Scholarly article on topic 'Economic Integration and the Two Margins of Trade: The Impact of the Barcelona Process on North African Countries' Exports'

Economic Integration and the Two Margins of Trade: The Impact of the Barcelona Process on North African Countries' Exports Academic research paper on "Economics and business"

0
0
Share paper
Academic journal
J Afr Econ
OECD Field of science
Keywords
{""}

Academic research paper on topic "Economic Integration and the Two Margins of Trade: The Impact of the Barcelona Process on North African Countries' Exports"

Journal of African Economies, Vol. 21, number 2, pp. 228-265 doi:10.1093/jae/ejr038 online date 13 November 2011

Economic Integration and the Two Margins of Trade: The Impact of the Barcelona Process on North African

Countries' Exports n

Sami Bensassi3-*, Laura Marquez-Ramosa-b and |

Inmaculada Martinez-Zarzosoa-b-c °

aDepartment of Economics, Universitat Jaume I, Campus del Riu Sec, 12071 Castellon, t

Spain j

bInstitute of International Economics, Universitat Jaume I, Spain .

cDepartment of Economics, University of Gottingen, Germany f

* Corresponding author: Sami Bensassi. E-mail: bensassi@eco.uji.es c

Abstract "S-

According to recently developed models of trade based on imperfect com- B

petition and heterogeneous firms, lower trade costs increase bilateral trade, not only through a rise in the mean value of individual shipments (the intensive margin of trade), but also through an increase in the number of exporting firms (the extensive margin of trade). The main aim of this paper is to provide new empirical evidence on the effects of the Euro-Mediterranean (EuroMed) agreements on both margins of trade. Using highly disaggregated export data for four North African countries (Algeria, Egypt, Morocco and Tunisia) over the 1995-2008 period, we estimate the impact of the EuroMed agreements on both trade margins, thus providing empirical evidence about the validity of theoretical predictions. Our results show that North African countries enjoyed significant positive returns from the Barcelona Process, through increased exports of manufactured products to the four most populated continental countries in the European Union.

JEL classification: F10

# The author 2011. Published by Oxford University Press on behalf of the Centre for the Study of African Economies. All rights reserved. For permissions, please email: journals.permissions@oup.com

1. Introduction

In recent times, a growing body of economic literature has highlighted the importance of export diversification as part of an export-led growth strategy (Das et al., 2007). For many developing countries, export diversification is conceived of as a shift from traditional to non-traditional exports. By providing a broader export base, diversification can lower instability in export earnings, expand export revenues and enhance economic growth through multiple channels (Acemoglu and Zilibotti, 1997; Gutierrez de Pineres and Ferrantino, 1997). In the light of the success of the Asian 'Tigers', it is now widely recognised that a successful development strategy involves adopting a number of selective measures aimed at reducing transaction costs and improving local business conditions and market access. Trade negotiations at the bilateral or regional level may help reduce market access constraints and open opportunities to tap into regional and global production and distribution chains.

The recent emphasis that trade theories based on heterogeneous firms put on the differentiation between the extensive (the number of varieties exported) and the intensive margins (the quantity of every variety exported) offers a new way of establishing the microeconomic foundations of trade diversification. Here, we explore those foundations along similar lines, and particularly the role played by preferential trade agreements (PTAs) on the diversification of exports. The main aim of this paper is to provide new empirical evidence of the effects of the Euro-Mediterranean (EuroMed) agreements on both margins of trade. To achieve this, we have estimated a theoretically justified gravity model, using data on exports from four North African (NA) countries (Algeria, Egypt, Morocco and Tunisia) to the four most populated continental countries in the European Union (EU): Germany, France, Italy and Spain between 1995 and 2008.

Regarding the existing literature, our main contribution is to disentangle the effects ofEuroMed agreements on the intensive and extensive margins of trade (Chaney, 2008). The empirical evidence shows that some EU trade preference regimes for developing countries, including the EuroMed agreements, have a positive effect on developing countries' exports (Persson and Wilhelmsson, 2006; Blanes-Cristobal and Milgram-Baleix, 2010). As Blanes-Cristobal and Milgram-Baleix (2010) show, the EuroMed trade liberalisation process has had a positive effect on trade between Spain and Morocco. However, other preference regimes, such as the Everything But Arms (EBA) regime, appear to have an insignificant, or even negative, effect on developing countries' exports (Gamberoni, 2007; Gradeva and

Martínez-Zarzoso, 2009). One of the explanations provided is that the Rules of Origin (RoO) are more restrictive than those of previous preference regimes applied to eligible countries, and therefore the regime is underutilised.

Most of the above-mentioned studies, with the sole exception of Gamberoni (2007), rely on trade theories that assume that all products are exported to all destinations. They do not take into account the new-new trade theories, which consider firm heterogeneity and productivity differences among firms to determine what firms are exporters (Das et al., 2007; Eaton et al., 2007). Based on these theories Gamberoni (2007) decomposes the total value of trade into the extensive margin and the intensive margin and then estimates the effects of trade preferences on each margin. Interestingly, the main findings indicate that the African, Caribbean and Pacific and the EBA regimes decrease trade (conditional on trade being present) by 11 and 19%, respectively, and also that both regimes decrease the number of products traded. This effect implies an anti-diversification bias effect in these preferences. Although our work is closely related to that of Amurgo Pacheco (2006), Gamberoni (2007) and Amurgo Pacheco and Pierola (2008), our empirical model is estimated using more recent data. It is built on a newer, more comprehensive methodology which follows the decomposition of trade proposed by Hillberry and Hummels (2008) and is based on Chaney's (2008) theoretical model.

Our main hypothesis is that tight RoO can be considered a hidden fixed cost to trade, as they can limit the use of intermediate goods from countries outside bilateral agreements and hence put a fixed and, sometimes, overwhelming price premium on these goods. According to the new-new trade theory, a decrease in the fixed cost associated with trade allows new firms (less productive than the ones already present on the export market) to engage in exporting activities, thus increasing the extensive margin. Additionally, a growth in the intensive margin of trade could also occur, as the lower the hidden fixed cost is, the higher the exports of each exporter (Hypothesis 1). Das et al. (2007) show that both the extensive and intensive margins change differently across industries in a dynamic structural model of export supply. Continuing in this line, we test whether the effect of the EuroMed agreements differs across sectors (Hypothesis 2). As there are significant differences among the four NA countries considered in the empirical analysis,1 country-heterogeneity is

1 These countries differ in several respects, among them the pace at which financial reforms have been implemented (Tahari et al., 2007).

tested with Hypothesis 3. We also analyse directly whether an improvement in the RoO promotes exports from EuroMed to EU countries (Hypothesis 4), and its effect in the different countries is taken into account with Hypothesis 5. Finally, we expect that the Barcelona Process may have significant consequences on vertical 'production networks', more specifically, an increase in imports of intermediates from EU countries may lead to an increase in exports from NA countries (Hypothesis 6).

The rest of the paper is organised as follows: Section 2 describes the two phases of the EuroMed agreements and related stylised facts. Section 3 presents the theoretical framework; and Section 4 presents the main hypotheses of the study. Section 5 presents the data, the estimated model and the main results. Section 6 concludes and presents a number of policy implications.

