Comparing utility-reported hours of piped water supply to households’ experiences

Intermittent water supplies (IWS), where consumers receive piped water for fewer than 24 hours per day, affect more than one billion people globally and are associated with adverse infrastructure, health, and economic impacts for households. Continuity of supply is a common utility performance benchmark, with 24/7 considered the ideal. Despite IWS prevalence and SDG guidance that water should be available when needed, there is no standard method to measure or report continuity in piped systems. Reported hours can vary substantially within and between systems due to hydraulics, elevation, demand, and leakage, leading to heterogeneous household experiences. The IBNET database systematically collects continuity data as hours per day from utilities, but does not capture whether delivery occurs daily, and there is no guidance on whether reported hours reflect plant operations, household receipt, citywide averages, or minima. Alternatively, national surveys (e.g., DHS) ask households about outages, though these are not tailored to piped continuity specifically. With climate change and urbanization likely to increase IWS, stakeholders need standardized continuity metrics to track progress and understand performance variation. This study compares utility-reported hours (IBNET) to household-reported outages (DHS) to triangulate continuity in Peru (2010–2014) nationally and regionally, tests thresholds aligning both sources, and applies the approach to three other countries.

Study site: Peru has nationwide piped water coverage managed primarily by 53 municipal utilities (EPS) serving cities and towns and by ~11,800 rural community organizations (JASS). Peru was selected for its high piped coverage and consistent IBNET reporting (48 of 50 utilities reported every year, 2010–2014). Data sources: Utility-reported continuity data (hours/day) were obtained from IBNET for 2010–2014. Household-reported data were obtained from Peru’s DHS (2010–2014). Analyses used only households whose primary drinking source was piped (in dwelling, on premises, or public tap). Two DHS continuity questions were available in 2013–2014: SH42 (availability the whole day) and SH43 (any day-long outage in past 2 weeks). SH43 was used for the main analysis as a better proxy for discontinuity. Region matching: IBNET utilities (city-level) were mapped to Peru’s regions by geolocating utility offices/cities (coordinates from Google Maps) and assigning each to a region using World Bank regional profiles. For regions with multiple utilities, mean hours of supply were computed for the region. In 25 of 27 regions, 1–5 utilities reported (none from Callao and Piura). Comparative framework: Two methods were applied.

• Method I (national): Household IWS (%) = households using piped water answering “yes” to SH43 divided by total piped-water households. Utility IWS (%) = population served by utilities classified as IWS divided by total population covered by reporting utilities. Utilities were classified as IWS under a primary threshold of <12 h/day (consistent with JMP’s “available when needed”), varied in sensitivity analyses (e.g., 16 h/day). The approach was also applied to other countries (Peru, Albania, Tanzania, Zimbabwe) where household data were available (DHS or LSMS).
• Method II (regional): For households, per-region IWS (%) = fraction of piped-water households reporting SH43 outages. Regions were classified as IWS if this fraction exceeded a threshold (varied, e.g., 25%). For utilities, per-region IWS (%) = population served by utilities below an hours/day threshold (varied; e.g., 12 or 16 h/day) divided by total covered population in the region. Regions were also classified as IWS if mean utility hours were below the threshold. Concordance between household- and utility-based regional classifications was assessed across threshold combinations. Sensitivity analysis: Thresholds for hours/day and household IWS% were varied to assess impacts on national IWS proportions and number of regions classified as IWS. Cumulative distributions of population vs hours thresholds and regional counts vs thresholds were evaluated. Data analysis: R was used for data extraction and regional clustering; Excel was used for regressions. Blank fields were excluded.

