Using publicly available satellite imagery and deep learning to understand economic well-being in Africa
Open accessAbstract
Accurate and comprehensive measurements of economic well-being are fundamental inputs into both research and policy, but such measures are unavailable at a local level in many parts of the world. Here we train deep learning models to predict survey-based estimates of asset wealth across ~ 20,000 African villages from publicly-available multispectral satellite imagery. Models can explain 70% of the variation in ground-measured village wealth in countries where the model was not trained, outperforming previous benchmarks from high-resolution imagery, and comparison with independent wealth measurements from censuses suggests that errors in satellite estimates are comparable to errors in existing ground data. Satellite-based estimates can also explain up to 50% of the variation in district-aggregated changes in wealth over time, with daytime imagery particularly useful in this task. We demonstrate the utility of satellite-based estimates for research and policy, and demonstrate their scalability by creating a wealth map for Africa’s most populous country.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Differences between urban and rural subpopulations have been well-documented.
“Differences between urban and rural subpopulations have also been well-documented (Jean et al., 2016; Yeh et al., 2020).”
From WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020) · cites this paper as Yeh et al., 2020
Yeh et al. (2020) split the ID train/val/test while ensuring that the spatial extent of images between each split never overlap, while we simply take uniformly random splits of the ID data.
“However, Yeh et al. (2020) split the ID train/val/test while making sure that the spatial extent of the images between each split never overlap, while we simply take uniformly random splits of the ID data.”
From WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020) · cites this paper as Yeh et al., 2020
We report a much larger drop in correlation due to spatial shift than in Yeh et al. (2020).
“We report a much larger drop in correlation due to spatial shift than in Yeh et al. (2020).”
From WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020) · cites this paper as Yeh et al., 2020
The PovertyMap-wilds dataset is derived from Yeh et al. (2020), which gathers LandSat imagery and DHS data from 19669 villages across 23 countries in Africa.
“The PovertyMap-wilds dataset is derived from Yeh et al. (2020), which gathers LandSat imagery and Demographic and Health Surveys (DHS) data from 19669 villages across 23 countries in Africa .”
From WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020) · cites this paper as Yeh et al., 2020
Our data splitting method for PovertyMap-wilds is slightly different from Yeh et al. (2020), allowing us to compare ID and OOD metrics simultaneously on one model as a more direct comparison.
“To explain this, we note that our data splitting method is slightly different from theirs. They have two separate experiments (with different data splits) to test in-distribution vs. out-of-distribution generalization. In contrast, our data splits on both held-out in-distribution and out-of-distribution points at the same time with respect to the same training set, thus allowing us to compare both”
From WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020) · cites this paper as Yeh et al., 2020
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
He, K., Zhang, X., Ren, S., & Sun, J. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.123 · Cited by 19,741 · see citing works
Measuring Economic Growth from Outer Space
Henderson, J. V., Storeygard, A., & Weil, D. N. (2012) · American Economic Review · doi:10.1257/aer.102.2.994 · Cited by 2,596 · see citing works
Flagged first, then by citations
Research works printed but not linked: 30
- [1]Osgood-Zimmerman A Mapping child growth failure in Africa between 2000 and 2015 Nature 2018 555 41 10.1038/nature25760 29493591 PMC6346257 · DOI
- [2]Reiner RC Jr Variation in childhood diarrheal morbidity and mortality in Africa, 2000–2015 N. Engl. J. Med. 2018 379 1128 1138 10.1056/NEJMoa1716766 30231224 PMC6078160 · DOI
- [3]Graetz N Mapping local variation in educational attainment across Africa Nature 2018 555 48 10.1038/nature25761 29493588 PMC6346272 · DOI
- [4]Burke M Heft-Neal S Bendavid E Sources of variation in under-5 mortality across sub-Saharan Africa: a spatial analysis Lancet Glob. Health 2016 4 e936 e945 10.1016/S2214-109X(16)30212-1 27793587 · DOI
- [5]Tusting LS Mapping changes in housing in sub-Saharan Africa from 2000 to 2015 Nature 2019 568 391 394 10.1038/s41586-019-1050-5 30918405 PMC6784864 · DOI
- [8]Jean N Combining satellite imagery and machine learning to predict poverty Science 2016 353 790 794 10.1126/science.aaf7894 27540167 · DOI · Paper page
- [9]Babenko, B. et al. Poverty mapping using convolutional neural networks trained on high and medium resolution satellite images, with an application in Mexico. in Proc. NIPS 2017 Workshop on Machine Learning for the Developing World. https://arxiv.org/abs/1711.06323. (Neural Information Processing Systems Foundation, San Diego, CA, 2017). · arXiv
- [11]Head, A., Manguin, M., Tran, N. & Blumenstock, J. E. Can human development be measured with satellite imagery? in Proc. Ninth International Conference on Information and Communication Technologies and Development, ICTD ’17, 8:1–8:11 (ACM, New York, 2017). https://dl.acm.org/citation.cfm?doid=3136560.3136576.