2. The EuroMed agreements

In 1995, the EU and fourteen countries of the Mediterranean basin decided to commit to a deeper economic integration by signing 'new generation' integration agreements. This commitment was named the Barcelona Process. Fourteen years later, seven of the signing countries have already enforced the agreement (Table 1), two of them have joined the EU (Malta and Cyprus), two are candidates (Croatia and Turkey), and a new country, Libya, joined the process in 2000. Table 1 presents a summary of the trade integration process between the EU and the EuroMed countries.

The first cooperation agreement including a PTA between the EU and the Middle East and North Africa (MENA) countries dates from the end of the 1970's. The PTA signed was asymmetric since the EU removed taxes on industrial products imported from the signatory countries, whilst the latter maintained their trade barriers in order to protect their developing industries and to obtain revenues from customs duties. Trade relations between Mauritania, Libya and Syria and the EU remain currently regulated by these agreements signed 30 years ago. For the remaining countries, the new agreements add two important novelties. First, the agreements open MENA markets to EU products. Signatory countries must relax all tariffs paid on industrial products imported from the EU over a 12-year period.2 The agreement specifies a calendar for each type of product. Second, RoO applying to signatory countries have been

2 With the exception of Israel, which opened its market to EU's industrial products in 1989.

Table 1: Evolution of Trade Integration in the EuroMed Region

Country PTA Med to EU Commitment to the Barcelona Process Enforcement of the new Cooperation Agreement PTA EU to Med

Algeria 1978 1995 2005 2017

Egypt 1978 1995 2004 2016

Israel — 1995 2000 1989

Jordan 1978 1995 2002 2014

Lebanon 1978 1995 2006 2018

Libya 1978 2000 No No

Mauritania 1978 1995 No No

Morocco 1978 1995 2000 2012

Palestinian — 1995a 1997a 2001a

territories

Syria 1978 1995 No No

Tunisia 1978 1995 1998 2008

aThe agreement with the Palestinian Authority is a transitory agreement which, due to the political situation, has not been applied.

modified. In prior agreements, these rules were particularly tight (Hoekman, 1998; Francois et al., 2005), as only products produced entirely in signatory countries or incorporating EU-made spare parts could enter the EU duty-free.

To understand the impact of the EuroMed process on exports from the Mediterranean countries to the EU, we must take a closer look at these agreements and their evolution over time. We differentiate between the effects resulting from an increase in EU openness to Mediterranean products inherent to the process (first stage) and the effects resulting from increased openness of Mediterranean countries to European products (second stage).

Industrial products originating in the Mediterranean countries have been authorised to enter the EU free of customs duties since 1978. Only marginal changes have occurred thereafter. For example, certain provisions established in 1978 concerning the UK and Ireland have disappeared. It is mainly regarding agricultural products (raw products and processed foods) that important changes may be expected from both sides of the Mediterranean in terms of tariffs. The 1978 agreements between the EU and the MENA countries excluded agricultural products and processed agricultural products from tariff reductions. The EU still imposes tariffs and quotas on a number of important NA agricultural products (tomatoes,

olive oil, fruits and vegetables) (Protocols 1, 2 and 3 of the EuroMed agreements stipulate the exclusion of agricultural products from tariff liberalisation). It is only very recently that progress has been made with a number of partners. At the end of 2009, the EU and Morocco concluded an agreement concerning the liberalisation of trade in agricultural products (European Commission, 2010, COM 2010 485). A similar agreement is currently being negotiated with Tunisia. Given that the advances concerning trade Q

liberalisation for agricultural and processed food were scarce until very n

recently, in this study we focus on manufactures, leaving these important g

issues for further research. LL

Initial assessments of the economic effects of EuroMed agreements were |

particularly controversial. Deardorff et al. (1996), Deardorff (1999) and t

Hoekman and Konan (1999, 2005) held that the outcome of these agree- j

ments could be negative for the MENA countries in terms of trade, £

growth and revenue, at least in the short term. The authors argued that r

lower import duties would lead to revenue losses for MENA countries, U

and that domestic consumption would be diverted towards increased EU §

imports. It is worth noting that these authors disregarded the effects of r

the new and more flexible RoO included in the agreements. As exports "

of industrial goods to the EU were already free of tariffs before 1995, we £

pay particularly close attention to the effects of changes in the RoO adopted in the new agreements. The determination of the geographic origin of the products is crucial and could hinder any attempts at real integration. In this sense, the RoO adopted during the Barcelona Process have modified the prior 1978 agreements. According to the new RoO, a product wholly obtained from, or entirely produced in, a given country is considered originating from that country. For a product produced in more than one country, the product is considered originating in the country where the last substantial transformation took place. The EU's most commonly used rule is that a substantial transformation takes place when there is a change in the product tariff classification line. An alternative criterion is that the value of the intermediate good originating outside the FTA must remain below a certain percentage (often between 40 and 50%) of the value of the final good, or that a particular production process is used to transform the product. The main novelty introduced by the new agreements is the so-called diagonal cumulation, which is one of the three main types of cumulation. The other two forms are bilateral and full cumulation. Bilateral cumulation means that two countries within the agreement can use each other's materials without any limitation. All FTAs allow for bilateral cumulation. Diagonal cumulation means that materials originating in

a third country also linked by an agreement to one of the signatory countries could be used without any limits by the other signing country. If Spain, for example, has an FTA with Iceland and signs an FTA with Morocco which includes the possibility of diagonal cumulation between Iceland and Morocco, intermediate products from Iceland used as intermediates in a Moroccan good are considered originating from Morocco. Finally, full cumulation allows intermediate processing to be split in any way between the parties of the FTA, provided that, when added together, all inputs used are sufficient to fulfil the RoO (Karray, 2003; Augier et al., 2005). Full cumulation is currently operated by the European Economic Area (EEA) and between the EU and Algeria, Morocco and Tunisia. Table A4 shows how the rules concerning cumulation possibilities have evolved over time for the Mediterranean countries (Protocols 3, 4 and 6 of the EuroMed agreements).

Moreover, the Barcelona Process encourages the Mediterranean countries to further integrate their service sectors (transport and finance sector, for example) and to homogenise their procedures (standardisation, metrology, quality controls and conformity assessment) with EU members. These measures should lower transaction costs between the EU and its partners. Except for the signing of the open sky agreement between the EU and Morocco in 2007, little progress has been made in this area (European Commission, 2009, Country Reports on Neighbourhood Policy 2008). Finally, European products will have duty-free access to the Southern and Eastern-Mediterranean markets once the negotiated transition period is over—this is a 12-year period during which customs duties are progressively abolished.

3. Heterogeneous firms and the two margins of trade

A major concern in the traditional literature on the formation of PTAs has been whether these areas generate welfare gains for the individual countries that engage in these processes. Since the 1950s (Viner, 1950), many authors have contributed to this debate, especially in the 1990s when studies based on the gravity model proliferated (Eichengreen and Frankel, 1995; Frankel et al., 1996, 1998; Soloaga and Winters, 2001). Indeed, the effect of FTAs on trade has been commonly analysed using the gravity model of trade, with the dependent variable being the aggregate value of trade between two countries and modelling the agreements with dummy variables. Some recent studies for aggregated trade are Carrere (2006), Magee (2008) and

Martínez-Zarzoso et al. (2009). Most of these recent papers rely on a model that assumes iceberg trade costs3 and symmetric firms. In this setting, consumers buy positive quantities of all varieties, and aggregated trade values react to trade cost reductions in exactly the same way as firm-level quantities and values.

