Talented employees discriminating against female entrepreneurs

How to cite

Hu, Zhongchen. 2025. "Talented employees discriminating against female entrepreneurs." AEA RCT Registry. AEARCTR-0016400. https://doi.org/10.1257/rct.16400-2.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Zhongchen Hu*

* Primary investigator

Registry ID
AEARCTR-0016400
DOI
https://doi.org/10.1257/rct.16400-2.0
First registered
Last updated
Status
Completed
Trial dates
2025-03-13 – 2025-06-01
Countries
United States of America ()
Registry record
http://www.socialscienceregistry.org/trials/16400

Abstract

We conducted a randomized field experiment in which an entrepreneur reached out to high-skilled prospective employees regarding a job opening. The gender of the co-founders was randomly altered. We aim to test whether high-skill employees exhibit a differential response rate to a meeting request from a female entrepreneur. Our pilot, which ran in March 2024, indicates that high-skill employees are 50% less likely to respond to a female entrepreneur. To identify the mechanism, we primed a subset of applicants with a newsletter from an unrelated non-profit organization, sent a few days before the entrepreneur’s outreach. The newsletter provided information on large initiatives aimed at supporting female entrepreneurs with venture capital funding. In the pilot, this intervention completely eliminated the observed discriminatory behavior. Our results support a theory in which applicants internalize the discrimination of other market participants that might affect the successful outcomes of entrepreneurs.

We submitted our AEA registry after the field experiment concluded because we were concerned that participants could learn about the study while it was still taking place. Given how easily information can diffuse on social media and the divide on topics like discrimination, we feared there was a small chance our project could become publicly visible before it began. While low-risk, in the current environment, this possibility is not negligible. We only learned after submitting the registration that AEA provides an option to keep the registration private, and the authors can decide when to make it public.

We understand that a post-registration might raise concerns about adapting our hypotheses post-experiment, but to demonstrate that our hypotheses were established well before data collection began in March 2025, we highlight three facts:

1. The original IRB application, initially submitted on October 16, 2023, and approved on October 23, 2023, clearly stated our core hypothesis (workers discriminate against female entrepreneurs). In January 2025, we amended that application to incorporate the priming treatment that identifies the mechanism through which workers discriminate against female entrepreneurs. The IRB applications are attached to this registration.

2. From Feburary 8, 2024 to March 22, 2024 and from November 27, 2024 to December 12, 2024, we ran two pilots in which we (i) tested whether workers discriminate against female entrepreneurs, and (ii) assessed whether priming participants with information about venture-capital initiatives supporting female entrepreneurs reduces that bias. The full experiment followed the pilots precisely. We describe the pilot below.

3. We shared these pilot findings (workers discriminate against female founders, and priming applicants with a newsletter reduces this bias) throughout 2024 to solicit feedback and confirm our theoretical framework:

- University College Dublin Seminar – February 29, 2024
- Research Institute of Industrial Economics (IFN) Seminar – March 13, 2024
- Chicago Entrepreneurship Workshop – June 7, 2024
- University of Melbourne Seminar – November 7, 2024

We attached below the presentation slides at Chicago Entrepreneurship Workshop. These steps confirm that our hypotheses were specified prior to the field experiment. We postponed the AEA registration not to adapt our hypotheses post-experiment, but only to maintain the experiment’s validity.

Intervention

The intervention design is inspired by the correspondence study approach. However, unlike a traditional correspondence study, in our setting, an employer reaches out to potential employees. Male and female entrepreneurs send emails to potential applicants, offering a new job opportunity and the chance to schedule a 5-to-10-minute phone or Zoom call, or an in-person campus meeting to discuss further details about the job opening.

We create 9 identical pairs of mock founders and startups (9 founded by males and 9 founded by females). Each male startup has a corresponding identical female startup. These startups are modeled after real startups. Our companies span various industries, including green tech, healthcare tech, prop tech, and others. Each mock startup has an online presence, including a website, a unique email domain, and a presence on LinkedIn. For each type of startup, we create two distinct startups with different websites and company names. The websites are identical except for the company name, founder pictures, and founder names. Most startups are represented as having two founders, either both male or both female.

