Essays on Market Microstructure

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

This thesis builds on the background of the critical changes in financial markets
for the past decade: the proliferation of fragmentation and high-frequency trading,
and machines have been increasingly involved in the process of trading decisions and
execution.
Chapter 1 briefly introduces the motivation, research questions, and findings of the
subsequent three chapters.
In Chapter 2, we study the impact of market fragmentation on market liquidity
through an exogenous event, where the lit trading of Swiss equity trading shifted from
fragmented to consolidated markets. We find that market consolidation deteriorates
consolidated liquidity, resulting in higher spreads and lower depth. However, market
consolidation improves the inside depth of the primary exchange. The effects are
predominantly focused on large stocks, while the smaller stocks are almost unaffected.
In Chapter 3, we investigate the impact of differential access to consolidated price
information in the landscape where the mandated consolidated tape does not exist across
Europe. We find that access to a consolidated market view has the potential to reduce
execution costs – approximately 15% lower spreads and over 40% more depth, and price
improvement about 13% of the time. Moreover, we demonstrate that incorporating the
consolidated data increases the predictability of high-frequency price movement in the
primary exchange by 2.66% and significantly enhances the performance lower bound.
However, the value of consolidated data can be susceptible to latency. The results in this
chapter provide valuable insights into the consolidated tape discussion in pan-European
equity markets.
In Chapter 4, we study intraday stock return prediction using market microstructure
features and machine learning techniques. We find that while the linear models already
perform reasonably well, more sophisticated machine learning techniques further enhance the prediction performance. Microstructure features provide strong predictability,
and we identify past returns, imbalance measures, and order arrival rates as the main
predictors, and uncover significant interaction effects between predictors. Moreover, we
reveal evidence that microstructure features could capture systemic effects, however,
tree-based models are needed to effectively capture these systemic patterns. The models’
prediction performance is economically significant and provides promising potential to
further develop profitable trading strategies.
Date of Award7 May 2024
Original languageEnglish
Awarding Institution
  • University of Bristol
SupervisorSylvain J Friederich (Supervisor) & Liyi Zheng (Supervisor)

Cite this

'