Moving run time from fast trips to slow ones: a pre-registered test on MBTA bus schedules
Objective. Bus agencies set scheduled run times in advance; for its headway-managed trips, the MBTA treats a trip as not on time at its last stop when its run time is more than 120% of its scheduled run time. I ask whether moving scheduled run-time hours toward slow trips, with total hours unchanged, lowers failures. Data. Public MBTA bus arrival and departure data. A pre-registered rule sets each route, direction, terminal pair and start-hour budget from its April to June 2026 average run time, scaled to equal the published schedule's hours, and was scored once on 64 held-out July to September 2026 weekdays (683,346 scored trips). Findings. On paper (a schedule simulation: actual run times are held fixed and only the budgets change), the share of trips failing a 120% run-time test (the MBTA's headway-trip test, applied here to all trips) fell from 13.10% to 10.86% (-2.24 percentage points; 95% interval -3.62 to -0.97). Five caveats limit this. The gain vanishes when hours must stay equal within each route, direction and period (+0.12). It is inconclusive in September 2026, which mixes 4 summer-schedule weekdays with 17 after the MBTA's fall schedule change (-0.37; 95% interval -1.87 to +0.93). And on the headway-standard trips that the MBTA's test actually governs it is inconclusive (-1.19; 95% interval -3.65 to +0.88). Training on July to September 2025 instead gives -1.00 (95% interval -2.41 to +0.39), so part of the gain may come from fresher data. And 15.7% of test trips lacked an end time and were not scored. The cost of moving hours was not measured. What is new. A network-wide, equal-hours re-score against the 120% test on a held-out, pre-registered quarter. Practical use. A screening check for schedule planners, not evidence that the MBTA should change its schedules.
Authors
- Jayden Ling
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-16
- DOI
- https://doi.org/10.5281/zenodo.24280937
- Primary Topic
- Transportation Planning and Optimization
- Type
- preprint