Flat versus spatially variable tolling: A case study in Fresno, California

• This paper examines the effects of spatially-variable pricing along road corridors.
• Applying spatially-variable tolls could effectively increase system performance and profits.
• Volumes over capacity (V/Cs) on different segments are crucial factors.
• Target V/Cs – rather than VMTs, MCs, or elasticities – determine the optimal tolls.
• The target V/C patterns are similar for peak and off-peak, even using different toll rates.

A number of studies have examined the feasibility of temporal variations in tolls. However, spatial variation in tolls has not received much attention, especially in practice. Spatial variation could effectively reduce congestion and increase profits. To fill this gap, we conduct an empirical application on 3 different road segments using the Fresno, California's transportation planning model. Our modeling results in a number of interesting insights. First, the derived optimal flat toll values are very close to the average variable tolls, but the effects of applying spatially variable tolls on improving total revenues (from 4% to 24%) and total improved travel time (from 18% to 1083%) measures are significant. Second, spatially variable tolls are more effective, but more costly, particularly for arterials, which can be attributed to the higher number of access points for arterials. Third, spatial variations in tolls are more effective for peak hours than for off-peak hours and for social optimization than for profit maximization. Fourth, to improve throughputs for both profit maximization and social optimization, the prevalent tolling pattern along a corridor induces lower final volumes per capacity (V/Cs) (after pricing) at the mainline flow sections and relatively higher final V/Cs at the entrance and exit (boundary) points. Finally, optimal toll patterns are not dependent on vehicle miles traveled (VMTs) or volumes but, rather, are related to targeted V/Cs. Therefore, flow-dependent charges along a corridor should be based on V/Cs rather than on volumes or VMTs.