Non-linearities in the dynamics of oil prices

Examining stationarity is of particular importance and represents the first step in empirical time-series research. Non-stationarity invalidates many of the results obtained from standard techniques and, therefore, requires special treatment. Because oil prices play an important role in affecting economic variables, this paper examines the stationarity of real oil prices (Brent, Dubai, WTI and the World) over the period 1973:2-2011:2. Real oil prices are expressed in the currencies of seven Asian countries (Indonesia, Japan, Korea, Malaysia, the Philippines, Singapore and Thailand) and in the U.S. dollar. While using linear unit root tests without structural breaks shows no evidence of stationarity, allowing for breaks shows very limited evidence of stationarity. We argue that these results are attributed to the presence of nonlinearities in the behavior of oil prices. Testing for nonlinearity shows significant evidence of nonlinearity in all the cases with evidence of exponential smooth transition autoregression (ESTAR) nonlinearity-type in most cases. Applying unit root tests that account for two types of nonlinearities (smooth transition and nonlinear deterministic trends) reveals evidence of stationarity in all the cases.

► The paper utilizes non-linear models to examine the stationarity of real oil prices. ► Linear unit root tests with structural breaks show limited evidence of stationarity. ► Using unit root tests of two types of nonlinearities shows evidence of stationarity. ► Results show that linear models are not appropriate in modeling oil price behavior.