This case discusses the use of data analysis and instrumental variables to study the drivers of corruption. Drawing on a quantitative analysis of the link between municipal flood assistance and local corruption in Bulgaria between 2004 and 2005, it explores the challenges the authors faced when collecting and coding the data used in their work, and how they managed to overcome or minimize them. It also discusses the advantages of corruption measured with objective data as opposed to survey-based perceptions. It highlights the issues of reverse causality and omitted variable bias which arise in a simple regression of corruption on flood assistance. It then explains how the instrumental variables technique can overcome these issues and applies it to the context under study. Finally, it takes a broad view on the practical lessons learned while conducting this research project and concludes that while the research project was full of challenges and frustrations, it was also very useful and rewarding
