Abstract
In financial markets, every investor seeks and receives information to decide how they should act (e.g., buy or sell a certain asset). In certain social circles, investors also learn about the decisions of other investors and they might sometimes ignore their own information and take the same decisions as other investors. This phenomenon is known as ”herding effect”. Many believe that herding can be one of the main causes of crashes and bubbles in financial markets.In this thesis, we adopt empirical methods to explore why investors try to imitate others, the impact of herding on financial markets and whether the trading mechanism used in the market affects herding.
Towards this goal, we connect opinion formation dynamics with herding in financial markets. We model social connections between the traders in different market environment as a graph and adopt a well-established opinion diffusion dynamics. Opinions are translated to trading positions and market prices evolve accordingly. We relate the shape of the graph social network to the equilibria of a game defined as follows. The players are traders that can strategically decide whether to follow the wisdom of the crowd or act upon their own beliefs.
Their payoffs are defined as the wealth they accumulate from trading. We adopt Empirical Game-Theoretic Analysis (EGTA) to compute the equilibria of our games.
We first explore the impact of social connections between market participants on herding and market stability in a hypothetical market environment, where orders are always executed at the desired price. We show that the larger the traders’ neighbourhood in the social network, the more the traders are willing to imitate others and the less volatile the stock price is. However, when every trader in the market has perfect knowledge of the opinions of all the other traders, the market will still exhibit crashes and bubbles. The definitions of crashes and bubbles in our research are based on changes in stock prices and are inspired by the financial concept of Maximum Drawdown.
The mechanics of trading in an order-driven market environment can influence the behaviour of traders and the idealised setting in our simulated market environment is too simplistic to model real markets. We then investigate opinion formation and herding in order-driven financial markets, which are widely used for many asset classes. We concentrate on Continuous Double Auctions, the principle trading mechanism in this class, and consider two forms of order queuing mechanisms: price-time priority, the de-facto standard, and spread-price/time priority, an alternative recently defined in literature to reduce toxic order flows due to latency arms race. We find that our conclusions are robust and hold in both these realistic market environments; the stronger the social connections between the agents, the more pronounced the herding. Furthermore, our empirical research shows that as the market gives more weight to spread, it becomes more stable thus confirming the findings of related work in our setup.
We conclude our work by enlarging the set of strategies that agents use. We use a meta-game to simplify the actual large game and explore herding of different types of investors in the market with different social connections. The results show that the herding is more pronounced among long-term investors than short-term investors. We see our work as the introduction of a framework that can be used to study more questions about herding in financial markets and other complex systems.
| Date of Award | 1 Jan 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Carmine Ventre (Supervisor) |