BIAS IN TAX PROGRESSIVITY ESTIMATES

Abstract

Tax progressivity is central in public and political debates when questions of vertical equity are raised. Applied, structural research demands a simple way to capture it. A power function approximation delivers one parameter that captures the residual income elasticity — a summary measure of progressivity. This approximation is accurate, tractable, and interpretable, and hence immensely popular. The most common procedure to estimate this parameter, a log ordinary least squares specification, produces biased and inconsistent estimates. A nonlinear estimator solves this issue and, using different data sets, I find differences in estimates between 6 and 14 percent.