Informational Cascades and Herds: Aspects of Dynamics and Time

Abstract

The currency crisis models of the last chapters necessarily assumed that spec-ulators have to decide at the same point in time whether or not to attack the fixed exchange rate parity. The models therefore analyzed currency crises as static coordination games. In reality, however, market participants are free to decide when, if ever, to short-sell the domestic currency, thereby attack-ing the fixed peg. Such a setting is inherently dynamic, encompassing several time periods. Even though a very important aspect, questions of dynamic coordination games in currency crisis models have only recently attracted scholarly attention. However, there exists a vast theoretical literature on dy-namic aspects of financial markets in general.¹ Typically, these models allow for backward-looking behavior, analyzing the decisions of agents who can ob-serve their predecessors’ actions. At the center of attention in these models is the question whether market participants act according to their own pri-vate information, or if they are willing to completely neglect their individual information and base their decisions solely on the observed behavior of their predecessors. The emphasis therefore is on aspects of herding behavior and informational cascades. Market observers on financial markets frequently note excessively optimistic or pessimistic behavior of market participants. Often it is suspected that decision makers simply imitate the behavior of others, ob-viously making no effort to gather and process information about underlying fundamentals. In December 1996, Alan Greenspan referred to such herding behavior as “irrational exuberance”, thereby criticizing markets for not effi-ciently using available information. Theoretical work on this kind of behavior, however, emphasized that herding not necessarily has to be irrational. Rather, it has been found that, once stuck in an informational cascade, agents ratio-nally decide to neglect their information. The rationality question notwith-standing, the market outcome will most certainly be inefficient, since herding behavior is characterized by the fact that market participants can no longer learn from their predecessors’ choices. Hence, the amount of information aggregated by the market process remains at a constant level, so that the market outcome is very likely to be inefficient at some point.