Thomas Piketty has now come out with a substantive response to the criticisms issued by Giles and Giugliano in the Financial Times. I am glad that he has done so and I suspect that there is much to chew over in the response, so I will look forward to reading the response of others to Piketty’s defense. For the time being, let me offer the following somewhat random observations (I am not going to comment on every paragraph or sentence in the letter, though I have read it all. I certainly encourage readers to read it all as well):

— Piketty tells us that he “certainly agree[s] that available data sources on wealth inequality are much less systematic than what we have for income inequality.” I am glad he states so; it is nice to establish that the data sources for wealth inequality are sketchy and incomplete. But while Piketty tells us that he is sure the data set can be improved, he also claims that he “would be very surprised if any of the substantive conclusions about the long run evolution of wealth distributions was much affected by these improvements.”

I am surprised that Piketty can admit both (a) that the data set is incomplete and can be improved upon; and (b) that it really doesn’t matter because, supposedly, even after it is improved, Piketty’s conclusions will somehow hold. Admitting the shortcomings should also mean admitting that the “substantive conclusions” in Piketty’s book might be improved by better data sets, but Piketty is unwilling to concede this (obvious) point.

— Piketty once again tries to defend his findings by pointing out that others have found even greater degrees of wealth inequality than he has, and that “if anything, my book underestimates the rise in wealth inequality.” I am going to quote myself in response: “[J]ust because other studies find widening inequality ‘by using different sources’ does not mean that they are right or that, even if they are, Piketty is justified in finding widening inequality through a flawed data set.”

— Piketty argues that “our ability to measure the most recent trends in wealth inequality is limited, partly due to the huge rise in cross border financial assets and offshore wealth.” I don’t think that having access to some accounts in the Caymans and/or Switzerland in order to augment our data set is going to do much to advance Piketty’s thesis concerning wealth inequality, but I am anxious to hear and read what others have to say on this topic.

— Concerning one particular argument on data sets, Piketty informs us that he prefers using decennial averages “because wealth series often display a lot of short-run volatility (in particular due to sharp movements in asset prices).” Well, okay, but some of the decennial averages only partially cover a particular decade (“1873-1877,” for example), so I am left wondering whether or not a better method can be used.

— Concerning the data set from the United States, Piketty concedes that “there are very large uncertainties regarding US historical sources on wealth inequality, and I certainly agree that the series that are provided in the book can be improved.” With all of these concessions, we may safely conclude that, if they have done nothing else, Giles and Giugliano have shown us that the data sets for wealth inequality are incomplete and in need of substantial buttressing. Dayenu, as me and mine are wont to say from time to time.

But again, Piketty falls back on the argument that (a) if anything, he understated the degree of wealth inequality he found and that others have found greater amounts of wealth inequality; and (b) improvement in the data sets “has little impact on the overall long-run pattern.” To be entirely fair to Piketty, it should be noted that he cites this paper by Saez and Zucman, which he says was done after his book was written, and which he believes “should be used as reference series for wealth inequality in the United States.” According to Piketty, Saez and Zucman “actually find a larger rise of top 10% wealth shares and especially top 1% and top 0.1% wealth shares than what I report in my book.”

— Regarding the data set from Britain, Piketty concedes — are you ready for this? — that there are “major uncertainties and limitations in our collective ability to measure [the] recent evolution of wealth inequality in developed countries, particularly in Britain.” Additionally, Piketty accuses the FT of cherry-picking wealth inequality estimates based on what particular decade of British data they want to study. He claims that “such a methodological choice is bound to bias the results in the direction of declining inequality” because “in every country wealth surveys [which the FT uses for later decades in Britain–ed.] tend to underestimate top wealth shares as compared to estimates based upon administrative fiscal data.”

There is certainly more to the Piketty response, but there still seem to be some unexplained contradictions in his arguments. I will be especially curious to see what kind of response Giles and Giugliano come up with.

Speaking of Giles, he has compiled an impressive list of “follow up problems” with Piketty’s data (this appears to be a response to Piketty’s response, though Giles does not link to the Piketty reply). Read it all, but let me note that Giles has taken the Saez-Zucman paper referenced above into account and has the following to say about it:

. . . Like any data, it has its shortcomings: in particular, a lot of imputation is required for pension and owner-occupied housing wealth (two of the largest sources of wealth), for which income streams are not taxed. As Prof Martin Feldstein has also noted, income tax data can be deeply problematic because the tax system itself affects the data supplied.

