Research
This thesis examines if and when Official Development Assistance (ODA) reduces intrastate conflict deaths in recipient countries, drawing on a dataset of 156 countries from 1989 to 2022. Two mathematical methods are used: Dynamic Panel GMM and a Causal Forest model.
The analysis finds that ODA, on average, does not reliably reduce conflict deaths. The strongest predictor of deaths in any year is the prior year’s deaths, reflecting a self-reinforcing “conflict trap” that no form of ODA consistently breaks. Beneath the ineffective averages, four conditions that shape ODA’s effects were found:
- Political stability matters most (below a threshold, effects turn harmful and unpredictable)
- Aid fragmentation into smaller projects reduces conflict risk by presenting less attractive targets for sabotage
- Multilateral channels outperform bilateral ones in fragile settings
- Inconsistent levels of conflict follow purpose-specific ODA.
Five policy implications follow:
- First, detailed local assessments should precede any disbursement, since country-level indicators alone cannot determine whether aid is safe to deliver.
- Second, political stability should be heavily considered in determining ODA allocation.
- Third, aid in fragile contexts should be channelled through multilateral institutions, which are more resilient to geopolitical perceptions and aid shocks.
- Fourth, prevention should be prioritised over de-escalation, since entrenched conflict rarely responds to aid.
- Fifth, these aggregate findings should inform, but never replace, context-specific, project-level analysis.
Link to the paper: No Bang for Buck | TU Delft Repository
Author: Emma Bokel, Data Scientist HCSS
This thesis was published on the TU Delft Repository
First supervisor: Yilin Huang, TU Delft
Chair and Second supervisor: Hans de Bruijn, TU Delft
Third supervisor: Jesse Kommandeur, Datalab Manager HCSS
Related Content
