Study title

India

Household

For Siruguppa samples we applied the two-stage randomization procedure. From 25 GP in Siruguppa, 12 are randomly chosen and then are split equally into treatment and control. Randomization is done in Excel. A similar condition is applied in random sampling: none of the control and treatment GP should be neighbors. Figure 1 is a schematic depiction of the stages and the GP sample chosen in the process. In Siruguppa, we plan to have only one type of treatment viz. T2. Figure 4 depicts the location of 12 selected GPs in Siruguppa Taluka and their classification in treatment and control groups.Another interesting research question which can be addressed is the magnitude of spillover effect. It has been observed in the field that farmers often collect information from other farmers in the village. Thus, farmers may pass the information provided by e-SAP and our extension agent to others in the village and the recipients may benefit as well. To measure this indirect benefit, it is decided to take some additional farmers in each treatment GP. They will not receive any direct information from the project but they will be surveyed.The subsequent question was how would these additional farmers be chosen? Initially, it was thought that each selected GP, select four or five villages and for each village select farmers randomly. To capture internal (intra-village) spillover effect, for each treated village by selecting 2 additional farmers from the village itself. To capture external (inter-village) spillover effect, for each treated village select two more from outside the village. But, as GP in Siruguppa have a smaller number of villages (sometimes only one village) and they are bigger too. So, it is decided to select farmers randomly from the GP, ignoring his/her village of residence. Ideally, these farmers should be chosen simultaneously with the farmers receiving treatment. However, in practice, a randomly selected and identified farmer may refuse to respond to our survey. Considering that possibility, it was decided that those 10 farmers would be randomly selected after completing the baseline survey in a treatment GP. The data generated was based on recall method from household interviews using structured questionnaires.