The prevalence of Braess' paradox

In a noncongested transportation network where each user chooses his quickest route, the creation of an additional route between some origin-destination pair clearly cannot result in an increase in travel time to users traveling between that o-d pair. It seems reasonable to assume the same can be said of congested networks. In 1968, D. Braess presented a remarkable example demonstrating this is not the case: a new route can increase travel time for all. The present paper gives, under reasonable assumptions, necessary and sufficient conditions for "Braess' Paradox" to occur in a general transportation network. As a corollary, we obtain that Braess' Paradox is about as likely to occur as not occur.