How Bad Are Trade Wars? Evidence from Tariffs

Petr Polák, Nikol Poláková, Anna Tlustá

We use more than 1,600 estimates from 71 studies to investigate the relation between international trade flows and tariffs. Our results suggest that the empirical literature suffers from the presence of publication bias, which has exaggerated the effect (the true elasticity is closer to zero). After accounting for publication bias, we estimate the trade elasticity with respect to tariffs to be between  -0.9 and -2.0. The results of Bayesian model averaging, which takes into account model uncertainty, show that the differences among estimates are systematically driven by the type of data (panel and level of aggregation), the data source (WITS vs. other databases), control variables (distance and trade agreements dummy), and estimation techniques (use of country-level fixed effects). The effect is also diminishing over time.

JEL codes: F12, F13, F14, O24, O30

Keywords: BMA, international trade, meta-analysis, publication bias, tariffs, trade costs

Issued: December 2020

Download: CNB WP No. 15/2020 (pdf, 1.6 MB)