Closing Strategies That Work: Evidence-Based Interventions

The patent gender gap has a large literature and a much smaller body of tested interventions. This page is about the second.

Four interventions have published quantitative evidence of impact. They share a structural pattern: each targets access to expert help at a specific stage of the patent system. None is a campaign or an awareness initiative. All of them produce measurable percentage-point movement in the relevant outcome metric.

This page walks each in detail with the measured effect, the mechanism, and what the evidence does and does not generalize to. It is a playbook for TTOs, corporate IP teams, and policy makers. It is also a useful test for distinguishing interventions with evidence from interventions that sound good.

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 four interventions at a glance

These four interventions all show measurable signal, but they sit at different evidentiary levels. We mark each with the strongest claim the published evidence actually supports.

Reading the evidence column: Causal (RCT) means a randomized trial established the intervention caused the effect. Institutional case study means an organization made a change and measured the result, but no counterfactual control was held. Accountability / transparency means the mechanism is voluntary disclosure rather than a measured intervention; the WIR numbers reported are real but causality to any specific action is not established.

Each one operates at a different stage of the patent system. Each one targets a different population. None is a turn-key fix. Together, they form a recognizable pattern about what works.

1. Washington University WIT: redesign the disclosure interface

What it is

The Women in Innovation & Technology (WIT) program at Washington University in St. Louis was founded in 2014 by Nichole Mercier in the Office of Technology Management. The program targeted women faculty in biomedical research. The motivating observation was a baseline mismatch: roughly 30% of relevant biomedical faculty were women, but fewer than 4.5% had ever participated in an invention disclosure.

The intervention

• Pre-populated disclosure forms. Instead of asking PIs to nominate inventors, the form pre-listed all lab members. Inventors were marked as default-included unless explicitly removed.
• Proactive outreach. The OTM team reached out to women faculty directly rather than waiting for disclosures.
• Women-led mentorship. Senior women researchers and IP attorneys were paired with faculty as patenting mentors.
• Data publication. The program publicly tracked and reported women's filing and grant numbers.

The effect

Between 2011-13 and 2013-16:

• Patents filed on behalf of women faculty rose 129%.
• Women faculty interactions with the Office of Technology Management rose 27%.

Generalizes to

University tech transfer offices, particularly in biomedical fields. The pre-populated disclosure form is the single most replicable component and the easiest to implement elsewhere. The mentorship piece is harder to scale.

Does not generalize to

Industrial filers with established disclosure pipelines. The intervention's leverage comes from a baseline where disclosure is the bottleneck. Corporate filers with mandatory disclosure processes face different stage-specific bottlenecks.

2. USPTO Pairolero RCT: expert help at examination

What it is

Pairolero et al. (American Economic Journal: Economic Policy, 2025) conducted the first randomized controlled trial inside the USPTO. The intervention was straightforward: senior examiners were trained and assigned to assist pro se (unrepresented) applicants through the prosecution process.

The intervention

• Random assignment of pro se applicants to receive supplemental examiner assistance versus standard prosecution.
• Treated applicants received clearer rejection explanations, interview prompts, and procedural guidance.
• No change to substantive examination standards or claim scope expectations.

The effect

Women applicants in the treatment arm had a probability of receiving a patent that was more than 12 percentage points higher than the control arm. Secondary press coverage has reported a related 17% effect for first-time applicants, though that specific figure is not in the AEA paper abstract and should be checked against the full paper before citing precisely.

Why it matters

This was the first credible causal evidence that an examiner-side intervention could close part of the grant gap. The mechanism is not bias correction; it is procedural clarity. The same procedural clarity that helped women applicants disproportionately also raised grant rates for the entire pro se population.

Generalizes to

Examiner training, structured response coaching, and any intervention that reduces the procedural complexity of prosecution. The USPTO Council for Inclusive Innovation has institutionalized parts of this work in its broader programs.

