EU Market Access for Mediterranean fruit and vegetables A gravity model assessment

Our model, based on the new developments of the gravity trade model, focuses on thedifficulties Mediterranean countries face in entering the EU market, compared to other EUpartners, cons

EU Market Access for Mediterranean fruit and vegetables:

A gravity model assessment

Charlotte EMLINGER1, Emmanuelle CHEVASSUS LOZZA2, Florence JACQUET 1

1 IAMM/UMR MOISA, Montpellier, France

2 INRA Nantes, France

Abstract Since 1995, a liberalization process - the so-called Barcelona Process - has taken

place in the Mediterranean area Its aim is to establish by 2010 a free trade area in theMediterranean Basin For the moment full liberalization concerns trade in industrialproducts, but agriculture remains sensitive Among agricultural products, the fruit andvegetables (F&V) sector is essential for Mediterranean countries, and the EU is their majortrading partner In this context, two questions arise: firstly, to what extent does protectioninfluence trade for the Mediterranean countries compared to other countries? Secondly,what impact would greater liberalization in the F&V trade between the EU andMediterranean Countries have?

Our model, based on the new developments of the gravity trade model, focuses on thedifficulties Mediterranean countries face in entering the EU market, compared to other EUpartners, considering the relative impact of the different trade costs The model is estimated

at the product level, in a sector which is highly specific: some products may be veryperishable and thus particularly time-sensitive The Mediterranean basin appears as a highlyheterogeneous country bloc Beside the actual level of preferences allowed by the EU, twomain elements vary according to the exporting country: its tariff sensitivity and its “non-tariff” trade resistance Thus, with respect to Euromed liberalization, the higher the tariffsensitivity, the higher the impact liberalization has on trade, and this impact can be limited

by a high trade resistance (NTB, logistic constraints…)

Key Words: Fruit and Vegetables, EU-Med agreement, gravity models, transport cost,

tariffs

JEL: F13 F17 Q17 Q18

Since 1995, a regional liberalization process - the so-called Barcelona Process - has takenplace in the Mediterranean area It aims to establish by 2010 a free trade area in theMediterranean Basin However this trade liberalization is progressing very slowly, notablybecause the agricultural sector has been largely exempted from the process In November

2005, a new impetus was given to the Barcelona Process, with the establishment of astrategy to reinforce liberalization in the agricultural sector

For the Mediterranean Partner Countries, the main concern is better access for their fruitand vegetable exports to the European market These products represent the main exports ofthese countries, and the European Union is their major trading partner On the other side,for the European Union, the main issue is not only the promotion of its cereal and dairyexports but also the protection of its producers of fruit and vegetables Indeed, theregulation of trade with third countries is the key element of the Common MarketOrganization of the sector It has several objectives, the first being of course the protection

of European producers in a sensitive sector, where production is most often highlyseasonalized and where perishable products are difficult to stock

For agricultural products, the agreements define preferences at entry to the EU market onlyfor fresh and processed F&V and they provide limited concessions for each partner forprecise products and limited quantities and calendars However, despite those preferences,Mediterranean countries still have to face important trade barriers when exportinghorticultual products to the European market

Within this context, an important issue is to assess the likely impact of greater liberalization

on Mediterranean F&V exports to the EU (Garcia Alvarez Coque 2002) Most of thecurrent studies on the topic are based on simulation models with general (Kuiper 2004,Doukkali 2003) or partial (Eruygur and Cakmak 2005, Chemmitz Grethe 2005) equilibriumframework and are the most often calibrated on one specific country (Morocco, Egypt orTurkey) However, in order to catch in the simulations the likely impact of a decrease intariffs, it is first necessary to evaluate the extent to which European protection influencesthe F&V trade from MED countries Thus, the objective of the paper is to analyse the maindeterminants of European market access for fruit and vegetables by using a gravityequation It focuses on the constraints faced by each Mediterranean country at entry to the

