Error correction models impose few prior restrictions on dynamic model specification and allow the data to determine model structure. Despite this obvious advantage, few applications have adopted the error correction model to explain trade flows. An error correction model of cotton import demand is estimated for France, Japan, and Hong Kong. A variety of tests are applied to determine the dynamic structure of the model. We find the most general models are those that best fit the data for cotton import demand. Long-run elasticities from these general models are significantly different than elasticities derived from a comparable static model.