The Heterogeneous Effects of P2P Ride-Hailing on Traffic: Evidence from Uber’s Entry in California

Abstract

Despite their promise, popularity, and rapid growth, the transit implications of ride-hailing platforms (e.g., Uber, Lyft) are not altogether clear. On the one hand, ride-hailing services can provide pooling (i.e., traffic reductions) advantages by efficiently matching customer demand (i.e., trips) with resources (i.e., cars) or by facilitating car-sharing. On the other hand, ride-hailing may also induce extra travel because of increased convenience and travel mode substitution, which may create crowding (i.e., traffic increases). We seek to reconcile these divergent perspectives here, exploring the heterogeneous determinants of ride-hailing’s effects. Taking advantage of Uber’s staggered entry into various geographic markets in California, we execute a regression-based difference-in-differences analysis to estimate the impact of ride-hailing services on traffic volumes. Using monthly micro data from more than 9,000 vehicle detector station units deployed across California, we show that Uber’s effect (either pooling or crowding) on traffic depends on various contextual factors. We find some evidence of pooling effects on weekdays; however, Uber’s entry leads to significant crowding effects on weekends. Furthermore, the crowding effect is amplified on interior roads and in areas characterized by high population density. Although ride-hailing seems to have a substitution effect on public transportation, we find ride-hailing services may have a complementary effect for carpooling users. Finally, we show that premium ride-hailing services (e.g., Uber Black) almost exclusively lead to a crowding effect. We conduct a battery of robustness tests (e.g., propensity score matching, alternative treatment approaches, placebo tests) to ensure the consistency of our findings.