Domestic Trade and its Measurement: A Simulation Analysis

The inclusion of domestic (intra-national) trade when estimating structural gravity models is an important topic and has been shown to solve empirical puzzles such as the missing globalization puzzle (Yotov, 2022). Despite efforts to construct intra-national trade data, its measurement still remains a challenge, in particular for historical trade data. Campos et al. (2021a) recently claimed that the exact definition of domestic trade flows (gross vs. net production) is not important for the estimation of free trade agreement (FTA) effects in a gravity framework: a surprising finding, since the size differences between these measures are large, and their correlation of about 80%, while high, is not perfect. In a simulation framework, we revisit the question of the (non)-importance of whether to include and how to measure domestic trade and find some support for the conclusion of Campos et al. (2021a) concerning the estimation of trade agreement or tariff effects. However, with regard to the point estimates, marked differences arise depending on how domestic trade flows are calculated: the GDP-based domestic trade flows clearly distort the coefficient estimates. Interestingly, using only inter- national trade flows yields results that are unbiased and as precisely estimated as those obtained when correctly measured (GO-based) domestic trade flows are included. Depending on what effects researchers are interested in, the basic gravity model without inclusion of domestic trade flows may thus be the preferred alternative after all.