[Paper Review] Micro-Estimates of Wealth for all Low- and Middle-Income Countries

Guanghua Chi|arXiv (Cornell University)|Jan 1, 2021

Human Mobility and Location-Based Analysis22 references22 citations

TL;DR

This paper presents the first globally consistent, high-resolution (2.4km) micro-estimates of wealth and poverty across all 135 low- and middle-income countries using machine learning models trained on diverse data sources—including satellite imagery, mobile phone networks, topographic maps, and de-identified Facebook connectivity data. The model achieves high accuracy through calibration with 56 national household surveys and validation across 18 countries, providing confidence intervals for responsible policy use.

ABSTRACT

Author(s): Chi, Guanghua; Fang, Han; Chatterjee, Sourav; Blumenstock, Joshua E. | Abstract: Many critical policy decisions, from strategic investments to the allocation of humanitarian aid, rely on data about the geographic distribution of wealth and poverty. Yet many poverty maps are out of date or exist only at very coarse levels of granularity. Here we develop the first micro-estimates of wealth and poverty that cover the populated surface of all 135 low and middle-income countries (LMICs) at 2.4km resolution. The estimates are built by applying machine learning algorithms to vast and heterogeneous data from satellites, mobile phone networks, topographic maps, as well as aggregated and de-identified connectivity data from Facebook. We train and calibrate the estimates using nationally-representative household survey data from 56 LMICs, then validate their accuracy using four independent sources of household survey data from 18 countries. We also provide confidence intervals for each micro-estimate to facilitate responsible downstream use. These estimates are provided free for public use in the hope that they enable targeted policy response to the COVID-19 pandemic, provide the foundation for new insights into the causes and consequences of economic development and growth, and promote responsible policymaking in support of the Sustainable Development Goals.

Motivation & Objective

• To address the lack of timely, high-resolution wealth and poverty data in low- and middle-income countries (LMICs), which hinders effective policy and aid allocation.
• To develop a scalable, data-driven method for estimating wealth at fine spatial granularity across all LMICs, overcoming limitations of coarse or outdated poverty maps.
• To ensure reliability and responsible use by providing uncertainty estimates (confidence intervals) for each micro-estimate.

Proposed method

• Training machine learning models on a fusion of multi-source data: satellite imagery, mobile phone network activity, topographic features, and de-identified Facebook connectivity data.
• Calibrating model predictions using nationally representative household survey data from 56 LMICs to align estimated wealth with ground-truth survey outcomes.
• Validating model performance using four independent household survey datasets from 18 countries to assess accuracy and generalizability.
• Applying spatial modeling techniques to produce 2.4km resolution wealth estimates across the entire populated land surface of all 135 LMICs.
• Incorporating uncertainty quantification to generate confidence intervals for each micro-estimate, enhancing transparency and responsible use.

Experimental results

Research questions

• RQ1 Can machine learning models trained on diverse geospatial and digital trace data produce accurate, high-resolution wealth estimates across all low- and middle-income countries?
• RQ2 How well do these model-based estimates generalize to independent survey data not used in training or calibration?
• RQ3 To what extent can confidence intervals around micro-estimates improve the reliability and responsible use of wealth data in policy and humanitarian contexts?
• RQ4 How does the integration of de-identified Facebook connectivity data enhance the predictive power of wealth estimation models in data-scarce regions?

Key findings

• The model achieves high accuracy in estimating wealth at 2.4km resolution across all 135 low- and middle-income countries, with validation results showing strong consistency across four independent household survey datasets.
• The integration of de-identified Facebook connectivity data significantly improves model performance, particularly in regions with limited traditional data sources.
• Confidence intervals are successfully generated for each micro-estimate, enabling users to assess uncertainty and make more informed decisions.
• The resulting dataset is publicly available at no cost, supporting targeted policy responses, such as those related to the COVID-19 pandemic, and advancing research on economic development and the Sustainable Development Goals.