Rational herds, speed of learning and contagion in financial markets.

摘要

This dissertation consists of three chapters. The first chapter studies whether imitative behavior and contagion, two well-documented regularities of financial markets, can occur in a two-asset economy where rational agents trade sequentially. It shows that, when traders do not trade only for pure speculative reasons, but arrive in the market because they have some gains from trade, informational cascades and herding arise. After a long enough history of trades, traders disregard their private information and act according to their gain from trade. Thus, all traders of the same type choose the same action. Our results contrast with those of Avery and Zemski (1998), who have argued that, in financial markets, the presence of a price mechanism rules out the possibility of informational cascades. Moreover, sequential trading helps to explain financial contagion, since herds can spill over from one asset to the other. When this happens, the asset prices can remain forever far away from the fundamental value.; The second chapter shows that herding and cascades can also occur when the proportion of informed traders in the market is unknown. This happens because traders and the market maker interpret the past history of trades differently and traders choose the same action, independently of their private information.; Finally, the third chapter studies the price path of a financial asset whose fundamental value changes over time. It focuses on the speed of learning, as measured by the expected distance between the price and the fundamental value. It shows that the speed of learning decreases in the probability that a shock hits the economy and in the unconditional volatility of the fundamental. On the other hand, it increases in the percentage of informed traders and in the accuracy of private information. Moreover, the chapter shows that an increase in the unconditional variance of the price is ambiguously related to the speed of learning. If it stems from an increase in the variance of the fundamental or in its persistence, the speed of learning decreases. In contrast, if it stems from an increase in the informativeness of each trade, learning is faster.