Missing Tariffs

Many studies use tariff data to measure changes in trade policy. This paper shows that a standard source of tariff data used in the literature suffers from substantial measurement error and selection bias due to missing tariffs. Many countries fail to report their tariff rates every year, with preferential tariffs more likely to be unreported than most favored nation (MFN) tariffs. The World Bank's World Integrated Trade Solution (WITS), the main provider of tariff data, falsely interpolates the missing preferential tariffs with MFN tariffs. This practice leads to artificial spikes in bilateral time series data and hence induces massive measurement error. Additionally, WITS tariff data are available only for importers reporting trade to UN Comtrade, such that low-income countries are underrepresented and tariffs for products with no trade flows are disregarded. I propose a new interpolation algorithm that takes this misreporting into account and combine five different sources of tariff data to introduce a new global tariff dataset at the six-digit product level covering 197 countries and 30 years. I illustrate the practical importance of the missing tariffs by replicating papers that were recently published in leading economic journals and that rely on WITS data. I find that existing estimates of the trade elasticity are particularly sensitive to the mishandling of the tariff data.