Richard Sutch attempted to reproduce the data found in Thomas Piketty’s Capital in the 21st Century. The abstract is brutal.
This exercise reproduces and assesses the historical time series on the top shares of the wealth distribution for the United States presented by Thomas Piketty in Capital in the Twenty-First Century….. I conclude that Piketty’s data for the wealth share of the top 10 percent for the period 1870 to 1970 are unreliable. The values he reported are manufactured from the observations for the top 1 percent inflated by a constant 36 percentage points. Piketty’s data for the top 1 percent of the distribution for the nineteenth century (1810–1910) are also unreliable. They are based on a single mid-century observation that provides no guidance about the antebellum trend and only tenuous information about the trend in inequality during the Gilded Age…This article does not question Piketty’s integrity
That’s from the published version. I don’t have easy access to the text, but the text from a 2015 Working Paper version is even more harsh. The working paper abstract is slightly different, and in particular doesn’t contain the disclaimer at the end, so there may have been some working out of differences.
Nonetheless, the following seems worth quoting from Sutch 2015, on aggregating one level up from the “bedrock” series on the percentage of wealth in the U.S. held by the top one percent in 20th century:
Since [Piketty] felt that the SCF data is more reliable than the estate data he chose to adjust the Kopczuk-Saez series upward to link with the SCF data and then to switch from the adjusted estate data to the SCF data at the earliest possible date, 1962. Kopczuk and Saez warn in this context that “patching together data from different sources is a perilous exercise” [2004: 479]. However, characteristic of his bold approach to the topic, Piketty did exactly that in a process he described as “homogenization” [Piketty, “Technical Appendix,” 2014b: 56-58].
Piketty’s series on wealth in the 19th century is interpolated from two data points. Sutch describes how the first data point is derived:
Piketty cites the source for the 1870 observation as Lee Soltow [1975: Table 4.2, 99] as reported by Peter Lindert [2000: Table 3, 188]. Soltow’s findings were based on an idiosyncratic “spin sample” drawn from the physical microfilms of the census enumerations. Soltow marked a spot on the glass screen of the microfilm reader, turned the crank a half turn, and sampled the individual whose name fell on the marked spot provided it identified a male 20 years old or older. He proceeded in this fashion through all 1,761 rolls of microfilm for the 1870 census [Soltow 1975: 4-5]!
And the second:
Piketty cited “Shamas (sic) 1993 and Lindert 2000” as his sources [Piketty spreadsheet TS10.1, also see Piketty and Zucman 2014: 17]. But neither Carole Shammas nor Lindert give a figure for 1810 [Shammas 1993, Lindert 2000]. Here is my best guess. Piketty started with Alice Hanson Jones’ estimate for 1774 for all households, 16.5, which is found in Lindert [2000: Table 3, 188].17 Piketty then rounded that off to 17. He then turned his attention to the 1860 free adult males figure for total assets in Lindert (referencing Soltow again). To obtain an estimate for net worth of a household he applied the ubiquitous 1.2 adjustment multiplier. He then read a figure for 1810 off a straight-line interpolation between 1774 and 1860. Following that procedure I get 24.5, which Piketty rounded to 25.
Sutch is even less forgiving regarding Piketty’s 10-percent writing:
When examining Piketty’s data for the trend of the top ten percent of the distribution it is difficult to be forgiving. The first point to make is that Piketty apparently thinks that his trend for the top 10 percent is more reliable than his trend for the top one percent.
at least in part because:
For 1870 Piketty reports Soltow’s number taken from Lindert, 71 percent, but without applying the 1.2 multiplier that he used on the top one-percent wealth share, which consistency suggests he should have done.19 Had he done so, his 1870 figure would be 85.2 percent of the wealth in the hands of the top 10 percent. But that figure implies there was a fall in the wealth share between 1870 and 1910. So much for the “well-established fact that wealth in the United States became increasingly concentrated over the course of the nineteenth century” [Piketty 2014: 347].
While I am sure that Piketty genuinely intends to paint an accurate picture, I have long shared Sutch’s frustrations.