Europe Without Borders?
The Effect of the EMU on Market Segmentation
Hisham Foad
Department of Economics
Emory University
Atlanta, GA
December 21st, 2003
Abstract
Proponents of monetary union argue that establishing a single currency area will decrease market segmentation, lowering relative price volatility. Has this been the case? Numerous studies have shown that purchasing power parity either fails to hold or that deviations from PPP are glacially slow to revert to trend. Similar evidence has been found for the “law of one price,” the notion that the same good should sell for the same price in all locations. Given this, what factors determine market segmentation? A study by Engel and Rogers (1996) looked at how distance and the presence of national borders affected relative price volatility between cities in the U.S. and Canada, and found that crossing the border is equivalent to 1700 miles of distance between cities in the same country. This paper extends this analysis to cities in Western Europe and finds that both distance and national borders are key determinants of relative price volatility. The “width” of the border is not nearly as large, however, ranging from 20 – 70 miles of distance. Various explanations for the continued significance of the border effect (uneven sampling bias, idiosyncratic shocks, exchange rate volatility) are tested, but the border effect remains significant. Given this, the analysis then turns to how monetary union has reduced the distance and border effects. Although the evidence is mixed, the results do indicate that monetary union has decreased the border and distance effects for EMU members while it has increased these effects for non-members.
Keywords: Law of one price, economic integration, exchange rates, European Monetary Union
JEL F15, F40
* I would like to thank Robert Chirinko, Debdulal Mallick, Elena Pesavento, and seminar participants at the Emory University Macro Lunch series for their invaluable comments and help in this project. This is still very much a work in progress, and all errors within are my own. Please direct all inquiries to Hisham Foad at or 404-272-2982. All correspondence may be sent to Emory University, Department of Economics, 1602 Fishburne Dr, Atlanta, GA 30322.
Table of Contents
I. Introduction………………………………………………... 3
II. Literature Review…………………………………………. 6
III. The Data…………………………………………………… 11
IV. Theoretical Model…………………………………………. 13
V. The Border Effect
a. A Comparison to Engel and Rogers……………………. 17
b. Uneven Sampling Bias…………………………………. 20
c. Idiosyncratic Shocks…………………………………… 22
VI. The Impact of Monetary Union…………………………… 26
VII. A Role for Exchange Rates?………………………………. 31
VIII. Conclusion……………………………………………….. 34
IX. Appendices………………………………………………… 38
Europe Without Borders?
The Effect of the EMU on Market Segmentation
1. Introduction
Much has been made about the increased globalization of the international economy. Free trade zones such as NAFTA and currency unions such as the EMU have purportedly reduced the economic significance of national borders. With this, markets have become less segmented, with physical distance between locations becoming less of a factor. But is this really the case? If indeed it is, then we should not only see purchasing power parity hold between countries, but also the same good should sell for the same price in two different locations, the “law of one price.”
Many empirical studies have refuted the law of one price, or shown that relative price convergence is glacially slow, indicating that there is a significant amount of market segmentation across borders. Why is it that such an economically plausible theory does not seem to hold? Two culprits that emerge are market segmentation brought on by physical distance and the presence of national borders. Distance increases segmentation mainly through transportation costs, which limit the arbitrage opportunities that drive both theories. Furthermore, markets that are further apart tend to be less similar than those that are closer together, creating different demands and therefore prices for a good. Certainly, local market conditions make the demand for housing more inelastic in Japan than in Australia. National borders could be significant for several reasons. The presence of trade barriers such as tariffs would tend to limit arbitrage in the same way as transportation costs. Another possibility is that residents of one country may have an affinity for domestically produced goods, the “home bias” in trade[1]. Differential tax schemes across countries may also have an effect (the same good may receive a different tax treatment in one country than it does in another.) Finally, there may be national standards that create natural market segmentation. Such examples include right side steering wheels, 220-volt outlets, or a warranty that is void outside the country of purchase.
A seminal work by Engel and Rogers (1996) examined the role of distance and national borders in explaining these phenomena. They examined deviations from the law of one price for 14 categories of goods, looking at the relative prices between nine Canadian and 14 U.S. cities. They find that while both distance and the border are significant determinants of relative price volatility, crossing the border alone has the same effect as over 1,700 miles of distance between cities. Several explanations for this effect emerge, most notably incomplete pass through of exchange rates. Changes in nominal exchange rates are often the result of factors in financial markets and not in goods markets. These exchange rate changes are not completely passed through to nominal prices. Because prices are more rigid than nominal exchange rates, there is a natural tendency for cross border prices to be more volatile once they have been converted into a common currency. However, filtering out this effect only reduces (but does not eliminate) the border effect.
This paper is similar in spirit to the work by Engel and Rogers. Whereas their study focused on the U.S. and Canada, my work will examine prices in a large number of cities across Western Europe. The price variable is the daily per diem rate for two categories of goods published monthly by the U.S. State Department for 201 cities in 16 countries. As all prices are quoted in the same currency (U.S. Dollars), the problem of sticky prices/volatile exchange rates is mitigated[2]. Constructing several measure of relative price volatility, both distance and the border are significant determinants. Furthermore, the magnitude of the border effect relative to that of distance is very similar to that found by Engel and Rogers, indicating that markets in Europe are still quite segmented[3]. One might argue that monetary union should reduce this, and restricting the sample to the post-euro period does reduce both the distance and border effect, although these effects are still significant. The evidence indicates that the adoption of the euro has decreased market segmentation when comparing markets in the euro zone to those outside it.