EU market These “trade costs” (Anderson and Van Wincoop 2005) include both transportand border-related costs (tariffs and non-tariff barriers, information costs, border formalitycosts…)

Our econometric estimations include 55 fruit and vegetable products This verydisaggregated level of commodity observation allows the heterogeneity among products to

be taken into account This heterogeneity appears at two levels: first, the nature and theintensity of EU protection and, second, the degree of product perishability Perishability isone of the most significant specificities of the sector, and it has been assumed that the moreperishable the products the higher the transport costs Thus, contrary to the majority ofempirical literature using gravity equation, in this paper, transport costs vary not only withthe distance but also according to the products Finally, another originality of our approach

is to compare the determinants of trade for the different Mediterranean countries This is all

the more important because the Barcelona process is not a regional agreement per se, but

the sum of ten bilateral agreements signed between EU and each Mediterranean country,and the conditions of access to the European market differ from one country to another.The remainder of the paper proceeds as follows: Section 1 first presents the Mediterraneancountries position as suppliers of fruit and vegetables to the European Union (EU15), andthen compares tariffs and preferences allowed by the EU for these different suppliers.Section 2 presents the theoretical foundation of the gravity model, based on Anderson andVan Wincoop (2003) This model allows the Mediterranean Countries access to the EUmarket to be compared with access for European producers and to other third countries.After a presentation of data and econometric methodology implemented in the third part,the fourth part provides results, a major result being the heterogeneity amongMediterranean countries concerning access conditions to the European Market Finally,section 5 concludes.

1 Market access for fruit and vegetables coming from Mediterranean countries

The Mediterranean countries involved in the Barcelona Process (Algeria, Egypt, Israel,Jordan, Lebanon, Libya, Morocco, Syria, Tunisia and Turkey1) are important tradingpartners of the EU with a market share of 4.8% (Table 1) Their main exports are hazelnuts,dried fruits and citrus, but also tomatoes and other vegetables Among the 10 MEDcountries, alongside very small exporters such as Algeria or Lebanon, four countries -Turkey, Morocco, Israel and Egypt – play a major part in the F&V trade (Table 2) Theyaccount for more than 95% of the F&V exports of the area Beside MED countries, those ofthe Southern Hemisphere (Chile, Uruguay, Argentine, South Africa, Kenya, New Zealandand Australia) are also important non-EU exporters (3.14%) to the European Union (apples,grapes)

However, the EU members remain the main suppliers of the European market, providing77% of the EU F&V imports Is this phenomenon only due to the abolition of tariff barriersbetween EU countries or do other determinants explain the trade within the Europeanmarket? In other words, what explains the EU border effect (as Mayer, 2002, calls it) fornon-EU countries at entry to the European market? More precisely, concerningMediterranean countries, is their market share explained by the level of preferencesprovided by the framework of the EU-MED agreements?

Table 1 World and European Union suppliers of fruit and vegetables in 2003

dollars

percentag e

Million dollars

percentag e

Source: COMTRADE database

Table 2 Mediterranean World and European Union suppliers of fruit and vegetables

in 2003

Exporters Million dollars percentage Million dollars percentage

Source: COMTRADE database

Despite the Barcelona process, Mediterranean countries don’t seem to have, on average,high preferences for their access to the European market In 2003, European tariffs applied

to Med Countries were, on arithmetical average, a little higher (8.7%) than those applied tothe NMS (8.4%) and to countries from the rest of the world (5.2%) (Graph 1) However,analysing preferences at the country level reveals heterogeneity among the countries andhuge preferences for certain countries

Graph 1 : European tariffs applied by area Arithmetical Average tariffs - for fruit and Vegetables 2003

