From Taxi to Transit: Causal Effects of Washington DC's Night Buses on For-Hire Travel Demand

This study evaluates whether Washington, DC's night bus service expansion (officially launched on December 17, 2023) fundamentally altered traditional for-hire travel demand—specifically taxi usage—within newly exposed transit corridors. Recognizing the complex interplay between public transit and for-hire vehicles, the empirical design combines high-frequency administrative taxi records from OpenDataDC with a spatial Difference-in-Differences (DiD) framework spanning a full two-year panel (January 2023 to December 2024). Treated exposure is constructed by intersecting 200-meter walking buffers around active night-bus stops with 2020 Census Block Groups, mapped to individual trip origins and destinations. Crucially, this study addresses severe baseline scale discrepancies between treated areas (~5,200 daily trips) and control areas (~400 daily trips). We demonstrate that a standard linear DiD model violates the parallel trends assumption in absolute terms, misattributing a massive localized drop of 891 daily trips solely to the transit intervention when it is largely driven by macro-level trends. To correct this, the primary specification relies on a log-linear DiD model, which identifies a broad 12.8% background decline in taxi usage across the entire city. Net of this citywide trend, the localized proportional impact of the night bus intervention is an estimated 5.0% relative decline in treated-area taxi trips (p = 0.080). These findings highlight a critical methodological insight for transport geography: while raw transit substitution may appear massive in absolute terms, appropriately scaling for baseline market size reveals a much more nuanced, context-dependent substitution effect.