| Summary: | We model crude oil price differentials as a two-regime threshold autoregressive (TAR) process using Caner and Hansen’s (2001) method. While standard unit root tests, such as the Augmented Dickey–Fuller (ADF), are inconclusive in some instances on whether oil price differentials follow a stationary process, the null hypothesis of unit root can be strongly rejected based on the threshold unit root test, even for crude oils with very different qualities. Our results also indicate that the adjustment process is different depending on whether one considers the differentials between crude oils of similar quality or between oils of different quality and whether a crude oil is part of a complex that involves a highly liquid tradable futures contract. These findings suggest that the different oil markets are linked and thus, at the very general level, the oil market is ‘one great pool’. However, differences in the dynamics of adjustment suggest that within this one great oil pool, oil markets are not integrated in every time period and, although the presence of an active futures market has helped make some distant markets more unified, arbitrage across the different markets is not costless or risk-free.
|
|---|