School Subsidies as a Commitment Device

This study evaluates whether school-fee subsidies function as a commitment device relative to unconditional cash transfers, and whether the timing of liquidity affects educational investment. The study is implemented in 30 government-run primary schools in Jinja District, Uganda, with a sample of 2,000 pupils completing Primary 4 or Primary 6 and their primary caregivers.

At baseline, caregivers complete an incentive-compatible preference elicitation exercise designed to measure willingness to pay for (i) a direct school-fee subsidy relative to unconditional cash delivered at the time school fees are due, and (ii) delayed cash delivered at the start of the school term relative to immediate cash. The elicitation uses adaptive multiple price list tasks in which caregivers make repeated choices between a fixed transfer of UGX 35,000 (approximately equal to Term 1 school fees) and varying alternatives. To ensure truthful reporting, a randomly selected 2 percent of caregivers receive a real payout based on a Becker–DeGroot–Marschak mechanism; these households are excluded from the main analysis.

After completing the preference elicitation module, households are randomly assigned at the individual level to one of four groups:
(a) Subsidy (Treatment 1): A direct payment of UGX 35,000 made to the child’s school at the start of Term 1, conditional on the child re-enrolling in the same school.
(b) Cash at Start of Term (Treatment 2): An unconditional cash transfer of UGX 35,000 delivered to the caregiver at the beginning of Term 1.
(c) Cash Now (Treatment 3): An unconditional cash transfer of UGX 35,000 delivered to the caregiver immediately following the baseline survey.
(d) Control: No transfer.

All cash transfers are delivered via mobile money, and transfer amounts are equal across treatment arms. The design allows comparisons between (i) receiving any financial support versus no support, (ii) subsidies versus equivalently sized cash transfers delivered at the same time to isolate the value of commitment, and (iii) cash delivered immediately versus at the school-fee deadline to assess the role of timing and saving ability.

The intervention builds on a pilot conducted in January 2025 in nine schools in Jinja District, which documented strong stated demand for school-fee subsidies but did not experimentally compare subsidies to cash. The full trial expands this design by randomizing across subsidy, delayed cash, and immediate cash arms, enabling causal tests of both effectiveness and underlying preferences for commitment and liquidity.

2025-12-13

2026-05-02

The study pre-specifies five primary child-level outcomes, measured during Term 1 and at endline:


(1) School fees paid (same school): Share of required school fees paid by the end of Term 1 at the original school, constructed using school administrative records and including all cash payments, in-kind contributions, and any subsidy received.
(2) Enrollment in any school: Indicator equal to 1 if the child is enrolled in any primary school at endline, based on caregiver report.
(3) Enrollment in the same school: Indicator equal to 1 if the child is enrolled at any point during the term in the same school attended in the previous academic year, measured using school administrative records.
(4) Attendance (same school): Proportion of unannounced enumerator visits during Term 1 in which the child is observed present at the same school.
(5) Learning Index: A composite standardized index combining literacy and numeracy performance from independently administered child assessments. The index is constructed as the average of standardized English and Numeracy scores and normalized using the control group distribution.

The study includes five primary child-level outcomes capturing learning and schooling investment. Constructed outcomes are defined as follows:

(1) School fees paid (same school): This outcome measures compliance with required school fees at the original school. For each child, all fee payments made during Term 1 are aggregated using school administrative records, including cash payments, in-kind contributions, and any subsidy received through the study. In-kind contributions are converted to monetary values using local market prices. The total amount paid is divided by the school- and grade-specific required fee amount to construct the share of required fees paid.

(2) Enrollment in any school: This outcome is a binary indicator equal to one if the caregiver reports that the child is enrolled in any primary school at endline, regardless of whether it is the same school attended in the previous academic year, and zero otherwise.

(3) Enrollment in the same school: This outcome is a binary indicator equal to one if the child appears at least once in the administrative enrollment records or class registers of the same school attended in the previous academic year during Term 1, and zero otherwise.

(4) Attendance (same school): Attendance is measured using unannounced enumerator visits conducted throughout Term 1. For each visit, enumerators record whether the child is physically present in the classroom. The attendance outcome is constructed as the proportion of visits in which the child is observed present among children who appear in the school’s records at any point during the term.

(5) Learning Index: The Learning Index is a composite measure combining children’s literacy and numeracy skills, assessed independently of schools using standardized instruments. Separate continuous scores are constructed for English literacy and numeracy based on performance across tasks of increasing difficulty. Each score is standardized using the control group mean and standard deviation, and the Learning Index is defined as the average of the standardized English and Numeracy scores. The index is re-standardized using the control group distribution.

