Capital in the Twenty-First Century

by Thomas Piketty

  The point I want to emphasize, however, is that historical reality is more complex than the idea of a completely stable capital-labor split suggests. The Cobb-Douglas hypothesis is sometimes a good approximation for certain subperiods or sectors and, in any case, is a useful point of departure for further reflection. But this hypothesis does not satisfactorily explain the diversity of the historical patterns we observe over the long, short, or medium run, as the data I have collected show.

  Furthermore, there is nothing really surprising about this, given that economists had very little historical data to go on when Cobb and Douglas first proposed their hypothesis. In their original article, published in 1928, these two American economists used data about US manufacturing in the period 1899–1922, which did indeed show a certain stability in the share of income going to profits.18 This idea appears to have been first introduced by the British economist Arthur Bowley, who in 1920 published an important book on the distribution of British national income in the period 1880–1913 whose primary conclusion was that the capital-labor split remained relatively stable during this period.19 Clearly, however, the periods analyzed by these authors were relatively short: in particular, they did not try to compare their results with estimates from the early nineteenth century (much less the eighteenth).

  As noted, moreover, these questions aroused very strong political tensions in the late nineteenth and early twentieth centuries, as well as throughout the Cold War, that were not conducive to a calm consideration of the facts. Both conservative and liberal economists were keen to show that growth benefited everyone and thus were very attached to the idea that the capital-labor split was perfectly stable, even if believing this sometimes meant neglecting data or periods that suggested an increase in the share of income going to capital. By the same token, Marxist economists liked to show that capital’s share was always increasing while wages stagnated, even if believing this sometimes required twisting the data. In 1899, Eduard Bernstein, who had the temerity to argue that wages were increasing and the working class had much to gain from collaborating with the existing regime (he was even prepared to become vice president of the Reichstag), was roundly outvoted at the congress of the German Social Democratic Party in Hanover. In 1937, the young German historian and economist Jürgen Kuczynski, who later became a well-known professor of economic history at Humboldt University in East Berlin and who in 1960–1972 published a monumental thirty-eight-volume universal history of wages, attacked Bowley and other bourgeois economists. Kuczynski argued that labor’s share of national income had decreased steadily from the advent of industrial capitalism until the 1930s. This was true for the first half—indeed, the first two-thirds—of the nineteenth century but wrong for the entire period.20 In the years that followed, controversy raged in the pages of academic journals. In 1939, in Economic History Review, where calmer debates where the norm, Frederick Brown unequivocally backed Bowley, whom he characterized as a “great scholar” and “serious statistician,” whereas Kuczynski in his view was nothing more than a “manipulator,” a charge that was wide of the mark.21 Also in 1939, Keynes took the side of the bourgeois economists, calling the stability of the capital-labor split “one of the best-established regularities in all of economic science.” This assertion was hasty to say the least, since Keynes was essentially relying on data from British manufacturing industry in the 1920s, which were insufficient to establish a universal regularity.22

  In textbooks published in the period 1950–1970 (and indeed as late as 1990), a stable capital-labor split is generally presented as an uncontroversial fact, but unfortunately the period to which this supposed law applies is not always clearly specified. Most authors are content to use data going back no further than 1950, avoiding comparison with the interwar period or the early twentieth century, much less with the eighteenth and nineteenth centuries. From the 1990s on, however, numerous studies mention a significant increase in the share of national income in the rich countries going to profits and capital after 1970, along with the concomitant decrease in the share going to wages and labor. The universal stability thesis thus began to be questioned, and in the 2000s several official reports published by the Organisation for Economic Cooperation and Development (OECD) and International Monetary Fund (IMF) took note of the phenomenon (a sign that the question was being taken seriously).23

  The novelty of this study is that it is to my knowledge the first attempt to place the question of the capital-labor split and the recent increase of capital’s share of national income in a broader historical context by focusing on the evolution of the capital/income ratio from the eighteenth century until now. The exercise admittedly has its limits, in view of the imperfections of the available historical sources, but I believe that it gives us a better view of the major issues and puts the question in a whole new light.

  Capital-Labor Substitution in the Twenty-First Century: An Elasticity Greater Than One

  I begin by examining the inadequacy of the Cobb-Douglas model for studying evolutions over the very long run. Over a very long period of time, the elasticity of substitution between capital and labor seems to have been greater than one: an increase in the capital/income ratio β seems to have led to a slight increase in α, capital’s share of national income, and vice versa. Intuitively, this corresponds to a situation in which there are many different uses for capital in the long run. Indeed, the observed historical evolutions suggest that it is always possible—up to a certain point, at least—to find new and useful things to do with capital: for example, new ways of building and equipping houses (think of solar panels on rooftops or digital lighting controls), ever more sophisticated robots and other electronic devices, and medical technologies requiring larger and larger capital investments. One need not imagine a fully robotized economy in which capital would reproduce itself (corresponding to an infinite elasticity of substitution) to appreciate the many uses of capital in a diversified advanced economy in which the elasticity of substitution is greater than one.

