Content area
Full Text
Abstract - The Suits Index is often used in tax policy analysis to measure the degree of progressivity of a tax, or to analyze changes in progressivity under alternative tax regimes. As a point estimator, however, the Index provides researchers with no assistance in assessing whether changes are in fact statistically significant. We present a bootstrap methodology by which researchers can estimate confidence intervals for differences in Suits Indices. We also illustrate the use of that methodology with an application for the U.S. income tax, simulating the effects of removing housing deductions.
INTRODUCTION
Tax policy researchers often employ the Suits Index in order to measure the degree of progressivity of a tax system. They also compute Suits Indices to analyze changes in the progressivity of a tax over time, or under alternative tax regimes. While this index is widely used and very helpful for analysis of tax regime changes, it has one fundamental limitation. The index is a point estimator and provides no confidence interval estimates for use in assessing whether changes in the index are in fact statistically significant. Given the popularity of the Suits Index and its pervasive use in policy analysis, it is important to move beyond simple use of the index as a point estimator and begin to compute confidence intervals for the index.
The Suits Index is a summary measure of the distribution of a tax. Like any average, the index theoretically has a sampling distribution. If the exact finite sampling properties of the Index were known, including its standard deviation, we could construct confidence intervals in the usual manner. In reality, however, we have no information about the distributional properties of the Suits Index (just as we have no knowledge of the distributional properties of the Gini Coefficient). Furthermore, we typically do not have sufficient data to compute a large number of Suits Indices for a particular problem and appeal to the Central Limit Theorem for a convenient application of the normal distribution. Hence, we cannot rely on the usual methods of constructing confidence intervals. The purpose of this paper is to present a methodology by which tax policy researchers can estimate confidence intervals for changes in the Suits Index. We present a step-by-step method by which confidence...