Error: Reference source not found

source : TARIC Database

Indeed, despite the fact that the Barcelona process is commonly presented as a regionalagreement, it is really a set of bilateral agreements with each of Mediterranean countriesand the state of progress in negotiations differs from one country to another For instance,the agreement with Tunisia was signed as early as June 1995, Libya has for the moment anobserver status but no trade agreements have been signed, and negotiations with Syria areongoing Finally, other countries such as Morocco, Egypt and Israel have alreadyrenegotiated their initial trade agreement Within the framework of the negotiations for EUmembership, Turkey has signed a Customs Union agreement with the EU, in continuation

of association agreements signed as early as 1963

Moreover, even if association agreements have been signed, not all products are concernedand some of them may benefit from other types of preferences (allowed notably in the

framework of the GSP regime) (Chevassus-Lozza et al, 2005) Thus, for the MED countries

taken as a whole, the EU-Med preferences account on average for only 26% of the tarifflines in the F&V sector, while 47.3% of their tariff lines may benefit from the GSP regime.Consequently, to estimate MED countries access to the EU market, all the preferencesallowed to these countries may be taken into account

Since preferences are negotiated for each product and each Mediterranean country, there isgreat heterogeneity in tariff regimes among Mediterranean countries (Table 3) Turkey andLebanon essentially have bilateral preferences (85% and 67% of tariff lines), and Turkeydoes not benefit from any GSP preferences The situation is very different with Israel,where 83% of tariff lines are submitted to the MFN regime without any preference

Table 3 Repartition of tariff lines (CN10) by country and tariff regimes for fruit andvegetables 2003

Graph 2.

European tariffs applied to Med countries Arithmetic

average for fruit and vegetables- year 2003

0 2 4 6 8 10

COMTRADE and MEDITAR

Finally, if these very first results on trade and tariffs are rapidly put side by side, it appearsthat they are not systematically linked For instance Israel, which is a major exporter to theEuropean market still has to face high tariffs and doesn’t benefit from huge preferences;whereas Lebanon benefits from high preferences and low tariffs but has a very lowEuropean market share and its exports are largely not EU-oriented On the other hand,Morocco and Turkey both have large European market shares and low tariffs on average.Other components should explain trade to the EU To answer these questions, a gravity-type model has been used, the derivation of which is presented in the following section.Another key point of this part is the fact that preferences are negotiated for each countryand each product separately For this reason our empirical work is conducted at adisaggregated level, product level and country level

2 The Gravity Model

The Gravity-type model is a widespread model in international trade analysis which permitsanalysis of bilateral trade volume and nature It is applied for various purposes, but it isparticularly used to assess market access, trade resistance and the impact of regionalagreements Indeed, it allows trade creation or diversion to be estimated in the case of aregional agreement (Nahuis 2004, Soloaga and Winter 1999), and thus makes an importantcontribution to the regionalism debate On the other hand, “border effect” methodology(Chen 2004, Head and Mayer 2004, Mayer and Zignago, 2005) provides analysis of amarket access measurement comparing imports from foreign countries to intra-nationaltrade in order to have a benchmark of best possible market access, that faced by nationalproducers

Our model is based on the new developments of the gravity equation made by Andersonand van Wincoop (2003) It has been assumed that consumers have identical andhomothetic preferences and that products are differentiated by origin The representativeconsumers in country i maximize a CES utility function Uik:

1 1

U (1)Under the following budget constraint:

  

j

ik j

ijk ijk

p (2)

We denote i the importing country, j the exporting country, k the product, cijk theconsumption by i of product k from j and bjk consumers’ preference for products k comingfrom j σ corresponds to the elasticity of substitution of imports of j Pijk is the price of good

k coming from country j paid by consumers in country i, mik is the country i expenditure forgood j Pijk differs from price in country of origin pjk due to trade cost tijk that are not directlyobservable We follow the iceberg assumption about trade costs, giving:

ijk jk

p  (3)The maximization of (1) under constraints (2) gives the bilateral imports by country i fromcountry j for a given good k:

ik ik

ijk jk jk

P

t p b m

Pik is the country i’s CES price index for good k:

