Research proposal: Corrupt Political Elites in Nigeria: Who Gets Exposed and Who Gets Prosecuted? A Social Network Analysis

Research proposal: In Nigeria, corruption is pervasive and undermines both the economy and state capacity. It penetrates the police and judiciary, producing a suboptimal and often hypocritical form of anti-corruption – “the pot calling the kettle black.” In a system of competitive clientelism, corruption becomes effectively quasi-legal as leaders consolidate power by redistributing rents to supporters and clients (Oladipo Ojo et al., 2019). In this low-capacity context, key links in the corruption chain lie in informal institutions, which fill gaps left by weak formal rules but do so in ways that entrench clientelism rather than accountability. In Nigeria’s system of competitive clientelism, anti-corruption efforts therefore face strong resistance from vested interests. Many political elites are accused of corruption by civil society actors, international organizations, foreign observers, and domestic journalists, yet only a subset is actually investigated or sentenced. This raises core questions: who is publicly exposed, who ends up in court, and how are these outcomes related to elites' positions within political networks? Existing work disagrees on whether certain network positions protect or expose elites, so I do not impose a strong ex ante hypothesis. Instead, I treat elite networks as empirical objects whose observed structures can illuminate selective enforcement in a weak, clientelist state. This research examines the “grabbing hand” of the state (Shleifer & Vishny, 1998) by focusing on the social networks of corrupt elites. It analyses the emergent outcomes of these networks, showing how corrupt ties form informal institutional patterns that shape political trajectories (Luna-Pla & Nicolás-Carlock, 2020). It also explores selective punishment and factional conflict under weak state capacity, asking whether and how corrupt networks circumvent judicial sanctions and perpetuate themselves. Finally, it considers collusion, localism, and nepotism as mechanisms operating both within corruption and within anti-corruption actions. Methodologically, the project extracts political elites’ social networks from two primary sources: court records and media reports. Previous research has used scraping algorithms to capture co-mentions in news data (Mahdavi, 2017). However, this study is particularly interested in the direction and weight of ties between individuals. Large Language Models provide an opportunity to identify such relationships based on the frames, phrases, sentiments, and argumentative structures within texts. I will use court records from the Corruption Cases Database and news reports from mainstream media outlets inside and outside Nigeria. Comparing the networks extracted from these two sources will help identify patterns of exposure, prosecution, and political protection. This research examines how corrupt political elites are embedded in corrupt networks, and how these networks shape who is exposed and who is punished in a corrupt system. Theoretically, it contributes to understanding whether elite networks shield members from sanctions under competitive clientelism and how informal institutions compensate for weak formal capacity in distributing incentives and constraints. Methodologically, it develops an approach for using LLMs to extract elite social networks with finer granularity from text data, addressing data scarcity in contexts where digitalization is low, political pressure is high, or field access is dangerous, as in Nigeria.