Does choosing BART station closures by fares beat the staff's "10 lowest ridership" list? A pre-registered test finds only a small gain

BART staff proposed closing the ten lowest-ridership stations if a regional tax measure fails. We ask whether choosing the ten closures by origin-destination fares loses fewer list-fare dollars. Using public BART ridership data and published list fares, we pre-registered the rule on 2025 data and scored it once on nine held-out 2026 months (January to September). The fare-based set differs from the staff list by one station (Colma closed instead of San Bruno) and loses 0.63% fewer list fares (95% interval 0.48% to 0.80%, about $0.23 million), far below the pre-registered 5% bar. An exact solve finds no better set of ten. The swap reverses if more than about 6% of riders at a closed station switch stations, and the gain fell from 1.47% in 2025 to 0.63% in 2026. Operating costs and real revenue were not measured. Eight of 13 predictions hit. The regional measure passed in November 2026 (unofficial results), so closures are now unlikely; the comparison is about the ranking rule, not an imminent decision. Data credit: ridership data from the San Francisco Bay Area Rapid Transit District (BART) monthly origin-destination workbooks (bart.gov/about/reports/ridership, Creative Commons Attribution License), used in aggregated form; list fares and station locations from BART's GTFS feed under BART's Developer License Agreement. This work is not endorsed by BART. Paper: CC BY 4.0. Code in supplement.zip: MIT licence.