This paper is concerned with the assessment of generalized user cost reductions in the cost-benefit analysis (CBA) of transport policies that aim at reducing unreliability. In particular, we investigate the implications of railway passengers' anticipating departure behavior when train services were unreliable. A simple model is established to describe and predict such anticipating behavior. Our numerical example shows how travelers' anticipating departures and scheduling costs depend on the level of unreliability. The possible bias incurred by ignoring the reliability and schedule delay costs reductions in the traditional CBA can be quite substantial. Given our assumptions and parameterization, the underestimation of ignoring these costs could range from 33 to 75% of total generalized user cost reductions. © 2011 The Author(s).