Soutenir les Enfants à la Maison et à l'École

Intervention Factor 1: Unconditional cash transfer of €8 (5,245 CFA, local currency) every week for 100 weeks as well as group treatment coaching consisting in group support of the village savings and loan association to which the family belongs
Intervention Factor 2: PEC/DIA: Teacher professional development via “Teaching at the Right Level” program supplemented with electronic-coaching and support.
Control group: Village savings and loan associations

2021-04-11

2023-07-31

Outcome construct 1: Child Language and Literacy Skills
Outcome construct 2: Child Numeracy Skills
Outcome construct 3: Child labor

Outcome 1: Child Language and Literacy is operationalized as the index of scores on five measures of emergent literacy and literacy skills adapted from the Early Grade Reading and Math Assessments (EGRA and EGMA; RTI International, 2009a, 2009b) and the International Development and Early Learning Assessment (IDELA; Pisani, Borisova, & Dowd, 2018) inventories:
Initial sound suppression: Children repeat a word without the first phoneme (initial sound), e.g., “pomme” becomes “om” (adapted from EGRA)
Initial sound matching: Children hear a target word and select another word (between three options) that starts with the same sound. (adapted from IDELA)
Antonyms: Children hear a word and generate a word with the opposite meaning (adapted from Woodcock Johnson Test of Cognitive Ability)
Knowledge of letters and groups of letters: Children give the name or sound of graphemes and grapheme clusters in 60 seconds. Score is the number of items correctly named (out of 100 possible). (adapted from EGRA)
Familiar words: Children read familiar French words for 60 seconds. Score is the number of correctly read words (out of 50 possible). (adapted from EGRA)

Outcome 2: Numeracy is operationalized as the index of scores on six measures of numeracy concepts and skills adapted from the EGMA, IDELA, and CTOPP inventories:
Number Identification: Number of single-digit numbers child knows (1-9)
Number Order: The child sees a line of numbers and selects the smallest number (adapted from EGMA)
Complete Number Sequences: The child looks at a line of numbers and responds which one is missing, e.g., 3,_,5,6. (adapted from EGMA)
Addition: Child performs simple addition problems for 60 seconds (adapted from EGMA)
Subtraction: Child performs simple subtraction problems for 60 seconds (adapted from EGMA)
Counting / One to one correspondence: Child views sets of objects and must select the correct set containing a certain quantity, e.g., ‘Find 8 beans’ (adapted from IDELA)

Outcome 3: Child Economic Activities is operationalized as the index of scores on surveys relating to children’s engagement in different economic and farming activities. We use a one-on-one questionnaire with children that was implemented in the recent NORC report (Sadhu et al., 2020). We ask children to respond “yes” or “no” to whether they have participated in the activities described in the survey within the last 12 months. We will create an index variable to include all three types of activities, and conduct post-hoc impacts on each of the three separately.
a. Sum of all agricultural activities (non-hazardous) to which the child responded “yes” (all items except hazardous activities). Complete list of questions is included in the survey materials attachment.
b. Sum of all hazardous activitieies. Hazardous agricultural activities are grouped in six categorie: (1) Conducting land clearing, (2) Carrying heavy loads, (3) Using agro-chemicals; (4) Using sharp tools, (5) Engaging in long working hours, or (6) Engaging in night work.
c. Sum of the domestic and other economic activities to which the child responded “yes” (e.g. laundry, child care, cooking, cleaning, selling at the market, working on a construction site).

Secondary Outcome 1: Child School Participation
Secondary Outcome 2: Children’s non-academic outcomes
Secondary Outcome 3: Family investments in education
Secondary Outcome 4: Family well-being
Secondary Outcome 5: Teacher knowledge on TaRL practices
Secondary Outcome 6: Teacher professional well-being

Secondary Outcome 1: Child School Participation is operationalized as the index of scores on questions about the number of days of school that a child missed in the last week for economic or other reasons, as well as random polling of teachers for child attendance.
Children self report the number of days of school they have missed in the past week due to domestic activities, economic activities, or agricultural activities: to learn a job, to work for pay, to help in a plantation or family business, to do field work, to do domestic work
Children self report the number of days of school they have missed in the past week due to: sickness, injury, physical inability, school too far, inability to pay school fees, parents did not want child to go to school, weak in studies, lack of interest, education considered worthless, lack of security in school, bad weather conditions, family emergencies, travel (with parents), obstacle on the way to school,
Children self report the number of days of school they have missed in the past week because the school was closed or the teacher was absent.
Children self report the number of days in the past week in which they attended school in the morning but did not return to school for the afternoon lessons.

Secondary Outcome 2: Children’s non-academic outcomes. Each domain will be examined separately as a scale.
Motivation (child-report using the Elementary School Motivation Scale (ESMS; Guay, Marsh & Dowson, 2005)
Social-emotional skills (including stress management and empathy measured using the International Social-emotional Learning Assessment (ISELA); Pisani et al., 2018)
Children’s executive function, measured using the Hearts and Flowers task (Diamond et al., 2007) and the forward and backward digit span.

