Quiet Where the Sign Has No Name

Probabilistic Linkage of 6,142 Driver Occurrence Reports to a 2,388-Crossing Register, and the Near-Miss Rate Left Behind by What Would Not Resolve

A paper from the School of Continuous Improvement re-estimates the attribution step behind a state rail safety regulator's published level-crossing occurrence rate, applying a Fellegi-Sunter linkage model with kilometrage blocking and clerical review to 6,142 driver-filed occurrence reports lodged against 2,388 public crossings over five financial years. At the operating point that reproduces the published counts, 71.4 per cent of reports resolve to a unique crossing, and the resolved share moves 21.6 points across three defensible thresholds. Reports naming a road and a locality resolve at 94.6 per cent; reports giving only a rail kilometrage, as the signage at 1,043 of the crossings does, resolve at 34.7 per cent. Those crossings are 43.7 per cent of the network and 11.2 per cent of the published occurrences, and against a 400-report hand-resolved audit sample their occurrence rate rises from 0.06 to 0.33 per crossing-year, above that of the named-road crossings. Recomputed on the corrected rate, 51 of the 120 crossings the education campaign and current upgrade tranche were directed at change.

@misc{slop_ba3177,
  author       = {Thandiwe Solberg and Sten Okwuosa and Bram Ntuli},
  title        = {Quiet Where the Sign Has No Name: Probabilistic Linkage of 6,142 Driver Occurrence Reports to a 2,388-Crossing Register, and the Near-Miss Rate Left Behind by What Would Not Resolve},
  year         = {2026},
  publisher    = {Slop University},
  doi          = {10.5555/slop.ba3177},
  url          = {https://slop.university/outputs/slop-paper-the-boom-gates-ba3177/},
  version      = {1.0},
  note         = {Research paper},
}