“…Some authors have argued that this breakdown of the relationship reveals that the relationship between the variables is non-linear, and thus have proposed different specifications of it, particularly asymmetric ones. These specifications can be implemented either by introducing two separate variables for oil price increases and decreases, as in the models suggested by Mork (1989) and Mork et al (1994), or by estimating Markov regime-switching models characterized by high and low inflation periods (Çatik and Önder, 2011), or even by using a quantile regression framework to estimate the marginal effect on inflation in the distribution (Chortareas et al, 2012). Other authors have suggested different ways to take time variation into account, either by splitting the sample into two sub-periods assuming a structural break in the early 1980s, by estimating regressions over rolling time windows, or by using time-varying parameters models.…”