When Course Evaluation Becomes an Employability Audit, It Misprices Most of What a Degree Does

Bottom line up front: The move to align course evaluation with graduate employability and earnings is a genuine improvement over asking only whether students were satisfied — but it contains a trap. The best economic evidence finds that the non-pecuniary returns to education (better health, civic participation, patience, family and life outcomes) are at least as large as the earnings returns. An evaluation regime that judges courses primarily by their labour-market signal will systematically undervalue — and, through resource decisions, quietly defund — exactly the teaching that produces the other half. The fix is not to abandon employability metrics; it is to stop letting one proximal indicator crowd out everything a course is for.

The reasonable case this piece is arguing with

Start by conceding the strong version of the opposing view, because it is largely right. For too long, course evaluation asked whether students enjoyed the module and stopped there — a satisfaction reading with a weak link to whether anyone learned anything transferable. Reorienting toward outcomes that matter after graduation is a real advance. We have argued for exactly this in pieces on graduate employability and the skills gap and on measuring skill development over student happiness. Employers'' feedback, graduate destinations, and earnings data are legitimate evidence. Nothing below is a plea to retreat into "we cannot measure anything, so let us measure comfort."

The argument is narrower and, we think, more uncomfortable: the employability lens is a partial lens, and partial lenses become distorting lenses the moment they are treated as complete. When graduate salary becomes the headline proxy for course quality — and in several national systems it increasingly is — the metric stops describing value and starts defining it. What it cannot see, it cannot value; what it cannot value, budgets eventually stop funding.

What the earnings number cannot see

The economics here is settled enough to be quotable. In their 2011 Journal of Economic Perspectives review "Priceless: The Nonpecuniary Benefits of Schooling," Philip Oreopoulos and Kjell Salvanes assembled evidence across many outcomes and concluded that the non-pecuniary returns to schooling — effects on health, decision-making, patience, marriage and parenting, job enjoyment, and reduced unemployment risk — are at least as large as the well-studied earnings returns. Walter McMahon''s 2009 book Higher Learning, Greater Good reached a compatible conclusion from a different direction, cataloguing the large private and social benefits of higher education that never appear on a payslip.

This is not soft advocacy; it shows up in hard population statistics. The OECD''s Education at a Glance 2025 reports that around 74% of tertiary-educated adults say they enjoy life all or most of the time, against 61% of adults with below-upper-secondary attainment, and that civic participation rises with education — roughly 10% of tertiary-educated adults had taken part in a public demonstration in the previous year, versus about 6% of those with upper-secondary attainment. Health, trust, volunteering, and mental wellbeing track education in the same direction. Some of this is selection rather than causation, and honest analysis must say so — but the causal core survives the careful studies, and its magnitude rivals the wage effect.

Now ask the operative question: which of those returns does a graduate-earnings metric register? Essentially none. A course that turns hesitant students into confident citizens, that builds the patience and judgement to make better health and financial decisions across a lifetime, that seeds the civic participation a democracy runs on — that course can look mediocre on a salary dashboard, particularly if its graduates enter socially valuable, modestly paid work. Teaching, nursing, social care, public-interest law, the arts, much of the public sector: these are precisely the destinations where the earnings proxy is weakest and the non-market contribution is highest. Judge the course that feeds them by starting salary and you will mark down the teaching society most needs.

But isn''t this just special pleading to dodge accountability?

This is the objection that deserves the most respect, so let us state it at full strength. Every under-performing programme would love an argument that its value is real but unmeasurable — "you cannot put a number on what we do." Appeals to intangible public goods have been used to shield genuinely weak teaching from scrutiny. If "non-market returns" becomes a rhetorical smoke machine, it is worse than the narrow metric it replaces.

Three responses.

First, the non-market returns are not unmeasurable — that is the whole point of citing Oreopoulos and Salvanes, McMahon, and the OECD. Wellbeing, civic participation, health behaviours, self-efficacy, and confidence are measured routinely in large surveys. The claim is not "value is ineffable" but "your current proxy measures the wrong slice of a measurable whole." That is a demand for better measurement, not less.

Second, the argument is explicitly additive, not substitutive. Keep the employability evidence. Keep graduate destinations and, read carefully, earnings. The failure mode is monoculture — letting a single proximal indicator become the target, at which point Goodhart''s law does its work and programmes optimise for the salary signal at the expense of everything it fails to capture. Triangulation is the discipline that keeps any one metric honest — the case we make in Triangulation in Teaching Evaluation.

Third, note the direction of the accountability. Reducing a public institution''s worth to the private earnings of its graduates is itself a contestable political choice dressed as neutral measurement — the marketisation move that reframes education as a private investment good. Insisting the evaluation also see the public and personal returns is not evading accountability; it is refusing to let accountability be quietly redefined as return-on-tuition. As the historian Stefan Collini put it, universities answer to more than the market alone — a "wider social justification" than graduate wages.

What an honest, wide-lens evaluation measures instead

The practical route is to evaluate the proximal mediators of the full return — the capabilities a course can plausibly build and a student can plausibly report on — rather than either satisfaction or a distant salary the course cannot control. That means asking about:

• Confidence, agency, and self-efficacy — the precursors of both employment and civic participation, and the theme of career adaptability.
• Judgement, patience, and decision-making — the durable dispositions Oreopoulos and Salvanes tie to non-market returns.
• Civic and ethical capability — mapped, where useful, onto frameworks like EU LifeComp rather than a single satisfaction number.
• What the course cannot claim. Evaluation should be precise about the gap between what a module builds and what only a labour market decides — the proximal-mediators discipline that stops a survey from over-promising causal reach it does not have.

A worked example makes the stakes concrete. Imagine two versions of the same social-work degree. Programme A drills interview technique and case-management software; its graduates place quickly into agency roles at a predictable starting salary. Programme B spends more of its hours on ethical reasoning, reflective practice, and the judgement to sit with ambiguity; its graduates take a few months longer to place and earn no more at first. On a graduate-earnings dashboard, A dominates and B looks like a laggard to be reformed or cut. Yet the non-market evidence predicts that B's graduates will make better decisions under pressure, burn out less, stay in the profession longer, and contribute more to the communities they serve — none of which the salary proxy can see, and much of which is precisely the public good the training exists to produce. Optimise the evaluation regime for the visible metric and you will, over a few funding cycles, quietly convert every Programme B into a Programme A and call it improvement. That is not a hypothetical risk of narrow metrics; it is their defining tendency once they become targets.

Where Koji fits

An earnings dashboard cannot ask a graduate whether a course changed how they reason, participate, or make decisions. A conversation can. Koji for Education runs AI-moderated conversational interviews that probe capability and agency in students'' own words, and its automatic thematic analysis surfaces the non-market gains — confidence, judgement, civic disposition — that a Likert average or a salary proxy renders invisible. With six structured question types alongside open-ended probing and programme-level reporting, an institution can hold the employability evidence and the wider-return evidence in one instrument, instead of letting the metric that happens to have a currency symbol win by default. The same interview engine powers the main Koji platform for user and customer research, where the identical failure — a single monetisable metric standing in for the whole of value — is just as common.

Employability is a real return, and course evaluation should measure it. It is simply not the only return, and it is not even the larger half. An evaluation system that forgets that is not being rigorous. It is being precise about the wrong thing.