Thread by @VincentGeloso

Thread: My paper on revisiting income inequality in the USA is now online at the Economic Journal @RoyalEconSoc. Let me explain why it matters here. But first, just look at this graph which summarizes it all

doi.org/10.1093/ej/uea… #econtwitter #econhist #inequality #Economics

1) In the paper, we find that the estimates of income inequality in the US over the 20th century suffer from three massive flaws pre-1960.

2) First, because there are so few tax filers in the early years, some adjustments have to be made to get a steady share of the top fractiles. However, the adjustment used by Piketty and Saez was based on ratios of married to unmarried filers in 1942...

3) The problem is that 1942 was a war year marked by labour market disruptions caused by the massive enlistment of single and military-age men. It is therefore unlikely that the 1942 ratio accurately reflects the tax filing conditions of the previous two
decades

4) Second, Piketty and Saez make an adjustment for deductions because the IRS was reporting only net income pre-1943. However, these adjustments are fixed proportions with arbitrary year-changes. In spite of the fact that there are sufficient data to create deductions straight-up

5) To estimate top incomes as a share of all income, Piketty and Saez created a measure of all incomes compatible with tax data. However, the assumption they use was too conservative and was also fixed. The result was that it underestimated total income on avrage.

6) This is in spite of the fact that data exists since the 1950s and 1960s to create "tax reconciled" measures of total income (thanks to Selma Goldsmith).

7) So, we went about correcting all of these flaws. And they matter a lot. Here is a breakdown for selected years. In total, they knock off between 15% and 20% of the estimated inequality levels. That is *not* trivial.

8) The second implication is related to the "Great Leveling" (term borrowed from the great Jeffrey Williamson and the amazing Peter Lindert -- two econs I admire immensely). In the Piketty and Saez story, the War and the tax reforms that came during it caused inequality decline

9) With our data, the tax changes of the war are much smaller. In fact, a sizable most of the wartime fall in inequality in Piketty and Saez was a result of a typo in their file and an arbitrary income definition change.

10) Our data shows that the Great Depression was a much stronger leveller than the tax policy changes.

11) We had already hinted at that in an earlier article in Economic Inquiry where we used state-income taxes with richer datasets to assess reliability of IRS-federal data. onlinelibrary.wiley.com/doi/abs/10.111…

12) So, not only do we find lower inequality level and a much smaller role for tax policy, we can also connect with the work of economists like Splinter and Auten (heroes in my opinion)

13) Indeed, they found that there was a much smaller inequality increase from 1960 to today. Connecting theirs to ours suggests that there is no strong "U-curve" of inequality. Instead there is a "Tea-saucer" (which U-ish shaped if you want) of inequality in the US.

14) This is rife with implications because it changes our reading of American econ history. It de-emphasizes tax policy and points to other forces (which I have called elsewhere "mundane" because it is simply price-theory stuff like capital/labor mobility) books.google.ca/books?hl=en&lr…

15) This was long and arduous. I am proud of this work that started in 2016 and I hope you will share this widely.