This paper analyzes the provisions of the new Agreement on Government Procurement (GPA), drawing insights from trade theory and recent developments in the economics of information and law. A central conclusion is that in a world where imperfectly informed procurers purchase from imperfectly competitive firms on behalf of imperfectly informed tax-payers, it is not easy to devise rules which would be optimal in all situations. Nevertheless, the non-discriminatory provisions of the GPA seem to approximate closely the rules which would maximize global welfare. A significant benefit of the GPA is in helping to overcome national agency problems in procurement by creating mechanisms for reciprocal international monitoring supported by multilateral enforcement. There is, however, scope for improvement. First, the GPA does not equip bidders for government contracts to vault over trade restrictions, so the creation of genuine international competition for government procurement remains crucially dependent on the liberalization of trade. Secondly, weaknesses remain in the enforcement mechanism. These include the low level of compensation to a successful challenger, the absence of restrictions on settlements, and the lack of provision for challenge and review of bail-outs. The paper proposes certain improvements.