Statistical Theory for Network-Based Contagion Models with Application to Financial Systemic Risk in the Nigerian Banking System

Systemic risk in the Nigerian banking system arises from interbank network exposures where the failure of one institution can cascade through the network, and its statistical quantification requires contagion network models with rigorous theoretical foundations that move beyond correlation-based approaches. This dissertation develops statistical theory for network-based contagion models applied to financial systemic risk in the Nigerian banking system, making original theoretical contributions to financial network statistics. A random graph contagion model with heterogeneous node fitness and time-varying edge weights is specified and its cascade failure probability distribution derived analytically under large-network asymptotic conditions. New estimators for network contagion parameters using observable interbank exposure data are introduced and their asymptotic properties established. Identifiability conditions distinguishing contagion from common shock explanations of simultaneous bank stress are derived. The theoretical framework is applied to interbank bilateral exposure data for 22 Nigerian commercial banks from the CBN credit register (2017 to 2022). Estimated cascade failure probability under the loss of the three most systemically important banks was 0.37, indicating substantial systemic fragility. Bayesian credible intervals quantified estimation uncertainty. Network topology measures showed that the Nigerian interbank network became significantly more concentrated following the 2018 CBN recapitalisation requirement. The dissertation provides CBN financial stability department with a statistically rigorous systemic risk measurement framework and recommends network-adjusted capital surcharges for systemically important Nigerian banks. Keywords: network contagion model, financial systemic risk, Nigerian banking, random graph model, systemic risk statistics