  It is obviously quite difficult to predict how much greater than one the elasticity of substitution of capital for labor will be in the twenty-first century. On the basis of historical data, one can estimate an elasticity between 1.3 and 1.6.24 But not only is this estimate uncertain and imprecise. More than that, there is no reason why the technologies of the future should exhibit the same elasticity as those of the past. The only thing that appears to be relatively well established is that the tendency for the capital/income ratio β to rise, as has been observed in the rich countries in recent decades and might spread to other countries around the world if growth (and especially demographic growth) slows in the twenty-first century, may well be accompanied by a durable increase in capital’s share of national income, α. To be sure, it is likely that the return on capital, r, will decrease as β increases. But on the basis of historical experience, the most likely outcome is that the volume effect will outweigh the price effect, which means that the accumulation effect will outweigh the decrease in the return on capital.

  Indeed, the available data indicate that capital’s share of income increased in most rich countries between 1970 and 2010 to the extent that the capital/income ratio increased (see Figure 6.5). Note, however, that this upward trend is consistent not only with an elasticity of substitution greater than one but also with an increase in capital’s bargaining power vis-à-vis labor over the past few decades, which have seen increased mobility of capital and heightened competition between states eager to attract investments. It is likely that the two effects have reinforced each other in recent years, and it is also possible that this will continue to be the case in the future. In any event, it is important to point out that no self-corrective mechanism exists to prevent a steady increase of the capital/income ratio, β, together with a steady rise in capital’s share of national income, α.

  FIGURE 6.5. The capital share in rich countries, 1975–2010

  Capital income absorbs between 15 percent and 25 percen
t of national income in rich countries in 1970, and between 25 percent and 30 percent in 2000–2010.

  Sources and series: see piketty.pse.ens.fr/capital21c

  Traditional Agricultural Societies: An Elasticity Less Than One

  I have just shown that an important characteristic of contemporary economies is the existence of many opportunities to substitute capital for labor. It is interesting that this was not at all the case in traditional economies based on agriculture, where capital existed mainly in the form of land. The available historical data suggest very clearly that the elasticity of substitution was significantly less than one in traditional agricultural societies. In particular, this is the only way to explain why, in the eighteenth and nineteenth centuries, the value of land in the United States, as measured by the capital/income ratio and land rents, was much lower than in Europe, even though land was much more plentiful in the New World.

  This is perfectly logical, moreover: if capital is to serve as a ready substitute for labor, then it must exist in different forms. For any given form of capital (such as farmland in the case in point), it is inevitable that beyond a certain point, the price effect will outweigh the volume effect. If a few hundred individuals have an entire continent at their disposal, then it stands to reason that the price of land and land rents will fall to near-zero levels. There is no better illustration of the maxim “Too much capital kills the return on capital” than the relative value of land and land rents in the New World and the Old.

  Is Human Capital Illusory?

  The time has come to turn to a very important question: Has the apparently growing importance of human capital over the course of history been an illusion? Let me rephrase the question in more precise terms. Many people believe that what characterizes the process of development and economic growth is the increased importance of human labor, skill, and know-how in the production process. Although this hypothesis is not always formulated in explicit terms, one reasonable interpretation would be that technology has changed in such a way that the labor factor now plays a greater role.25 Indeed, it seems plausible to interpret in this way the decrease in capital’s share of income over the very long run, from 35–40 percent in 1800–1810 to 25–30 percent in 2000–2010, with a corresponding increase in labor’s share from 60–65 percent to 70–75 percent. Labor’s share increased simply because labor became more important in the production process. Thus it was the growing power of human capital that made it possible to decrease the share of income going to land, buildings, and financial capital.

  If this interpretation is correct, then the transformation to which it points was indeed quite significant. Caution is in order, however. For one thing, as noted earlier, we do not have sufficient perspective at this point in history to reach an adequate judgment about the very long-run evolution of capital’s share of income. It is quite possible that capital’s share will increase in coming decades to the level it reached at the beginning of the nineteenth century. This may happen even if the structural form of technology—and the relative importance of capital and labor—does not change (although the relative bargaining power of labor and capital may change) or if technology changes only slightly (which seems to me the more plausible alternative) yet the increase in the capital/income ratio drives capital’s share of income toward or perhaps beyond historic peaks because the long-run elasticity of substitution of capital for labor is apparently greater than one. This is perhaps the most important lesson of this study thus far: modern technology still uses a great deal of capital, and even more important, because capital has many uses, one can accumulate enormous amounts of it without reducing its return to zero. Under these conditions, there is no reason why capital’s share must decrease over the very long run, even if technology changes in a way that is relatively favorable to labor.

