ADBI Working Paper Series

International Price Dispersion and Market Segmentation in Japan and the United States: Theory and Empirics

K.C. Fung, Alicia GarciaHerrero, and Francis Ng

No. 417 April 2013

Asian Development Bank Institute

K.C. Fung is professor of economics at the Department of Economics, University of California, Santa Cruz, CA. Alicia Garcia-Herrero is chief economist for emerging markets at BBVA in Hong Kong, China. Francis Ng is senior economist of the International Trade and Integration Team in the Development Economic Research Group of the World Bank, Washington, DC. The authors benefitted from useful comments by two anonymous referees, a co-editor and a researcher at ADB. Suggestions from Maria Socorro G. Bautista were particularly relevant. They are also grateful to Katelyn Fung for her constructive editing and appreciate insights and data provided by the staff and researchers at the World Bank, US Department of Commerce, US Justice Department, Japan Fair Trade Commission, and Japan Economic Planning Agency. The authors are grateful for research and funding support from the World Bank, ADB, and BBVA. Views contained in this paper reflect solely the views of the authors and not those of the World Bank or BBVA. The views expressed in this paper are the views of the author and do not necessarily reflect the views or policies of ADBI, the ADB, its Board of Directors, or the governments they represent. ADBI does not guarantee the accuracy of the data included in this paper and accepts no responsibility for any consequences of their use. Terminology used may not necessarily be consistent with ADB official terms. The Working Paper series is a continuation of the formerly named Discussion Paper series; the numbering of the papers continued without interruption or change. ADBIs working papers reflect initial ideas on a topic and are posted online for discussion. ADBI encourages readers to post their comments on the main page for each working paper (given in the citation below). Some working papers may develop into other forms of publication. Suggested citation: Fung, K.C., A. Garcia-Herrero, and F. Ng. 2013. International Price Dispersion and Market Segmentation in Japan and the United States: Theory and Empirics. ADBI Working Paper 417. Tokyo: Asian Development Bank Institute. Available: http://www.adbi.org/workingpaper/2013/04/19/5602.price.dispersion.market.segmentation.jp.us/ Please contact the authors for information about this paper. Email: alicia.garcia-herrero@bbva.com.hk; kcfung@ucsc.edu Asian Development Bank Institute Kasumigaseki Building 8F 3-2-5 Kasumigaseki, Chiyoda-ku Tokyo 100-6008, Japan Tel: Fax: URL: E-mail: +81-3-3593-5500 +81-3-3593-5571 www.adbi.org info@adbi.org

2013 Asian Development Bank Institute

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

Abstract

This paper focuses on the pricing behavior of Japanese and United States firms selling their identical products in New York City, Chicago, Osaka, and Tokyo. The authors utilize some
simple models of international price dispersion and market segmentation that generate predictions about testable prices. The dataset, which consists of prices of identical products in the Japanese and American cities, was collected and accepted by both governments. Using this data, versions of international price dispersion theories are tested and some empirical evidence to support the view that simple international price dispersion models can partly explain the observed prices is found. JEL Classification: F12, F14, L11, L13

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

Contents
1. 2. 3. 4. Introduction ..................................................................................................................... 3 A Simple Stylized Model of International Price Dispersion and Market Segmentation ..... 4 An Expanded Model of International Price Dispersion and Market Segmentation ........... 9 Conclusion .....................................................................................................................12

References ...............................................................................................................................14 Appendix ...................................................................................................................................16

