Workers in developing countries waste significant time commuting, and gaps in public transit constrain access to productive jobs. In many cities, privately-operated minibuses provide 50–100% of urban transit, at the cost of long wait times and poor personal safety for riders. Should developing-country cities follow the typical recommendation of bus rapid transit or subway investments or rather optimize this existing, home-grown network? I build a micro-founded model of privatized shared transit subject to externalities in matching between buses and passengers. I then estimate the model with newly collected data on minibus and passenger queues in Cape Town and stated user preferences for exogenously-varied commute attributes. I find that Cape Town’s existing bus rapid transit decreased welfare, net of costs, but socially-optimal minibus fares and commuter taxes correct matching externalities, particularly benefit low-skill workers, and reduce spatial misallocation. Government actions to improve security bring even more substantial welfare gains.