The Value of Reliability: An Equilibrium Approach

This paper extends the theory on the valuation of travel time reliability, currently limited to the case with a single traveler, by considering interactions between individuals. To be more specific, we examine how the Nash equilibrium in departure times is modified when stochastic travel times are introduced in the standard bottleneck model with N morning commuters. For individuals with (α,β,γ) preferences, we find that the primary impact of travel time variability, modeled here under the form of a uniform random delay, is peak spreading. Departures are spread more evenly around the peak than when travel times are perfectly reliable, which reduces congestion. This mitigating phenomenon leads to a lower value of reliability (VoR) than the one given by non-equilibrium approaches. When the maximum random delay is small enough compared to the total length of the rush hour, the VoR is even null. These results entail two caveats as far as cost-benefit analysis is concerned: first, VoR based on non-equilibrium approaches overestimate the cost of unreliability as they fail to capture equilibrium mechanisms. Conversely, VoR based on (α,β,γ) preferences likely underestimate this same cost.