Computational models of financial markets

The three chapters of this thesis share the common theme of computational approaches to modeling financial markets. Chapter 1, “Market Ecologies: The Interaction and Survival of Technical Trading Strategies”, finds its place in the boundedly-rational heterogeneous agent literature. Market prices result from the interaction of fundamental and technical trading strategies. We show that the way in which traders process information is critical in determining the long-run profitability of individual strategies. More realistic auction settings— in which price information is incorporated into trading methods in “real time”—demand computationally demanding techniques. The main conclusion of the chapter is that contrarian technical traders inadvertently mimic the role of arbitrageurs in more realistic auction settings. Chapter 2, “Capital Allocation in a Delegated Trading Model”, develops a model of capital allocation that removes the need for full mean-variance optimization of the firm-level portfolio. The strategies explored within the artificial setting of Chapter 1 are used to test the model against empirical foreign exchange data. We observe that the proposed capital allocation scheme yields economically and statistically significant returns, even when traders choose rules without the benefit of hindsight. Chapter 3, “Portfolio Choice: The Costs and Benefits of Asymmetric Information”, continues the theme of artificial markets, with the auction process departing from the fictitious auctioneer of Chapter 1, toward a market making model in which risk-neutral dealers quote bid-ask spreads to compensate them for the losses incurred by trading with informed agents. We obtain the intriguing result that, in multiple markets, there is an “optimal” level of inside information. In individual markets, portfolio managers incur higher transaction costs as asymmetric information increases, but benefit from an externality at the portfolio level, as inside information aids price discovery.