中国通货膨胀预期及其影响因素分析——基于混频无套利Nelson-Siegel利率期限结构扩展模型

综合国债市场的利率期限结构信息以及不同频率的宏观信息,本文构建混频无套利Nelson-Siegel利率期限结构扩展模型,在对不同期限债券进行一致性定价理论约束下,提取了中国通货膨胀预期的期限结构并对其进行影响因素分析。研究结果表明,本文模型提取的通胀预期期限丰富、结果稳健,具有较好的参考价值。通胀预期水平和变动响应主要受货币增长率、通胀率及全球食品价格变动等国内外相关宏观变量的影响,与国债收益率因子之间的关系不显著;国债收益率因子对中长期通胀预期的方差波动具有较强解释力,表明国债定价反映了未来通胀的不确定性。本文研究有助于充分利用我国宏观与金融市场信息条件,有效发现和锚定通胀预期,一方面,研究结果可为政策制定者和市场投资者提供科学的决策参考,另一方面,研究方法对丰富宏观金融领域的分析框架具有参考价值。

Analysis of Inflation Expectations in China and Their Determinants: Based on a Mixed Frequency No-arbitrage Nelson-Siegel Extended Model

Stabilizing inflation expectations is a prerequisite for the effective implementation of monetary policy. Scientifically measuring inflation expectations and analyzing the influencing factors are fundamental to the effective management of the macroeconomy and financial markets for investors, entrepreneurs, and policy makers. Both domestic and international economic and policy environments have recently changed drastically, particularly due to the 2020 COVID-19 pandemic, which is ongoing at the time of writing. A balance must be struck between the prevention and control of the pandemic and the development of the domestic economy. Inflation components such as industrial product prices and food prices have become structurally differentiated. Thus, accurately capturing the degree of inflation expectations and their changes attracts extensive attention from market investors and policy makers. Chinese scholars take two main approaches to exploring the nature, management, and determinants of inflation expectations. The first approach is to extract information on inflation expectations from modeling yield data and the actual inflation rate, but this ignores the influence of many macro factors and the underlying mechanisms. Quarterly frequency data can also be considered, but this approach is limited by the availability of low-frequency macro data. To utilize data more efficiently, we construct a mixed frequency no-arbitrage Nelson-Siegel (NS) extended model that jointly models the yield factors and macro variables of different frequencies under theoretical consistency, decomposing information on inflation expectations with various maturities from the Treasury yield curve. Our study makes two main novel contributions to the literature. First, we theoretically derive a mixed frequency no-arbitrage NS extended model containing GDP growth and other macro variables. The model has the characteristics of theoretical consistency and information efficiency. Under the theoretical constraint of consistent pricing of bonds of different maturities, we extract the term structure of inflation expectations, which can well reflect the level of and changes in inflation in China from the bond market and macro-financial variables with different frequencies. Second, in terms of application, we construct a framework for analyzing the factors affecting the term structure of inflation expectations. Unlike the traditional no-arbitrage model, our model guarantees the consistency of theoretical pricing through no-arbitrage conditions in combination with the simplicity of the traditional NS model. We also use the Markov chain Monte Carlo Gibbs sampling method for parameter estimation, and the results are relatively stable, with a good fit for the yield curve. In addition, the model considers and estimates mixed frequency data and effectively utilizes information from quarterly data at a monthly frequency, which overcomes the double defects of sample information loss and time delay encountered when analyzing quarterly data. In light of previous research, our macro variables include various important macroeconomic indicators with both monthly and quarterly frequencies, which may affect inflation expectations. We model these macro variables jointly with the Treasury yield curve. The sample period is from April 2006 to February 2020, with data obtained from the Wind Database. The key empirical results of this paper are as follows. First, the model fits the yield curve of Treasury bonds well. Second, the model infers a reasonable expected term structure of inflation, which is consistent with the Lang-Run forecast at quarterly frequencies. The agreement between the model's implied expectations and the survey data is greater after 2012, indicating that the model can extract more accurate macro information from the financial market as China's bond market gradually develops and improves. Third, our analysis of factors affecting inflation expectations shows that 1) the level of inflation expectations is mainly determined by the money supply growth rate, inflation rate, and global food price changes; 2) the short-to medium-term inflation expectation significantly responds to various macroeconomic variables; and 3) yield factors contribute most of the variance in future mid-to long-term inflation expectations, indicating that future inflation uncertainty is reflected in bond pricing.