Piketty, data and the scientific method

Economics super star of the day, Thomas Piketty has been taken to task about his empirical data and his method of compiling it for his best seller, Capital in the 21st century.  FT economics editor Chris Giles noted that Piketty’s measurement of the inequality of wealth in the UK as provided by the official statistical office did not match that of Piketty’s (http://blogs.ft.com/money-supply/2014/05/23/data-problems-with-capital-in-the-21st-century/).

So Giles went through the wealth of data used by Piketty to reach his conclusion about the rising inequality of wealth and income in the major capitalist economies. Giles found that Piketty had made a simple mistake in transcribing some of his data on Sweden.  He also noted that, in filling in gaps with the raw data on France and the UK, Piketty made ‘arbitrary’ changes in the assumptions for his estimated data without explanation.  And when constructing the time series for the average wealth inequality for Europe using just three countries: the UK, France and Sweden, he used a simple mean average even though Sweden is a much smaller country.  The average should have been weighted to, say GDP or population. Giles also reckons that Piketty ‘cherry-picks’ his data sources,using different measures in different countries at different times.  Giles made new calculations with other data sources and found that there is no “obvious upward trend” in inequality of wealth in Europe.

All this has the whiff of that old controversy about the work of Carmen Reinhart and Kenneth Rogoff who concluded from a pile of historic data on sovereign debt, that if the debt to GDP ratio rose above 90% in any country, real GDP growth would subsequently be lower by several percentage points.  The idea that an economy could have too much debt for its own health was used by right-wing supporters of fiscal austerity for policies of cuts in government spending.  This was anathema to Keynesian economists and a researcher at the more left-wing University of Massachusetts, Amherst, delved through the data provided by the two RRs and found of heaps of mistakes that if corrected seemed to refute the claims that there was a debt threshold that caused slower growth (see my post,
Now in this case of Piketty’s work, we have a right-wing economics journalist trying to refute the claim that inequality of wealth is getting worse by ‘picking’ Piketty.

So do these latest revelations about the inconsistencies and weaknesses in Piketty’s data do the trick?  Well, as someone who often labours over raw data in order to get empirical evidence of what is happening to an economy over time, I have sympathy with Piketty.  Data are always inadequate and often inconsistent and it is also easy to make simple mistakes.  But it is better to try and provide evidence and, above all, release sources and your workings for all your data so that others can check and, even better, try and replicate your results.  That is the scientific method.  Above all, showing your data sources and workings means that any fakery or trickery can be avoided.  Fakery has not been missing from even eminent scientists’ work, and not just in the social sciences – the tobacco and drugs industry have much to answer for in this category.

Now Giles is not accusing Piketty of this. He is just finding errors and unexplained arbitrary assumptions in Piketty’s workings. But do these refute Piketty’s conclusions?  As Piketty says in his reply to Giles, that at least he has put all his data and workings on line for people to analyse (http://blogs.ft.com/money-supply/2014/05/23/piketty-response-to-ft-data-concerns/).  I have used his online appendix already (see my post, https://thenextrecession.wordpress.com/2014/04/23/a-world-rate-of-profit-revisited-with-maito-and-piketty/) for a wealth of information.  And that is more than we can say about the bulk of mainstream economics, which either offer no evidence to back up theoretical claims or fail to provide any workings or both.  As Piketty says, he has been more transparent that most with his evidence.

He also argues that more recent work on inequality of wealth by his colleagues, Emanuel Saez and Gabriel Zucman, using different measurement methods “confirm and reinforce my findings” SaezZucman2014Slides. and PikettyZucman2014HID . So Piketty reckons that any mistakes or biases in his own data “will not have much of an impact on the general findings”.  However, the problem with these latest studies, at least for the US, is that they are based on tax returns to America’s Inland Revenue Service.  As such, they exclude non-cash income as well as most of the social security benefits.  So an increasing percentage of what poorer Americans receive in income is not included.  This probably does not make much difference to the trend of rising inequality income and wealth, but it does undermine the consensus argument coming from the majority of the left that the Great Recession was a product of rising inequality as labour income was squeezed and an ‘underconsumption’ crisis ensued when the great credit bubble burst (see my post https://thenextrecession.wordpress.com/2014/03/11/is-inequality-the-cause-of-capitalist-crises/).

