Faced with the imperfection of econometrics in terms of potential endogeneity bias, there are generally three options in research methodology for empirical test in economics. The most scientific option should be based on a comparison of all possible analysis methods and a choice of the optimal model based on objective criterion rather than subjective value judgment. This paper illustrates these three options by testing a Neoclassical growth model with the institutional effect of government size based on panel data of OECD countries from 1960 to 2013. The author does not argue that empirical finding of this paper is conclusive but argues that the estimation result obtained from Difference GMM estimator developed by Arellano and Bond (1991) is more reliable than those from the estimators of Within-group Fixed Effect (WG), Chamberlain’s π-matrix or Ordinary Least Square (OLS) as it can remove more endogeneity than the latter three methods. Despite differences in dataset, model specification and estimation methodology, the finding of a negative growth effect of government size for developed economies in this paper is generally consistent with other recent empirical growth studies since 2000, however, this paper gives much higher magnitudes of estimates for growth elasticity of government size than previous studies, possibly because Difference GMM estimator removes more endogeneity than other estimators.
Read full abstract