Can we bring together the two arguments for preferences: one based on the inclusion of domestic firms profits in the government's objective function and the other on cost differences between firms? The analysis of McAfee and McMillan (1989) and Branco (1994a) suggests that the second consideration may reinforce, but never offsets, the case for preferences based on the first consideration. Thus, the optimal extent of discrimination against foreign firms may be greater in comparative disadvantage industries than in comparative advantage industries.12
How realistic are the assumptions on which these arguments are based, and how do they compare with the world in which the GPA exists? Consider first the argument for preference based on cost differences between firms. There are at least three important qualifications. First, this is not an argument for preferences being given to all domestic firms since a country is unlikely to have relatively high costs in all industries from which the government purchases. It would seem, therefore, that the government would wish to give preferences to some local industries, while in others it would give preference to foreign industries.
This proposition can be put to a simple test. As noted above, the GPA allows parties to pick and choose services sectors that will be subject to its disciplines, which prohibit any preference being given to domestic firms but do not prevent preferential treatment of foreigners.13 Now according to the above argument, each Member would wish to retain the right to grant preferences in sectors in which it had a cost disadvantage, but be willing to foreclose this option in industries in which it had a cost advantage. Thus, somewhat ironically, Members would tend to include the industries in which they have comparative advantage and are anyway unlikely to award contracts to foreigners; and exclude the industries in which they have a comparative disadvantage and where foreign competition could play a meaningful role. In fact, the lists of included services sectors of the Members are fairly similar, at least at an aggregative level, so there is little obvious support for this argument.
(Branco (1994b) extends the analysis of procurement preferences to a dynamic context, and examines the argument that procurement preferences may reduce incentives for domestic firms to invest in new and more efficient technologies. He finds that if the cost of adopting a new technology is low, favouring the domestic firms is harmful, in the sense that it makes welfare improving adoption less attractive to the firms. However, if the cost of adoption is high from the firm's point of view but not the government's, then some degree of favouritism may be needed to induce the firm to adopt the new technology. Branco (1995) comes to similar conclusions, and demonstrates that non-discriminatory treatment is usually not too far from the optimal discriminatory mechanism.
(Exclusions are also permitted in the case of goods, but these are more the exception than the rule.