Community toilet use in slums - willingness to pay and the role of informational and supply side constraints

Two interventions will be evaluated:

(1) Supply-side intervention:
In this intervention, cleanliness and maintenance of CTs will be promoted. A grant scheme will be introduced that offers standardized (i.e. same monetary value) packages including either (i) deep cleaning (i.e. septic tank sewage removal, unclogging latrines and sewerage pipes and cleaning walls, floors and inside toilets), (ii) sanitation/water connection repairs and/or infrastructure refurbishment or (iii) cleaning tools and agents and cleaning training, from which CT caretakers can choose according to their needs and their expectations to increase usage. In addition (two month after baseline) a financial reward aimed at incentivizing caretakers to keep the CT clean will be announced. At this point, CT caretakers will be informed of their baseline cleanliness performance. After two months (and every two months from then on), caretakers will receive a financial reward conditional on achieving the following: (i) availability of soap in hand-washing facilities; (ii) latrines free of visible faeces; and (iii) bacteria count in defecation cubicles kept to a minimum standard (to be computed from the baseline distribution). We aim to disentangle the effects of releasing resource constraints from the announcement and payment of the reward scheme, relying on the phased-in design of the supply-side intervention.

(2) Supply side intervention plus Information provision:
In this intervention, the supply side intervention (see intervention (1) above) will be complemented with a household-level information provision component. Participants in this group will receive information about private and public risks of open defecation. Furthermore, information on the relevance of ensuring females’ safety will be provided and male household members will be encouraged to accompany females when using CTs, particularly at night. In addition, information on the benefits of using CTs and the importance of ensuring that everybody pays the CT fee will be provided. Emphasis will be placed on providing information on the rights of the users when paying fees, including clean and well-maintained toilets, as well as promoting users to ensure maintenance and cleanliness of CTs. Particular attention will be paid in providing information that can be processed by participants. All household members will be targeted, especially household heads and spouse. Information will be provided in four forms:
(i) door-to-door information campaign using a flipchart with cartoons and messages;
(ii) leaving a leaflet with a summary of the flipchart;
(iii) posters placed in the CT, highlighting messages provided during the door-to-door campaign;
(iv) monthly remainders in the form of voice messages, through which we aim to cross-randomize the salience of public and private health risks.

2018-04-18

2019-04-18

1. Willingness to pay;
2. Sanitation behaviour, with a focus on CT use and open defecation;
3. Demand for cleanliness;
4. CT quality.

1. Willingness to pay: Elicited separately from the two primary decision-makers per household (hence there are two observations of each outcome per household).
- Incentivized WTP for bundle of 10 tickets to use the nearest CT (using multiple price list and random draw)
- Hypothetical WTP for a ticket to use an improved toilet

2. Sanitation Behaviour: Sanitation practices of the respondents and their family members. In particular, we will focus on survey reports of CT use and open defecation to construct the following indicators :
- All members above 5 years old used CTs the 2 last times
- At least 1 member above 5 years old used CT at least once the last 2 times
- At least 1 member above 5 years old practiced OD at least once the last 2 times
We will further measure CT usage through tallies at the CT at specific times of the day as well as by collecting tickets acquired by HHs in the WTP game.
- Number of users (all, male, female)
- % users that pay (all, male and female)

3. Demand for cleanliness: Eliciting willingness to contribute to cleanliness of the CT through a “donation game”.
- Amount donated out of 50Rs (continuous variable 1-50)

4. CT quality: Quality will be proxied using observations and lab results from samples taken at the CT and captured through:
- Dirtiness index
- Bacteria count
- Bad infrastructure quality index
(The attached pre- analysis plan details the variables to be used in the indexes)

1. Hygiene
2. Health status;
3. Sanitation attitudes, expectations, and knowledge;
4. CT management

1. Hygiene: Hygienic practices of the respondents and their family members. In particular, we will focus on survey reports and observations to construct the following indicators
- All members >5 years old washed hands with soap after defecating the last time
- HH dirtiness index
(The Pre- analysis plan, to be published on the lie and IFS website will detail the variables to be used in the indexes)

2. Health status: Health situation of household members (illness status) reported in household surveys:
- At least 1 member had diarrhoea during last 15 days
- At least 1 member had fever during the last 15 days

3. Sanitation attitudes, expectations and knowledge: Priors about sanitation practices and the connection with illnesses and safety, attitudes towards CTs and subjective expectations reported in household surveys:
- Positive CT attitudes index
- Percentage points (ppts) difference in (adults and children) expected illness likelihood when passing from OD to CT
- Ppts difference in (adults and children) expected illness likelihood when passing dirty to clean CT
- Ppts difference in expected female unsafe incidents likelihood when passing from OD to CT
- Sanitation knowledge/awareness of public health risks of inadequate sanitation index
(The attached pre- analysis plan details the variables to be used in the indexes)