Secondary Outcome 3: Family investments in education. We will create an index of the following scales:
Parent involvement in child’s educational activities
Cognitive stimulating activities with children
Parents’ perceptions of educational value and opportunity
Parents’ expectations and beliefs about children’s school attendance

Secondary Outcome 4: Family well-being. We will create an index of the following scales:
Parental stress ratings
Household wealth
Women’s empowerment and bargaining power
Maternal mental health

Secondary Outcome 5: Teacher knowledge on TaRL practices. Teachers will respond to a survey assessing their knowledge of the TaRL content, implementation practices, and purpose behind the approach. We will create an index to assess one scale of teacher knowledge.

Secondary Outcome 6: Teacher professional well-being. Teachers will respond to a survey assessing their professional well-being in three domains. Each domain will be examined separately as a scale score.
Motivation
Burnout (measured by the Maslach Burnout Inventory (MBI))
Job satisfaction

Data collection is stratified across regions, clustered by communities (villages). The sampling plan has three levels:
Level 1: Child and Family (15 per community)
Level 2: Community or school
Level 3: Region

1. Cash transfers only (34 communities)
2. PEC/DIA: Technology-supported teacher professional development via “Teaching at the Right Level” program, and coaching control (34 communities / schools) (34 communities)
3. Cash transfers + PEC/DIA (34 communities)
4. Control (34 communities)

To enhance our power for RQ2, 58 additional communities will be randomized to one of these two treatment conditions:
1. PEC/DIA (29 communities)
2. Control (29 communities)

In the event that a community has more than one one school, all the schools in the community are treated with the education intervention and one of them is randomly selected for data collection.

UCT: Random selection of communities performed by the research team; In each community, the poorest households are nominated by the local selection committee, made of village leaders. To be eligible, the family must be recognized to be among the neediest families in the village, with a child aged 5-15 living in the household. This latter additional condition was included to account for the education component of the intervention. We sample 15 families per community ("HOME SAMPLE") and survey both the family head and one focal child in the school age range.

PEC/DIA: Communities randomly assigned to TaRL receive the intervention. In each school, we randomly sample a total of 15 focal kids ("SCHOOL SAMPLE") to measure the impact of the education intervention.

Randomization unit is Level 2: Community

Yes

Cash transfer treatment contrast: 136 communities
PEC/DIA treatment contrast: 194 communities
Combined treatment contrast: 68 communities

Cash transfer contrast: Within each community, we will sample 15 children (and their 15 mothers) in the "HOME SAMPLE". Approximately 2,040 parent-child pairs. TaRL treatment contrast: Within each community, one school will be randomly selected for data collection. Within each school, 2 teachers will be randomly sampled, and 15 children in the "SCHOOL SAMPLE". (Note that 58 communities will only have a "school sample" and be included in this treatment contrast). Approximately 2,910 children. In total, there will be approximately 7,378 observations (4,950 children, 2,040 mothers, 388 teachers).

1. Cash transfers only (34 communities)
2. PEC/DIA (63 school communities)
3. Cash transfers + PEC/DIA (34 communities)
4. Control (34 communities + 29 school communities)

Our power analysis assumes 80% power at the 5% significance level. We test two samples for treatment impacts: first, focal women and children who are selected through the 100WEEKS criteria to be eligible for cash transfers (“the vulnerable sample”). For RQ1 and RQ2, with the “vulnerable sample”, our assumptions are: • Clustering of families within communities, assuming an ICC of 0.10 at the community level for child and parents’ outcomes • R-squared of 0.4 at the individual level (given baseline data collection). • 68 communities per individual treatment arm (J = 136 for each treatment contrast). In other words, we would have 68 communities who received cash transfers (PEC/DIA respectively), and 68 who did not receive cash transfers (PEC/DIA respectively). • 15 families per community (n = 136 * 15 = 2040) This is sufficient to detect effect sizes of 0.17 standard deviations (minimum detectable effect size, MDES) for the focal parent and child outcomes for (i) cash transfer communities, and (ii) PEC / DIA communities. For the combined treatment impacts (RQ3), where J = 68, n = 1020 (15*68), our MDES = 0.21. Second, we will also sample an additional 15 children per school to enhance our power to examine impacts of the PEC/DIA program at the school and village level, given that it is a school-wide program (“School sample”). We will also add an additional 58 communities that focus exclusively on the implementation of PEC/DIA and the school sample of children (called “school communities”). Here, our assumptions are: • We keep our clustering at the village level and now assuming an ICC = 0.15 for child learning outcomes • 97 communities per treatment arm (J= 194; 136 communities + 58 school communities) • Combining focal children and school-sampled children, n = 4950 (15 children from the cash transfer group, plus an additional 15 per school) [15 * 194] This is sufficient for an MDES = 0.15 for impacts of PEC/DIA program versus control (RQ2). Our MDES is almost the same as with the “vulnerable sample” because (i) ICC is assumed to be slightly higher for this sample and (ii) more importantly, increase in power is faster when we increase cluster rather than individuals in clusters. For the combined treatment impacts (RQ3), where J = 68, and n = 2040 (30*68), our MDES = 0.24.