Economics 201

Case of the Day: The Education Premium in Wages

It has always been the case that more educated workers earn higher wages and salaries in modern economies. This makes good common sense and good economic sense. Highly educated workers are both valuable to their employers and relatively scarce, assuring that they will command a higher price for their labor. The prospect of raising your future earnings is probably one of the reasons, if not the main reason, that you attend college.

However, the last century has experienced remarkable fluctuations in what economists call the "education premium"—the earnings ratio between more educated workers and those with less education. In this case, we examine movements in the education premium over the 20th century using the evidence compiled by Claudia Goldin and Lawrence Katz in their influential 2008 book The Race Between Education and Technology.

The diagram below (from Goldin and Katz) shows the changes in two education premia in the United States over the 20th century. The higher line with filled dots is the percent gap of the average income of college graduates vs. high-school graduates; the lower line with open dots is the gap between high-school graduates and those finishing eighth grade. (The most reliable data are available only in census years, which is why most data points are ten years apart.)

Wage premia

From 1910 to 1950, the premium declined substantially; the gap between the incomes of college graduates and high-school graduates narrowed from over 60% to barely 30%. Since then, the gap rose from 1950 to 1970, narrowed again until 1980, and has risen rapidly since 1980, which is one of the underlying reasons for the current historically high level of income inequality. Clearly, from these data, the education premium can both rise and fall depending on the balance of economic forces in the labor market.

At its most basic level, Goldin and Katz's analysis is based on the relative supply and demand for educated and uneducated labor. When the supply of educated labor goes up relative to uneducated labor, the education premium falls. When the demand for educated labor increases relative to uneducated labor, the premium rises. So we explain the education premium (sometimes called the skill premium, though this is really a broader concept) in terms of the demand for and supply of the skills that education provides.

Supply of Educated Labor

Economists view education as investment in human capital. As with other investments, families or individuals sacrifice income and consumption during the education period expecting to be rewarded with higher income in later years. Willingness to invest in education depends on the returns they can expect from this investment, so we would expect a higher education premium to lead to more investment in education and, as these more educated individuals enter the labor force, an increase in the supply of educated labor over the following decades. The figure below (also from Goldin and Katz) shows the changes in the rate of return to investing in a college education in the United States over the years. (Think of the rate of return as like an interest rate. If you took the money invested in education and instead bought a financial asset such as a bond, this is the interest rate that would give you the same overall return.) The pattern clearly mirrors that of the education premium above. Most of the variation in the rate of return comes from earnings, not from the cost of investment in education.

Returns to schooling

But also in common with other kinds of investment, education requires the investor to be able to finance both the explicit cost of schooling and the reduced income during the school years. With many kinds of physical investment the investor can obtain loans more easily by posting the investment asset (house, factory, machine, vehicle, etc.) as collateral, promising that the lender can repossess the asset if the investor fails to repay the loan. Human capital cannot be repossessed, so it is much more difficult for the free market to provide funding for education investments. That's why access to financial aid, government-supported student loans, and work study programs have evolved to help students offset these initial investment costs. For other students, including most primary and secondary students, all that is required to finance the educational investment is sufficient family income to allow the children to attend school rather than work.

The simplest way of summarizing the availability of labor with various levels of education is to examine the education distribution within the overall population. This is not a perfect measure because there may be variations in labor-force participation across groups in the population. People who are choosing not to work (for example, retirees) are not part of the labor supply. But the broad pattern of change in the population education distribution is a pretty good indicator of the distribution in the labor force. The diagram below is Figure 1.8 from Goldin and Katz and shows the population shares with various levels of education in the United States from 1915 to 2005.

Education shares

The change in the distribution has been truly remarkable. In a bit less than a century, the share of the population with some college has increased from about 5% to almost 60% and the share with less than a high-school degree plummeted from 88% to about 10%. And the rate of change has not been constant. Increases in college attainment were gradual until 1940 and accelerated after that.

Increases in the supply of educated workers should, other things being equal, depress their relative wages. We can see this to some extent in the data, for example, the decline in the education premium accompanying increases in education levels before 1940 and the lack of growth in the wage premium from 1960 to 1980 at the time when the share of college-educated workers begins to soar. But the premium has spiked upward since 1980, despite the rapid increase in the share of college graduates in the population, which would tend to push the premium downward. Clearly supply factors cannot tell the whole story. Something must have been happening to labor demand.

But before we leave the supply side, we must consider one additional supply phenomenon. The organization of workers into labor unions increased in the early post-World-War-II period, but then began to decline in the 1970s. One of the principal goals of unions is to raise the wages of their employees through collective bargaining, exerting monopoly power in the labor market. If unionization is more common among less-educated workers (which it is), then the decline in unions could explain the rising education premium after 1980. While this is probably a contributing factor, Goldin and Katz (along with most labor economists) do not think this effect is nearly strong enough to explain the large change in the wage structure. For that, we must look at the demand side.

Demand for Educated Labor

If the education premium is rising at the same time that the supply of educated workers is rising, then the relative demand for educated labor must be rising even faster than the supply. To explain the widening of the wage gap since 1980, we must think about the factors that affect the demand for skills and how these might have changed.

Economic theory tells is that the demand for any particular kind of labor depends fundamentally on its marginal revenue product. Since categories of labor with higher marginal products will have correspondingly higher marginal revenue products, they will have higher demand curves that will make firms willing to hire these workers at higher wages. If the demand for skills has increased enough to offset the increased supply, then the MP (and MRP) of educated workers must be rising much faster than the MP (and MRP) of those with less education.

