Abstract
This dissertation seeks to understand how urban commuters adjust their schedules and modes to congestion, as well policy implications of this adjustment. An equilibrium simulation model of commuting traffic on a hypothetical, urban highway corridor is developed. The demand side is a discrete choice model of mode and time of day, estimated with data from the San Francisco Bay Area. The supply side is a speed-flow function that predicts travel time from flows leaving the corridor. The research has three objectives: to simulate the effects of capacity expansion, optimal toll, and six other pricing policies; to test hypotheses relating to schedule shifts in response to congestion and policy changes; and to estimate biases in policy effects when schedule shifts are ignored. An iterative procedure is developed to compute optimal tolls that vary with time of day. Policies are examined from five perspectives: welfare (consumer surplus, toll revenue, and total benefits), peaking (traffic counts and share in the peak 15-minute period), congestion (average and peak 15-minute travel delays), schedule delay (average variable schedule delay), and mode mix (mode shares, average occupancy, and total traffic). Five results emerge. First, although an optimal toll can achieve substantial benefits, savings in travel delay are accompanied by increases in schedule delay. Second, a toll equal to the marginal social externalities of an additional trip at different times of day at a base case can achieve benefits equivalent to those of optimal toll, which is equal to the marginal social externalities of an additional trip at different times of day at a social optimum. Third, schedule delay has variable and constant components. The constant component is the equilibrium level at a base case when travel is free-flow. The variable component changes with congestion and policies. Fourth, urban commuters shift their schedules in response to congestion and policy changes. Heavy congestion forces people away from the peak; capacity expansion attracts people back to the peak; an optimal toll discourages people driving alone in the peak. Fifth, the benefits of capacity expansion and an optimal toll are substantially overestimated if trip scheduling is ignored.
