Harvesting Rainfall: Randomized Cistern Deployment in Northeast Brazil
Last registered on March 06, 2019


Trial Information
General Information
Harvesting Rainfall: Randomized Cistern Deployment in Northeast Brazil
Initial registration date
December 02, 2014
Last updated
March 06, 2019 2:04 PM EST
Primary Investigator
Other Primary Investigator(s)
PI Affiliation
PI Affiliation
University of Toronto
PI Affiliation
University of California-Berkeley
Additional Trial Information
Start date
End date
Secondary IDs
Water scarcity and water quality are acute problems for millions of rural households across the world, affecting health, education, and economic outcomes. Access to safe and reliable water remains a challenge for 780 million people worldwide, most of whom live in low density rural areas. While in urban settings piped water networks working with large reservoirs and underground aquifers are viable solutions for water access, for families living in low-density rural settings an economical solution to water supply is yet to be found. In the Brazilian Northeast region we study, comprising 9 states home to 22 million rural inhabitants spread over 974,752km2, 96% of households in rural areas have inadequate or semi-inadequate water and sanitation. The region is populated by small landowners dedicated to rain-fed agriculture. For the majority of these families the main sources of water are seasonal rivers and small community reservoirs at the foot of hills which replenish during the rainy season. Vulnerability to drought is very high, and families cope with water scarcity by restricting use, depleting savings to purchase water, and migrating temporarily or even permanently. Even when water is available during the rainy season, in terms of health, drinking surface water has a high likelihood of being contaminated causing children to suffer from chronic diarrhea and other illnesses that prevent absorption of micronutrients essential to physical growth and cognitive development.

In this context, we conduct an impact evaluation of a novel water technology being rolled out to address water access in the rural Northeast. It consists of providing households (free of charge) with a large residential rain-fed water cistern to households in rural areas lacking other water sources (No piped water, no pre-existing water cistern). The cistern intervention we study consists of a 16,000 liter reinforced concrete cistern for residential use, which is filled during the rainy season with water running from the house’s rooftop, and – in theory – should provide families enough water for cooking and drinking during the dry season. The cistern technology was a proposal of the NGO community in Brazil as a solution to water access in the region after a major drought in the late 1990’s exposed profound structural water vulnerabilities. Since 2003, the Brazilian government has been working with a network of organizations on a program to install one million roof-fed rainwater cisterns that will serve roughly five million rural people in the region. With support from AECID (Spanish International Development Fund for Water and Sanitation) we implement a rigorous randomized control trial of this technology and its impacts at household level. We address key questions about effects of the cistern on water availability, water quality, children’s health, labor market outcomes, and educational outcomes.

An additional data collection effort was implemented on the same sample in 2012 and 2013 financed by SSHRC to study how the cisterns affected political interactions between households and local politicians.
External Link(s)
Registration Citation
Bobonis, Gustavo et al. 2019. "Harvesting Rainfall: Randomized Cistern Deployment in Northeast Brazil." AEA RCT Registry. March 06. https://www.socialscienceregistry.org/trials/561/history/42527
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
The intervention consists of the random allocation of residential rain-fed water cisterns among rural households in Northeast Brazil. Developed by Articulação no Semi-Árido Brasileiro (ASA), these cisterns are constructed from reinforced concrete and hold up to 16,000 liters of water. They are constructed semi-buried next to the main housing structure with a conical cover, a small metal door to allow for its maintenance, and a manual pump made of PVC tubes. The cisterns require the resident’s house to have a roof size of minimum 40m2 to have the minimum required catchment area. The cost of building one of these cisterns is around US$1,000.
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
i. Water access (quantity and quality)
ii. Time use
iii. Labor market participation and earnings
iv. Spillover effects onto non-treatment families
v. Savings (Durable goods, wealth accumulation)
vi. Health outcomes
vii. Children’s educational outcomes
viii. Citizen requests of private goods from local politicians in exchange for votes (clientelism)
ix. Voting outcomes at the voting machine level
Primary Outcomes (explanation)
i. Ultrasonic water level sensors had been installed into the constructed cisterns. These provide hourly data on the level of water and also increases or decreases in water level in the cistern. This information allows us to measure water collection and water use patterns. Water quality is measured through EColi presence absence tests

ii. Using a visual basic time use module, we obtain information from the main water fetcher of the household, providing us with time allocated to various activities during the previous day

iii. Standard modules about labor supply and earnings

iv. Direct questions about frequency and amount of water that is being requested and shared, as well as participation of neighboring families without cistern assignment in the study

v. Standard modules about durable goods (and values), livestock holdings (and values)

vi. Height, weight, anemia, reports of diarrhea and illness symptoms in the last seven days to measure health outcomes.

vii. Standard modules on school attendance school enrollment

viii. Two dedicated survey rounds were fielded on the same sample to study political interactions of the individuals in the household with local politicians. The main outcome is citizens' requests for private goods.
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
We use randomized treatment assignment.
Experimental Design Details
Randomization Method
Randomization by a computer
Randomization Unit
Households for the study were selected using clustered random sampling. Clusters of eligible households range in size from 1-5. Eligible households were assigned to one of three treatment conditions:
1-All households receive cisterns
2-Half of households in the cluster receive treatment
3-None of the households in the cluster receive treatment (Control)
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
1000 clusters of households with an average size of four households each
Sample size: planned number of observations
1,000 clusters with 4,000 households
Sample size (or number of clusters) by treatment arms
1-All households receive cisterns (1000 households)
2-Half of households in the cluster receive treatment (1000 households with cistern, 1000 households without cistern)
3-None of the households in the cluster receive treatment (Control, 1000 households)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
The schedule of minimum detectable effect size (MDES) vs Power and MDES vs Intracluster correlation using Optimal Design for continuous variables is in Figure 8. Assuming 12 individuals per cluster and a 5% significance level, the MDES for standard power ranges between 0.8 and 0.85 are between 0.15 and 0.2 SDs. Similarly, the design is robust to a range of large intra-cluster correlation values (Figure 9) as a result of the large number of independent clusters we built into the design. We also have power for binary outcomes. For example, diarrhea incidence last week in our baseline survey was 19%. Assuming a power of 0.9, significance of 0.05, we find a MDES of 4 percentage points (a reduction to 15% prevalence) for our sample size.
IRB Name
IRB Approval Date
IRB Approval Number
IRB Name
Innovations for Poverty Action IRB - USA
IRB Approval Date
IRB Approval Number
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers