Demand for Sanitation in Kenyan Urban Slums
Last registered on May 02, 2017


Trial Information
General Information
Demand for Sanitation in Kenyan Urban Slums
Initial registration date
May 02, 2017
Last updated
May 02, 2017 1:59 PM EDT
Primary Investigator
University of California, Berkeley
Other Primary Investigator(s)
PI Affiliation
World Bank
PI Affiliation
U Maryland
PI Affiliation
UC Berkeley
PI Affiliation
Additional Trial Information
In development
Start date
End date
Secondary IDs
We propose to study the demand for household connection to municipal sewage systems in informal slums in Nairobi Kenya. Governments are investing in expensive sewerage systems to bring sanitation services to the household door. The cost-effectiveness of these investments depends on the number of households that connect the sanitation systems. However, there are large fixed costs to connect to sewage systems including both the costs charged by the utility investment in household sanitation facilities and pipes to connect from the house to the network. We propose to use an RCT to estimate price elasticity of the demand for connections and the extent to which price elasticities depend on information about the relationship between sanitation and health. We also consider complications related to collective action multi-household compound connections and resident versus non-resident landlords. Results from this study are critical to developing pricing/subsidy and information campaign policies to cost-effectively improve connectivity.
External Link(s)
Registration Citation
Coville, Aidan et al. 2017. "Demand for Sanitation in Kenyan Urban Slums." AEA RCT Registry. May 02.
Experimental Details
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
The experiment seeks to answer the following research questions:
• What is the overall price elasticity of the demand for sewerage connections?
• How does the price elasticity of demand change with better knowledge of the role of sanitation and preventing illness?
• How does the price elasticity differ as the number of residents in the compound grows?
• Is the price elasticity different when the landlord is resident versus non-resident?
• How much of the cost of connection is passed onto tenants and how does connection to the sewage systems affect housing rental prices?
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
In June 2014, a census listing of 2,200 compounds will be done to form the sampling framework and randomization for the 2,200 households. The census will collect basic demographic information; number of households per compound; number of children; if landlord is resident or not; contact details of landlord; toilet type; etc. From this listing, two households per compound will be randomly selected for baseline.
A baseline survey will be administered to all 2,200 compound landlords to collect socio-economic information, wealth, credit history, ownership and value of the compound, rents and policy for late payment, knowledge and attitudes towards hygiene and health, and interaction with and knowledge of tenants.
A baseline survey will be conducted with two household heads (male or female) per compound, resulting in about 4,400 household surveys. Baseline surveys will collect data on socio-economic characteristics; sanitation facilities, use, practices, and satisfaction; water sources, use, storage, treatment, and cost; hygiene and hand-washing knowledge and practice; altruism, reciprocity, and risk preferences; and limited health outcomes such as diarrhea and child weight for height to be able to assess power for the possible health impact follow-on study.
When available, administrative data will be used to cross-check information on take-up rates. Administrative data will also be collected and used to measure costs of the interventions. Direct costs include the subsidies, the administrative costs to deliver them, and the cost of the information campaign. We will use these costs to measure the cost-effectiveness of alternative subsidy and promotion campaigns in terms of take-up.
We will also conduct a short follow-up survey approximately 6 months after the intervention to verify the proper installation and use of sanitation facilities, satisfaction with the new sanitation facilities, other investments in housing that might have been made, changes in rent, and changes in tenants.
We will study two interventions deigned to promote take-up of household connections to the sewage system:
1. Three subsidy amounts – a low subsidy, a medium subsidy, and a large subsidy. The low subsidy amount is the subsidy being offered by GPOBA. The medium and high subsidy amounts will be decided based on the results of formative research on willingness to pay for sanitation services, which will be conducted as part of this study. The subsidies will be delivered in the form of vouchers that will be difficult to sell in a secondary market. The operational aspects of the voucher program will be designed and implemented by a local NGO with oversight from the research team.
2. An information campaign that promotes hygienic practices such as sanitation facility use, hand washing, and drinking water. The promotion campaign is being designed and delivered by the Water and Sanitation Program (WSP) in the World Bank and will explicitly incorporate learning from previous WSP activities including a recent series of impact evaluations exploring the effectiveness of these campaigns.

