August 1997 This paper explores a world in which regional trade agreements help reduce security tensions between neighbors. Regional integration agreements (RIAs) are examples of second best and have an ambiguous impact on welfare, contend Schiff and Winters. They build a model in which RIAs unambiguously raise welfare by correcting for externalities. It assumes that trade between neighboring countries increases trust between them and reduces the likelihood of conflict. The optimum intervention in that case is a subsidy on imports from the neighbor. The authors show that an equivalent solution is for the neighboring countries to tax imports from the rest of the world- is, to form an RIA- with imposing some domestic taxes. In fact, security threats have moved neighboring countries to form RIAs. Examples include the creation of the European Coal and Steel Community (1951) and the European Economic Community (1957) to reduce the threat of war in Europe, as well as various RIAs among developing countries. Schiff and Winters show, among other things, that: * The optimum tariffs on imports from the rest of the world are likely to decline over time. * Deep integration implies lower optimum external tariffs if it is exogenous. * But if deep integration is endogenous, it implies higher optimum external tariffs before it occurs and lower ones thereafter. * Enlargement of a bloc (in terms of symmetric countries) has an ambiguous impact on external tariffs but improves welfare, and some form of domino effect exists in the sense that enlargement increases the incentive for nonmembers to seek accession. Although externalities associated with security matters imply that an RIA may maximize welfare, this model suggests that the RIA is a transitory arrangement in the sense that optimum trade preferences are highest at the time the RIA is formed (when security is low) and tend to decline over time. In other words, the RIA's external trade policy becomes increasingly open over time (as well as following deep integration). This paper-a product of the Development Research Group-is part of a larger research program on regionalism and development (directed by the authors).