Quotas on Clothing Imports: Impact and Determinants of EU Trade Policy

Quotas on Clothing Imports: Impact and Determinants of EU Trade Policy Juliette Milgram Baleix* Abstract We assess the impact of the phasing-out of quotas on European clothing imports within the framework of the phase-out of the Multi-Fiber Agreement and the accession of the CEEC.We use 1996 data on trade barriers for 20 categories of clothing products and 22 exporters. The estimation of a standard gravity equation concludes that tariffs have a large and negative impact as expected but seldom corroborated in the literature.The negative impact of non-tariff barriers also appears clearly after controlling for an endogeneity bias by the instrumental variables method. The phasing-out of quotas should increase EU imports by 20%. 1. Introduction Trade theory has recently focused on the difficulties to assess the impact of trade barriers on imports. In most cases, the difficult part of these predictions consists in gathering detailed and reliable information on tariff and non-tariff barriers.When such data are used, the estimation of the impact of trade liberalization on imports is surprisingly low, suggesting that estimation methodology or theoretical prediction should be reviewed. The main goal of this paper is to assess the impact of the removal of quantitative restrictions on EU clothing imports within the phase-out of the Multi-Fiber Agreement (MFA) and the accession of the Central and East European Countries (CEEC). Another goal of the paper is to focus on methodological concerns that could lead to estimates of the impact of trade protectionist measures more consistent with theoretical predictions. On the one hand, we argue that the use of tariff data at a broad disaggregation level, since it offers a more adequate measurement of prices, allows to estimate more easily demand elasticity in relation to customs duty. We also focus on the necessity to take into account that trade protection is influenced by factors of economic policy as shown by the theory of endogenous protection. Considering trade barriers as exogenous could lead to downward estimates of their impact on imports. For this purpose, we estimate a gravity equation that uses cross-sectional data for imports of 20 different categories of clothing for the 22 largest exporters of these articles into the EU market for the year 1996 and tariff and non-tariff barriers thanks to an original gathering of data. Finally, we performed a simulation of the abolition of these quantitative restrictions in order to evaluate its impact on EU imports of clothing. This paper is organized as follows:the next section presents the theoretical framework of the empirical model tested further on. Section 3 briefly describes the dimensions and Review of International Economics, 13(3), 445–460, 2005 © Blackwell Publishing Ltd 2005, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA *Milgram: CentrA and Universidad de Granada Facultad de Ciencias Empresariales y Económicas Departamento de Teoría e Historia Económica, Campus de la Cartuja 18011 Granada, Spain. Tel: +34 958 24 99 95; Fax: +34 958 24 40 46; E-mail: [email protected]. I am grateful to CentrA for its financial support, to CEPII for its hospitality and to Marta Castilho for useful suggestions. I also thank Lionel Fontagné, Guillaume Gaulier, Amina Lahreche, Daniel Mirza, Carlos Sanchez, and two anonymous referees for very helpful comments. the scope of tariff and non-tariff barriers on EU imports of clothing. Section 4 presents the empirical model and the econometric results. Section 5 presents the results of the simulation process. Finally, the last section summarizes the main conclusions. 