Are oil and gas futures markets efficient? A multifractal analysis

ABSTRACT

Energy futures markets have shown high volatility, giving rise to challenges regarding their pricing and efficiency. This study investigates the weak-form efficiency hypothesis for two major energy futures markets (gas and oil) over both calm and crisis periods, using a multifractal approach and intraday data to deal with the flexible econometric framework and rich information. Our results are threefold. First, we show that multifractality in lower frequency might be more biased than in intraday data, which motivates the use of the fractal approach when testing the intraday efficiency. Second, we highlight that high frequency data are characterized by a true long memory with a higher degree of persistence during the post-crisis period for the oil market. However, for lower frequencies, the long memory becomes spurious for both markets. Finally, our forecasting results show that, for the oil market, the proposed multifractal approach outperforms conventional methodologies, especially during turmoil periods.

KEYWORDS:

• Weak form efficiency
• energy futures markets
• intraday data
• multifractal approach

JEL codes:

• C1
• G14
• Q4

Acknowledgments

We would like to thank the editors and three anonymous referees for their helpful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ See Fama (1965) for details on the other efficiency forms (semi-strong and strong forms).

² The DMA method removes trends by subtracting local means, whereas the DFA approach removes trends based on polynomial fitting. Gu and Zhou (2010) showed that the MF-DMA is more robust than the MF-DFA.

³ See Serletis and Rosenberg (2007) for details on the DMA approach.

⁴ The general test idea is based on the different behaviours of the fitted spectral density and the periodogram of the frequencies close to the origin .

⁵ The MF-DMA approach doubly differs from the MF-DFA. First, it is not based on the box-splitting procedure as is the MF-DFA. Second, the MF-DMA assumes a power-law scaling of covariances with an increasing moving average window size (n). Gu and Zhou (2010) showed that the backward MF-DMA algorithm outperforms MF-DFA based on numerical experiments.

⁶ See Kantelhardt et al. (2002) for more details on these procedures.

⁷ According to Jiang, Xie, and Zhou (2014), shuffling the original series 1,000 times might provide a more robust test.

⁸ For more applications and details on this approach, see Cristescu et al. (2012), Zhuang, Wei, and Zhang (2014), Li, Lu, and Zhou (2016), Ruan, Jiang, and Ma (2016), and Ftiti et al. (2019).

⁹ We would like to thank the three anonymous referees for their suggestions to add this section.

¹⁰ The MSM and ARMA(1,1) models fit the crude oil and gas realized volatility well. The in-sample results are not reported due to space considerations but are available upon request.