The basic issue examined in this paper is the role of national borders in segmenting markets. If borders were irrelevant, we would expect that the only impediment to fully integrated markets would be transportation costs. One way to examine this issue is to look at price levels in given markets. Figure 1 shows a relative price index for two pairs of cities, one pair separated by a border and the other pair within the same country[4]. If the border were not an issue, we would expect that the pair of cities located closer together would exhibit a greater degree of price stability than the pair located further apart. In this example, Hamburg is 366km (229 miles) away from Amsterdam and 611km (382 miles) from Munich. Thus, the relative price index between Hamburg and Amsterdam should exhibit at least the same stability as the price index between Hamburg and Munich. As is evident from Figure 1, however, the Amsterdam to Hamburg relative price exhibits greater volatility than the Munich to Hamburg price. Interestingly, this difference is volatility is reduced when restricting attention to dates after 1999, perhaps indicating the impact of monetary union on reducing border effects.
2. Literature Review on the Failure of LOP and the PPP puzzle
One of the most basic and fundamental theorems of economics is the law of one price (LOP). Once prices are converted to a common currency, the same good should sell for the same price regardless of where it is being sold. Otherwise, an arbitrage opportunity would exist and the resulting market forces would lead to price convergence. Given the logic of the LOP, it is a surprising result that most empirical evidence has refuted this theory. Furthermore, numerous studies have documented the failure of purchasing power parity (PPP), the aggregate corollary to the law of one price. PPP states that the exchange rate should adjust to equate national price levels. In other words, the following relation should hold:
In other words, the real exchange rate should be equal to one. A less stringent theory is that of relative purchasing power parity. This requires only that growth in the exchange rate offset the differential between the rate of growth in home and foreign aggregate prices. Prices may differ across locations, but the exchange rate adjusts so that they move together. This weaker version of PPP implies the following relation:
The real exchange rate need not be equal to one, but it must be constant. Any movements in the real exchange rate are synonymous with deviations from PPP. Two of the earliest works examining the failure of PPP – Samuelson (1964) and Balassa (1964) - looked at the dichotomy between traded and non-traded goods. They argued that PPP fails because the general price level has both traded and non-traded components. The arbitrage opportunities necessary for PPP exist in the traded goods sector, but since non-tradables are difficult to arbitrage, we will not see PPP in this sector. Consider the following two-country example. Let the general price level P be decomposed into tradable and non-tradable components:
Where α is the share of tradables in the price index. The foreign price index can be written as:
Let s be the log of the nominal exchange rate. The log real exchange rate can be written as:
If PPP holds among traded goods, the first term in the above equation will be constant. Deviations from PPP are entirely the result of relative price movements within countries. Furthermore, the correlation between the aggregate real exchange rate and that of traded goods will be small since the tradable real exchange rate displays little variation (as PPP holds for tradables.)
A necessary condition for relative PPP to hold is that the real exchange rate be stationary. Otherwise, any deviation from PPP would be permanent and convergence would not occur. Numerous studies have shown that real exchange rates follow a random walk, indicating no tendency toward purchasing power parity[5]. In fact, the failure to reject the null hypothesis of a unit root has been one of the defining features of this literature[6]. Some studies have shown mean reversion in real exchange rates, but with a half-life of between three to seven years, hardly keeping pace with fluctuations in nominal exchange rates or prices for that matter[7]. Another theory is that within a certain band of deviation from PPP, transportation costs limit arbitrage and there is no tendency for revert to PPP. Outside this band, the benefits of arbitrage outweigh the transportation costs, and we observe convergence[8].
Estimates of relative PPP are highly sensitive to the base year chosen. Rogoff (1996) gives the example of the appreciation of Mexico’s real exchange rate during the early 1990’s. Given this information, one might have correctly concluded that the peso was overvalued, leading to its collapse in 1994. However, during Mexico’s debt crisis in the 1980’s, the real value of the peso fell sharply. Thus, depending on the base year chosen for comparison, one could draw two vastly different conclusions about the true valuation of the peso.
Though not as celebrated as the PPP puzzle, there have been many significant contributions documenting the failure of the law of one price. Among the first was Isard (1977), who examined data on U.S., German, Canadian, and Japanese exports for a range of highly traded goods. Deviations from LOP were large and persistent. He concluded that these deviations reflected nominal exchange rate movements. A key assumption regarding LOP is that the goods in question really are identical across locations. If not, there will be imperfect arbitrage (depending on the degree of substitutability between the goods), and LOP will be less likely to hold. Therefore, we could expect homogenous goods to be more uniformly priced than more differentiated goods. This relates to the tradable vs. non-tradable dichotomy in the Samuelson-Balassa theory. However, Giovannini (1988) found sharp differentials even in highly tradable commodity manufactures (screws, nuts, bolts, etc.) and found that LOP deviations were highly correlated with exchange rate movements. Knetter (1989, 1993) examined 7-digit export unit values from a single source to multiple destinations and found significant evidence of pricing to market. He found for example, large and volatile differences in the price of German beer shipped to the United States as opposed to that shipped to the U.K. Rogers and Jenkins (1995) perform unit root tests on the real exchange rate defined by 54 different commodities for the U.S. and Canada. They are only able to reject a unit root in one-sixth of the cases, mostly for highly tradable goods. This would suggest that while the Samuelson-Balassa models may be logically correct, they are fairly limited in scope. Enforcing this claim, Jenkins (1997) found that while the tradability of a good has an effect on the persistence of LOP deviations, it does not seem to influence the likelihood of rejecting a unit root null for real exchange rates.