A0921
Title: Estimating spatial decomposition of income inequality via constrained Bayes method
Authors: Yuki Kawakubo - Chiba University (Japan) [presenting]
Kazuhiko Kakamu - Nagoya City University (Japan)
Abstract: The class of generalized entropy (GE) inequality measures, which includes the widely used Theil index as a special case, has the property of additive decomposability. When the population (entire country) is divided into non-overlapping and exhaustive subpopulations (regions), the GE of the entire country is decomposed into the weighted average of the GE of each region (within-region inequality) and the GE of the mean incomes of the regions (between-region inequality). In this research, the GE of the entire country is estimated, and those of the regions based on grouped data in a way that yields estimates that are compatible with the decomposition. First, the GE of the entire country is estimated by assuming some suitable parametric income distribution. Next, a parametric income distribution is fitted to each region, but as the sample size of each region is often not very large, the parameter vectors are modeled by linking to region-wise auxiliary variables in order to borrow strength from other regions. Based on the model, the GE of each region is estimated using the constrained Bayes method under the constraint that the decomposition holds. The proposed method is applied to the Japanese income data, and the results are compared with those of several existing methods.
