In this paper I draw lessons from two quasi-natural experiments (the transition process in former Communist countries and the rapid globalization of food chains) on the optimality of farms and agricultural structures more generally. I argue that (a) the farm structures that have emerged from the transition process are much more diverse than expected ex ante; (b) this diversity is to an important extent determined by economic mechanisms which are influenced by initial conditions (eg technology) and reform policies; (c) non-traditional farm structures have played an important role during transition since they were optimal to address the specific institutional and structural constraints imposed by the transition process; (d) there is more diversity than often argued in the farms that are integrated in global food chains; (e) endogenous institutional (contracting) innovations in food chains may lock existing farm structures in a long-run institutional framework; and (f) indicators based on farm structures are not a good measure of welfare effects of the globalization of food chains.