Discussion of Matthew Rognlie: "Deciphering the Fall and Rise in the Net Capital Share"
J. Bradford DeLong :: University of California at Berkeley
Let me begin by thanking Matt Rognlie for doing some very serious and thoughtful digging into this set of factor-payments data. That digging leaves me in an ideal position for a discussant: There are interesting and important numbers. These numbers have not been put together in this way before. The author is wise enough not to believe he has nailed what the numbers mean to the floor. Thus I am in an excellent position to, if not add intellectual value, at least to claim a lavish intellectual-rent share of Matt Rognlie's product.
I was weaned on the education-deficit explanation of recent trends in US inequality, perhaps best set out by the very sharp Claudia Goldin and Larry Katz (2009) in The Race Between Education and Technology. In their view, the bulk of U.S. inequality trends since the 1980s were driven by education's losing this race. In the era that had begun in 1636 the United States-to-be had made increasing the educational level of the population a priority. But that era came to an end in the 1970s, while skill-biased technological change continues. That meant that the return to education-based skills began to rise. And it was that rise that was the principal driver of rising income inequality.
But, recently, reality does not seem to agree with what had once seemed to me to be a satisfactory explanation. First, to get large swings in the income distribution out of small changes in the relative supply of educated workers requires relatively low substitutability between college-taught skills and other factors of production. As inequality has risen, the required substitutability to fit the data has dropped to what now feels to me an unreasonably low magnitude. Second, while it is true that we have seen higher experience-skill premiums and sharply higher education-skill premiums, the real action in inequality appears now to be unduly concentrated in the upper tail. The distribution of the rise in inequality does not seem to match the distribution of technology-complementary skills at all.
Looking simply at my own family history, my Grandfather Bill reached not just the 1% or the 0.1% but the 0.01% back in the late 1960s in the days before the rise in inequality by selling his construction company to a conglomerate back in 1968. A good many of those of us who are his grandchildren have been very successful--consider my cousin Phil Lord's LEGO movie, and the other franchises for which he gets "director" but not, or not yet, "producer" and "created by" credits. But even should any of us be as lucky as my Grandfather Bill was in terms of our peak income and wealth as a multiple of median earnings, we would still be a multiple of his rank further down in the percentile income distribution.
Today, you need roughly 3.5 times the wealth now in the U.S. and 8 times the wealth worldwide to achieve the same percentile rank in the distribution (see Atkinson, Piketty and Saez (2011), Piketty (2014), Saez and Zucman (2014)). I find it simply impossible to conceptualize such an extreme concentration as in any way a return to a factor of production obtained as the product of "hours spent studying" times "brainpower", even when you also multiply by a factor "luck" and a factor "winner-take-all-economy".
So what, then, is going on and driving the sharp rise in inequality, if not some interaction between our education policy on the one hand and the continued progress of technology on the other? Thomas Piketty (2014) has a guess. Piketty guesses that the real explanation is that 1914-1980 is the anomaly. Without great political disturbances, wealth accumulates, concentrates, and dominates. The inequality trends we have seen over the past generation are simply a return to the normal pattern of income distribution in an industrialized market economy in which productivity growth is not unusually fast and political, depression, and military shocks not unusually large and prevalent.
What about what John Maynard Keynes (1936) called the “euthanasia of the rentier”? Eighty years ago, Keynes guessed that, as accumulation proceeded and the capital-output ratio rose, the relative rate of profit would decline, and would decline by more. Thus more and more concentrated wealth meant a smaller and smaller share of income received by pure rentiers--as opposed to entrepreneurs and risk-bearers? Keynes strongly believed that the returns to investment at the margin were likely to drop rapidly enough to make this "euthanization of the rentier" the most likely possibility. Matt Rognlie agrees. And, indeed, it is difficult to see how, if investment takes the form of the accumulation of useful physical capital, it could be otherwise.
In an anticipatory response to this part of the Rognlie critique, Thomas Piketty (2014) points to a remarkable constancy in the rate of profit. His data show it as stuck between 4% and 5% per year across centuries with very different capital-output ratios. Piketty, however, appears agnostic as to whether the cause is easy capital-labor substitution, rent-seeking via control of the government the rich, or social structures that set 4-5%/year as the "fair" rate of profit. That this question is left hanging by Piketty (2014) is the reason why it is truly excellent that Matt Rognlie has written this paper, and so brought well-ordered and insightfully-organized data to these questions.
Rognlie focuses on the net rather than the gross capital share. That focus on the net capital share is surely right. I never understood why, in the Solow model, gross savings was supposed to be a function of gross output anyway.
There are the standard big worries over the data. But let me skip over those: I cannot resolve them here, and it will take a great deal of additional thought and work before we can even begin to think of resolving them.
Let me focus, instead, on the big news. As Matt stressed, the big news in the post-World War II net capital share is the surge in housing: from 3% to 8% of private domestic value added. How much of this is a real increase in housing intensity? How much reflects congestion driven by exhaustion of the low-hanging superhighways? How much is rent-extraction via NIMBYism? How much trust do we give to these “imputed rent” imputations anyway? And what that does this mean, anyway?
