Investigation of the Burst Hypothesis and Intraday Volatility Behaviour

Specialeforsvar ved Mathias Wilken

Titel: Investigation of the Burst Hypothesis and Intraday Volatility Behaviour

Abstract: This thesis expands the theory and mathematical foundation behind the existence of short-lived explosive trends in the financial markets, split into the explosion of the volatility and explosion of drift as first described Christensen et al. (2016). We add finite sample corrections and extra parameter tuning to the estimators, allowing us to improve the detection of periods with extremely high volatility and possibly drift. With fast computational algorithms, we create thorough simulation studies showing that these periods cannot be described by the standard Heston model with a jump component. However, we also provide evidence that conclusions drawn from the Heston model does not necessarily extend to financial markets, particularly S&P 500. To remedy this, a thorough study of volatility is conducted. The empirical behaviour of volatility is investigated and volatility is modelled utilizing Brownian semi-stationary processes as introduced in Barndorff-Nielsen & Schmiegel (2009) in an attempt to capture both roughness and persistence along with intraday seasonality. Ultimately, the simulation study is extended to a model featuring a Brownian semi-stationary process, which strengthens the hypothesis of the existence of the short-lived explosive periods as even such models fall short of explaining these extreme events, denoted as bursts. This allows us to show the presence of bursts on S&P 500 but question whether drift bursts are present or the bursts are in fact all explained by an explosion in the volatility. We also apply the theory on Bitcoin and justify the existence of both volatility bursts and drift bursts in this market, though high-frequency frictions on the exchanges make these conclusions uncertain.

Vejleder: Rolf Poulsen
Censor: Kim Christensen