electricsheepafrica/africa-poverty-all

---
annotations_creators:
- no-annotation
language_creators:
- found
language:
- en
license: other
multilinguality:
- monolingual
size_categories:
- 10K<n<100K
source_datasets:
- original
task_categories:
- tabular-regression
- other
task_ids: []
tags:
- africa
- humanitarian
- hdx
- electric-sheep-africa
- economics
- poverty
- socioeconomics
- dza
- ago
- aia
- atg
- arg
pretty_name: "Relative Wealth Index"
dataset_info:
splits:
- name: train
num_examples: 66276
- name: test
num_examples: 16569
---

# Relative Wealth Index

**Publisher:** AI for Good at Meta · **Source:** [HDX](https://data.humdata.org/dataset/relative-wealth-index) · **License:** `hdx-other` · **Updated:** 2026-03-26

---

## Abstract

The Relative Wealth Index predicts the relative standard of living within countries using de-identified connectivity data, satellite imagery and other nontraditional data sources. The data is provided for 93 low and middle-income countries at 2.4km resolution. Please cite / attribute any use of this dataset using the following:
Microestimates of wealth for all low- and middle-income countries
Guanghua Chi, Han Fang, Sourav Chatterjee, Joshua E. Blumenstock
Proceedings of the National Academy of Sciences Jan 2022, 119 (3) e2113658119; DOI: 10.1073/pnas.2113658119

More details are available here: hhttps://ai.meta.com/ai-for-good/datasets/relative-wealth-index/

Research publication for the Relative Wealth Index is available here: https://www.pnas.org/content/119/3/e2113658119

Press coverage of the release of the Relative Wealth Index here: https://www.fastcompany.com/90625436/these-new-poverty-maps-could-reshape-how-we-deliver-humanitarian-aid

An interactive map of the Relative Wealth Index is available here: http://beta.povertymaps.net/

Each row in this dataset represents geolocated point observations. Data was last updated on HDX on 2026-03-26. Geographic scope: **DZA, AGO, AIA, ATG, ARG, ABW, BHS, BGD, and 118 others**.

*Curated into ML-ready Parquet format by [Electric Sheep Africa](https://huggingface.co/electricsheepafrica).*

---

## Dataset Characteristics

| | |
|---|---|
| **Domain** | Poverty and economic vulnerability |
| **Unit of observation** | Geolocated point observations |
| **Rows (total)** | 82,845 |
| **Columns** | 7 (5 numeric, 2 categorical, 0 datetime) |
| **Train split** | 66,276 rows |
| **Test split** | 16,569 rows |
| **Geographic scope** | DZA, AGO, AIA, ATG, ARG, ABW, BHS, BGD, and 118 others |
| **Publisher** | AI for Good at Meta |
| **HDX last updated** | 2026-03-26 |

---

## Variables

**Geographic** — `quadkey` (range 12032231223323.0–12211210020301.0), `latitude` (range 35.8267–42.0738), `longitude` (range 25.697–44.8132).

**Identifier / Metadata** — `esa_source` (HDX), `esa_processed` (2026-04-20).

**Other** — `rwi` (range -1.248–1.917), `error` (range 0.346–0.922).

---

## Quick Start

```python
from datasets import load_dataset

ds = load_dataset("electricsheepafrica/africa-poverty-all")
train = ds["train"].to_pandas()
test = ds["test"].to_pandas()

print(train.shape)
train.head()
```

---

## Schema

| Column | Type | Null % | Range / Sample Values |
|---|---|---|---|
| `quadkey` | int64 | 0.0% | 12032231223323.0 – 12211210020301.0 (mean 12187910438678.678) |
| `latitude` | float64 | 0.0% | 35.8267 – 42.0738 (mean 39.0437) |
| `longitude` | float64 | 0.0% | 25.697 – 44.8132 (mean 34.5419) |
| `rwi` | float64 | 0.0% | -1.248 – 1.917 (mean 0.0139) |
| `error` | float64 | 0.0% | 0.346 – 0.922 (mean 0.495) |
| `esa_source` | object | 0.0% | HDX |
| `esa_processed` | object | 0.0% | 2026-04-20 |

---

## Numeric Summary

| Column | Min | Max | Mean | Median |
|---|---|---|---|---|
| `quadkey` | 12032231223323.0 | 12211210020301.0 | 12187910438678.678 | 12210301021001.0 |
| `latitude` | 35.8267 | 42.0738 | 39.0437 | 38.9338 |
| `longitude` | 25.697 | 44.8132 | 34.5419 | 34.5959 |
| `rwi` | -1.248 | 1.917 | 0.0139 | -0.034 |
| `error` | 0.346 | 0.922 | 0.495 | 0.495 |

---

## Curation

Raw data was downloaded from HDX via the CKAN API and converted to Parquet. Column names were lowercased and standardised to snake_case. Common missing-value markers (`N/A`, `null`, `none`, `-`, `unknown`, `no data`, `#N/A`) were unified to `NaN`. The dataset was split 80/20 into train and test partitions using a fixed random seed (42) and saved as Snappy-compressed Parquet.

---

## Limitations

- Data originates from AI for Good at Meta and has not been independently validated by ESA.
- Automated cleaning cannot correct for misreported values, definitional inconsistencies, or sampling bias in the original collection.
- This dataset spans 126 countries; geographic and methodological inconsistencies across national boundaries may affect cross-country comparability.
- Refer to the [original HDX dataset page](https://data.humdata.org/dataset/relative-wealth-index) for the publisher's own methodology notes and caveats.

---

## Citation

```bibtex
@dataset{hdx_africa_poverty_all,
title = {Relative Wealth Index},
author = {AI for Good at Meta},
year = {2026},
url = {https://data.humdata.org/dataset/relative-wealth-index},
note = {Repackaged for machine learning by Electric Sheep Africa (https://huggingface.co/electricsheepafrica)}
}
```

---

*[Electric Sheep Africa](https://huggingface.co/electricsheepafrica) — Africa's ML dataset infrastructure. Lagos, Nigeria.*