In this paper we present the macroscopic model for pedestrian flows proposed by Colombo and Rosini [10] and show its main properties. In particular, this model is able to properly describe the movements of crowds, even after panic has arisen. Furthermore, it is able to reproduce the so called Braess' paradox for pedestrians. From the mathematical point of view, it provides one of the few examples of non classical shocks motivated by real problems, for which a global existence result is available. Finally, its assumptions were experimentally confirmed by an empirical study of a crowd crush on the Jamarat Bridge in Mina, Saudi Arabia, near Mecca, see [17].