선형패널자료 분석방법에 관한비교연구

Abstract

This study discusses and applies many linear panel estimation methods from the latest literatures to a panelized dataset to see which method might be better. Actually, the panel estimation model can be divided into two categories, which is related effect model and unrelated effect model from the relationship between time-invariant variable and the explanatory variables. The unrelated effect model estimates the generalized Least Square (GLS) using the covariance matrix of the error term. The related effect model estimates panel estimators with three types-WGE, BGE, panel GMM. We estimates the determinant factors of the income in the household whose heads are above 50 years-old with the panelized Urban Household Income and Expenditure Survey. The empirical evidences say as follows: first, there seems to be no specific optimal approach for the panel estimation. Second, the best fitted one of the panel estimation methods depends on the research purpose and the data characteristics.