Maimer and Persson (1994) provide a direct estimate of to what extent people used asset transactions to lower their tax on labor income. In Table 1 we reproduce their results based on the 1980 cross-section of HINK. Each individual in the 20-64 age group is sorted into deciles by reported labor income.
Column 1 reports the average labor income (including certain social benefits) in each decile. Column 2 shows the tax (‘simulated tax’) that should have materialized if people’s only source of taxable income had been the labor income reported in the first column (allowing for the 1980 tax brackets, basic deductions, etc.). Column 3 shows the tax (‘final tax’) that materializes when Maimer and Persson also consider the actual taxes on all kinds of taxable asset income. Columns 4 and 5, finally, show the implied average tax rates.
Starting at the bottom row, which shows the figures for the average individual in the sample, we see that the deductions for interest expenses, realized capital losses, and private pension savings exceeded reported asset income. As a consequence, the final tax falls short of the simulated one, and the average tax on labor income is lowered from 38.1 to 33.9 percent.
But these tax reductions are very unevenly distributed. While the discrepancy between simulated and final tax rates is modest in the lower deciles, it is huge in the upper ones. Through various capital transactions the average individual in decile 10 lowered the average tax on labor income from 52.6 to 42.1 percent. Even more strikingly, Ма1тёг and Persson report that a third of the individuals in the highest decile managed to reduce the final tax on labor income by more than 25 percent, and that a tenth of them halved the same tax rate.