Revenue and investment data for rural drinking water infrastructure in Africa 2021-2024
Data files
Sep 16, 2026 version files 2.53 MB
-
analysis_dataset.csv
2.51 MB
-
README.md
4.82 KB
-
SI-DFI_commitments.csv
12.41 KB
Abstract
Dataset Documentation (analysis_dataset.csv)
Overview
analysis_dataset.csv contains the anonymized analytical dataset used to generate the descriptive figures and to train/evaluate the predictive models reported in the manuscript.
Each row corresponds to a single annual observation of a rural water service (aggregated at the service-year level, e.g., waterpoint-year or scheme-year depending on infrastructure type).
All monetary values are in USD.
Development finance investment data
Development finance data from the African Development Bank (AfDB) and the World Bank has been collected from public sites with web links to wider documentation (see SI). The data represent values of financial commitments in USD converted from an IMF currency basket, called Special Drawing Rights (SDR), by a standard annual calculation for equivalence. Projects were selected based on a) financial commitments made between 2006 and 2025, b) a project commitment of over USD10m, and c) a specific focus on rural drinking water services. Of the 67 projects included, 15 were funded by the World Bank and 52 by AfDB across 27 countries. Equivalent data from bilateral donors is not available in a standardized format.
Overview
analysis_dataset.csv contains the anonymized analytical dataset used to generate the descriptive figures and to train/evaluate the predictive models reported in the manuscript.
Each row corresponds to a single annual observation of a rural water service (aggregated at the service-year level, e.g., waterpoint-year or scheme-year depending on infrastructure type).
All monetary values are in USD.
Empty cells should be treated as NULL values.
SI-DFI_commitments.csv contains records of Development Finance Institutional funding commitments linked to rural water with applicable years, funding amounts, institution name, and link to documentation.
Column Dictionary (analysis_dataset.csv)
A) Identifiers and basic descriptors
year
Calendar year of the observation.wp_type_cat
Infrastructure type. Options: "Handpump", "Scheme".country_name
Country identifier used in modeling. In the public dataset, this field may be label-encoded or otherwise anonymized (see “Anonymization” below).wp_revenue_model
Categorical descriptor of the revenue model (as recorded in source data; e.g., tariff model/payment structure). Options: "waterpoint", "volumetric".volume_method
Method used to estimate or measure volume (e.g., metered, estimated, etc., as recorded). Options: "estimated", "unmeasured", "manual", "automated".quarters_used
Number of quarters within the year used/available for aggregation (data completeness indicator).
B) Location and scale variables
lat,lon
Geographic coordinates. In the public dataset these are anonymized placeholders (e.g., set to 0.0) for privacy/contractual reasons.num_taps
Number of taps associated with the service (where applicable).num_hh
Number of households served/associated (where available).wp_pop_last
Population served estimate used in analysis (latest available for the observation).num_kiosks
Number of kiosks (where applicable).num_scheme_tanks
Number of scheme tanks (where applicable).
C) Revenue and volume outcomes (primary outcomes)
wp_revenue_USD
Annual revenue collected by the service (USD). Primary outcome used in descriptive analysis and modeling.wp_m3
Annual volume (m³). Used for volumetric normalization and derived metrics (e.g., revenue per m³).
D) Rainfall covariates (previous-year monthly rainfall)
month_1…month_12
Monthly rainfall covariates (mm). These represent rainfall from the previous year relative to the observation year (e.g., for an observation in 2024, months correspond to 2023), as described in the manuscript.
E) Local service density (nearby waterpoints counts)
wp_count_within_500mwp_count_within_1000mwp_count_within_2000mwp_count_within_10000mwp_count_within_20000m
Counts of other waterpoints within the specified radius.
F) Nearby service coverage (population served by other waterpoints)
pop_served_last_within_500mpop_served_last_within_1000mpop_served_last_within_2000mpop_served_last_within_5000mpop_served_last_within_10000mpop_served_last_within_20000m
Sum of populations served by other waterpoints within the specified radius (derived from the server dataset).
G) Background population context (population within radius buffers)
Population_Within_500mPopulation_Within_1000mPopulation_Within_2000mPopulation_Within_5000mPopulation_Within_10000mPopulation_Within_20000m
Population estimates within the specified radius buffers (derived from gridded population data, as described in the manuscript).
Anonymization and Public-Release Notes
To protect sensitive infrastructure locations and comply with contractual constraints:
latandlonare anonymized placeholders in the public dataset (e.g., set to 0.0).country_namemay be label-encoded in the public dataset.
As a result, while the released data and code reproduce the full analytical workflow and model structure, exact numerical replication of location-dependent results may differ slightly from those reported in the manuscript.
Related code is openly available in the Zenodo repository under DOI 10.5281/zenodo.20908456 (https://doi.org/10.5281/zenodo.20908456)
Intended Use
This dataset is intended to support:
- Reproduction of descriptive figures and tables
- Replication of the modeling workflow (two-stage model, LOCO calibration, drift analysis, learning curves)
- Transparency of analytical methods used in the paper
It is not intended to replace raw contractual or operational datasets.