Robust Inference on Income Inequality: t-Statistic Based Approaches

Empirical analyses on income and wealth inequality those in other fields economics finance often face the difficulty that data is heterogeneous, heavy-tailed or correlated some unknown fashion. The paper focuses applications of recently developed \textit{t}-statistic based robust inference approaches analysis measures their comparisons under above problems. Following approaches, particular, a large sample test equality two parameters interest (e.g., regions countries considered) conducted as follows: samples dealt with partitioned into fixed numbers $q_1, q_2\ge 2$ $q_1=q_2=2, 4, 8$) groups, (inequality with) are estimated for each group, standard two-sample $t-$test resulting q_2$ group estimators. Robust $t-$statistic result valid general conditions estimators measures) considered asymptotically independent, unbiased Gaussian possibly different variances, weakly converge, at an arbitrary rate, to independent scale mixtures normal random variables. These typically satisfied empirical even pronounced heavy-tailedness heterogeneity possible dependence observations. The methods complement compare favorably available literature. use illustrated by across Russia.