The theoretical models used to generate the gravity equation usually assume homogeneous firms within a country and consumer's love of G

variety. These two assumptions imply that all products are traded to all n

destinations. However, empirical observation indicates that few firms g

export and exporting firms commonly sell in a limited number of LL

countries. This empirical fact has led to the development of the so-called |

new-new trade theories based on firm heterogeneity in productivity and t

fixed export costs (Melitz, 2003). These newer theories predict, for each j

country, a productivity threshold that firms must exceed in order to £

become exporters. As a result, two margins of trade emerge: the extensive r

margin and the intensive margin. Chaney (2008) shows that a higher elas- U

ticity of substitution makes the intensive margin more sensitive to changes §

in trade barriers, whereas it makes the extensive margin less so. The reason- r

ing is as follows: when goods are highly differentiated (low elasticity of sub- "

stitution), demand for each individual variety is relatively insensitive to £

changes in trade costs, therefore trade barriers have little impact on the intensive margin of trade. Conversely, as trade barriers decrease, firms with lower productivity levels are also able to enter these markets. The extensive margin is, therefore, strongly affected by trade barriers when elasticity of substitution is low. The opposite holds when it is high.

We may express the quantity of a variety from origin country i to destination country j (q¡j) as:

E ((prtij)-a\ m

qij = j P )' (1)

where Ej represents country j's total expenditure on the differentiated product, pitij is the price of product i at destination j, it varies across destinations due to positive iceberg transport costs, tij. Pj = i (Pitij)<'1-sS is a

3 Iceberg trade costs mean that for each good that is exported, a certain fraction melts away during the trip, as if an iceberg were shipped across the ocean.

4 For example, in the case of Colombia, Eaton etal. (2007) show that aggregate export levels are primarily accounted for large established firms and that once firms have penetrated other Latin American destinations they are more likely to reach larger OECD markets, but not vice versa.

price index and s is the elasticity of substitution, which is constant across varieties5 (CES).6

Since traded quantities of each variety are generally not observable, adding the two following assumptions, (a) all varieties at the point of origin are symmetric and (b) destinations will consume all the varieties in equal quantities, allows us to multiply the quantity per variety (qj times prices (pi) and times the number of varieties («¿), to obtain total trade values. The resulting formula is:

r „ u (pi( pitv)-s\ Tij = nipiqij = EjnA-P-I. (2)

In equation (2), the quantity per variety is the only component of Tij exhibiting bilateral variation. Following Hillberry and Hummels (2008), we can examine each of the components of total trade values in a more flexible way, as not only are data on quantities available, but prices and the range of products also vary across origins and destinations. Therefore, we need to relax some of the assumptions made above. Prices may vary 0

across destinations if the elasticity of substitution is not constant or if /

transport costs are not iceberg costs (Hummels and Skiba, 2004). ^

Consequently for a given year t, we can assume:

Tij = nijpijqij. (3) §

At least three reasons have been suggested in the literature to explain why SS

the range of traded products might vary with trade costs. First, goods pro- d

duced in different locations (origin and destination) maybe homogeneous. o In this case, if production costs in origin and destination are very similar

trade costs are sufficiently high, these goods will not be traded. S Additionally, the higher the transport costs, the more likely products are

to be non-traded goods. Second, if goods are differentiated by country §

of origin, each country producing a different variety has to incur a fixed §

cost to sell the product in each destination country. Therefore, not all y

the varieties will be shipped to each destination and the number of varieties §§

traded will be inversely proportional to trade costs. Finally, not all varieties 1 are consumer goods. Intermediate inputs that are used in the production of final goods would only be exported to destination j if country j produces the final good. Due to 'just in time' production processes, intermediates

5 Varieties refer to different products that are substitutes in consumption.

6 CES assumption is made in order to obtain a simple model that is easily derived and the implications of which may be tested.

are more likely to be traded over short distances. We focus on the first and second explanations and assume that both the number of varieties and the quantity traded are negatively affected by trade costs.

The methodology we use to decompose aggregate value of trade into its various components is based on Hillberry and Hummels (2008). Unique shipments are indexed by s and the total value of shipments from country i to country j is given by:

Tj = £ PjQj, (4)

where Nij is the number of unique shipments (extensive margin of trade) and PQij is the average value per shipment (the intensive margin). Hence, total trade value is decomposed first into extensive and intensive margins:

Tj = NijPQij, (5)

(e2i PjQj)

Since there can be multiple unique shipments within any origin-destination country pair, the number of shipments can be further decomposed into the number of distinct Standard International Trade Classification (SITC) products shipped, Nk, and the number of average shipments between a country of origin and a destination country, NF. NF > lmeansthatwe observe more than one unique shipment per commodity travelling from country i to country j:

Nj = nn . (6)

The average value per shipment can also be further decomposed into average price and average quantity per shipment:

(e2i pjQj)(rNii Qj)

PQij = --- = PjQij. (7)

j Li Qij Nij j j

By substituting equations (6) and (7) into equation (5), we can decompose total trade between two countries into four different components:

Tij = NjNjPijQ ,j. (8)

j N- ' c

For all commodities, quantity is measured in tons. Using a common unit allows us to aggregate over different products and compare prices (import unit values) across all commodities.

We now have two decomposition levels. The first is given by equation (5) and decomposes total trade value into the range of products traded and the average value per product. The second level decomposition, given by equation (8), decomposes each precedent element into two again. The first element, the range of products traded, is decomposed into the number of distinct SITC goods shipped and the number of average shipments between a country of origin and a destination country. The second element, the average value per product, is decomposed into average price and average quantity. Taking logs for the first- and second-level decompositions and adding the time dimension, t, we obtain:

ln Tjt = ln Nljt + ln PQtjt, (9)

ln Tjt = ln Nj + ln NFt + ln hjt + ln Qtjt. (10)

In the empirical analysis, we analyse how each of the components of /

equation (10) co-varies with distance and with other trade-related costs. U

The variable of interest is trade cost reductions induced by trade liberalisa- B. tion between the EU and the Maghreb countries considered. Before

specifying the empirical model in the next section, we will advance a number of d hypotheses that are based on recent theories of international trade under g^ conditions of imperfect competition with heterogeneous firms. d

4. Main hypotheses

The first hypothesis to be tested is that the EuroMed process has positive effects on the extensive and intensive margins of trade. Melitz (2003) introduced firm heterogeneity in a general equilibrium model of international trade. Chaney (2008) extended Melitz's model to multiple countries with asymmetric trade barriers. This model predicts that, for aggregated bilateral trade flows, the elasticity of exports with respect to trade barriers is larger than that in the absence of firm heterogeneity and larger than the elasticity for each individual firm. A reduction of variable costs has two effects. First, it increases the size of exports for each exporter, and second, it allows new firms to enter the market. Therefore, the extensive margin amplifies the impact of a reduction of variable costs on trade. The results obtained in the present paper support this first hypothesis

and show that the EuroMed process has a positive effect on both the extensive and intensive margins of trade.

The second hypothesis is that the effect of the EuroMed agreements may differ for different sectors. In more homogeneous sectors, exports are very sensitive to changes in transportation costs because many firms enter and exit when variable costs change. The elasticity of exports with respect to variable costs does not depend on the elasticity of substitution between goods. However, the elasticity of exports with respect to fixed costs is negatively related to the elasticity of substitution. This is in contrast to models with a representative firm, according to which the elasticity of exports with respect to transport costs equals the elasticity of substitution minus 1. Furthermore, regarding the two margins of trade, Chaney (2008) shows that, in the presence of firm heterogeneity, the extensive margin and the intensive margin are affected in different directions by the elasticity of substitution. The impact of trade barriers is strong in the intensive margin for high elasticities of substitution (homogeneous products), whereas the impact is mild on the extensive margin. The author proves that the dampening effect on the extensive margin dominates the magnifying effect on the intensive margin. The results obtained in the present paper support this second hypothesis, as the effect on the intensive margin is stronger for product categories for which the elasticity of substitution is higher.