2. Methodological Framework Specialization is at the source of the gravity force in trade and this explains why the imports of a country are positively correlated with this country’s income and with the production of the exporting country. Bilateral trade volume may also be negatively correlated with trade barriers such as transportation costs. The standard gravity equation of trade establishes these relations using the geographical distance as a proxy for transportation costs. The gravity equation is highly effective at explaining bilateral flows as proven at a very early date by the works of Linnemann (1966) and Leamer and Stern (1970). Later, the gravity equation was justified in the context of various theoretical frameworks (Bergstrand, 1989, for the factorial model; Deardorff, 1998, for the Heckscher–Ohlin (HO) model; and Anderson, 1979, for goods differentiated according to their origin). More recently, empirical validations of the gravity equations derived from these theoretical models1conclude that the HO model would better explain the success of the gravity equation when the partners have very different factorial endowments, while increasing-returns models would better explain the exchanges between similar countries precisely because the exchanges of differentiated goods represent a significant share of their trade. The specification of the gravity equation was refined in many studies in order to account for factors that could limit or strengthen trade relations and thus obtain a more complete empirical model.2Surprisingly, we can only find few attempts to reflect trade policies in the gravity equation. A first generation of models focus on the influence of regional agreements on trade flows. Their presence is generally integrated by means of dummy variables representing the regions’ affiliation to some kind of agreement.3 But the use of dummy variables can lead to an overestimation of the impact of such agreements, if they reflect other elements not specified in the model. Only a few recent studies propose integrating finer estimates of the trade barriers, opening the way to completely innovating and highly promising research. Fouquin and Gaulier (1999) and Wall (1999) used a qualitative variable, determined exogenously that expresses the restrictiveness of the trade policy. Harrigan (1993), Haveman and Hummels (1998), Hummels (1999), and Castilho (2002) explicitly took into account customs duties and non-tariff barriers (NTBs). Estimates are carried out at the industry level due to the heterogeneity of the barriers and establish the impact of trade policies much more precisely. These studies also evidence that the tariff and NTB coefficients do not always display the expected sign.We argue that these surprising results can often be explained within the framework of the theory of endogenous protection. Trefler (1993) and Lee and Swagel (1997) offer strong support for this view. Their estimations prove that the non-endogenization of the NTB could lead to an undervaluation of their effects on imports and even, in certain cases, to a change in the coefficients’ signs.4These findings can clearly be explained in the framework of the political-economy literature. Baldwin (1985),Magee et al.(1989),and Grossman and Helpman (1994) argue that high levels of import penetration result in a more intensive mobilization of private interests,who tend to organize in lobbies in support of protectionism.In this sense,when the NTBs are postulated as exogenous, their impact on imports is necessarily underestimated. 446 Juliette Milgram Baleix © Blackwell Publishing Ltd 2005 3. Scope of the Study The countries that export clothing articles to the EU are confronted with the wellknown double problem of the European trade policy. Indeed, these articles belong to the group of products classified as “very sensitive,” and therefore the customs duties imposed by the EU are higher than those for other categories. Moreover, as final consumer goods, they are subject to “tariff escalation” which consists in applying lower tariffs to raw or primary materials than to more elaborated products. An MFN tariff applied to these products is thus the highest of all the tariffs applied to industrial products: the average customs duty of the EU for industrial products was 6% in 1995 and 4.9% in 1997 against 13% and 12% for garments (OECD, 1997, p. 46). The common trade policy for “sensitive products” is also characterized by the presence of NTBs. These are established at the category level where each category is composed of similar clothing products at the eight-digit level and defined by the common external policy and the MFA. The main exporting countries of clothing to the EU are the members of the EU themselves and the newly industrialized countries (NICs) of Asia (the share of these two groups is decreasing as they are progressively disengaging from this type of specialization).As far as