Market Efficiency in the Crude Oil Futures Market - an Empirical Study after the Shale Oil Revolution

Abstract

This thesis has studied efficiency in the crude oil futures market for WTI and the Brent Blend for a period including the “shale oil revolution”. The main objective was to provide new information by investigating a period in time not much explored in already published articles. Furthermore, the thesis sought to close a gap of earlier empirical studies performed, by combining the two crude oil types and including up to 6 months maturities for futures contracts, while at the same time having a precise definition of an efficient market. The raw data applied was daily closing prices for spot and 1-, 2-, 3-, 4-, 5-, and 6- months (M) futures contracts from January 1986 to March 2016 for WTI and from October 2003 to March 2016 for the Brent Blend. The data was converted to monthly observations, and the Johansen cointegration analysis was performed while imposing the restrictions of the unbiasedness hypothesis, α = 0 and β =1. Tests were performed pairwise, both for the same blends and across blends. The cointegration analyses were further performed both for the total period and separately for the sub period (January 2012- March 2016), as the produced amounts of shale oil started its sharp incline in 2012. Finally, tests of weak exogeneity were performed. For the total period, the market was concluded efficient for the cointegration pairs Brent spot/Brent 1M, WTI spot/WTI 1M, WTI spot/Brent 3M, WTI spot/Brent 4M, WTI spot/Brent 5M and WTI spot/Brent 6M. For the sub period, the market was concluded efficient for the pairs Brent spot/Brent 4M, WTI spot/WTI 2M, WTI spot/WTI 4M, WTI spot/WTI 5M, WTI spot/Brent 2M, WTI spot/Brent 3M and WTI spot/Brent 4M. In general, futures prices were found to have led spot prices for most of the cointegrated pairs. The hypothesis of the spot price leading the futures price was rejected for all cointegrated pairs. For the total period, the finding of efficient markets for the contract with the shortest maturity for WTI and Brent was similar to previous empirical findings. The amount of cointegrated pairs however implied a higher degree of cointegration than previously. There was found little cointegration and no market efficiency for the Brent spot/WTI futures- pairs, implying possible speculation opportunities. The assumption of “normal backwardation” did not seem applicable for the periods investigated, as most relationships showed signs of contango. Suggestions were made that a non-linear method or a method including a structural break might better model the crude oil futures market. Risk-varying premiums, convenience yields and investors with different investment horizons were among the factors discussed as possible explanations to why the unbiasedness hypothesis was rejected in most cases. Finally, the results implied that using crude oil futures as a risk management tool might not be efficient for all maturities.