Rethinking Inequality Decomposition, with Evidence from Rural China

We introduce a new, integrated regression-based approach for decomposing inequality indices with household-level data, and we examine the strengths and weaknesses of inequality decompositions by income source in light of the way that they are commonly interpreted. The approach uses estimated income flows from variables in linear income equations to decompose aggregate inequality indices. The integrated approach provides an efficient and flexible way to quantify the roles of variables like education, age, infrastructure, and social status in a multivariate context. These tools are applied to a new data set with rich information on incomes in Zouping County in Shandong Province, China. The evidence from China illustrates the sharp differences that can result when using decomposition methods with varying properties, and it demonstrates advantages of the proposed, integrated method. We trace the differences to how the decomposition methodologies treat equally-distributed sources of income. The empirical results show the importance that spatial segmentations play in increasing inequality: village of residence strongly drives inequality in the sample. This force is counter-balanced in part by the relatively equitable distribution of human capital, especially demographic variables. Contrary to other recent findings, affiliation with the Communist Party and measures of social status have a very limited role in explaining inequality.