Sampling Procedure
The trial was a pragmatic cluster (by village) randomized controlled trial with an additional matched control group. Sampling included 310 villages in the following 9 districts:
1. Bengkulu, Bengkulu Province
2. Sarolangun, Jambi Province
3. Lampung Timur, Lampung Province
4. Majalengka, Jawa Barat Province
5. Rembang, Jawa Tengah Province
6. Kulon Progo, DI Yogyakarta Province
7. Ketapang, Kalimantan Barat Province
8. Sidenreng Rappang, Sulawesi Selatan Province
9. Lombok Tengah, Nusa Tenggara Barat Province
It was planned that batch 1 would receive the first block grants at the start of the project. Block grants for batch 2 and batch 3 were to follow after nine and eighteen months respectively. A selection of villages was randomly allocated to either batch 1 or batch 3 (within each district), this sampling feature was used during the evaluation design.
Of the 310 villages, 100 were originally allocated to the intervention arm, 20 were originally allocated to a nine month delay staggered start, 100 were originally allocated to an 18 month delay staggered start, and 90 villages were allocated to a matched control group (no intervention).
Since this study aimed to track the same children over three rounds of data collection, the study team made every attempt to interview the baseline children (aged 12 to 23 months and the other aged 48 to 59 months at baseline) at the midline survey in 2010.
Deviations from the Sample Design
The most common reasons a child was unavailable for the 2010 midline interview were that the household/child moved (101+ children), the household was impossible to find (17 children), the child had died (9 children) or the family refused (5 children).
Response Rate
The resultant participation rates at both the district and village levels were 90% (extra villages were added in Middle Lombok, these have not been included in this calculation). At the child level, the participation rate was 99.92%. The retention rate at the child level at midline was 99.67%.
