A new discrete formulation of the most commonly used tax progressivity
indexes is suggested. This approach allows one to rigorously prove th
at in some cases the Dalton/Musgrave-Thin and the Kakwani/Suits indexe
s may rank tax systems differently. Moreover it enables one, when comp
aring tax progressivity in two countries (or over time) to precisely d
etermine which part of the gap between the indexes is due to differenc
es in pretax incomes, in (standardized) tax rates, and eventually also
in the average tax burden. An illustration based on U.S. and Canadian
data for the year 1985 confirms the empirical relevance of this formu
lation of the tax progressivity indexes.