Why Annual Earnings Can Misread Business Careers
An annual-earnings profile does not measure only the price an employer pays for an hour of skill. This chapter uses the business-major pattern to examine how graduate education and labor-market attachment enter the outcome, and why the same distinction matters when interpreting engineering and computer-science profiles. The practical judgment is to separate a change in annual earnings from a change in wage growth before calling either one human-capital appreciation or decay.
Annual earnings bundle several pathways
The study’s dependent variable is log yearly earnings. That is a useful measure of what a worker receives over a year, but it is a composite. It can rise because pay for work rises, because a person works more quarters, because they work more intensely while employed, or because they move into a different employment pattern. It can fall for the reverse reasons. A profile built from annual totals therefore describes a combination of wage and labor-supply outcomes unless the analysis separates those components.
This matters for the business comparison because the initial cross-sectional result looks like a story about skill. Business graduates begin with an earnings advantage over the broad other-major group, but their advantage narrows with age. If one assumes that the annual total is a clean measure of the market value of accumulated knowledge, the natural explanation is that business skills appreciate more slowly or depreciate sooner. But that explanation competes with a simpler possibility: business graduates may work more consistently at the beginning of their careers, while people in other fields increase their labor-market attachment later.
Graduate education creates a related mixture. The undergraduate major is recorded, but the earnings profile may include years after a worker earns a master’s, professional degree, or doctorate. If people with the same undergraduate major have different probabilities of obtaining advanced degrees, and advanced-degree holders have different age profiles, a single major-by-age line combines several paths. The model can include a general indicator for having a graduate degree and still miss the possibility that the difference associated with that degree changes with age.
The paper addresses that issue by allowing age profiles to differ by advanced-degree status. The distribution is not uniform across major groups: about 54.5 percent of life-and-physical-science workers have an advanced degree, compared with 35.9 percent of engineering-and-computer-science workers, 22.3 percent of business workers, and 36.3 percent of the broad other group. Those shares are descriptive, not a measure of what a degree causes. They show why a common undergraduate-major slope can be misleading when the post-college education pathways attached to the major are different.
The adjustment has its largest conceptual effect on the life-and-physical-science comparison, because advanced degrees are especially common in that group and advanced-degree holders tend to have steeper age profiles. Business moves in the other direction: business majors are less likely than the comparison group to obtain a graduate degree, so giving advanced-degree holders their own age profiles makes the business estimate less negative.