these more highly developed Asian countries are concerned, they must face MFN tariffs, and as signatories of the MFA quantitative restrictions are imposed on their exports—see Table 1. In general, the poorer Asian countries are granted a preferential status (lower customs duty and a higher quota) as LDCs (least developed countries).5However, China, India, and Vietnam, in spite of their low per capita incomes, do not benefit from any preference. Thus, exports from these latter partners have been,together with those of the CEEC,the most dynamic in recent years. Customs duties on EU-imported clothing articles originating in the CEEC have been gradually reduced. By 1996, only a few quotas remained. Among the most significant exporters of clothing articles,6we also find Turkey, Morocco, and Tunisia. These three Mediterranean countries are important suppliers of the EU. Indeed, their industrial products have already enjoyed free access to the EU since the 1976 Cooperation Agreements for the North African countries and, from 1998, within the framework of the Customs Union for Turkey. The progressive phase-out of the MFA means the suppression of these NTBs.7In addition, the implementation of the European Agreements with the CEEC resulted in an almost immediate tariff reduction, while the quantitative restrictions are being dismantled only gradually.The EU trade policy in the textile and clothing sector has thus been subject to considerable changes for several years for these reasons. It is likely to affect negatively nearby EU partners such as Turkey,Tunisia, and Morocco, since these changes will mean a reduction in their margin of preference. Benefiting from the favorable treatment which was granted to them by the EU, these Mediterranean countries have increased the volume of their textile and clothing exports in their foreign exchanges, as well as the weight of the European market as recipient of their exports. 4. Econometric Results Standard Gravity Equation It is very widely accepted that the exchange of clothing products between the developed and the developing countries is explained by a Heckscher–Ohlin-type model.We study here the EU countries’ imports coming from the Mediterranean countries, the IMPACT OF QUOTAS ON CLOTHING IMPORTS 447 © Blackwell Publishing Ltd 2005 CEEC, and Asia. The endowments of the importer countries (the EU members) and the exporters of our sampling are sufficiently different for a considerable degree of specialization to take place.We use the most general specification of the gravity model described by the following equation: 448 Juliette Milgram Baleix © Blackwell Publishing Ltd 2005 Table 1. EU Clothing Market: Characteristics of Main Exporters Simple Average Number Number Per customs quota of quotas of quotas capita Share of duty utilization (on 21 fully GDP EU Exporters (1996) rateacategories) uilizedbin ECUs importsc Turkey 0.0 0 0 2,290 13.4 China 12.2 89.2 20 11 530 13.3 Hong Kong 12.2 64.7 16 3 19,060 9.8 Tunisia 0.0 0 0 1,696 7.1 Morocco 0.0 0 0 1,032 6.6 Poland 0.0 41.9 6 0 2,749 6.4 India 12.2 87.0 11 6 299 6.1 Bangladesh 0.0 0 0 257 5.0 Romania 0.0 41.4 9 0 1,231 4.6 Indonesia 10.4 73.8 7 1 868 4.0 Hungary 0.0 25.2 9 0 3,440 3.1 Thailand 12.2 56.6 10 0 2,431 2.1 Macao 0.0 71.2 15 6 13,701 1.8 Sri Lanka 10.3 57.5 5 1 592 1.7 Croatia 0.0 0 0 3,226 1.6 Czech 0.0 33.6 12 1 4,318 1.5 Republic Pakistan 12.2 63.0 8 1 367 1.5 Vietnam 10.4 88.0 21 8 248 1.5 Slovenia 12.4 0 0 7,534 1.4 Malaysia 10.4 57.8 6 1 3,905 1.4 Slovakia 0.0 34.9 10 0 2,754 1.4 Korea 12.2 16.9 20 0 8,402 1.3 Bulgaria 0.0 60.5 6 1 929 1.2 Philippines 10.3 40.6 12 1 877 1.0 Taiwan 12.2 28.7 18 0 9,978 1.0 100.0 Notes: aThe average quota utilization rate has been obtained as the average of the quota utilization rate (UR) of each of the 21 categories of clothing products. UR is the ratio between the quantity of EU imports and the amount of the quota. EU imports in units for each category have been obtained as the sum of EU imports in units for each eight-digit product included in the category. bWe consider as fully utilized a quota with a utilization rate superior to 90% as Nagarajan (1995). It is important to recall that the utilization rate can be superior to 100%. Indeed, the bilateral agreements involve always a certain percentage of flexibility which allows