Causality in quantiles and dynamic relations in energy markets: (De)tails matter

Highlights: • Causal relations between energy markets are investigated. • This is performed within the framework of a dynamic quantile regression model. • Causal effects are found heterogeneous across quantiles and market states. • Evidence of significant causality is mainly found under extreme market conditions. • Prevailing market conditions should be considered by stakeholders. -- Abstract: In this article we investigate the dynamic relations between crude oil price returns and a set of energy price returns, namely diesel, gasoline, heating, and the natural gas. This is performed by means of Granger non-causality tests for US closing spot prices over the period from January 1997 to December 2017. In previous studies this has been done by testing for the added predictive value of including lagged returns of one energy price in predicting the conditional expectation of another. In this paper we instead focus on different ranges of the full conditional distribution, and thus market states, within the framework of a dynamic quantile regression model, and identify the quantile ranges from which causality arises. The results constitute a richer set of findings than what is possible by just considering a single moment of the conditional distribution, which can be useful for implementing better substitution investment strategies and effective policy interventions. We find several interesting uni-directional dynamic relations between the employed energy price returns, especially in the tail quantiles, but also bi-directional causal relations between energy price returns for which the classical Granger non-causality test suggests otherwise. Our results are robust to alternative measures of the price of oil and different data frequencies.