I have a new working paper with Phil Magness. In this paper, we show a major methodological inconsistency in the estimates of income inequality produced by Thomas Piketty and Emmanuel Saez for the period from 1917 to 2020. The error is not just a level error but also a trend error. Below is the image that summarizes the paper. The abstract is right after and the link to the SSRN page can be found here.
Abstract: We present a new series for top income concentrations in the United States, using a consistent data construction methodology for the entire range of available data. This is meant to connect efforts that have separately considered pre-1960 and post-1960 inequality measures. Our series improves upon the series of Piketty and Saez (2003), correcting for data discontinuities induced by arbitrary choices that introduce distortions to their fiscal income denominator between the mid-1980s and the present. We then link our series with previous corrections to the Piketty-Saez series, creating a yearly estimate of top income concentration from 1917 to 2020. The results indicate more a more tempered rise in inequality in recent decades than the previous literature.
Vincent Geloso 