Abstract

Disentangling the causal effect of structural barriers — electoral violence, polling-unit inaccessibility, and institutional distrust — from intrinsic voter disengagement requires credible identification strategies rarely applied in the African electoral literature. This paper exploits two quasi-experiments: a regression discontinuity design around court-ordered election nullifications in Nigerian governorship elections (which randomise institutional disruption), and an instrumental variable strategy using colonial-era road networks as instruments for current polling-unit proximity. Spatial econometric models further identify the geographic spillover of electoral violence on neighbouring constituencies, revealing that turnout suppression radiates well beyond directly affected polling units. The results imply that supply-side institutional failures account for a larger share of observed non-participation than demand-side apathy, with important implications for policy sequencing in electoral reform programmes.