The third hypothesis predicts that the impact of the EuroMed agreements will differ across countries. Production structures and specialisation patterns differ across Mediterranean countries. If one country is specialised in homogeneous or referenced goods, then it should be more dependent on price competition and the new agreements should have an effect on export volumes. On the other hand, countries specialising in more sophisticated products may find more opportunities to diversify with the new

agreements.

The fourth hypothesis predicts that the RoO adopted with the Barcelona Process could have a positive effect on finished goods trade, or could affect intermediate goods trade, promoting the development of production networks. Augier et al. (2005) point out that moving to a system of diagonal cumulation of origin widens the possible source of intermediate suppliers to all those countries which are part of that system. Therefore, exporters of the Mediterranean countries could use intermediate goods from more efficient partners inside the agreement or from the rest of the

7 See Table A1 for specialisation patterns.

world (RoW).8 Consequently, if RoO adopted with the first phase of the Barcelona Process are more flexible than those that existed previously, exports from southern Mediterranean countries to the EU should increase. Furthermore, the new RoO adopted may also have consequences for the imports of intermediate products from the RoW. In relation to this, Hypothesis 5 states that the effect of the EuroMed trade liberalisation process may differ across sectors. d

With regard to the second phase of the Barcelona Process, the entry into n

force of FTA agreements between the EU and the Mediterranean countries °

means that EU products will have duty-free access to the southern and LL

eastern Mediterranean markets after the negotiated transition periods. If |

trade barriers applied to intermediate goods imported from the EU into t

these countries are reduced and eventually phased out completely, these j

products will became less expensive and finished goods produced by o

Maghreb and Mashrek exporters could be sold at more competitive r

prices. Consequently, the end of customs duties at the border of the U

southern and eastern Mediterranean markets could lead to increased §

exports from these countries due to the lower costs of imported inputs. r

Hypothesis 6 analyses the impact of the second phase of the Barcelona "

Process and states that an increase in intermediate imports from the EU i

has a positive effect on Mediterranean countries' exports.

Figure 1 summarises the main expected effects of the Barcelona Process on trade between the EU and the southern and eastern Mediterranean countries.

In the next section, we start by estimating the overall impact of the Barcelona Process. As we are particularly interested in the means by which the Barcelona Process might create trade, we investigate whether the process impacts exports by creating new trade (more varieties exported) or by exploiting existing comparative advantages (increase in average export volumes of existing flows) or both. Thereafter, we aim to specifically disentangle whether those liberalisation effects are due to a change in the RoO or to the liberalisation of imported inputs from the EU. These results should contribute to our understanding of the potential effect of the new series of bilateral EuroMed agreements.

8 As Mediterranean countries may use intermediate goods imported from one of their agreement partners free of tariffs in the production process, there is more room for using intermediate goods from the RoW.

Production network

Reformed Rules of Origin inside the FTA

Impact on the quantity of imported inputs from RoW

Decrease of European inputs'price

Figure 1: The Effect of the Barcelona Process on EuroMed Trade. Note: RoW stands for Rest of the World (meaning all the countries outside the Barcelona Process)

5. Empirical analysis

5.1 Data, sources and variables

Our main data source is Eurostat. We used the external trade detailed database which covers both extra- and intra-EU trade. External EU trade statistics include data for the trade in goods between the NA countries and the four EU Member States considered in this study (France, Italy, Germany and Spain).9 Products are classified according to the SITC codes at the five-digit level. Only manufactured products are taken into consideration (categories 5 to 8, see Table A2). Income and population data are taken from the 2009 World Development Indicators Database and distance and colonial links from CEPII. Table A3 provides a summary of the data and sources used in this paper.

The extensive and intensive margins, average price and average quantity of products exported from the NA to France, Italy, Germany and Spain over the 1995-2008 period are calculated by using export values and quantities. We count the number of products (five-digit SITC) exported within

9 Libya and Syria, countries that did not enforce bilateral free trade agreements with the EU, are used as a control group in our estimations.

each two-digit SITC sector from each exporter to each importer yearly. For example, out of a total of 2,678 products categories listed in 1999, Algeria exported only forty-six different products to Germany. This number increased to sixty-three in 2008. Therefore, only a 2.3% of the existing products were traded. The total number of products exported to Spain rose from 79 to 93 over the period 1995-2008, representing a 3.4% of the products listed.10 To better assess the effect of the Barcelona Process on the Q range of products traded, we present different graphs for each NA n country, which show the evolution of the number of zero-traded categories ° with each EU partner over time. Figures 2-5 show a decreasing trend for & the number of zero-traded categories over the period considered. | The variables input_EU and input_RoW are used as proxies for inter- t mediate inputs imported from the main countries of the EU (France, j Germany, Italy, Spain and the UK) or alternatively, from the main produ- £ cers of the RoW (Japan, South Korea, Hong Kong, USA). The source for r these two variables is the OECD exports database. We focus more particu- U larly on machinery and equipment exports from the main countries of the § EU and the RoW to each of the Mediterranean countries (Sector 84 of the r harmonised system of commodity classification). Summary statistics for " these variables are presented in Table 2. n

5.2 The estimated model

We are interested in knowing whether the hypotheses that have been described hold for trade flows in the Mediterranean region. In order to test some of these predictions, the estimating equation takes the following form:

lnXljkt =a> + ai ln GDPJt + «2 ln GDPjt + «3 ln POPJf

+ a4 ln POP jt + a5 ln Dij + a6FTAjt + a7Colony + gk (11)

+ 1 + 1 ijkt > sijkt = mijk + nijkt'

where gk and 1t are industry (at two-digit level) and year fixed effects, respectively. eijkt is a two-component error term, and ln Xjkt is in turn

10 The following example illustrates the importance of the new products exported in monetary terms: in 1999, Algeria exported forty-six products worth €4,591,102 to Germany. In 2008, the number of products exported increased to sixty-three and their value amounted to €19,390,231, of which only €4,120,152 corresponded to products already traded in 1999, with the rest corresponding to new export products.

Figure 2: Evolution over Time of the Number of Zero Trade Categories: Algeria

Figure 3: Evolution over Time of the Number of Zero Trade Categories: Egypt

Figure 4: Evolution over Time of the Number of Zero Trade Categories: Morocco

the log of the average value per shipment (intensive margin), and the log of the range of shipments (extensive margin), as described in equation (9). GDPit and GDPjt are for gross domestic product of the importer and the exporter country in year t, respectively. Dj is the geographical distance

Figure 5: Evolution over Time of the Number of Zero Trade Categories: Tunisia

Table 2: Summary Statistics

Variable Observations Mean Standard deviation Minimum Maximum

Export value (€) 10,640 1.18E + 07 6.88E + 07 0 2.29E + 09

Number of 10,640 9.807049 16.4289 0 155

products

Average quantity 8,526 16,772.06 93,794.28 0 2,122,389

(tons)

Average value 8,526 902,920.3 3,605,443 0 1.53E + 08

(€/ton)

Average price (€) 8,214 3,906.157 17,929.67 2.258772 734,315

GDP exporter (€) 10,640 4.58E + 10 2.84E + 10 8.71E + 09 1.18E + 11

GDP importer (€) 10,640 1.41E + 12 5.46E + 11 4.56E + 11 2.48E + 12

Distance 10,640 2,025.284 944.1976 595.3532 3,610.572

Inputs imported 10,640 1.03E + 09 5.36E + 08 2.52E + 08 2.56E + 09

from EU (€)

Inputs imported 10,640 2.50E + 08 2.60E + 08 3.97E + 07 1.21E + 09

from ROW (€)

between the trading countries' capitals, and FTAjt represents free trade agreements dummies that take the value of 1 when both countries have implemented a cooperation agreement in year t, 0 otherwise. Finally, Colonyj is a dummy that takes the value of 1 when the trading partners had a colonial relationship in the past, 0 otherwise. Since OLS is linear, the coefficient on total imports will be equal to the sum of the coefficients on the two margins. This can be broken down even further, using each of the components in equation (10) as a dependent variable in equation (11).

