Grain prices, oil prices, and multiple smooth breaks in a VAR

Ignored structural breaks in a VAR result in a misspecified model such that Granger causality tests are improperly sized; there is a bias towards a rejection of the null hypothesis of non-causality even when the null is correct. Instead of modeling structural breaks as being sharp, changes in the relationship between the maize and petroleum markets are likely to have occurred gradually. We show the flexible Fourier form has good size and power properties in testing for smooth structural change in a VAR. When applied to a VAR including maize and oil prices, we uncover important linkages between the two markets.