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

1. INTRODUCTION
There have been many theoretical models yielding observations of differing prices for identical products. In this paper we focus on two related strands of literature: international price dispersion and market segmentation. 1 Many of the international market segmentation papers incorporate theories that feature firms with market powers operating in different international markets (see e.g., Bagwell and Staiger 2009; Aminian, Fung, Garcia-Herrero, and Lin 2012; Helpman and Krugman 1985, 1989; Bernhofen 1999; Christos, Clerides, Ioannou, and Senft 2007; etc.). These papers tend to use international oligopolistic models to analyze international trade patterns and the effects of trade policies. Some of these papers exhibit features associated with international price differences. In contrast, in the literature of price dispersion, Belleflamme and Peitz (2010) provides us with a clear and recent discussion on the phenomenon of price dispersion, a notion related to whether the law of one price holds. In Engel and Rogers (2004), a dataset of prices of consumer goods across European cities from 1990 to 2003 is collected and the paper tests if the introduction of the euro increases the integration of European markets. The authors conclude that there is no evidence of price convergence. In Bayes et al. (2006), firm and pricing information are collected from online sites in seven countries (four in the eurozone and three outside). The paper concludes that the introduction of the euro raises eurozone prices by at least 3%, and that with a currency union, online pricing seems to act like a clearinghouse. In Goldberg and Verboven (2005), detailed automobile prices are collected in five European countries from 1970 to 2000, a period during which the policies associated with the European Single Market were intensifying. The paper examines versions of the law of one price and concludes that contrary to the existing literature, there is surprising evidence indicating the convergence of prices. Finally, in Maier (2009), prices of homogenous goods auctioned on eBay across euro countries and the United Kingdom are collected; the paper finds that in countries with the same currency, price variations are smaller. The models we will use here do exhibit features of price dispersion and market segmentation. But our paper differs in focus from these two strands of literature. As previously mentioned, in the international market segmentation literature, the models are utilized mainly to study the trade pattern and the impact of trade policies in imperfectly competitive markets. For our focus, we adopt versions of a simple, stylized oligopolistic model of international market segmentation to see if they can explain the observed prices across markets for almost identical products. In the price dispersion literature, the main interest seems to be on examining the law of one price and the rate at which prices tend to converge. Another major concern of this literature is how policies such as the creation of a currency union (introduction of the euro) or a Single Market Program can affect the convergence of prices. In this paper, our focus is not the law of one price. We simply want to ask if the observed prices are consistent with the predictions of a popular, stylized model of international oligopoly that sells across different markets. One further potential contribution of our work is that our data are neither from different countries within the European Union (even though European markets clearly provide a natural experiment to study market integration and the law of one price), nor from different states within an integrated United States (US). Instead, our data consists of prices from four cities in two different major Pacific economies: Japan and the US. Furthermore, these surveyed price data are jointly collected and publicly accepted as valid by both governments. One difficulty of empirically examining price variations across different countries such as Japan and the US is that the usual observations are often associated with highly
1

We are indebted to a referee for pointing out the relationship of our paper to the price dispersion literature and for suggesting a better review of the price dispersion literature.

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

differentiated products. Even for products that have very similar characteristics, one is unsure if the data of prices collected are from expensive department stores (stores that inflate prices due to the brands of the department stores, location, etc.), from discount stores or other outlets. In other words, the researchers cannot be sure if the observed outcomes are due to international pricing behavior by oligopolistic firms or to products having different attributes (including different locational or geographic characteristics). In this paper, the dataset we will use provides us with prices of products that are as close to being identical as possible, and sold in different cities and different markets in Japan and in the US. These surveyed prices are for goods of the same brands, and for the same volumes and sizes. Retail prices were also collected according to the types of stores at which they were sold (supermarkets, discount stores, specialty stores). These data are highly suitable to test price dispersion and international market segmentation models. 2 In the next section, we will present a stylized model of international price dispersion and derive a testable hypothesis associated with the model. Econometric tests will then be conducted to see if the predictions of the models can be rejected. From the data, we observe price differences of identical products in different international markets. Initially, it appears that there are no particular reasons why these different prices in Japan and in the US should be compatible with some simple stylized price dispersion models. However, there does indeed seem to be some empirical support. In section 3, we will present our empirical results. In the last section, we conclude.

2. A SIMPLE STYLIZED MODEL OF INTERNATIONAL PRICE DISPERSION AND MARKET SEGMENTATION

We first assume there are two international firms, one in Japan and one in the US 3, each selling an identical product at home and abroad. They compete as duopolistic quantitysetting firms. They have identical constant marginal cost c i . The retail price of good i in the domestic market is p i , and the price of the same good in the foreign market is p* i . For prices generated by US producers, p i is the price observed in the US market, while p* i is the price in Japan. Conversely, for prices generated by Japanese manufacturers, p i is the price observed in Japan, while the foreign price p* i is the price in the US. To get good i from the domestic factory to the retail store in the domestic market, an ad valorem distribution cost of d i must be incurred. To get good i from the factory of the domestic market to the foreign retail store, an ad valorem international transport cost of tr* i must first be paid. 4 When it
2

However, we do not have data on individual firms. We simply know that US firms were selling their almost identical goods across markets and cities, and Japanese firms were selling their almost identical products across Japan and the US. We are grateful to a referee for suggesting the use of a matched-group econometric technique. In future work, we will attempt to collect firm-level data as well as price data, and adopt the more appropriate econometrics techniques.
3

We will extend the assumption of having only two international firms to multiple firms in Section 3.

As pointed out by a referee, we are making an assumption that transport costs are ad valorem instead of being specific. This has the implication that as the price of the product increases, the absolute amount of the transport cost rises. This can partly be explained by the observation that as the product becomes more expensive, producers may want to purchase more insurance, raising the effective transport cost (or cost, insurance and freight, c.i.f.). In addition, some more expensive items such as high quality computer chips may be better transported via air transport, which would also increase transportation costs.