There are two major changes in the U.S. economy since 1980 that are thought to have had major impacts on the demand for skills. Many people (especially populist politicians) blame globalization, which surely plays a role. The advancement of foreign economies and the increase in our trade with them has allowed production processes that rely on unskilled labor to be relocated to places where this labor is cheaper. The reorientation of the U.S. economy away from these production processes has shifted the structure of jobs in the United States away from low-skilled jobs and toward jobs for highly educated workers, which should increase the education premium. Although this is a phenomenon that has highly visible and identifiable effects (for example, the cities in the Rust Belt that experienced large-scale job losses), the evidence suggests that, when averaged over the whole economy, globalization is much less important than the other major cause: technological change.

When technological change affects the economy, the primary effect is that we can produce more output with the same amount of inputs. The production function shifts upward and the isoquants telling us how much of the inputs are required to produce any given amount of output shift down toward the origin. The marginal products of factors of production rise as productivity increases. This makes all of us better off because (assuming initially no change in factor prices) firms' costs fall, their output supply curves shift to the right, and we can get products at lower prices.

But here we must consider a possible secondary aspect of technological change: Is technological change balanced in increasing the productivity of all factors of production or is it "biased" in the direction of some factors whose marginal products rise more than others? There is a lot of evidence that technological change in the computer age has been strongly "skill-biased," raising the marginal products of highly educated workers (programmers, for example) by much more than it raises the productivity of those with less education. This makes intuitive sense. While many white-collar professionals are made ever more productive through advances in computing hardware, software, and communication, manual workers often see machine-controlled tools (operated by a smaller number of skilled workers) taking over their factory jobs. Skill-biased technological change seems to be the dominant factor in explaining why the education premium has risen dramatically at a time when college-educated workers have become ever more abundant.

Is technological change always skill-biased? Economists differ on this question. Goldin and Katz, on whom this case study focuses, argue that even during the earlier years of the 20th century technological advance favored those with more education. This led them to the title of their book: the race between education and technology. Increases in educational attainment tend to drive the wage premium down; advances in technology tend to drive it up. In their view, the outcome of the race between them determines the direction that the premium moves. In the first half of the 20th century, education was racing ahead but in the last half technology has surged back into the lead.

Does the Education Premium Affect Quantities Supplied and Demanded?

So far we have treated changes in the supply and demand for educated labor as "exogenous," arising from changes outside the labor market. But shouldn't the education premium itself affect the amount supplied and demanded just as the price of any commodity is expected to move buyers and sellers along their demand and supply curves?

Basic economic theory tells us that increases in the education premium will raise the rate of return to education, which should induce more people to become educated: the supply curve should slope upward. Similarly, firms may be able to direct their research and development toward new technologies that allow them to hire less of the relatively expensive input: replacing educated workers when the premium is high and replacing manual workers when it is low.

But both of these responses to relative labor costs will occur over many years or decades. The educational decisions of your generation will be reflected in the composition of the labor force during your entire working lives, for perhaps 40 or 50 years. Similarly, the education levels of today's working population are the result of education decisions made over the last 40 to 50 years. It takes a long time for overall population educational attainment to adjust to changes in incentives.

Similarly, research and development stimulated by today's relative wages will lead to new technologies that will be put into production well into the future. Implementation times vary greatly across technologies, but it is not unusual for the R&D process to take ten years, then for the new technology to take another decade or even much more to diffuse through the economy as leading-edge firms are eventually followed by technological laggards. (It is also important to note that there are many other factors influencing technological advance; firms do not perform R&D exclusively in response to wage differentials.)

So even if today's very high education premium stimulates massive R&D to replace high-priced educated workers, and even if it causes an ever-greater proportion of young people to get college degrees, it will likely take decades for these responses to have their full moderating effect on the education premium.

Will We All Be Replaced by Machines?

What about the long-anticipated advances that are beginning to occur in artificial intelligence? Will these advances someday allow machines to replace college-educated workers? In a 2013 study, Oxford economists Carl Benedikt Frey and Michael Osborne estimate the probability that 702 different jobs are "computerizable" in the coming decades. According to their calculations, many white-collar professions are very much at risk.

The probabilities range from 0.0028 for recreational therapists to 0.99 for telemarketers. (A quick look at the list will reinforce the value of critical thinking in the future job market.) The results are, of course, controversial (are economists really 43% likely to be replaced?), but these questions are very important to individuals contemplating how they will fare in the job market of the mid-21st century.


1. For each of these historical events, discuss how it should affect the subsequent supply of educated labor and whether it is consistent with the pattern of educational attainment shown in the third diagram above (Goldin and Katz's Figure 1.8):

a. The Great Depression of the 1930s, when family incomes were very low.

b. The G.I. Bill, offering tuition support for World War II (and other) veterans after 1945.

c. The fall in the education premium leading up to 1940.

d. The beginning of Pell Grants in 1965.

e. The Vietnam War of the late 1960s and early 1970s, in which educational deferments allowed many men to avoid being drafted.

2. Not everything you learn in college is equally valuable in the labor market. What college majors would you expect to lead to higher incomes and what majors would be less well rewarded by the market? Why?

3. Evaluate the following two (related) statements:

a. "A larger education premium will cause more people to become educated. A smaller education premium will cause fewer people to become educated. In the long run, supply and demand should reach an equilibrium in which the lifetime value of the education premium matches the opportunity cost of being educated."

b. "In long-run equilibrium, the rate of return on education should match the rate of return on other similarly risky assets in the economy."

Answer the questions on the Moodle site