Pipeline vs. Choice: gender gaps in STEM applications

10

Centralized admission systems: Australia, Brazil, Chile, China, Finland, Greece, Spain, Sweden, Taiwan, Uganda.

23.7 percentage points

How much less likely a top-10% woman is than a top-10% man to rank a STEM program first, averaging across the ten settings.

57%–75% male

What the high-achieving STEM applicant pool would still be if the top 10% were gender balanced, holding application rates at their observed values, in every setting.

Pipeline and choice gaps across settings

The pipeline gap is who reaches the top of the academic distribution. It varies markedly across settings and in sign, from female disadvantage in Uganda to female advantage in Sweden, and is positively associated with labor-market gender parity. The choice gap compares women and men who have already cleared the same bar. It is negative in every setting under both “STEM first” and “STEM anywhere on the list” definitions. The pooled first-choice gap changes by less than one percentage point as the high-achieving group widens from the top 5% to the top 30%. The choice gap exceeds 20 percentage points in eight of the ten systems. 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.

Pipeline gap and choice gap, ten settings, shared percentage-point axis

Both quantities are female-minus-male differences in percentage points among top-10% students, plotted on the same scale so their dispersion can be compared directly.

Below zero: female disadvantage
Above zero: female advantage

Pipeline and choice gap by setting among top-10% students, in percentage points. Negative values indicate women below men.
Setting	Pipeline gap (pp)	Choice gap (pp)
Australia	+7.3pp	−16.1pp
Brazil	−1.2pp	−20.2pp
Chile	−12.0pp	−23.1pp
China	+10.7pp	−36.5pp
Finland	+3.3pp	−26.2pp
Greece	+22.1pp	−22.5pp
Spain	+25.5pp	−29.7pp
Sweden	+31.7pp	−26.3pp
Taiwan	−20.9pp	−24.7pp
Uganda	−22.7pp	−11.5pp

The zero line is drawn because the sign is the point: every choice gap sits below it, while the pipeline gap crosses it. Each dot is one setting, labelled with its ISO country code; the bracket over each row runs from its lowest to its highest value. Chile's choice gap (−23.1) uses full top-decile totals, as in Online Appendix Table B.I and the decomposition; the paper's Figure 2(b) conditions on students who submit a list (−24.6).

Women's share at each stage, pooled across settings

Share of each group that is female, %. The rest of each group is men.

Women's share of each group, pooled across the ten settings, in percent.
Group	Women's share (%)
Top 10% of scorers	51.2%
List STEM anywhere	38.6%
Rank STEM first	34.0%
Assigned to STEM	34.7%

The four shares are the pooled values the manuscript reports (Results, Section 4.5, pp. 13–15; Online Appendices E, F and B.8). The column at the right is the change from the stage before, in percentage points.

Both gaps against the wage parity index

Top-10% students, female minus male, percentage points, against each setting's wage parity index (women's to men's earnings).

The wage parity index is from Online Appendix Table A.I. The pipeline gap broadly tracks wage parity, while the choice gap stays below zero at every level of it. The two lines are least-squares fits across the ten settings, computed on this site to describe the association; they are not estimates reported in the paper.

The ten settings on the map

Each map shades a setting by its gap among top-10% students, on the same scale. On the pipeline map the shading runs from terracotta to teal; on the choice map every setting is terracotta.

Pipeline gap

Women's minus men's share of the top 10%, percentage points.

Choice gap

Gap in ranking a STEM program first among top-10% students, percentage points.

Female disadvantageFemale advantage−36 to +36 pp

Australia: Victoria only. Brazil: federal institutions on the SISU platform. China: the Ningxia Hui Autonomous Region. Values are printed on the maps, so the shading is never the only cue; the same values are in the data table of the first figure on this page. Boundaries: Natural Earth, 1:110m.

Every threshold, three fields

The panels repeat the STEM first-choice gap for each setting from the top 1% to the top 30% of its performance distribution, on one scale. The second figure sets STEM beside business and law and health at the top 10%.

STEM ranked first, from the top 1% to the top 30%

Female minus male, percentage points. One scale for every panel, 0 to −45 pp.

Australia−16.1 at 10%

Brazil−20.2 at 10%

Chile−23.1 at 10%

China−36.5 at 10%

Finland−26.2 at 10%

Greece−22.5 at 10%

Spain−29.7 at 10%

Sweden−26.3 at 10%

Taiwan−24.7 at 10%

Uganda−11.5 at 10%

STEM first-choice gap by setting and threshold, in percentage points.
Setting	Top 1% (pp)	Top 5% (pp)	Top 10% (pp)	Top 20% (pp)	Top 30% (pp)
Australia	−11.9pp	−14.7pp	−16.1pp	−18.9pp	−21.1pp
Brazil	−17.0pp	−19.7pp	−20.2pp	−19.5pp	−18.9pp
Chile	−25.0pp	−23.1pp	−23.1pp	−22.8pp	−22.3pp
China	−17.0pp	−32.8pp	−36.5pp	−38.0pp	−38.5pp
Finland	−22.5pp	−25.3pp	−26.2pp	−25.8pp	−24.6pp
Greece	not available	−17.8pp	−22.5pp	−24.5pp	−24.3pp
Spain	−30.6pp	−30.0pp	−29.7pp	−29.7pp	−29.1pp
Sweden	−20.4pp	−23.5pp	−26.3pp	−26.0pp	−25.5pp
Taiwan	−11.5pp	−20.7pp	−24.7pp	−28.8pp	−31.9pp
Uganda	−8.0pp	−12.8pp	−11.5pp	−15.6pp	−17.4pp

Greek data are reported by decile and by top 5%, so a top-1% group cannot be formed; that point is marked n/a rather than estimated. Some groups are not exact in every setting; the results explorer gives the details.

First-choice gap by field, top 10%

Female minus male, percentage points. Each dot is one setting; the lowest and highest are labelled.

Below zero: female disadvantage
Above zero: female advantage

First-choice gap at the top 10% by setting and field, in percentage points.
Setting	STEM (pp)	Business & law (pp)	Health (pp)
Australia	−16.1pp	−13.5pp	+4.9pp
Brazil	−20.2pp	−0.8pp	+19.3pp
Chile	−23.1pp	−1.9pp	+19.1pp
China	−36.5pp	+13.5pp	+0.9pp
Finland	−26.2pp	−3.8pp	+16.2pp
Greece	−22.5pp	not available	−1.9pp
Spain	−29.7pp	+3.9pp	+13.3pp
Sweden	−26.3pp	+2.7pp	+8.9pp
Taiwan	−24.7pp	+14.5pp	−1.4pp
Uganda	−11.5pp	+7.3pp	+2.8pp

Greece does not report business and law counts at the top 10%. Business and law and health are reported at the top 5%, 10% and 20% only.

Why ranked applications are informative

All ten systems rank students on academic performance and assign places through variants of deferred acceptance. Under that class of mechanism, ranking programs in true preference order gives applicants little incentive to misreport. The submitted list is therefore informative about the ordering of programs an applicant was willing to enter. It is not interpreted as a preference parameter: list limits, score timing, and information held before application remain relevant to interpretation.

Read the decomposition in full →

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