摘要
准确把握宏观经济发展趋势有利于前瞻性地调控经济运行,防范外部冲击。当前广泛用于宏观经济预测的MF-VAR模型,虽然能胜任常态情形的预测任务,但其参数估计过程多以传统Minnesota形式分布作为推断先验,难以贴合现实中异方差性的非理想预测环境。文章引入稳态随机先验对模型进行改进和优化,并通过湖北省主要宏观经济指标进行实例验证,发现稳态先验的“均值调整”信念驱使预测向均值回归,在简化估计程序的同时还能提高远期视野下的预测精度;随机先验的时变方差设定能有效捕捉序列的结构变动,使模型能同时适应常态和不确定性冲击的情形;分级稳态和因子随机波动可以牺牲部分样本信息而兼顾降维能力与计算优势。稳态随机先验的延展性和灵活性拓展了MF-VAR模型的应用场景,放宽了模型的应用条件,并进一步提高了预测精度。
Accurately grasping the macroeconomic development trend is conducive to forward-looking control of economic operation and prevention of external shocks. Currently, the frequency-mixed vector autoregression(MF-VAR) model widely used in macroeconomic forecasting is capable of predicting the normal situation, but its parameter estimation process mostly takes the traditional Minnesota form distribution as the inference priori, which is difficult to fit the non-ideal forecast environment of heteroscedasticity in reality. This paper introduces the steady-state stochastic priori to improve and optimize the model, verifies the main macroeconomic indicators in Hubei Province by examples, and finds that the steady-state prior“mean adjustment”belief drives the forecast to the mean, which not only simplifies the estimation procedure but also improves the forecast accuracy in the long-term perspective. The time-varying variance setting of stochastic priori can effectively capture the structural changes of the sequence, so that the model can adapt to the situation of normal and uncertain shocks at the same time. Hierarchical steady-state and factor stochastic fluctuation can sacrifice part of the sample information while taking into account dimensionality reduction ability and computing advantages. The ductility and flexibility of the steady-state stochastic priori expand the application scenarios of the MF-VAR model, relax the application conditions of the model, and further improve the prediction accuracy.
作者
刘洪
王丹阳
高跃伟
Liu Hong;Wang Danyang;Gao Yuewei(School of Statistics and Mathematics,Zhongnan University of Economics and Law,Wuhan 430073,China)
出处
《统计与决策》
CSSCI
北大核心
2023年第5期22-26,共5页
Statistics & Decision
基金
国家社会科学基金重大项目(20&ZD132)
湖北省第四次经济普查重点项目(HBJP2020-3)。
