US–Canada border effect between 1993 and 2007: smaller, less asymmetrical, and declining

We revisit the US–Canadian border effect on trade from 1993 through 2007. For 1993 data, we find that traditional log-transformed gravity models make the border effect larger, while data construction processes make the border effect smaller. Overall, we find that the border effect in 1993 is 4.1, which is 15–20 % lower than the previous estimates of approximately 5 by Anderson and van Wincoop (Am Econ Rev 93(1):170–192, 2003) and Feenstra (Scott J Polit Econ 49(5):491–506, 2002). Using the same data construction and specification methodologies, we find that the effect has shrunk to 3.2 by 2007; namely, the border effect in 2007 reduces state-province trade by two-thirds. Thus, between 1993 and 2007, the border effect declined by 22 %. Based on the structural estimator, we also find that the border effect is stronger for Canadian provinces as shown by Anderson and van Wincoop (2003), but the asymmetry in the border effect between the US and Canada shrank between 1993 and 2007.