Abstract

The penetration of BRT systems has been increasing fast, although there have been many reports of heavy queuing to board the buses. We propose a dynamic congestion approach that endogenously models queuing both on the road and at BRT stations, which are the center of our interest. We show analytically that, if capacity is perfectly divisible, implementing a BRT is always efficient (it decreases total social cost), while we show numerically that if capacity is not perfectly divisible, a BRT is in most cases efficient. Moreover, BRT can induce a Pareto Improvement where both time costs and public transport operating costs decrease. Compared to the optimum when buses run in mixed traffic, the optimal BRT system has: (i) Shorter period of bus operation and car-peak period, (ii) larger frequency and, very importantly, (iii) more boarding delays, i.e. longer queues at bus stops. Point (ii) implies that, while for some level of demands it may be optimal not to provide any public transport service under mixed traffic, with a BRT it may well be worthwhile. (C) 2019 Elsevier Ltd. All rights reserved.