“The stability of the proportion of the national dividend accruing to labor […] is one of the most surprising, yet best-established, facts in the whole range of economic statistics [….] Indeed, […] the result remains a bit of a miracle.” J. M. Keynes, “Relative Movements of Real Wages and Profit,” The Economic Journal, March 1939
“It was known for some time that the share of wages and the share of profits in the national income has shown a remarkable constancy in “developed” capitalist economies.” Nicholas Kaldor, “A Model of Economic Growth,” The Economic Journal, December 1957.
“The object of this paper is to suggest that, like most miracles, [the stability of labor’s share] may be an optical illusion.” Robert M. Solow, “A Skeptical Note on the Constancy of Relative Shares,” The American Economic Review, September 1958.
Economists build models around simple facts to isolate what drives behavior. In macroeconomics, perhaps the most famous of these facts has been the observed stability of the shares of income paid to labor and to capital. At least since Kaldor wrote 60 years ago, this pattern of income distribution has been at the top of the list of regularities to be explained by theories of economic growth.
Well, it turns out that what was stable for much of the 20th century looks as if it is unstable in the 21st. For at least the past 15 years, and possibly for several decades, labor’s share of national income has been declining and capital’s share has been rising in most advanced and many emerging economies.
Why should we care? Well, one possibility is that, if we understand the decline in labor’s share, it will help us to understand recent increases in income inequality. That is, since the ownership of capital is highly concentrated, it would naturally follow that if capital income rises, the distribution of income would become less equal. While this is surely the case, it turns out that a big part of the increase of inequality can be tied to the rise in labor income of the very highest ranked―the top 1 percent, as it were.
Another reason we could be concerned about movements in national income shares is their impact on the inferences we make from growth accounting exercises. Growth accounting is the primary tool used to measure how much changes in capital, labor, and technology contribute to economic growth. Assessing the contributions requires estimates of the various shares. Again, based on a simple simulation, it turns out that the impact of the decline thus far in labor’s share (up to six percentage points depending on the measure used) on this accounting is quantitatively small (see here).
Other reasons to care are even more subtle. For example, depending on how easily firms can shift the relative quantities of capital and labor used in output, a shift of tax policy (say, from taxing labor income to taxing consumption) could boost both capital formation and labor income (see Lawrence). Separately, because capital faces lower taxes than labor, a decline in labor’s share will naturally lead to a fall in government revenue.
Ultimately, however, it is important that we understand why labor’s share has declined; and whether that decline will continue. The answer could influence a range of policies, from education and training to taxation and transfers. In what follows, we describe what we know about the evolution of labor’s share (including key measurement issues) and highlight several explanations that have recently been proposed for the observed decline.
We start with a simple definition: labor’s share equals the ratio of labor payments to total income or output. This is the product of wages times the quantity of labor input divided by the price of output times the quantity of output. In practice, the Bureau of Labor Statistics’ headline measure of labor’s share examines the payments to labor out of nonfarm business product (see black line in chart). (Nonfarm business accounts for about three-fourths of total gross domestic product.) From about 1970 to 2000, the BLS measure declines very gradually, before turning sharply lower. Since 2010, the share has averaged 57.2 percent, down nearly 6 percentage points from the 1947 to 2000 average of 63 percent.
United States: Measures of Labor’s Share, 1947-1Q 2017
Perhaps surprisingly, labor income (the numerator in the standard measure) is difficult to gauge, not least because reporting by the self-employed tends to understate income. The key measurement problem is that payments to sole proprietors include both the wage payments for their time and the returns to their capital investment. Consequently, any estimate of labor’s share must take a view on the proportion of proprietors’ payments that reflects their work. If this fraction is changing, it is easy to get the shares wrong. After a detailed analysis, Elsby, Hobijn, and Sahin conclude that one third of the apparent decline in the headline measure reflects “the methods employed by the BLS to impute the labor income of the self-employed.” After adjustment, they estimate that the average labor share dropped by a bit more than 4 percentage points between 1948-87 and 2010-12. (We note that Giandrea and Sprague provide counterevidence in a recent BLS Monthly Labor Review article.)
For some purposes, such as growth accounting, using nonfarm business product appears superior to using GDP. The most important reason is that nonfarm business output is measured at market prices, whereas substantial portions of GDP (like the government sector) are measured at cost. Nevertheless, as a check, we can look at a measure based on the income version of the GDP measure. Using gross domestic income as the denominator, and following Gomme and Rupert, we assign a proportion of proprietors’ income to labor consistent with each quarter’s unambiguous payments to labor (compensation of employees) and to capital (profits, rents and interest). The result is a more stable labor share in the period through 2000 (see red line in the chart above), but a decline in the average share between the 1947-2000 and 2010-2017 periods of 3 percentage points. So, while the precise timing and size of the decline may be difficult to ascertain, the fact that there was one is not in dispute.
As Solow highlighted in 1958, the aggregate labor share conceals enormous sectoral heterogeneity, both in levels and in changes. The next chart, based on Table 1 in the Giandrea and Sprague article, displays recent employee-only labor shares for 14 sectors of nonfarm business. As of 2014, the latest available data, this measure ranged from 22 percent in mining to 74 percent in professional and business services, with the aggregate nonfarm business share at 45 percent. Furthermore, of these 14 sectors for which we have data, 8 showed a decline in labor’s share since 1997 (resulting in a dot to the left of the 45-degree line), 5 showed an increase, and 1 did not change. The red dots highlight the sectors with the declines greater than 5 percentage points, while the blue dot denotes the entire nonfarm business sector. On a weighted basis, the declines in durable and nondurable goods sectors’ labor shares more than accounted for the decline in the nonfarm business share, and were partly offset by the rising share in professional and business services.
Employee-only labor shares by sector of nonfarm business, 1997 vs. 2014
The fall in the labor share in the United States is part of a global phenomenon. According to the Penn World Table, in 95 of 130 countries, the latest reported labor share was below the average that prevailed until 2000 (the data begin as early as 1950 for some countries, but start later for others). Countries with declining labor share appear as dots to the left of the 45-degree line in the following chart. (The red dot is the United States.) On average, the latest observation (usually 2014) is 2.8 percentage points lower than the pre-2000 average.
Labor Shares Across Countries, 2014 vs. pre-2000 average
Does the global decline in labor’s share in recent decades reflect a general shift of activity away from industries or sectors that have higher labor shares? The answer generally is no. Conventional analysis of the data shows that most of the decline occurs within-industry or within-sector, rather than resulting from a compositional shift of activity. This finding is common to studies of the United States alone and to cross-country analyses. (The IMF’s April 2017 World Economic Outlook mentions China, with its dramatic shift away from agriculture, as a notable exception.)
So, after decades of apparent stability, what has been causing the recent decline of the aggregate labor share? Will the decline continue? And how is it affecting welfare? A truly compelling explanation (and prediction) must be consistent with the timing of the measured decline (especially the larger drop in the past two decades), with its geographic breadth (including both advanced and emerging economies), with its sectoral and geographic diversity, and with the within-industry pattern observed. This is a very tall challenge to which researchers are beginning to respond.
Chapter 3 of the IMF’s April 2017 WEO explores a range of possible global drivers, including technological progress and economic integration (in markets for goods, services, and capital). Technological advances that lower the relative price of investment goods encourage the substitution of capital for labor. If the substitutability is sufficiently high, then labor’s share will decline. (In 1932, Hicks famously showed that this occurs if the elasticity of substitution between labor and capital exceeds one. For recent—notably divergent—estimates of this key elasticity, see here and here.)
In the case of international trade, theory tells us that increased cross-border activity will raise labor’s shares in countries where labor is abundant, and lower it in countries where labor is scarce. Yet, labor’s share is falling in both advanced and emerging economies. One possible explanation highlighted by the WEO is the deepening of global value chains, which usually involve offshoring from advanced economies of what is locally labor-intensive production. If it is difficult to substitute between labor and capital, and if the offshored tasks are more capital intensive than the norm in the emerging economy, then labor’s share will decline everywhere―and it has.
How might one account for the predominance of within-industry changes in the decline of the labor share? Recent research by Autor et al (here and here) emphasizes the heterogeneity of firms and the role of imperfect competition in explaining the differences. One possibility is that “superstar” firms arise, leading to “winner-take-most” concentration and increased price mark-ups over marginal cost. (Google and Facebook immediately come to mind, but the research shows a much broader pattern beyond the tech sector.) This hypothesis is consistent with two related facts: in recent decades, sales concentration has increased in the private sector in the United States; and the more concentrated the industry, the more labor’s share has fallen. A similarly negative relationship is evident in other advanced economies. In addition, at least in the U.S. manufacturing sector (where data are available), concentration rose more in those industries exhibiting greater increases in productivity, consistent with the “superstar” notion.
So, where does this all leave us? Keynes’ “miracle” and Solow’s “optical illusion” of constant labor and capital shares have given way at every level: for firms, for industries, for sectors, for countries, and for the world. But the race for a better understanding of what drives the income distribution is just getting started, so we still have a great deal to learn about what determines labor’s share. How important are measurement issues? What are the linkages to trade, technology and inequality? What are the policy implications? In the end, some analytic tools built on the assumption that labor’s share is constant likely will continue to be useful. But even that will depend on the outcome of the ongoing debate about the sources and implications of the labor share declines that we are clearly seeing.