Experimental Design
At the school level, we cross-randomized farmers at the school level into two main treatment conditions. First, to lower costs of communicating and sharing knowledge about fertilizer and other agricultural practices, we encouraged farmers to form farmers' cooperatives to talk about agriculture. While we facilitated organizing the groups and coordinated the fi rst few meetings, we did not provide any information directly to farmers.
Second, in previous work, we fou nd that providing farmers with small incentives to invest in fertilizer when they have money (right after harvest) can substantially increase usage (Duflo et al. 2011). Hence, to increase usage exogenously, we implemented a scalable version of a program to provide farmers with small, time-limited discounts which were valid within a short window (3 to 4 weeks) right after harvest, redeemable at a local shop. Farmers received coupons for a discount of 15% of the price of fertilizer. The coupon was valid for discounts to either diammonium phosphate fertilizer (DAP), used at planting, and calcium ammonium nitrate fertilizer (CAN), used at top dressing, when the maize plant is knee high, approximately one to two months after planting. Farmers could choose any combination of DAP and CAN up to 25 kilograms in total. Moreover, to evaluate the e ectiveness of text message reminders, we randomly selected half of the schools that received fertilizer discount coupons into a text message reminder program. In these schools, we randomized a subset of individuals who either owned a cellphone or had access to a cellphone to receive a text message reminder two days before the expiration date of their time-limited discount.
At each school meeting, we asked about 25% of farmers in each school to provide names and contact information of up to 3 individuals outside their own household with whom they discussed agriculture whom we refer to as agricultural contacts" below. About two thirds of these individuals were randomly selected to receive a "Bluespoon," a half-teaspoon, painted blue, which farmers could use to apply fertilizer to their maize. We chose this simple technology because in earlier work we had found 1/2 teaspoon of CAN to yield the highest profits on average among four di fferent quantities (Duflo et al. 2008). We delivered these spoons in an additional short meeting
(at the same schools) to which we had invited the randomly selected subset of farmers. In addition, farmers were given the information that in earlier work we had found that this quantity of CAN resulted in the highest pro ts on average. Again due to logistical constraints, the timing of these meetings was randomized. Bluespoons were also made available to anyone for a nominal
fee ($0.05) in the local market center at the same shops which handled the coupon redemption.