Exploratory Graphics for Volatility in Financial Time Series
Tony Lawrance (University of Warwick)
Friday 17th November, 2017 15:00-16:00 Seminar room 311B
Although volatility is in the main stream of financial time series, I claim that there is little available data analysis methodology for exploring the evidence and form of volatility in time series data. In the usual paradigm of statistical modelling, this is something one would wish to do before working with particular volatility models. The talk presents a framework for non-parametric volatility graphics which allows exploration of the time progression of volatility and a function showing the dependence of volatility on past behaviour, mainly the latter. Plotting techniques have been devised using a base non-parametric non-linear volatility model, and include bootstrap assessments of their reliability. They are demonstrated on daily FTSE100 time series data, both covering and subsequent to the 2008 recession, with their contrasting asymmetric and symmetric volatility behaviour. Simulated volatile and non-volatile series are used to statistically validate the methodology, via use of both conditionally heteroscedastic and stochastic volatility models. Except for a comment or two, this is very much an applied statistics methodology talk on volatility diagnostics.