Vector autoregression (VAR) models of agricultural prices and exports have appeared in the literature over the past decade. These models are commonly interpreted as dynamic alternatives to structural econometric models. Most VAR models, however, assume a recursive structure between the "equations" in the models (or the variables of interest). In this paper we compare recursive VAR models of interaction between macroeconomic factors and the agricultural economy with a variety of non-recursive (or structural) VAR models. The results of the non-recursive models are strikingly different from the recursive models. In particular, the impact of the exchange rate on agricultural exports is far mroe pronounced in the non-recursive models and the interactions among macroeconomic and agricultural variables appear to be stronger. The results suggest that research effort should concentrate on the identification problem in structural VAR models rather than concentrating on their recursive counterparts.