The Breakthrough Penalty: Why Women's Most Original Patents Are the Ones the System Loses

Women inventors and men inventors have nearly identical grant rates on conventional patents. On the most unconventional patents (the kind that combine ideas in surprising ways), the grant rate gap is 14.9 percentage points.

That finding comes from Sowrirajan, Whalen, and Uzzi's 2026 paper in PNAS, the largest published study to date on what drives the US patent gender gap. The headline finding is not that women face universal bias. It is something narrower and more useful: the gap exists almost entirely on a specific kind of invention, the kind the patent system was designed to reward most.

This page walks the study's methodology, the mechanism behind the gap (which is not what most readers expect), the dollar value of what's being lost, and what the finding implies for patent policy. The mechanism is institutional. It is also closeable.

This article presents published research data for educational purposes. It is not legal, policy, or investment advice. Patent gender statistics depend on name-based inference and methodology choices: see How we measure.

The study

Sowrirajan, T., Whalen, R., & Uzzi, B. "The institutional dynamics of inequality for women inventors who break with conventional thinking." Proceedings of the National Academy of Sciences 123(16): e2500343123, April 17, 2026. DOI: 10.1073/pnas.2500343123.

• Sample: 6,185,556 USPTO applications, 2001-2018.
• Replication: 280,128 Canadian and 224,365 UK applications, with the same interaction effect confirmed at P<0.001.
• Methodology: "Unconventionality" of each application is measured by the z-score of CPC code-pair co-occurrence versus random expectation — adapting the Uzzi et al. (2013) approach from scientific publications to patent classifications. Inventor gender is inferred from given names using WGND.

The headline finding: the gap lives on unconventional patents

Across all 6.2 million applications, the authors found:

• No meaningful gender gap on conventional patents. The patent system grants conventional applications at similar rates regardless of inventor gender composition.
• A 14.9 percentage point gap on highly unconventional patents. Men-led teams and women-led teams diverge sharply when the application combines patent classification codes in surprising ways.
• Women-led teams file fewer unconventional applications to begin with. Male-majority teams file 31.8% unconventional applications; women-majority teams file 25.9%. The lower filing rate partly reflects the lower grant rate that follows.

The most extreme version of the gap appears in paired-examiner data:

The mechanism: examiner experience, not examiner bias

Here is the part of the study that surprised the authors and that frames the finding correctly.

The gap is not driven by male examiners discriminating against women applicants. When examiner experience is held constant, men and women examiners produce substantially similar grant outcomes. The bias-only story does not survive the data: the assignment pattern itself does most of the work, with women inventor shares and women examiner shares correlating strongly within CPC classifications (r² = 0.46 unweighted, 0.80 inventor-volume-weighted).

Two structural facts drive the gap instead:

1. Women inventors are 16.9% overassigned to women examiners, relative to random expectation. This is a property of how the USPTO's assignment system has historically routed applications.
2. Women examiners are systematically less experienced, because women examiners have higher historical dropout rates from the USPTO. Less experienced examiners reject unconventional applications at higher rates than experienced examiners do, regardless of who they are or who they are reviewing.

The mechanism, in one sentence: women applicants get matched to women examiners more often than random; women examiners are on average less experienced; less-experienced examiners are tougher on unconventional inventions; women's unconventional applications therefore disproportionately get rejected.

Why inexperience penalizes unconventionality

Unconventional patent applications combine technology areas in ways the examiner has not seen before. They typically require more analysis, more cross-disciplinary expertise, and more familiarity with the relevant prior art across multiple fields.

Experienced examiners are better at distinguishing genuinely novel and non-obvious combinations from combinations that fail Section 103 obviousness tests. Inexperienced examiners default to skepticism because the cost of erroneously granting a bad patent is more visible than the cost of erroneously rejecting a good one. The result is that unconventional applications are systematically over-rejected by inexperienced examiners, and women applicants disproportionately encounter inexperienced examiners.

The paper provides direct supporting evidence: inexperienced examiners are overturned significantly more often than experienced examiners when their rejections are appealed, particularly for unconventional patents. The rejections are wrong more often, but most rejections are never appealed (see Where women's patents disappear on the appeal-rate gap), so the wrong rejections stick.

What the lost patents are worth

The authors estimated the value of the unconventional patent applications by women inventors that would have been granted at equal rates:

• Approximately 2,238 additional patents would have issued to women inventors over the 2001-2018 study period under equal grant rates.
• The estimated value of those lost patents exceeds $234 million in 1992 USD. (The authors use a real-dollar reference year for cross-period comparability.)

Both figures should be read as lower bounds. They count only direct lost-patent value, not downstream effects: lost licensing revenue, foregone commercialization opportunities, missing collaborators who would have built on the rejected work, and the chilling effect on subsequent unconventional filings by women inventors who have seen the prior rejections.

The finding replicates in Canada and the UK

The interaction effect between inventor gender and patent unconventionality was replicated in two additional jurisdictions:

• Canada: 280,128 patent applications, same gender × unconventionality interaction, statistically significant at P<0.001.
• UK: 224,365 patent applications, same interaction, statistically significant at P<0.001.

The cross-country replication strengthens the finding considerably. The gap is not unique to the USPTO; it appears in any patent system that combines variable-experience examiners with non-random assignment of applications to those examiners.

Why this is the breakthrough penalty (and what the authors call it)

The "breakthrough penalty" framing is ours. The paper's own terminology is "innovation gap for unconventional inventions." Both phrases describe the same finding.

We call it the breakthrough penalty for one reason: unconventional patent applications are, on average, the breakthroughs. Patents that combine ideas across fields are systematically more disruptive on standard disruption indices. They are the ones that earn the most forward citations, get licensed at the highest rates, and define the technological frontiers of their decades.

When the patent system disproportionately rejects the unconventional applications from women inventors, it is rejecting precisely the most valuable category of women's inventive output. The cumulative effect over decades is to systematically thin women's representation among the inventions that define their fields.

What this implies for policy

The authors offer three actionable implications:

1. Train junior examiners to evaluate unconventional inventions. The over-rejection effect is concentrated in inexperienced examiners. Training, mentorship, and structured rubrics for evaluating cross-disciplinary combinations could close part of the gap.
2. Equalize examiner experience across inventor gender. If women applicants are 16.9% overassigned to women examiners, and women examiners are systematically less experienced, the assignment algorithm itself is a leverage point. Either change the assignment to be experience-balanced, or invest in retention so that examiner experience distributions equalize over time.
3. Redesign assignment algorithms. The non-random match between inventor gender and examiner gender is a property of the USPTO's current system. It is technically possible to design assignment to be gender-neutral on the examiner-experience dimension. The paper does not propose a specific algorithm but identifies the design problem.

The Pairolero et al. (AEJ:EP, 2025) randomized controlled trial showed that examiner-side interventions can shift grant rates by 12+ percentage points. The PNAS 2026 study identifies a more specific intervention target: not "all examiner training" but "training inexperienced examiners on unconventional inventions, especially when paired with women inventors."

The takeaway

The patent gender gap in the United States is not, on this evidence, a universal phenomenon driven by bias against women. It is a structurally specific phenomenon: institutional dynamics that route women's most original work to the parts of the examination corps least equipped to evaluate it.

That is, in one sense, encouraging. Universal-bias problems are hard to close because they require changing many minds. Institutional-assignment problems are closeable because they require changing one assignment algorithm and one training program. The leverage points are concrete.

It is also a more urgent framing. As long as the assignment dynamics persist, the women inventors most likely to define the technological frontiers of the next decade are the ones most likely to be filtered out by the system designed to recognize them.

Sources

More from the patent gender gap series