Herd Behaviour in Humans

As we all know from experience, and as emphasized above, humans do not act independently but are often directly influenced in their decisions by the behaviour of others. This collective or herd behaviour should naturally be taken into account in any realistic model of social systems as it is likely to be a crucial ingredient. Consider for example the distribution of 'price returns' (logarithm of the difference between the purchase and sale price of a share). An analysis of the empirical data reveals a power-law tail to the distribution in contrast to the Gaussian distribution predicted by models of independent agents. In Ref.[17] the authors present a self-organised model to describe the herd behaviour in the financial market. The model is an evolving network in which nodes represent traders that can be linked to form clusters of like-minded agents. The action of any individual in a cluster is imitated by others in the cluster. The links in the cluster (and hence also the size and number of clusters) are dynamical quantities in contrast to the fixed topology networks we studied earlier in the course. The authors find that the model can show power-law results for values of a herding parameter below a critical value and also discuss the increase in probability fo crashes for large values of the herding parameter. (See also [18] for a discussion of stock market crashes).