To take into account particular aspects of the agreement, an extended model is specified in which the FTA variable in equation (11) is replaced

by two variables that describe the type of cumulation rule and incorporate imports of intermediate products from the EU and from the RoW. The extended model is given by,

lnXykt =a0 + a1 lnGDPit + a2lnGDPjt + a3lnDij+ a41D_Cumulationi7f + a42Pan-EuroMed_RoOj + a5Colony,j + a6ln input_euit + a7\n input-row^ + yk + It + lijkt, lijkt = mijk + vijkt,

where D_Cumulationi-;t takes the value of 1 when the RoO allows for diagonal cumulation with the other MENA countries, 0 otherwise; Pan-EuroMed_RoOijt takes the value of 1 when a country has full cumulation RoO, 0 otherwise; input_euit stands for imported machinery from the EU and input_rowit for imported machinery from the RoW. Having done this, we estimated an extended model for all countries and for each sector. We were not able to estimate the extended model for each |

country, as the variables imported inputs from the EU and imported .

inputs from RoW are country specific. /

5.3 Main results

Integration effects are estimated jointly for the four Mediterranean countries considered. The three-dimensional structure of our data allows us to control for unobservable heterogeneity in several ways. After testing a number of competing specifications, the selected specification includes exporter, importer and sectoral effects jointly considered random, and year and industry effects specified as fixed effects. In order to control for the remaining unobserved heterogeneity, we added the averages of the time-variant variables as additional explanatory variables, as suggested by Mundlak (1978).11 We also considered adding population variables or GDP per capita for exporters and importers, and the results concerning the effect of our target variable remain unchanged. The estimated model is also corrected for first-order autocorrelation by adding a first-order autoregressive term. The fixed effects results are also shown in Table A5. In addition, in the spirit of Baldwin and Taglioni (2006), we also estimated the model with country-and-time fixed effects in addition to the dyadic bilateral effects. The results show positive and significant

11 This model has been labelled in the econometric literature as the 'correlated random-effects model'.

coefficients that are even greater in magnitude (Table A6). We are thus able to control for changes that are specific to each country and common for all sectors but vary over time, such as the financial reforms mentioned in the introduction, which may impact exports. As a final robustness check and to address the possibility that the FTA variable could be endogenous, Table A7 shows the results obtained by estimating the model in first differences and using instrumental variables.12 The coefficient of the Barcelona-Process G

variable is statistically significant at the 1% level and only slightly lower n

in magnitude. Hence, we have chosen the above-mentioned specification g

as our benchmark model, which is somewhat conservative as it shows LL

lower magnitudes than the model with country-and-time fixed effects, |

but slightly higher than the instrumental variables estimates. t

Table 3 shows the results for total trade and for each margin of trade. The j

dependent variable in column (1) is the logarithm of the total value of £

exports from the NA countries to the four importing European countries. r

In columns (2) and (3), the dependent variable is each of the components o

of equation (9), respectively, that is, the extensive and the intensive margins §

of trade. In columns (4) and (5), the dependent variables are the two last r

components of equation (10) that represent the decomposition of the "

intensive margin into average quantity and average price, respectively. £

For the whole set of Mediterranean countries, Table 3 shows that the coefficient of our variable of interest, the implementation of an FTA between the NA countries and the EU, is positive and statistically significant for total trade (column 1), the extensive margin (column 2) and the intensive margin (column 3). Turning to the second-level decomposition of equation (10), the first component of average value per shipment (column 4—Table 3), average quantities shipped are higher after the FTA entered into force, whereas the FTAvariable is not significant when the average price component is used as a dependent variable (column 5—Table 3). These results show a significant diversification effect (Hypothesis 1). Such diversification involves shipping existing exports to new foreign destinations or shipping a good abroad for the first time. Regarding the additional explanatory variables, exporting countries' GDP was found to have a positive statistically significant impact on total trade, as expected. Geographical distance presents a negative and significant coefficient, except for the average price, which shows a positive distance coefficient (this result has also been obtained in results for a sample of Latin American countries, see Hillberry and Hummels, 2008; Martínez-Zarzoso 2009; Martínez-Zarzoso and Wilmsmeier, 2010). The

12 We used the second lag of the FTA variable as instrument.

Table 3: Main Results for All Countries and Sectors

N„ AVj AQj APj

Ln GDP, 2.233*** (8.009) 0.212*** (2.803) 2.051*** (8.184) 1.702*** (5.705) 0.183 (1.008)

Ln GDPy 0.022 (0.131) -0.011 (-0.229) 0.016 (0.107) -0.231 (-1.304) 0.300*** (2.675)

Ln Distance -0.956*** (-6.322) -0.272*** (-5.751) -0.678*** (-5.902) -0.837*** (-6.635) 0.162*** (2.974)

FTA (Barcelona Process) 0.255*** (4.134) 0.044** (2.309) 0.230*** (4.129) 0.300*** (4.637) -0.059 (-1.445)

Colony 1.462*** (7.345) 0.530*** (8.516) 0.929*** (6.167) 0.695*** (4.216) 0.144** (2.065)

Average ln GDP, -1.915*** (-5.547) 0.133 (1.349) -2.070*** (-7.037) -2.129*** (-6.212) 0.307 (1.578)

Average ln GDPj 0.471** (2.173) 0.125* (1.942) 0.364** (1.979) 0.742*** (3.511) -0.440*** (-3.603)

Constant term -2.656 (-0.416) -9.897*** (-4.960) 6.953 (1.435) 10.998** (2.062) -5.916*** (-2.591)

R-squared 0.319 0.502 0.323 0.519 0.648

Number of observations 8,517 8,526 8,517 8,214 8,214

LBI 1.424 1.716 1.451 1.425 1.610

d1 1.095 1.469 1.118 1.086 1.295

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of export value (current €). Incomes (GDPs) and distance are

also in natural logarithms (ln). The estimation uses White's heteroscedasticity-consistent standard errors. LBI and d1 express the Baltagi and

Wu, and the Bhargava et al. tests for autocorrelation, both tests reject the null of no first-order autocorrelation.

*Significance at 10%.

"Significance at 5%.

