Food Policy 33 (2008) 595606

Contents lists available at ScienceDirect

Food Policy
journal homepage: www.elsevier.com/locate/foodpol

The importance of EU-15 borders for CEECs agri-food exports: The role of tariffs and non-tariff measures in the pre-accession period
Emmanuelle Chevassus-Lozza a,*, Karine Latouche a,1, Darja Majkovic b,2, Manuela Unguru c,3
a

INRA, UR 1134, Rue de la Graudire, BP 71627, F-44316 Nantes, Cedex 03, France University of Maribor, Faculty of Agriculture, Vrbanska 30, 2000 Maribor, Slovenia c CASA ACADEMIEI ROMANE, Institute for World Economics, 13, Calea 13 Septembrie nr. 13, Sector 5, Romania
b

a r t i c l e

i n f o

a b s t r a c t
This paper assesses the evolution of the EU-15 market access for the agri-food products, originating from CEECs. A gravity model technique has been used to assess the overall trade resistance (border effect), and to weight its various components such as tariff and non-tariff measures (sanitary and phytosanitary standards, other quality measures) in the pre-accession period (19992004). The ndings reveal, despite the undertaken integration and trade liberalisation processes, a persisting and signicant trade resistance for the CEECs agri-food exports to the EU market, even just prior their accession in 2004. Still present difculties in market access at the time are partially explained by tariff and non-tariff measures, while a large part of border effect remains in the domain of other, presumably non-trade policy related factors (home bias, consumer preferences, etc.). These results indicate that despite the accession process (customs union and adoption of the EU standards), the trade integration of the enlarged EU market was not yet completed in 2004. 2008 Elsevier Ltd. All rights reserved.

Article history: Received 28 April 2005 Received in revised form 19 June 2008 Accepted 19 July 2008

Keywords: Eastern EU enlargements Gravity equation Border effect Tariffs and NTBs Trade resistance Agri-food products

Introduction Since the Single European Market (SEM) has been established, the free movement of goods has been facilitated not only by removing border formalities, but also by the technical harmonisation of national legislation of each member state. For the agri-food sector a particular concern is to guarantee the safety and integrity of products. In this respect, the European Commission has developed a stringent policy regarding food safety and consumer information. Hence, strict regulation is imposed for all agri-food products commercialised in the SEM, whether of European or third country provenance. In the case of EU enlargement, accession to the SEM is conditional upon the candidate countries acceptance of internal market obligations, and therefore making their national legislation compatible with the body of laws and regulations of the EU, the socalled acquis communautaire. The impact of EU accession on the candidate countries agri-food trade is related, on the one hand, with the adoption of production and commercialisation norms

* Corresponding author. Fax: +33 240675074. E-mail addresses: chevassu@nantes.inra.fr (E. Chevassus-Lozza), latouche@ nantes.inra.fr (K. Latouche), darja.majkovic@uni-mb.si (D. Majkovic), m.unguru@ gmail.com (M. Unguru). 1 Fax: +33 240675074. 2 Fax: +386 2 229 60 71. 3 Fax: +40 213182454. 0306-9192/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodpol.2008.07.005

stipulated in the acquis communautaire (namely in Chapter 7 Agriculture and its chapter 23 Consumer and health protection) and, on the other hand, with the liberalisation of trade (according to the rst chapter of the acquis Free movement of goods). Fullling the requirements of EU accession means for the acceding countries not only costs related to adjustments of their production technologies in order to be consistent with the acquis communautaire, but also benets in terms of market access. Concerning the latter, one may expect that the new member states (NMS) will gain in terms of trade, since they should benet, on the one hand, from abolishment of the tariff barriers and, on the other hand, from the likely reduction of the barriers to trade in the pre-accession period, when the production and commercialisation standards were catching-up with those of the EU. Although several articles evaluate the impact of enlargement on trade, only few studies focus on the level of market access and its evolution related to the process of EU accession. This paper focuses on the impact of trade barriers on agri-food exports towards the EU from Central and Eastern European Countries (CEECs). More precisely, it attempts: (i) to investigate if, overall, the process of preparing for EU accession has improved the access of CEECs agri-food products on the European markets; (ii) to highlight the role of specic determinants of trade resistance, such as tariffs and non-tariff barriers, over the period immediately before their EU accession. For this purpose, we will measure the impact of EU borders (called hereafter border effect) on CEECs agri-food exports on EU

596

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606 Table 1 Tariffs applied by the European Union 1999 Czech Republic Estonia Hungary Lithuania Latvia Poland Slovenia Slovak Republic Bulgaria Romania MFN rate 14.91 15.81 15.05 16.13 16.15 13.98 16.36 14.93 14.62 14.58 18.1 2004 5.49 5.16 4.93 5.14 5.36 5.83 5.95 5.42 5.5 8.22 16.3

markets and stress the role that tariff and non-tariff measures play in explaining this effect. That means answering the following questions: just prior to joining the SEM, do NMS still face trade resistance, measured in terms of border effect when selling on EU market? Has this effect been reduced over the period of preparation for enlargement? During this period tariffs have been gradually eliminated and, due to the harmonising of their production and commercialisation norms with the EU standards, the non-tariff barriers are supposed to impede less the exports of these countries towards EU. For assessing the border effects of CEECs exports towards, we use a gravity model based mainly on the model developed by Anderson and van Wincoop, 2003. The originality of our work is twofold. First, the model is dened at a highly disaggregated level. Second, in addition to classical determinants for bilateral trade ows, our model includes separately variables for tariff and nontariff measures. The latter indicate whether products are regulated when entering the SEM, through either technical measures meant to ensure the compliance with products quality requirements (e.g. labelling standards) or measures for human and animal protection (e.g. sanitary and phytosanitary standards). The model is applied to agri-food imports of each EU-15 member state, from all its trading partners, by distinguishing among them four groups of exporting countries: (i) EU-15; (ii) the NMS84 (Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovak Republic and Slovenia); (iii) the NMS-2 (Romania and Bulgaria)5; (iv) third countries. The cross-section estimation was assessed for two years: rstly 1999, when despite the Europe Agreements, the liberalisation of agri-food trade between NMSs-8 and the EU was not completed (before the 2000 zerozero agreements), and secondly 2004, the year of the rst wave of Eastern enlargement.6 The paper proceeds as follows. The next section goes over trade developments and main policy instruments for trade between the EU and CEECs in the pre-accession period. The third section gives an outline of the model specication and presents the econometric method used. The fourth section deals with empirical assessment. In order to answer the questions asked in the previous part of the paper, a three step assessment is implemented. In the rst step the overall trade resistance faced by CEECs is carried out. The role played by tariffs and non-tariff measures in this global resistance is then highlighted and nally, the role of non-tariff measures according to the origin of the products is analysed. Conclusions are drawn in the nal section. EU15-CEECs agri-food trade development in the pre-accession period The Europe Agreements: a path towards trade liberalisation An important step in the preparation of the EU Eastern enlargement was the launch of Europe Agreements (EA), the association agreements between the EU and individual candidate countries from Central and Eastern Europe, signed after 1991. The EAs provided free trade in industrial goods and only a gradual and slight reduction in tariff for agricultural and food products. It was only with the so called double zero agreements, signed in 2000, that a progressive liberalisation of trade in agricultural products was

Arithmetic average for agricultural and food products (%). Source: computations based on TARIC database. MFN rate: tariff applied to third countries that do not benet from a preferential agreement.

