward is mainly from the trend rather than the cycle. The basic conclusions still hold under the bivariate model, alternative data and other classical univariate methods. It provides an empirical basis for macro-control and “supply-side” reform.