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

arrives at the foreign shore, the good faces an ad valorem tariff rate of t* i . Then, before the good ends up on the shelf of the foreign retail store, a foreign ad valorem distribution cost of d* i is incurred. We can similarly trace how a good originating in the foreign country ends up on the shelves of the foreign retail store (a foreign distribution cost of d* i has to be paid) and in the domestic retail stores (international transport cost tr i , tariff rate t i and domestic distribution cost of d i must be incurred). With these various costs in mind, we can define the profit functions of the two international firms, one Japanese and one American, as: H i = x i p i /(1 + d i ) c i (x i + x* i ) + x* i p* i /(1 + d* i )(1 + t* i )(1 + tr* i ) H* i = y i p i /(1 + tr i )(1 + t i )(1 + d i ) + y* i p* i /(1 + d* i ) c i (y i + y* i ) (2) where x i is the output of the domestic firm for the domestic market, x* i is the output of the domestic firm for the foreign market, y i is the output of the foreign firm in the domestic market, and y* i is the output of the foreign firm in the foreign market. Market segmentation and price dispersion imply that each firm chooses its output in each market separately. From the profit maximization first order conditions, the retail price of product i in the domestic market is: p i = [e i /(2e i 1)]c i (1 + d i )[(1 + tr i )(1 + t i ) + 1] (3) where e i is the domestic elasticity of demand and c i is the marginal cost of production. Note that to determine the domestic retail price of a product manufactured domestically, the domestic tariff rates imposed by the domestic country t i and the international transport costs to the domestic country tr i enter in the expression for p i . This reflects the interdependence between the Japanese and US firms, with x i and y i (which has to incur t i and tr i in order to sell in the domestic market) competing in the domestic market. Similarly, the firms maximize profits in the foreign market, generating an expression for p* i : p* i = [e* i /(2e* i 1)]c i (1 + d* i )[1 + (1 + t* i )(1 + tr* i )] (4) with e* i being the foreign demand elasticity. Using the retail prices of the same good sold in Japan and the US, we have: p* i /p i = {[e* i /(2e* i 1)](1 + d* i )[1 + (1 + t* i )(1 + tr* i )]}/ {[e i /(2e i 1)](1 + d i )[1 + (1 + t i )(1 + tr i )]} (5) Note that the retail price ratio depends on both countries distribution costs, both countries elasticities of demand, tariff rates imposed by both countries, and international transport costs to and from the domestic or foreign country. Recall that the domestic price is the domestic retail price of the product manufactured in the domestic country, and the foreign price is the retail price of the domestically manufactured good. Intuitively, the price ratio should include domestic distribution costs, tariffs imposed by the foreign country and international transport costs from the domestic country to the foreign country. However, the fact that the price ratio is also determined by the tariffs imposed by the domestic country and the international transport costs to the domestic country is a special feature of this stylized international oligopolistic price dispersion and market segmentation model. Taking logs on both sides, the regression equation is given by: ln (p* i /p i ) = a + b 1 ln [(1 + d* i )/(1 + d i )] + b 2 ln {[1 + (1 + t* i )(1 + tr* i )]/[1 + (1 + t i )(1 + tr i )]} + b 3 ln {[(2 1/e i )/(2 1/e* i )]} + u i (6) where u i is the disturbance term. To test the international price dispersion and market segmentation model, we subsume the positive markup ratio (the term involving the price elasticities) in the intercept. Then, to test for the international price dispersion model is to test: 5 (1)