4 The 10 CEECs have been separated into two groups: those acceding the EU in 2004 were called NMS-8 and those joining the EU in 2007 NMS-2. 5 We denote NMS-8 and NMS-2 together as CEECs. 6 In this paper we focused on the Central and Eastern European Countries (CEECs), out of which eight have acceded to the EU in 2004 (called in the paper NMS-8). The other two CEECs that joined the EU in 2007 are called NMS-2. Although Cyprus and Malta were involved in the 2004 EU enlargement, we do not include them in the estimations.

undertaken. Hence, as a result of these agreements, between 1999 and 2004, EU tariff protection for agri-food products coming from NMS-8 diminished much more than tariffs applied to imports originating from third countries that did not benet from a preferential trade regime, i.e. countries submitted to MFN regime (Table 1). However, even though tariffs were totally removed on May 1st 2004, for the new incumbents, the annual average tariff for 2004 was far from being null. As evidence, in the four months before accession of NMS-8, these countries still faced tariff barriers: the protection being particularly high for meat products, dairy products and tobacco (more than 12%). Nevertheless, these accords are not only simple trade agreements reducing tariffs, but also political agreements seeking to bring candidate countries legislation into line with the acquis communautaire. In the agri-food sector, for ensuring the smooth functioning of the internal EU market, the candidate countries had to comply with the EU in several elds of the acquis, such as: market regulation, veterinary and sanitary controls, animal welfare, the administrative structures needed to implement the Common Agricultural Policy. Among others, CEECs had to both upgrade the production technologies and make some institutional changes in the eld of sanitary inspection methods and facilities of border checkpoints. To this end, given that border controls no longer exist for the intra-EU trade, the European Commission requested candidate countries to draw up a veritable food safety strategy for integrating EU standards. The lack of full compliance with the acquis communautaire impede CEECs exports towards EU markets. Prior to the two waves of Eastern enlargement, yet remained some establishments which do not conform to EU norms. For those of the agri-food companies that have been appreciated by EU evaluators as capable of upgrading to the EU standards, a transition period has been granted to comply with the EU standards, during which they are allowed to sell their products exclusively on the local markets. When this period expires, if they are still not successful in upgrading to EU standards, these establishments will be closed down. The rest of this article will focus more specically on sanitary, phytosanitary and quality standards (relating to product commercialization). Within the context of adopting the acquis communautaire, it might be expected that, at the end of the pre-accession period, EU norms no longer inhibit CEECs agri-food exchanges on the SEM. Increasing orientation of CEECs agri-food exports to the EU, but much below the average intra-EU trade share In this context, can we observe a change in the trade patterns of CEECs? Over the period 19992004, the growth of exports to the

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606

250% 232% 235% Exports to EU Total Exports

597

200%

187%

187%

161% 150% 122% 110% 100% 92% 73% 65% 69% 142% 135%

158% 139%

83% 74% 58%55%

50%

41%

0% Czech RepE stonia Hungary Lithuania Latvia Poland Slovenia Slovakia Bulgaria Romania

Chart 1. Changes in the CEECs agri-food exports between 1999 and 2004.

EU has exceeded the pace of total CEECs export growth (Chart 1). The highest changes in volume of exports, of over 100%, were recorded by the Baltic States, the Slovak Republic, the Czech Republic and Poland, and the lowest by Slovenia. EU share in CEECs exports has also increased, except for Slovenia (Chart 2). For Hungary, Poland and Romania, exports to the EU represent more than the half of their total exports, while for Lithuania, Slovak Republic, Slovenia and Latvia, the EU accounts for less than a third of their exports in the analysed period. However, compared to old members, this share is quite low: on average, intra-EU exports represent 75% of their total agri-food exports for all EU-15 members. However, this rate also varies signicantly from one EU member to another (for instance, intra-EU exports account for 85% of total Portuguese or Dutch agri-food exports, while they represent only the half of Finnish exports). Among CEECs, Poland, Hungary and Czech Republic are the three main exporters of foodstuffs to the EU, with respectively 36%, 22% and 12% of total CEECs agri-food exports. Meat (processed products and live animals) and dairy products are the main exports of CEECs. It is also noteworthy that their fruit exports to EU remained stable over the period and consequently their share in the overall CEECs agri-food exports was reduced (from 11% to 6%). This short overview of CEECs trade illustrates a signicant and increasing weight of the European market for the newcomers, but the orientation of CEECs agri-food exports towards EU is still much lower than the share of intra- EU-15 exports in the EU15 agri-food trade. Moreover, in comparison to non agri-food products, this ratio is also very low.7 The above observations suggest that the level of agri-food trade integration of CEECs into the EU markets is not yet so explicit, despite the efforts towards economic (and trade) integration in the pre-accession period. In other words, they suggest that a non-negligible trade resistance for the agri-food exports towards the EU markets is persisting in the pre-accession period.

Measuring the trade resistance: the theoretical framework Among the studies on the border effects between new acceding countries and the EU, checking for the hypothesis that lifting of barriers at EU border induces a potential trade gain, there are Manchin and Pinna (2003), Nahuis (2004) and Olper and Raimondi (2008). Using gravity modelling, Nahuis shows that for CEECs the highest border effects at entry to the EU occur in the agricultural and food sector, and he consequently anticipates the highest trade gain for these products. When estimating the impact technical barriers have on border effects between the EU and selected CEECs, Manchin et al. show that the less harmonised the standards are, the higher the border effect they have. The Olper and Raimondi empirical estimates of the border effects derived from a gravity model reveal heterogeneous, but strong market integration among all the OECD groups of countries, except for the EUCEECs trade, as an indication that trade liberalisation between CEECs and EU has not yet produced the expected results for agricultural products. The model Gravity models represent one of the most common options for assessing bilateral trade ows, either in the classical specication or in their further developments within the new international economics theory.8 The equation estimated hereafter is based on Anderson and van Wincoops model (2003). It is assumed that all goods are differentiated by origin, consumer demand (in the importing country) being dened by a CES utility function. Anderson and Van Wincoops specication allows both components of total demand imports and domestic production to be taken into account. At the disaggregated level of this current study (i.e. at product level),9 the empirical estimation faces a serious data constraint:
8 Among others, see Anderson and van Wincoop, 2003; Evenett and Keller, 2002; Head and Mayer, 2000; Mayer and Zignago, 2005. 9 Usually these models are dened at a very aggregated level, but they have been also employed for studies at industry-level (Head and Mayer, 2002; Chen, 2004; Nahuis, 2004; Manchin and Pinna, 2003; Subramanian and Wei, 2007) and even at product level (Haveman and Thursby, 2000; Henry de Frahan and Vancauteren, 2006; Disdier et al., 2008; Olper and Raimondi, 2008) for the agricultural sector.

7 According to Papazoglou et al. (2006) the EU-15 account for 75% of total CEECS exports of goods.

598
60 .00% 56% 54% 50.00%

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606

56% 55%

1999

2004

54% 51 %

46 45% % 41% 40 .00% 38 % 33 % 31% 30 .00% 29% 25% 29% 27% 25% 43% 39% 35%

20 .00%

10.00%

0.00 % Czech RepE stonia Hungary Lithuania Latvia Poland Slovenia Slovakia Bulgaria Romania

Chart 2. EU share in total CEECs agri-food exports.

production data are not available for all products and all countries. Therefore, our proposal is to focus exclusively on the EUs import ows10 and the model is synthesised in the following equation:

xijk

  xik mjk Pijk =Pjk 1r Y wk Wik

where xijk is the nominal value of exports from i to j for product k; mjk is the total expenditure of j for the imported product k (the total imports of j); xik is the total exports of i for product k; Ywk is the total world trade in product k (sum of total imports for product k for all countries j); Pijk is the delivered price of product k from country i, paid by consumer in j; Pjk is the CES index import price of product k in country j; thus Pijk/Pjk is the price competitiveness of i on mar   1=1r ket j. P Pijk 1r And Wik hjk is a CES index of price competj P
jk