***Significance at 1%.

decomposition of the influence of distance on trade shows a greater effect on the intensive margin (column 3—Table 3), for all industrial products. About 28% of the distance effect on trade works through the extensive margin [i.e., 0.272/(0.272 + 0.678)]; 72% of the increase in disaggregated trade flows comes from larger average shipments. Prior research found the opposite picture, with the extensive margin being more important than the intensive margin (Hillberry and Hummels, 2008; Mayer and Ottaviano, 2007). Our results are very different to Mayer and Ottaviano (2007), who analyse French and Belgian individual export flows and show that 75% of the distance effect on trade comes from the extensive margin. Finally, sharing colonial links fosters exports from NA countries to the EU; 36% of the increase in disaggregated trade flows comes from the extensive margin (a wider variety of products traded), whereas 64% of the increase in disaggregated trade flows comes from larger average shipments (row 5, Table 3).

Summarising, these results consistently show that the new FTA agreements signed between the NA countries and the EU have fostered exports from these countries to some of their main European partners. Furthermore, we find that this increase in exports has been mainly channelled by an increase of the intensive margin of trade. The NA countries export more of the products they already exported in the past. This fact is in line with what we know of the industrial structure of these countries and with the explanation proposed by Chaney (2008) concerning how reductions in trade costs influence the two margins of trade. NA countries are mainly producers of low technological content goods, which are highly substitutable on the international market. In this case, Chaney (2008) argues that the main impact of a decrease in trade barriers ought to be through the intensive margin.

In order to test Hypothesis 2, the effect of the bilateral FTAs on trade is also estimated for each sector (at one digit-level SITC). Table 4 shows the main results for the FTA variable for each sector of the SITC.13 The different sectors are not equally impacted by the agreements. Concerning the effect on total trade (column 1), the coefficient of the FTA variable is significant and positive for Sectors 5 and 6 (chemicals and manufactured goods classified chiefly by material), whereas it is positive and nonsignificant for Sectors 7 and 8 (machinery and transport equipment and miscellaneous manufactured articles). The effect on the extensive margin is only positive and significant for Sector 5, and on the intensive margin it is positive and significant for Sectors 5 and 6. The second-level

13 Full results are available upon request from the authors.

Table 4: FTA Effects for Each Product Category

Sector X, N, AV¡,- AQ¡,

5 - Chemicals and related products 0.387*** (2.645) 0.072* (1.579) 0.331** (2.425) 0.379** (2.308) -0.165* (-1.667) t

6 - Manufactured goods classified chiefly by material 0.266** (2.304) 0.001 (-0.004) 0.237** (2.302) 0.233* (1.911) -0.013 (-0.179) g

7 - Machinery and transport equipment 0.148 (1.229) 0.002 (0.051) 0.138 (1.279) 0.199* (1.649) -0.124 (-1.392) i

8 - Miscellaneous manufactured articles 0.019 (0.184) 0.002 (0.065) 0.021 (0.226) 0.129 (1.199) -0.026 (-0.331) §

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of exports in value (current €). Income, population and o.

distance are also in natural logarithms. The estimation uses White's heteroscedasticity-consistent standard errors. he

*Significance at 10%. T

"Significance at 5%. °

***Significance at 1%. 2

Table 5: FTA Effects for Each Country

Countries Xj Nj AVj AQij APj

Algeria 0.280** 0.066* 0.213* 0.232* -0.084

(2.296) (1.957) (1.953) (1.649) (-0.969)

Egypt 0.249*** 0.175*** 0.143* 0.519*** -0.430***

(2.972) (7.877) (1.862) (5.724) (-8.057)

Morocco 0.281*** 0.156*** 0.188** 0.281*** -0.064

(3.354) (6.749) (2.407) (3.083) (-1.117)

Tunisia 0.545*** 0.167*** 0.444*** 0.501*** -0.152**

(5.312) (5.332) (4.798) (4.806) (-2.159)

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of export

value (current €). Income is also in natural logarithms. The estimation uses White's

heteroscedasticity-consistent standard errors.

*Significance at 10%.

"Significance at 5%.

***significance at 1%.

decomposition shows that it is through an increase in the average quantity of goods exported that the intensive margin increases.

The results are in line with the idea that the main changes induced by the FTA arise through the intensive margin of trade.

These results seem consistent with the explanation proposed by trade theorists: for more homogeneous products (Sectors 5 and 6), a reduction in trade costs is expected to have a greater impact through the intensive margin. Two of the exporters considered, Tunisia and Morocco, are important exporters of fertilisers (SITC 56) as well as of textile and leather products (SITC 61, and 65), and Algeria is a major exporter of organic chemicals derived from petroleum (SITC 51). None of these products is highly differentiated products.

In order to test Hypothesis 3, which predicts that the effect of the EuroMed differs across countries, we estimated equation (11) for every NA country as an exporter. Table 5 depicts our main results. The effect of the FTA is positive and significant for total exports for the four countries. For Egypt and Morocco, the total effect seems to occur through both margins of trade, whereas for Algeria and Tunisia most of the effect goes through the intensive margin. Interestingly, for Egypt and Tunisia, the second-level decomposition shows a significant and negative effect of the agreement on the average prices, which could be the effect of greater competition in local markets.

Table 6: Main Results Extended Model

N AVj AQij APj

Ln GDP, 2.270*** (8.172) 0.225*** (3.006) 2.077*** (8.315) 1.777*** (5.99) 0.139 (0.765)

Ln GDPy -0.018 (-0.104) -0.058 (-1.098) 0.009 (0.057) -0.247 (-1.323) 0.266** (2.199)

Ln Distance -0.838*** (-4.436) -0.201*** (-2.682) -0.622*** (-4.236) -0.894*** (- 4.473) 0.254** (2.048)

D_Cumulation 0.220*** (3.498) 0.019 (0.992) 0.198*** (3.474) 0.253*** (3.807) -0.101** (-2.295)

Pan_EuroMed_RoO 0.098 (1.294) 0.021 (0.839) 0.097 (1.402) -0.025 (-0.316) 0.104* (1.888)

Ln input_eu 0.016 (0.131) 0.058* (1.518) -0.013 (-0.124) -0.236* (-1.846) 0.169** (1.982)

Ln input_row -0.265*** (-3.636) -0.113*** (-4.675) -0.164** (-2.521) -0.158** (-2.025) 0 (-0.009)

Colony 1.432*** (5.973) 0.504*** (5.219) 0.927*** (5.051) 0.720*** (2.84) 0.132 (0.844)

Average ln GDP, -2.016*** (-5.432) 0.096 (0.774) - 2.146*** (- 6.861) -2.252*** (- 5.703) 0.327 (1.349)

Average ln GDPj 0.835*** (3.271) 0.256*** (2.949) 0.602*** (2.787) 1.204*** (4.443) -0.598*** (-3.528)

Constant term -3.96 (-0.514) -9.971*** (-3.231) 5.795 (0.98) 8.652 (1.059) -3.651 (-0.725)

R-squared 0.123 0.104 0.085 0.076 0.013

Number of observations 8,517 8,526 8,517 8,214 8,214

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of exports in value (current €). Incomes (GDPs), distance 2

(Dist) and imported inputs from the EU (input_eu) and from the RoW (input_row) are also in natural logarithms (ln). The estimation uses i

