Skip to main content
Pipeline vs. ChoiceOnline materials

Paper & citation

The Global Gender Gap in STEM Applications: Pipeline vs. Choice

A Short Communication by 13 authors across 13 institutions, using administrative microdata from ten centralized university admission systems.

Abstract

Women make up only 35% of STEM graduates worldwide. We use ranked applications and assignments from ten centralized university admissions systems to distinguish gender differences in academic preparation (the pipeline) from differences in application decisions (the choice gap). Among students in the top 10% of each setting's academic performance distribution, women are about 24 percentage points less likely than men to rank STEM first, averaging across settings. The pipeline gap varies widely and broadly tracks wage parity; the STEM choice gap remains large and negative everywhere. Linking applications to assignments shows that women's underrepresentation largely precedes allocation. Pooled across settings, women are 51.2% of high achievers, 38.6% of those listing STEM anywhere, 34.0% of those ranking it first, and 34.7% of those assigned to STEM. Holding application rates fixed, a gender-balanced pipeline would still leave high-achieving STEM applicants between 57% and 75% male in every setting. A strong female academic pipeline can therefore coexist with substantial underrepresentation in STEM.

Download

Title page of ConsiliumBots Working Papers issue 03

Working-paper versions

Issued as ConsiliumBots Working Papers no. 03. Each version keeps a permanent URL and is never rebuilt, so a citation resolves to the document that was cited. Version 2 (2026-09-05) is the latest issued working-paper version, an earlier version of the text this site documents. The pages on this site and the manuscript PDF below follow the October 2026 revision, which the series has not yet issued.

Working paper v2 (PDF)The series

Earlier version: v1 (2025-11-15).

Manuscript PDF

The manuscript as revised in October 2026, the version this site documents, online appendix included (pp. 23–42).

Download PDF

Replication materials

Code and the derived, non-identifying aggregates that generate every figure on this site.

pending · repository link

Authors

Affiliations are reproduced from the manuscript title page.

Citation

@article{ahimbisibwe_gender_stem,
  title   = {The Global Gender Gap in STEM Applications: Pipeline vs. Choice},
  author  = {Ahimbisibwe, Isaac and Altmejd, Adam and Artemov, Georgy and
             Barrios-Fern\'andez, Andr\'es and Bizopoulou, Aspasia and Kaila, Martti and
             Liu, Jin-Tan and Megalokonomou, Rigissa and Montalb\'an, Jos\'e and
             Neilson, Christopher and Otero, Sebasti\'an and Sun, Jintao and Ye, Xiaoyang},
  year    = {},
  journal = {},
  note    = {Short Communication}
}

Year, journal and DOI fields are intentionally empty until the manuscript record exists.

Data access

The analysis uses individual-level administrative records from the ten centralized admission systems listed in Online Appendix Table A.II. These records are confidential and were obtained under separate agreements with each agency, so the authors are not able to redistribute them. Researchers seeking access should apply to the agency named for that setting in the same table.

What can be shared from this project:

  • The derived aggregates behind every figure and table on this site — shares, gaps and counts by setting, threshold, outcome and field — as CSV.
  • The code that draws the figures and tables from those aggregates.
  • The documentation on each country page: rules, calendars, performance-measure construction, and sample definitions.

What cannot:

  • Student-level records, ranked lists, or any file from which an individual could be identified.
  • Any agency extract. The records were obtained under separate agreements with each agency and cannot be redistributed by the authors.

The per-agency conditions differ, and the exact terms and contact routes belong on each country page under “Data provenance and access”. pending · per-agency access terms