Uber versus Trains? Worldwide Evidence from Transit Expansions

We study how ride-hailing and public transit interact – an open question with importantimplications for transportation policy and urban form. We develop a model of mode choice that features three key mechanisms: (i) substitution, whereby riders switch from Uber to transit near stations; (ii) last-mile complementarity, whereby Uber is used to access transit; and (iii) a dynamic choice-set effect, whereby outbound mode choice constrains return options, givingUber option value. Using proprietary Uber trip data around 650 rail station openings worldwide, we estimate how ride-hailing responds when public transit improves. We document three patterns consistent with these mechanisms: Uber usage rises sharply within 100 meters of new stations (last-mile complementarity), declines locally among nearby residents and workers (substitution), and falls modestly overall for these users as local reductions are partially offsetby increased Uber usage elsewhere (option value). Estimating the model on these moments, we find that all three channels are quantitatively important, with substitution the largest individualchannel but last-mile complementarity and option value together more than offsetting it, so that Uber and transit are net complements. Policy implications include: (a) a ride-hailing tax induces users to switch to private vehicles rather than transit; (b) a subsidy for last-mile ride-hailing trips is more effective at increasing transit use when it targets the per-kilometercost rather than the fixed cost; and (c) increases in transit use are best induced by a congestion charge that subsidizes transit fares rather than by policies relying on ride-hailing taxes and/or last-mile subsidies.