(proxied by distance) also reects other border variables (Bijk) stimulating bilateral trade (a common border, a common language, etc.), and a remaining trade resistance effect (Oijk) inhibiting trade (cultural differences, home bias, etc.). Among these other factors, Chen (2004) or Disdier and Mayer (2007) investigate informal trade barriers, such as product-specic information costs, and their role in explaining the trade reduction effect. Therefore Tijk becomes:

T ijk dij t h NTBk B1 Om ijk ijk ij ijk

itiveness of i on the world market global competitiveness, where m hjk Y jk is the share of the market j in the world market. wk The delivered price Pijkdiffers from the exporters supply price Pik, because of trade costs between i and j. These trade costs are broadly dened to include all costs incurred in getting a good to the nal user, excluding the production cost of the good itself (Anderson and van Wincoop, 2004). This so-called bilateral trade resistance encompasses mainly tariffs, transport costs, and nontariff barriers. Taking into account the trade costs (Tijk), the relation between the two prices becomes: Pijk = PikTijk. Thus (1) may be rewritten as follows:

where q > 0, (the higher the transport costs, the higher the trade costs) and h > 0. The sign of k is not dened (see OECD, 2002; Bureau and Beghin, 2001). It depends on the nature of the regulation. Some regulations may act as transaction costs and hinder trade. For others, the regulations act as a barrier if products do not meet standards, but once standards are met they may facilitate trade. f is expected to be negative (trade costs are reduced when countries share a common border or a common historical link). m > 0: the higher the unobserved trade resistance, the higher the trade costs. While checking for possible endogeneity of total exports (xik), total imports (mjk) and world demand (Ywk), these three variables have been moved to the left side of the equation.11 This solution constrains the coefcient of total exports, total imports and world demand to be equal to 1. In this way it is not the value of the bilateral ow which is regressed but the coefcient of relative bilateral trade intensity (RIijk) (Freudenberg et al., 1998).

xijk

 1r xik mjk T ijk Wijk Y wk Wik

RIijk

xijk =mjk xik =Y wk

P where Wijk Pik may be dened as the cost competitiveness of i in jk comparison to that of j (without any trade barrier). Following Peridy (2005) it is assumed that Tijk in addition to policy-related barriers (e.g. tariffs and NTBs) and transport costs

10 This implies adding a major supplementary hypothesis to Andersons model, i.e. that the consumer follows a two-step budgetary procedure. In the rst step, the importing countrys consumers dene the import demand, choosing between domestic and imported products in order to satisfy the total demand. In the second step, the import demand is differentiated by country of origin.

This index compares the proportion of country i exports of good k in the total imports of j, to the market share of the exporter i in the international market. An index equal to 1 means that the ow of good k between j and i is determined only by the size of the partners. A coefcient other than 1 means that trade is determined by other factors than country size: if it is greater than one, it denotes privileged trade links between i and j for good k, whereas an index less than one refers to trade resistance between the two countries

11

This approach has been used also by Mayer and Zignago (2005).

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606

599

which could be explained by the low competitiveness of j, by trade costs, etc. The econometric method Insofar as one of our objectives is to assess the impact of different trade barriers and, more precisely, to identify those which prohibit trade, both actual bilateral trade and zero values i.e. all potential bilateral ows must be taken into account. OLS regressions of the relative intensity index would not take into account the high proportion of zero trade ows, and so would lead to biased estimators.12 The zero values found in the trade database in fact correspond either to genuinely zero ow, or to a ow below a certain reporting threshold. The latter are very low and are therefore assimilated to absence of trade. Consequently the Heckman procedure is used to take such a selection bias into account (Heckman, 1979; Greene, 1981). This method attempts to control for the effect of non-random selection by incorporating both the observed and unobserved factors that affect non-response (Sales et al., 2004). Hence, the effects of the different trade barriers are tested at two stages of the export process: rst, the probability of a country i to export towards country j; second, when export occurs, the effects of trade barriers on the volume of traded products (more precisely the value of the log of the relative bilateral intensity of trade RIijk).13 In order to compare the dynamics of the role played by various trade barriers (between 1999 and 2004) a cross-section analysis is used14 and the data for the two years are pooled. The signicance level of the difference between coefcients obtained for the two years is tested through a Wald test,15 while the residuals of the model estimations have been graphically checked for normality. As explained in previous section, we have regressed RIijk to check for possible endogeneity of total exports xik, and total imports mjk on xijk. As an alternative, to check the robustness of the obtained results, the instrumental variable approach has been used. While assessing xijk as dependent variable, xik and mjk were instrumented to take into account possible endogeneity problem. For xik the instruments are GDP of each exporting country i and nc2 product dummies. For mjk, the instruments are GDP, Population, non tariff measures and nc2 product dummies. The two estimation methods have both converged towards the same main conclusions, although slight differences were occurring in results for certain variables according to the estimation method that has been used. These differences are discussed when appropriate. Empirical assessment of the trade resistances for CEECs agrifood exports on EU markets As highlighted by Mayer and Zignago (2005), most of impediments to trade developed in Eq. (3) are hard to measure individually and an assessment of global trade resistance is therefore useful. For this purpose, we use the border effect methodology initiated by McCallum (1995) and subsequently widely employed,16 which captures all impediments to trade related to the existence

of borders through their impact on trade ows. In our case, the principle of the analysis is extended from national borders to the external frontiers of the European Union. Once the global trade resistance is assessed (called here after overall border effect), we investigate, in a second step, the role of tariffs and non-tariff measures in the market access. Third, we analyse more precisely how much NTMs explain the border effect for each group of countries. What is the magnitude of the overall border effect faced by trading partners at entry to the Single European Market? The equation and the data used In this rst step, the regressed equation is thus:

X X ln RIijk a1 ln wijk a2 ln dij a3 Bij ai4 T i ak K k 5 X

j i k

aj6 F j eijk model 117

Data have been aggregated at a 4-digit level in the Harmonized System, resulting in about 200 products; and for each product k, we focus on each country-pair where j is a EU-15 importing country and i denotes the country of origin of imports (European or not). Our sample includes all the 212 EU partners. However, for the two years, our sample includes 587,941 observations. The latter include trade ows in which, for each k product: exporting countries i export at least to one country in the world (i.e. Xik > 0); EU15 importing countries j import at least from one country of the world (i.e. Mjk > 0). Observations include also trade ows between i and j, when simultaneously i and j fulll the above described conditions. The index of relative bilateral intensity of trade RIijk has been computed using the COMTRADE database. In order not to mix imports and exports data in the same index, exports from country i towards country j have been measured by the imports of j from i. In the same way, total exports of country i are the total world imports from i. wijk is the bilateral cost-competitiveness: the ratio between is export price (Pik) and the js import for product k (Pjk). Pjk should normally be a CES index of import price. The average import price of j has been used as a proxy of this index (i.e. the unit value of the total imports of j). For Pik the FOB unit value of the export of i to the market j is used. Because of the lack of data about quantities in the COMTRADE database, these unit values have been computed from the COMEXT database at 4-digit level. The variable dij is a proxy for transport costs between two countries. It represents the distance, calculated by the CEPII18 between capitals of i and j. Two variables have been considered to reect bilateral afnity between trading partners (Bij): the contiguity variable which is a dummy variable equal to 1 if the two trading partners have a common border and equal to 0 otherwise. The common history has been caught through a dummy variable which is equal to 1 if the exporting country was a colony of its trading partner. Ti catches the overall border effect faced by EU partners i, after controlling for transport costs and cultural proximity. More precisely, Ti are four country-group dummies, each of them corresponding to the area the exporting country i belongs to: NMS-8,
17 Recalling Eq. (3) of the theoretical part, (T ) becomes . In regards with theoretical equations (2) and (3), structural value of several coefcients can be deduced: 1=1-; 2=(1- ); 3= (1-) . 18 Available on the CEPII website: http://www.cepii.fr/.
1 ijk

12 For the discussion about zero values in trade matrix see among others Hillberry (2001) or Helpman et al. (2005). 13 As an alternative for capturing the censored data effect (the zero trade ows explained by censoring trade data below a threshold), the model was also estimated by a Tobit procedure. Concerning the volume equation results obtained from the two estimations are similar, although some slight differences were occurring. These differences are discussed when necessary. Detailed results are available upon request. 14 Baldwin and Taglioni (2006) argue, that applicability of Anderson and van Wincoop (2003) multilateral trade resistance factor (see its derivation in Appendix) is limited to cross section data only. 15 Results of the tests are available upon request. 16 See, among others Chen, 2004; Mayer and Zignago, 2005.

