Not All Transfers Are Equal: Frequency, Transfer Quality, and Ridership Response in Denver's Transit Network

When a transit agency consolidates or cuts routes, some riders lose a one-seat ride and must transfer. Counting added transfers is an incomplete measure because a short connection between frequent services differs from a long walk followed by an infrequent bus. We define transfer burden as the demand disadvantage associated with changing vehicles, including scheduled waiting, walking, and difficulty navigating the connection. We estimate this burden in Denver using TransitApp journeys and an origin–destination (OD) flow model that accounts for relationships among nearby origins and destinations. Go journeys are trips users tracked while traveling; non-Go journeys are proposed itineraries whose completion is not observed. Both are joined to General Transit Feed Specification timetables. The model separates riding time, scheduled wait, walking, transfer count, and service frequency. Frequency is measured by headway, the scheduled minutes between vehicles. Modeled flow on one-transfer itineraries is roughly 70 percent below flow on otherwise comparable zero-transfer itineraries in both samples. We also examine RTD's May 28, 2023 consolidation of Routes 34 and 38. Completed Go demand in the corridor rises 17.9 percent. After adjustment for regional app use and changes on similar routes, the estimated increase remains positive but imprecise. The findings support evaluating the conditions of a connection when screening service changes, while the corridor's aggregate response still requires an episode-specific network forecast. Working conference draft prepared for the North American Regional Science Council (NARSC) 2026 meetings. This preprint has not undergone journal peer review. Comments welcome.