January 1, 2008
Seth H. Giertz
This paper applies the methods of Gruber and Saez (2002) to a panel of tax returns spanning years 1979 through 2001 in order to examine the sensitivity of the elasticities of taxable and broad income to an array of factors. The paper finds that Gruber and Saez’s approach yields an estimated elasticity of taxable income (ETI) for the 1990s that is about half the size of my corresponding estimate for the 1980s. The paper finds that weighting regression results by income not only has a substantial impact on the estimates, but also results in overall estimates that are influenced by a small number of predominately high-income filers. For example, excluding 100 of the most influential observations (just 0.2 percent of the sample) lowers the estimated ETI for the 1980s from 0.37 to 0.08. Estimating the elasticities on the large SOI sample does not change the elasticities much, but does appear to yield results that are more robust to other specification changes.