600

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606

Table 2 Results of the regression equation of the relative bilateral index ln(IRijk) Model 1 1999 Bilateral competitiveness Distance Contiguity Colony Border effect ROW NMS-8 NMS-2 Tariffs ROW NMS-8 NMS-2 Non-tariff measures Sanitary Phytosanitary Quality Number of observations: Log likelihood Rho Sigma R2 Mc Fadden** 0.16 (0.01) 0.37 (0.01) 1.92 (0.03) 1.33 (0.03) 0.88 (0.02) 1.43 (0.04) 0.43 (0.09) 2004 0.14 (0.01) 0.35 (0.01) 1.85 (0.03) 1.46 (0.03) 1.06 (0.02) 1.44 (0.04) 0.56 (0.08) Model 2 1999 2004

Constrained 0.29 (0.02) 0.92 (0.06) 0.3 (0.13) 0.40 (0.01) 0.30 (0.03) 0.36 (0.07) 0.24 (0.02) 0.21 (0.03) 0.1 (0.02) 0.33 (0.02) 1.06 (0.06) 0.40 (0.1) 0.45 (0.01) 0.21 (0.03) 0.38 (0.06) 0.13 (0.02) 0.07 (0.03) 0.2 (0.02)

Beside these classical variables, the border effect at entry to the EU market is captured separately for the three non-EU geographical zones, in comparison to intra-EU trade. For the three groups, the coefcient is signicantly negative, showing that third countries exchange less with EU countries than EU members among themselves. Moreover, contrary to what might be expected, in both years the border effect for NMS-8 is higher than that for the third countries whilst for Romania and Bulgaria it is much lower. In other words, despite the accession process, trade resistance is still high for NMS-8 just prior to their EU accession. However, this border effect remains stable for NMS-8 over the period while it increases for all other third countries. Does this result reveal a trade diversion effect due to the enlargement process? Actually, the increase of CEECs exports towards the EU has led simultaneously to a decrease in the ROWs share in total EU agro-food imports. Is border effect totally explained by tariffs and NTBs? In order to answer this question, this section seeks to weight different components of the border effect, namely tariffs and various non-tariff measures. Thus, the equation becomes [model 2]20,21

587,941 (uncensored obs.: 114,468) 427370.89 424235.71 0.59 0.58 2.34 2.30 0.26 0.27

X X ln RIijk a1 ln wijk a2 ln dij a3 Bij a0i T i a0k K k 4 5 X

j

Notes: * Non-signicant coefcients. All the other coefcients are signicant at less than 10%. Standard errors are reported in brackets. ** The R2 Mc Fadden is calculated using the reference likelihood (model with only constant). From an econometric point of view, the two modelling steps (selection and regression of Bilateral relative intensity) are not independent (value of rho and of Chi2). The value of Rho justies the use of maximum likelihood method.

0j 6Fj

X
i

i 7 t ik

T i a8 Sank a9 Phytok 6

a10 Qualk eijk

NMS-2 (Romania and Bulgaria), EU-15 (the 15 EU member states) and ROW (the Rest of the Word).19 The product dummies (Kk) are dened at the 2-digit level of the nomenclature (i.e. 24 dummies for agricultural and food products) and thus capture the specicity of the groups of products. Fj are EU importing country xed effects. The index of global competitiveness (Wik) mentioned in the theoretical equation could not be kept for the estimation, due to the lack of necessary data. Hence the impact of this variable is assumed to be spread among the other variables, notably country dummies Ti (border effect). Results Results are shown in Table 2, which presents the second stage of the export process, i.e. the regression on the value of the relative intensity index. Results of the rst step (probit) are shown in Appendix. Main differences between the results of the selection and the regression part will be presented in the text when necessary (notably when discussing the role of non-tariff measures as trade barriers). Results obtained in [model 1] for classical variables are in line with expectations for a gravity model. Distance restricts trade between two countries. Conversely, having a common border and a common history (colony) stimulates trade between partners. Moreover, the bilateral price competitiveness has a signicant impact on trade: the higher the export price of the exporting country compared to the importing price, the lower the volume of exports (Table 2).

where tik is the tariff applied by the EU to its partner i for product k, and Sank, Phytok and Qualk are dummy variables indicating the existence of non-tariff measures for product k (respectively sanitary, phytosanitary and quality measures). The coefcients a0i on coun4 try-group dummies (Ti) represent the estimated impact of the border effect, after tariffs and NTMs have been taken into account. More precisely, they reect the impact of the other trade resistance that includes inertial type factors, such as history, home bias, etc. Tariffs and non-tariff measures data Tariffs are available from the TARIC Database (DG Taxation and Customs Union). Because tik is an ad-valorem equivalent for all the taxes country i has to pay at entry to the EU market, the calculation takes into account preferential agreements where they exist (notably the European agreements in view of the CEECs accession to the SEM) and all the measures applied (i.e. not only the ad-valorem part of the tariffs, but also the specic duties).22 In this step, the impact of tariffs on trade ows is estimated by distinguishing between the different exporting zones. The benchmark for the estimations is the intra-EU trade, which is not concerned by tariffs. Sank, Phytok and Qualk are dummy variables indicating existence for product k non-tariff measures. When traded or entering the EU market, each specic product should be submitted to a wide-ranging set of regulations in addition to tariffs.23 There are about 60 such rules applying to agri-food products, classied by the customs ofces into 15 categories. Among these 15 categories, some are out of the scope of our study: hunting games regulations, protection of wild fauna or ora, etc. Therefore, we kept measures

19 To capture the CEECs heterogeneity and check the robustness of results, border effect was separated country by country for NMS and acceding countries. The results did not change signicantly and are available upon request.

20 In regards with theoretical equations (2) and (3), structural value of several coefcients can be deduced: a1 = 1 r; a2 = q(1 r); a3 = f(1 r); a4 = m(1 r); a7 = k1(1 r), a8 = k2(1 r), a9 = k3(1 r) with ki individual NTB effect. 21 In order to be able to compare coefcients from one model to another and because the coefcients were stable, the different models have been nested by constraining a1, a2, a3 to be equal to those obtained in [model 1]. 22 In order to avoid eliminating from the estimation products not subject to tariffs (ln tik = 8), the variable is transformed in the following way: ln tik = ln(tik + 1). 23 Available on the French customs website (http://www.douane.gouv.fr/).