There has always been a problem with using our GDP estimates as social accounts. In GDP, we measure each unit's contribution to production at the final unit’s marginal cost and each unit's contribution to societal well-being at the final unit's money-metric marginal utility. In the presence of anything like near-satiation in consumption, or of near-exhaustion in productive capacity, this does not convey a true picture.
These are very hard questions. We do not have any very good answers to them.
The fact that the big news since World War II is a rise in housing as a share of value added is striking. It raises the question of whether this surge--the rise of valuable urban housing now--is the only such shift in value-added shares. Was there another significant shift a century and more ago? With the coming of the railroad, the iron-hulled steamship, and the first era of globalization, the value of European farmland and European mine installations crashed under competition from what were then developing resource-abundant economies. How significant was this crash then for aggregate wealth and income distributions? Was it large enough to drive a significant share of the great equalization Piketty sees between 1900 and 1930? We do not know.
Moreover, to the extent that shifts in land values are the drivers of shifts in the capital-output ratio, is this really a problem. It is a problem, or rather a reflection and a consequence of a problem, to the extent that it is driven by NIMBYism. But is it a problem otherwise?
As Matt Rognlie has also also rightly stressed, a secondary piece of big news in his numbers is the pre-1990 fall in the net capital share. This fall is driven by a rise in calculated depreciation rates. And this rise in depreciation is real. Our capital stock today contains many more machines that are rapidly obsoleted by Moore’s Law, and so are not built for durability.
However, this raises a puzzle. The pre-1990 fall in the net capital share is not matched by a decline in the relative capitalization of the corporate sector. Matt points out a steady rise in capitalization up to the late 1960s, followed in the 1970s by a “negative bubble”--truly absurdly high earnings yields on equities--that lasts well into the 1980s. Since the start of the 1990s we see a bubbly rise in the relative capitalization of the corporate sector--a rise that persists in spite of sub-par business-cycle performance. There is thus a strong dissonance between what the production-function and depreciation logic says the value of claims to ownership of capital should be, and what financial markets say the value of claims to ownership are. Once again: these are very hard questions, and we do not have any very good answers to them.
Let me conclude quickly.
I strongly endorse what I take to be Matt Rognlie’s bottom line. The post-WWII variation in the observed net capital share is not explainable via returns on the underlying assets. Instead, the decomposition in Rognlie attributes most of the variation in the factor distribution of income to shifts in markups and pure profits, with accumulation and returns playing, outside of housing, a distinctly secondary role if any role.
It is equally hard to find any role for the race between education and technology. There should be such a role For we do think the factors of production are labor, education skills, machines, and buildings (including residences). Variations in factor supplies should show themselves in factor returns.
Likewise, variation in income inequality is hard to attribute to wealth ownership or to human capital investment or to differential shifts in rewards to factors like raw labor, experience-skills, education-skills, and machines.
Matt thus concludes that: "Concern about inequality should be shifted away from the split between capital and labor, and toward other aspects of distribution, such as the within-labor distribution of income."
The only dissent I wish to make is this: Matt Rognlie is correct now. But this may not still be the case in 50 years, if Piketty is right.
Matt Rognlie’s conclusion is bad news for us economists. It leaves us in the same position as those trying to explain an earlier large puzzle in the production function, the twentieth-century retardation of the British economy. It was Robert Solow who said: "Every discussion among economists of the relatively slow growth of the British economy compared with the Continental economies ends up in a blaze of amateur sociology…" But this time, I really would like for us to be able to do better than we did then.
 These big worries are:
- Worries about depreciation allowances in these accounts. (Mine are perhaps bigger than most.)
- Worries about how much of the value that comes from installing capital comes from (local) learning about how to handle the technology, and is something that does not depreciate from the point of view of the individual firm that is not captured.
- Worries about, from the societal point of view, how much of the value that comes from installing capital comes from global learning about how to handle the technology.
- Perennial worries about what in high-end labor incomes are really incomes earned by raw labor and human capital, and what are rent-extraction and thus sharing in the returns to capital.
Anthony B. Atkinson, Thomas Piketty, and Emmanuel Saez (2011), "Top Incomes in the Long Run of History", Journal of Economic Literature 49:1 (Mar.): pp. 3-71 https://eml.berkeley.edu/~saez/atkinson-piketty-saezJEL10.pdf
Claudia Goldin and Lawrence Katz (2009), The Race Between Education and Technology (Cambridge: Belknap Press: 0674035305) https://amzn.to/1f70tMF
John Maynard Keynes (1936), _The General Theory of Employment, Interest and Money (London: Macmillan) https://amzn.to/1HA7a50
Thomas Piketty (2014), Capital in the Twenty-First Century (Cambridge: Belknap Press: 067443000X) https://amzn.to/1FYDXvf
Emmanuel Saez and Gabriel Zucman (2014), "Wealth Inequality in the United States since 1913 (Berkeley: University of California) https://gabriel-zucman.eu/files/SaezZucman2014Slides.pdf