Testing Mobile Phone-based Sales to Increase Use of Safe Water Filters in Kenya

Kenyan woman carrying water from pond


Contaminated drinking water contributes to the deaths of some 750,000 children under the age of five every year due to diarrheal diseases. However, a number of randomized controlled trials have shown that low-cost, point-of-use (POU) ceramic water filters are microbiologically effective and, when used, improve health.

In several studies, consumers preferred water filters to chlorine-based POU treatment, which may lead to higher and more consistent usage over time. Gravity-driven filters that feature built-in safe storage also are easier to use than many other POU treatment options: with filters, the user simply adds water.

Like many durable health products, however, water filters face several barriers to their initial uptake. Filters have a relatively high upfront cost; many consumers face liquidity constraints and may not be able to pay for them. Some consumers may also have "present bias," which reduces willingness to spend today for future benefits, even when those benefits are significant. Most consumers are also unfamiliar with how well filters work and with their effectiveness in improving water quality.


The RAND research team is working with a local implementing partner, the Safe Water and AIDS Project (SWAP), a Kenyan NGO that is an umbrella organization of more than 300 community health promoters (CHPs). These CHPs sell health-promoting products such as handsoap, bednets, and water treatment products as an income-generating activity that also benefits the wider community.

We developed and tested an innovative sales model the CHPs can use to overcome these barriers: A free product trial lets consumers learn about the product, and installment payments ease liquidity constraints. By collecting payments via Kenya's vast mobile banking network, M-PESA, we reduced transaction costs.

This project takes advantage of mobile health and banking innovations that have accompanied the exploding growth in mobile phone access throughout the developing world to make this model potentially sustainable and scalable in reaching poor rural populations. In Kenya the mobile banking system M-PESA is widespread, with 73% of the poor using its services in 2012.

We implemented a randomized controlled trial of 300 households in western Kenya to test this model.

Preliminary Findings

Initial acceptance rates of the filter sales offer were quite high, with 52% of customers agreeing to pay back on installment. After six months, 29% of all households had completed their purchase.

The deteriorating payments were largely due to husbands and wives not agreeing on whether to follow through on their commitment. The initial sales offers were generally made to wives, who tend to be responsible for water treatment. However, the wives accepted without their husbands' input, and husbands usually hold primary custody of a shared mobile phone and have input (if not sole discretionary power) on the household’s finances.

In future work we plan to expand upon these promising results to refine the sales model based on lessons learned, including taking the dynamics of gender into account; test this model on a larger scale to measure cost-effectiveness and potential for scalability; and analyze vendor considerations on the supply-side to understand the potential for private sector expansion.


This project is sponsored by USAID Development Innovation Ventures (DIV)

Principal Investigator

Jill Luoto, Associate Economist