Agent-Based Models of the Financial Markets

 Introduction The study of traditional financial theory relies on the assumption that the financial markets are made up of fully rational agents, whom’s decision making is based on utility function with beliefs. However, this assumption has been questioned and challenged in the past decades. This is because financial markets arise through the interactions of people. Human behaviour is therefore key to understanding how financial markets change over time and react to various events. But our best attempts to model financial markets have not been particularly successful, in part because human behaviour is not adequately described by simple rules. More importantly, the prospect theory suggests that the investors’ psychological drives will affect their decision-making process and they may deviate from rationality. One response to this is the promotion of agent-based models (ABMs). These are computational models in which individual agents and their interactions are simulated. With ABM, financial markets can be modelled from bottom up, allowing the incorporation of interaction and evolution of agents. One of the appeals of agent-based modelling is that one’s intuition for how things work can be coded up relatively easily, allowing various scenarios to be tested ‘in silica’. Whilst ABMs can be used to model more complex systems, they too rely on simplified models of human behaviour, but it is much harder to understand the effects of particular model mechanisms on the emerging dynamics. There is also rarely the micro-level data available to properly calibrate agent behavioural rules. 2 Part A — Literature review and programming Students who choose this project will all start from a common root given in this section. This will form the foundation of the project and lead to the research topics in the next section. The project description includes a list of reference at the back. They should help get you started. These should not be the only papers you read. You should also look for other relevant papers yourself, e.g. using Google Scholar. A good way to do this is to look at the papers cited by from the list below. 2.1 Literature review Provide a critical literature review which discusses the development of agent-based modelling in the literature, in particular the financial markets. What motivates researchers to adopt ABM over existing models? What are the strengths and weaknesses of agent-based modelling? Summarise the agent-based models of financial markets, assess any drawbacks and identify any ‘gaps’ or ‘parts’ that can be further improved / modified. 2.2 Implementation Implement a basic model from an existing paper using computer programme. A two-type model is the simplest model with respect to heterogeneity. Model the financial markets with two different types of agents, namely the fundamentalists and the chartists, and replicate the relevant results from the paper. 2.3 Modification Make modification to the model if necessary. 2 3 Part B — Further work You should then address an additional topic in your report related to the model you simulate. This will depend on the model that you attempt to implement in Part A and therefore this will be discussed and agreed on at a later stage. Below is a list of suggested research direction that you can consider. 3.1 Switching mechanism Traders in the financial markets do not necessarily stick to the same trading strategy, therefore traders should be able to switch between different trading strategies instead of simply assuming that the ratio of fundamentalists over chartists remains constant over time. A simple implementation of this can be done by assuming that agents change their strategy by random switching. A more sophisticated implementation will involve inductive learning, which is related to computation models of autonomous agents. For example, Hessary and Hadzikadic[6] implemented the social interactions and adaptation of agents and allowed agents to switch to the other strategy based on certain conditions. This can be one of your starting point reference. 3.2 Behaviour bias In this research direction, you can incorporate other human behaviours into your model. This is because agents’ behaviour plays a crucial role in making your model more complex yet realistic. For example, Pruna et al.[12] tested the implication of agents trying to avoid regret by holding onto their investment for a longer period despite the fact that they are suffering from a loss. This is observed in the financial markets because some traders may not want to accept the fact that they have made a bad decision in the trading strategy. It was shown that the incorporation of this behaviour bias into the ABM of financial markets has a direct impact on the returns’ distribution. 3.3 Other possible further topics Other possible further topics of study could investigate the dynamics of certain models in more mathematical detail and how these depend on parameter values, or you could look at how your model can be calibrated and validated. 3