Paying off construction is not a good argument for tolls (althouogh if demand is elastic enough the right toll will pay for the right amount of construction); regulating traffic flow is. There's no good reason to have a toll on an empty stretch of new road that hasn't been paid off; there's plenty of reason to have one on a congested road that has been paid off. Since roads are generally uncongested right after they're built and become more congested as they become older (it takes time for people to find out about new roads, learn their advantages, and then adjust where they live and work), the idea that tolls should be used to pay off a road's costruction cost leads to the worst possible allocation of resources: high tolls at first that discourage trips that should be made, and no or low tolls later that encourage waiting that should not be done. It's like taking a vile-tasting medicine when you're young and healthy, and stopping it when you get sick and it could help you.
This is from City Economics by Brendan O'Flaherty, a professor of urban economics at Columbia. The entire textbook is written in this conversational style. It's fun to read, and it's the only textbook I've ever read cover to cover. (I admit that I enjoy books that confirm my preconceived opinions.)
