Why do Estimates of the EMU Effect on Trade Vary so Much?

Larger data sets, with more countries and a longer span of time, exhibit systematically larger effects of European monetary union on trade. I establish this stylized fact with meta-analysis and confirm it by estimating a plain-vanilla gravity model. I then explain this finding by examining systematic biases in "multilateral resistance to trade" manifest in time-varying country fixed effects; bias grows as the sample is truncated by dropping small poor countries.