Trends and Levels of Income Inequality in Wisconsin, 1919-1941: The Effect of Using Different Data Sources

Abstract

Long term measures of income inequality must grapple with the challenges presented by incomplete historical records. In this paper we examine one such problem affecting the quality of federal income tax return data in the period between the two World Wars, which form the basis for the widely used Piketty-Saez (2003) inequality series for the United States. To test the accuracy of IRS records in estimating historical top income shares, we compare them to a parallel series of independently tabulated state income tax returns in Delaware and Wisconsin, two states with unusually thorough records from this period. These state data sources offer a combination of advantages over the IRS including broader tax base coverage, more thorough income bracket records, and lower and more stable tax rates over time. By comparing estimates to federal IRS data, we find that state tax records yield highly different levels and patterns for the evolution of top income concentration. These findings attest to the high sensitivity of modern distributional estimation techniques to issues of quality with their underlying data sources, and suggest that the federal tabulations as found in Piketty-Saez systematically overstate the concentration of top incomes during the interwar period.