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

a>0, b 1 = 1 and b 2 = 1 (7) In addition, we can conduct a joint F-test for the model. 5 Alternatively, we can also divide the products into product groups, with category-specific intercepts D 1 , D 2 , D 3 and D 4 and allow the markup ratios of different product groups to differ, in which case, we are testing: D 1 >0, D 2 >0, D 3 >0, D 4 >0, b 1 =1, b 2 =1. The regression equation can be written as: ln (p* i /p i ) = D i + b 1 ln [(1 + d* i )/(1 + d i )] + b 2 ln {[1 + (1 + t* i )(1 + tr* i )]/[1 + (1 + t i )(1 + tr i )]} + u i (8) where the markups have been subsumed into D i (i = 1, 2, 3,4), corresponding to the four different product groups. We next turn to the data, which includes two price surveys jointly conducted by the staff of the International Trade Administration of the US Department of Commerce (DOC) and the staff of the former Ministry of Trade and Industry (MITI, now METI) of Japan. The surveys also received some help from other Japanese ministries such as the Ministry of Agriculture. The objective of the surveys is to observe how identical goods (or goods that are as close to identical as possible) are priced in Japan and in the US. The surveys cover a wide range of goods, including consumer goods, capital goods, and services. There are, however, only two years of data: The first survey was conducted in October 1989 and the second survey in April 1991. A set of prices were identified as generated by US manufacturers, while other prices were generated by Japanese producers. The results of the surveys were publicly accepted as valid by both governments. Although these data are not new, they have not yet been used to test models of price dispersion and market segmentation. Furthermore, our theories of international price dispersion are abstract, timeless models that should apply to useful data surveyed from a different decade. For the 1989 survey, 121 products (both consumer and capital goods) and 18 services were examined; for the 1991 survey, the corresponding figures were 112 products and 14 services. In this paper, we ignore the observations about the services, as they tend to be non-tradables in the 1990s and may not be suitable for testing our international price dispersion models. The surveys were conducted in four cities: Tokyo, Osaka, Chicago, and New York City. The retail prices of identical goods (controlling for brand names, volume, and/or size) in a variety of outlets (including supermarkets, discount stores, and specialty stores) were surveyed. Each survey was conducted over a two-week period. Prices of the products observed in Japan were given in yen, while prices of products observed in the US were given in US dollars. The averages of the prevailing exchange rates during the survey periods were used to convert the prices from yen to US dollars. For both consumer and capital goods, the objective was to obtain actual prices to the end-users. For consumer and electronic goods, the prices observed in the stores were used. For autos, interviews were conducted to ascertain the actual selling prices. For capital goods, distributors and manufacturers sales representatives were interviewed to determine the discounted prices. The surveys also indicate and differentiate prices generated by US manufacturers versus those generated by Japanese manufacturers. To implement the tests of the model, we also collected data on ad valorem distribution costs, ad valorem international transport costs, and ad valorem tariff rates. Different surveyed prices are also identified as belonging to different sectors or subsectors (e.g., as auto parts or perfume or tobacco). We identify similar subsectors using trade data, tariff lines, and input-output tables. These costs are then matched up with the different product prices. It should be noted that the distribution costs are taken from the US and Japan input-output tables, which are classified by fairly aggregate
5

We are indebted to a referee for suggesting the additional joint F-test to test the model.

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

sectors and are not in sufficient details to match the product prices. Because of this, the distribution costs may be less accurate than other data used in this paper. The definitions and the sources of all the relevant data are listed in the appendix. The following tables provide some summary statistics. Table 1 highlights summary statistics concerning retail prices in the domestic market relative to those in the foreign market. Recall that the domestic retail price p i is the retail price of good is observed in the same country where it is manufactured, and the foreign retail price p* i is the retail price of good i observed in a country different from where it is manufactured. For example, the retail price of a made-in-Japan good in Japan is an observation p i , while the retail price of the same good sold in the US is an observation p *i . The (p i /p* i ) ratios vary by product groups and by whether Japan or the US is the domestic market. Table 2 and 3 provide summary statistics for the independent explanatory variables, including ad valorem domestic distribution costs, ad valorem international transport costs, and ad valorem tariff rates. Table 4 shows the results of our regressions in testing the international price dispersion and market segmentation model.

Table 1: Summary Statistics of the Dependent Variable: Ratios of Domestic Retail Prices to Foreign Retail Prices
Average Domestic Price/Foreign Price (p i /p* i ) by Product Groups Auto Parts Autos Capital Goods Electronic and Optical Goods Food Products Liquor Other Consumption Goods Ratio of Prices with Products Made in the US as Domestic Prices p i (1989) 0.6198 0.6793 0.6930 0.7284 0.6082 0.3574 0.6305 Ratio of Prices with Products Made in Japan as Domestic Prices p i (1989) 1.5715 0.9977 1.0282 0.9611 0.4480 0.5415 0.5414 Ratio of Prices with Products Made in the US as Domestic Prices p i (1991) 0.4641 0.7364 0.6409 0.8961 0.5876 0.6826 0.7804 Ratio of Prices with Products Made in Japan as Domestic Prices p i (1991) 1.2947 0.8595 0.9646 0.9222 0.4050 1.4545 0.9655

Source: Authors' calculations and estimations.

Table 2: Summary Statistics for Independent Variables for the Japanese Market
Ad valorem Distribution Cost 0.2180 0.1066 0.6790 0.0496 Ad valorem Transport Cost 0.0569 0.0380 0.2100 0.0000 Ad valorem Tariff 0.0305 0.0584 0.2672 0.0002

Mean Standard Deviation Maximum Minimum

Source: Authors' calculations and estimations.

Table 3: Summary Statistics for Independent Variables for the US Market

Ad valorem Distribution Cost 0.00698 0.00711 0.03496 0.00003 Ad valorem Transport Cost 0.0904 0.1076 0.6309 0.0000 Ad valorem Tariff 0.0841 0.1117 0.9600 0.0078

Mean Standard Deviation Maximum Minimum

Source: Authors' calculations and estimations.

