Nudging parents and teachers to improve learning and reduce child labor in Cote d'Ivoire

The study addresses whether low-cost, mobile message-based nudges to parents and/or teachers can improve children’s literacy and numeracy and reduce child labor in rural cocoa-farming communities in Cote d’Ivoire, where enrollment has risen but learning remains low and child labor is widespread. Drawing on behavioral economics, the intervention aimed to overcome informational, salience, cognitive load, and identity-related barriers to parental investment; and motivation and pedagogical support constraints for teachers. The context features low adult literacy, large classes, low teacher support, and high opportunity costs for children. The study posed four preregistered research questions: (1) Impacts of messages to parents on children’s literacy, numeracy, and child labor; (2) Impacts of messages to teachers on these outcomes; (3) Impacts of messages to both parents and teachers; (4) Heterogeneous impacts by child sex, baseline ability, and parent education.

Evidence from high- and middle-income settings shows SMS nudges can improve parental engagement and student outcomes (e.g., United States, Brazil, Costa Rica), though effects and contexts vary. In sub-Saharan Africa, SMS combined with phone calls or meetings (Botswana, Zambia) improved foundational skills, but pure SMS evidence is scarce. Teacher-focused mobile messaging has mixed evidence; while SMS/calls can complement teacher training and motivation (e.g., Niger), disentangling their standalone effects is rare. The Ivorian context differs substantially from settings with positive results (e.g., literacy rates, absenteeism), making external validity uncertain. Gender and baseline-ability heterogeneity, and parental misperceptions of children’s abilities, suggest potential differential effects by subgroup.

Design: School-randomized controlled trial in two regions (Aboisso and Bouaflé), Cote d’Ivoire. Target grades: CP2 (grade 2) and CE2 (grade 4). Sampling: 100 public schools selected by district offices (one school later inaccessible; 99 schools analyzed). From each school, 25 students were sampled (13 CP2, 12 CE2) for child assessments; their parents were surveyed at home. Baseline: Sept–Oct 2018; Endline: May–Jun 2019. Preregistration: AEA RCT Registry (AEARCTR-0003385); ethical approval: University of Zurich OEC IRB #2018-035. Intervention (Eduq+): Two messages per week (Dec 2018–Jun 2019) following a behavioral sequence (motivating fact, suggested activity, interactive prompt, growth message). Parents: simplified French SMS or audio messages in local language; content on social-emotional support and school engagement (no curricular knowledge required). Teachers: French SMS on pedagogy (e.g., group work in large classes, discourage corporal punishment, support low performers) and motivation. Arms: (i) Parents only (text or audio; pooled), (ii) Teachers only (text), (iii) Parents+Teachers, (iv) Control. Original assignment included five arms (parent text; parent audio; parent text+teacher text; parent audio+teacher text; control). For analysis, parent delivery modes were pooled yielding three treatment indicators. Sample sizes: Baseline N=2475 children; Endline N=2246 (90.6%). Parents N=2475 baseline; +25 at endline. Schools: Parents-only 26; Parents+Teachers 24; Control 50 (with 99 accessible schools total at endline). Outcomes: Primary learning outcomes were standardized literacy and numeracy indices using EGRA- and IDELA-based tasks (a≈0.85 literacy; a≈0.86 numeracy). A learning summary averaged standardized literacy and numeracy within grade relative to control. Child labor: child-reported index of four activities (≥1 hour in past month): domestic chores/caregiving; work on family fields/garden; construction/major repairs; cocoa plantation work (child-reported emphasized for accuracy). Mediators: Parental autonomous motivation (9 items; a=0.81), social-emotional engagement (9 items; a=0.72), school/educational engagement (9 items; a=0.78). Covariates: Baseline grade fixed effects; child sex; baseline standardized learning, parental engagement, student effort, child labor, socio-emotional skills, working memory, visual attention, impulsivity, self-esteem, mindset. Analysis: ANCOVA with endline outcomes regressed on treatment indicators, baseline outcome values, covariates; SEs clustered at school; grade fixed effects. Attrition handled via inverse propensity weighting derived from control-group tracking probabilities predicted by baseline characteristics. Heterogeneity by sex, baseline learning (below/above median), and parental education (any vs none). Power: With 80% power and α=0.10, MDES ≈0.18 SD for learning summary and 0.22 SD for child labor.

