Structural Estimation of a Flexible Translog Gravity Model

How large are the gains from trade? Do all trade models have the ‘same old gains’? Arkolakis, Costinot, and Rodriguez-Clare (2012) show that many quantitative trade models that summarize trade responses via a single elasticity have the same welfare implications. I develop a flexible approach to estimating trade responses using a translog expenditure function, and find welfare results that differ starkly from conventional trade models. In my model, trade responses can vary bilaterally, and the link between own- and cross-price elasticities of trade to trade cost is broken. I apply my approach to inter-regional trade flows in North America and international trade flows between OECD and BRICS countries. I structurally estimate the parameters and conduct counterfactual analyses. Canada’s border effect is at least three times smaller than estimates in previous literature. Compared to those implied by the formula in Arkolakis et al., welfare responses are larger and more heterogeneous. The welfare losses from raising trade barriers are underestimated by eight times for China, France, India, and the United Kingdom, and underestimated by more than ten times for Australia, Brazil, Canada, and Russia.