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

Table 4: Testing the Basic International Price Dispersion and Market Segmentation Model
All Products, Both Years 0.209 (7.688) No 0.669 (-3.101) Yes 1.593 (0.681) No 5.303 Yes 0.188 168 Japanese Products Only -0.154 (-1.987) Yes -1.065 (-5.696) Yes 1.202 (0.146) No 16.382 Yes 0.074 87 US Products Only 0.330 (3.787) No 0.371 (-2.087) Yes 2.329 (1.285) No 3.075 No 0.050 81 All Products, 1989 0.213 (5.449) No 0.770 (-1.388) No 1.332 (0.258) No 0.982 No 0.201 85 All Products, 1991 0.207 (5.360) No 0.592 (-2.844) Yes 1.676 (0.554) No 4.846 Yes 0.156 83

Reject a>0? b1 Reject b 1 =1? b2 Reject b 2 =1 Joint FStatistics Reject F-Test? Adjusted R Squared Number of Observations

Note: The numbers in parentheses are t-statistics. For b 1 and b 2 , the t-statistics are for the hypothesis b 1 =1 and b 2 = 1 Source: Authors' calculations and estimations.

Table 4 shows that at least for some cases, our empirical tests cannot reject the predictions of our stylized international price dispersion and market segmentation model. For the intercept term, almost all the hypothesis testing (except for the one using data with Japanese products only) conforms to the model. Tests involving the coefficient on the distribution margins (b 1 ) are less successful. All, except the one involving data from 1989 only, reject the price dispersion model. However, for the coefficient involving tariffs and transport costs (b 2 ), the price dispersion model does very well. None of the tests can reject the prediction of the model. As discussed earlier, the term b 2 is a special feature of the stylized oligopoly model and it is noteworthy that b 2 =1 is statistically significant in all tests. The results of the joint Ftest are mixed, with two out of five tests not able to reject the model. The adjusted R squared is also low. Yet overall there is some evidence supportive of the international price dispersion model. The model is strongest for two cases: data with US products only (column 4) and data for 1989 only (column 5). In the former case, all but one of the predictions are consistent with the tests, while in the latter, none of our tests can reject any prediction of the basic international price dispersion and market segmentation model. We interpret results from the above table to mean that there is at least mild evidence in support of the international price dispersion and market segmentation model. In tests so far, we have imposed the restriction that all products have the same constant intercept. We next want to relax this assumption. To do so, we divide the products into four groups: consumer and household electronics (group 1); automotive and machinery (group 2); consumer goods and miscellaneous (group 3); and food, liquor, and tobacco (group 4). The next table shows the results of our tests. By examining Table 5, we see that very few tests can reject the predictions that the intercept terms of different product groups are positive (D 1 >0, D 2 >0, D 3 >0, D 4 >0). For the coefficient on the term with the distribution margins (b 1 ), however, the prediction of the basic international price dispersion model is uniformly rejected. On the other hand, the prediction that b 2 =1 is almost uniformly consistent. Again, the fact that the term involving b 2 =1 cannot 8

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

be rejected provides an important piece of evidence in favor of the model. In general, there is at least some evidence in favor of the basic model with different product group intercepts. This is true for the Ds and for b 2 . Indeed, for the case of US products only, all the predictions except one cannot be rejected by the empirical tests. Furthermore, the F-test cannot reject the joint test of the model. We interpret the tests for the case with different product groups to indicate that while the results are certainly mixed, there is nonetheless some evidence in favor of the basic model.

Table 5: Testing the Basic International Price Dispersion and Market Segmentation Model with Different Product Groups
All Products, Both Years 0.151 (3.251) No 0.130 (2.849) No 0.338 (4.738) No 0.350 (5.456) No 0.488 (-4.159) Yes 1.427 (0.482) No 8.976 Japanese Products Only -0.060 (-0.745) Yes -0.186 (-1.969) Yes 0.247 (1.088) Yes -0.209 (-1.368) Yes -0.870 (-4.662) Yes -0.253 (-2.781) Yes 14.464 US Products Only 0.275 (0.139) No 0.299 (2.813) No 0.368 (2.673) No 0.450 (3.423) No 0.243 (-2.159) Yes 1.207 (0.139) No 2.388 All Products 1989 0.124 (1.843) No 0.104 (1.405) Yes 0.340 (3.368) No 0.415 (4.719) No 0.480 (-2.620) Yes 1.659 (0.516) No 3.443 All Products 1991 0.176 (2.689) No 0.152 (2.273) No 0.340 (3.273) No 0.273 (2.718) No 0.452 (-3.139) Yes 1.469 (0.347) No 5.434

D1 Reject D 1 >0 D2 Reject D 2 >0 D3 Reject D 3 >0? D4 Reject D 4 >0? b1 Reject b 1 =1? b2 Reject b 2 =1? Joint Fstatistics Reject F-Test? Adjusted R Squared Number of Observations

Yes 0.224 168

Yes 0.122 87

No 0.033 81

Yes 0.254 85

Yes 0.153 83

Note: The numbers in parentheses are t-statistics. For b 1 and b 2 , the t-statistics are for the null hypothesis of b 1 =1, b 2 =1, respectively. Source: Authors' calculations and estimations.

