The second argument for preferences is less obvious. McAfee and McMillan (1989) show that even if the government were indifferent to the distribution of profits between domestic and foreign firms, welfare could be improved by discriminatory procurement polices. In their model, there is imperfect competition in the bidding for a contract, and each bidder is better informed about his own costs than either rival bidders or the procurer. If foreign firms have cost advantages and the government is interested in minimizing the expected procurement cost, then the government should discriminate in favour of domestic firms. In the absence of such discrimination, foreign firms may choose to bid just below what they expect domestic firms to bid, which will be higher than their actual costs. A price preference policy increases the "effective" competition from domestic firms and forces the foreign firms to lower their bids.10
This argument can be illustrated by a simple numerical example in which we assume away uncertainty and asymmetric information. Say there are only two firms, a domestic firm with constant marginal costs of 121, and a foreign firm with constant marginal costs of 100. In the absence of preferences, the foreign firm would bid slightly less than 121 (say 120), and be sure to win the contract. Now the domestic government announces that it will grant a 20 per cent preference to domestic firms - i.e. domestic bids will be accepted as long as they are not more than 20 per cent higher than foreign bids. This would force the foreign firm to bid close to its cost of 100 in order to be sure of winning the contract - only then will its preference-inflated price be lower than the domestic firm's cost of production. The preference policy means that the government can procure from the foreign firm at a price of around 100 rather than at a price close to 121. The gain in domestic welfare, and the reduction in foreign rents, would be around 20.
(When optimal policy requires discrimination in favour of domestic firms, it is important to know how to implement such a policy. Branco (1994) shows that the correct way to implement the discrimination policy varies according to the mechanism used. Most procurements are organized (due to legal requirements?) as first price sealed-bid auctions, with the payment to the winner depending on its own bid only. Discrimination issues are more complex in these mechanisms than in mechanisms where the payment depends on the losing bids. Furthermore, a constant price preference ratio is usually not optimal policy.