What We Can Responsibly Say About Underemployment

Underemployment among US workers with bachelor’s degrees is a real labor-market concern, but the size and meaning of the problem cannot be separated from the way it is measured. The final question is not which published percentage sounds most convincing; it is what each percentage establishes about the relationship between graduates and the occupations where they work. This series supports a disciplined judgment: compare the classifier, population, timing, observed variable, and blind spots before treating an estimate as evidence about policy or about an individual career.

The title’s claim, made precise

To say that underemployment depends on the definition is not to say that underemployment is imaginary. It means that “using a degree” contains several questions that national data observe only indirectly.

One question concerns entry access. How much education do workers typically need to enter an occupation? The BLS-based approach answers that question by comparing a worker’s bachelor’s degree with the occupation’s typical entry-level education assignment. A graduate in an occupation assigned below the bachelor’s level is counted as underemployed under this rule.

A second question concerns realized hiring. What education do the people currently working in an occupation actually have? A realized-matches approach uses that distribution. It may classify an occupation as bachelor’s-degree level when graduates are a majority, or when they are the plurality—the largest single education group—even if they are not more than half.

A third question concerns economic return. Do bachelor’s-degree workers in the occupation earn more than workers with lower levels of education? The earnings approach treats a premium as evidence that some graduates may be contributing valued capabilities or occupying roles that the entry assignment does not reveal.

These are not three names for the same measurement. They are three ways of describing the graduate–occupation relationship. Entry access is about the usual route into a broad occupational category. Realized hiring is about the education composition that has emerged among workers. Earnings are about a group-level economic outcome. Each lens can add information while leaving a different part of the relationship unobserved.

The distinction also explains why the report can describe underemployment as a meaningful social problem while resisting a single settled rate. If graduates’ education is genuinely underused, individuals may lose earnings and the economy may misallocate human capital. But if a broad entry label treats every graduate in an occupation as equally mismatched, the estimate can exaggerate the problem by ignoring specialized roles, employer preferences, and the returns graduates receive within that occupation. The dispute is therefore not between concern and indifference. It is about which evidence should carry the conclusion.

One comparison model for three lenses

The easiest way to compare the approaches is to hold the worker and occupation in place and change only the question.

The table is useful only if its final column stays attached. A reader who sees that the realized-matches estimate is lower than the BLS estimate should not conclude that the first method was simply wrong. The two methods use different definitions of an occupation’s education boundary. A reader who sees a graduate earnings premium should not conclude that every graduate in the occupation is productively matched. The premium is a signal about economic value, not a complete account of work.

The same worker can cross different classification boundaries without changing jobs. Under the BLS-based rule, a bachelor’s-degree worker in an occupation assigned to an associate’s degree or less is counted as underemployed. Under a plurality rule, that occupation may be treated as bachelor’s-degree level if graduates are the largest group. Under an earnings adjustment, the worker may be reclassified if their earnings reach the specified premium. The worker’s situation has not become three different jobs. The analyst has asked three different questions of it.

That is why the definition is not a cosmetic choice. It determines whether the decisive evidence is an expert judgment about entry, an observed distribution of workers, or an earnings relationship. It also determines which people are counted and which are left outside the category. Changing the definition changes the estimated boundary of the problem.

What the reported numbers actually say

The report’s own estimates illustrate the difference. Using the BLS-based approach, 39 percent of prime-age workers ages 25–54 with bachelor’s degrees are underemployed when all workers are included, and 37 percent are underemployed among full-time, full-year workers. For workers ages 22–23, the comparable figures are 47 percent and 43 percent. The younger group approximates people within one year of graduation; the older group represents prime-age workers. The all-worker and full-time/full-year populations also answer different questions.

For prime-age, full-time, full-year workers, the realized-matches versions produce lower estimates: 32 percent when graduates must form a majority of the occupation and 24 percent when they need only be the plurality. Those results show that the occupation’s observed education distribution classifies some workers differently from its BLS entry assignment. They do not prove that the plurality threshold is the correct standard. They show what happens when “degree-level occupation” means largest education group rather than more than half of workers.

The earnings-adjusted model produces another result.