This article tests the Protection for Sale (PFS) model using detailed data from U.S. food processing industries from 1978 to 1992 under alternative import demand specifications. All empirical results support the PFS model predictions and previous empirical work qualitatively. Although welfare weights are very sensitive to import demand specification, a surprising result is that we obtain weights between 2.6 and 3.6 for domestic welfare using import slopes or elasticities derived from domestic demand and supply functions. In contrast, results based on import slopes or elasticities from directly specified import demands (including the Armington model) yield the usual, unrealistically large estimates for the domestic welfare weight. We contend that the latter empirical paradox arises mainly because the explanatory variables tend to be extremely large for industries with low import ratios and/or low estimated elasticities or slopes resulting from relatively volatile import prices. The results with derived import parameters point to a much stronger role of campaign contributions within the PFS model than previously found. They also suggest that the commonly-used Armington estimates may not be appropriate for estimating the PFS model.