Does OPEC influence crude oil prices? Testing for co-movements and causality between regional crude oil prices

Abstract

Using high-frequency data the co-movements among crude oil prices are analysed in order to address the question of regionalization of the world crude oil market. Time-series econometrics in the form of error-correction modelling is applied for daily crude oil price data covering the time period 1988 to 2004 and in this framework topics like weak and strong exogeneity among three major oil prices – represented by Brent, OPEC and Texas (WTI) – are addressed. The empirical results are that causality is most likely bi-directional among these crude oil prices – and hence rejecting a regionalization hypothesis of the global oil market – and also an influence from the OPEC oil price towards Bent and WTI, which are usually claimed to have a benchmark role.

Notes

¹Additionally, the Perron (1997) test for unit roots allowing for a structural break (in the intercept) has also been applied -- and the unit root hypothesis is not rejected in any of the cases when including this type of break.

²Furthermore, Chaudhuri (2001) shows oil prices to be cointegrated with primary commodity prices.

³Applying the Johansen multivariate cointegration test to the oil prices Brent, WTI and OPEC indicates a similar conclusion, i.e. crude oil prices are cointegrated.

⁴The Bai and Perron (2003) test for multiple structural breaks in time series has also been applied to the oil price data (the test is performed with break(s) in the intercept term, cf. Bai and Perron, p. 16, using the Bai and Perron procedure from RATS/Estima). The most likely time of one structural break is 5 November 1999, rather close to the March 2000 sub-division of the data in the analysis. The results concerning more structural breaks do not seem sensible from an economic point of view and also appear (too) sensitive to the number of breaks investigated for, cf. Fig. 1 consistent with a break around 1999–2000. All the models presented have also been (re)estimated assuming a break in 1999 but as this only has a marginal influence on the parameter estimates and conclusions the results are not reported.