To mirror or not to mirror : modeling relationships in social trading

Abstract

Many scholars have applied ecological principles to study the financial market. As early as 1940s, John Maynard Keynes coined the term "animal spirits" to describe human decision making under uncertainty. In modern economic terms, "animal spirits" are often used to describe the psychological factors that drive investors decision making during volatile market. Many scholars used Darwin's evolutionary theory to explain evolution of investment strategies [5] However, few studied leader election, individual adaptation, and social dynamics in the financial market. This lack of research is mostly due to a lack of centralized research entities to implement large-scale experiments. Luckily, a new investment mechanism, social trading, where investors can interact with each other by mirroring and commenting on each other's trade ideas, provided a new avenue to study evolution of a new financial system. We are able to observe how leaders become leaders, how followers choose their leaders, and how different groups interact with each other. Our research takes place on one of the biggest platforms of this kind, eToro, a retail social trading platform in foreign exchange and other asset markets. Treating this economic system almost as a new ecological environment, we begin with understanding who are the different players and how they interact with each other. We categorize traders based on their investing styles and observe how their types change over time. Interestingly, these profiles resemble major players in the financial market: diversified institutional investors, speculators, and specialized strategy (macro and value) funds. Then we try to understand why some leaders have more followers than others and train a model to predict whether a leader will get a new followers/unfollowers on a particular day. Build upon existing literature, we found that not only can leader's style factors predict whether he gets new followers/unfollowers, popularity rank, average performance of his followers, and recent maximum gains also have predictive power. Our models are trained using SMOTE-balanced training sets and are able to achieve roughly 80%-90% accuracy. Lastly, we take a microscopic view of how followers follow. We claim that followers exhibit "foraging" pattern when choosing their leaders. Followers create people-portfolios, and foraging is essentially equivalent to diversification. By foraging, followers can prevent significant losses regardless of which type of investor they are. However, foraging would not lead to outstanding gains, or alpha per se. Traders who forage are analogous to index-following investors who track the market.