• National IWS proportions were similar between sources but with households reporting slightly more IWS: households 23–26% vs utilities 20–21% (2010–2014), a difference within ≈5 percentage points.
• Utility continuity and coverage: 50 utilities reported; mean continuity 15.7 h/day (SD across years ≈0.7 h/day), increasing by almost 1 h/day over 2010–2014. Hours ranged 1.6–24 h/day. All but 2 utilities were <24 h/day; 13/50 were <12 h/day. Using <12 h/day threshold, ≈3.5 million people (≈21% of Peru’s population) were served by IWS (2010–2014). No association was found between a utility’s hours of supply and its population served, nor between improvement/decline and baseline hours.
• Household outages: Mean 24.8% (SD 1.2%) of piped-water households reported at least one day-long outage in the past 2 weeks; regional values spanned 3.3–49.8%. Household-reported outages decreased on average by 0.32 percentage points per year.
• Regional discordance: Despite similar national levels, household- and utility-based IWS classifications rarely matched regionally: only 2 of 23 regions displayed close agreement. Many regions showed IWS by one measure but not the other. Utility-reported IWS spanned 0–100% of regional population, while household-reported IWS per region ranged ≈2.65–49.3%.
• Threshold effects: Using a more stringent utility threshold of 16 h/day yielded national IWS estimates consistent with households (both ≈26%). Thirteen utilities were <12 h/day; 24 utilities were <16 h/day. More than 90% of the population experienced at least 2 h/day of outage at some threshold. A discontinuity at 21.8 h/day (Lima’s continuity) shifted IWS population from 52% to 71% due to Lima’s large population (>9.2 million). The highest regional concordance occurred around household thresholds ≥16% and utility thresholds near 21.8 h/day, but concordant regions remained few.
• Cross-country comparison (12 h/day threshold): Albania: utilities 73% IWS vs households 30%; Tanzania: utilities 50% vs households 67%; Zimbabwe: utilities 18% vs households 13% (similar).
• DHS question consistency (2013–2014): SH42 (all-day availability) vs SH43 (day-long outage in past 2 weeks) had 65% and 64% concordance, respectively. In 2014, 26% reported not available all day but continuous over 2 weeks; 10% reported available all day but not continuous over 2 weeks, indicating substantial inconsistency and differing constructs.

Nationally, household- and utility-reported IWS in Peru were close (within ≈5 percentage points), but this appears coincidental given marked regional mismatches. Threshold selection materially influenced alignment; a 16 h/day utility threshold better matched household-reported outages nationally in this context. Household reports likely capture intra-system variability (hydraulics, elevation, demand, valve operations) that utility summaries miss, while utilities lack standardized guidance on how to measure/report hours (e.g., plant operations vs household receipt, averages vs minima). Seasonal and temporal variability further complicates comparisons. Household-reported measures varied more year-to-year (sampling, timing) than utility reports, whose national aggregates are stabilized by large systems (e.g., Lima). The two DHS questions capture different constructs (availability vs severe intermittency), explaining low concordance; SH43 better reflects severe discontinuity but may miss moderate intermittency and seasonality. Cross-country comparisons likewise showed no consistent relationship between utility and household measures, underscoring methodological and contextual differences. Overall, discrepancies highlight the need for standardized, comparable metrics of continuity that integrate frequency, duration, and predictability of supply.

This study triangulated continuity of piped water supply in Peru using utility-reported hours (IBNET) and household-reported outages (DHS), demonstrating similar national IWS proportions but poor regional concordance and strong sensitivity to threshold choices. A utility threshold of 16 h/day aligned best with household-reported outages nationally, suggesting perceived intermittency may become salient around this level in Peru. Results from Albania, Tanzania, and Zimbabwe further showed inconsistent relationships between data sources. The work highlights the need to standardize continuity metrics and reporting practices to enable fair performance benchmarking and better understand user experiences. Future directions include: (1) standardizing utility continuity reporting to include hours/day and days/week (and seasonal ranges), as well as population actually experiencing IWS; (2) capturing predictability (scheduled vs unscheduled) and adherence to schedules; (3) enhancing data collection via customer surveys, complaints, and feasible sensor deployments; and (4) refining household survey questions to differentiate continuous, moderately intermittent, and severely intermittent service, including frequency and duration of deliveries/outages.

IBNET covers only reporting utilities; households may be served by providers outside IBNET, potentially biasing comparisons, though national coverage figures were broadly consistent across sources. Regional and utility-level classifications obscure intra-city variability in continuity. Household and utility denominators differ (households vs population). Surveys vary in timing and samples year-to-year, masking seasonality; utilities may change service areas or structures over time. Utility reporting methods for hours/day are not standardized and may reflect plant operations or target schedules rather than household receipt. Geographic matching lacked fine resolution, limiting precise utility–household linkage. DHS question inconsistencies indicate construct differences and recall issues, complicating interpretation.