New research out of Emory University delivers a pointed message for talent acquisition leaders in STEM: rewriting a job description is not enough to move female-applicant conversion rates. Women who believed a role had higher female representation were more likely to apply, even when the posting required abilities typically stereotyped as masculine. The implication for HR strategy is structural. The pipeline problem is not at the job description. It is at the visibility of who already holds the job.

The study was authored by Stella F. Lourenco, Kylee Novick, Ryno Kruger, and Arber Tasimi at Emory University, with Lourenco's public statement issued September 14. For People and Culture leaders benchmarking STEM hiring outcomes, this research surfaces a retention-and-acquisition risk that diversity statements and inclusive language guides cannot neutralize on their own.

Representation is signaling climate, not symbolism

Stella F. Lourenco, Emory psychology professor and lead author, pushed back directly against the framing most employer-branding teams operate under. Gender representation, she argued, is not a decorative outcome of past hiring. It is a primary input into whether the next woman applies.

"Some people may think of gender representation as mostly symbolic, that it's simply nice for women to see themselves represented in a field. What we found is that gender representation is more consequential. It is perhaps a primary reason women decide whether to enter a particular occupation in the first place."

Lourenco's working hypothesis is that gender representation functions as a proxy for workplace climate: a shorthand that candidates use to read what an organization actually expects around support, inclusivity, and competitiveness. When a candidate scans a team page and sees no women in the function she is being recruited into, she is not reading it as a coincidence. She is reading it as a forecast about day-to-day reality.

For CHROs and HR directors, that framing has a direct cost implication. A recruiting funnel that consistently loses qualified women at first-contact because of what those candidates observe on team pages, LinkedIn footprints, and engineering blogs carries a compounding opportunity cost that no amount of sourcing spend can offset downstream.

The brilliance trap in STEM recruiting

The Emory study connects to the field-specific ability beliefs theory, which holds that women may be underrepresented in fields associated with intrinsic aptitude because women are stereotyped as not possessing that quality. The theory helps explain why the gender gap persists in software engineering, quantitative finance, machine learning, and applied mathematics: functions where the folk model of role success centers on innate cleverness rather than acquired, trainable skill.

The practical implication for talent acquisition teams is that removing masculine-coded language from a job posting is necessary but not sufficient. If the team photograph, the LinkedIn profiles of the hiring manager's function, the engineering blog authors, and the conference speaker roster are predominantly male, the ad copy is working against evidence a candidate can verify in under a minute. The Society for Human Resource Management (SHRM), which represents approximately 340,000 HR professionals across the United States, has consistently flagged the gap between diversity-language investment and structural representation gaps as one of the unresolved tensions in STEM talent strategy.

A pattern that shows up across the funnel

The Emory finding does not stand alone. Two other research threads suggest female candidates read pay and progression signals differently at multiple points in the recruiting funnel, not just at first contact.

A Cornell University report published earlier in 2026 found that women study participants felt more discouraged than men when facing job postings with wide salary ranges. This matters for US-based HR teams because pay-transparency legislation in California (SB 1162, effective January 2023), New York, Colorado, and a growing number of states now requires employers to post salary ranges. Wide bands introduced to preserve negotiation flexibility or to comply with disclosure rules may be generating an unintended friction point for female applicants. What reads as flexibility to the employer reads as ambiguity, and pay ambiguity reads as risk to the candidate evaluating it.

A Rockwool Foundation Berlin Institute for the Economy and Future of Work report published earlier in 2026 found men were more comfortable than women using outside job offers to negotiate a higher salary. In the Rockwool study of Swedish workers, women moved jobs as often as men, which led researchers to conclude the pay outcome gap was not a gap in ambition or mobility. It was a gap in the willingness to reopen the salary conversation. US-based compensation research from WorldatWork's annual Salary Budget Survey reflects a comparable pattern: offer-stage negotiation behavior diverges by gender in ways that, left unmanaged, embed pay equity gaps before a new hire clears their first performance cycle.

Read together, the three studies describe a funnel in which women self-select out at three distinct points: at the decision to apply (Emory), at the interpretation of the salary range in the posting (Cornell), and at the leverage moment during renegotiation (Rockwool). Every stage where the recruiting process treats candidates as interchangeable is a stage where friction compounds asymmetrically by gender. For organizations monitoring Equal Employment Opportunity Commission (EEOC) charge trends or managing proactive pay-equity audits, these are the upstream variables that downstream metrics tend to obscure.

What HR can audit this quarter

Diversity statements on the careers page carry the lowest production cost and the lowest credibility with the candidates talent teams most need to convert. The Emory research points toward higher-leverage actions that are structural and verifiable, not rhetorical.

• Publish representation data at the function level, not the company level. Aggregate statistics about overall workforce gender balance do not answer the question a data-engineering or machine-learning candidate is actually asking: how many women are already in the specific team she would be joining. Where function-level disclosure is feasible, publish it. Where it is not, ensure interview panels are substantively mixed rather than tokenistic.
• Audit the first-contact surface systematically. Team pages, LinkedIn profiles of the hiring manager's direct function, engineering blog authorship, and conference presentations the team has given in the past twelve months. A candidate evaluating a posting will find these in the same session she reads the job ad. If they contradict the ad, the ad loses.
• Tighten salary bands or annotate the range. Where state pay-transparency laws, including California SB 1162 and New York's pay range requirements, require a posted range, add a plain-English explanation of what determines placement within it. The Cornell data indicates that unexplained range width discourages the exact applicant group organizations are trying to broaden.
• Neutralize the renegotiation gap at the offer stage. An offer process that assumes candidates will counter has built an equity gap into the compensation architecture before the hire starts. Structured approaches, including best-and-final offers or a defined second conversation where every candidate is invited to raise concerns about the package, can reduce the behavioral asymmetry the Rockwool study identifies. Organizations tracking compliance with the Equal Pay Act of 1963 alongside Title VII of the Civil Rights Act of 1964 should treat offer-stage renegotiation design as a pay-equity control, not just a negotiation courtesy.
• Monitor employer review platforms as a climate-signal channel. Written testimony from current female employees carries credibility that a branded diversity statement cannot replicate. Review platform content either confirms or contradicts what the recruiting team communicates, and candidates treat it accordingly.

The self-reinforcing cycle HR has to break

Lourenco's closing observation is the one with the most direct strategic relevance for talent planning.

"Gender representation is often spoken of simply as a result of women choosing not to work in a particular field. It's really a self-reinforcing cycle. Today's gender representation likely plays a role in shaping tomorrow's gender representation."

For HR leaders, the cost of inaction is not stasis. It is compounding underrepresentation, because each cohort that self-selects out narrows the visible signal available to the next. Bureau of Labor Statistics (BLS) occupational data consistently shows that computer and mathematical occupations remain below 30% female representation at the national level. The functions with the widest gaps today are the ones where the climate signal has been consistently male for longest, and where a rewritten job posting will produce the least measurable change.

The lever is not in the ad copy. It is in what a candidate observes when she looks past the ad: the team composition, the interview panel, the salary language, and what current employees report when asked without a company representative in the room. HR teams that treat representation as a live climate signal, measured and managed alongside other talent metrics, are the ones positioned to shift female-applicant conversion rates this recruitment cycle. For organizations benchmarking against peers, SHRM's annual State of the Workplace report and WorldatWork's compensation data provide the named reference points against which function-level progress can be calibrated.