Roughly 3,800 years ago, an administrator at a textile workshop in Elephantine, on Egypt's southern frontier, recorded a month's grain payments to 28 adult women and 11 girls by name. The women received between 50 and 57 heqat of grain. The girls received between 3 and 7 heqat, scaled to age and experience. Nobody negotiated these figures. They were set by a fixed, role-based pay schedule, the same administrative logic used to pay the male tomb-builders working across the Nile at Deir el-Medina.
Ancient Egypt gets invoked often in conversations about gender equality, usually for the same handful of facts: women could own property independently, initiate divorce, and occasionally rule as pharaoh. All of that is true and unusual for the period. It is also not really the most useful fact this civilisation left behind for anyone trying to understand today's gender pay gap. The more useful fact is administrative rather than legal. It sits in payroll records rather than court documents. It says something sharper than “ancient Egyptian women had rights.”
That detail matters more than it first appears. It means one of the oldest documented pay systems in human history was, in its mechanics, already gender-neutral. The system did not care who stood in front of it. What it cared about was the role, which raises an uncomfortable question for anyone working in reward today: if role-based pay-setting has been solvable for millennia, why does a gender pay gap still exist almost everywhere on earth?
The answer, then and now, has very little to do with how fairly a single role gets paid. It has everything to do with which roles exist, who gets to hold them, and how the value of those roles gets decided in the first place. None of what follows required anyone to sit down and decide that women's work was worth less, which is exactly what makes the pattern worth taking seriously.
A fair formula, an unfair map
Deir el-Medina is the best-documented workplace of the ancient world. For roughly 450 years, a state-employed crew of artisans built and decorated royal tombs in the Valley of the Kings and left behind an extraordinary paper trail of ostraca recording exactly what they were paid. Ordinary workmen received a fixed monthly grain allotment. Foremen and scribes, sitting above them in the hierarchy, received noticeably more. Pay scaled with rank, not with who asked for what.
That workforce was almost entirely male. The state issued the ration to the male workman; his household, wife and children included, shared in that single allocation rather than each earning one independently. Much of the physical labour of turning grain into food, the grinding, the baking, and the brewing, was carried out by servant and enslaved women attached to those households, work that does not appear as an independent line in the wage record at all.
The most complete pay data we have from this period, then, describes a job category, tomb-building, that women were simply not part of. That is a different problem from unequal pay for equal work. It is occupational segregation. It produces the same downstream effect: men, as a group, held the roles that carried the highest fixed rations, largely because those roles required literacy.
Literacy itself was tightly rationed. Only around one percent of the population could read and write. Formal scribal schools took in the sons of scribes, priests and officials and passed the profession down through male family lines almost exclusively. Evidence for individual literate women exists, mostly among royal and elite households. It is rare enough that Egyptologists still debate how to interpret it. The credential that opened the door to the best-paid tier of Egyptian society was, structurally, closed to almost every woman in the country before she ever applied for a role.
Where women did hold formally paid positions, the record shows the same non-negotiated logic applying to them without adjustment. The Elephantine workshop is one example. Weaving more broadly was one of the largest industries in the Egyptian economy. It was staffed predominantly by women, organised in workshops that supplied temples, estates and the royal household with a commodity as economically vital as grain itself.
Even so, weaving does not carry the same status in the historical record as scribal or administrative work. Bob Brier and Hoyt Hobbs, summarising the broader pattern of gender relations in ancient Egypt, described women as equal to men in almost every respect except occupation itself.
That is the throughline. Ancient Egypt's ration system was, by design, blind to gender. Its occupational structure was not. Literacy gated access to the highest-paid tier; literacy access was itself gendered. The system never had to discriminate, because the map of who could hold which role had already done the work for it.
The bug survived the operating system
It would be comforting to file this under ancient history and move on. We no longer ration grain by rank; we no longer restrict literacy along gender lines; and Egyptian women, notably, held stronger independent legal and property rights than women in many societies thousands of years later. Progress happened. What did not go away, however, is the specific mechanism: occupational segregation quietly deciding whose work counts as valuable. It just got a professional coat of paint.
Modern job evaluation exists precisely to replace ad hoc pay decisions with an objective, role-based assessment method, the same instinct that produced the Deir el-Medina ration tables. Most schemes score jobs against a set of compensable factors, typically some version of skills, effort, responsibility and working conditions, then use the resulting points to build a pay structure. In principle, this should be immune to the ancient failure mode. A measurement model does not know or care about the gender of the person in the seat.
In practice, the factors themselves can carry the bias forward. Many point-factor schemes developed during the twentieth century, particularly those originating in industrial settings, placed greater emphasis on physical effort, equipment responsibility and environmental hazards than on interpersonal or emotional demands.