4. CT Management: Management of CTs as reported in CT surveys by caretakers:
- % time allocated to clean and/or supervise cleaner
- % time allocated to collect fee
- CT cleaned more than twice per day
- Adequate cleaning index
(The attached pre- analysis plan details the variables to be used in the indexes)

We will address these research questions in the context of Indian slums in the cities of Lucknow and Kanpur (Uttar Pradesh, India) using a randomized-controlled trial (RCT) design. Figure 1 shows the study area. We will randomly select 110 catchment areas of CTs in both cities and we will allocate them to one of three experimental arms. Catchment areas will then be allocated to one these three groups:
1. Control
2. Supply-side intervention
3. Supply side intervention plus Information provision

Randomization into these groups will be carried out at the cluster level, specifically the CT catchment area level. Randomizing at the cluster level has the advantage that we can limit contamination of the control group, especially considering the possible spread of the information that will be provided. This is of particular relevance for answering research questions around longer-term behaviour (particularly usage of sanitation facilities).

To allocate clusters to treatment arms, we will stratify the sampled clusters by the main company who owns the CTs (versus other owners) and by city of study (Lucknow and Kanpur). We will then build blocks of 3 CTs using m-distance (Mahalanobis) relative proximity. To construct m-distances, we will make use of the rich census information we collected, including CT and slum-dweller characteristics. After forming blocks of similar clusters (CTs), we will randomly allocate each CT in a block to either treatment 1, treatment 2 or the control group. Each one of the three possibilities will have the same probability.

In our study, a cluster is a CT catchment area. We identified catchment areas by first mapping all CTs in our study cities. Out of these, we chose a subset of CTs to become part of the study, based on the following criteria:
1. The CT has to be pay-to-use;
2. It has to be located close to a residential area (slum) and used by residents.
We drop CTs for which the distance to another CT is limited. In particular, there should be sufficient distance between two CTs to avoid users switching between CTs (possibly driven by their treatment status). We drop CTs that are closer than 300 meters to each other, and CTs that have two other CTs closer than 350 meters.
In addition, we drop CTs in whose catchment areas fewer than eight eligible households are living. A household is considered eligible if the following conditions are respected:
1. The household lives in the catchment area of a selected CT, which is broadly defined as slum area within 250 meters from the CT building. Households are linked to CTs based on geo-coordinates collected during the census.
2. At least one household member uses a CT or shared toilets (i.e. neighbours, makeshift, work, school), or practices open defecation.
3. The household must have reported during the census interview not to intend to migrate during the following 18 months (i.e. until the planned study endline survey). We will focus on these households to reduce the risk of attrition.

Within each of the selected study CTs and their catchment areas, we will sample up to 17 eligible households (if available), aiming for an average of 15 households per cluster. Given that distance is a major determinant of CT usage, we will focus on eligible households living closer to the CT (within 150 meters). Since some CTs have more dispersed populations, we will conduct a two-step sampling procedure. First, in large-population catchment areas (where 10 or more households are available within 150 meters), we will sample only from households that are located within this bound. Second, in small-population CTs (where less than 10 households are available within 150 meters), we will first sample all households within 150 meters and randomly selected the remaining households from those that are located between 150 and 250 meters from the CT. We aim to interview 1,650 households in 110 randomization units (catchment areas of CTs) and one CT caretaker per randomization unit.

The main respondent for the household survey will be an adult household member – likely the household head, spouse or other knowledgeable household member – who should fall in the age range of 18-64 years. We will further interview the spouse of the household head, aiming to collect information on WTP from one male and one female household member. In households with children aged 5 years or younger, we will further interview the primary caregiver to collect information on child health and sanitation practices.
The main respondent for the CTs (supplier survey) will be the caretaker.

The statistical software Stata, and specifically the random number generator, will be used to apply this procedure.

Randomization into the three experimental groups will be carried out at the cluster level, specifically the CT catchment area level. We provide more detail on the definition in the next section. Randomizing at the cluster level has the advantage that we can limit contamination of the control group, especially considering the possible spread of the information that will be provided. This is of particular relevance for answering research questions around longer-term behaviour (particularly usage of sanitation facilities).

To allocate clusters to treatment arms, we will stratify the sampled clusters by the main company who owns the CTs (versus other owners) and by city of study (Lucknow and Kanpur). We will then build blocks of 3 CTs using m-distance (Mahalanobis) relative proximity. To construct m-distances, we will make use of the rich census information we collected, including CT and slum-dweller characteristics. After forming blocks of similar clusters (CTs), we will randomly allocate each CT in a block to either treatment 1, treatment 2 or the control group. Each one of the three possibilities will have the same probability.

Yes

110 CTs and their catchment area

We aim to interview 1,650 households in the 110 randomization units (catchment areas of CTs) and one CT caretaker per randomization unit.

Control: 39 clusters, 585 households, 39 CT caretakers
Supply only: 36 clusters, 540 households, 36 CT caretakers
Supply+Info: 35 clusters 525 households, 35 CT caretakers