exporting more products of one category if the exports of another category are reduced within certain limits. cThese are the sum of imports of the EU coming from selected partners, amounting to approximately onehalf of the total EU imports and more than 80% of the imports coming from third countries. Source: Calculations by the author from: TRAINS database of UNCTAD (1996) for simple customs duty, Comext (1997) for quantities and values of EU imports, World Development Indicators, World Bank, 1997, for per capita GDP, OJ EU L 275 of 8.11.93 and OJ EU L 307 of 28.11.96 for amounts of quantitative restrictions. where irepresents the importing EU member country (i=1,...,14); j, the exporter (j=1,...,22,the 22 main exporters of garment articles towards the EU); M, the bilateral imports of the various clothing products; Y, GDP; y, the per capita GDP; dist,the geographical distance (in kilometers) between the capitals of countries iand j; t, the average duty,8and NTB, an indicator of the incidence of the NTBs.9The indicators of the trade barriers at accessing the EU are calculated at the level of 20 categories of clothing products since NTBs are established at this aggregation level. According to gravity principles, the per capita GDP of the exporting countries is a proxy of capital intensity. It is thus negatively correlated with its exports when the sector is labor intensive as it is in the present case. Likewise, countries relatively abundant in capital tend to import labor-intensive products. The per capita GDP of the importing countries is thus supposed to have a positive impact on the imports of these products. As GDP of the exporter is used as a measure for its potential supply, it must have a positive impact on exports. Finally, imports are supposed to grow with the import demand, which in this case means with the importer’s GDP used as a proxy for the import potential.10 Obstacles to trade should obviously have a negative coefficient. This is the case of geographical distance, but also of tariff and non-tariff barriers. The standard model is tested in its logarithmic form. Two specifications were considered: one specification without a fixed effect (specification 1a): (1a) where Crepresents the product categories (C=1,...,21). The NTB variables indicating the presence of quotas or the utilization rate of the quota are available only at the level of the member countries and by product categories, which combine many products defined at the eight-digit level of the combined nomenclature and another one with a fixed effect (specification 1b): (1b) where DCrepresents a dummy category.11 For each of the two specifications tested, all variables are significant, at the 1% level (Table 2).The explanatory capacity is of 30% while using the method of ordinary least squares (OLS) (specification 1a), and of 38% when following the fixed-effects method (specification 1b). It should be noted that these coefficients are relatively high when dealing with such a disaggregated estimate. The standard variables of the gravity models show the expected signs, since the exporters’ and the importers’ GDP, as well as the per capita GDP of the importer show a positive coefficient, whereas the exporters’ per capita GDP coefficient is negative. Moreover, distance has a negative impact on imports, as one would expect. The variables of trade policy are particularly deserving of our attention and will constitute the most original part of this study.12 Customs duty has the expected negative sign, which is not always the case when estimates are carried out at the sectorial level— see, for example, Castilho (2002). The coefficient of this variable, which represents the demand elasticity in relation to one of the price components such as customs duty, is, in fact, rather high (between -3 and -4.5 according to the specifications). It is true, ln ln ln ln ln ln ln , M Y Y y y dist t NTB D ij Cijij ij j Cj CCC C ij =+ + + + + ++ () +++ = Â aa a a a a aa be 01 2 3 4 5 67 1 20 1 ln ln ln ln ln ln ln , M Y Y y y dist t NTB ij Cijij ij j Cj Cij =+ + + + + ++ () ++ aa a a a a aae 01 2 3 4 5 67 1 M Y Y y y dist t NTB ij i jij ijj j =a aaa aa a 1234 56 7 .. , IMPACT OF QUOTAS ON CLOTHING IMPORTS 449 © Blackwell Publishing Ltd 2005 however, that empirical studies often obtain inferior values which lead us