What Causes Herding:Information Cascade or Search Cost ?

We analyze in this study cause of herding in a stock market. Information cascades have often been considered as a primary choice. However, we present evidences inconsistent in this study. Employing intraday order book data, our analysis supports the inclusion of of an alternative theory based on search cost of investors, in addition to the information cascade argument. Specifically, previous works studied daily data or those with lower frequency based on a herding measure of Lakonishok, Shleifer, and Vishny (1992). We adopt instead the measure of Patterson and Sharma (2006) and argue that the search model of Vayanos and Wang (2007) characterizes herding phenomenon better at market open. Our analysis is also consistent with the information competition equilibrium of Back, Cao and Willard (2000) and dynamic friction model of Hu (2006).We find that stronger order flow herding is driven by lower transactions cost and shorter time to fill an order. Herding tends to occur in trading of high-cap, high turnover stocks, which contradicts prediction of the information cascade hypothesis. Search cost effect is stronger at market open, while information cascade effect is stronger at market close. Therefore our study suggests that herding should be related both to intrinsic search cost structure of investors as well as information related factors.