According to a new study (http://www.brookings.edu/blogs/up-front/posts/2014/05/20-rising-inequality-1920s-measuring-income-burtless), if social benefits are included in family income, then real net incomes for the bottom 20% rose 50% from the 1970s and were not stagnant, as claimed.  Sure, the top 1% did even better, so inequality grew.  But the crisis was not caused by falling or stagnant real incomes for the majority.  The study confirms the argument presented by Andrew Kliman in his book, The Failure of Capitalist Production, against the dominant underconsumption thesis

The point I am making is that the scientific method is full of pitfalls: human mistakes; inadequate data; unrealistic assumptions; inconsistent conclusions.  And these pitfalls are probably greater in the social sciences, given less data and where the political and ideological pressures are greater.  After all, much of modern mainstream economics is really just an apology for confirming capitalism as the best of all possible ways for human development – it’s vulgar economics, as Marx put it.  The scientific method allows researchers to build a body of evidence from different sources and using a variety of methods to reach a general conclusion.  Such a process has happened with global warming and the overwhelming evidence that it is man-made. The data in some of the work of the two RRs over debt proved less than robust, but several other studies using different methods have still supported the view that increased debt or leverage does damage the ability of a capitalist economy to grow until that debt is written off or devalued (see my post, https://thenextrecession.wordpress.com/2013/04/24/the-two-rrs-and-the-weak-recovery/).

In my view, there are more important deficiencies in Piketty’s work than inconsistencies in the data (https://thenextrecession.wordpress.com/2014/04/30/piketty-in-french-its-worse/).  For one, it’s the key difference between ‘wealth’ and ‘capital’.  As Keynesian Brad de Long has recently argued (http://equitablegrowth.org/2014/05/22/honest-broker-mr-piketty-neoclassicists-suggested-interpretation-week-may-17-2014/), there is a difference (de Long calls it the ‘wedge’) between what wealth holders get as a return on their investments and the rate of accumulation in new investment.  The rate of profit on accumulated investment (capital) can differ from the rate of return on wealth.  De Long reckons that the rate of profit can be falling (he uses the usual neoclassical marginalist reasons) and this will lower Piketty’s rate of return on wealth, unless the wedge rises to compensate.

In other words, in my Marxist view, Piketty’s r actually depends on the rate of profit on capital.  And it’s the law of the tendential fall in the rate of profit that is the central contradiction of capitalism, not the rising inequality of wealth.

6 thoughts on “Piketty, data and the scientific method

  1. I am an admirer of your columns. This is the first time that I have offered a comment. I am struck by the willingness of Piketty to reveal his sources and promote transparency, and by the fact that you have so openly praised Piketty for this. Perhaps in the future you would consider going more in that direction yourself. For example, in your book The Great Recession, most if not all of the graphs do not list at the bottom of each graph the sources for the data or other explanatory information.

    Thank you for your columns!

    1. Robert, point taken. Mea culpa. My excuse was that the Great Recession book was really a compilation of a series of articles from 2005 to 2009 and was self-published with little expectation that anybody would read it – and I was right. In the appendix there is some explanation of my methodology for the rate of profit measures but not sources or workings. In various academic papers since then, I have put sources and methods in and offered excel workings to anybody who wanted them. I am publishing a new book in early 2015, I hope, and will ensure that excel data is made available.

  2. Marx’s Capital is loaded with data, and almost all of it is “at hand” – direct government figures on production, employment, etc.

    Little of the data in Piketty’s book is direct. Raw numbers have been worked through considerable transformations, choices of conversion ratios, and approximate corrections. Furthermore, most of this work is buried in layers of files on a website, not in the book.

    Is the world more complicated than it was in Marx’s time? Perhaps. It is still possible today to analyze economic processes in the manner of Marx. Piketty offers shabby explanations, purely as a matter of establishing material necessity. That is the flip side of the way he grapples with data. Of course, the Financial Times is not interested in scientific explanation of capitalist processes. Haggling over the numbers is a great way to stir up a cloud of dust.

  3. Mr Roberts I have a problem with Estaban Maito’s fig 2 (UK) in ‘Piketty against Piketty’. How is it that in about 2003 the two graphs coincide and stay together for the next few years? The upper graph (say Marx) is P/C (profit/capital) and the lower flattish graph (say Piketty) is P/W (profit/wealth).

    Now Estaban’s whole case is that Piketty got it wrong by not realizing that W>C. But how come the distinction disappears from 2003 on-wards for the UK? [Figure 1 for Germany does not have this problem]. Have you any thoughts on this?

    1. This is what Maito says in conclusion: “Across the relative stability on Piketty´s rate, his “capital” shares changed in the long run. In XIX century and first decades of the XX, “land” and other items, which does not take part in a Marxian definition of capital, has a longer share. In last decades, machinery and non-residential buildings showed a larger relative increase. For this reason, levels of Piketty´s and Marxian rates show some convergence in the long term. And for that reason too, Piketty´s “rate of return” will probably continue to show a “Marxian” downward trend like it showed since the sixties (see Figures 1 and 2).” So Maito says that as the non-productive parts of Piketty’s ‘wealth’ (land and financial assets) decline as a share, Piketty’s r tends to converge with a Marxian ROP which is based on productive capital (machinery, plant etc in production). But as you say, that convergence does not apply to Germany after the mid-1980s. But Germany is the exception to the downward trend in ROP for most economies after the late 1990s on my measures. But no doubt more to discuss.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: