基于贝叶斯结构突变的随机系数动态面板模型研究  

Bayesian Inference of Structural Break in Random Coefficient Dynamic Panel Models

作　　者：王小刚[1,2] 王黎明[1,3] 

机构地区：[1]上海财经大学统计与管理学院 [2]北方民族大学信息与计算科学学院 [3]上海财经大学浙江学院

出　　处：《数量经济技术经济研究》2013年第8期149-160,共12页Journal of Quantitative & Technological Economics

基　　金：国家统计局全国统计科研计划重点项目(2011LZ035);上海财经大学科学研究项目(CXJJ-2011-439)的资助

摘　　要：针对目前随机系数动态面板模型中存在内生变量初始值固定、个体自回归系数平稳以及不存在结构突变的种种限制，本文提出用分层贝叶斯方法首次检测和估计了含未知结构突变的随机系数动态面板模型。在容许初始值与个体相关和自回归系数服从Logitnormal分布的假设下，得到了未知结构突变和随机系数的后验密度估计。对1995-2012年中国5省份出口总值月度数据进行实证分析，检测出4个结构突变，得到了出口总值存在的三大特征。This study is concerned with detecting and estimating structural breaks in random coefficient dynamic panel model using hierarchical Bayesian approach. Our approach allows for the initial values of each unit＇s process to be correlated with the unit specific coefficients and AR （1） coefficients drawn from a logitnormal distribution in order to hold stationary assumption. The posterior density for structural break and random coefficients were established. We apply our approach to analyze gross export value of five provinces in China from 1995 to 2012, and the results show that four breaks were detec- ted. There are three characteristics in gross export value. Firstly, it continues to show steady growth, but the gap gradually expand in five provinces; Secondly, large shocks of external demand has a significant impact; Finally, structural breaks of exports have sea- sonal characteristic.

关 键 词：贝叶斯方法 结构突变 GIBBS抽样 动态面板 

分 类 号：F064.1[经济管理—政治经济学]