Study site
Peru had a population of ~32 million with a per capita GDP of US$211 Billion in 201734. Drinking water in Peru is provided by 53 Municipal Service Providers (las Entidades Prestadoras de Servicios de Saneamiento, EPS) that provide surface or groundwater to cities and towns in Peru, serving >62% of the country’s population35. Approximately 11,800 community-based organizations, known as Sanitation Services Administrative Committees (Juntas Administradoras de Servicios de Saneamiento, JASS), provide water to the remaining 29% of Peru’s rural population and 9% of Peru’s population that live in 490 small municipalities35. The National Sanitation Services Supervisory (SUNASS), a regulatory agency funded through a surcharge on water bills, regulates and oversees the EPS35,36,37. EPS are spread throughout the country; the median elevation for the utilities is 783.6 m and the country as a whole has an average elevation of 1555 m (Supplementary Fig. S1). We selected Peru for this analysis due to its high coverage by piped water (85–91% of the Peruvian population had access to piped water in 2017)35,37, and continuous reporting by utilities: 48 of the 50 utilities in Peru reporting continuity to IBNET for every year from 2010 to 2014 (Supplementary Table S1).
Data sources
Utility-reported water supply continuity were obtained from the IBNET, which compiles global data on water and sanitation utilities performance (ib-net.org)18. IBNET data from utilities in Peru from 2010 to 2014 were obtained from the online database. The accuracy of these data depends on reporting by utilities and/or regulators38.
The DHS Program collects, analyzes, and disseminates representative longitudinal data on population and health in over 90 countries39. These surveys documented coverage and continuity of drinking water in household (Table 3). Household survey results for Peru were obtained for 2010–201419,20,21,22,23. For all analyses, we used the subset of households surveyed who reported using piped water as their primary drinking water source (including piped water, piped into dwelling, piped outside dwelling, and public tap) (Question HV 201, Table 3).
Matching regions
Data from IBNET were reported by utility (which, in Peru’s case, corresponds to a city), while DHS records reported data from individual households by region. To harmonize these, we first identified the longitude, latitude, and elevation of the utilities using Google Maps based on the IBNET utility or city name40. When we could not locate a utility office on Google Maps, the coordinates of the center of the served city were used instead. We used these coordinates to group the utilities into their respective regions by matching each utility’s location to the list of regions from the country profile in World Bank publications (Supplementary Table S4)34. In 25 of the 27 regions in Peru, there were between 1 and 5 utilities reporting supply continuity to IBNET (no utilities reported from Callao and Piura). To arrive at an average utility-reported continuity (h/day) for a region with more than one utility, we calculated the mean hours of supply reported for utilities within that region.
Comparing utility and household hours of supply
Our analysis was divided into two parts: (1) Method I compared the two data sources at the national level; (2) Method II compared the two data sources at the regional level (Fig. 5). The analysis converted the utility- and household-reported data into comparable values by assigning binary values to utilities or households—each either supplied water intermittently or did not—and calculating percentages of the population using each type of piped water service. This was performed for each year from 2010 to 2014 for each region.
X = threshold of the minimum % population with IWS to deem a region intermittent. The numbers shown are from 2010 and used as an illustrative example; the analysis was performed for each year. Pop population, HH households, PWS piped water supply, IWS intermittent water supply.
Method I: national comparison
We expressed household-reported continuity as the percent of households experiencing IWS, estimated as the sum of the households reporting using piped water as their primary drinking water source and answering “yes” to questions SH43 (that water was discontinued for at least a day in the last 2 weeks) (Table 3 and Supplementay Table S2), divided by the total number of households using piped water. We expressed utility-reported continuity in Peru as the percent of people served by a utility providing IWS. To estimate this, we categorized utilities as providing IWS if their mean hours of supply between 2010 and 2014 was <12 h/day. We selected 12 h/day as the threshold, as we postulated that households will report an outage of a day if it is out for at least 12 h and to be consistent with current definitions of available when needed used by the JMP24. We summed the population served by the utilities categorized as IWS (<12 h/day) and divided by the total population covered by reporting utilities. This threshold was varied as part of the sensitivity analysis.
To compare the trends seen in Peru to global data, we collected available data for other countries reporting to IBNET and performed the same comparison as described above. Household surveys for Tanzania and Zimbabwe were available through the DHS Program, while those for Albania were obtained from Living Standards Measurement Study41.
Method II: regional comparison
To compare intermittency within regions, we compared the regions where utilities reported IWS according to thresholds of hours of supply and the percentage of households with piped water reporting water outages.
We calculated the percent of surveyed households using piped water supply who reported having discontinuous supply during the past 2 weeks (“yes” to questions SH43) for each region (Supplementary Table S2). We then categorized regions as supplying IWS based on whether this percent was greater than a threshold (e.g., if the threshold was set as 25%, then a region was classified as an IWS region if >25% of households reported an outage in the last 2 weeks; this threshold was varied). This yielded two values to compare with utility-reported data: (1) the percent of the population within each region with household-reported intermittent supply; and (2) whether a region was intermittent or not as defined by households.
We obtained utility-reported continuity for each region by grouping utilities into regions (described previously) and classifying each as IWS depending on their hours of supply (varying thresholds were used; e.g., <16 h/day). We calculated the total population with IWS by summing the number of people in each region served by utilities classified as IWS divided by the total population served by utilities in that region. We also calculated the mean hours of supply per day for the utilities in each region to categorize regions as IWS if the mean hours of supply for utilities in that region was less than a threshold (e.g., if the threshold was 16 h/ day, then a region was classified as IWS if the mean of utility supply hours in that region was <16 h/day; thresholds were varied). This yielded two values to compare with household-reported data: (1) the percent of population within each region with utility-reported intermittent supply; and (2) whether a region was intermittent or not as defined by utilities. In the regions where all reporting utilities supplied an average greater than the threshold, no populations were classified as receiving IWS; in regions where all the reporting utilities supplied less than the threshold, 100% of the population was classified as receiving IWS.
Sensitivity analysis
We performed a sensitivity analysis to determine the effect of varying thresholds throughout the analysis for Peru. We plotted the cumulative density function of the mean hours of supply of each utility against the average populations served over the 5 years, the number of regions with IWS according to household reports, and a range of minimum percent of households reporting outages (0–100%) per region against the number of regions with IWS.
Data analysis
R statistical analysis packages were used to extract and analyze household surveys and cluster utility data by regions42. Microsoft Excel statistical functions and Data Analysis tool were used to perform linear regressions. Blank fields were excluded from analysis.
Source: Resources - nature.com
