It's called emergent behavior. The conclusions derived by the authors do more to show their own biases. They see a pattern and ascribe "hierarchy" and "leaders" to that pattern, then wave away those very concepts by admitting that it's all very fluid and organic.

My bet is they're ascribing "leader" to which ever sheep happens to be in the "front" of the motion. That/those sheep are probably emotionally "pushed" by those behind it, and seeks an area to move to. The sheep behind it probably focus on their closest peer and match their trajectory to that peer. Each sheep probably sticks with matching that peer as long as that peer stays within distance and visual range. When the peer is lost, another peer is elected to follow. This all gives the effect of a hierarchy with a leader. The sheep at the front of the flock will just as readily follow another if given the chance.

Source: I've run this flocking simulation before.