dc.description.abstract | Abstract
This paper is written to highlight a rather novel statistical methodology called Quantilogram/ Cross-Quantilogram/ Partial Cross-Quantilogram. The purpose is to make the method more accessible for someone without a master’s degree in statistics or a phd in finance, e.g. a master student of economics or someone working within finance. The method is very useful when working with financial data, since it does not require data to be normally distributed. Financial data are frequently known to not have finite fourth moments due to heavy tails. The cross-quantilogram can reveal nonlinear and/or asymmetric relationships under varying market conditions. It can detect directional predictability and tail dependency between two time series, for arbitrary lags, and model how the dependency varies over time. The method is based on a quantile hit process, where the quantilogram is the correlogram of this quantile hit process.
The paper uses quantilograms to explore two cases from empirical finance. The first case examines the cross-quantile dependence structure between Brent crude oil, S&P 500 and OSEBX, to see which of the former two has the most spillover effect on the latter. The paper reveals that both Brent and S&P 500 have spillover effects on OSEBX, with S&P 500 being the strongest influencer. S&P 500 shows positive predictability for OSEBX for most quantiles at lag 1. A partial cross-quantilogram reveals that S&P 500 has a moderating effect on the spillover effects from Brent to OSEBX, whereas Brent has negligible effect on the relationship between S&P 500 and OSEBX. In general, the effects are not very persistent.
The second case study explores the directional predictability between 3 stocks from the aerospace industry; Lockheed Martin, Intuitive Machines and Astrotech. The industry is very diverse, and this is reflected in the results from the analysis. There is a surprising lack of cross-quantile correlation between the three. We find the strongest connectedness between Lockheed Martin and Intuitive Machines, which makes sense considering that their business models have the most in common. A lack of positive correlation in the medium-to-lower quantiles for Astrotech and Intuitive Machines at lag 1 makes them good hedges for each other. | |