High-Frequency Trading: Insights from Analytical Models and Simulated Agent-Based Models

Abstract

This dissertation studies the influence of High-Frequency Trading (HFT) on market characteristics and participants using analytical dynamic limit order book models and agent-based models. An attempt to converge the assumptions and results of the two types of models is made, which helps to overcome the current divergence between these models. It was shown that entry of fast traders damages slow traders' results, but if the volatility is high enough, their entry might improve some important market qualities, such as trading rate or the aggregate market welfare. The informational advantage of fast traders is detached from their speed advantage and it is investigated how severely HFT influences the market compared to usual informed trading. As the results show, a fast trader is not always worse than an informed slow trader: it depends on initial market conditions, market criteria, and the perspective chosen for the analysis. Moreover, the heterogeneity of the market was enriched further by introducing a random trader, an HFT based on analytical rules, and an informed slow trader based on analytical rules. The more realistic agent-based model with a sophisticated fast fundamentalist provides a further testbed for many possible configurations. For example, different degrees of market latencies, various market compositions and versions of order cancellation rights were investigated.