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606

601

which imply specic compliance by the exporting country, either in terms of production or commercialisation. We have gathered these remaining regulations into three groups, called: (1) sanitary measures (San); (2) maximum residue level of pesticides (Phyto); (3) quality measures (Qual). Each custom regulation was then attributed to one of the three groups, according to its topic. In order to meet sanitary requirements, veterinary controls imposed inside the EU should be accompanied by strict border controls. Sanitary measures therefore include veterinary measures for animals (breeding and production) designed to protect both animal and public health. They are comprised of different categories: animal health, hygiene of food of animal origin, animal feeds and veterinary pharmaceuticals. The rst condition that needs to be met by a third country trying to sell products on the European market is to be on the list of authorized countries.24 Then, when crossing the EU border there are veterinary controls. The phytosanitary measure refers in this paper to standards for the maximum residue level of pesticides (MRL). In order to protect animal and human health, foodstuffs intended for human or animal consumption in the European Union (EU) are now subject to a MRL of pesticides in their composition. Up until 2005, different directives25 were applied to different sectors (fruit & vegetables, cereals, foodstuffs of animal origin meat & some dairy products and products of vegetable origin). The objective was to ensure that pesticide residues in foodstuffs do not constitute an unacceptable risk for consumer and animal health. Quality measures cover the set of standards relating to product quality control, other than sanitary and phytosanitary requirements, for instance commercial characteristics such as freshness, calibration, and conditioning. The products must be accompanied by a quality control certicate. All information must appear on the label, especially for products such as meat, eggs, fresh fruit and vegetables, prepared vegetables, etc. These three categories of regulation have been introduced into the model as dummy variables,26 with a value of 1 if the product is subjected to the regulation, and 0 otherwise. Results Impact of tariffs in line with expectations. Concerning the impact of tariffs, the results of [model 2] in Table 2 highlight the reduction in tariff coefcient for NMS-8, whilst this coefcient remains stable for the ROW. According to the probit estimation, tariffs no longer act as a trade barrier in 2004 (Appendix, Table 4). This is an expected result, since tariffs for NMS-8 were abolished in May 2004, the date of the EU enlargement. However, in 1999 the impact of tariffs was still high because the liberalisation process was far from complete. For NMS-2, the effect of preferential trade agreements is also reected in the Probit estimation. But in the regression part, tariff coefcient is positive for these two countries: the higher the tariff, the higher the export volume. This nding suggests that tariff barriers are not the main determinants of their

agri-food exports towards EU markets. There are at least two sets of explanations for this situation. The rst one is related with agri-food external supply of Romania and Bulgaria, which is low. For Romania, for instance, agro-food exports account for a very small proportion of the national agro-food production, around 5% in 2004. Decreasing the EU tariffs is not enough to encourage CEECs exports, these countries facing rst of all challenges for the production structure, which need more time and important nancial assets to recover. A second explanation is that NMS-2 could take advantage of price competitiveness, since the costs of labour force in these countries were relatively low. Hence, despite tariff duties, products originating from Romania and Bulgaria were still competitive on the European market due to relative lower border prices.27 Specic impacts of non-tariff measures. Regarding non-tariff measures (NTMs), coefcients in the probit selection part equation (2) (Appendix, Table 4 [model 2]) show a signicant negative impact on decision to trade of such variables. Hence non-tariff measures act as a signicant trade barrier for the concerned products. Nevertheless, in the regression part of the estimation procedure, differences among these variables appear. Thus, the impact of sanitary measures is signicantly positive on the volume of trade ow (Table 4 [model 2]), indicating that once the barrier has been overcome, sanitary measures enhance trade. Conversely, phytosanitary and quality measures seem to reduce trade for the products concerned.28 The explanation of this contradictory result (between sanitary and phytosanitary/quality measures) lies more from the ways standards are harmonised across the EU than in the actual content of the measures themselves. The use of regulation in the sanitary eld has led to a higher degree of EU market integration (Ugland and Veggeland, 2006) because each member state has the obligation to apply the regulation text strictly without any national adaptation. Conversely, regarding residue level of pesticides (phytosanitary measures), pan-EU harmonisation was, up until to 2005, carried out through directives allowing member states a certain latitude to transpose this text in their law. Thus the standard on level of MLR may differ from one country to another, which leads to a certain extent trade barriers remain inside the SEM despite the principle of Mutual Recognition and causes a fragmented market, for the products and period concerned. The situation is similar for quality measures which encompass commercialisation and labelling considerations. To sum up, the results obtained suggest that sanitary measures act as trade barriers at entry to the EU market, but once standards are met by exporters they facilitate trade. This result is in line with the assumption made by the OECD (2002), as well as by Barret and Yang (2001) who argue that internationally accepted product standards can facilitate international trade by reducing search and adjustment costs. On the contrary, as far as phytosanitary and quality measures are concerned, national differences in the transposition of regulations increase transaction costs for exporters and in this way increase trade barriers. A remaining border effect after controlling for tariffs and NTMs. Results from [model 2], when tariffs and NTMs have been taken into account, show a reduced, but still signicant, border effect (Table

In the pre-accession period all CEECs were on this list, and were therefore able to export to any European country. 25 Directive 76/895/EEC; fruit and vegetables (Directive 76/895/EEC), cereals (Directive 86/362/EEC), foodstuffs of animal origin (Directive 86/363/EEC) and plant products (Directive 90/642/EEC). In order to harmonise them, these directives have been replaced by a unique regulation in 2005: Regulation (EC) No. 396/2005 of the European Parliament and of the Council of 23 February 2005 on maximum residue levels of pesticides in products of plant and animal origin. 26 Henry de Frahan and Vancauteren (2006) or Haveman and Thursby (2000) did not use a dummy variable, but a trade weighted coverage ratio for each category of standards. Working at our detailed level, choosing the alternative measure would make little difference, because for the majority of products at 4-digit level all the goods within a category are subject to the same standards.

24

27 It should be noted here that in alternative estimation, using Tobit procedure, positive impact of tariffs for Romania and Bulgaria is not conrmed. Tariffs are not signicant in 1999, but they appear as signicant and negative in 2004. 28 These results are conrmed by the test for endogeneity and the Tobit estimation. Nevertheless, in the tobit model, the coefcient of quality measure is signicant at 10%.

602

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606

2). In other words, determinants other than tariffs and non-tariff measures explain trade resistance at entry to the EU market. This is once again particularly true for NMS-8 for which the border effect remains high (1.12 in 2004) though it is heterogeneous among CEECs.29 For instance, for Hungary which is EU market oriented (EU-15 accounted for 56% of its total agri-food exports in 2004), remaining trade resistance was very low (0.24 in 2004). On the contrary, for Lithuania, Slovak Republic, Slovenia and Estonia, the remaining border effect increased between 1999 and 2004 and was much higher than for Hungary. This remaining high level of trade resistance for the CEECs might be an indication of an underperformed agricultural sector in CEECs, especially in some of them. For this hypothesis, further research is needed to compare trade resistance in manufacturing with that of agricultural trade, in order to assess the specicity of agricultural sector in these countries. Nevertheless, one should reveal also the role of history for the geographical orientation of CEECs traders away from the EU (an aspect of global competitiveness or multilateral trade resistance we were not able to measure directly). As shown in Fidrmuc and Fidrmuc (2000), the newly independent countries (of former Czechoslovakia and Yugoslavia) traded much more intensively with their previous counterparts than with other countries, despite greater hidden trade barriers (transaction costs, exchange rate uncertainty and political instability). These results for the former federations in Eastern Europe thus indicate a very strong bias, explaining the trade resistance towards the EU. Hence, as shown in Chart 1, only 29% of Slovene agri-food exports were sold to the EU in the analysed period (27% for the Slovak Republic). How much do NTMs affect the border effect for each group of countries? The aim of this third step is to evaluate whether the considered NTMs have different impact on trade ows with the respect to the product provenance. Thus, the variable, reecting presence or absence of a NTM, is combined with the country-group dummies that indicate the origin of the product. Due to colinearity, the analysis must be done successively for the three NTMs. The following equation presents [model 3a] dealing with sanitary measures impact according to the origin of the product (model dealing with phytosanitary and quality measures are derived in the same manner [model 3b and 3c]). The estimated equation becomes: [model 3a]30