P (5)The general equilibrium structure of the model imposes market clearance Bothinternational and intranational trade are being considered, so with xjk production of good k

by country j, market clearance gives:



ijk ijk

x (6)Applying the equation (5) to this market clearing condition, we obtain:

ijk jk jk

P

t p b x

 1

ijk jk

jk jk

m P

t p

x

1 1

(8)Substituting this expression of bjk (8) in (7), it yields

ijk jk ik

ijk jk ik jk ijk

m P

t p P

t p m x m

We differ from Anderson and van Wincoop in not simplifying this equation assuming that

ijk jk

m m P

p A



1

with mwk total expenditurefor product k in the world It corresponds to a CES index of price competitiveness of j forthe good k This index assesses the global competitiveness of country j on the totaldestination markets 

ijk jk ik jk ijk

m A P

t p m x

That gives:

ijk jk ik

ijk jk wk

jk ik

ijk

RI A P

t p

of i to the market share of the exporter j in the international market An index equal to 1means that the flow of good k between i and j is only determined by the size of the partners

A coefficient different from 1 means that trade is determined by other factors than the size(equation 11): if it is greater than one, it denotes privileged trade links between i and j forgood k whereas an index less than one refers to trade resistance between the two countrieswhich could be explained by the low competitiveness of i, but also by the trade costs Trade costs tijk are defined to include all costs incurred in getting a good to a final user otherthan the production of the good itself (Anderson van Wincoop 2004) These costs comprisetransport costs, tariffs and non-tariff barriers, but also information costs, the use of differentcurrencies or marketing cost The main problem is to measure these costs for which data arenot always available So this means that trade cost needs to be cpatured by observable costproxies

We follow Péridy (2005) or Chevassus-Lozza et al (2005) and decompose trade costs intodifferent factors: the distance dij between i and j (proxy of transport costs), tariffs applied by

i towards j for good k tijk and other border variables Bijk that are traditionally used in gravitymodels in order to take into account information costs and other elements that cannot bemeasured, such as a common language, common frontier, and common history Aperishability variable is added in order to catch this major specificity of fruit and vegetablesproducts

3 Data and econometrics

The above theoretical development leads to the estimable gravity equation:

) ln

)(ln 1

( ) ln(

) ln(

)

1

(

) ln(

) ln(

)

ln(

j k

ij ij

ij ijk

jk ik

jk

wk

jk ik

ijk ijk

zone Periss

Colony Contig

d tariffs

A P

p

m

x m

m RI

dependent variable includes both international (m ijk ) and intra-national flows (m iik );however, the latter are not available at so disaggregated a level It was thus necessary togenerate these flows from the data on production (coming from FAOSTAT) and trade(coming from COMTRADE database) For this, the balance sheet between supply anddemand for each product and country has been computed This needs specific attention to

the consistency between the two databases, taking into account the problem of exportation which entails, for example, that some countries without production can present

re-a lre-arge re-amount of exports for some products

Bilateral relative prices (

ik

jk

P

P ) are calculated from production price data of FAOSTAT

database for each country and product Nonetheless, as data needed to calculate the index ofglobal competitiveness (Ajk) are not available, this variable has not been introduced into ourestimation

For the transport costs between two countries – we have taken as proxy the distance

between the capitals of i and j d ij and the internal distance calculated by the CEPII 2 Because of the time sensitivity of fruit and vegetable, these transport costs must be a hugeconcern in this sector; and the more perishable the product, the higher the costs Thus, inaddition to distance a multinomial variable has been introduced, corresponding to thedegree of perishability of the products Four groups have been created, using data on timekeeping, respiratory intensity, and fragility from the least (group 1) to the most perishable(group 4) (Appendix 1)

