In designing anti-corruption interventions, a key challenge is to ensure that corruption is reduced rather than merely displaced, yet there is a risk that improving controls in one area will simply prompt corrupt actors to shift focus to other areas where controls are weaker. Such displacement effects, although difficult to measure precisely (Levi & Maguire, 2004; Vijlbrief, 2012), are often observed as a result of law enforcement actions against organised crime (Guerette & Bowers, 2009; Levi & Maguire, 2004; Smith, Wolanin, & Worthington, 2003; Welsh & Farrington, 2002). Because an intervention is successful in increasing the risks associated with criminal activity in one area, actors switch to activities which are less vulnerable to detection, sometimes utilising new technologies, exploiting any remaining loopholes more aggressively, or focusing on a different geographical location. Such effects seem likely to occur as a result of anti-corruption interventions too (Fisman & Golden, 2017). Indeed, some research suggests that efforts to control corruption in EU funding have perverse outcomes (Fazekas & King, 2018). However, there has been very little systematic analysis of this question. Our contribution is to analyse corruption risks in the procurement process, using a dataset of World Bank-funded national procurement over two decades in >100 developing countries. With data points from multiple stages of the procurement process as well as key outcomes, we observe the effect of interventions and assess whether risks shift to a different stage of the procurement process as a result of specific changes in Bank rules.
