SPPAT-Recorded Road-Traffic Fatalities in Ecuador, 2016–2021: Denominator Audit, Rate Uncertainty and Typology Indeterminacy as Proposition-Based Evidence Coordinates

Resumen

Administrative registries of road-traffic deaths support territorial comparison only when their denominators, periods and missing values are audited. This study reanalysed the 16,526 records of victims registered by Ecuador's Public Service for Traffic Accident Payments (SPPAT) for fatal protection coverage (24 provinces; 208 territorial labels, many of them cantonal capitals rather than canton names). Every figure of the original analysis was reproduced: 1,767 records in Quito and 1,398 in Guayaquil; 163 DESCONOCIDO and 88 empty crash-type values (1.52 %); and, with the authors' denominators, annualised rates of 56.45 per 100,000 in Machachi (Mejía), 49.39 in Naranjal and 47.21 in Yaguachi against 10.79 in Quito and 8.71 in Guayaquil. The denominator audit found that the period used (5.89 years) does not correspond to the registry's creation window (5.64 years) or its crash-date window (5.85 years), that 163 events pre-date 2016, that several denominators are rounded approximations, and that the reported ranking was computed over seventeen selected cantons. Recomputing the rates as person-years from the contemporaneous INEC cantonal projections for 2010–2020, which matched every one of the 16,363 events in the period, Machachi reaches 54.9 per 100,000 per year (95 % Poisson 49.2–61.1) and ranks 6th of 207, behind Amazon and highland cantons such as Loreto, Patate and Colta that the original selection did not consider. Pre-specified rate ratios of the three corridor cantons to Quito and Guayaquil ranged from 4.5 to 6.2 and were robust to the denominator scenario. A documentary check against the published route of the two trunk roads confirms the corridor reading for Machachi, Latacunga, Naranjal and Quevedo and does not confirm it for Yaguachi, Babahoyo or Guano. Typology indeterminacy was analysed as proposition-based evidence coordinates: for each canton and crash type k, ,  and  are the shares of records that assert k, are indeterminate (sentinel or empty), or assert another type; normalisation to one is a design choice here, and no independence is claimed. Indeterminacy propagates into identification bounds on typology rates that are narrow (30.2–31.8 % for run-overs nationally) and are reported separately from the Poisson sampling layer. The chi-square evidence for territorial clustering of indeterminacy survives a permutation test but with 56 % of expected cells below five and a Cramér's V of 0.17 (0.07 after grouping); the corridor membership of the high-rate cantons is asserted, not verified geospatially. Rates are SPPAT-recorded fatality rates, not mortality or risk.