This paper investigates the robustness of CGE models with respect to the elasticities of substitution in demand between domestically produced goods and foreign goods – the so-called Armington elasticities. The Armington-type modeling of trade is still one of the most extensively used specifications in CGE modeling. For a long time the choice of the respective elasticities of substitution has not been given much attention. The most frequently used procedure was to adopt the elasticities from the literature, which meant using elasticities that had been estimated (or guessed) for a different country and often also for a different degree of data aggregation. However, recently, some authors have shown that the elasticities 1) vary more substantially over countries than had been expected and 2) are higher in more recent estimations than in those which have been published in the 1980s and 1990s.