I have yet to read this paper, by Antoni Estevadeordal and Alan M. Taylor, but it appears to be of considerable interest for recent debates on free trade. The main argument is that trade really does raise growth rates by a noticeable amount:
According to the Washington Consensus, developing countries` growth would benefit from a reduction in tariffs and other barriers to trade. But a backlash against this view now suggests that trade policies have little or no impact on growth. If "getting policies right" is wrong or infeasible, this leaves only the more tenuous objective of "getting institutions right" (Easterly 2005, Rodrik 2006). However, the empirical basis for judging recent trade reforms is weak. Econometrics are mostly ad hoc; results are typically not judged against models; trade policies are poorly measured (or not measured at all, as when trade volumes are spuriously used); and the most influential studies in the literature are based on pre-1990 experience (which predates the "Great Liberalization" in developing countries which followed the GATT Uruguay Round). We address all of these concerns — by using a model-based analysis which highlights tariffs on capital and intermediate goods; by compiling new disaggregated tariff measures to empirically test the model; and by employing a treatment-and-control empirical analysis of pre- versus post-1990 performance of liberalizing and nonliberalizing countries. We find evidence that a specific treatment, liberalizing tariffs on imported capital and intermediate goods, did lead to faster GDP growth, and by a margin consistent with theory (about 1 percentage point per annum). Endogeneity problems are considered and other observations are consistent with the proposed mechanism: changes to other tariffs, e.g. on consumption goods, though collinear with general tariffs reforms, are more weakly correlated with growth outcomes; and the treatment and control groups display different behavior of investment prices and quantities, and capital flows.
Here is a non-gated version, though it is more than a year older. If there is any piece that can get Dani Rodrik blogging again, this is it.
Addendum: New updated, ungated version.