Improving Women’s Participation in STEM in Tajikistan: A Role-Model Intervention

The core intervention is a structured 30-minute classroom session in which girls in grades 9 and 10 are shown a sequence of three videos: (i) a general video on STEM careers; (ii) a video in which three Tajik women describe their professional journeys; and (iii) a deeper profile of a single role model who trained internationally. A randomly selected subset of sections additionally receives the same videos with parents over WhatsApp. Randomization is at the level of the section. Because treatment is assigned at the section level while outcomes are measured at the student level, all inference will account for clustering at the level of randomization.

2026-05-14

2026-05-23

The primary outcomes for this study are students’ STEM aspirations, aspired education level, and attitudes toward STEM.

The primary outcomes for this study are students’ STEM aspirations, aspired education level, and attitudes toward STEM. STEM aspirations will be measured using indicators of intended track or stream choice and desired or expected sector, and will be coded as a binary outcome equal to one if the student selects a STEM-related track or sector, and zero otherwise. Aspired education level will be measured using students’ reported desired education level under both no-barrier and current-condition scenarios, coded on an ordinal scale ranging from basic general education to doctorate-level education. Attitudes toward STEM will be measured using multiple Likert-scale survey items capturing students’ perceptions, interest, and orientation toward STEM fields. Where outcomes are measured using multiple related survey items, we will construct standardized indexes by first aligning item direction so that higher values correspond to stronger STEM aspirations or more positive STEM attitudes, standardizing each item relative to the control group, and then averaging across non-missing items within each outcome domain. Higher values of all primary outcomes and indexes will be interpreted as indicating stronger STEM orientation or aspiration.

Secondary outcomes span six families covering the normative, informational, and academic channels through which the intervention is expected to operate.
Social-gender norms are measured in two dimensions: girls' own beliefs about the appropriateness of STEM and technical work for women (items 5.1–5.4) and their perceptions of what their parents believe (items 5.7–5.10), both constructed as standardized Likert indices. Complementing these, perceived prevalence of women in STEM (items 5.13–5.15) captures girls' factual beliefs about how common female participation in STEM and the workforce is, measured on a 0–100 scale and standardized continuously.
Career perceptions (items 7.1–7.5) capture beliefs about the returns, stability, and family-life compatibility of STEM careers, constructed as a standardized index with item 7.3 reverse-coded. The subjective probability of completing a STEM degree (items 4.7, 4.8.1, 4.16) measures girls' forward-looking assessments of their own likelihood of entering and succeeding in STEM fields, indexed by field.
Academic outcomes include math grade (item 3.2, third-quarter) and a science composite grade averaging across math, physics, chemistry, and biology (item 3.5), both standardized within grading-scale type using item 3.6.
Mechanism outcomes include self-efficacy and growth mindset (items 8.1–8.11), a Likert index with mindset items reverse-coded where applicable; STEM social networks and role-model exposure (items 9.1, 9.3, 9.4), capturing whether girls know a woman working in STEM and the size of their STEM-relevant peer ties; and encouragement and mentoring in STEM (items 3.18, 3.19, 6.12, 6.13), measuring teacher and family support through a mix of binary, categorical, and Likert items.

These secondary outcomes capture students’ social-gender norms, perceptions of women’s participation in STEM and work, career perceptions, perceived likelihood of pursuing and succeeding in a STEM degree, and academic performance in STEM-related subjects. Outcomes measured using multiple related survey items will be combined into standardized indexes. For these indexes, item direction will first be aligned so that higher values consistently represent more gender-equitable norms, more positive STEM perceptions, stronger STEM-related confidence, or greater support for STEM participation. Items will then be standardized and averaged within each outcome family, using non-missing responses where applicable. Reverse-coded items, such as item 7.3 and relevant growth mindset items, will be recoded before index construction. Math and science grades will be standardized within grading-scale type to ensure comparability across respondents. The final three outcomes—self-efficacy and growth mindset, STEM social networks and role-model exposure, and encouragement and mentoring in STEM—will be interpreted primarily as mechanism outcomes, since they are intended to help explain potential pathways through which the intervention may affect students’ STEM aspirations, attitudes, and choices.

The core intervention is a structured 30-minute classroom session in which girls in grades 9 and 10 are shown a sequence of three videos: (i) a general video on STEM careers; (ii) a video in which three Tajik women describe their professional journeys; and (iii) a deeper profile of a single role model who trained internationally. A randomly selected subset of sections additionally receives the same videos with parents over WhatsApp. Randomization is at the level of the section. Because treatment is assigned at the section level while outcomes are measured at the student level, all inference will account for clustering at the level of randomization.
Arm Description
Control No video exposure; standard schooling. Receives baseline and endline surveys only.
T1: Classroom 30-minute structured classroom screening of the three-video sequence
T2: T1 + parent WhatsApp All elements of T1 plus continued WhatsApp engagement with parents through school groups. Same videos are shared with parents via WhatsApp over the following week.

Not available

Done using code on computer using the baseline schools and sections

Section is the unit of randomization. We stratified the ranodmization by school and grade

Yes

We have 25 schools and 213 sections across grades 9 and 10

We expect to have around 1500-2000 students answering the survey

73 sections Control, 71 sections T1 and 69 sections T2