- Primary outcomes: Parents-only arm showed positive but statistically insignificant effects on the learning summary (d≈0.081, p=0.158) and similarly sized, non-significant effects on literacy and numeracy (p=0.236 and 0.108). Teacher-only and Parents+Teachers arms had small negative, non-significant learning effects (d from −0.019 to −0.073). Control group learning gain over the year was ≈0.33 SD, so the parents-only point estimate is meaningful (~one-quarter of a school year), though imprecisely estimated. - Child labor: Parents-only arm led to a marginally significant increase in child-reported labor (d≈0.113, p=0.091). Other arms showed no effect; point estimates near zero. Comparing treatment arms showed a larger increase in child labor in Parents-only vs Parents+Teachers (difference d≈0.138, SE=0.076, p=0.072). - Heterogeneity: (a) Baseline ability: Larger positive learning impacts for children below the median at baseline in the parents-only arm (e.g., literacy positive and marginally significant). (b) Sex: Teacher-only had negative effects on girls’ learning; significant interaction for teacher-only with sex on the learning summary (B≈−0.13, p<0.05). Overall effects more positive for boys. (c) Parent education: No consistent differential impacts by parental schooling. - Mechanisms: No significant average impacts on parental autonomous motivation, social-emotional engagement, or school engagement. No differences between text vs audio delivery for parents. - Cost: Approx. USD 796 per school, USD 6.68 per student (development and operations included). Scaled implementation could lower costs substantially (e.g., bulk/free SMS via ministry/carriers or RapidPro).

The trial indicates that standalone SMS/audio nudges in a rural West African setting yield limited average improvements in learning and may inadvertently increase child labor (notably domestic/construction tasks) when targeting parents alone. Nonetheless, parents-only messaging showed more favorable patterns than teacher-only or combined arms, particularly for lower-performing children, suggesting salience and simple guidance can help engage parents to support the most struggling learners. Teacher-only messaging did not improve outcomes and was detrimental for girls’ learning in some analyses, implying message-based support alone is insufficient and might interact negatively with classroom dynamics or perceived monitoring. Absence of effects on measured parental mechanisms and no difference between text versus audio suggest either limited engagement pathways were captured or that additional supports (e.g., calls, in-person components) are needed. Given modest control-group learning (≈0.33 SD/year), even small effect sizes could be meaningful if made reliable; however, precision and power constraints limit firm conclusions. The increase in child labor, concentrated in domestic activities rather than cocoa-specific tasks, may reflect shifts in parents’ perceptions of children’s capabilities or trade-offs in time use. Accurate measurement of child labor remains complex; child-reported indices were used based on prior validity evidence. Overall, demand-side nudges may need to be coupled with supply-side teacher professional development to translate into consistent learning gains and to avoid unintended consequences.

This study contributes experimental evidence on parent- and teacher-targeted SMS/audio nudges in rural Cote d’Ivoire. On average, messages to parents alone produced small, non-significant learning gains and increased child labor; teacher-only and combined arms showed no learning benefits, with negative effects for girls in the teacher-only arm. Benefits were larger for children with low baseline skills, pointing to potential equity gains if interventions are optimized. Future research should: (1) integrate nudges with more intensive teacher professional development and/or parent support (e.g., coaching, calls); (2) unpack message components, timing, frequency, and interactivity to identify active ingredients; (3) deepen measurement and theory of parental mechanisms and how identity, salience, and cognitive load shape behaviors; (4) rigorously measure types, frequency, and hazardousness of child labor and its interplay with schooling; and (5) explore gender-sensitive designs to avoid adverse impacts on girls.

- Statistical power: Four-arm school-level randomization left the study underpowered to detect small effects typical of SMS nudges; observed learning effects (~0.08 SD) were meaningful but imprecise. - External validity: Conducted in two regions and in non-randomly selected schools; results are not nationally representative of all cocoa-growing areas. - Mechanisms: Limited mediator set and measurement restrict inferences about pathways of change. - Intervention components: Unable to isolate effects of specific message components (motivating fact, activity, interactivity, growth) or delivery parameters (frequency, timing, intensity). - Outcomes scope: No child-level data on repetition/dropout; child labor measures were brief and did not capture frequency/duration or hazardousness, constraining interpretation of labor-schooling trade-offs. - Potential interactions: Combined parent-teacher nudging may have unintended effects (e.g., on teacher motivation/attendance) not fully captured within this study.