EcoSta 2026: Start Registration
View Submission - EcoSta2026
A1372
Title: Connected trade flows via trade costs: A spatial autoregressive framework Authors:  Hanbat Jeong - Macquarie University (Australia) [presenting]
Jieun Lee - Emory University (United States)
Abstract: A new gravity specification grounded in countries' network-leveraging features is proposed. Building on spatial autoregressive models, trade costs are endogenized as a function of a network of country proximities. The resulting specification induces interdependence among trade flows and pair-specific heterogeneity, and nests the conventional iceberg cost as a special case when countries do not leverage network connections. For estimation, PPML is used and robust inference accommodating heteroskedasticity and unknown error correlations is developed. A fast algorithm for the large-dimensional network multiplier transformation is also proposed. Empirically, evidence of endogenous trade costs and substantially improved fit relative to the conventional gravity equation (up to 30\%) is found, implying pair-specific heterogeneity missed by existing models. Counterfactual analysis shows that a bilateral increase in trade costs generates sizable reallocations of import shares across many third countries once network interactions are accounted for.