An Autoregressive Market Model of Trader Herding and Communication

Abstract

Chaudhary studies noise traders that communicate and trade with each other in a market. He begins by examining a classical t-statistic to identify a boom and uses it on the NASDAQ 100 dot-com “bubble.” He next generalizes the classical geometric Brownian motion stock model to incorporate communication among traders. He represents individual traders that observe each others’ past n daily returns using a nonlinear vector autoregressive NLVAR(n) process. He models traders endogenously creating a market price. With this model, Chaudhary measures autocorrelation and herding as functions of traders’ communication level (a) and number of past daily returns (n) that the traders rely on. He finds that autocorrelation and herding increase with communication level, a, and decrease with the length of returns being examined, n. Under this model, he observes traders forming spontaneous herds without specific leaders, which leads to price booms.