Applicants can identify the gender of the founders either by reading on the startup’s website or by visiting the founders’ LinkedIn profiles. In both cases, gender is signaled through pictures and names.

The founders’ pictures are generated using artificial intelligence. We create images of non-existent individuals through online tools and use AI to generate matching counterparts of the opposite gender, preserving most key facial traits, such as eye shape, mouth width and height, the distance between eyes, smile, etc. Using online surveys, we verify that the male and female pairs have similar attributes, including attractiveness, trustworthiness, and other relevant characteristics.

The male and female founders also have LinkedIn profiles that perfectly match their employment and education history. The only differences between these profiles are the founders’ pictures and names.

Our sample of potential employees consists of individuals with Bachelor’s and Master’s degrees in computer science and business administration from more than 30 of the top 100 U.S. universities. We include only universities that maintain publicly accessible directories of student emails. We use LinkedIn to gather students who are graduating in 2025 with Bachelor's or Master's degrees from our target majors and universities. We include individuals whose names can be unambiguously matched to their university email listings.

To provide some indirect compensation to individuals who responded to the mock entrepreneurs, we shared their email contact information and online resumes with real entrepreneurs in similar industries. When sharing this information, we emphasize that these individuals are interested in startups like those managed by the real entrepreneurs. The real entrepreneurs will reach out to these individuals about potential employment opportunities if they align with their experience and the needs of their businesses.

Experimental design

We targeted a sample of 8,605 individuals who are graduating in Spring 2025.

There are four treatment arms. The first treatment arm received no priming and was contacted by a male entrepreneur. The second treatment arm received no priming and was contacted by a female entrepreneur. The third treatment arm received priming via a newsletter and was contacted by a male entrepreneur. The fourth treatment arm received priming via a newsletter and was contacted by a female entrepreneur.

Each entrepreneur reaches out to a single individual, who only receives one contact. Additionally, only one firm in a matching pair reaches out to graduates from a specific university-major, thereby ensuring no information spillovers that could allow participants to uncover the experiment. We aim to prevent two acquaintances from receiving emails from identical firms with founders of different genders, as it would allow participants to easily conclude that the companies are not real. Entrepreneurs contact participants via email, introducing the company, describing the job opportunity, and offering a software engineer or product manager position to individuals with a computer science or business background, respectively. The entrepreneur invites the candidate for a short meeting to discuss the job opportunity.

Participants receive a priming treatment that aims to mitigate their beliefs of gender-based discrimination in the venture capital market. They receive a newsletter from a non-profit organization that we established specifically for this project and has no connection with the entrepreneurs. The non-profit’s mission is to connect young entrepreneurs with financing programs. The inaugural issue highlights four real initiatives that uniquely support female-led ventures, including one sponsored by Google and another by Mastercard. The newsletter is sent two to five days before the entrepreneur’s outreach.

Design details

Randomization method
Randomization by computer
Randomization unit
Individual
Clusters
No clusters
Observations
8,605 individuals
Arms
The first treatment arm received no priming and was contacted by a male entrepreneur: 2,083 individuals The second treatment arm received no priming and was contacted by a female entrepreneur: 2,167 individuals The third treatment arm received priming via a newsletter and was contacted by a male entrepreneur: 2,068 individuals The fourth treatment arm received priming via a newsletter and was contacted by a female entrepreneur: 2,287 individuals
Primary outcomes
The outcomes of interest are: ⁃ Likelihood of wanting to schedule a phone or Zoom call, or a campus meeting to discuss the job opportunity. We code a binary variable that equals one if the applicant replies yes to scheduling, and zero if they decline to schedule or provide no response. ⁃ Textual characteristics of the reply to the entrepreneur (i.e., tone, length, additional information about the applicant). ⁃ Likelihood of requiring additional verification (i.e., whether the applicant requests additional information about the startup before wanting to schedule a meeting)

Cited by

No citation data available for this registration yet.

Keywords