NMS-2, or rest of the world), estimations provide information on the impact of non-tariff measures separately for each region. Simultaneously, the comparison between 1999 and 2004 of such measures highlights the role of the acquis communautaire on CEECs agri-food exports. The impact of sanitary measures Table 3, model 3a, shows the impact of sanitary measures on the EU border effect faced by non-EU countries. Coefcients are estimated with regard to intra-EU trade for which sanitary measures do not act as a trade barrier.31 For NMS-8 countries, conversely to other third countries, sanitary measures do not increase the barriers faced at entry to the EU market (the coefcients of the probit results are not signicantly different for products with and without sanitary measures, Appendix, Table 5). Therefore, in the decision for NMS-8 to export towards the EU, the border effect is not attributable to the presence of sanitary measures. Moreover, for products, submitted to these measures, the volume of trade seems to be higher than for other products, with the impact reinforced between 1999 and 2004. Hence, sanitary standards seem to be met and consequently to stimulate trade. There are two possible, not mutually exclusive, interpretations of these ndings. Firstly, as explained above, a uniform standard across the SEM reduces transaction costs; and secondly, EU importers place increasing condence in NMS-8 products guaranteed by their EU-type standards. Analysing estimations32 for individual NMS-8 country shows that this is particularly true for Hungary, whose products with sanitary measures no longer face trade resistance at entry to EU markets. Given that Hungarian public policy imposes EU regulation as a guideline for its own regulation (Fehr, 2002), this result is not surprising. Results for Poland, Latvia, Estonia and Lithuania indicate the signicantly reduced border effect for products covered by sanitary measures. For NMS-2, these measures still act as important barriers at entry to the EU market (probit regression) while, in the volume equation, the difference between the two coefcients, for 1999 and 2004, is not signicant; meaning that these standards have no (or little) impact on the volume of trade. These results are conrmed by the conclusions of the Standing Committee on the Food Chain and Animal Health (European Commission, 2006), which underline the low level of compliance with EU standards in Bulgarian and Romanian meat and milk establishments (a considerable proportion of raw milk was in 2006 not yet in compliance with EU requirements). Therefore, from 1 January 2007, these two countries were allowed a transitional period to upgrade their production processes. During this period, the produce from noncompliant rms carried a special identication mark and were sold only on the domestic market. A more important border effect would therefore be expected for these NMS-2 products, but taking into account the very low proportion of produce exported, we may assume that Romanian and Bulgarian rms exporting to the EU are in general compliant with EU standards. The impact of phytosanitary measures Phytosanitary measures (MRL) act as a trade barrier at entry to the markets (probit results, Appendix, Table 5) for intra-EU trade

X X j ln RIijk a1 ln wijk a2 ln dij a3 Bij ak K k a6 F j 5 X

i

i 7 t ijk

X
i

0 i;0 8 Sanik

X
i

1 i;1 8 Sanik

a9 Phytok a10 Qualk eijk

San1 ik

Sanik T i Sank . Thus, T i for products subject to a sanitary measure. Hence, ai1 measures the border effect for products subject 8 to sanitary measure imported from i; while ai0 concerns products 8 not subject to sanitary measures originating from i. The difference between ai1 and ai0 evaluates therefore the impact of the sanitary 8 8 measure on the border effect. If ai1 < ai0 , the irrespective measure is supposed to reduce 8 8 trade; if the contrary holds, the trade is stimulated by that measure. Taking also into account the origin of the product (NMS-8,

29 The model 2 was also estimated using country xed effects rather than countrygroup xed effects for the CEECs. Results conrm our general ndings. Results are available upon request. 30 ai0 and ai1 are combinations of previous a8 and a4. 8 8

31 It is noteworthy that according to our estimations, sanitary measures have no impact on intra-EU trade whereas phytosanitary and quality measures hinder intraEU trade. These results are in line with those obtained by Henry de Frahan and Vancauteren (2006) who show that EU harmonisation of food regulation has a major positive effect on intra-EU trade. 32 Available upon request.

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606 Table 3 Results of the regression equation of the relative bilateral index ln(IRijk) Model 3a 1999 Bilateral competitiveness Distance Contiguity Colony Tariffs RoW NMS-8 NMS-2 Non-tariff measures San ROW No San San NMS-8 No San San NMS-2 No San San Phytosanitary ROW NMS-8 NMS-2 EU Quality ROW NMS-8 NMS-2 EU No Phyto Phyto No Phyto Phyto No Phyto Phyto Phyto 0.10 (0.02) No Qual Qual No Qual Qual No Qual Qual Qual 424,139 0.58 2.30 0.27 0.19 (0.02) 0.16 0.37 1.92 1.33 0.43 (0.01) 0.31 (0.03) 0.39 (0.07) 2004 0.14 0.35 1.85 1.46 0.44 (0.01) 0.24 (0.03) 0.37 (0.06) Model 3b 1999 2004 Model 3c 1999 2004

603

Constrained 0.42 (0.01) 0.30 (0.03) 0.37 (0.07) 0.22 (0.02) 0.43 (0.01) 0.21 (0.03) 0.37 (0.06) 0.14 (0.02)

Constrained 0.42 (0.01) 0.30 (0.03) 0.37 (0.07) 0.22 (0.02) 0.43 (0.01) 0.21 (0.03) 0.37 (0.06) 0.14 (0.02)

0.21 0.20 1.00 0.75 0.30 0.19

(0.03) (0.03) (0.08) (0.07) (0.17) (0.15)

0.36 0.44 1.31 0.81 0.50 0.33

(0.03) (0.03) (0.07) (0.07) (0.12) (0.12) 0.23 (0.03) 0.21 0.51 0.92 1.10 0.32 0.43 0.21 (0.03) (0.04) (0.07) (0.09) (0.15) (0.17) (0.04) 0.39 0.45 1.12 1.06 0.54 0.30 0.13 (0.03) (0.04) (0.07) (0.08) (0.11) (0.14) (0.04) 0.20 (0.03) 0.37 (0.04) 0.88 (0.06) 1.04 (0.11) 0.28 (0.14) 0.20 (0.22) 0.01* (0.04) 424,190 0.58 2.30 0.26 0.36 0.57 1.06 1.32 0.38 0.70 0.13 (0.02) (0.03) (0.06) (0.10) (0.10) (0.18) (0.04) 0.06 (0.03)

0.19 (0.03)

0.03* (0.03)

0.10 (0.02)

0.18 (0.02)

Number of observations: 587,941 (uncensored obs.: 114,468) Log likelihood 424,574 Rho 0.58 Sigma 2.30 R2 Mc Fadden 0.26

Notes: * Non-signicant coefcients. All the other coefcients are signicant at less than 10%. Standard errors are reported in brackets. ** The R2 Mc Fadden is calculated using the reference likelihood (model with only constant). From an econometric point of view, the two modelling steps (selection and regression of bilateral relative intensity) are not independent (value of rho and of Chi2). The value of Rho justies the use of maximum likelihood method.

