Many nonmarket valuation models, such as the Ricardian model, have been estimated using cross sectional methods with a single year of data. Although multiple years of data should increase the robustness of such methods, repeated cross sections suggest the results are not stable. We argue that...