The Off the Wall Manner of Doing Data the Thomas Piketty Way

Salim Furth writes at WSJ withe regard to Thomas Piketty's Capital in the Twenty-First Century:

Economists Jesper Roine and Daniel Waldenstrom did the painstaking work of compiling and publishing Swedish wealth data for 37 individual years between 1873 and 2006. Mr. Piketty uses just 15 of the 37 available data points to construct a data series that departs from the original enough to show a small, steady rise in Swedish wealth concentration from the 1980s to the 2010s. Steadily rising wealth concentration supports Mr. Piketty’s hypothesis about capitalism — but contradicts the source from which he drew his data.

In an article for the Swedish think tank Timbro, Malin Sahlén and I compared the data that Mr. Piketty reports in his book to Mr. Roine and Mr. Waldenstrom’s original data. We found two questionable choices and one embarrassing goof.

The most questionable choice is Mr. Piketty’s publication of a data point for the 2010s. The source he cites ends in 2006, yet he arbitrarily decided that 2005 and 2006 are a good representation of the 2010s. But the results gainsay this “source”: Mr. Piketty’s measures of wealth concentration for the 2010s are higher than the measures Mr. Roine and Mr. Waldenstrom note for both 2005 and 2006. Mr. Piketty did not document any reason for departing from Mr. Roine and Mr. Waldenstrom’s data in this instance.

Without the fudge factor, Mr. Piketty’s data would have shown Swedish wealth concentration dropping below its levels from the 1990s and 2000s. Instead, his data show a smooth rise from the 1980s to the 2010s....

Mr. Piketty’s second questionable choice is to ignore 22 of the 37 data points in the source..

Finally, Mr. Piketty’s goof took place in splicing his series from data on wealth tax to estate-tax data in 1920. He notes that the adjustment is “standard practice,” but he accidentally adjusted in the wrong direction.