to believe that the price effects have been underestimated with regard to theoretical forecasts.13 On the one hand, there may be certain factors that influence both the prices and the amounts in demand (when, for example, quality and technical progress are not included in the model this will lead to an underestimation of the elasticities). On the other hand, estimating the price elasticity is often carried out at aggregate levels (geographically and sectorially) and thus often requires the use of inadequate price measurements (indices, average unit values). Erkel-Rousse and Mirza (2002) propose to instrument for the price variables and carrying out estimates on sectorial data in order to control these two types of bias. In so doing they obtain elasticities which are more in accordance with those envisaged by the theoretical literature (between 1 and 7, depending on the sector). The coefficients obtained in our study are thus in harmony with the theoretical forecasts (strong price elasticity) since we are dealing with relatively homogeneous goods and with exports from countries which can be regarded as “price-takers” towards a “large importer.” They confirm that the disaggregated estimates and the quality of the price measurement (we are dealing here with customs duty, which is a component of the price but does not entail a quality effect) make it possible to improve elasticity 450 Juliette Milgram Baleix © Blackwell Publishing Ltd 2005 Table 2. Impact of Tariffs and NTB on EU Imports Explained variable: bilateral imports of EU countries, 1996 Specification 1a 1b 1c 1d Exporter’s GDP 0.488*** 0.566*** 0.526*** 0.547*** (15.72) (19.13) (18.16) (18.73) Exporter’s per capita GDP -0.199*** -0.187*** -0.133*** -0.202*** (-7.05) (-6.98) (-5.23) (-7.64) Importer’s GDP 1.125*** 1.17*** 1.165*** 1.171*** (36.53) (40.1) (39.81) (40.07) Importer’s per capita GDP 1.413*** 1.484*** 1.54*** 1.471*** (12.84) (14.25) (14.78) (14.12) Distance -0.211*** -0.292*** -0.413*** (-4.28) (-6.18) (-11.81) Customs duty -4.506*** -3.178*** -6.653*** (-5.24) (-3.81) (-10.74) QR 0.803*** 0.306*** 0.34*** 0.241*** (12.0) (4.23) (4.7) (3.43) Constant -24.455*** -25.457*** -28.103*** -24.229*** (-19.85) (-21.69) (-25.62) (-21.44) Fixed effect by category X X X Number of observations 4634 4634 4634 4634 R20.308 0.385 0.38 0.383 Notes: ***significant at 1%; **significant at 5%; *significant at 10%. tof Student in parentheses. No indications of heteroskedasticity were verified after performing the Cook–Weisberg test nor of multicollinearity when using the inflation factors of the variance. Source: Calculations by the author using data from Comext (for imports), TRAINS for customs duties, Chelem (for the GDP data), and the European Commission (1994) (for the NTB). estimates. Integrating the tariff data in this type of estimate thus opens up a highly promising research field. The variable indicating the presence of quantitative restriction (QR) does not show the expected negative sign. This problem also appears in other studies that take into account NTB indicators—Haveman and Hummels (1998), Hummels (1999), and Castilho (2002).The variable is, however, very significant. Since we have cross-sectional data, this result suggests that, on average, those countries whose exports are subject to quotas are the largest exporters, in spite of the fact that the size effect is taken into account by the GDP variable. This paradox could be explained by the presence of an endogeneity bias14 which would lead to an erroneous estimation of the parameters. Indeed, one would tend to think that the quotas are imposed precisely on those countries whose clothing exports are already very significant, in order to prevent a further increase in EU imports. Lastly, in the case of our study, distance is shown as correlated with the customs duty at 67%, which is explained by the fact that countries close to the EU benefit from a preferential access. Since this correlation could lead to a distorted estimate of the parameters, we tested two other specifications without including the distance (specification 1c) or the tariff (specification 1d). The explanatory character of the model is unquestionable (the R2decreases only slightly) and the variables are very significant. In the same way, the signs and values of the coefficients