Experimental Design
We have two main research questions: the first tests the role of information in eliciting contributions of labour towards a group public good, the second tests the additional impact of the identity of the messenger. To identify these effects, we will vary the assignment of SHGs along two dimensions. First, we will randomize whether the group receives SBCC on the importance of the public good. Second, among the groups that receive information, we will randomize the ethnicity of the messenger vis-à-vis the ethnicity of the SHG.
Our outcome of interest is the individual member’s willingness to contribute labour hours towards the creation and maintenance of a community nursery and kitchen garden- a public good. In our model the community nursery will be used to raise plants which can be taken home by the SHG women to be planted on their homestead. An attached community kitchen garden will be used to raise other vegetables to maturity, with the resultant produce shared among all SHG women. Both the garden and the nursery are models that PRADAN (the NGO we are partnering with) has developed for other interventions. Also, PRADAN, elicits voluntary contributions of labour hours for other projects, so women in these areas are familiar with these methods. Fruits and vegetables will be chosen keeping in mind local conditions, availability of food from other sources, nutrient values of foods, and local diets.
We want to exploit differences in caste groups between the messenger and the women in the SHG. In particular, we want to assess the impact of co-ethnicity in two separate scenarios - (1) if the group is “high” caste and the messenger is “low” caste, and (2) if the group is “low” caste and the messenger is “high” caste. We believe these two scenarios are likely to generate different behaviors on the part of the group. Table 1 (in Annex, attached in Docs & Materials) summarizes some of our hypotheses on the direction of the effects. Here a “high” SHG or agent means a group or agent who is of a higher caste (analogously for the term “low”).
From a preliminary scoping exercise in Baghmundi block of Puruliya district, we ascertained that the distinction between the Other Backward Classes (OBCs) and the Scheduled Castes (SCs) is stark, and the hierarchy is well recognized and agreed upon by community members. We have used these two groups as the high and low group in West Bengal.
Treatment groups: In order to identify the two types of variation, our research design has six distinct treatment categories. There are two sets of “pure control” groups that do not receive the SBCC treatment – “high” (“low”) groups that are then matched to a “high” (“low”) agent who simply describes the public good and conducts the WTC experiment, but does not provide any SBCC.
There are then four additional groups that receive the “information treatment”. Two of these groups will receive information from an agent who is of the same ethnic group as the group members – “high” (“low”) group matched to a “high” (“low”) agent. Two additional groups will receive information from an agent who belongs to a different ethnic group. Since both groups receive the same information, this will minimize concerns that the public good we have chosen is inherently more valuable to one group rather than another. All six groups are depicted in Table 2 (in Annex, attached in Docs & Materials).
This experimental design will allow us to isolate the impact of information on group members’ willingness to contribute labour hours, as well as the added effect of ethnic affiliations between the group members and the community agent. To illustrate, the difference in WTC between (H, H, info) and (H, H, noinfo) tells us the added effect of information, while the difference between (H, H, info) and (H, L, info) tells us the impact of the messenger being from a lower caste than the caste of the group. In addition, the difference between (H, H, info) and (L, L, info) tells us whether the good is valued differently by each of the two types of SHGs.