Act I · The Question
Does Development Aid Shift Local Wealth Trajectories?
This research explores how development projects from China and the World Bank are associated with changes in local wealth across African neighborhoods.
The results show, across our models, that treated areas are already on declining wealth trajectories before projects begin. This means that aid is often directed toward places that are falling behind.
23 donor-sector combinations
4 statistical models
Data spans more than a decade
Geographic scope: African neighborhoods from 35 countries observed between 2002 and 2013.
Data sources: International Wealth Index (IWI), a household asset-based measure of living standards built from DHS and satellite-linked data, combined with machine learning–based wealth maps from Pettersson (2023) with colleagues, who use satellite imagery and neural networks to predict neighborhood-level living standards across Africa.
Event time: Estimates use three-year windows around first exposure within 5 km: t−1 covers the three years before project start, and t+1, t+2, t+3 track years 0–3, 3–6, and 6–9 after first exposure.
Explore the Evidence
How We Compare Areas
A Simple Comparison Strategy
Projects start at different times in different places. The strategy compares outcomes before and after aid begins, while also comparing against similar places where aid has not yet started.
When project timing varies across locations, that variation helps create fairer comparisons. We therefore use a method designed for settings where projects begin at different times.
The outcome is the International Wealth Index (IWI), a local asset-based wealth measure.
Projects begin at different times across locations. This variation in timing allows comparison between areas before and after projects begin.
Act II · What We Find
What We Find
Systematic Targeting Toward Declining Areas
Before projects begin, many treated areas were already on declining wealth paths. This shows that aid is often directed toward places that are falling behind rather than already growing.
Most treated locations are already on declining wealth paths before projects begin.
Pre-project estimates show that aid is often directed toward areas experiencing economic downturn.
The targeting pattern is consistent across donors and models.
Negative pre-project trends are observed for both China and the World Bank and across estimators.
Both funders show comparable targeting dynamics.
Pre-treatment declines are visible in the majority of panels for both China and World Bank.
Aid Is Directed Toward Declining Areas
Share of projects where treated areas were already declining before aid arrived. Declining means the pre-project estimate (t-1) is negative and statistically distinguishable from zero.
A high share suggests aid is systematically directed toward weaker areas.
Wealth Trajectories
Changes in local wealth before and after projects begin.
t-1 shows pre-project trends.
t+1 to t+3 show temporal changes after projects start.
If lines rise after t=0, projects are associated with improving wealth trajectories.
Funder Comparison
Average long-term wealth changes associated with China and World Bank projects.
Comparing donor portfolios: long-term average changes by pre-project and post-project windows.
Act III · Average Effects
What do the average temporal effects show?
This section compares average temporal effects across alternative estimators. Because projects begin at different points in time, we rely on methods designed for staggered treatment adoption and assess whether the estimated patterns remain stable across specifications.
Act IV · What This Means
What This Tells Us — and What It Doesn’t
These estimates summarize average patterns. They do not prove that any individual project caused a specific local outcome.
What it shows
- Average wealth trajectory changes after projects begin.
- Differences across sectors and donors.
- Evidence that aid targets weaker areas.
What it does not show
- Whether every individual project works.
- National-level economic growth effects.
- A ranking of countries.
Important Assumptions
- The analysis assumes treated and comparison areas would have followed similar trends without aid.
About This Research
AI & Global Development Lab
The AI & Global Development Lab fuses AI with Earth Observation to illuminate the causes and consequences of human development across time and space.
The lab is an interdisciplinary team that develops AI-based methods to study global challenges such as poverty, conflict, sustainability, and policy effectiveness.
Using satellite imagery and large-scale data, we reconstruct development trajectories to generate new insights into global change.
