Intervention (Hidden)
To conduct the experiment, we partnered with BRAC Uganda, a large non-profit organization. BRAC has a network of community health providers (CHPs) from which we recruited our “agents”. Effectively, there is one CHP per village; and prior to our intervention, these CHPs worked as vendors of health related consumable goods such as soap, sanitary pads, and malaria pills that they acquired from BRAC. We visited 8 BRAC branches in rural Uganda.
The trial in the first four branches ran from April 2013 to April 2014 and in the second wave of 4 branches from January to June 2014. Each branch was selected based on having low penetration of grid connections and limited distribution of low-cost solar lights. A BRAC branch has a few dozen microfinance groups, each with 20 or so women. We divided each branch into 4 zones, each of which typically had 10 or more microfinance meetings.
A BRAC credit officer escorted us to four microfinance group meetings per zone. The meetings were geographically dispersed so that each vendor would have a catchment area of 200 or so households. Our goal was that each catchment area would have enough residents to support one vendor. At the microfinance meetings we presented the solar lights and explained we were recruiting vendors to sell these lights. We provided vendors with several different models of lights. The most popular model was the Firefly Mobile, produced and distributed by Barefoot Power. The FireFly Mobile easily provides enough light for reading and studying as well as completing basic household chores. It is also capable of charging mobile phones.22 The FireFly Mobile was packaged in a small box which included a solar panel, a lamp with a built-in battery, necessary cables, and multiple adapters to enable charging for the most popular mobile phones in the region. We invited one woman per microfinance meeting to a recruitment meeting. If more than one woman at a recruitment meeting was interested, we gave preference to one who had access to SMS text messaging. In a few cases we asked the BRAC credit officer privately for a recommendation. If a recommendation from the credit officer was not possible, we selected the vendor who first expressed interest. During the recruitment meeting, we trained vendors on how to use the lights and informed them of their economic benefits. Our training involved explaining features of the light, the operation of the light, and the terms of the retail sales offer (such as the one-year warranty). We showed vendors how the light can save customers money, where the savings on kerosene can quickly sum up to more than the cost of the light.
We anticipated that many of the vendors’ customers would be liquidity constrained. Thus, we explained to the vendors the advantages of several sales offers that overcome liquidity constraints: layaway (which Guiteras et al. (2014) found worked well selling water filters in Bangladesh), installments (which Levine et al. (2013) found worked well selling cookstoves in urban and in rural Uganda), and selling via a rotating savings and credit association (ROSCA). In a ROSCA, a group of customers pool their funds each meeting to purchase one light. The group continues until all customers have purchased a light. ROSCAs are common in this part of Uganda.
After completing training, we gave each interested woman a solar light. She was then asked to pay for the light with mobile money over the next several weeks. The purpose of this exercise was two-fold: first, we hoped to familiarize potential vendors with the type of sales offer to use with their own customers and second, to partially screen out vendors that were unlikely to perform well. All vendors completed payments, though several of them took longer than was originally specified.
After the field staff made invitations and prior to the recruitment meeting, we randomized half of the zones to receive an arrangement that included a line of trade credit and the other half of the zones received an arrangement that did not include trade-credit line. We offered all vendors the right to purchase lights at a fixed price during the first meeting or at any point in the future. In order to test Hypothesis 2 and Hypothesis 3, we conducted two additional (and orthogonal) randomizations. First, we randomized over whether the agent was provided the right to return unsold inventory. Second, we randomly selected a subset of 23 the vendors to which we provided a loaner light. The light was clearly labeled as “Property of BRAC” and the vendor was instructed to use the loaner light in order to give potential customers a free-trial period.
We held a separate recruitment for all of the potential vendors who received the same vendor arrangement (that is, all the women in the same zone who had the same arrangement). At the recruitment meeting we discussed strategies for how to sell lights. We emphasized offering the customer a free-trial period and time payments. We then introduced the vendors to their designated arrangement. We also emphasized to vendors that the arrangement could continue as long as they were successful and that by reinvesting they could grow their business eventually generate substantial income. At the end of the recruitment meeting we took initial orders. Vendors who were not allocated credit were required to pay cash for their initial orders upon delivery, while vendors treated with a credit line paid cash only for units in excess of the credit limit.
Vendors sold lights throughout the month. Once a month Barefoot sent a text message asking vendors to SMS back with their next order. Barefoot made a delivery a week after the text message requesting orders. Vendors met the delivery driver at the BRAC branch headquarters to make their payment and accept delivery of the lights they had ordered.