Credential inflation
A Socratic walk-through of credential inflation — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why do employers start demanding degrees for jobs that were done well for decades without one?
A job specification for an administrator, a technician, a junior manager now says "degree required". The same job, twenty years ago, said nothing of the kind, and was done competently by people who had none. Nobody in the hiring chain believes the tasks now demand three years of study.
The tempting explanation is that jobs got harder, and for many of these roles it fails on its face. The second — employers are being snobbish — explains nothing, because snobbery does not have a start date. So what changed, and why does the requirement only ever move one way?
Reasoning it through
REASONING #Start with the employer's actual problem, which is not education but ignorance. They cannot observe whether an applicant will be good at the job; they can observe things correlated with it. A degree is one such thing, with a property that matters enormously in practice: it is free to check. A line on a form filters a stack of applications at zero cost.
Now ask what adding that filter does to the pool. Put illustrative numbers on it. A hundred people apply and thirty could do the job well; sixty hold a degree; the degree is genuinely but weakly predictive, with forty per cent of holders capable against fifteen per cent of the rest. Check the arithmetic — 0.40 x 60 = 24 capable graduates, 0.15 x 40 = 6 capable non-graduates, 24 + 6 = 30, which matches. Screening on the degree lifts the hit rate in the interview pile from 30% to 40%. It also discards six capable applicants, a fifth of every capable person who applied.
Both are true at once, and that is the whole engine. The improvement is captured by the employer; the loss is borne by six people who never learn why. A rule can be individually sensible and collectively wasteful without anyone behaving badly.
But this still does not explain movement: a weak predictor was weak twenty years ago too. So add the thing that changed — the share of the population holding degrees rose substantially across the rich world. What does that do to the signal?
Two things, pulling the same way. First, the degree separates less, because there are fewer non-holders left to separate from. Second, and less obvious, the absence becomes more informative — when most people have something, not having it is unusual, and unusual things invite an inference. The filter that once said "unusually educated" comes to say "unusual", which is a different and harsher claim.
As the signal weakens, an employer wanting the same discrimination must raise the bar: a degree where none was asked, a master's where a degree was. Applicants, seeing it rise, acquire more. Each side responds rationally to the other, and the requirement ratchets while the job stands still. Thurow's job-competition framing puts it sharply: if education fixes your position in the queue rather than your productivity, everyone acquiring more leaves the order unchanged and only raises the entry price.
That makes a testable prediction, and it is what distinguishes this story from the ordinary one. If credentials add skill, jobs newly requiring degrees should have changed content and should pay more. If credentials sort, the job pays what it paid and only the ticket got dearer.
The analogy
THE ANALOGY #Think of a crowd at a parade in which one person stands on a box to see. It works, so the person behind gets a box, then the next, until everyone stands on one and everyone sees exactly what they saw before, at the cost of a box each. Nobody erred: for each individual at the moment of deciding, the box was worth it and still is, because stepping off now is worse than never having had one.
a box gives nothing but height, whereas a degree plausibly does teach something, so the education is not pure deadweight the way the boxes are. And boxes do not sort people — the parade captures the arms race but not the screening, which is why the employer wanted the boxes there at all.
Clarifying the model
THE MODEL #Three refinements, then the honest weakness.
First, this is a screening account, not a claim that education is worthless. The strong version — degrees teach nothing and signal everything — is not supportable, and most economists hold a mixed view. The evidence usually cited for a signalling component is the sheepskin effect: earnings jump on completing a qualification, beyond what the years of study predict, which is hard to explain if the value lay wholly in the learning. That argues for a signalling component, not a pure one.
Second, this belongs beside grade inflation but runs from the opposite end. Grade inflation is senders — teachers, institutions — devaluing a signal each has a local reason to strengthen. Credential inflation is receivers — employers — raising the level of a signal each has a local reason to demand. One erodes the currency; the other raises the price. Both ratchet because unilateral restraint is punished: an employer who drops the requirement simply takes a weaker pile of applications.
Third, the falsification. The load-bearing claim is that the requirement rose without the work changing, so the refuting observation is available: find roles where a degree requirement was added and check whether the tasks, the tools, or the real wage moved with it. If those jobs turn out to have been substantially re-specified, rising skill demand is doing the work and this account is redundant for them.
A natural experiment is running now, since several large employers and governments have publicly removed degree requirements in recent years. My recollection of the follow-up analyses is that the announcements moved the stated requirement far more than the share actually hired without degrees — which, if it holds, suggests the line was partly a formality standing in for a screen operating through other channels. I quote no figures.
The deepest caveat is that separating signalling from skill has resisted decades of effort: the two predict similar things in most data, and the studies pulling them apart rely on assumptions themselves argued over.
A picture of it
THE PICTURE #How to readRead left to right through the illustrative hundred applicants. The first split is the filter; the second shows what each side of it contained. The employer's gain is the ribbon into "Capable" — twenty-four of the sixty interviewed, a better rate than the thirty-in-a-hundred they started with. Their cost is the thin ribbon into "Capable but lost": six people, a fifth of all capable applicants, discarded by a rule that was still, on the employer's own terms, an improvement. The two ribbons that matter belong to different people, which is why the filter is neither a mistake nor cheap.
What became clearer
WHAT CLEARED #Credential inflation is not employers becoming fussier. It is a cheap screen losing power as the thing it screens on becomes common, and the only available response — raising the bar — being one applicants must then match, which lowers the power again. The job never enters the loop. It is self-sustaining because every step is correct for whoever takes it: the filter really does improve the employer's pile, the extra year really does protect the applicant's place in the queue, and the cost lands on people outside the transaction.
Where to go next
ONWARD #- Why removing a degree requirement from a job advertisement changes hiring so much less than expected.
- What screening devices employers reach for when the credential stops discriminating, and whether they sort more fairly.
Key terms
TERMS #| Term | What it means |
|---|---|
| Screening | using an observable characteristic to infer an unobservable one, here a qualification standing in for likely job performance. |
| Sheepskin effect | the earnings premium attached to completing a qualification, over and above the premium for the years of study themselves. |
| Job competition model | Thurow's framing in which education fixes a worker's position in a queue for jobs rather than their productivity in one; the rival account, human capital, holds that education raises earnings by raising what a worker can do. |
Every term the collection defines is gathered in the glossary.