Affiliations: Department of Statistics, Rome Hall (Room 558), 801 I
St. NW. George Washington University, Washington, DC 20052, USA. Tel.: +1 202
994 6548; Fax: +1 202 994 6917; E-mail: jlgast@gwu.edu
Abstract: The increase in income inequality, due to a shift in favor of the
upper end in the United States and other countries, has become a public policy
concern. This paper shows that the Gini index, G, underestimates the rate of
increase in inequality because a shift of income towards the top incomes
increases both the numerator and denominator of G. A modified index
(G_2), which replaces the mean in the denominator by the
median, indicates that income inequality in the United States grew at about
twice the rate as the Gini index from 1967 to 2012. The analysis accounts for
the effect of the changes made in the survey collection process in 1994. In
contrast with the United States, while income inequality increased in Sweden,
the shift in favor of the upper income region was less pronounced. The index
G_2 is readily computed from the mean, median and Gini
index, published by many national statistical agencies. Although
G_2 may not be appropriate for some analytic purposes, it is
an easily calculated summary descriptive measure that is more sensitive to
important changes in the income distribution than the Gini index. Most other
indices placing greater weight on the upper end of the income distribution must
be calculated from micro-data.
Keywords: Atkinson index, coefficient of dispersion, coefficient of variation, Gini index, Lorenz curve, mean to median ratio, polarization, robustness, skewed distributions, Theil index
