Experimental Design
Recruitment of Study Communities
In 2014, the Tropical Agriculture and Higher Education Center (CATIE) executed a large Canadian government-funded research project focusing on climate change adaptation and water scarcity in Central America. Part of this project included work in rural communities in western Costa Rica. In particular, the project was interested in exploring the role that water saving technologies could play in reducing the pressure on scarce water resources. In 2014, CATIE’s staff identified community-based water drinking organizations (CBWMO) that had metered household water use and applied variable rate pricing. Staff called the 66 CBWMOs that met these criteria, and asked their management committees if: (1) they had monthly water records of households dating back to 2012 and would share these data by sending them to CATIE, and (2) they were interested in having the project team install water-efficient technology in a randomly chosen subset of their residential customers and in sharing the post-installation water data. Ten CBWMOs transferred the requisite pre-treatment data within 2-3 months after the invitation and nine of them were chosen at random to meet the RCT’s sample size target.
Treatment Intervention
Households were offered water-efficient technologies installed by professional plumbers: (1) 1.5 gpm (gallon per minute) shower heads; and (2) 1 gpm faucet aerators for bathroom and kitchen faucets. During the study period, the offered technologies were not common even in urban areas of Costa Rica. No surveyed home in the study area had such technologies prior to the experiment.
The communities reported 2,246 billed customers in March 2015. Based on the pre-treatment billing data and a pre-treatment field visit, 348 customers were eliminated: customers that had zero consumption between December 2014 and March 2015 (assumed vacant), shared a water meter with another house, or were commercial establishments. The exclusion exercise left 1,898 households for contact.
To contact the households, CATIE had four teams, each with an interviewer and a plumber. Interviewers had bachelors’ degrees and survey experience and were trained to implement the randomization protocol (see below). The four teams, overseen by a field manager, went to the nine communities sequentially. Communities in rural Costa Rica do not have maps with the location of houses and houses are not numbered. Thus, in order to facilitate the randomization procedure and ensure measurement fidelity over time, CATIE created community maps with the location of all houses and placed identification number labels on every water meter in the community. Using the community maps, CATIE divided the community into four equally populated sectors and assigned each team to one of them. Interviews were conducted using a tablet.
The team was able to contact 1,346 heads of households. The interviewer read a short script that comprised: 1) an introduction of team members; 2) information from a CATIE climate study about recent and future weather changes in the region and the implications of these changes for water conservation, 3) a description of the technologies and a video of them in use, and 4) an offer to install the water-efficient technologies for free if their home was selected at random. Households were only randomized to a treatment arm if a head of household indicated he or she was interested in accepting the installed technologies for free that same day.
Of the 1346 households, 1310 agreed to have the technologies installed should they be selected to receive them. Among these households, 395 were visited in May 2015 and the other 915 in June 2015. They were randomized into one of three treatment arms:
1. Control Group: Residents who agreed to install the technologies but did not receive the technologies.
2. No Bonus Group: Residents who agreed to install the technologies and received the technologies.
3. Bonus Group: Residents who agreed to install the technologies and received the technologies. After they had agreed to install the technologies, they were also offered a performance bonus of $38 if they still had all technologies installed when the team returned unannounced sometime in the following six months. A proportional bonus was paid if some, but not all, of technologies were still installed at the time of the audit.
After installation, the team took away the technologies that were replaced.
Audits
CATIE conducted two audits of treated households (November 2015, October 2016) and an audit of a random sample of 63% of the control group four to five months after treatment assignment to check if the technologies had been transferred to control households. None of the control households had the technologies and thus we assume zero non-compliance with the treatment assignment.