Figure 19

Distributions \(\operatorname{Pr}(F_{ij})\) (black ‘◯’) and \(\operatorname{Pr}(F_{ij}^{(s)})\) (red ‘□’) for \(M=10^{3}\) Monte Carlo realizations. The distributions differ widely, with \(\operatorname{Pr}(F_{ij}^{(s)})\) considerably steeper than \(\operatorname{Pr}(F_{ij})\), illustrating the way that the condition to fix \(\tau _{i}\) for each node has the effect of spreading flows evenly over the network, thereby eliminating large flow between node pairs