As far as the contiguity variable (Bij) is concerned a dummy variable has been introduced, equal to 1 if the two trading partners have a common border, otherwise equal to 0 The

common history has been caught through the dummy colony equal to 1, if the exporting

country was a colony of its trading partner

In order to take into account all the preferences granted, duties included in the estimation

are tariffs applied by the EU to each of its trading partners The data come from TARIC

database (DG Taxud) Although the model is estimated on annual data and for the FAOnomenclature, the protection needs to be measured at the most disaggregated level in order

to have a comprehensive picture of the protection: i.e monthly data at the 10-digit level ofthe combined nomenclature This allows variations in tariffs during the year due to theseasonality of protection and the different calendars of preferences to be included.Moreover, the calculation of ad-valorem equivalent may be problematic in the F&V sector,due to the so called “entry price system” applied to some sensitive products such astomatoes, cucumbers or citrus fruits This system implies that the level of protectiondepends on the level of the import price If the import price is greater than a threshold – thetrigger price – the exporter pays the ad-valorem part of the duty only If the price is belowthe trigger price, the exporter also has to pay a specific duty This duty is highest when theprice falls below a certain level, equal to 92% of the trigger price Consequently, themeasurement of the ad-valorem equivalent requires an import price to be chosen Here, inthis paper, we have chosen to measure the protection at its maximum level, i.e when theimport price is 92% of the trigger price Finally, for these specific products, preferencesallowed by the EU may be either an exemption or a reduction of the ad-valorem tax, thelevel of the specific duty remaining the same Morocco has negotiated lower entry pricesfor some products (tomatoes and oranges), and preferences allowed for this country arehigher In order to catch this preferential advantage for these products, the ad-valoremequivalent has been calculated on the Morocco prices Finally, once the ad-valoremequivalent is calculated at the most disaggregated level for each country, it is necessary toaggregate this monthly data calculated at the 10-digit level of the combined nomenclature

in annual data defined in the FAO nomenclature

Insofar as one of our objectives is to assess the impact of different trade barriers and, moreprecisely, to point out those which prohibit trade, we must take into account not only the actualbilateral trade but also “zero values” i.e all potential bilateral flows The zero values found in thetrade database in fact correspond either to genuinely zero flow, or to a flow beneath a certainreporting threshold The latter are very low and therefore assimilated to an absence of trade.Consequently, the effects of the different trade barriers are tested at two stages of the exportprocess: firstly, the fact that the country exports or not in a specific market; secondly, in thecase of export, the log of the Relative Bilateral Intensity of trade RIijk From theeconometric point of view, these two stages are not independent The log of the RelativeBilateral Intensity of trade is a truncated variable There is a selection bias

In order to correct this bias, HECKMAN (1979) has built a two-step estimation in which heassumes that there is an underlying regression relationship He defines two equations:

1

u V RI

log ijk ijk [Regression equation]

The dependent variable RIijk (the Relative Bilateral Intensity of trade), is observed for thetriplet (i= importing country, j=exporting country and k=product) if:

Z ijk u2  0 [Selection equation]

Where

) ,u corr(u

N u

N u

2 1 and

) 1 , 0 (

) , 0 (

2 1

Heckman proceeds in two steps:

First of all, he estimates the following probit equation

EXP ijk  1 if Z ijk u2  0

In the second step - i.e the regression on the Relative Intensity - the Mills ratio isintroduced along with the other explicative variables This second step consists, inHeckman’s seminal study (1979), of ordinary least squares But although it leads toconsistent estimates, they are inefficient and there is heteroskedasticity in the errorvariances In this paper, instead of OLS estimations we have used maximum likelihoodestimations when regression estimates using the Mill’s ratio provide starting values forMLE (GREENE, 1997) Finally, the Huber-White estimator of the variance is used instead ofthe conventional MLE estimation in order to obtain robust estimations of the variances.The explicative variables of our regression estimate are those presented in equation 12 Forthe probit, we explain the fact that the countries do or do not export in a specific market.The probability of exporting depends upon the production level of the exporting country