- [12]Watmough GR Socioecologically informed use of remote sensing data to predict rural household poverty Proc. Natl Acad. Sci. USA 2019 116 1213 1218 10.1073/pnas.1812969116 30617073 PMC6347693 · DOI
- [13]Sahn DE Stifel D Exploring alternative measures of welfare in the absence of expenditure data Rev. Income Wealth 2003 49 463 489 10.1111/j.0034-6586.2003.00100.x · DOI · Paper page
- [14]Filmer D Scott K Assessing asset indices Demography 2012 49 359 392 10.1007/s13524-011-0077-5 22135117 · DOI · Paper page
- [15]Alkire, S. et al. Multidimensional Poverty Measurement and Analysis (Oxford University Press, USA, 2015).
- [16]Filmer D Pritchett LH Estimating wealth effects without expenditure data-or tears: an application to educational enrollments in states of India Demography 2001 38 115 132 11227840
- [17]Blumenstock J Cadamuro G On R Predicting poverty and wealth from mobile phone metadata Science 2015 350 1073 1076 10.1126/science.aac4420 26612950 · DOI
- [18]Pokhriyal N Jacques DC Combining disparate data sources for improved poverty prediction and mapping Proc. Natl Acad. Sci. USA 2017 114 E9783 E9792 10.1073/pnas.1700319114 29087949 PMC5699027 · DOI
- [19]Burke M Hsiang SM Miguel E Global non-linear effect of temperature on economic production Nature 2015 527 235 10.1038/nature15725 26503051 · DOI · Paper page
- [20]Nordhaus WD Geography and macroeconomics: New data and new findings Proc. Natl Acad. Sci. USA 2006 103 3510 3517 10.1073/pnas.0509842103 16473945 PMC1363683 · DOI
- [21]Fick SE Hijmans RJ Worldclim 2: new 1-km spatial resolution climate surfaces for global land areas Int. J. Climatol. 2017 37 4302 4315 10.1002/joc.5086 · DOI
- [22]Dell M Jones BF Olken BA What do we learn from the weather? the new climate-economy literature J. Economic Lit. 2014 52 740 798 10.1257/jel.52.3.740 · DOI · Paper page
- [24]Sheehan, E. et al. Predicting economic development using geolocated Wikipedia articles. In Proc. 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2698–2706 (Association for Computing Machinery, New York, NY, 2019).
