Model Choice based on Bayesian Theory in Spatial Panel Data Models: an Example of Spillovers of Chinese Telecom Services

ZENG Zhao-you,LONG Zhi-he,DONG Da-yong

DOI: https://doi.org/10.3969/j.issn.1007-3116.2010.03.013

2010-01-01

Abstract:A whole family of different spatial panel model has been developed,but how to choose the "best" model using appropriate criteria to meet the need of empirical research is still left as an unsolved problem.There is no uniform and coherent framework for such model choice.Furthermore,existing test for model choice aim at large sample-sizes mostly,which result in the biased estimation from finite sample.This paper develops a comprehensive and coherent Bayesian theory framework for spatial panel model choice.Finally,this framework applied to Spatial Econometrics model choice on spillovers of Chinese Telecom services.Findings indicate that spatial panel data model choice framework based on Bayesian theory is coherent in nest model and non-nest model;there are robust performances both in finite-sample and large-sample sizes.Moreover choice criterion can be computed directly,do not need simulation.Research discovered that the spatial autoregressive error model integrated network neighbor effect and economic relation is more suited the researches on spillovers of Chinese Telecom Services than the others,which is consistent with classical methods in spatial econometrics.