Does not generalize to

Represented applicants (those with patent counsel). The Pairolero study population was specifically pro se. Effect sizes for represented filers are likely smaller because the bottleneck is different.

3. Stanford OTL: 50 years of institutional learning

What it is

Tseng et al. (Patterns, Cell Press, 2022) published a 50-year systematic analysis of Stanford's Office of Technology Licensing. The dataset covered 4,512 inventions from 6,557 inventors, 1970-2020.

The finding

Women's share of inventors at Stanford rose from roughly 7% in the mid-1990s to roughly 20% by 2020. The growth was concentrated in biomedical fields, where Stanford's faculty composition was most balanced. Women's representation remained substantially below women's share of Stanford faculty overall (~30%+).

What it reveals

Stanford did not run a single intervention. The 50-year arc reflects compounded institutional learning: an OTL that was actively recruiting faculty inventors, a faculty composition that was slowly diversifying, a research portfolio that tilted increasingly toward biomedical fields where women were more represented, and policy changes (Bayh-Dole in 1980, federal funding shifts) that affected what got patented.

Generalizes to

Other research-intensive universities with active TTOs. The Stanford trajectory is roughly what other Tier 1 research universities should expect over multi-decade horizons.

Does not generalize to

Anywhere expecting fast gains. The 50-year arc shows that institutional change happens over institutional time. The dramatic interventions (WIT, Pairolero RCT) are exceptions; the default trajectory is slow.

4. USIPA Innovation Diversity Pledge: corporate disclosure

What it is

The US Intellectual Property Alliance launched its Innovation Diversity Pledge in 2021. Signatories commit to measuring and disclosing women inventor rates and to implementing internal interventions to close the gap. As of 2025:

• 50+ company signatories, including Meta, Microsoft, Google, Lenovo, P&G, Adobe, AT&T, HP, Cisco.
• 25+ supporting organizations (law firms, consultancies, universities).

Disclosed effects

Meta has reported a 2023 inventorship rate of 17.6%, with a fractional inventorship rate of 16.8%. Both are above the US baseline of 12.8% (2019). Several other pledgees have disclosed rates in the 12-18% range.

What the pledge actually does

The pledge's mechanism is measurement-and-disclosure plus voluntary internal action. Companies that publicly track their numbers are more likely to invest in internal pipeline programs. The disclosure cadence is annual or biannual, which creates accountability.

Caveats

Pledgee disclosures are voluntary and self-reported. The pledgees that have disclosed are not a random sample of US filers; they likely outperform the un-disclosed median (selection effect). Treat the pledgee-disclosed rates as an upper-bound proxy for "what large companies that are actively measuring this look like."

The pattern across all four

The interventions with measured evidence share a common shape:

1. They target access to expert help at a specific stage. (WIT: TTO interaction. Pairolero: examination support. Stanford: institutional outreach. USIPA: company-internal measurement.)
2. They produce measurable percentage-point movement within years, not decades. Even Stanford's 50-year arc is faster when broken into component decades.
3. They are infrastructural rather than rhetorical. None of them is a campaign. All of them change a process.
4. They do not depend on changing the inventor pool first. All four operate downstream of pipeline composition, by raising conversion rates among the existing pool.

The contrast with what does not work is sharp. Awareness campaigns, exhortations, and one-off events do not produce measurable WIR changes in published evaluations. The literature is essentially silent on their effects, which is itself a finding: nothing has been measured because there is nothing to measure.

What an organization can do right now

Three steps, ranked by evidence base:

• Pre-populate invention disclosure forms with all team member names. WashU's 129% effect came largely from this single change.
• Proactively reach out to women researchers rather than waiting for disclosures to arrive. Pair with the disclosure-form redesign.
• Measure and publish gendered patent data within the organization. USIPA pledgees who publish their numbers tend to act on them.

For policy actors: the Pairolero RCT showed that an examiner-side intervention can shift outcomes by 12+ percentage points. Funding more interventions of this type would produce more evidence on what generalizes.

Sources

More from the patent gender gap series