Are crude oil markets globalized or regionalized? Evidence from WTI and Brent

Abstract

This study applies a novel quantile unit root with structural breaks approach to explore whether the international crude oil markets are better characterized as ‘globalized’ or ‘regionalized’. By using the spreads between WTI and Brent crude oil prices as a benchmark, we find that the spreads contain a unit root in the lower quantiles but display mean reversion behaviour in the upper quantiles. However, instead of focusing on some selected (local) quantiles, the quantile Kolmogorov–Smirnov tests over a range of quantiles suggest that the price differentials are universally mean-reverting and, thus, provide strong support to the ‘globalization’ view.

Keywords:

• oil price spread
• quantile unit root
• structural breaks
• globalized
• regionalized

JEL Classification:

• C22
• F20
• F41

Notes

¹ Ewing and Harter (2000) show that both Alaska North Slope and UK Brent oil prices follow a random walk and that these oil markets share a long-run common trend, suggesting that the two markets are ‘unified’ (globalized).

² Otherwise, if the price spreads contain a unit root and are nonstationary, the oil markets are better regarded as regionalized.

³ In the following, all oil price spreads are calculated by subtracting logarithmic prices of Brent from logarithmic prices of WTI.

⁴ The approach has also been applied to analyse the time series property of nominal interest rates (Koenker and Xiao, 2004), real exchange rates (Nikolaou, 2008), consumption (Gomes, 2011) and inflation rates (Tsong and Lee, 2011), to name a few.

⁵ The description below follows closely that of Tsong and Lee (2011).

⁶ Although not shown, we also find similar patterns (structural break) for the weekly and daily data.

⁷ Please see http://www.eia.gov/

⁸ Please see Fattouh (2010) for some explanations about this crossover phenomenon.

⁹ We also follow Tsong and Lee (2011) to calculate the half-lives in each quantile with the formula . The half-lives are calculated when the null hypothesis of is rejected; otherwise, half-lives are set to be infinite (not shown). The results are shown in column 6 of Table 2. It is found that the higher the quantile, the shorter the half-lives, and this finding indicates that when hit by a large positive shock, the price spreads return to its long-run equilibrium in a relatively quick way (say 2.261 months at the 90% quantile).

¹⁰ The bootstrapping critical values at 10%, 5% and 1% significance level are 2.942, 3.224 and 3.783, respectively.

¹¹ However, unlike the cases of monthly and weekly data, we find that the (individual) unit root null hypothesis can be rejected within each quantile considered.