Gordon H. Hanson & Chong Xiang suggest that fixed export costs are multilateral rather than bilateral, meaning that it’s costly for businesses to “go global” but not particularly costly to enter particular countries:
In this paper, we develop a simple empirical method to test two alternative versions of the Melitz (2003) model, one with global fixed export costs and one with bilateral fixed export costs. With global costs, import sales per product variety (relative to domestic sales per variety) are decreasing in variable trade barriers, as a result of adjustment occurring along the intensive margin of trade. With bilateral costs, imports per product variety are increasing in fixed trade costs, due to adjustment occurring along the extensive margin. We apply our approach to data on imports of U.S. motion pictures in 46 countries over 1995-2006. Imports per product variety are decreasing in geographic distance, linguistic distance, and other measures of trade costs, consistent with adjustment to these costs occurring along the intensive margin. There is relatively little variation in the number of U.S. movies that countries import but wide variation in the box-office revenues per movie. The data thus appear to reject the bilateral-fixed-export-cost model in favor of the global-fixed-export-cost model.
On the other hand, motion pictures are not exactly the typical manufacturing product, so I would hesitate to extrapolate from these results to other sectors.