Schooling and Human Capital Formation under Weather Stress: Experimental Evidence from Bangladesh

The intervention provides eligible grade 8 students with a foldable umbrella intended for daily commuting. The umbrella is lightweight, easy to carry with a built-in strap, and small enough (when folded) to fit in a school bag. The umbrella is accompanied by a light-touch information package consisting of two stickers intended to be placed in a visible location at home: (i) a reminder sticker (“Rain or shine, this umbrella is mine!”); and (ii) a brief note to parents stating that the umbrella is intended for the student’s use and should remain available to the child. None of these materials includes educational encouragement or messages about schooling.

The umbrellas were provided by the Sub-district Administrations from their annual development budget.

2026-04-21

2026-10-11

Absenteeism rate on adverse-weather school days.

The adverse-weather absenteeism rate is defined for each student as the number of adverse school days absent divided by the total number of adverse-weather school days.

Seasonal sickness (self-reported)
Standardized test scores
Girls' mobility

School Selection
The study is a cluster-randomized controlled trial with randomization at the school level. A list of all registered schools was obtained from the District Education Office in Pabna. The list contains 325 secondary schools, of which 241 are located within the five study subdistricts.
Schools were considered ineligible if they met any of the following criteria:

1. The school was designated as a center for the Secondary School Certificate (SSC) public examination.
2. The school was single-sex (i.e., not co-educational);
3. The school had fewer than 200 enrolled students in total, or
4. The school was not a standard secondary school (i.e., it was a junior secondary school serving up to grade 8 or a higher secondary school serving up to grade 12).

After applying these eligibility criteria, 134 schools remained eligible for participation. From this pool, 106 schools were randomly selected and subsequently assigned to either the treatment or control group. Randomization was conducted prior to baseline data collection.

Student Selection
Within each study school, the target population is Grade 8 students. Student eligibility was determined through a short screening survey administered to all, or nearly all, Grade 8 students present on the survey day. Students were eligible for inclusion if they met both of the following criteria:

1. The student walked to school or used a rickshaw van as the primary commuting mode; and
2. The student lived in a household with no more than one usable umbrella.

In treatment schools, I aimed to enroll approximately 30 eligible students per school. If more than 30 students were eligible in a treatment school, up to 30 students were randomly selected from the eligible list through an on-the-spot lottery. This cap was imposed because the number of umbrellas available for distribution was limited. In control schools, all eligible students were included in the study sample.
 
Data Collection

Baseline survey
To assess eligibility, I conducted a short paper survey consisting of three questions. The eligible students completed a structured survey in a classroom setting on a tablet computer. They also attended a standardized cognitive test covering English, mathematics, and general aptitude. In parallel, I administered a short school questionnaire to the headteachers to capture key school characteristics. I extracted the attendance data up to the survey date (Jan- April 2026) from the school records.

Endline survey
At the endline, I will re-survey students and headteachers. Students will complete a cognitive assessment comparable in structure and difficulty to the baseline test. There will also be a survey module on relevant behaviors and experiences over the study period.

During-intervention data
During the intervention period, I will collect implementation and monitoring data to assess treatment intensity and compliance. First, I will conduct unannounced school visits in both treatment and control schools to measure (i) student attendance on the day of the visit and (ii) umbrella use. I plan four visits per school: two during the monsoon and two during the hot season. Second, I will compile high-frequency attendance outcomes by digitizing school register data. Third, I will compile high-frequency weather data for the study area, including daily rainfall (mm) and daily temperature (°C). I will align these measures to the finest feasible geographic and temporal resolutions.

Not available

Randomization was done at the office using Stata.

Randomization was done at the school level

Yes

105 schools

2000

54 schools in the control group
51 schools in the treatment group

MDE of 0.15–0.20 SD in absenteeism during adverse-weather days