The Impact of Cash and Food Transfers Linked to Preschool Enrollment on Child Nutrition and Cognitive Outcomes

Allocation: Randomized

This evaluation is conducted in communities surrounding Early Childhood Development (ECD) centers in the Karamoja sub-region of Uganda, in Napak, Kotido, and Kaabong districts. The catchment area of ECD centers serve as the clusters for randomization and analysis. After stratifying ECD centers (by district for Napak and Kotido, and by sub-district for the more spatially-diverse Kaabong), clusters were assigned to the treatment groups (FOOD, CASH and CTRL) using block randomization, to ensure as equal a distribution of ECD centers across treatment arms within each stratum as possible. Block randomization yielded 33 ECD centers in the FOOD group, 32 centers in the CASH group, and 33 centers in the CTRL group. Subsequently but prior to the start of the interventions, WFP re-assigned one CTRL center to FOOD, due to its proximity to a nearby FOOD center, in order to avoid migration of children which posed contamination concerns. WFP also re-assigned one CASH center to FOOD, because the cash distribution systems required use of mobile phones, and mobile phone signals were not readily available in the center's catchment area. The final assignment included 35 ECD centers in the FOOD group, 31 centers in the CASH group, and 32 centers in the CTRL group.

Statistical Analysis:

• Sample Selection and Sample Size: Prior to the baseline survey, power calculations were conducted using existing data from Uganda on cognitive measures (Mullens scores) and food security measures (share of food expenditures out of total expenditures) to estimate the necessary sample size for detecting a minimum effect size of 10 percentage points with 80 percent power. These calculations indicated a target minimum sample of 30 ECD centers per treatment arm and 20 households per ECD center. 98 ECD center clusters were determined eligible for transfer receipt by WFP. ECD enrollment data collected in August-September, 2010, were used to identify households with children ages 3-5 years enrolled in these ECD centers. In each of the 98 ECD centers, out of all households with children ages 3-5 years enrolled, roughly 25 households were selected at random for inclusion in the baseline sample.
• Data Collection: The evaluation uses a longitudinal design, with the same households interviewed at baseline (September-November, 2010, prior to start of the food or cash program) and at endline (March-May, 2012). In the intervening months, treatment households receive 15 months of exposure to the intervention. Direct assessments of children's cognitive development, anthropometry, and hemoglobin levels are conducted in a randomly-selected roughly 80% of sampled households in each cluster, due to field time constraints. Cognitive development is assessed using a battery of test items drawn primarily from the Mullen Scales of Early Learning, adapted by psychologists for the Karamoja context. Anthropometry is measured following standard protocol using height boards and weighing scales. Hemoglobin concentration is assessed using capillary blood obtained by finger prick and read using a HemoCue® analyzer. Hemoglobin readings are adjusted as necessary for the effects of varying altitude in the sample.
• Analysis: Differences in baseline characteristics are assessed using pairwise t-tests and binomial probability tests. Analyses of impacts focus on outcomes for children age 3-5 years (36-71 months) at baseline, as well as for children aged 36-53 months at baseline who remain in the targeted age range throughout the course of the intervention. Intent to treat (ITT) effects are estimated using single-difference and ANCOVA specifications on children at baseline and endline. Estimates are calculated using Stata 12; standard errors account for stratification and clustering in sampling and assignment.

Ethics Review:

• Side Effects: Potential risks from participation in this study are low. Many of the questions used in the survey have been used before in the Uganda National Household Survey. Cognitive tests consist of simple games played between a trained enumerator and the child and are unlikely to pose any risk to the children. Anthropometry measurements are non-invasive and also pose no risk. Risks associated with hemoglobin data collection are minimized through use of well-trained staff, a well-designed data collection protocol, and safety lancets. The study poses no social, emotional, or psychological risks to subjects.
• Health provisions: If any child is found to be severely anemic (Hb<7 g/dL) or wasted (weight-for-height less than two standard deviations below the mean), that child and the child's parents are notified and the child is referred to the nearest clinic for medical treatment.
• Informed consent: Details about the study are provided verbally to local leaders at the LC1 (roughly the equivalent of a village) level. The same information, including the right to opt out at any time, is provided verbally to survey respondents in sampled households. Additional informed consent is sought before each direct interaction with children, including for cognitive tests, anthropometry, and hemoglobin measurement.
• IRB: This study underwent internal review with the International Food Policy Research Institute; permission to conduct the data collection was granted by Uganda National Council for Science and Technology (SS 2470).

Funding:

This evaluation was funded by the World Food Programme, with support from the Spanish government and other sources through the Strategic Impact Evaluation Trust Fund at WFP, and by the United Nations Children's Fund.