of the other variables are not altered, and the coefficient of variable QR is not affected. Since the relative correlation between distance and tariff do not affect the results of the other variables, we can affirm that estimates 1a and 1b are not skewed. As the fixed-effects method (1b) offers a better explanatory capacity, we retained this specification to carry out other estimates which attempt to detect and to correct any possible endogeneity bias which would lead to an incorrect estimate of the QR variable coefficient. Endogenization of the NTB Trefler (1993) and Lee and Swagel (1997) simultaneously estimate import and NTB equations.Their results are consistent with the political-economy theories of the determination of trade protection. Thus, modeling protection as endogenous also appears to be the most adequate issue here. Following political-economy literature, measures of import penetration, differences in labor costs between importer and exporter, wages, and comparative advantage in thousandths of GDP should be introduced as determinants of NTBs. Since we focus on NTBs imposed by the EU in only one industry, and in their average effect on various products and partners, we are unable to obtain these data.15 Thus it does not make sense to estimate an NTB equation but it appears necessary to consider NTBs as endogenous in the import equation using instruments for this variable. The difficulty consists in choosing instrumental variables which must be correlated with our QR indicator but not correlated with the residuals of the main equation (gravity equation). What can explain the presence of NTBs for different categories of clothing products and partners? Several solutions were considered here. Since the EU has lost competitiveness in relation to developing countries, the most competitive partners (those whose real labor costs are low) undoubtedly are more severely affected by the QR. One option would be to take into account the difference in labor costs between the importing and exporting countries, but it was not possible to gather these data. However, it is possible to use the real exchange rate as a macroeconomic indicator of price competitiveness. On IMPACT OF QUOTAS ON CLOTHING IMPORTS 451 © Blackwell Publishing Ltd 2005 the other hand, the country fixed-effects can be included, which would take into account other competitiveness effects than those caused by exchange rates. The lagged value of the independent variable is often used as an instrument variable. In this case, the growth of past exports is an additional indicator of competitiveness (an indicator of the same dimensions as the explained variable) and a candidate to being a good instrument (not correlated to the error of the gravitational model) and it is only natural that those partners whose past imports were especially dynamic will enjoy a higher protection.This is why we also used the growth rate of past imports as an instrument. Finally, all the explanatory variables used in the main equation are deemed instrumental, because, as they are not correlated with the residuals, these variables are “the best candidates to be good instruments” (Kennedy, 1999, p. 165). In the first stage we regress the QR.j Cvariable on the instruments.The predicted value of the dependent variable in that regression (QRj C,pred) is then used in the second-stage regression to explain imports. The estimated equations are thus as follows: where Djrepresents a dummy partner country; RERij is the real exchange rate between the importing country iand the exporting country j;16 mij Cis the growth rate of the imports of country ifrom country jfor the product category C. It has been calculated for three different periods: 1988–96, 1988–92, and 1993–96. The exporter fixed-effects are therefore common to all the estimates. Six specifications are presented in Table 3: • 2a: equation (2) without the RER and without the imports growth rate; • 2b: equation (2) with the RER and without the imports growth rate; • 2c: equation (2) with the RER and the imports growth rate over the 1988–96 period; • 2d: equation (2) with the RER and the imports growth rate over the 1993–96 period; • 2e: equation (2) with the RER and the imports growth rate over the 1988–92 and 1993–96 periods; • 2f: equation (2) with the RER and the imports