The demand for sanitation services is also likely to be confounded with collective action problems. Moral hazard and free riding problems typically arise when the sanitation infrastructure is shared. In many urban slums, multiple households live in compounds that share a single connection to the sewage network. Landlords may be unwilling to pay the cost of connection if they cannot recoup all or some of the cost from tenants either through increased rent or cost sharing. The per-household cost sharing needed to cover costs will be smaller in larger compounds. Hence, larger compounds might require a smaller subsidy in order to induce the landlord to connect. In this case, a policy that has smaller subsidies for larger compounds would achieve higher connection rates than uniform subsidies for the same total budget for subsidies. On the other hand, larger compounds may lead to more free riders so that some tenants may refuse to participate in the payments and thereby freeze collective action to pay for the connection. In this case, a policy that had larger subsidies for larger compounds would achieve higher connection rates than uniform subsidies for the same total budget for subsidies.
We also propose to measure how the cost is shared between tenant and landlord, and whether the connection results in increased rent that the landlord might be able to get from the market for the improved housing. If the sewage connection allows landlords to raise rent to tenants and current tenants are not willing to pay the increased rent, then they may be forced to move, thereby further disadvantaging the poor. We will also assess the extent to which altruism and reciprocity traits on the part of both landlord and the tenants affect the distribution of costs, and the extent to which the landlord living in the compound affects the distribution of the costs. Altruistic landlords may not pass on much of the cost, where reciprocal landlords will likely pass them on. If most or all of the costs are passed onto tenants and rents raised, then rather than subsidizing the landlord, it might be better to subsidize tenants’ cost-sharing conditional on being able to maintain residency in newly connected compounds.
There also may be differences in willingness to pay when the landlord is resident versus absentee. Absentee landlords might benefit from increased rent that they could charge by connecting their property to the sewage network, and resident landlords should additionally benefit from the health value of having sanitation in their homes. Moreover, an absentee landlord is probably less likely to be able to solve the collective action problem to be able to pass on the costs to the tenants. If the demand for sanitation is substantially different, pricing policies might have to be different not only depending on the size of housing compounds but also based on whether the landlord is resident.
Experimental Design Details
Randomization Method
The identification strategy is to generate exogenous variation through random assignment of the two interventions. The study site consists of 2,200 compounds that each have between 6-10 dwellings. The connection to the sewage system is at the compound level. Given the density of the slum, there may be spillovers in both the information campaign and knowledge of subsidies that may affect take-up and health outcomes in a follow-up study. In order to minimize potential contamination we will group the compounds into neighborhood clusters called blocks (about 10 compounds each). The unit of intervention will be the compound and we have 220 blocks.
The evaluation design consists of randomly assigning blocks into the two interventions, creating five treatment groups and one pure control group. There will be 3 different levels of subsidies—low, medium and high—randomly offered to landlords. The different levels of subsidies will be accompanied with a hygiene promotion campaign. Specifically, the blocks will be randomly assigned into arms of the study. This process will be done in the office by a computer
Since treatment assignment will be random we assume treatment and control groups will be balanced and comparable. Thus, the causal impact of the intervention will be estimated by using differences in mean outcomes between treatment and control groups. We will also estimate a series of models that condition on landlord and tenant characteristics to attempt to reduce residual variation and increase statistical power.
For sample size calculations we use latrine connection rates as our main indicator of interest for take-up. At baseline none of the households are connected to the sewage system. Assuming intra-cluster correlation of 0.01, a significance level of 0.05, and 0.80 power, we can detect .05 differences between each of the arms of the study. The landlord and up to two households in the compound will be surveyed.
We expect the demand for sanitation to be price elastic. One reason for such high price elasticity might be that households do not sufficiently value proper sanitation because they don’t understand the relationship between proper sanitation and health outcomes. In fact, there is evidence that awareness of potential benefits from a product may interact with the associated demand curve (Ashraf et al., 2013), as well as the identified importance of appropriate hygienic behavior (Luby et al., 2013), and water quality in influencing health outcomes (Kremer et al., 2010).
This hypothesis will be tested by examining whether an information and promotion campaign raises the demand for sanitation at each subsidy level. In this case, subsidies combined with an information campaign would best encourage take-up. However, another reason for low take-up could be that households value sanitation but have hyperbolic discount rates causing them to procrastinate. In this case, even though they know sanitation is good for them, people may put off the investment because they believe that they will have time to do so later since the benefits only accrue to them later. In this case, the information campaign will add little.
Randomization Unit
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
2200 Compounds/Landlords
10 compounds per Block/Cluster
220 Clusters of 20 compounds each
2 Households from each of the 2200 compounds (4400 Households)
Sample size: planned number of observations
6600 Observations (4400 Households + 2200 Landlords)
Sample size (or number of clusters) by treatment arms
The random assignment into these groups will be stratified by whether the landlord is resident or absentee and by number of households in the compound. We will define large compounds as those with more households than the median compound and small ones as the remainder. Specifically, the blocks will be randomly assigned into arms of the study as presented in the table below. The sample sizes are also specified.

Low subsidy (no info: 366 households, Info: 366 households); Medium subsidy (no info: 366 households, info: 366 households); High subsidy (no info: 366 households, info: 366 households)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
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