The Muddle of Labor and Capital

What if we took human capital and other forms of intangible capital seriously?

In the movie of economic thought, the point where I walked in was after World War II. When I was in graduate school at MIT in the late 1970s, Paul Samuelson taught a one-quarter course on capital theory, in which he gave us the flavor of 19th-century Austrian thinking on the topic. But otherwise, professors dealt only with the latest theory. A few prominent papers from the 1940s were still fresh enough for the professors to teach. I am hazy on how economists thought about capital before then, although of course I have some awareness of Marxist economics.

The central idea in MIT economics was the neoclassical production function. The idea is that economic output depends on two factors of production: labor and capital. One way to express it is as follows:

output per worker depends on capital per worker

The more capital per worker, the more output per worker. What is capital? The classical concept is machinery that is used in production. If you run a small lawn care business, the capital consists of mowers, leaf blowers, and the truck to carry the equipment to the lawns that you service.

In this example, it seems plausible that output per worker would depend on capital per worker. If the job is to clear leaves, and the workers have no tools to use, the job will take a long time. If you give them rakes, it will go much faster. If you give them leaf blowers, it will go faster still.

But in today’s economy, many jobs are more complicated than clearing leaves. Nowadays, most factory jobs require a lot of training. And of course, technical and professional jobs require extra years of education as well as training.

Economists, following Nobel Laureate Gary Becker, apply the term “human capital” to describe skills acquired by workers that make them more productive. Conceptually, this is useful. But when national statistics are calculated, human capital is not counted as capital. Instead, all of the income of trained factory workers, as well as professional and technical workers, is included in “labor” income.

In the National Income Accounts, capital income is associated with business profits, including the profits earned by the owners of commercial buildings. But this assumes that profits are returns to investment in machinery. But the earnings of Microsoft, Goldman Sachs, Amazon, and other major enterprises do not come from machinery. They come from a combination of engineering creativity, management strategy and execution, and successfully manipulating and navigating the regulatory environment.

One estimate is that as of 2006 Microsoft’s physical assets were only one percent of its market value. Most of Microsoft’s capital is intangible. We have “capitalism without capital,” as Haskel and Westlake put it. See my review of their book.

Because human capital and intangible business capital typically are not measured, we do not really know the ratio of capital per worker in modern economies. As a result, we have no real idea what the true relationship is between output per worker and capital per worker. The problem is exacerbated by the fact that there is so much intangible capital that is embodied in workers (human capital).

This in turn means that we cannot accurately interpret trends in the distribution of income. If the statistics that we use to distinguish labor income from capital income are deeply flawed, then so is the purported analysis that is based on those statistics.

It might make sense to include intangible capital when we measure capital. Moreover, suppose that we count all of “human capital” as capital, so that only the compensation of an unskilled worker is measured as “labor” income. Every increment above that earned by skilled workers and professionals is measured as “capital” income. If an unskilled worker can earn $25,000 a year and a software engineer can earn $200,000, then $175,000 of the software engineer’s income should count as capital income. Thinking this way about the economy as a whole, labor’s share of income is not the 60 percent that you get by putting all of the software engineer’s income (and that of other skilled workers) into “labor” income. Labor’s share would probably be closer to 5 percent.

For consistency, the ratio of “capital” to “labor” would be very high, because the denominator would be unskilled labor and the numerator would be enlarged by all of human capital. Finally, the ratio of output to “labor” also would be extremely high, raising our estimate of labor productivity. In fact, this is probably correct, but it would differ by orders of magnitude from conventional measures.

This essay is part of a series on human interdependence.