空间面板模型协整检验的Bootstrap改进算法及功效分析  

Improved Bootstrap Algorithm and Power Analysis for Cointegration Test of Spatial Panel Models

作　　者：曾召友 ZENG Zhao-you(Hunan Academy of Social Sciences(Development Research Center of Hunan Provincial People's Government),Changsha 410003,China)

机构地区：[1]湖南省社会科学院(湖南省人民政府发展研究中心),湖南长沙410003

出　　处：《系统工程》2024年第2期151-158,共8页Systems Engineering

基　　金：湖南省社会科学院(湖南省人民政府发展研究中心)哲学社会科学创新工程资助项目(23ZYB54);湖南省社会科学成果评审委员会课题(XSP18YBC340)。

摘　　要：忽略面板模型横截面之间的空间相关性易导致其协整检验误拒率偏高。本研究采用模型变量的空间矩阵变换和回归残差Sieve Bootstrap技术相结合,既能使数据结构不受重抽样影响,又可发挥自助抽样(Bootstrap)技术优势,由此发展出基于Sieve Bootstrap技术的空间面板计量模型协整P检验,可在建模过程中充分考察空间效应的同时防止出现虚假回归。模拟及实证结果均表明:当面板数据模型出现内生回归元或横截面之间存在空间相依关系时,该检验不仅明显降低误拒率,还能保持高检验势,同时结果更具稳健性。Ignoring the spatial correlation between the cross sections of the panel model can easily lead to a high rejection rate in its cointegration test. This study combines the spatial matrix transformation of model variables with the regression residual Sieve Bootstrap technique, which can not only prevent the data structure from being affected by resampling but also leverage the advantages of Bootstrap technique. Therefore, a spatial panel econometric model cointegration P-test based on Sieve Bootstrap technique has been developed, which can fully examine spatial effects while preventing false regression. Both simulation and empirical results indicate that when there is a spatial dependency between endogenous regression elements or cross-sections in the panel data model, this test not only significantly reduces the rejection rate, but also maintains a high test potential. At the same time, the results are more robust.

关 键 词：空间面板协整检验 Sieve Bootstrap法 虚假回归 

分 类 号：C812[社会学—统计学]