A market behavior known as herding is not as important a trend as economists previously assumed – which comes as a “big surprise” – say SFI Professors Doyne Farmer and Fabrizio Lillo and their colleagues in a recent paper.
The agent-based models the researchers are developing can give economists another tool to explore policy scenarios and avoid a repeat of the financial crisis that began in 2008, which traditional economic models failed to predict. Their new models evaluate the relative contributions of herding and other behaviors. Data they analyzed from the London Stock Exchange did not shore up the assumption that herding plays a major role.
“There is some herding, especially for very short time lengths,” says Fabrizio in describing the intraday persistence of order ow, a trend in which buy orders follow buy orders or sell orders follow sell orders. “But the dominant component is order splitting,” he concludes, “at least at the broker level.”
Herding occurs when investors imitate others or act together in response to a signal such as a press release or price change. Order splitting by brokers minimizes the impact of big orders on price by splitting the orders into pieces and gradually executing them.
The researchers’ new technique for distinguishing between the behaviors and identifying which is dominant in market trends may be important for examining other data sets as well, and in developing more accurate agent-based models of the economy.
A preprint of their paper is published on arXiv.