White's heteroscedasticity-consistent standard errors. s

*Significance at 10%. f

"Significance at 5%. a

***Significance at 1%. e

5.4 Disentangling FTA effects

In this section, we present and discuss the results obtained from estimating the extended model given by equation (12). With this model, we aim to test for Hypothesis 4, which states that RoO adopted in the first phase of the Barcelona Process could have had a positive effect on exports of finished goods due to an increase in imports of intermediate goods. Table 6

shows that the effect of diagonal cumulation on total trade is significant o

and positive and most of this effect operates through the intensive l

margin of trade. The coefficients for the Pan_EuroMed_RoO variable are e

significant and positive only for the average price of goods exported. r

Regarding the variables related to a possible 'production network' effect, 3

the coefficient of the inputs imported from the EU variable has a positive :

effect on total trade, but the latter is not statistically significant at conven- e

tional levels. It is significant at the 10% level and positive for the extensive f

margin of trade. Looking at the second-level decomposition of the inten- jj

sive margin, this variable has a positive and significant effect on average |

prices, which is compensated by a negative and significant effect on .

average quantities exported. This neutralises the effect on the intensive g

margin of trade. These elements may reveal that the integration of U

European intermediate goods has led to the exporting of new products,14 v

with more value added, to European markets. The coefficient of the inputs t

imported from the RoW variable is negative for total trade and also for the e

number and quantity of goods exported. This could be interpreted as indi- §

cating that a displacement of third-country imports in favour of EU e

imported inputs has the effect of reducing the diversity of exports to the o

EU. This result may also be pointing towards the co-existence of other §

regional integration agreements, such as the FTA signed between S

Morocco and the USA in 2005, which entered into force in 2006. The i US-Morocco FTA eliminated tariffs on 95% of bilateral trade

in consumer i

and industrial products, with all remaining tariffs to be eliminated within a §

9-year period. Since its entry into force until 2010, bilateral trade between JJ

the two countries has increased 147%.15 Therefore, some of the potential ,

exports to Europe might have been diverted to the USA, as the USA is 1 an important trade partner for Morocco (Table A1). The agreement could also have effects on export dynamics in Morocco.

14 As our level of decomposition is a five-digit one, it may be that new categories of goods are exported at our level of decomposition—as well as new products within a category, representing a shift towards sophisticated goods.

15 Source: Bureau of Near Eastern Affairs, US Department of State (2010).

As a further refinement of our estimation, we also considered a model that controls for possible unobserved sector-specific heterogeneity. We added fixed effects for each exporter-time-sector and importer-time-sector, at one-digit levels. This allows us to control for possible misspecifications due to the exclusion of sector- and country-specific variables that are time variant and can be correlated with the error term, such as: sectoral value added in case of the exporter-time-sector effects or consumer biases in case of the importer-time-sector effects. The results of the extended model are presented in Table 7. They show a greater effect for the diagonal cumulation variable and confirm that trade variations arise mainly through the intensive margin. Regarding the coefficients for the Pan_EuroMed_RoO variable, they are significant and positive not only for the average price of goods exported, but also for the extensive margin, in both cases at the 10% level. As the results show that sector-specific effects are important, in the next set of results we present sector-specific estimates.

Hypothesis 5 states that the effect of the change in RoO may differ across sectors. Results presented in Table 8 show that the coefficients for the diagonal cumulation variable are significant and positive when total trade, the

Table 7: Extended Model with Dyadic Fixed Effect and Country-Sector-and-Time Effects

X Ni AVj AQi APj

D_Cumulation 2.718** 0.695* 2.019** 1.709 (1.48) 0.846

(2.463) (1.799) (2.222) (1.321)

Pan_EuroMed_RoO 0.345 0.293* 0.052 -0.454 0.675*

(0.562) (1.73) (0.097) (-0.639) (1.527)

Ln input_eu 0.586 -0.018 0.603 0.925 - 0.284

(1.064) (-0.097) (1.27) (1.411) (-0.841)

Ln input_row 0.043 -0.825* 0.868 0.645 - 0.591

(0.028) (-1.785) (0.646) (0.352) (-0.654)

R-squared 0.371 0.432 0.278 0.302 0.343

Number of 8,517 8,526 8,517 8,214 8,214

observations

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of export

value (current €). Imported inputs from the EU (input_eu) and from the rest of world

(input_row) are also in natural logarithms (ln). The estimation uses White's

heteroscedasticity-consistent standard errors.

***Significance at 1%.

"Significance at 5%.

*Significance at 10%.

extensive and the intensive margins are the dependent variables for Sector 5 exports. The impact on the intensive margin is notably higher than the impact on the extensive margin of trade. Turning to the second-level decomposition, the coefficient is significant and positive for the average quantity, and significant and negative for the average price. The results for exports of Sector 6 are very similar, but the coefficients are significant only for total trade, the intensive margin and the average quantity. The coefficients for the diagonal cumulation variable show the same sign as in Sector 5. Concerning exports of Sectors 7 and 8, the coefficients of the Pan_EuroMed_RoO variable are significant and positive for total trade and the intensive margin. In Sector 7, the coefficient for the inputs imported from the EU variable is also significant for the extensive margin. The coefficients for the inputs imported from the RoW variable are significant and negative, mainly for Sectors 7 and 8. These results show that the adoption of diagonal cumulation between the NA countries has an impact on their exports to Europe mainly through the intensive margin of trade, more specifically through the average quantity exported. The results for entering in the unified regime of RoO of the EU are slightly different; here the main effect on exports comes through the intensive margin, but with a positive impact on the average price of exported goods. The effect of an increase in European imported inputs has an interesting effect for lower technological content sectors (5 and 6). The effect on the average quantity of exports is negative, but it is positive for average prices. These results may be interpreted as an increase in the quality of the goods produced by NA countries. Additionally, for Sector 7, which has the highest technological content, an increase in inputs imported from the EU has a positive and significant effect on the extensive margin of trade. Regarding the results for inputs imported from the RoW, the above-mentioned counter-diversification bias is found for Sectors 6, 7 and 8, thus indicating that an increase in imported inputs from the RoW has a negative effect on the number of varieties exported to our four EU countries and may also lead to a reduction in total exports to these countries.

6. Conclusions and policy implications

In this paper, the effect of EuroMed agreements on international trade is evaluated by using disaggregated trade data. These agreements should contribute to modifying trade patterns between the two shores of the

Table 8: Sectoral Results Extended Model

Sector 5 D_Cumulation Pan_EuroMed_RoO Ln input_eu Ln input_row Sector 6 D_Cumulation Pan_EuroMed_RoO Ln input_eu Ln input_row Sector 7 D_Cumulation Pan_EuroMed_RoO Ln input_eu Ln input_row Sector 8 D_Cumulation Pan_EuroMed_RoO Ln input_eu Ln input_row

0.430*** (2.88) -0.152 (-0.854) -0.422 (-1.393) -0.302 (-1.64)

0.266** (2.279) -0.076 (-0.561) 0.021 (0.094) -0.195 (-1.472)

0.145 (1.181) 0.358** (2.301) 0.319 (1.355)

0.021 (0.193) 0.256** (2.04) 0.141 (0.731)

0.082* (1.743) 0.02 (0.328) -0.031 (-0.319) -0.058 (-1.023)

0.019 (0.509) -0.004 (-0.076) 0.092 (1.235) -0.186*** (-4.003)

0.002 (0.05) 0.058 (1.256) 0.150** (2.219)

-0.296** (-2.12) -0.082* (-1.937)

0.024 (0.674) 0.051 (1.099) 0.037 (0.555)

0.373*** (2.674) -0.129 (-0.768) -0.358 (-1.266) -0.289* (-1.687)

0.245** (2.343) -0.061 (- 0.505) -0.077 (-0.392) -0.01 (-0.086)

0.132 (1.196) 0.331** (2.349) 0.203 (0.961) -0.206* (-1.685)

0.013 (0.139) 0.221** (1.978) 0.19 (1.11)

