Three Essays on the Application of Second Generation Cointegration Analysis in Time Series Econometrics

Abstract

Frank (2009) constructed a comprehensive panel of state-level income inequality measures using individual tax filing data from the Internal Revenue Service. Employing an array of cointegration exercises for the data, he reported a positive long-run relationship between income inequality and the real income per capita in the US. This paper questions the validity of his findings. First, we suggest a mis-specification problem in his approach regarding the order of integration in the inequality index, which shows evidence of nonstationarity only for the post-1980 data. Second, we demonstrate that his findings are not reliable because the panel cointegration test he used requires cross-section independence, which is inappropriate for the US state-level data. Employing panel tests that allow cross-section dependence, we find no evidence of cointegration between inequality and the real income.