Again, to be entirely fair to Piketty — not to mention Saez and Zucman! — Giles does say that the Saez-Zucman paper “appears reasonable and further work on the reasons why its results differ from those published in Capital in the 21st Century is worthwhile.” But:

. . . in a separate slide of their presentation, the two authors make the identical point I have made, that evidence from estate taxes and the Survey of Consumer Finances (the sources used by Prof Piketty in his book) do not show the rise in wealth concentration for the top 1 per cent that Prof Piketty shows in his work.

This suggests that my comment that Prof Piketty’s own sources do not obviously show rising wealth inequality for the top 1 per cent is a point we should all be able to agree upon.

Concerning Piketty’s preference for estate tax data over wealth and asset surveys, Giles states the following:

. . .I believe this is the wrong choice, since the ONS data was specifically designed to measure wealth, correcting for potential biases. The survey oversamples the top decile to address known non-response biases, samples between 21,000 and 30,000 households in each wave and bases its sample on stratified techniques, not just bumping into people in the street and asking them questions.

I was therefore surprised for two reasons by Prof Piketty’s response to Bloomberg that the ONS wealth and assets survey is “very low quality”.

First, the UK research community does not share that view in general. Comparing different data on British wealth, Prof Sir John Hills, director of the Centre for the Analysis of Social Exclusion at the London School of Economics, writes about the difficulties of inconsistent data in his book Wealth in the UK: Distribution, accumulation and policy. There, he writes that the Wealth and Assets Survey is “the most comprehensive survey” and lists its drawback as its relatively short life. Just to avoid some confusion, the Wealth and Assets survey has the National Statistics kitemark and is not experimental.

Second, in Capital in the 21st Century Prof Piketty favours the US Survey of Consumer Finances, conducted by the Federal Reserve, which is a very similar voluntary cross-section survey with effort made to ensure known non-response biases among the rich are addressed. One big difference between the ONS survey and the SCF is that the US version is much smaller, even though the US is much larger, sampling 6,500 families in the latest wave.

It is inconsistent to accept a cross-section survey for the US when it gives high numbers, but reject one for the UK which gives low numbers. Prof Piketty needs to explain why he has made this choice and rejected the specific advice of HM Revenue and Customs not to use its latest survey of personal wealth survey in estimating the wealth of the UK as a whole.

What Prof. Piketty also does not explain is how he derives an estimate for the top 1 per cent for Britain in 2010 on the basis of these published estimates shown below. Sir John Hills in his table on the question did not attempt such a calculation.

Finally, regarding the issue of whether the corrections in data undermine Piketty’s point, Giles offers this (to my mind) entirely reasonable explanation of why it does:

My point is that if someone is claiming to have found a fundamental contradiction of capitalism, predicts the result is a rising share of wealth inequality, and uses apparent recent rises in recent wealth inequality as evidence that his theory is correct, those data lie at the core of the book’s argument.

Without the rising wealth inequality data across all advanced economies, I would submit that Prof Piketty has a theory without all the necessary evidence.

This post has gone on for quite a while, so let me close it by citing James Galbraith, who is no one’s idea of a right-wing economist, but who takes issue with Piketty’s claim that capitalism is responsible for wealth inequality:

In his feted book Capital in the Twenty-First Century , Thomas Piketty contends that wealth inequality rises inexorably under capitalism. The Financial Times has reported that, once various apparent errors are corrected, the European numbers show no tendency to rising wealth inequality after 1970. A judgment on this issue must, of course, await Prof Piketty’s response. Whatever the outcome, we are reminded that economic knowledge, especially in this area, is forged in long hours of work with sketchy and ambiguous statistical sources.

For the past two decades, the University of Texas Inequality Project has been contributing to this work. We did not set out to make, prove or disprove any grand theoretical claim.

Our main goal was to provide information, to clarify a factual record that was sparse, inconsistent and noisy. Our data on pay inequality, with estimates of income inequality, now cover most of the world from the early 1960s until (in some cases) as recently as 2012. We also find that inequality, measured within countries, rose in recent decades. But we do not find anything inexorable about this. We think that global circumstances and national policies were largely at fault. When policies and circumstances change, the rise of inequality can be stopped. Our data show that rich countries are more equal than poor countries. No surprise: to be rich is to have a large middle class. Communism defied this rule for a while, but not any more.

In rich countries “market income”, which reflects the income people receive from paid work, is very unequal, but this is because strong welfare states permit many households to exist without market income. Once taxes and transfer payments are factored in, disposable incomes are much more level. In poor countries the distinction is less clear. In some countries with weak welfare states and light taxation, measures of market and disposable income inequality do not appear to differ very much.

Piketty is now doing a better and more substantive job of responding to his critics. I am glad of that. But his critics have hardly run out of ammunition and Giles’s responses in particular still pack quite a wallop, one that may well be fatal to Piketty’s argument.