The within-state inequality ratios are defined as the average income per household accruing to each fractile divided by state i’s average income per household. State i’s average income data at the denominator cannot come from the IRS publications because the number of income tax returns does not represent more than the richest 10 percent of the population in the early years of the sample. In other words, IRS data cannot represent the state average income consistently from 1913 to 2003, hence the resort to another data source for a comparison of top incomes (IRS) with the mean income (BEA).
Unlike the IRS’s Statistics of Income, the BEA’s State Personal Income data are not displayed by income classes, and both differ in definition. 23 As mentioned in the introduction chapter, adjustments needed to be made on the BEA income series so that the BEA definition fits the IRS income (itself modified for time-consistency purposes as described in Section 3.3.2). The details of the adjustments made on the BEA personal income series are presented in Appendix A.1.2. Overall, to divide the top income levels by the state average yields an indicator measuring within-state inequality.
“In addition to non-taxable government transfers, non-taxable personal income includes imputed rent; interest and dividends received by pension plans, life insurance carriers and non-profit institutions; non-taxable employer and employee contributions to pension plans, health insurance, day care, etc.; capital and inventory adjustments (NIPA capital consumption is generally smaller than IRS capital consumption, so that NIPA entrepreneurial income is generally larger than IRS entrepreneurial income); etc. See Park (2000) for a detailed description of the differences between NIPA personal income and individual tax return income.” (Piketty and Saez, 2004, p. 28, note 49.)