Despite significant investments by governments, international donors and NGOs in bringing sanitation to low-income communities, the adoption of improved sanitation by low-income households has remained stubbornly low. Building on a recent but flourishing literature stressing the importance of social networks on influencing individuals’ behaviour, this study will address two questions: What are the magnitude and the nature of peer effects on the adoption of a new sanitation technology? Given what we will learn about these effects, how can policy-makers and NGOs design their interventions to maximize take-up?
In order to answer these questions in the Zambian context, this study partners with Water and Sanitation for the Urban Poor (WSUP) to carry out research to test different strategies to encourage the inhabitants of Chazanga, a poor urban neighbourhood in Lusaka, Zambia to connect to modern sewage services. By mid-2016, WSUP plans to establish a supply chain for pour-flush toilets and septic tanks in this compound, in collaboration with the Lusaka Water Trust, a water provider to peri-urban areas under the management of Lusaka Water and Sewage Company. This solution, which will be adopted by plots (usually 3-5 households, managed by a landlord who may or may not live on site), represents one step above the current sanitation situation. As part of their program, WSUP will conduct community education and marketing to promote the construction of pour-flush latrines and septic tank connection, and has consented to submit aspects of this project to evaluation.
In this study, we will elicit the distribution of the willingness-to-pay for pour-flush toilets and unpack the dominating explanations of low adoption. We will document how social networks link households in Chazanga and how network ties affect adoption decisions.
We will examine how subsidies, access to credit, and social incentives compare in encouraging households to adopt the pour-flush/septic-tank solution supplied by WSUP. To reach causal inference, we will use a randomised controlled trial. The randomisation unit is the plot and most of the interventions will target the landlords of these plots.
The first stage will consist in randomly allocating the population into 4 treatment arms. The control group will receive no particular intervention. In the subsidy group, the landlords will be proposed a voucher to buy the sanitation solution at a lower price. In the credit group, the landlords will be offered a micro-credit solution to buy the product with instalments. The coordination group will be offered a coordination intervention that increases communication between landlords and tenants.
The distribution of the voucher will take the form of a Becker-DeGroot-Marschak (BDM) game, which offers the benefit to both elicit the landlord’s willingness to pay and randomise whether the landlord is offered the subsidy. Landlords who have been randomly assigned to play the BDM game will reveal the maximum amount they would pay to have the pour-flush toilet installed (their willingness to pay). Then, they will draw a price between zero and the true price of the pour-flush latrine. If the draw price is equal to or below their stated willingness to pay, the landlord will pay the amount of the draw price to have the latrine installed. The difference between the true cost and the draw price will be the subsidy voucher. If the price drawn is higher than the maximum amount stated by the landlord, there is no transaction.
Landlords in the credit group will be offered a credit solution, coupled with a subsidy, to pay for the improved sanitation, in the form of a BDM game. To make it easy to understand, the previous game will be played except that landlords will now bid on a monthly amount they will have to repay each month during 12 months. A price will be drawn: if the price is higher than the monthly amount bid, no transaction will occur; if it is lower, the landlord will commit the price drawn each month and will receive the latrine in exchange. The credit arm will be done in collaboration with a local microfinance institution.
The coordination intervention will take place at the plot level and involve both the landlord and his tenants. Anecdotal evidence suggests that low take-up is as a result of a lack of communication between tenants and landlord about sanitation and rents. If the new sanitation system is valued by tenants and the landlord does not have strong credit constraints, we expect agents to reach a solution in which sanitation is improved in exchange for a rent increase. The idea of this intervention is to propose the deal explicitly to all agents of the plot so that they can start to discuss and eventually take up.
Since the health benefits from adopting improved sanitation solutions also likely depend on the adoption rates of neighbouring households, we also intend to measure the externalities to sanitation improvement. In particular, to examine how households are affected by having a greater (or lesser) proportion of their neighbours with improved sanitation. This second stage does not involve any intervention; we only intend to measure how adoption spreads through social networks. Our experiment is designed so that we can disentangle two competing hypotheses underlying peer effects: (i) learning (as neighbours get equipped, a household learns about the benefits it could experience with adoption), and (ii) social norm (a household may incur a cost in being the only one to use or not use a new technology). Separating the two explanations will lead to richer policy recommendations, which can in turn improve the cost-effectiveness of future programs.