3. AN EXPANDED MODEL OF INTERNATIONAL PRICE DISPERSION AND MARKET SEGMENTATION

In Section 2, we use a simple, stylized model of international price dispersion and test its predictions. The basic model assumes that we have two international quantity-setting firms, one Japanese and one American. In this section, we adopt a more expanded model and examine the case with n identical firms in one country and n* identical firms in the other country. All firms produce an identical product using the same constant marginal costs, 6 and 9

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

each firm sells both at home and abroad. Domestic firm i produces x i for the domestic market and x* i for the foreign market, while foreign firm i produces y i for the domestic market and y* i for the foreign market. Following the same derivations as before, the ratio of the retail prices of identical products at home and abroad is given by: p* i /p i = {[e* i /((n + n*)e* i 1)](1 + d* i )[n* + n(1 + t* i )(1 + tr* i )]}/ {[e i /((n + n*)e i 1)](1 + d i )[n + n*(1 + t i )(1 + tr i )]} (9) Note again that the domestic tariff rates and the international transport costs to the domestic market are included in the expression for p i . But compared to the similar expression for the basic international price dispersion and market segmentation model, the term involving t i and tr i also involves n and n*. Incorporating the number of firms and testing the coefficient on this term is potentially a way to distinguish between the basic and the expanded version of the model. Thus the regression to be run is: ln (p* i /p i ) = a + b 1 ln [(1 + d* i )/(1 + d i )] + b 2 ln {[n* + n(1 + t* i )(1 + tr* i )]/[n + n*(1 + t i )(1 + tr i )]} + u i (10) where we again subsume our positive markups into the intercept term. To test the expanded version of international market segmentation, we test the hypothesis that a>0, b 1 =1 and b 2 =1. The table below shows the results. Our testing of the expanded version of the international market segmentation model exhibits results that are consistent with those from testing the basic model. The prediction for a>0 is almost always supported (row 1). Results for testing b 1 =1 and b 2 =1 are more mixed. Since the coefficient associated with the term containing tariff rates and transport costs (b 2 ) is what distinguishes the basic model from the expanded one, we can loosely interpret the results from Table 4 and Table 6 to mean that the basic model performs better. In the basic price dispersion model, we cannot reject b 2 =1 in all cases, whereas in the expanded model, b 2 conforms to the prediction in two out of five cases. Nonetheless, in two cases, US products only and All products, 1989 (columns 4 and 5, respectively), we cannot reject the F-test or the two separate tests b 1 =1 and b 2 =1. In fact, even though the R2 tends to be low, for these cases, none of the predictions of the expanded international market segmentation can be rejected. In both the basic and expanded versions of the model, tests using the subsample US products only yield a greater degree of support of the international price dispersion model than tests using Japanese products only. This may be due to the different perceptions of market segmentation by Japanese and US producers in the 1990s. Perceptions of market segmentation form an integral part of the international price dispersion models (Helpman 1984, Fung 1991a). For instance, it is often perceived that Japanese firms have stable, exclusive and long-term relationships with their distributors (Fung 1991b, Fung 2002), and in the past, US producers often had the impression that they could not easily find suitable Japanese distributors to carry their products in Japan. If US firms felt that the Japanese market was difficult to penetrate for institutional or structural reasons, they might regard the Japanese market and the US market as segmented. If Japanese firms felt that they had access to both group-affiliated distributors in Japan and the arms-length, competitive distributors in the US, they would be less likely to consider the two markets segmented. Thus, the different perceptions about market segmentations may explain why the international price dispersion model works better if we use the subsample US products only. Next, we turn to testing the expanded model by allowing for different product groups. Tables 6 and 7 show that in a majority of cases, we cannot reject D 1 >0, D 2 >0, D 3 >0, or D 4 >0. For tests involving All products, both years, US Products Only, and All Products, 1991, all the product groups have positive intercepts. With All Products, 1989, we can only 10

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

reject the hypothesis that D i >0 for product group 2 (D 2 ). Results involving b 1 and b 2 are more mixed. For the case of US products only, all the predictions of the model conform to the tests. For the case of All products, 1989, we cannot reject the hypothesis b 2 =1. Support for the expanded version of the international price dispersion model seems weaker, but again, when we use data for US product prices only, none of the predictions of the model can be rejected.

