Title: Multiscaling in finance
Authors: Tiziana Di Matteo - Kings College London (United Kingdom)
Giuseppe Brandi - Kings College London (United Kingdom) [presenting]
Abstract: The multiscaling behaviour of financial time series is one of the acknowledged stylized facts in the literature. The source of the measured multifractality in financial markets has been long debated and it has been attributed to mainly two sources: the power law tails and the non linear autocorrelation of the analysed time-series. We will discuss the origin of multiscaling in financial time-series and investigate how to best quantify it. In particular, we will show results on the application of the Generalized Hurst exponent tool to different financial time series and we will show the powerfulness of such tool to detect changes in markets behaviours, to differentiate markets accordingly to their degree of development, to asses risk and to provide a new tool for forecasting. We will also show an empirical relationship, to our knowledge the first on in the literature, which links a univariate property, i.e. the degree of multiscaling behaviour of a time series, to a multivariate one, i.e. the average correlation of the stock log-returns with the other stocks traded in the same market and discuss its implications.