Multi-Market Trading: Simulations and Empirical Analysis

Careful construction of properly scoped agent-based models can offer meaningful insights about financial markets and lead to improved decision aides for financial regulators and other stakeholders. However, it is of critical importance to back these models with sound empirical analysis and careful consideration of market institutions. Analysis of financial data is increasingly difficult due to the sheer volume of data, but also due to complex market structures. Open out-cry markets have been replaced with electronic distributed-access systems, and trading strategies are highly automated, no longer subject to human response times. This dissertation presents a new agent-based model for analyzing fragmented markets, and provides an empirical analysis that demonstrates the sea change in financial markets.

In the first part of this dissertation, we survey the literature on constructing agent-based models of financial markets. The survey provides a comprehensive review of a number of topics regarding the design, implementation and analysis of agent-based financial markets. The survey also contributes to the discussion of the methodological distinction between several classes of models that are commonly referred to as agent-based models.

Next, we introduce a model designed to analyze strategic behavior in a multi-market setting. In our model, the prevalence of order routing is an experimental variable. The model allows us to study the causal relationship between certain types of strategic behavior and the statistical properties of a multi-market system. The unique approach to the specification of strategic behavior in a multi-market system is a methodological contribution to the literature on agent-based financial markets.

Finally, we provide an empirical analysis of the fragmented market for a sample of Nordic equities. The institutional setting is ideal in that the platforms are identical and centrally located, but securities trade in different currencies, which provides enough of an economic friction for us to observe the evolution of market inefficiencies as market structure changes throughout the sample period. Correlations, arbitrage opportunities and trade throughs are analyzed. We document the drastic transformation in markets with respect to the speed of exchanges and the response times of market participants. The analysis highlights the need for models that explicitly account for the mechanics and structure of markets.

In summary, this dissertation contributes to the literature on agent-based modeling, multi-market systems and and market efficiency with a new model of multi-market trading and an empirical analysis of a fragmented market for equities.