growth rate over the 1988–92 period. The endogenization of variable QR, no matter which specification is chosen (Table 3), provides negative coefficients for this variable, whereas they were positive in the traditional estimate according to the OLS method and the fixed-effects method. The coefficients of determination for the first equation (estimate of the endogenous variable QR) are approximately 0.6 in all cases. The stability of the coefficients from one specification to the other suggests that the instruments common to all the specifications (specific effects of partner country and category) are an important determinant of the restrictive character of the QR.In general, there are some categories that receive more protection from the EU, as well as partners for whom the restrictions are more effective than for others, independently of their competitiveness or the growth of their exports in the past.17 ln ln ln ln ln ln ln , , M Y Y y y dist tQR D ij Cijij ij j C j C pred CC C ij =+ + + + + ++ () +++ = Â aa a a a a aa be 01 2 3 4 5 67 1 20 1 QR Y Y y y t RER m D D j Cijij j C ij ij CCC C jj j C j =+ + + + + + () ++ () ++ () +++ = ÂÂ bb b b b b bbbbm 01 2 3 4 5 67 1 20 1 11 ln ln ln ln ln ln ln 452 Juliette Milgram Baleix © Blackwell Publishing Ltd 2005 (2) IMPACT OF QUOTAS ON CLOTHING IMPORTS 453 © Blackwell Publishing Ltd 2005 Table 3. Impact of Tariffs and NTB on EU Imports with Endogenization of NTB Explained variable: bilateral imports of EU countries, 1996; 2SLS estimates with instrumented NTB Specification AB C D E F (RER +m93–96 + No RER RER (RER +m88–96) (RER +m93–96) m88–92) (RER +m88–92) Exporter’s GDP 0.606*** 0.581*** 0.545*** 0.522*** 0.511*** 0.540*** (19.96) (18.56) (12.58) (14.14) (11.95) (12.49) Exporter’s per capita GDP -0.181*** -0.159*** -0.120*** -0.154*** -0.120*** -0.122*** (-6.71) (-5.74) (-3.52) (-4.89) (-3.61) (-3.60) Importer’s GDP 1.167*** 1.172*** 0.922*** 0.946*** 0.875*** 0.922*** (39.69) (39.24) (23.18) (26.23) (21.96) (23.21) Importer’s per capita GDP 1.490*** 1.470*** 1.453*** 1.437*** 1.247*** 1.454*** (14.19) (13.89) (9.32) (11.47) (8.06) (9.34) Distance -0.322*** -0.417*** -0.543*** -0.402*** -0.549*** -0.534*** (-6.75) (-7.39) (-5.79) (-5.29) (-5.93) (-5.70) Customs duty -1.474** 0.994 -2.576 -1.098 -3.006* -2.845* (-1.68) (0.856) (-1.540) (-0.750) (-1.81) (-1.70) QR -0.317*** -0.521*** -0.514*** -0.560*** -0.320* -0.455** (-2.74) (-3.97) (-2.57) (-3.54) (-1.64) (-2.28) Constant -25.703*** -24.764*** -19.341*** -20.590*** -16.254*** -19.353*** (-21.72) (-20.41) (-10.41) (-14.44) (-8.74) (-10.43) Fixed-effects categories XXXXXX Number of observationsa4634 4505 2458 3117 2376 2458 (1st equation) R20.585 0.585 0.589 0.587 0.59 0.589 Notes: ***significant at 1%, **significant at 5%, *significant at 10%. tof Student in parentheses. aThe number of observations varies according to specifications, since the growth rate of the imports could not be calculated for all countries (in particular for the Czech Republic, Slovakia, Croatia, and Slovenia). In addition, the RER was not available for three countries. Finally, for certain pairs of countries, all categories are not imported. Source: See Table 2. 16. The RER between iand jhas been obtained dividing the RER of iby the RER of jdefined in relation to the EU (15 members). They have been obtained from the CHELEM database (CEPII, France). 17. It should be noted that by including the country fixed-effects in the standard gravity equation (equation (1)), it would have been impossible to solve the problem of the sign of the variable QR.The results are not presented in order not to overburden the discussion, and in any case, the inclusion of these effects would only improve the explanatory capacity of the model very slightly, and it does not modify in any way the scope and the significance of the results. 18. Thus, one can easily calculate the importing potential of the EU from each country j (Mj pot =S iSCMij C,pot) or for each category of considered products (MC,pot =S iSjMij C,pot). 19. The simulated imports (Mij C,pred) are obtained as follows: MYYy y dist t M e ij C pred iji jij j CCij C pot QR j C pred , ,. exp . . ln . ln . ln .ln . ln .ln . . , =- + + + ( -- -+ ()) += 25 703 1 167 0 606 1 490 0 181 0 322 1 474 1 0 317 b 460 Juliette Milgram Baleix © Blackwell Publishing Ltd 2005