  A second reason for caution is the following. The probable long-run decrease in capital’s share of national income from 35–40 percent to 25–30 percent is, I think, quite plausible and surely significant but does not amount to a change of civilization. Clearly, skill levels have increased markedly over the past two centuries. But the stock of industrial, financial, and real estate capital has also increased enormously. Some people think that capital has lost its importance and that we have magically gone from a civilization based on capital, inheritance, and kinship to one based on human capital and talent. Fat-cat stockholders have supposedly been replaced by talented managers thanks solely to changes in technology. I will come back to this question in Part Three when I turn to the study of individual inequalities in the distribution of income and wealth: a correct answer at this stage is impossible. But I have already shown enough to warn against such mindless optimism: capital has not disappeared for the simple reason that it is still useful—hardly less useful than in the era of Balzac and Austen, perhaps—and may well remain so in the future.

  Medium-Term Changes in the Capital-Labor Split

  I have just shown that the Cobb-Douglas hypothesis of a completely stable capital-labor split cannot give a totally satisfactory explanation of the long-term evolution of the capital-labor split. The same can be said, perhaps even more strongly, about short- and medium-term evolutions, which can in some cases extend over fairly long periods, particularly as seen by contemporary witnesses to these changes.

  The most important case, which I discussed briefly in the Introduction, is no doubt the increase in capital’s share of income during the early phases of the Industrial Revolution, from 1800 to 1860. In Britain, for which we have the most complete data, the available historical studies, in particular those of Robert Allen (who gave the name “Engels’ pause” to the long stagnation of wages), suggest that capital’s share increased by something like 10 percent of national income, from 35–40 percent in the late eighteenth and early nineteenth centuries to around 45–50 percent in the middle of the nineteenth century, when Marx wrote The Communist Manifesto and set to work on Capital. The sources also suggest that this increase was roughly compensated by a comparable decrease in capital’s share in the period 1870–1900, followed by a slight increase between 1900 and 1910, so that in the end the capital share was probably not very different around the turn of the twentieth century from what it was during the French Revolution and Napoleonic era (see Figure 6.1). We can therefore speak of a “medium-term” movement rather than a durable long-term trend. Nevertheless, this transfer of 10 percent of national income to capital during the first half of the nineteenth century was by no means negligible: to put it in concrete terms, the lion’s share of economic growth in this period went to profits, while wages—objectively miserable—stagnated. According to Allen, the main explanation for this was the exodus of labor from the countryside and into the cities, together with technological changes that increased the productivity of capital (reflected by a structural change in the production function)—the caprices of technology, in short.26

  Available historical data for France suggest a similar chronology. In particular, all the sources indicate a serious stagnation of wages in the period 1810–1850 despite robust industrial growth. The data collected by Jean Bouvier and François Furet from the books of leading French industrial firms confirm this chronology: the share of profits increased until 1860, then decreased from 1870 to 1900, and rose again between 1900 and 1910.27

  The data we have for the eighteenth century and the period of the French Revolution also suggest an increase in the share of income going to land rent in the decades preceding the revolution (which seems consistent with Arthur Young’s observations about the misery of French peasants),28 and substantial wage increases between 1789 and 1815 (which can conceivably be explained by the redistribution of land and the mobilization of labor to meet the needs of military conflict).29 When the lower classes of the Restoration and July Monarchy looked back on the revolutionary period and the Napoleonic era, they accordingly remembered good times.

  To remind ourselves that these short- and medium-term changes in the capital-lab
or split occur at many different times, I have shown the annual evolution in France from 1900 to 2010 in Figures 6.6–8, in which I distinguish the evolution of the wage-profit split in value added by firms from the evolution of the share of rent in national income.30 Note, in particular, that the wage-profit split has gone through three distinct phases since World War II, with a sharp rise in profits from 1945 to 1968 followed by a very pronounced drop in the share of profits from 1968 to 1983 and then a very rapid rise after 1983 leading to stabilization in the early 1990s. I will have more to say about this highly political chronology in subsequent chapters, where I will discuss the dynamics of income inequality. Note the steady rise of the share of national income going to rent since 1945, which implies that the share going to capital overall continued to increase between 1990 and 2010, despite the stabilization of the profit share.

  FIGURE 6.6. The profit share in the value added of corporations in France, 1900–2010

  The share of gross profits in gross value added of corporations rose from 25 percent in 1982 to 33 percent in 2010; the share of net profits in net value added rose from 12 percent to 20 percent.

  Sources and series: see piketty.pse.ens.fr/capital21c.

  FIGURE 6.7. The share of housing rent in national income in France, 1900–2010

  The share of housing rent (rental value of dwellings) rose from 2 percent of national income in 1948 to 10 percent in 2010.

  Sources and series: see piketty.pse.ens.fr/capital21c.