Unwinding Financial Market Complexity

Student thesis: Doctoral ThesisDoctor of Philosophy

Abstract

Complex systems are characterised by dierent distinguishing aspects often associated with completely separate behaviours. In nancial markets, paramount example of complex systems, two of these aspects stand out in characterising the statistical properties of the many constituents: one is multifractality, a feature which describes the departure of nancial time series from purely random processes and is therefore a measure of complexity of the prices; the other is the cross-correlation structure between assets, which encloses information about the market organisation and can reveal dominant factors as well as hierarchical properties.

In this thesis I have studied the relationship between these two distinctive properties of financial markets. I have first unveiled new empirical properties of stock returns, casting new light on the latent mechanism governing price dynamics and interactions, and I have then proposed a model which reproduces the observed properties.

I have investigated multifractality dynamically on stock returns after having introduced the weighted generalised Hurst exponent, a study that has revealed remarkable increasing trends in the dynamical scaling exponents for rms bailed-out after the 2008 nancial crisis. I have then tested the signicance of dynamical fluctuations of multifractality against a well-established multifractal model, the Multifractal Random Walk (MRW). The hypothesis of constant multifractality in financial markets has been rejected in many cases revealing a much more complex behaviour of financial time series. I have then linked the multifractal behaviour in financial markets to the cross-correlation structure, showing that the two properties are indeed related. I have investigated the relationship between a proxy of multifractality and cross-correlation hierarchical properties on different markets which have confirmed the result. After having thoroughly reviewed the existing
literature on multivariate models, I have proposed a dynamical multivariate model able to reproduce the empirical facts reported in this thesis along with an array of other well-established stylised facts, thus unifying correlation and multifractality in a unique coherent framework.
Date of Award2014
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorTiziana Di Matteo (Supervisor)

Cite this

'