Morning Must-Read: Daniel Kuehn Reads Howard Reed on Piketty vs. Giles: “It’s All About the Discontinuities”
Daniel Kuehn: Facts & other stubborn things: It’s all about the discontinuities: “The other day I posted a suspicion of mine…
…that Giles was very wrong about Piketty based on his treatment of discontinuous data series…. It was just a suspicion, but it was due to a problem that I think anyone who has worked with disparate data sources would immediately recognize. ]Howard Reed has gone through the data sources and he agrees….](http://www.theguardian.com/news/datablog/2014/may/29/piketty-chris-giles-and-wealth-inequality-its-all-about-the-discontinuities) This is one of the best commentaries on this argument that I’ve seen yet. He does a great job walking you through the problems in Giles’s work. What’s more amazing is that Giles seems to recognize the difference between what he did and what Piketty did, but he does not recognize it’s significance. btw – some comments by Phil Magness on Facebook suggest he doesn’t get this point either…”
Howard Reed: Piketty, Chris Giles and wealth inequality: it’s all about the discontinuities: “Piketty uses six different data sources…
… b) Atkinson, Gordon and Harrison (1989) from Inland Revenue estates data, published in “Trends in the Share of Top Wealth Holders in Britain, 1923-81”, Table 1 (1923-81). c) Atkinson, Gordon and Harrison (1989) Data from Inland Revenue series C, Table 2 (1966-81). d) Inland Revenue Statistics 2005, Table 13 series C (1976-2005)…. f) ONS Wealth and Assets Survey (2006-10)I have identified three major discontinuities between the series…
…for the years 1974-81 where both series (b) and (c) are available, series (c) gives an estimate of the top 10% wealth share that is on average 6 percentage points lower than series (b); for the years 1976-81 where both series (c) and (d) are available, series (d) gives an estimate of the top 10% wealth share that is on average 6 percentage points lower than series (c); the first estimate from series (f) for 2006 is around 11 percentage points lower than the final estimate from series (d), for 2005. Taken as a whole, these discontinuities imply that the estimate of the top 10% share of wealth is 22.5 percentage points lower by 2010 than it would have been if the wealth statistics had been collected on a consistent basis after 1974 as they were before 1974.
As I demonstrate below, the main difference between the Piketty time series for UK inequality and the Giles time series for UK inequality, is that Piketty corrects his data series to allow for this 23 percentage-point drop (caused by changes in the methodology used to measure the wealth distribution), whereas Giles does not. While Giles has made it clear to me in private correspondence that he was fully aware of the discontinuities in the data series, he chose not to correct his final published data series to allow for them.