Cognitive Outcomes:
(a) Expanded Literacy Score, a standardized English literacy measure combining Uwezo and higher-difficulty assessment items.
(b) Expanded Numeracy Score, a standardized numeracy measure combining Uwezo and higher-difficulty assessment items.
(c) Uwezo Literacy Z-Score.
(d) Uwezo Numeracy Z-Score.
(e) School-based examination outcomes, including a standardized composite School Learning Index and standardized subject-specific scores in English, Mathematics, Science, and Social Studies.

Schooling Outcomes:
(a) Official attendance rate, defined as the proportion of school days attended as recorded in class registers.
(b) School transfer indicator, equal to one if the child transfers to a different school during the study period.
(c) Caregiver-reported school fee payments, defined as the standardized total value of school-related payments reported by caregivers.

Socio-Emotional Outcomes:
(a) Resilience Index, a composite index based on child survey items capturing perseverance, belonging, and coping.
(b) CES-DC Scale, a standardized measure of depressive symptoms in children.
(c) Depression indicator, equal to one for scores above the CES-DC clinical threshold.

Food Security:
(a) Food Security Index, a standardized composite index based on child-reported food access questions.
(b) Food Insecurity Experience Scale (FIES), a standardized caregiver-reported measure of household food insecurity.

Child Labor:
(a) Child labor hours, defined as total hours spent in labor activities during the previous week.
(b) Child Labor Index, a standardized index capturing participation in different types of labor activities.

Educational Aspirations:
(a) Educational Aspirations Index, a standardized measure of the highest level of education the child aspires to attain.

All secondary outcomes and indices are standardized using the control group distribution where applicable. Multiple-hypothesis adjustments are applied within outcome families as specified in the pre-analysis plan.

The study will be conducted in government-run primary schools in Jinja District, Uganda. The target sample consists of pupils who were enrolled in Primary 4 or Primary 6 during Term 3 of the 2025 academic year, as recorded in official class registers. Focusing on pupils enrolled in Term 3 allows the study to capture dropout and enrollment decisions that occur during the transition between school years.

Thirty government primary schools will be selected from the set of eligible schools in Jinja District. Within these schools, a total of 2,000 pupils will be selected using stratified sampling by school, grade, and gender to ensure balance across treatment groups. Treatment is assigned at the individual (pupil/household) level.

Baseline data collection will take place during the school holiday preceding the new academic year. During this period, caregivers complete surveys and a structured preference elicitation exercise measuring tradeoffs between school-fee subsidies and unconditional cash transfers, as well as between immediate and delayed cash. Treatment assignment is revealed only after completion of the baseline survey and preference elicitation module.

Following baseline data collection, households are randomly assigned to one of four groups: (i) a direct school-fee subsidy paid to the child’s school at the start of Term 1, (ii) an unconditional cash transfer delivered at the start of Term 1, (iii) an unconditional cash transfer delivered immediately after the baseline survey, or (iv) a control group that receives no transfer. Transfer values are equal across treatment arms.

Endline data collection will occur during Term 1. Outcomes are measured using a combination of school administrative records, unannounced school visits conducted throughout the term to measure attendance, and independent child assessments and caregiver surveys administered at endline.

Not available

Schools and pupils are selected using computer-generated randomization conducted by the research team prior to data collection. Government-run primary schools in Jinja District are randomly selected from the set of eligible schools. Within selected schools, pupils are randomly sampled from Primary 4 and Primary 6 class registers using stratified random sampling by school, grade, and gender.

Once the study sample is identified, treatment assignment occurs at the individual (pupil/household) level. Pupils are randomly assigned by computer to one of four groups: (i) a school-fee subsidy, (ii) an unconditional cash transfer delivered at the start of Term 1, (iii) an unconditional cash transfer delivered immediately after baseline data collection, or (iv) a control group receiving no transfer. Randomization is stratified by school, grade, and gender to ensure balance across treatment arms.

Separately, to ensure incentive compatibility in the preference elicitation exercise, a randomly selected subset of caregivers receive a real payout determined by a random draw from the underlying valuation schedule implied by their responses. The realized payout corresponds to the respondent’s revealed preferences but does not necessarily coincide with any single choice shown during the elicitation. These households are excluded from the main treatment analysis.

Pupil level

No

0

2,000 pupils

500 pupils