Empirical evidence on the Generalized System of Preferences (GSP) for low-income countries generally portrays a rather stark prediction: the program has produced virtually no impact on intended low-income beneficiaries’ exports to high-income countries. This result, based on total merchandise trade, is misleading because it masks three underlying heterogeneities in the program: i) preference structure across countries, ii) pre-existing distortions across sectors, and iii) rules of origin. Using a theoretically consistent gravity equation for sector- and product-level trade over 1962–2010, we illustrate that the GSP has delivered significant positive effects for lowincome countries’ agricultural exports (but not necessarily for their nonagricultural exports) to developed countries.