Load shedding and crime in South Africa: causal estimates and distributional welfare of the energy-apartheid externality

South Africa's continuous load shedding since 2007 — peaking in FY2022 with national outage hours exceeding 7,000 and independent GDP-cost estimates of 1–5 percentage points — generates a national natural experiment in electricity-supply disruption in a middle-income developing economy. Using a panel of 1,140 SAPS police stations over FY2008–FY2025 joined to per-area outage hours and per-station daily VIIRS nighttime radiance, and a Borusyak–Jaravel–Spiess imputation estimator addressing differential pre-trends, 100 additional outage hours per quarter reduce residential burglary by 1.26% and theft from motor vehicle by 1.23% (both p<0.0001) on the per-area metropolitan subsample. A direct VIIRS first stage (β=-2.08% mean radiance per +10 hours, F=121, N=2.0 million station-day cells) identifies the lighting-channel mechanism. These crime-reduction benefits accrue regressively: applied to FY2022's peak intensity, the ~70,000 fewer property crimes flow 1.6× more to higher-income, predominantly white-majority, rooftop-solar-adopting suburbs than to lower-income, black-majority, grid-dependent counterparts — an energy-apartheid externality that places the post-apartheid distributional cost of infrastructure failure on the same households who already bear South Africa's largest inequality burdens, nationally extending Imelda and Guo (2024).