Investigating Nonlinearities in WTI Crude Oil Markets using Threshold Cointegration

Investigating Nonlinearities in WTI Crude Oil Markets using Threshold Cointegration

We tried to apply one of nonlinear models, a bivariate 2-regime Threshold Vector Error Correction Models to WTI crude oil spot and futures price series from September 1, 1998 to August 31, 2005 for the analysis of the interaction between these two markets. One difference from the previous studies was to divide the whole sample into two sub-samples which showed disparities in mean return level and volatility. We adopted TVECMs instead of VECM in both sub-samples although there was rather doubtful Hansen and Seo(2002) threshold cointegration test result on sub-sample 1. In relatively less volatile period with negative mean return, sub-sample 1, the upper regime(contango, positive basis, ft-1 ＞ st-1) had faster adjustment speed to the long-run equilibrium than the lower regime(backwardation, negative basis, ft-1 ＜ st-1) in spot markets. There were active short-run feedbacks between WTI crude oil spot and futures markets, but the shocks from futures markets tended to affect spot and futures markets adversely while the shocks from spot markets favorably. In relatively more volatile period with positive mean return, sub-sample 2, the lower regime(backwardation, negative basis, ft-1 ＜ st-1) had faster adjustment speed to the long-run equilibrium than the upper regime(contango, positive basis, ft-1 ＞ st-1) in spot markets. There was price information leadership in futures markets because usual short-run dynamics happened in spot markets. The shocks from futures markets tended to affect spot and futures markets adversely but the shocks from spot markets favorably.