Anti-corruption in aid-funded procurement: Is corruption reduced or merely displaced?

Given a widespread sense among donors that mainstream anti-corruption reforms over the past 25 years have failed to yield results, there is a move towards more targeted interventions. Such interventions should, in principle, overcome implementation gaps and make it easier to evaluate impact, supporting learning. However, when interventions are narrowly targeted, there is a risk that corrupt actors simply adapt, shifting their focus to areas with weaker controls, so that overall corruption is not reduced but merely displaced. We analyse data points from World Bank-funded development aid tenders over 12 years in >100 developing countries, and observe the heterogeneous effects of a 2003 anti-corruption reform aimed at increasing oversight and opening up competition. Our tight matching estimations suggest that the reform is effective in the targeted area: it decreases corruption risks due to low competition (the share of single bidding falls from 22% to18%). But we also find that evasive tactics largely cancel out these positive direct effects: buyers switch to non-treated less competitive procedure types (whose share increases from 7% to 10%) and exploit them more intensively (single bidding goes from 61% to 81%). Our results demonstrate how data analytics can be used to observe public procurement at the system level to inform more adaptive and effective anti-corruption programming. More broadly, we underline that technical interventions might not represent the best way to tackle systemic corruption, instead strategies should target the root causes of corruption and contribute to building a culture of integrity.