and imports from the ROW. However, for NMS-8 and NMS-2, these phytosanitary measures have no signicant impact on the decision to export to the EU market; and results in the probit estimations are not signicantly different for the two years. That means that imposed phytosanitary standards would not represent major constraint at entry to the EU market for both NMS-8 and NMS-2 in the considered period. Although this result is not totally conrmed by alternative tests for endogeneity, it does not sound inconsistent in the sense that CEECs were small consumers of pesticides in the analysed period.33 According to the European Environment Agency (2004), the mean consumption of pesticides (insecticides, herbicides, fungicides and others) in 1997 was less than 0.75 kg/ha of UAAL for CEECs, while for the EU-15 it was more than 2.25 kg/ha. This situation, due mainly to economic necessity rather than environmental awareness, should give these countries a comparative advantage regarding the pesticide standards. As these standards are not uniform across the Single European Market, their presence still in-

creases trade costs and reduce the volume of trade in the case of intra-EU trade. The impact of quality measures The results show that, for intra-European trade, quality measures have a negative impact mostly on the selection equation, but this impact tends to disappear in the volume equation. Nevertheless, such measures have less impact on intra-EU trade than on other exporters. This result was expected since intra-EU exporters are supposed to have best access to the single market. For third countries as well as NMS countries, applying the quality measures reduces both the decision to export and the volume traded. The same conclusion can be drawn in the case of Romania and Bulgaria. The lack of unication of quality measures, and likely also differences in description of products due to different national languages of EU-15, thus inhibits the volume of internationally traded products in the SEM, whatever the country of provenance. This impact is widely recognised e.g. food labelling, as a part of quality measures is denoted by OECD (2003) as a contentious issue concerning agri-food products, for which particular attention is needed in order to minimise the risk of disruptions of international trade.

33 As recently published by Eurostat (2006), the quantity of commercial fertilisers in 2001 was highest in France, followed by Germany, Spain, UK, Italy, Poland, Ireland, Greece, Netherlands, Czech Republic.

604

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606

Conclusion A specic feature of this study is to identify the hindrances to European market access for agricultural and food products according to the origin of products, notably from Central and Eastern European countries. The rst result of the paper indicates that despite the expected improvements in market access of CEECs agrifood products on the SEM over the pre-accession, CEECs still faced impediments to trade at entry to the EU market just prior their accession. In this context, the aim of this paper is to evaluate how much tariffs and non-tariff measures explain this border effect by using a gravity model and the border effect methodology. The results by components of border effects reveal several interesting ndings. First, concerning the impact of tariffs, the results highlight the reduction in tariff elasticity for CEECs, whilst this coefcient remains stable for the rest of the world. It is noteworthy that tariffs no longer act as a trade barrier anymore in 2004 for NMS-8. This is an expected result, since tariffs for NMS-8 were abolished in May 2004, the date of the EU enlargement. However, in 1999 the impact of tariffs is still high because the liberalisation process is far from complete. Second, the estimations of non-tariff measures coefcients suggest that for NMS-8 sanitary measures do not act as a barrier to trade at entry to the EU market and even signicantly stimulate traded volume for NMS-8 rms fullling sanitary requirements. For NMS-2 these measures still act as barrier to trade, and once the barrier has been overcome, traded volume is slightly increasing. Phytosanitary measures do not seem to act as barrier to trade at entry to the EU market for CEECs products (both from NMS-8 and NMS-2) but still limit traded volume. For third countries as well as NMS-8 and NMS-2, applying the quality measures reduces both the decision to export and the volume traded. The lack of unication of quality measures thus inhibits the volume of internationally traded products in the SEM, whatever the country of provenance. The impact of NTBs on the degree of European market access is less a matter of the specic nature of the NTB than of the degree of SEM harmonisation among countries of the SEM. These results call for two comments. The rst comment deals with the CEECs fullment of the EU requirements (in other word, the efciency of the implementing the acquis communautaire). Sanitary measures have been fullled in NMS-8 leading to a nonsignicant impact of such measures on decision to trade, whereas for NMS-2 it was not the case, justifying therefore the transition period imposed by European authorities. Concerning phytosanitary measures, they do not act as trade barriers for CEECs product rather because the use level of pesticide is structurally low in those countries than because of the implementation of the acquis communautaire. The second comment deals with the opposite impacts of sanitary and phytosanitary measures on traded volume. Results concerning sanitary measures is easily explained because the high degree of food safety policies integration in the EU. Phytosanitary and quality measures, relatively far from unication on the SEM, still limit the volume of trade. Even the intra EU-15 trade is concerned by the negative role of those measures both on the decision to trade and on the traded volume. Finally, results reveal that the border effect is not totally explained by tariffs and non-tariff measures and remaining trade resistance is still signicant. This is particularly true for NMS-8 for which the border effect remains high. This is a signal that other factors beside tariffs and non-tariff measures had played a major role as barrier to CEECs agri-food exports towards EU in the preaccession period, despite the trade liberalisation coming with the preparation of the EU Eastern enlargement. In other words,

trade-related policies are not the only determinants of trade relations between countries and the trade integration of the enlarged EU market was not yet reached in 2004. The understanding of this remaining resistance calls for further research on the specicities of agricultural sector in CEECs and other determinants of trade like information networking, cultural similarities, product image, etc.

Appendix. First Heckman step: the probit model on the decision to trade The Heckman selection model is a two-step estimation in which it is assumed that there is a latent process underlying regression. Hence two equations are dened

lnRIijk a1 ln wijk a2 ln dij a3 Bij Regression equation

X
i

ai4 T i

X
k

ak K k eijk 5

The dependent variable RIijk (the relative bilateral intensity of trade), is observed for the triplet (j = importing country, i = exporting country and k = product) if:

b1 mjk b2 xik b3 dij b4 Bijk Selection equation

X
i

b5 T ijk lijk > 0

where eijk ? N(0, r); lijk ? N(0, r) and q = corr(eijk, lijk). When q 0, the two equations are not independent and standard regression techniques applied to RIijk would provide biased results. The left hand side represents the decision to trade. If it is positive then trade occurs. mjk is the total imports of product k of the EU member j and xik is the total exports of product k of the partner i. These variables aim at controlling for country sizes in the decision to trade. Tables 4 and 5 show the probit results of our models estimations.

Table 4 Results of the probit model on the decision to export Probit Model 1 1999 Total imports Total exports Distance Contiguity Colony Border effect ROW NMS-8 NMS-2 Tariffs ROW NMS-8 NMS-2 Non-tariff measures Sanitary Phytosanitary Quality 0.07 0.25 0.57 0.28 0.68 (0.001) (0.001) (0.003) (0.02) (0.01) 2004 0.06 0.24 0.52 0.25 0.77 (0.001) (0.001) (0.003) (0.02) (0.01) Model 2 1999 2004

Constrained 0.18 (0.01) 0.91 (0.02) 0.63 (0.04) 0.03 (0.003) 0.09 (0.01) 0.04 (0.02) 0.21 (0.01) 0.26 (0.01) 0.08 (0.01) 0.19 (0.01) 0.93 (0.02) 0.58 (0.03) 0.04 (0.003) 0.01* (0.01) 0.03* (0.02) 0.29 (0.01) 0.11 (0.01) 0.16 (0.01)

0.23 (0.01) 1.09 (0.01) 0.77 (0.03)

0.29 (0.01) 0.99 (0.01) 0.70 (0.03)

Notes: * Non-signicant coefcients. All the other coefcients are signicant at less than 10%. Standard errors are reported in brackets.

