Grade inflation
A Socratic walk-through of grade inflation — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why do grades drift upward decade after decade without students getting any better?
Grades exist to distinguish. If a top grade is awarded to nearly everyone, it has stopped doing the job it was invented for — and yet the drift upward continues, decade after decade, in systems that all know this and all say they do not want it. Nobody sat down and decided that an A should mean what a B used to. So who is choosing this, and why does knowing it is happening not stop it?
Reasoning it through
REASONING #Start with what a grade actually is. It is not a measurement in the way a length is; it is a signal sent to people who cannot observe the student directly — employers, admissions committees, funders. And a signal's value comes entirely from what it lets a reader distinguish. A grade that everyone receives carries no information, however accurate it may be about the work.
Now ask what happens to a signal when the sender benefits from sending a favourable one. Consider a single lecturer deciding a borderline mark. Awarding the higher grade produces a satisfied student, fewer appeals, better teaching evaluations — which in many places affect promotion — and costs the lecturer essentially nothing. The cost of that decision is a fractional loss of meaning in a signal used by thousands of people the lecturer will never meet. What should we expect a rational, well-meaning person to do, repeatedly?
Now scale it up. A department's results affect its recruitment. A university's degree classifications affect its league-table position and its graduates' prospects, which affects applications, which affects income. So the same asymmetry holds at every level: the gain from grading a little more generously is captured locally and immediately, and the loss is spread thinly across everyone who reads grades, in the future, invisibly.
That structure has a name. It is a collective action problem — individually rational, collectively self-defeating — and it has the signature property that no participant needs to be cynical for the outcome to occur. It is also self-accelerating in a way worth noticing: once other institutions have moved, holding standards firm actively harms your own students, since their marks are read against everyone else's. Being the only honest sender is a penalty, not a virtue. That is why the drift is one-way, and why unilateral correction almost never survives.
But I should stop and be honest, because "it is all incentive-driven inflation" is an overclaim, and the measurement question is genuinely contested. Grades could rise for real reasons. Student populations change — selectivity at many institutions rose over the same period, so a given cohort may genuinely be stronger. Teaching improved in places, and study resources multiplied. Assessment changed shape: more coursework, more resubmission, more explicit marking criteria, all of which legitimately raise marks by making the target clearer. Course choice shifts between disciplines that grade differently. And the raw statistics are awkward to interpret, since a rising average can come from the composition of who is enrolled rather than from any change in standards.
So how would anyone tell the difference? The honest answer is: only partially. The most persuasive attempts hold the observable characteristics of students constant and ask whether the remaining rise is explained — work of this kind in England, examining the sharp increase in first-class degrees, found a substantial portion of the increase not accounted for by student characteristics or prior attainment. That is evidence, not proof, and it is exactly where the argument sits: some of the rise is real, some is not, and the split is disputed rather than settled.
The analogy
THE ANALOGY #Grades work like currency. Each individual issuer has an incentive to print a little more, since the new notes buy today's goods at today's prices, while the loss of value lands on everyone holding the currency later. Nobody intends the devaluation, everybody contributes to it, and the more of it there is, the more foolish it becomes to be the one issuer holding back.
money has a central bank that can restrain issuance, whereas grading authority is distributed across thousands of independent institutions and individual markers with no one able to bind the others — and unlike prices, grades hit a ceiling, so the signal does not simply reprice, it compresses until it cannot distinguish at all.
Clarifying the model
THE MODEL #Three refinements.
First, the ceiling is what makes this different from ordinary inflation, and it explains the observed consequence. When most marks are near the top, readers of the signal must look elsewhere — to the institution's name, to internal ranking, to interviews, to tests set by the employer. The information does not vanish; it migrates to proxies that are frequently less fair than the grades were, since institutional prestige tracks family background more closely than coursework does.
Second, the mechanism does not require anyone to lower standards deliberately, and mostly nobody does. It requires only that every marginal judgement be made under a slight asymmetry of consequences, thousands of times a year, over decades. Slow drift with no decision point is exactly what such a structure produces.
Third, a caution against a single-cause story. Real improvements, compositional change, assessment reform, consumer pressure from fee-paying students, evaluation-linked promotion, and rankings all plausibly contribute, and their relative weights are not established. What is well documented is the direction, its persistence across many systems, and the compression at the top.
A picture of it
THE PICTURE #How to readEach point is a party to the decision, placed by how much it gains from grading generously and how much of the resulting loss of meaning it escapes. The top-right quadrant is where the drift is generated: large gain, no cost borne. Read down the left edge for the parties who pay — employers and later cohorts — and note that they have no vote in the decision at all. The cohort sitting low but far right is the tell-tale of a collective action problem: it gains this year and is devalued by the same behaviour later.
What became clearer
WHAT CLEARED #A grade is worth something only while it is scarce, and every party in a position to award one gains from awarding it generously while bearing almost none of the resulting loss. That asymmetry, repeated at every level and compounded by the fact that holding out penalises your own students, produces a one-way drift that nobody chose. The complication is that not all of the rise is inflation — some reflects genuine changes in who is being taught and how — and separating those is the part that remains genuinely contested.
Where to go next
ONWARD #- What replaces grades as a filter once they compress, and whether the replacements are fairer.
- Why external examinations and centrally moderated marking drift less, and what that costs.
Key terms
TERMS #| Term | What it means |
|---|---|
| Signal | an observable indicator whose value depends on its ability to distinguish; a grade awarded to everyone conveys nothing regardless of its accuracy. |
| Collective action problem | a situation where each participant's individually rational choice produces an outcome worse for all, with no participant able to fix it alone. |
| Grade compression | the bunching of awarded grades near the top of the scale, which removes the ability to distinguish rather than merely shifting the average. |
| Signal erosion | the loss of informational value in a credential as it becomes more widely and uniformly awarded. |
Every term the collection defines is gathered in the glossary.