Table 6: Testing the Expanded International Price Dispersion and Market Segmentation Model
All Products, Both Years 0.208 (7.371) No 0.638 (-3.064) Yes 0.110 (-3.228) Yes 7.587 Japanese Products Only -0.132 (-1.716) Yes -1.025 (-5.599) Yes -0.130 (-2.463) Yes 19.134 US Products Only 0.313 (3.174) No 0.457 (-1.638) No 0.391 (-1.787) No 2.730 All Products, 1989 0.204 (5.024) No 0.783 (-1.188) No 0.252 (-1.696) No 1.983 All Products, 1991 0.214 (5.362) No 0.522 (-3.072) Yes -0.018 (-2.600) Yes 6.292

Reject a>0? b1 Reject b 1 =1? b2 Reject b 2 =1? Joint Fstatistics Reject F-Test? Adjusted R Squared Number of Observations

Yes 0.156 157

Yes 0.087 87

No 0.014 70

No 0.176 80

Yes 0.120 77

Note: numbers in parenthesis are t-statistics. Those for b 1 and b 2 are t-statistics for b 1 =1 and b 2 =1. Source: Authors' calculations and estimations.

11

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

Table 7: Testing the Expanded International Price Dispersion and Market Segmentation Model with Different Product Groups
All Products, Both Years 0.151 (3.420) No 0.108 (2.078) No 0.333 (4.561) No 0.377 (5.238) No 0.450 (-3.999) Yes 0.132 (-3.220) Yes 10.635 Japanese Products Only -0.060 (-0.745) Yes -0.186 (-1.969) Yes 0.247 (1.088) Yes -0.209 (-1.388) Yes -0.870 (-4.662) Yes -0.253 (-2.781) Yes 14.484 US Products Only 0.274 (1.972) No 0.243 (2.157) No 0.352 (2.400) No 0.528 (3.936) No 0.287 (-1.894) No 0.354 (-1.927) No 3.294 All Products, 1989 0.134 (1.986) No 0.079 (1.018) Yes 0.325 (3.196) No 0.437 (4.406) No 0.529 (-2.245) Yes 0.269 (-1.925) No 3.488 All Products, 1991 0.191 (2.818) No 0.136 (1.572) No 0.344 (3.163) No 0.301 (2.783) No 0.383 (-3.291) Yes -0.017 (-2.573) Yes 7.136

D1 Reject d 1 >0? D2 Reject d 2 >0? D3 Reject d 3 >0? D4 Reject d 4 >0? b1 Reject b 1 =1? b2 Reject b 2 =1? Joint Fstatistics Reject F-Test? Adjusted R Squared Number of Observations

Yes 0.200 157

Yes 0.120 87

Yes 0.052 70

Yes 0.237 80

Yes 0.121 77

Note: numbers in parenthesis are t-statistics. Those for b 1 and b 2 are t-statistics for b 1 =1 and b 2 =1. Source: Authors' calculations and estimations.

4. CONCLUSION
In this paper we examine different stylized versions of international price dispersion and market segmentation models, and we derive specific testable predictions of these models. To test the models, we utilize a dataset collected and publicly accepted by both the Japanese and US governments. The dataset consists of surveyed prices of individual, identical products in both Japanese and US cities. The retail prices of the goods were obtained by controlling for brand names, volumes, and sizes, as well as the types of stores in which they were sold. Thus we have observations of retail prices of products across markets that can be viewed as identical. For both goods manufactured by US producers and goods manufactured by Japanese producers, we are using prices of products that are as close to being identical as possible to test the predictions of models of international oligopolistic market segmentation and price dispersion. Our tests found evidence for both the basic international price dispersion model as well as the expanded model. The theories seem to have very strong support when we focus on the use of prices of US products. There also seems to be stronger support for the basic model 12

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

as compared with the expanded model. Thus a more stylized, simple and parsimonious model of international oligopolistic price dispersion may also is of greater empirical relevance. We believe that our study is important as it is done by using consistent micro data surveyed by both the Japanese and the US governments. There are several broad policy implications from our analysis. While Japan is indeed a technologically advanced high-income country, the last two decades demonstrate clearly that Japan will need more revitalizations and more liberalizations, particularly in its distribution and service sectors. As such, the international price dispersion highlighted in this paper is just a symptom of the much larger underlying problem facing Japan. How can Japan become even more competitive and generate more growth? Opening up to more international competition (particularly in services) and embracing globalization, including inviting more foreign direct investment and sending more students abroad to study, will seem a logical way for Japan to plot a path to reclaim some of the lost advantages that Japan used to have in the 1980s and in the early 1990s. International price differences seem to be a symptom associated with a traditional, stagnating Japan. Instead we believe that the world and in particular Asia need a stronger, more vibrant, more open and more competitive Japanese economy. Linking the economy of Japan with that of the US and other Pacific countries more closely will generate higher productivities. Joining in free trade arrangements like the Trans-Pacific Partnership (TPP) can thus be a useful way to improve the competitiveness and growth of Japan, the US, and other Asian economies.