- [26]Azzari G Lobell DB Landsat-based classification in the cloud: an opportunity for a paradigm shift in land cover monitoring Remote Sens. Environ. 2017 202 1 11 10.1016/j.rse.2017.09.013 · DOI
- [27]Hsu F-C Baugh K Ghosh T Zhizhin M Elvidge C DMSP-OLS radiance calibrated nighttime lights time series with intercalibration Remote Sens. 2015 7 1855 1876 10.3390/rs70201855 · DOI
- [28]Elvidge CD Baugh K Zhizhin M Hsu FC Ghosh T Viirs night-time lights Int. J. Remote Sens. 2017 38 5860 5879 10.1080/01431161.2017.1342050 · DOI
- [29]Gorelick, N.et al. Google earth engine: Planetary-scale geospatial analysis for everyone. in Remote Sensing of Environment (Elsevier, 2017). 10.1016/j.rse.2017.06.031. · DOI
- [30]He, K., Zhang, X., Ren, S. & Sun, J. Identity Mappings in Deep Residual Networks. in Computer Vision—ECCV 2016, (eds Leibe, B., Matas, J., Sebe, N. & Welling, M.), 630–645 (Springer International Publishing, 2016). 10.1007/978-3-319-46493-0_38 https://arxiv.org/abs/1603.05027. · DOI · Paper page
- [31]Russakovsky O ImageNet large scale visual recognition challenge Int. J. Computer Vis. 2015 115 211 252 10.1007/s11263-015-0816-y · DOI · Paper page
- [32]Perez, A. et al. Poverty prediction with public Landsat 7 satellite imagery and machine learning. in Proc. NIPS 2017 Workshop on Machine Learning for the Developing World, (Long Beach, Neural Information Processing Systems Foundation, San Diego, CA, 2017). http://arxiv.org/abs/1711.03654. · arXiv
- [34]Kingma, D. P. & Ba, J. Adam: A method for stochastic optimization. arXiv. Preprint at http://arxiv.org/abs/1412.6980 (2014). · arXiv
- [35]Molina I Nandram B Rao J Small area estimation of general parameters with application to poverty indicators: a hierarchical bayes approach Ann. Appl. Stat. 2014 8 852 885 10.1214/13-AOAS702 · DOI
- [36]Elbers C Lanjouw JO Lanjouw P Micro-level estimation of poverty and inequality Econometrica 2003 71 355 364 10.1111/1468-0262.00399 · DOI
Cited by
Sorted by citations
Generalizing to unseen domains: A survey on domain generalization
Wang, J., Lan, C., Liu, C., Ouyang, Y., Qin, T., Lu, W., Chen, Y., Zeng, W., & Yu, P. (2022) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2022.3178128 · Cited by 1,046 · see citing works
WILDS: A Benchmark of in-the-Wild Distribution Shifts
Koh, P. W., Sagawa, S., Marklund, H., Xie, S. M., Zhang, M., Balsubramani, A., Hu, W., Yasunaga, M., Phillips, R. L., Gao, I., et al. (2020) · arXiv · doi:10.48550/arxiv.2012.07421 · Cited by 518 · see citing works
“PovertyMap-wilds is based on a dataset collected by Yeh et al. (2020), which assembles satellite imagery and survey data at 19,669 villages from 23 African countries between 2009 and 2016 (Figure 4).”
InfluentialPredicting and explaining with machine learning models: Social science as a touchstone
Buchholz, O., & Grote, T. (2023) · Studies in History and Philosophy of Science · doi:10.1016/j.shpsa.2023.10.004 · Cited by 15 · see citing works
Under the Radar -- Auditing Fairness in ML for Humanitarian Mapping
Kondmann, L., & Zhu, X. X. (2021) · arXiv · doi:10.48550/arxiv.2108.02137 · Cited by 3 · see citing works
- Authors
- Yeh, C., Perez, A., Driscoll, A., Azzari, G., Tang, Z., Lobell, D., Ermon, S., & Burke, M.
- Affiliations
- Stanford University (US)
- Stanford Institute for Economic Policy Research (US)
- National Bureau of Economic Research (US)
- Corresponding author
- Burke, Marshall
- ORCID
- Venue
- Nature Communications
- Published
- 2020-05-22
- Volume
- 11
- Issue
- 1
- Pages
- 2583-2583
- ISSN
- 2041-1723 (ISSN-L)
- PMC
- PMC7244551
- OpenAlex
- W3027716283
- Cited by
- 425
- References
- 2
- Citations by year
- 2026: 49 · 2025: 82 · 2024: 77 · 2023: 68 · 2022: 68 · 2021: 62 · 2020: 19
- FWCI
- 28.82
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- Nature Communications (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- RePEc: Research Papers in Economics (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Science Foundation (1651565)
- US Agency for International Development Stanford King Center for Global Development
- Topics
- Physical Sciences › Environmental Science › Ecology › Remote Sensing in Agriculture
- Physical Sciences › Environmental Science › Global and Planetary Change › Impact of Light on Environment and Health
- Physical Sciences › Engineering › Media Technology › Remote-Sensing Image Classification
- Keywords
- economic well-being, satellite imagery, multispectral imagery, deep learning, rural villages, Africa, household wealth measurement
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
