We have focused our discussion of the empirical evidence regarding the effectiveness of IP on infant-industry protection. But clearly there are many other forms of IP: countries could subsidize exports across the board or in particular industries, they could impose differential taxes, as well as differentiated production, credit and R&D subsi- dies. Since a comprehensive review of all forms of IP is not possible, in the rest of this section we focus our discussion on export subsidies. Consider again the case in which some sectors exhibit Marshallian externalities. An overall export subsidy would simply preserve the allocation associated with the current pattern of comparative advantage. If export subsidies are targeted to the sectors that exhibit Marshallian externalities, then they could also be effective in switching the economy to the equilibrium with higher welfare. Again, production subsidies are less distortionary than export subsidies but they impose stronger fiscal demands. Thus, for governments with great fiscal needs or where taxation is very distortionary at the margin, export subsidies could be a reasonable option, although import tariffs would be less costly. In any case, the advantages of export subsidies relative to import tariffs in improving productivity are threefold: (1) that by promoting exports, a country makes sure that firms are subject to the “discipline of the international market,” which forces firms to become more productive; (2) that by subsidizing only exporting firms, a country effectively limits the subsidy to firms with high productivity; and (3) that domestic markets may be too small to allow the protected industry to reap the full benefits of Marshallian externalities. All of these arguments are relevant but require some qualification. First, the discipline of the international market applies both to firms that export and to those that sell in domestic markets as long as there are no quantitative restrictions. Second, there is in principle no rea- son to subsidize highly productive firms over low productivity firms (see Demidova & Rodrı́guez-Clare, 2008), unless there are barriers that prevent resources from flowing from the latter to the former, in which case the most efficient policy would be to remove those barriers. Finally, if the economy is small in relation to the industry size needed to fully exploit the Marshallian externalities, this is not going to be fixed by an export subsidy. A different and more reasonable argument is that domestic demand is not sufficiently sophis- ticated, hence firms selling to domestic consumers will not develop the necessary level of sophistication needed for success in international markets. Export orientation for infant industries would avoid this problem.29 One could redefine the case for industrial policy as trying to change incentives to produce (or export) some goods and not others. What evidence is there that what a country exports (or imports) matters? There are a number of studies listed in Appendix Table 1 which suggest that the growth effects of openness hinge on the composition of trade. These include An and Iyigun (2004), Choudhri and Hakura (2000), Dodaro (1991), Giles, Giles and McCann (1992), Ghatak, Milner, and Utkulu (1997), Hansen