A set of rigorous diagnostic techniques is used to evaluate the forecasting performance of five multivariate time-series models for the U.S. cattle sector. The root-mean-squared-error criterion along with an evaluation of the rankings of forecast errors reveals that the Bayesian vector autoregression (BVAR) and the unrestricted VAR (UVAR) models generate forecasts which are superior to both a restricted VAR (RVAR) and a vector autoregressive moving-average (VARMA) model. Two methods for calculating a test evaluating the ability to forecast directional changes are implemented. The BVAR models and the UVAR model unambiguously outperform the VARMA model in the forecasting directional change