13

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

REFERENCES
Aminian, N., K.C. Fung, A. Garcia-Herrero, and C. Lin, 2012, The Political Economy of Exchange Rates: The Case of the Japanese Yen. Japan and the World Economy 24(3): 193196. Belleflamme, P., and M. Peitz. 2010. Industrial Organization: Markets and Strategies. Cambridge, UK: Cambridge University Press. Bagwell, K., and R. Staiger. 2009. Profit-Shifting and Trade Agreements in Imperfectly Competitive Markets. NBER Working Paper No.14803. March. Baye, M., J.R. Gatti, P. Kattuman, and J. Morgan, 2006. Did the Euro Foster Online Price Competition? Evidence from an International Price Competition Site. Economic Inquiry. 44(2): 265279. Bernhofen, D., 1999, Intra-Industry Trade and Strategic Interactions: Theory and Evidence, Journal of International Economics 47(1): 225244. Christos, C., S. Clerides, I. Ioannou, and D. Senft. 2007. A Textbook Example of International Price Discrimination. Economic Letters 95(1): 9195 Engel, C., and J. Rogers. 2004. European Product Market Integration after the Euro. Economic Policy 19(39): 347384. Fung, K.C. 1991a. Collusive Intra-Industry Trade. Canadian Journal of Economics 2: 391 404. Fung, K.C. 1991b. Characteristics of Japanese Industrial Groups and Their Potential Impact on U.S.-Japanese Trade. In: Empirical Studies of Commercial Policy, NBER Conference Volume, edited by R. Baldwin. Chicago, IL: University of Chicago Press. Fung, K.C. 2002. International Trade and Bank Groups: Welfare Enhancing or Welfare Reducing? Journal of Japanese and International Economies 16(2): 212226. Goldberg, P., and F. Verboven. 2005. Market Integration and Convergence to the Law of One Price: Evidence from the European Car Market. Journal of International Economics 65: 4973. Helpman, E. 1984. Increasing Returns, Imperfect Markets and Trade Theory. In: Handbook of International Economics, vol.1, edited by R. Jones and P. Kenen. Amsterdam: North Holland. Helpman, E., and P. Krugman. 1985. Market Structure and Foreign Trade. Cambridge, MA: MIT Press. Helpman, E., and P. Krugman. 1989. Trade Policy and Market Structures. Cambridge, MA: MIT Press. Japan Administrative Management Agency. 1985. Input-Output Tables. Tokyo: Government of Japan. Japan Administrative Management Agency. 1990. Input-Output Tables. Tokyo: Government of Japan. Japanese Government. 1989. Census of Manufactures: Report by Industries. Japanese Government. 1991. Census of Manufactures: Report by Industries. Maier, P. 2009. An Analysis of International Price Differentials on eBay. Contemporary Economic Policy 28(3): 307321. US Government and Japanese Government. 1989. Joint Price Surveys. 14

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

US Government and Japanese Government. 1991. Joint Price Surveys. World Bank. Data on International Transport Costs. Trade Policy Division, mimeo. Washington, DC.

15

ADBI Working Paper 417

Fung, Garcia-Herrero, and Ng

APPENDIX
p i /p* i d i (d* i ) Ratio of retail price of product i in the domestic market to the retail price of the same product in the foreign market Source: Japanese and US Governments, Joint Price Surveys ad valorem distribution costs (the sum of retail and wholesale costs divided by the producers price) in the domestic (foreign) market Source: US Bureau of Economic Analysis, Input-Output Tables, Department of Commerce; Japan Administrative Management Agency, Input-Output Tables ad valorem international transport costs from the domestic (foreign) market to the foreign (domestic) market Source: US international transport costs come from data supplied by the Trade Policy Division, World Bank, constructed from import data of the US Bureau of the Census; Japanese international transport costs come from data supplied by Japan Economic Planning Agency number of domestic (foreign) firms Source: US Department of Commerce, Census of Manufactures; Japan Census of Manufactures ad valorem domestic (foreign) tariff rates for product i Source: GATT tariff data

tr i (tr* i )

n (n*)

t i (t* i )

16