How progressive is the current U.S. tax system? Not very, if you read the new paper in the Journal of Economic Perspectives by Thomas Piketty and Emmanuel Saez. As Mark Thoma reported, they say that

The progressivity of the U.S. federal tax system at the top of the income distribution has declined dramatically since the 1960s.

But according to Alan D. Viard,

the federal tax system is highly progressive. The lowest quintile, the 20 percent of households with the lowest incomes, bears less than 1 percent of the federal tax burden. The next 20 percent bears less than 5 percent of the burden. The burden is higher for the next income groups, as the middle 20 percent bears 10 percent of the burden and the fourth 20 percent bears more than 17 percent.

That leaves the top 20 percent paying 66 percent of the cost of government; one fifth of the population pays two-thirds of the cost. The breakdown within that 20 percent is even more striking. Most notably, the top 1 percent of the population bears a staggering 25 percent of the tax burden.

Thanks to Greg Mankiw for the pointer.

What I find particularly fascinating is the different graphical ways of presenting data. Compare (a) the first chart in Mark Thoma’s blog post with (b) Viard’s figure one. They show exactly the same thing–overall tax payments by income category. In the lower part of the income distribution, people are grouped into quintiles. At the top end, we see the top one percent, which is a group that is one-twentieth the size.

1. (a) is a line chart, and (b) is a pie chart. In (a), it is less apparent that groups of different sizes are being compared.
2. (a) does not include the bottom quintile, and b does;
3. (a) includes the 1970 data, which spreads out the scale (because the tax system was more progressive then. As an aside, how much of the decrease in progressivity comes from greater breadth of stock ownership–if more people own stock, then more people are imputed to be be paying corporate income taxes–not that such imputations have anything to do with actual tax incidence? My guess is that the answer is “not much,” but it illustrates a potential hazard with this sort of analysis. There are many hazards, which is why I tend to shy away from these sorts of discussions.), and (b) does not

These three factors together mean that (a) tells a story of low progressivity and (b) tells a story of high progressivity. Can anyone say, “confirmation bias”?