The Importance of Trade Costs in Deviations from the Law-of-One-Price: Estimates Based on the Direction of Trade

Article excerpt

I. INTRODUCTION

We hold that economic theory places restrictions on Law-of-One-Price (LOP) deviations that are no less important than those placed on their changes. (1) At the same time, most of the prior macroeconomic literature on this topic had until recently typically focused on the time-series behavior of price indices rather than on cross-sectional price differences. It follows that closing the remaining gap between theory and empirics requires the use of product-specific price levels that enable exact comparisons across space. In line with this, Anderson and van Wincoop (2004) go on to propose the use of price levels comparable across locations at a point in time as a promising route for inferring trade costs, (2) and suggest that to achieve this "[a] natural strategy would be to identify the source country for each product," because "[u]nfortunately survey data often do not tell us which country produced the good."

In this article, we use microeconomic price levels, cross-sectional productivity indices, and bilateral trade flows between countries, in order to identify the likely source of each product. We then examine an empirical model where international price dispersion is determined by transport costs, local trade costs, taxes, and market size. Trade costs in the form of transportation and distribution costs are shown to be important in segmenting international markets.

Transport costs and broader trade costs are of central importance in many macroeconomic models, as in the recent examples of Bergin and Glick (2003) and Atkeson and Burstein (2007, 2008). However, assessing these at the macroeconomic level has proved problematic. For example, if one averages over a number of different goods exported and imported across a pair of countries, transport costs would have a tendency to cancel out. Thus, the impact of trade costs in segmenting individual product markets will typically be underestimated when considering aggregate (consumption-weighted average) prices or the simple average (over products) of price deviations. This is the case because countries are likely to both export and import some of the different goods that go into the construction of the aggregate or average composite price. As a result, the impact of transport costs on price differences would wash out on average even if these were important in segmenting markets as determinants of international price deviations for individual products. This is the "averaging-out property" put forth by Crucini, Telmer, and Zachariadis (2004). Here, we note that this can be going on even within narrow categories of products, a point we elaborate more on in the next paragraph.

Trade costs can also be mismeasured by price differences if the distance between two locations does not capture distance between exporter and importer. For example, if trade between two countries does not occur for a product, then that price difference will lie between the no-arbitrage bounds and will be less than the trade cost. On the other hand, if there is trade back and forth between two countries for the same category of goods, a feature which is not atypical of actual trade flows between countries, then the transportation cost included in both final goods prices will tend to cancel out when comparing the price for that category of goods across space. We address these concerns by utilizing information about the source of narrow product categories and including in our sample only those country pairs where we have information that trade takes place in a specific direction. A third related and largely unresolved concern arises if both countries export the same product to each other, so that the overall impact of trade costs on that product's price difference between the two locations can be zero even if these costs are large. That is, to the extent that the countries being compared export similar products to each other, the final price for such products incorporates about the same transportation cost in both countries so that there might be little or no impact of transportation costs on price differences for these products between the two countries. …