Abstract
This article solves the border puzzle that asks why countries have such a strong preference for consuming their own goods. After replicating the negative impact of the border on Canada–United States international trade using the methods of previous research, it is shown that past estimates that find the border to be a significant barrier to international trade are the result of statistical misspecification.
Acknowledgement
I woluld like to thank Trevor J. Barnes and Stephen T. Easton for helpful comments on earlier drafts of this paper.
Notes
1 Anderson and van Wincoop (Citation2003) have proposed a solution to the border puzzle but that solution is not complete. The Canada–United States border still decreases international trade flows and their specification is based on the assumption that international and interprovincial trade are perfect substitutes that is not supported empirically (Coulombe, Citation2003, Citation2004).
2 Most of the research investigating the border effect has not used panel data.
3 Wall (Citation2000) also claims that the border effect parameters need to be obtained in this manner because of perfect multicollinearity between border effects and fixed effects. This, however, is not the case. Fixed effects are present for all trading-partner-pairs (province–province and province–state), whereas the border effects are only present for cross-border trading-partner-pairs (province–state).
4 Canadian Territories constituting 0.4% of the population are excluded as the relevant measures of the independent variables are not available.
5 The period-specific dummy variables are not included in for interest in brevity.