Who earns more, and why: two new studies test the academic gender gap

A new study published on 20 July 2026 in the Proceedings of the National Academy of Sciences goes back to a familiar number, the gender pay gap in academia, and asks a less familiar question: what actually explains it. The hard part, as the authors frame it, is not the size of the gap but its persistence in institutions where formal equality policies have been on the books for decades.

The headline finding cuts against a popular narrative. After adjusting for a stack of measurable variables, research output, citations, hours worked, grants won, career length, and field, the gap shrinks dramatically. Most of what looks like discrimination, in this dataset, has a productivity-shaped explanation sitting underneath it. That is an awkward conclusion for a sector that has spent fifteen years treating structural bias as the default frame.

What the numbers actually say

Gender pay gaps in academia are well documented. The harder question is why they persist, and the PNAS paper takes a deliberate swing at every leading candidate answer in turn. The researchers tested the usual suspects, outright wage discrimination, differences in seniority, mothers taking longer to climb the ladder, the "motherhood penalty," and the work-intensity gap between men and women in research-active roles.

The result, in summary: the variables most often invoked by HR-style explanations account for only a fraction of the residual gap once you hold output constant. Men in the sample publish more, win more grant funding, and accumulate more citations, and those three variables alone do most of the statistical lifting. The gap that remains after all the controls is small, in some specifications statistically indistinguishable from zero.

This does not mean bias has vanished. It means the contribution of bias, in this dataset, looks harder to detect than the contribution of differing research intensity. The two are not the same thing, and the paper is careful about not collapsing them.

Why the productivity line is contested

The reading-down-to-productivity has its own pushback. Critics argue that output itself is shaped by invisible advantages, better lab space, more generous start-up packages, lighter teaching loads, faster promotion reviews, and old-boy networks that route collaborations and named-chair opportunities. If men's higher productivity is itself partly downstream of bias, then "controlling for productivity" is not neutral; it is baking the bias back into the model under a different name.

There is a second objection. The study, like most PNAS-scale meta-analyses, leans heavily on bibliometrics: how many papers, how many citations, which journals. None of those capture teaching quality, mentorship loads, or committee work, which fall disproportionately on women in many departments and which the academic labour market rewards invisibly if at all. The productivity that gets counted is not the only productivity that exists.

A third objection is structural. If women publish fewer papers because they are doing more of the work the institution does not count, then the right answer is to count that work and pay for it, not to use the citation gap as an excuse for the salary gap.

The structural backdrop, in plain terms

What sits underneath these results is a sector that has been quietly bifurcating for two decades. Research-intensive universities have pushed towards a publish-or-perish model in which career progression is auditable on paper-count and grant-count metrics. Teaching-intensive institutions have moved the other way, with heavier classroom loads and less room for the visible output that bibliometrics reward. The pay gap is largest, and most-studied, in the research-intensive tier.

Two quiet shifts are reshaping what the gap even measures. First, the post-pandemic hiring freeze at many Western universities has thinned the junior ranks, which is where the gap opens widest. Second, the share of total academic pay tied to external grants has grown, because base budgets have not kept pace with research costs. The people best placed to win those grants are, on average, men in the dataset, and the gap follows.

A fair reading is that the academic labour market is doing what its incentives tell it to do. An unfair reading is that the incentives were built on assumptions about who does what kind of work, and those assumptions have not been audited.

Stakes for the next bargaining round

The paper lands at an awkward moment. Several European unions are renegotiating collective agreements in research-intensive universities, and the gender pay gap is one of the named items on the table. If the structural reading of the new PNAS data holds, the actionable lever is not "stop discriminating at the offer stage", which most equity offices now claim to police, but "redesign the productivity metric so it counts the work that has been invisible."

That is harder, and slower, than drafting another equality plan. It also faces the institutional resistance of every department that likes its current ranking system. The next two years of negotiations will be a stress test of whether universities want to follow the data or pick the parts of it they find convenient.

On a different desk in the same journal week: a separate study from Heinrich Heine University Düsseldorf and the universities of Cologne and Frankfurt/Main shows how bee colonies divide labour without any central planner. The researchers tracked which tasks individual bees performed and read out the neural activity underlying the choice. The model that emerges is striking in its ordinariness: each bee reacts to local cues, what is needed nearby, what its body is wired for, and the colony-level pattern falls out for free. It is a useful counter-image to the PNAS finding. Insect nervous systems, working with far less information than a hiring committee, produce a division of labour that no one could call inequitable. The cost is they have no-one to sue.