From Overqualified to Underused: What the Evidence Actually Shows

This chapter interprets what education and skills mismatch can cost workers, firms, and economies, and asks a narrower question than “Is there a skills crisis?” The useful judgement is conditional: identify which mismatch was measured, compare the right group, and separate an observed association from a proven cause. That discipline makes the evidence more useful, not less urgent.

The stakes are substantial. But a headline number about overqualification, a report of an unfilled vacancy, and an employer’s account of a skill gap do not describe the same harm. They can point to different actors, different time horizons, and different remedies.

What current numbers can—and cannot—tell us

The OECD’s 2026 report, drawing on the Survey of Adult Skills 2023, reports that across the EU 21% of workers are over-qualified and 25% report being over-skilled. It also reports that 10% are under-qualified and 9% report being under-skilled. These figures indicate that formal-education mismatch and worker-reported skills mismatch are both widespread in the population covered by the evidence. They are not interchangeable totals: each uses a different comparison.

The outcome differences reported alongside them are meaningful. Over-qualified workers typically face an 11% wage penalty compared with well-matched peers in the same occupation and industry and are nearly three percentage points less likely to report high life satisfaction. Over-skilled workers earn 2% less than their well-matched counterparts. Under-skilling is associated with a 0.2-percentage-point decrease in life satisfaction. The word “typically” matters, as does the comparator. These are not claims that every over-qualified person loses exactly 11% of pay or that a degree causes lower satisfaction.

A wage penalty is always relative to a reference group. The ILO review found an average overeducation penalty of 13.5% across 61 estimates in the literature it surveyed and an average overskilling penalty of 7.5% across 38 estimates. Those are not an update to the OECD figures or a harmonised trend. They come from an older review of studies using different countries, periods, definitions, and statistical specifications. Their value is that they show a recurring association in a substantial body of research, while their variation shows why the estimate should not be detached from its design.

The reference group can even reverse the apparent meaning of education. An overeducated worker may earn less than a similarly educated worker in a better-matched job, yet more than a less-educated worker doing the same occupation. The first comparison suggests a cost of being unable to find work that uses the worker’s education; the second suggests that additional education may still raise the productivity or value of the work being performed. Neither comparison proves that education is harmful. They answer different counterfactual questions.

This is not a statistical trick. Imagine two workers doing the same broad job, one with a degree and one without. The more educated worker may earn a premium within that occupation because they bring additional knowledge or ability.