Association for Public Policy Analysis & Management

Increasingly, tax returns have become the primary source for tracking inequality trends in the United States, although limitations of these data have largely limited their use for analysis of the distribution below the top income shares.  We overcome these obstacles to observing the full distribution of income in tax data and present the first set of comprehensive tax-based income and inequality measures using this data.  In doing so, we first address the missing population and income of nonfilers using information returns. Second, we use the address fields in the universe of tax returns to combine “tax units” into households, thereby matching the sharing unit traditionally used by non-tax-based economic researchers which more closely represents the level at which consumption decisions are made.  When comparing income inequality levels and trends using the full distribution of tax data to the Census Bureau’s results for equivalent definitions of income, we find that results mirror those from the Census but with important differences in both tails of the distribution. We argue that while there is uncertainty in the lower-tail of both the Census data and the IRS data, the tax data can provide valuable additional information on incomes in the middle and lower tail of the distribution as researchers are increasingly cautious about survey based income data.