This paper presents a theory of government intervention that explains "industrial strategy" policies such as R&D or export subsidies in imperfectly competitive international markets. Each producing country has an incentive to capture a greater share of rent-earning industries through subsidies, but the subsidy-ridden international equilibrium is jointly suboptimal. The equilibrium in the strategic game involving firms and governments is modeled as a three-stage subgame perfect Nash equilibrium. The assumption that the government is the first player in this game allows it to influence equilibrium industry outcomes by altering the set of credible actions open to firms. The paper focuses on possible subsidization of cost-reducing or "process" R&D, such as the recent Japanese and French subsidization of robotics in automobile assembly. Export subsidies are also considered. It is shown that with export subsidies available, countries would not choose to subsidize R&D. However, because GATT codes effectively restrict direct export subsidies, the setting in which only R&D subsidies are available is viewed as the most relevant case. R&D is assumed to be undertaken before the associated output is produced, with firms anticipating the effect of R&D on the resolution of output shares. Thus, R&D serves as a commitment or credible threat, along the lines considered in Spence (1977, 1979), Friedman (1979), Dixit (1980), and Eaton and Lipsey (1980, 1981). In contrast to these papers, where an established firm acts first, the paper uses a model developed in Brander and Spencer (1983) where firms have equal opportunity in setting R&D levels. The efficacy of government policy in this paper arises from the assumption that a government can credibly commit itself to R&D (or output) subsidies before the R&D decisions are made by private firms. This is an example of a more general principle in understanding government policy: the government becomes the first player in a multi-stage game and can influence the equilibrium outcome of the game played by private agents by altering the set of credible actions open to them. The paper shows that a government which has the objective of maximizing domestic welfare has an incentive to introduce such subsidies. The model of firm behavior is based on a two-stage game played by two competing firms, which are imagined to be located in different countries. In the first stage, firms choose R&D levels, and in the second stage, output levels. The second stage equilibrium is a Nash equilibrium in outputs, taking R&D levels as given by the preceding stage. Using this second stage solution, the paper can then write down the payoff functions for the game played at the preceding stage: the profits of each firm are written as a function of the pair of R&D levels chosen. The paper seeks a Nash equilibrium in that game. This gives rise to a subgame perfect equilibrium in the two-stage game, which is characterized by inefficiently high levels of R&D forThis paper presents a theory of government intervention that explains "industrial strategy" policies such as R&D or export subsidies in imperfectly competitive international markets. Each producing country has an incentive to capture a greater share of rent-earning industries through subsidies, but the subsidy-ridden international equilibrium is jointly suboptimal. The equilibrium in the strategic game involving firms and governments is modeled as a three-stage subgame perfect Nash equilibrium. The assumption that the government is the first player in this game allows it to influence equilibrium industry outcomes by altering the set of credible actions open to firms. The paper focuses on possible subsidization of cost-reducing or "process" R&D, such as the recent Japanese and French subsidization of robotics in automobile assembly. Export subsidies are also considered. It is shown that with export subsidies available, countries would not choose to subsidize R&D. However, because GATT codes effectively restrict direct export subsidies, the setting in which only R&D subsidies are available is viewed as the most relevant case. R&D is assumed to be undertaken before the associated output is produced, with firms anticipating the effect of R&D on the resolution of output shares. Thus, R&D serves as a commitment or credible threat, along the lines considered in Spence (1977, 1979), Friedman (1979), Dixit (1980), and Eaton and Lipsey (1980, 1981). In contrast to these papers, where an established firm acts first, the paper uses a model developed in Brander and Spencer (1983) where firms have equal opportunity in setting R&D levels. The efficacy of government policy in this paper arises from the assumption that a government can credibly commit itself to R&D (or output) subsidies before the R&D decisions are made by private firms. This is an example of a more general principle in understanding government policy: the government becomes the first player in a multi-stage game and can influence the equilibrium outcome of the game played by private agents by altering the set of credible actions open to them. The paper shows that a government which has the objective of maximizing domestic welfare has an incentive to introduce such subsidies. The model of firm behavior is based on a two-stage game played by two competing firms, which are imagined to be located in different countries. In the first stage, firms choose R&D levels, and in the second stage, output levels. The second stage equilibrium is a Nash equilibrium in outputs, taking R&D levels as given by the preceding stage. Using this second stage solution, the paper can then write down the payoff functions for the game played at the preceding stage: the profits of each firm are written as a function of the pair of R&D levels chosen. The paper seeks a Nash equilibrium in that game. This gives rise to a subgame perfect equilibrium in the two-stage game, which is characterized by inefficiently high levels of R&D for