In this paper we investigate how vector autoregressive (VAR) models canbe used to study the soybean crush spread. By crush spread we mean a time se-ries marking the difference between a weighted combination of the value of soymealand soyoil to the value of the original soybeans. Commodity industry practitionersoften use fixed prescribed values for these weights, which do not take into accountany time varying effects or any financial market based dynamic features that can bediscerned from futures price data. In this work we address this issue by proposing anappropriate time series model with cointegration. Our model consists of an extensionof a particular VAR model used widely in econometrics. Our extensions are inspiredby the problem at hand and allow for a time varying covariance structure and a timevarying intercept to account for seasonality. To perform Bayesian inference we designan efficient Markov Chain Monte Carlo algorithm, which is based on the approach ofKoop et al. [2009]. Our investigations on prices obtained from futures contracts dataconfirmed that the added features in our model are useful in reliable statistical de-termination of the crush spread. Although the interest here is on the soybean crushspread, our approach is applicable also to other tradable spreads such as oil andenergy based crack or spark.