This paper empirically investigates the role of transport costs, trade and investment barriers, production scale economies, and firm-specific advantages in determining the use of overseas production relative to exports. The proximity-concentration hypothesis is robust in explaining the share of total sales accounted for by affiliate sales: this share is greater the higher are transport costs and trade barriers and the lower are plant scale economies and investment barriers. Although strictly speaking, the proximity-concentration hypothesis applies to the shares of affiliate sales and exports rather than the levels, the effects of trade and investment barriers on the levels are similar to their effects on the shares, controlling for simultaneity, and so is that of freight factors in the trade estimates. The elasticity of inward and outward net affiliate sales with respect to tariffs is around 0.45, and that with respect to NTBs is an additional 0.17. The elasticity of both imports and exports with respect to freight factors is -1. However, the effect of freight factors on the level of affiliate sales is not robust, and the probability of observing any affiliate sales in a market increases with proximity. The overall complementarity between trade and affiliate sales arises in part because relative income and intellectual property intensity increase both. In contrast, affiliate sales and trade move in opposite directions with increases in advertising intensity, suggesting that advertising-intensive products require a local presence. The paper empirically investigates the proximity-concentration hypothesis, using a variety of methods to avoid simultaneity between multinational sales and trade flows. Tests on the share of total foreign sales accounted for by affiliate sales provide strong support for the proximity-concentration hypothesis: affiliate sales comprise a greater share the higher are transport costs and trade barriers and the lower are plant scale economies and investment barriers. In addition, the effects of trade and investment barriers on the levels of trade and affiliate sales are similar to their effects on the shares, even though, strictly speaking, the hypothesis applies only to the shares. Similarly, transport costs have the predicted negative effect on the level of trade. However, the effect of transport costs on the level of affiliate sales is less robust, and the probability of observing any multinational sales in a market decreases with the distance between markets. These findings contrast with past research, which has found that distance is insignificant in explaining multinational activities. The difference is attributable in part to a more accurate measure of transport costs. In addition, the tests avoid the simultaneity problems encountered by earlier work by using the share of overall sales as the dependent variable and by using a two stage least squares specification for the levels estimates. Both the factor proportions and proximity-concentration hypotheses assume that firms in differentiated products (or oligopolistic) industries exploit their proprietary advantages internally and focus on the location decision. A third set of models (Ethier, 1986; Horstmann and Markusen, 1987; Ethier and Markusen, 1991; Raff, 1992)This paper empirically investigates the role of transport costs, trade and investment barriers, production scale economies, and firm-specific advantages in determining the use of overseas production relative to exports. The proximity-concentration hypothesis is robust in explaining the share of total sales accounted for by affiliate sales: this share is greater the higher are transport costs and trade barriers and the lower are plant scale economies and investment barriers. Although strictly speaking, the proximity-concentration hypothesis applies to the shares of affiliate sales and exports rather than the levels, the effects of trade and investment barriers on the levels are similar to their effects on the shares, controlling for simultaneity, and so is that of freight factors in the trade estimates. The elasticity of inward and outward net affiliate sales with respect to tariffs is around 0.45, and that with respect to NTBs is an additional 0.17. The elasticity of both imports and exports with respect to freight factors is -1. However, the effect of freight factors on the level of affiliate sales is not robust, and the probability of observing any affiliate sales in a market increases with proximity. The overall complementarity between trade and affiliate sales arises in part because relative income and intellectual property intensity increase both. In contrast, affiliate sales and trade move in opposite directions with increases in advertising intensity, suggesting that advertising-intensive products require a local presence. The paper empirically investigates the proximity-concentration hypothesis, using a variety of methods to avoid simultaneity between multinational sales and trade flows. Tests on the share of total foreign sales accounted for by affiliate sales provide strong support for the proximity-concentration hypothesis: affiliate sales comprise a greater share the higher are transport costs and trade barriers and the lower are plant scale economies and investment barriers. In addition, the effects of trade and investment barriers on the levels of trade and affiliate sales are similar to their effects on the shares, even though, strictly speaking, the hypothesis applies only to the shares. Similarly, transport costs have the predicted negative effect on the level of trade. However, the effect of transport costs on the level of affiliate sales is less robust, and the probability of observing any multinational sales in a market decreases with the distance between markets. These findings contrast with past research, which has found that distance is insignificant in explaining multinational activities. The difference is attributable in part to a more accurate measure of transport costs. In addition, the tests avoid the simultaneity problems encountered by earlier work by using the share of overall sales as the dependent variable and by using a two stage least squares specification for the levels estimates. Both the factor proportions and proximity-concentration hypotheses assume that firms in differentiated products (or oligopolistic) industries exploit their proprietary advantages internally and focus on the location decision. A third set of models (Ethier, 1986; Horstmann and Markusen, 1987; Ethier and Markusen, 1991; Raff, 1992)