Roles built around sustained interpersonal demand, emotional regulation, conflict management, multitasking under time pressure and care for other people's wellbeing were often captured only indirectly or given comparatively little weight, rather than left out of the scheme altogether. That is the daily reality of nursing, social care, early years education, and much of HR and customer-facing work. That is not because the work was easier. It is because many factor sets were originally built with different kinds of demand in mind.
Picture two roles scored on a legacy scheme built this way: a warehouse operative lifting stock and operating machinery, and a family support worker managing distressed clients under constant safeguarding risk and emotional strain. The first maps cleanly onto physical effort and equipment, both explicit, well-weighted factors. The second gets partial credit at best, because that kind of demand tends to be folded into a broader factor and scored well below its actual weight in the role. Neither role is objectively harder. The instrument simply is not built to register one of them as clearly as the other.
The result is a modern echo of the Elephantine problem. The grading structure can be internally consistent and completely rule-bound. It can still produce a lower grade for a role that happens to sit in a female-dominated sector. That is not because anyone decided women should earn less. It is because the demands that role depends on are rarely weighted to reflect how central they actually are.
What the fix looks like this time
This is where the current wave of pay transparency legislation earns its place in the conversation, because it is the first time this specific, narrow failure mode has been written into binding law rather than left to good intentions.
Article 4(4) of the EU Pay Transparency Directive, Directive (EU) 2023/970, requires employers to assess whether workers are in a comparable situation regarding the value of their work using objective, gender-neutral criteria, built around skills, effort, responsibility and working conditions, applied consistently and without direct or indirect sex-based discrimination. Buried inside that requirement is a single, deliberate instruction: relevant soft skills must not be undervalued.
That instruction is doing a great deal of work. It is not a general aspiration toward fairness. It is a direct legislative response to a specific pattern: interpersonal, communication and care-based skills have historically been treated as personality traits rather than compensable job requirements, while physical and technical skills have been treated as measurable and therefore valuable.
The EU Institute for Gender Equality's accompanying job evaluation guidance is explicit about the mechanics of this risk. Where weights on factors and sub-factors are self-selected by the evaluator, it warns, there is a real risk of unconsciously assigning more weight to something like physical strength than to psychosocial or interpersonal skill, producing a gender-biased outcome from a process that looks entirely objective on paper.
In practice, testing a job evaluation scheme against this standard starts with the factor plan itself. Each factor and sub-factor needs a genuine mechanism for scoring interpersonal, emotional and care-based demands as rigorously as it scores physical or technical ones, rather than folding that kind of demand into a generic catch-all worth a handful of points regardless of how central it is to the role. It also means checking where the points actually land once a scheme is applied to real roles, not just how balanced the factor plan looks on paper, since a scheme can read as fair in the document and still produce skewed grades in practice.
The deadline for EU member states to transpose this directive into domestic law was 7 June 2026, a date that has now passed. Only a small number of countries, Slovakia, Italy and Malta prominent among them, had comprehensive local legislation in force by that point, with most of the rest, including several of the largest economies in the bloc, running months or years behind. The European Commission has confirmed there will be no extension.
What that means in practice is that the substance of Article 4(4) is no longer a future obligation to plan around. It is a current one. Most employers have not yet been formally tested against it, which leaves a narrow window worth using rather than waiting out.
The pattern, stated plainly
Line the two periods up and the shape is the same. In Deir el-Medina, a fair pay schedule sat on top of a workforce that had already been sorted by gender before pay was ever calculated. The highest-value tier required a credential, literacy, that ran along gender lines.
In a great many job evaluation schemes still in use today, a fair-looking evaluation framework sits on top of a set of compensable factors that were built, decades ago, around one kind of work and one kind of worker. The highest-value tier still tends to require exposure to factors such as physical effort, equipment responsibility and budget accountability, all of which cluster in male-dominated roles by historical accident rather than genuine necessity.
Neither case required anyone to decide, explicitly, that women's work was worth less. That is precisely why the problem has been so durable. It does not live in a negotiation, where it would be visible and contestable.
It lives in the design of the measuring instrument itself: which demands get their own factor, which get downgraded to a minor sub-factor, and which get folded into a vague catch-all that nobody scores highly. A biased weighting scheme is far harder to see and far harder to challenge than a manager simply offering a woman less money for the same job title.
What has changed, finally, is that the law now names this specific mechanism and requires employers to test for it. That is a meaningfully different starting point from any point in the last 3,800 years of documented pay history.
It will not fix itself through good intentions any more than it did in the New Kingdom. It requires someone to open the job evaluation scheme, look directly at what each factor and sub-factor is actually measuring, and ask the question Deir el-Medina's scribes never had reason to ask: does this instrument know how to see the work that is mostly done by women?
For anyone responsible for reward, job evaluation or pay equity right now, that question is no longer optional. It is, for the first time, the law.
