Unequal darkness: Racial and socioeconomic disparities in nighttime pedestrian fatality risk in the United States

Background Pedestrian fatalities in the United States have reached a 40-year high, with approximately 78% now occurring at night. Darkness amplifies pedestrian crash severity, yet whether the concentration of fatalities in darkness is itself inequitable across race and income lines, and whether lighting infrastructure deficits explain any such disparity, has not yet been studied at the national level. Methods We linked 53,353 pedestrian fatalities from the Fatality Analysis Reporting System (FARS, 2015–2023) to census-tract demographics (American Community Survey), satellite nighttime radiance (Visible Infrared Imaging Radiometer Suited [VIIRS] annual composites, 2015–2023), the EPA National Walkability Index, Vision Zero (VZ) adoption status, and crash-level FARS characteristics including intersection type, driver alcohol involvement, and speed-related factors. Darkness was classified using astronomically computed civil twilight. Logistic regression models with cluster-robust standard errors (tract level) and state, year, month, road-type, and weather fixed effects estimated the associations between neighborhood disadvantage and the probability of a fatality occurring in darkness. Results Among 58,655 pedestrian fatalities with valid darkness classification, 47.8% occurred in darkness. The probability of a fatal crash occurring in darkness was significantly higher in tracts with greater poverty (OR = 1.07, 95% CI: 1.05–1.10) and higher minority concentration (OR = 1.08, 95% CI: 1.05–1.11), independent of fixed effects and crash-level controls. Brighter VIIRS radiance was associated with lower dark-crash odds (OR = 0.86), but adding radiance to the model did not attenuate the demographic gradient, providing no evidence of mediation through this tract-level proxy. Crashes at intersections were less likely to occur in darkness (OR = 0.83), while crashes involving alcohol-impaired drivers were more likely to occur in darkness (OR = 1.09). VZ adoption reduced overall nighttime risk by approximately 7% but did not narrow the equity gap. A negative control analysis of daylight crashes confirmed these associations are darkness-specific. Findings were robust to alternative darkness definitions, urban restriction, temporal subperiods, and geographic exclusion. Conclusions The disproportionate concentration of pedestrian fatalities in darkness in disadvantaged and minority neighborhoods persists independently of measured lighting levels and crash-level confounders. Current VZ frameworks provide uniform nighttime safety benefits without closing the equity divide. These findings can inform targeted infrastructure investments: lighting improvements in high-poverty corridors, enhanced intersection illumination, and nighttime alcohol-impaired driving enforcement may reduce disparities, though the equity gap likely requires comprehensive built-environment interventions beyond lighting alone.