Heterogeneous risk spillovers from crude oil to regional natural gas markets: the role of the shale gas revolution

ABSTRACT

This paper investigates the heterogeneous risk spillovers from crude oil to regional natural gas markets against the background of the North American shale gas revolution. The market risks are measured by the upside and downside VaR. We first confirm the existence of risk spillovers between oil and gas markets by using Granger causality in risk. Then, we employ the MVMQ-CAViaR model to reveal the tail-dependence patterns and propose the performance test to highlight the accuracy of the model for the construction of VaR. Furthermore, we construct quantile impulse-response functions and identify asymmetric features in the magnitude, duration, and direction of response by gas markets to extreme negative and positive oil price shocks. Our results show that the revolution actually affects the risk spillovers from oil to gas markets and exhibits the time-varying property. And the heterogeneous risk-transmission mechanisms depend on the regional characteristics and specific market scenarios. Finally, policy implications are discussed.

KEYWORDS:

• Risk spillovers
• crude oil
• natural gas
• shale gas revolution
• upside and downside VaR

Acknowledgments

This work was supported by the Fujian Social Science Planning Fund Program under grants No. FJ2016C092 and the National Natural Science Foundation of China under grants No 7157030562. We also thank Professor Simone Manganelli for providing Matlab codes of his method. All remaining errors are ours.

Notes

¹ Although the copula function can model the dependency structures and provide information on the probability that two variables experience extreme upward or downward motion together, the application of the copula function is usually performed based on the Inference Function for Margins (IFS). Particularly, the fitting of the marginal model requires attention since the standardized residual utilized in the copula estimation may provide below or above the fitting parameters if the model cannot interpret the time series changes..

² To save space, we do not introduce the definitions of specific risk spillover types and how to perform the test of Granger causality across quantiles. Please refer to Candelon and Tokpavi (2016) and Peng et al. (2018) for details..

³ Since the daily price for Asian natural gas is not available and the natural gas price mechanism in Asian markets is similar to the European market. The risk spillovers from oil to gas in Asia are not analyzed..

⁴ These empirical results seem to contradict the Granger causality results, but they are not. The concept of quantile Granger causality test is used in the monotonic nature of quantiles to divide the distribution of variables into multiple intervals. It is aimed to construct event vectors composed of event variables, and test whether there is Granger causality between variables in any distribution interval. Quantile Granger causality test is a joint significance test for multiple quantiles, which only examines the risk spillover effect between the crude oil and natural gas markets from a holistic perspective. However, the result of the MVMQ-CAViaR model is to estimate each quantile separately. Therefore, the estimations for the beta (2, 1) parameter are not contradicted with the Granger causality results. We appreciate the reviewer for pointing out this issue..

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [No. 7157030562];Fujian Social Science Planning Fund Program [No. FJ2016C092].