Unpacking Herding in Crypto Markets: An Agent-Based Study of True and Spurious Herding Dynamics

Herding behavior is a widespread phenomenon in human society, particularly evident in financial markets, often linked to unexplained price movements such as bubbles and crashes. While herding in financial markets is extensively documented, distinguishing between true herding (investors imitating others) and spurious herding (investors independently responding to similar information) remains challenging. Traditional statistical methods, such as cross-sectional standard deviation (CSSD) and cross-sectional absolute deviation (CSAD), commonly used to detect herding, fall short in effectively differentiating these two types during periods of low market return. Moreover, another main limitation is that they are market-wide metrics are not designed to detect herding in a single asset market. This study addresses these limitations by not only detecting the presence of herding behavior but also uncovering its specific dynamics under varying market conditions - bull or bear - and identifying whether it is driven by fundamental economic indicators or non-fundamental factors like social media, all within the context of a single financial market. To achieve this, I develop an agent-based model calibrated with real-world market data and validated using Bitcoin datasets. My findings reveal that during bear markets, herding tends to be spurious, heavily influenced by non-fundamental factors, while during bull markets, it is true herding, primarily driven by fundamental information. These results support existing literature on the asymmetry of herding behavior across market regimes and contribute to understanding herding dynamics within cryptocurrency single-market level.