Includes bibliographical references (p. 38-39).

Why do governments employ inefficient policies to redistribute income towards special interest groups (SIGs) when more efficient ones are available? To address this puzzle we derive and test predictions for a set of policies where detailed data is available and an efficiency ranking is feasible: tariffs vs. non-tariff barriers (NTBs). In our policy choice model a government bargaining with domestic SIGs can gain by constraining tariffs through international agreements even if this leads to the use of the less efficient NTBs. This generates two key testable predictions (i) there is imperfect policy substitution, i.e. tighter tariff constraints are not fully offset by the higher NTBs they generate and (ii) the decision to commit to constraints depends on the government's bargaining power relative to SIGs. Using detailed data, we confirm that tariff constraints in trade agreements increase the likelihood and restrictiveness of NTBs. We also provide a structural estimate that indicates NTBs are less efficient than the tariffs they imperfectly replace. Moreover, we find parametric and non-parametric evidence that the higher the government bargaining power relative to a SIG the more relaxed the tariff constraint it chooses. This result is stronger for organized industries, which further supports the theory. The main theoretical insights and empirical approach can be applied to other policies to provide additional evidence on inefficient redistribution.