中国通货膨胀波动的测算——基于TVP-AR-SV模型

Measure of China's Inflation Volatility:Based on TVP-AR-SV Model

准确测度通胀波动规律是研究通货膨胀及货币政策的重要基础。为克服现有随机波动(SV)模型隐含假设中存在的天然缺陷,平衡模型估计的准确性与操作的简便性,通过对比SV、UC-SV、AR-SV、TVP-AR-SV四种主要模型,认为其中时变系数自回归随机波动(TVP-AR-SV)模型最适用于中国通胀波动的测度。对中国的季度通胀数据建立该模型,利用马尔科夫链蒙特卡洛模拟(MCMC)方法对模型参数进行贝叶斯估计,得到通胀的波动序列,同时对通货膨胀进行预测,并与其他形式的SV模型的估计结果进行对比。结果表明,TVP-AR-SV模型估计的随机波动能够全面地解释中国通胀率的整体波动规律,参数估计精度及模型对未来通胀的预测能力均优于其他备选模型。

An accurate measure of inflation volatility lays foundation for inflation researches and monetary policy making.Considering hypothesis problems in existing SV models and trade-off of the accuracy and simplicity,a TVP-AR-SV model is finally chosen from 4 candidates for China's inflation volatility.MarkovChain Monte Carlo simulation is used to conduct the Bayesian inference of the time-varying parameters of the designed model.Thus,the stochastic volatility of the inflation rate series is obtained,and a forecasting can be done.Four other SV models are estimated at the same time just for comparison.The results show that stochastic volatility estimated by the model established here explains fluctuation of the inflation to a large extent,and the accuracy of parameter estimation and prediction for future inflation are better than those of other candidate models.