This study has used an empirical approach developed by Urga and Walters (2003) to examine the implications of the short-run specification of the standard translog cost specification along with the possible implications of non-stationarity. We have estimated a dynamic translog cost specification complete with dynamic share equations for U.S. agriculture and compared it to the static, long-run specification. We found that the dynamic translog specification yielded more significant parameter estimates, and yielded results that are consistent with economic theory. In particular, the coefficient m (the adjustment cost parameter) determines the overall autoregressive structure of the model. The fact that its estimated value (0.36) is statistically different from zero at any conventional level of confidence indicates that the dynamic structure of the model is important. This finding illustrates the superiority of the short-run, dynamic specification over the static, long-run model.