Cheap Talk and Coordination in the Lab and in the Field: Collective Commercialization in Senegal

We worked with 79 groundnut-producing cooperatives comprising close to 2800 individuals. These cooperatives are part of two umbrella federations (i.e., conglomerates of cooperatives) in the Bassin Arachidier, which is the main groundnut production zone of Senegal. From November to December 2013, two leaders of each farmer group attended a two-day training conducted by two development specialists. The training focused on the potential, pitfalls, and conduct of collective commercialization; in particular, strategies for identifying distant buyers, negotiating prices, and organizing transportation. Participants were instructed to conduct a briefing meeting with all cooperative members upon returning to their village, to report the gist of what was covered during the training. Trainees were also provided with standardized booklets to keep records of each member's contribution to the group's sales in the upcoming commercialization season. A reward of 10,000 CFA (approximately 25 US dollars) was promised for filling in the booklets with all the requested information. All groups eventually received such reward.

After the training, during January 2014, enumerators went to the villages in order to elicit commercialization intentions from all cooperative members. Prior to doing so, they made sure that the leaders who had taken part in the training had held the "briefing" meeting. For each farmer group, all members who produced groundnuts for the 2014 commercialization season were asked how they intended to use their production. They had to split their anticipated harvest into (1) individual commercialization, (2) collective commercialization (via the cooperative), (3) inventories, and (4) other uses. They were told that the purpose of this survey was to better understand their decisions with regard to groundnut production. They were also informed that a subsequent group meeting would be held, where a message would be delivered to them. They were thus invited to attend that meeting.

The 79 groups were randomly allocated to one of the following four conditions, depending on the information that would be disclosed in the subsequent meeting:

First, in Condition A (the control group), members' intentions were not revealed. Enumerators announced that a follow-up survey would be conducted after the end of the commercialization period. This was also announced in Conditions B, C, and D.

Second, in Condition B members' aggregate intentions were revealed.

Third, in Condition C, members' aggregate intentions as well as the distribution of intentions among members were revealed. I.e., how many members intended to contribute 100kg; how many intended to contribute 200kg, and so on. This was most comparable to the cheap-talk treatment in the lab-in-the-field experiments (LFEs). See experimental design discussion further below.

Finally, in Condition D, the same information as in C was revealed, but the distribution of intentions was disaggregated by ordinary members versus cooperative leaders (defined as all members of the management committee which typically comprises 8 to 12 individuals).

2013-05-01

2014-07-31

Quantity of groundnuts sold to the farmer cooperative in kilograms; revenues from quantity of groundnuts sold in FCFA; contributions to the group in LFEs.

As previously mentioned, the 79 groups were randomly allocated to one of the four conditions (A-D) described previously. An average of 10-12 randomly selected farmers per cooperative were surveyed before (time of eliciting intentions/cheap talk) and after commercialization. In addition, prior to the RCT, randomly selected members of 28 groups (26 of which are among the above 79) completed LFEs. Since those LFEs are not part of this registry, the experimental design details are not being included here. However, they are in the working paper and instructions at the following website, www.angelinoviceisza.com/papers.

Since the trial has already been finalized, this information is the same as the former.

The randomization was done in office by a computer.

For the RCT, the main unit of randomization was the group/farmer cooperative level. For the LFEs, the main unit of randomization was the individual/farmer/group member/subject level. When collecting the survey data before and after the RCT intervention, the main level of randomization was the individual/farmer/group member level.

Yes

79 farmer cooperatives.

The total number of individual responses to surveys is close to 900 in the RCT. In the LFEs, due to multiple rounds this number varies between 2000 and 3400 depending on the specification.

Condition A: 17 cooperatives, Condition B: 21 cooperatives, Condition C: 20 cooperatives, Condition D: 21 cooperatives.

Given the study has already been conducted, we have calculated the empirical equivalent of ex-post power. In the RCT, power is 0.99. We have also calculated the post-study probability according to Maniadis et al. (AER, 2014). For priors beyond 0.1, researchers can be fairly confident that our findings are accurate (i.e., with post-probability greater than 0.5). Additional details are in the working paper at www.angelinoviceisza.com/papers.

Coordination is central to social interactions. Theory and conventional lab experiments suggest that cheap talk/communication can enhance coordination under certain conditions. Two aspects that remain underexplored are (1) the interaction between the number of players (group size) and communication and (2) how existing findings might play out in the field. We address both of these by studying a typical naturally-occurring setting that requires coordination; that is, one where members of agricultural cooperatives seek to jointly sell their output. Combining artefactual/lab-in-the-field experiments (LFEs), natural field experiments (RCTs), surveys, and cooperative records, we find that (1) revealing farmers' intended sales (i.e., cheap talk/communication) yields enhanced collective commercialization (i.e., coordination), particularly in larger groups; (2) such cheap talk may lead to higher incomes for small-scale farmers; (3) participants transfer learning from the LFEs thus affecting subsequent behavior in the RCTs (i.e., the day-to-day environment). Our results contribute to existing literature by highlighting the potential for cheap-talk institutions to (1) boost coordination, particularly in settings with greater strategic uncertainty (e.g., larger farmer cooperatives), and (2) promote collective entrepreneurship and development.

Aflagah, Fo Kodjo Dzinyefa, Tanguy Bernard, and Angelino C. G. Viceisza. 2019 "Cheap talk and coordination in the lab and in the field: Collective commercialization in Senegal." NBER Working Paper (forthcoming)

https://www.angelinoviceisza.com/papers