AbstractThis research complements and combines market microstructure theory and mechanism design to optimize the market structure of financial markets systematically. We develop an agent-based model featuring near-zero-intelligence traders operating in a call market with a wide range of trading rules governing the determination of prices, which orders are executed as well as a range of parameters regarding market intervention by market makers and the presence of informed traders. The market structure which generates the best market performance is determined by applying the search technique Population-based Incremental Learning, guided by a number of performance measures, including maximizing trading volume or price, minimizing bid-ask spread or return volatility. We investigate the credibility of our model by observing the trading behavior of near-zero-intelligence traders with stylized facts in real markets. Based on computer simulations, we conform that the model is capable to reproduce some of the most important stylized facts found in financial markets. Thereafter, we investigate the best found market structure using both single-objective optimization and multi-objective optimization techniques. Our results suggest that the best-found combination of trading rules used to enhance trading volume may not be applied to achieve other objectives, such as reducing bid-ask spread. The results of single-objective optimization experiments show that significantly large tick sizes appear in the best market structures in most cases, except for the case of maximizing trading volume. The tick size is always correlated with the selection of multi-price rules. Though there is no particular combination of priority rule and multiprice rule achieving the best market performance, the time priority rule and the closest multi-price rule are the most frequently obtained rules. The level of market transparency and the extend of market maker intervention show ambiguous results as their representative parameter values change in a wide range. We also nd that the results of multi-objective optimization experiments are much similar to those obtained in the single-objective optimization experiments, except for the market transparency represented by the fraction of informed trader, which shows a clear trend in the multi-objective optimization. Using the results obtained from this research we can derive recommendations for exchanges and regulators on establishing the optimal market structure; for securities issuers on choosing the best exchange for their listing; and for investors on choosing the most suitable exchange for trading.
|Date of Award||31 Jul 2012|
|Supervisor||Andreas Krause (Supervisor)|
- agent-based model
- market design
Evolutionary mechanism design using agent-based models
Li, X. (Author). 31 Jul 2012
Student thesis: Doctoral Thesis › PhD