Measuring Rural Water Demand: A Price and Information Experiment in Kenya

The study includes two interventions. Communal water-tap clusters receive temporary prices of KES 5, KES 3, or KES 0.50 per 20-liter unit. KES 5 is the prevailing comparison price. Price conditions rotate across clusters over three weeks. Second, eligible households are individually assigned either to receive information about the availability and safety of water from the communal taps or to a no-information comparison group. The information intervention is assigned independently of the cluster-level price intervention.

2026-08-07

2026-08-31

Primary outcomes measure water demand during the first intervention week and across the full three-week price rotation using administrative transaction data: volume purchased, any purchase among registered customer accounts, activation among accounts with no purchases at baseline, and quantities and transaction frequency among baseline-active accounts. Endline survey outcomes measure reported water-source use, perception, and quantities.

The study uses a two-level randomized experimental design. Approximately 80 communal taps are grouped into 39 clusters constructed before randomization using overlap in their baseline customer populations. Households are associated with a tap cluster before the intervention on the basis of their registered or usual tap. The customer analysis sample includes registered customers who make no purchases at baseline, allowing the study to measure whether the intervention activates previously inactive households.

For the first intervention week, the 39 tap clusters are stratified by baseline water consumption and assigned in equal numbers to three price conditions: 13 clusters at KES 5 (control), 13 clusters at KES 3, and 13 clusters at KES 0.50 per 20-liter unit. When the price at a cluster changes, registered customers associated with that cluster receive an SMS communicating the applicable price. During weeks two and three, clusters follow randomized price sequences that rotate them among the three price conditions. Eligible households are separately assigned in equal proportions to receive information about tap-water availability and safety or to receive no information. Household assignment is stratified by tap cluster.

The analysis examines administrative outcomes during the first intervention week and across the complete three-week price rotation, as well as survey outcomes measured immediately after the first week. Heterogeneity analyses examine whether treatment effects vary with baseline water-storage capacity, elicited willingness to pay, administrative water purchases, survey-reported use of other water sources, and distance to study taps.

Not available

Random assignment is conducted in an office by computer. Tap clusters are stratified by baseline water consumption and assigned in equal numbers to the three price conditions. Clusters are subsequently assigned to randomized price sequences for weeks two and three. Eligible households are separately randomized in equal proportions to the information and no-information conditions, with randomization stratified within each tap cluster.

Price conditions are randomized at the tap-cluster level. The information intervention is randomized at the household/customer-account level within each tap cluster.

Yes

39 tap clusters for the price intervention, comprising approximately 80 communal water taps; approximately 3,000 eligible households or customer accounts are individually randomized for the information intervention.

Approximately 3,000 eligible household accounts observed in administrative transaction data. A subset of households also participates in the household surveys.

First-week price assignment: 13 tap clusters at KES 5 per 20 liters, 13 tap clusters at KES 3 per 20 liters, and 13 tap clusters at KES 0.50 per 20 liters. During weeks two and three, all 39 clusters follow randomized sequences that rotate them among the three price conditions.
Information assignment: eligible households are assigned in equal proportions within each tap cluster to the information and no-information conditions.