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606 Table 5 Results of the probit model on the decision to export detailing non-tariff measures impact Probit Model 3a 1999 Total importations Total exportations Distance Contiguity Colony Tariffs ROW NMS-8 NMS-2 Non-tariff measures Sanitary ROW No San San NMS-8 No San San NMS-2 No San San Phytosanitary ROW NMS-8 NMS-2 EU Quality ROW NMS-8 NMS-2 EU No Phyto Phyto No Phyto Phyto No Phyto Phyto Phyto 0.08 (0.01) No Qual Qual No Qual Qual No Qual Qual Qual 0.17 (0.01) 0.07 0.25 0.57 0.28 0.68 0.03 (0.003) 0.08 (0.01) 0.04* (0.02) 2004 0.06 0.24 0.52 0.25 0.77 0.03 (0.003) 0.001* (0.01) 0.03* (0.02) Model 3b 1999 2004 Model 3c 1999 2004

605

Constrained 0.03 (0.003) 0.08 (0.01) 0.03* (0.02) 0.20 (0.01) 0.03 (0.003) 0.01* (0.01) 0.02* (0.02) 0.28 (0.01)

Constrained 0.03 (0.003) 0.09 (0.01) 0.05 (0.02) 0.21 (0.01) 0.03 (0.003) 0.01* (0.01) 0.03* (0.02) 0.29 (0.01)

0.10 0.25 0.91 0.93 0.55 0.71

(0.01) (0.01) (0.02) (0.02) (0.05) (0.05)

0.11 0.38 0.99 0.96 0.47 0.77

(0.01) (0.01) (0.02) (0.02) (0.04) (0.04) 0.26 (0.01) 0.20 0.44 0.95 1.17 0.70 0.82 0.37 (0.01) (0.01) (0.02) (0.03) (0.05) (0.06) (0.02) 0.24 0.32 0.99 1.02 0.69 0.70 0.24 (0.01) (0.01) (0.02) (0.02) (0.05) (0.05) (0.02) 0.17 0.25 0.91 0.97 0.65 0.48 0.12 (0.01) (0.01) (0.02) (0.03) (0.04) (0.07) (0.02) 0.20 (0.01) 0.36 (0.01) 0.94 (0.02) 1.05 (0.03) 0.60 (0.03) 0.66 (0.06) 0.22 (0.02) 0.11 (0.01)

0.27 (0.01)

0.13 (0.01)

0.07 (0.01)

0.16 (0.01)

Notes: * Non-signicant coefcients. All the other coefcients are signicant at less than 10%. Standard errors are reported in brackets. ** The R2 Mc Fadden is calculated using the reference likelihood (model with only constant).

References
Anderson, J.E., van Wincoop, E., 2003. Gravity with gravitas: a solution to the border puzzle. American Economic Review 93 (1), 170192. Anderson, J.E., van Wincoop, E., 2004. Trade costs. National Bureau of Economic Research Working Paper # 10480. Baldwin, R., Taglioni, D., 2006. Gravity for Dummies and Dummies for Gravity Equations. NBER Working Paper No. 12516. 31 pp. Barret, C., Yang, Y., 2001. Rational incompatibility with international product standard. Journal of International Economics 54, 171191. Bureau, J.-C., Beghin, J.C., 2001. Quantitative policy analysis of sanitary, phytosanitary and technical barriers to trade. Economie Internationale, 107 130. Chen, N., 2004. Intra-national versus international trade in the European Union: Why do national borders matter? Journal of International Economics 63, 93 118. Disdier, A.-C., Mayer, T., 2007. Je taime, moi non plus: Bilateral opinions and international trade. European Journal of Political Economy 23, 11401159. Disdier, A.-C., Fontagn, L., Mimouni, M., 2008. The impact of regulations on agricultural trade: evidence from the SPS and TBT agreements. American Journal of Agricultural Economics 90 (2), 336350. European Commission, 2006. IP 06/1641. More veterinary and food safety decisions for after Enlargement agreed. Available at: http://europa.eu/rapid/press ReleasesAction.do?reference=IP/06/1641&format=HTML&aged=0&language= EN&guiLanguage=fr. European Environment Agency, 2004. Agriculture and the Environment in the EU accession countries. Environmental Issue Report No. 37. 55 pp. Eurostat, 2006. Europe in gures. Eurostat Yearbook 2005 online. http://epp. eurostat.ec.europa.eu/portal/page?_pageid=1334,49092079,1334_ 49092794&_ dad=portal&_schema=PORTAL. Evenett, S.-J., Keller, W., 2002. On theories explaining the success of the gravity equation. Journal of Political Economy 110 (2), 281316. Fehr, I., 2002. Meeting EU standards in Eastern Europe: the case of the Hungarian agri-food sector. Food Control 13, 9396. Fidrmuc, J., Fidrmuc J., 2000. Integration, disintegration and trade in Europe: evolution of trade relations during the 1990s. Centre for Economic Research Working Paper 2000-12. 27 pp.

Freudenberg, M., Gaulier, G., nal-Kensenci, D., 1998. La rgionalisation du commerce international: Une valuation par les intensits relatives bilatrales, document de travail 1998-05, Paris: CEPII, 101 p. Haveman, J., Thursby, J.G., 2000. The impact of tariff and non-tariff barriers to trade in agricultural commodities: a disaggregated approach. Purdue University, 21 p. Head, K., Mayer, T., 2000. Non-Europe: the magnitude and causes of market fragmentation in the EU. Review of World Economics 136, 286314. Head, K., Mayer, T., 2002. Effet frontire, intgration conomique et forteresse europe. Economie et Prvision, 7192. Heckman, J.-J., 1979. Sample selection bias as a specication error. Econometrica 47, 153161. Helpman, E., Melitz, M., Rubinstein, Y., 2005. Trading partners and trading volumes. Available at: http://ipc.umich.edu/edts/pdfs/TradingPartners05_0331.pdf. Henry de Frahan, B., Vancauteren, M., 2006. Harmonization of food regulation and trade in the single market: evidence from disaggregated data. European Review of Agricultural Economics 33 (3), 337360. Hillberry, R., 2001. Aggregation Bias, Compositional Change, and the Border Effect. Ofce of Economics Working Paper. U.S. International Trade Commission, No. 2001-04-B: 26 pp. Manchin, M., Pinna, A.M., 2003. Border effects in the enlarged EU area. Evidence from imports to applicant countries. Center for European Policy Studies 8, 79. Mayer, T., Zignago, S., 2005. Market Access in Global and Regional Trade. CEPII Working Paper No. 2005-02, January. McCallum, J., 1995. National borders matter: CanadaU.S. regional trade patterns. The American Economic Review 85 (3), 615623. Nahuis, R., 2004. One size ts all? Accession to the internal market; an industrylevel assessment of EU enlargement. Journal of Policy Modelling 26, 571 586. OECD, 2002. Vue densemble des obstacles non tarifaires: Rsultats des enqutes auprs des entreprises. OECD, Paris. OECD, 2003. Agro-food products and technical barriers to trade: a survey of issues and concerns raised in the WTOs TBT Committee. Olper, A., Raimondi, V., 2008. Agricultural market integration in the OECD: a gravity-border effect approach. Food Policy 33, 165175. Papazoglou, C., Pentecost, E.J., Marques, H., 2006. A gravity model forecast of the potential trade effects of EU enlargement: lessons from 2004 and pathdependency in integration. The World Economy 29 (8), 10771089.

606

E. Chevassus-Lozza et al. / Food Policy 33 (2008) 595606 Subramanian, A., Wei, S.-J., 2007. The WTO promotes trade, strongly but unevenly. Journal of International Economics 72, 151175. Ugland, T., Veggeland, F., 2006. Experiments in Food Safety Policy Integration in the European Union. Journal of Common Market Studies 44 (3), 607624.

Peridy, N., 2005. Towards a new trade policy between the USA and middle-east countries: estimating trade resistance and export potential. World Economy 28 (4), 491518. Sales, A.E., Plomondon M.E., Magid, D.J., Spertus, J.A., Rumsfeld, J.S., 2004. Assessing response bias from missing quality of life data: the Heckman Method. Health Quality Life Outcomes 2:49. Published online 2004 September 16.