The reaction of trade volumes to tariffs is far larger than what current models predict. One reason for this is that they abstract from endogenous productivity choices (“innovation”), which amplify this reaction. To show this, I develop a model of international trade with innovation, and calibrate it to Canada and United States before the Free Trade Agreement. Feeding in the tariff drops observed during the agreement, the increase in the trade volumes is within the empirical estimates. Without innovation, the change in trade volumes is too low, and similar to what current models without innovation have found.