Using machine learning and corruption risks indicators to identify politically exposed firms

Corruption in public procurement has a direct impact on the quality and quantity of public goods and services. One way this happens is when a Politically Exposed Person (PEP), interferes with a procurement process. Nonetheless, the role of PEPs in procurement processes is still poorly understood because we lack micro-level data on the relationship between corruption in public procurement processes and politically connected firms. In light of this, the goal of this paper is to identify political connections indirectly using machine learning methods, firm census data, and politically exposed firms' information. The objective is to use this indirect approach to create an indicator of what companies' characteristics make them more prone to be politically connected. The results of this indicator would then be validated by analyzing the relation with a Corruption Risks Indicator (CRI), built by considering red flags in procurement process data. The rationale of the paper is that given that it is so hard to identify a few links between PEPs and companies even after lots of meticulous manual research, some smart indirect ways to identify political connections might work more readily and might be more replicable with better results.