Comments and Discussion
- Comments and Discussion
Charles Engel:
Anderson and van Wincoop (AvW) have made an important contribution by analyzing border barriers in the context of Anderson's gravity model.1 AvW's earlier paper shows how border effects should be estimated in the context of the gravity model.2 In the present study, the authors take parameters estimated from the empirical work and use them in the gravity model to measure the welfare effects of borders.
This is truly excellent work, a perfect example of economics as a science. AvW's previous work clarifies exactly what is meant by a "border effect" in the context of a theoretical model.3 The theoretical model is used as a guide to quantifying the border effect. Because AvW have estimated a structural model, they can tell us the effects on welfare and on trade volumes of changes in trade policy or other border barriers, as they do in this study.
The paper is also ingenious because of a number of specific points it makes about border barriers. At some level these points must be obvious to a trade economist, but these are the types of observations that only become obvious once they are pointed out. AvW note that (a) much of the border effect must arise from barriers that do not create rents, so that their welfare impact is much greater than rent-producing barriers such as tariffs; (b) the impact of a border barrier on trade between two countries does not depend so much on the size of the barrier, but the size of the barrier relative to "multilateral" barriers; and (c) the welfare costs of these barriers are not directly related to the amount of trade that is diverted. [End Page 231]
I believe that the work of Anderson and van Wincoop is a starting point, and there are many avenues that lead from their recent papers. In this paper,AvW point to the relative simplicity and clarity of their gravity model as compared to computable general equilibrium models, which they say "are simulated rather than estimated, and … are almost always based on a very large black box consisting of dozens to hundreds of equations." But I shall argue that in some ways the model of AvW is surely too simple, and the quantitative answers provided by AvW are too rough at this stage to be useful as a guide to policymakers.
I shall also make some comments specific to AvW's calculations of the benefits of a currency union. My concerns there are with the interpretation of the empirical work of Rose and van Wincoop.4
As I have indicated, the paper is a major step in the literature because of its conceptualization. AvW show how we can estimate a border effect, and how to translate those estimates into structural parameters of the theoretical gravity model—and they do a beautiful job of it.
As AvW discuss, border effects may arise from policies that erect trade barriers. Policies such as tariffs or quotas are easily observable and quantifiable, and generally create rents. But Anderson and van Wincoop make a good case that many border barriers do not create rents in any obvious ways. Barriers due to language, customs, or product standards, for example, may contribute to border effects but do not generate rents. The deadweight loss from these barriers is much greater, they show, than from rent-creating barriers.
On the one hand, policy aimed at removing these barriers is likely to generate much higher welfare gains than reduction of rent-producing barriers. But on the other hand, these barriers are not so easily eliminated by policy choices. Anderson and van Wincoop focus on one perfect example of a nonrent barrier that policy can influence—the choice of currency. One of the greatest perceived economic advantages for Europe of the adoption of the euro is the increased ease of international transactions within Europe. Money's role is to coordinate buyers with sellers, and that role is more beneficial the larger the club that uses the money. Another area where policy might reduce nonrent barriers is the coordination of safety regulations. If AvW have accurately measured the size...
