Agentic Women and Workplace Bias

How to cite

Maheshwari, Prerana. 2025. "Agentic Women and Workplace Bias." AEA RCT Registry. AEARCTR-0015847. https://doi.org/10.1257/rct.15847-1.0

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

Prerana Maheshwari*

* Primary investigator

Registry ID
AEARCTR-0015847
DOI
https://doi.org/10.1257/rct.15847-1.0
First registered
Last updated
Status
In development
Trial dates
2025-04-29 – 2025-08-31
Countries
Private
JEL codes
J24, J71
Registry record
http://www.socialscienceregistry.org/trials/15847

Abstract

Despite advances, women remain underrepresented in masculine fields and higher-level positions. Agentic qualities (e.g., confidence, assertiveness, dominance) are often crucial for workplace success, yet women face a double bind: they may be stereotypically perceived as lacking these traits, but penalized for displaying them (backlash). This study experimentally investigates whether providing explicit information about women's agentic qualities mitigates stereotype-based hiring discrimination in masculine-typed roles or triggers norm-based backlash. We examine how effects vary by job level (entry vs. higher-level/influential) and the specific agentic trait, isolating the impact of dominance (a proscriptive trait) versus other prescriptive agentic qualities (confidence, assertiveness, diligence, independence). Using a two-part experiment, we first generate candidate profiles based on worker performance in tasks measuring competence (Target Guessing Game) and these agentic traits. Second, online evaluators review and rank sets of CVs—systematically varied by candidate gender, competence signal (noise level), and the presence/type of agentic information—for both entry-level and higher-level positions. We test whether agentic information reduces discrimination or increases it via backlash, whether dominance information triggers greater backlash compared to other agentic traits (especially for women and in higher-level roles), and whether agentic qualities are perceived as signals of competence. The findings aim to elucidate the complex interplay between statistical discrimination and taste-based backlash in driving workplace gender bias against agentic women.

Experimental design

This study employs a two-part experimental design. In the first part, participants (workers) complete a series of tasks designed to generate measures related to competence (via a dynamic number-guessing game) and several agentic qualities (including assertiveness, confidence, diligence, self-reliance, and dominance). Data from this part are used to construct candidate CV profiles for the second part.
In the second part, a separate group of participants (evaluators) engage in a hiring/selection process. Evaluators are presented with sets of four candidate CVs at a time. These CVs contain varying information about the candidates, including gender, a signal related to their competence derived from the first part, and information regarding their agentic qualities. Evaluators are asked to rank these four candidates from 1 (most preferred) to 4 (least preferred) for a hypothetical position requiring skills relevant to the competence task.
The design includes two key structural components for comparison. First, each evaluator will make ranking decisions for two distinct job contexts presented in random order: an entry-level position and a higher-level position. Second, evaluators will be assigned to different conditions regarding the specific agentic information displayed on the CVs; notably, some evaluators will see profiles including information on dominance, while others will see profiles where this specific trait information is omitted, focusing instead on other agentic qualities. The primary outcome variable is the rank assigned to each CV under these different informational and job-level contexts.

Design details

Randomization method
All randomizations will be done by computer softwares.
Randomization unit
Randomization will happen at individual level, as different individuals will be assigned to different treatments and order of decisions within a treatment.
Clusters
800 individuals
Observations
800 individuals
Arms
400 individuals
Primary outcomes
The central outcome variable in this study is the rank assigned by evaluators to each candidate's CV. We are primarily interested in examining the differences in these ranks when comparing candidates who are presented with systematically varied information regarding their gender, their competence signal, and their agentic qualities (including variations in the presence, level, and specific type of agentic information, such as the inclusion or exclusion of dominance). By analyzing how these ranks differ across CVs presenting distinct combinations of characteristics, we aim to measure the extent of baseline gender discrimination and determine whether providing specific informational signals about competence or agency mitigates this discrimination or potentially triggers backlash, leading to changes in hiring preference. Furthermore, we will investigate whether these observed differences in ranking patterns vary significantly depending on the job level context for which the evaluator is considering the candidate (entry-level versus higher-level). Supporting ratings of hireability, competence, and likability will also be collected to provide deeper insight into the evaluators' decision-making processes reflected in the rankings.
Secondary outcomes
Using the ranks assigned in a particular set of CVs, we will be able to determine if agentic qualities are perceived as signals of competence.

Cited by

No citation data available for this registration yet.

Keywords