I. INTRODUCTION During the past decade, several new models of the political economy of trade policy emerged. Most notable among these is the Grossman and Helpman (1994) for sale model. This model views trade policy as the product of a menu auction in which special interest groups present trade policy officials with contribution schedules that specify the contributions of each interest group under different trade policies. Officials then choose the trade policy that best balances the contributions of special interests against the deadweight loss generated by protection. The primary significance of this model is the simple but elegant results it derives using a sophisticated game-theoretic framework. The simplicity of these results is attractive because it provides an ideal framework for empirical analysis. Several recent empirical studies, including Gawande and Bandyopadhyay (2000), Goldberg and Maggi (1999), and Gawande et al. (2001), have tested the implications of the Grossman-Helpman model. The results of these studies have tended to confirm the model, but as Gawande and Krishna (2003) point out, these results are open to criticism at several levels. For example, the degree of protection, which is the primary dependent variable, is measured in most of these studies by the coverage ratio of nontariff barriers. Coverage ratios, however, provide a crude measure of protection at best. Another problem is that these studies assume that the trade barriers used to construct the dependent variable are fully controlled by elected officials and can thus be adjusted in response to changes in campaign contributions. As a factual matter, this is simply not true with respect to some of the most important nontariff trade barriers, including antidumping and countervailing duties. Many of these non-tariff trade barriers (NTBs) are either determined by administrative processes that are largely immune to political pressure or are constrained by multilateral trade treaties. Treating these trade barriers as if they were under the control of elected officials is thus incorrect. A third problem involves determining when an industry is organized in support of protectionism, which is essential for any test of the Grossman-Helpman model. Previous studies have based their measure of industry organization on an industry's aggregate political action committee (PAC) contributions. In the absence of any means of determining why particular PAC contributions are made, however, it is difficult to know whether a contributing industry is organized in support of protection or in support of something entirely different. This problem is aggravated by the high level at which the industry data used in previous studies are aggregated, producing a situation in which some firms in virtually every industry make campaign contributions. The result is that it is difficult to separate those industries that are organized in support of protection from those that are not. These problems and others cast doubt on the initial results confirming the model. The purpose of this article is to test the Grossman-Helpman model in a context that avoids many of the problems associated with earlier studies. The author tests the Grossman-Helpman model using presidential decisions involving product eligibility under the U.S. Generalized System of Preferences (GSP). The GSP is a system under which developed countries grant preferential tariff treatment to imports of certain products from certain developing countries. This context avoids the issue of nontariff barriers because the products involved are subject almost exclusively to tariffs and not other forms of protection. The president (through the U.S. Trade Representative, USTR) also has direct control over product eligibility decisions, so there is a clearer connection between lobbying and the exercise of trade policy. The issue of industry organization is also addressed because industry lobbying is more readily observable in the context of GSP eligibility decisions. …