Too spooked to talk?

Sharing information can feel spooky at times. Especially when you are not exactly sure you are correct and the stakes are high—besides finding you mildly irritating, would your friends and family hold you responsible for sharing bad advice later?

This is the mechanism Chandrasekhar, Duflo, Kremer, Pugliese, Robinson, and Schilbach go after in their recent paper, ``Blue Spoons: Sparking Communication about Appropriate Technology Use.’’ In this paper, they explore the possibility that being too spooked to talk may in part explain the low levels of communication about farming techniques observed in rural Kenya during previous rounds of experiments promoting the application of the correct amount of fertilizer at top dressing (Duflo, Kremer, and Robinson, 2008). This is perhaps what I like the most about the blue spoon experiment: it tells a compelling story of iterative learning over multiple cycles of experiments with farmers. First, the research team used experiments to figure out the right amount of fertilizer at top dressing (Duflo et al, 2008). Then they experimented with free delivery (Duflo, Kremer, and Robinson, 2011). Over the course of these first two experiments, they noted that, despite having huge returns, knowledge about the quantity of fertilizer to apply was not traveling very fluidly through the social network. Because these past rounds of experiments were not set up to understand the cause of these low levels of communication about the promoted technology, the research team set up the blue spoon experiment we are about to dive into.

A simple and transparent technology

So, what is this blue spoon all about? The key idea in the paper is that information that’s simple and transparent is more likely to travel, as reputation costs associated with sharing such information are likely lower. To ease transmission of knowledge on the amount of fertilizer to use along those lines, Chandrasekhar et al employ a kitchen spoon with a blue painted handle that measures the recommended quantity of fertilizer to apply in a planting pit. The spoons are distributed with the information that the spoon measures the correct quantity of fertilizer to apply in a planting pit. This is not a type of spoon that was available before, making it novel and easy to track adoption. A number of local stores source this new spoon over the course of the experiment, and the blue spoons are distributed with clear information on where to purchase a blue spoon in the village.

Is farmers not talking about farming consistent with a simple model of technology diffusion?

Chandrasekhar et al first consider that low levels of communication about farming may be consistent with a simple model of social learning in which farmers’ decision to join the network is based on their expected returns (is this going to increase their yields?) and their costs of participation (talking to people takes effort). A simple way to test predictions from this model is then to ask:

• 1/ Are farmers more likely to take part in information sharing as returns to or quality of the information go up?
• 2/ Does farmers participation in the network increase as communication costs go down?

To test these predictions, Chandrasekhar et al implement a large-scale RCT with maize farmers across 184 school catchments in rural Kenya. To test the first prediction, they randomize access to a voucher for fertilizer (increasing returns) and access to blue spoon (increasing quality). To test the second prediction, they set up farming coop that encourage farmers to share knowledge about farming and fertilizer use in a group (lowering communication costs). They also implement an active control group in which farmers meet to discuss issues unrelated to fertilizer use. All treatments are assigned at the school catchment level, except the blue spoon which is randomly assigned to 15 percent of farmers in each school catchment (so-called blue spoon farmers), including the control. The general idea is to track whether coops and vouchers may encourage farmers to share information about the blue spoon. After verifying that all treatments did what they were supposed to do, the authors can test how they affected diffusion of blue spoons to friends of blue spoon farmers and non-blue spoon farmers.

Blue spoons spread like fire while cooperatives and vouchers do not increase learning through the network

Knowledge and ownership of blue spoons increase sharply among both friends of blue spoon farmers and non blue spoon farmers, across all treatments. Yet, inconsistent with the simple model of technology diffusion, knowledge about the correct amount of fertilizer to apply at top dressing does not increase in cooperatives and voucher groups. Interestingly, while farmers in the cooperative treatment are more likely to know and possess a blue spoon, they do not seem to have more accurate knowledge of the recommended amount of fertilizer! In line with this, Chandrasekhar et al show that blue spoon farmers are less likely to talk to their friends about farming in cooperative groups. 

While the blue spoon treatment sharply increases communication about farming and fertilizer, cooperatives and vouchers do not—in other words, lack of interest in learning about fertilizers or high communication costs cannot explain why information failed to travel through the network in the first place. The overall stock of knowledge about fertilizer use is in fact lower in cooperative groups: people did not stop talking with their friends because knowledge was full. Hence, Chandrasekhar et al reject the simple model and propose an augmented version to rationalize their findings.

Sharing is costly (sometimes)

They augment the simple model with two added features: 1/ not all information is made equal—some is more reliable than other; and 2/ farmers are reputation concerned, learning them to hoarding information when sharing could backfire. In their model, some farmers are more concerned than others about spreading falsehoods and, thus, more or less likely to share unreliable information through the network. In this model, less concerned farmers are more likely to spread information regardless of reliability, while more concerned farmers are more likely to share as reliability of information increases. The blue spoons help information about fertilizer quantity diffuse faster, as more concerned farmers feel pretty good about spreading this reliable information. 

Things change a bit with the cooperative treatment: as more people observe the exchanges, concerned farmers get even more concerned about sharing! As less concerned farmers lower the overall stock of information in the network, more concerned farmers are less likely to share, even if they hold higher value information, for fear of being held responsible for the overall low quality of information in the network. This helps explains why reliable information may not travel well in general, and specifically why knowledge about fertilizer quantity may be lower in the cooperatives arm, even though blue spoons diffuse even faster. 

What can we do about this?

Clearly, targeting demonstration activities to more reliable farmers is hard to do at scale, so this is not the main takeaway from this paper. Instead, Chandrasekhar et al push us to think about spreading technologies to as many farmers as possible to remove the social filter on diffusion. Their results also show that simple and transparent technologies such as the blue spoon spread fast—this is an invitation to think of ways to improve the way we transmit information, while allowing more farmers to acquire first-hand experience with new technologies.