How reasonable are assumptions used in theoretical models? Computational evidence on the likelihood of trade pattern changes

In theoretical literature it is common to make the assumption that in a multi-country, multi-good world, the direction of trade (import and export by commodity) is predetermined and fixed for each good for each country. We consider a simple three-country, three-good, pure-exchange model with CES preferences. We compute free trade competitive equilibria, three-country non-cooperative Nash equilibria, and customs union equilibria for randomised parameterizations, and find that trade pattern changes between free trade and customs union equilibria in around 35% of cases.