of demand for ordnance and accessories, and over 40 per cent of demand for aircraft and parts, and a substantial part of demand in categories such as scientific equipment, transport equipment and computers. However, the exclusion of defence purchases, which are mostly outside the scope of the GPA, changes the picture substantially. It turns out that services like construction, maintenance and repair, computer and data processing services are some of the most important. A key variable, in terms of the test identified above, is the share of public purchases in total domestic production. Again, non-defence procurement accounts for a substantial share of domestic output only in the services identified above, and in the ophthalmic and photographic equipment sector.
The finding that government procurement is of the greatest relative importance in defence-related sectors probably holds, not just for the United States, but for many other countries . However, the finding that non-defence related procurement is relatively important only in a few sectors is probably less generalizable. In contrast to many other countries, the United States does not have a telecommunications monopoly, state-owned airlines, or full state ownership of airlines. Thus, Francois, et al. (1995) show that public ownership of the type found in certain European Union states would imply significant government presence in the markets for engines, turbines, transportation equipment, communications, pipelines, air transport services, communications equipment, and a number of utility-related sectors
A second qualification of the neutrality result is that domestic and foreign goods may not be perfect substitutes. Shifts in government expenditure towards domestic goods and the associated price changes may not then induce fully offsetting shifts in private demand towards foreign goods. Under assumptions of imperfect substitutability, Baldwin and Richardson (1972) estimate that the Buy American program (excluding agricultural commodities, minerals and armaments) reduced total imports in 1963 by between $76 million and $110 million (depending on the precise elasticity assumptions) from a base of approximately $20 billion. Furthermore, in markets for differentiated goods, discrimination need not always be through price preferences, it could also take the form of choice of inferior domestic goods, or choice of domestic goods which are more distant from the preferences of domestic consumers than foreign alternatives. The implications of such discrimination have not been fully explored, but some work in this area is described in the next section.