Currency unions and trade: the effect is large

This critique of my Economic Policy paper generates many emotions. It is certainly an honor for my research to be taken seriously by such an eminent scholar as my friend Torsten Persson. And while any publicity is good, a negative citation dominates a positive one. 2 There is irony; I came up with the original idea in a hallway conversation with Harry Flam a few feet from Persson’s office. There was indubitably surprise, especially at the AEA meetings where Persson’s research was first presented. There is fear; neither Economic Policy nor I want a reputation for publishing sloppy work. Persson is a ruthlessly efficient scholar with an almost fanatical devotion to the truth. But even if I wasn’t expecting a Spanish Inquisition, honor demands a satisfactory response. 3 Defense My original estimate of the effect of currency union on trade was large, at least compared to my intuition (there were no extant estimates in the literature). The surprising thing was how difficult it was to reduce this effect. More precisely, the coefficient (denoted γ) on a currency union (CU) dummy in an empirical “gravity ” model of bilateral trade seemed doggedly positive, and significant in both economic and statistical terms. Its value rarely fell below 1.2, implying an effect of currency union on trade of at least e 1.2 ≈ 300%. I tried hard – plenty hard – to reduce the size of the effect. I went to pains to perform sensitivity analysis, providing over fifty estimates of the effect of currency union on trade. I cut my sample in different ways, added many factors, took simultaneity seriously, measured