Buses operating in mixed traffic experience problems with delay and travel time variability in peak periods. Numerous transit priority measures are found in the literature but most rely on separating buses from other vehicles. Complete separation such as dedicated bus lanes might underutilize the system capacity when the frequency of buses is not high. We propose a "dynamic bus lane policy", where control adjusts the proportion of cars to buses in one lane of a multi-lane arterial instead of enforcing complete separation of vehicle types. This strategy is analyzed with aggregated macroscopic dynamics for different models of lane choice. It is shown to benefit both cars and buses by maintaining higher outflow during the peak period and helping the system to recover more quickly at the end of the peak. Greater reductions in travel time are possible with increased information available to vehicles. Analytical derivations and non-linear optimization are then used to obtain the system optimal allocation of cars, as well as the toll needed to achieve it. The system optimal tolling would be difficult to implement in practice. Another optimization which includes tolls paid in its objective is nearly as good at reducing vehicle hours traveled and produces a much simpler pricing structure amenable to feedback control. A microsimulation with feedback control confirms the results of the macroscopic model and shows additional benefits attributable to driver behavior. Future work will look into varying control parameters to achieve a global minimum travel time and will incorporate user heterogeneity to test pricing strategies.