Market Agents and Flash Crashes

This thesis studies the use of agent-based modelling to investigate factors that can affect the stability of an electronic trading venue. Automated trading strategies have contributed to tighter bid/ask spread and granular liquidity in the markets. However, over the past few years, there have been several incidents, notably the flash crash of 6 May 2010, that raised concerns about the effect of such automated trading systems on the stability of financial markets. There have been different views on why the flash crash happened and what can be done to prevent similar issues in the future. Several changes have been proposed by regulators and market operators to improve the stability of financial markets. It is essential that these suggestions are well understood and scientifically analysed to clarify their ability to improve market stability and understand any negative side-effects they may bring. The study proposes an agent-based modelling framework for financial market and market participants to allow analysing the effect of such proposed changes. The objective is to use agent-based models to simulate and analyse the behaviour of market participants in the event of a liquidity shock. Firstly, the study develops an agent-based model of a financial market and its participants and corresponding simulation platform. Using this simulation platform and market data from trading venues, the study examines the behaviour of artificial agents versus data from a real environment. Finally, this simulation platform is used to perform extensive experiments to understand the factors that have been claimed to contribute to a flash crash. It further analyses the effectiveness of solutions that are proposed to prevent it. • The study first investigates how the diversity of the trader population can change the market’s reaction to order-flow imbalance. In particular, it focuses on two types of market participants: high-frequency traders and fundamental traders. There have been concerns that a high ratio of high-frequency or fundamental traders can make markets more unstable. The experiments performed in the study confirm that having too many high-frequency traders contributes to higher market volatility. It is observed that not having high-frequency market makers also generates problems as long-term investors may not be able to provide short-term liquidity when needed causing price fluctuations. • The study examines electronic trading controls that are being implemented by trading venues to maintain an orderly market. Trading rules are designed to provide a framework in which a market participant is expected to behave. Circuit breakers are extra controls that can stop trading when some safety conditions are broken. Circuit breakers are commonly triggered by a large price move in a short period and they stop trading in the venue so that a human can review the status and decide whether to resume trading. Some of the trading venues have automated circuit breakers that enforce a short-term suspension of trading and automatically resume trading using an auction. There also have been suggestions for more complex measures to trigger circuit breakers than large price shift. The study examines how effective these measures are on increasing market stability. • The study finally investigates the interaction between markets. It investigates how a liquidity crisis from one security in one market can expand to other securities and other markets. The flash crash started in one contract in the future market but soon expanded to ETF market and equity markets. This experiment analyses the factors that can affect that interaction by either reducing the effects of such event on other markets or amplifying the crisis and make it worse. The first contribution of this thesis is the introduction of an agent-based framework to study the behaviour of a financial market with different classes of trading agents. An extensive study of this platform is performed and it is used to simulate the flash crash. The second main contribution of the thesis is the development of a simulation platform that is capable of simulating an electronic trading market with high-frequency traders. The study employs the same techniques and tools that are used in building high-frequency trading platforms to build a platform targeted for analysing regulation and market rule changes. The third contribution of this thesis is to analyse some of the factors that are claimed to contribute to flash crash or methods that are designed to prevent such events. The study analyses the effect of fundamental and high-frequency trader population on market stability. It examines studied the effect of circuit-breakers, minimum quote life and order-to-trade ratio limits. The final contribution of this thesis is to extend the base model further to more than one financial market and study how liquidity issues expand from one market to another