Panel Paper: The Sensitivity of U.S. Top Income Shares in Tax Record Data to More Comprehensive Measures of Income

Thursday, November 3, 2016 : 11:00 AM
Dupont (Washington Hilton)

*Names in bold indicate Presenter

Philip Armour1, Richard Burkhauser1, Gerald Auten2 and Jeff Larrimore3, (1)Cornell University, (2)U.S. Department of the Treasury, (3)Federal Reserve Board


Access to Internal Revenue Service (IRS) tax records improve researchers’ ability to track U.S. income and inequality, especially at the very top of the distribution not directly observable with survey-based data. However income measures on tax forms were designed to satisfy IRS administrative objectives, and are not always comparable with the income definitions economic researchers prefer. Using IRS tax record data since 1989 that we statistically match to Survey of Consumer Finances and Census data for income sources not available in tax data, we explore the robustness of levels and trends in inequality using the taxable market income definition often used in the top income literature to alternative income definitions that consider income more comprehensively. Additionally, we illustrate the importance of the choice of capital gains measures – demonstrating that focusing on taxable realized capital gains distorts the level and trend in top incomes relative to an accrued capital gains measure that is more consistent with the Haig-Simons income definition.  This is particularly important during the Great Recession, where the exemption of housing capital gains and losses from taxation meant that the substantial losses incurred by middle-income families were not reported on tax forms. As a result, while researchers incorporating only taxable realized capital gains observed a decline in top income shares in the initial years of the Great Recession, we observe that the top income shares actually rose sharply once fully including accrued capital gains and losses including those from housing.