Sustainable Water Supply and Sanitation Project Impact Evaluation 2019, Endline Survey

SAMPLE SIZE: The size of the eventual sample was 300 communities - 150 assigned to treatment and 150 assigned to control.

SELECTION PROCESS: The starting point was a universe of 6,862 communities across 148 municipalities that were in the SIASAR database and constituted almost the entire country. Community eligibility criteria were introduced in order to focus the evaluation on communities that would most likely improve their indicators with the intervention, given their initial status. The criteria for the selection process is described in more detail in the study report found under the ‘Documentation’ tab.

STRATIFICATION AND STAGES OF SAMPLE SELECTION: The sample of 300 communities was selected, across 75 municipalities and within each municipality the 4 eligible communities were randomized to either treatment or control (2 allocated to treatment and 2 allocated to control). This resulted in 150 treatment communities and 150 control communities evenly distributed across the 75 municipalities. Within the 150 treatment and 150 control communities, 100 communities come from 50 municipalities classified as "less-poor" and 50 communities across 25 municipalities that are "poor"

Municipalities were stratified (poor, less poor) and randomly ordered; the first 50 less poor and the first 25 poor municipalities in each list were selected for inclusion in the evaluation. Within each selected municipality all eligible communities were randomly ordered. The first two communities became control communities and were automatically enrolled in the evaluation. The remaining randomly ordered communities (3 through to n) were all assigned to receive the treatment from the municipality; two communities from the treatment list (all communities below 3 in the original random ordering) were randomly selected to be enrolled in the evaluation (data collection) in a manner consistent with the random ordering described.

LEVELS OF REPRESENTATION: As the original universe was from all the communities in the country registered in SIASAR (nearly every community nationally), and we then selected randomly half the municipalities in the country, our sample has a good level of representation. We also checked for balance on distribution across the treatment and control groups of the communities in the Pacific Coast, Central Region and Atlantic or Caribbean Coast.

During baseline fieldwork, owing to logistical issues and interviewer safety, the evaluation team had to make minor adjustment to the community list. The team replaced (1) one municipality (four communities), (2) two communities from one municipality with two from another, and (3) one community for another. For this reason, the final sample after baseline work had 76 municipalities instead of the 75 planned. These replacements and substitutions used randomization to ensure that the initial experiment design was preserved.