Debates about the pay gap concentrate on what happens at work: promotion, negotiation, caring responsibilities, discrimination. All of these matter.
A substantial part of the eventual difference, however, is determined before anyone enters employment, through decisions made in education and in the transition out of it.
Component one: field of study
Decompositions of the graduate pay gap consistently find that field of study is the single largest identifiable component.
The differences between fields are large. Analyses of graduate earnings by subject in several countries find gaps between the highest and lowest earning fields that exceed the entire gender pay gap by a wide margin.
The distribution of women and men across those fields is uneven and has changed only partially. Engineering, computing and physical sciences remain male-majority in most systems; education, social work, psychology and allied health remain female-majority.
This is not a claim that anyone chose badly. It is the observation that the earnings consequence of the choice is rarely presented at the time it is made.
Component two: the subject choice that precedes it
Field of study at university is constrained by subject choice at fifteen or sixteen, which is where the physics and computing gaps discussed elsewhere on this site originate.
The chain runs: subject choice constrains degree options, degree options constrain field, field is the largest component of the graduate gap.
Each link is well documented. What is rarely stated is the whole chain, which means the financial consequence of a decision made at fifteen is not visible to the person making it.
Component three: the first offer
Starting salary differences within the same field are small but consequential, for two reasons.
Raises are typically percentages, so an initial difference persists proportionally through a career.
And subsequent employers frequently anchor on prior salary. Studies of jurisdictions banning salary history questions have found associations with increased pay for women changing jobs, consistent with the anchoring mechanism.
A modest first-offer difference, compounded and anchored, does not remain modest.
Component four: unpaid entry work
In sectors where entry runs through unpaid internships, the requirement filters by family wealth and delays the start of paid employment.
Since several of the affected sectors are heavily female, this operates as a specific drag on early earnings in exactly the fields where earnings are already lower.
What this changes about the argument
The common framing of the pay gap as either pure discrimination or pure choice is unhelpful, and this decomposition explains why.
Like-for-like pay differences — same role, same employer, same experience — are real but small in most rigorous studies, usually a few percentage points.
The larger part is sorting: who ends up in which field. That is not discrimination in the direct sense, and calling it choice is also inadequate, because the choices are made at fifteen, under informational conditions that nobody would defend, with the consequences not disclosed.
A decision made in ignorance of its main consequence is a decision in a limited sense only.
The valuation question underneath
There is a further layer that the sorting explanation does not address.
Why does engineering pay more than teaching? Some of the answer is supply and demand. Some of it is not.
Occupational research has documented a pattern in which occupations lose relative pay as they become female-dominated, over periods where the work itself did not change. Studies of occupational transitions have found this effect in several sectors.
If that is right, then field of study is not a fully independent variable. The pay associated with a field is partly a function of who does it, which means advising individuals to choose differently addresses only half of a circular system.
What is actually actionable
For a student: know the earnings data for fields before choosing. It is published in most countries and is not part of standard careers advice. This is not an instruction to choose for money; it is an instruction to know what is being traded.
Keep the option open longer. The single most consequential act is not dropping the subjects that gate the high-earning fields, because reopening that door later is difficult or impossible.
Treat the first offer as a compounding decision.
For a school: state prerequisites explicitly at the point of subject choice. Many students genuinely do not know what physics is required for, and telling them is free.
For anyone arguing about the gap: be specific about which component is under discussion. The like-for-like gap, the sorting gap and the valuation question are three different arguments with three different remedies, and conflating them is why the debate makes so little progress.