0.433*** (2.592) -0.266 (-1.344) -0.547 (-1.611) -0.424** (-1.992)

0.263** (2.131) -0.133 (-0.939) -0.325 (-1.407) 0.048 (0.347)

0.221* (1.787) 0.235 (1.5) -0.028 (-0.114) -0.21 (-1.473)

0.122 (1.119) 0.056 (0.438) 0.139 (0.699)

-0.281** (-2.35) -0.148*** (-3.438) -0.192* (-1.827) -0.199 (-1.64)

-0.189* (-1.866) 0.12 (0.961) 0.194(0.939) 0.071 (0.559)

-0.036 (- 0.508) 0.059 (0.682) 0.250* (1.841) -0.097 (-1.229)

-0.148 (-1.617) 0.09 (0.746) 0.17 (0.94) 0.001 (0.005)

-0.033 (-0.404) 0.187* (1.904) 0.012 (0.084) 0.033 (0.372)

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of exports in value (current €). Imported inputs from the EU

(input_eu) and from the rest of world (input_row) are also in natural logarithms (ln). The estimation uses White's heteroscedasticity-consistent

standard errors.

*Significance at 10%.

"Significance at 5%.

***Significance at 1%.

Mediterranean. We apply recently developed trade models (Chaney, 2008) to depict the impact of FTA on the extensive and intensive margins of trade, focusing on exports from NA countries to the four biggest continental European economies: Germany, France, Italy and Spain.

Our results confirm a positive and significant effect of the new FTAs on exports of NA countries to their main European partners. Moreover, strong differences across countries in the effect of the new FTAs are found. Indeed, empirical evidence indicates that NA countries have enjoyed a significant increase in exports associated with the FTA, operating through the intensive margin for Algeria and Tunisia, and through both the extensive and intensive margins for Egypt and Morocco. Diverse trade patterns between NA countries could be at the origin of these differences. In further research, we aim to analyse whether FTAs are effective in diversifying the productive structure of their members and whether this is only the case for countries which already had strong commercial links before the agreement.

This positive effect of the new FTAs on trade could be due to the new RoO. A plausible explanation as to why the adoption of new RoO has resulted in increased trade is that the new rules have allowed the integration of better quality/less-expensive intermediate goods in production in NA countries, consequently enhancing the demand for these goods in European markets. The sectoral result partially confirms this hypothesis, since the effect of an increase in the inputs imported from the EU has a positive effect on NA exports of sophisticated manufactured products. This effect is channelled by an increase in the extensive margin of trade for machinery and transport equipment. Further research using more disaggregated trade data is needed to ascertain whether export diversification is actually a consequence of the change in RoO.

Other elements that should be taken into account to promote trade and development in the area are product standards and financial liberalisation. It would be desirable that NA countries and the EU converge in terms of standards and financial sector regulations, as the EuroMed agreements did not contain any specific measures in these domains. Progress on this depends mainly on the willingness of the NA governments to promote this kind of reforms in their own countries.

Although a growing body of literature has focused on the impact of standards and regulations on trade (Swann, 2010), most data sets available cover only developed countries. The countries covered in this study are still in the process of establishing national committees to monitor compliance with technical regulations, as well as sanitary agencies and

professional associations, in order to spur the adoption of new standards (European Commission, 2009).

Financial liberalisation has been identified as a possible source of export growth (Rajan and Zingales, 1998; Braun and Larrain, 2005; Berthou, 2010) because it increases external sources of financing for growing businesses. NA countries underwent important financial reforms in the nineties under the supervision of the International Monetary Fund (Jbili et al., 1997). Consolidation in the banking sector and the development of a functioning financial market have continued in countries such as Morocco and Tunisia. These measures should have served to improve the position of NA exporters in all markets, albeit in varying degrees, according to the depth of the reforms adopted.

One last point concerns the impact of measures taken through international bodies such as the WTO on trade, as these could weaken the exporting position of NA countries. The end of the Multifibre Agreement (MFA) in 2005 is a case in point. The MFA allows countries to establish exceptions to the general agreement on trade and tariffs for the textile sector. The consequence of this agreement was that trade between some partners was protected from competition through higher tariffs and quotas. Trading partners such as Tunisia, Morocco and Egypt were shielded by this agreement from competition with Asian and eastern Europe countries. The end of the MFA agreement led to a decrease in trade costs between Europe and NA countries' competitors, which could negatively affect trade between Europe and NA countries. In this paper, we focus exclusively on the direct effects of the Barcelona Process on Euro-North African trade and leave for further research this kind of indirect effects of trade policy on trade flows, which has been described as the 'multilateral resistance effect' by Anderson and van Wincoop (2003). Nevertheless, our results are not affected by the impact of the adoption of new standards and regulations since we controlled for multilateral resistance effects in different ways, obtaining similar results concerning the effects of the RoO.

This paper highlights the importance of non-tariff trade barriers, and in particular, of the RoO in a world where the classic weapons of trade protection—quotas and tariffs—are becoming less relevant in advanced economies. We strongly support the idea that further simplification of RoO through, for example, the adoption of a single convention on Pan-EuroMed preferential RoO, referred to as 'the Convention' by the European Commission (European Commission, 2010, COM 172), could

have a significant positive impact on EuroMed trade by simplifying red tape for firms in cases of diagonal cumulation.

Acknowledgements

The authors would like to thank two anonymous referees and the participants of the first CREMed Workshop for their helpful comments and suggestions and the Ministry of Science and Innovation (ECON2010-15867/ ECON) for financial support.

References /

Acemoglu, D. and F. Zilibotti (1997) 'Was Prometheus Unbound by Chance? Risk, o

Diversification, and Growth', Journal of Political Economy, 45: 709-51. r

Amurgo Pacheco, A. (2006) Preferential Trade Liberalization and the Range of o

Exported Products: The Case of the Euro-Mediterranean FTA, HEI Working a

Papers 18-2006, Economics Section. Geneva, Switzerland: The Graduate r

Institute of International Studies. /

Amurgo- Pacheco, A. and M. Pierola (2008) Patterns of Export Diversification in ^

Developing Countries: Intensive and Extensive Margins, Policy Research e

Working Paper Series 4473, The World Bank. t

Anderson, J.E. and E. van Wincoop (2003) 'Gravity with Gravitas: A Solution to e

the Border Puzzle', American Economic Review, 93: 170-92. ^

Augier, P., M. Gasiorek and C. Lai Tong (2005) 'The Impact of Rules of Origin on °

Trade Flows', Economic Policy, 43: 567-624. R

Baldwin, R. and D. Taglioni (2006) Gravity for Dummies and Dummies for g

Gravity Equations, National Bureau of Economic Research Working Paper a

12516, Cambridge, MA. P

Berthou, A. (2010). The Distorted Effect of Financial Development on n

International Trade Flows, Working Papers 2010-09. Paris, France: CEPII O

Research Center. M

Blanes-Cristobal, V. and J. Milgram-Baleix (2010) 'Impacto de la liberalizacion ^ comercial de Marruecos sobre las exportaciones por regiones', Revista de

Economía Aplicada, 18: 63-90. 1 Braun, M. and B. Larrain (2005) 'Finance and the Business Cycle: International,

Inter-industry Evidence', Journal of Finance, 60: 1097-128. Carrere, C. (2006) 'Revisiting the Effects of Regional Trade Agreements on Trade Flows with Proper Specification of the Gravity Model', European Economic Review, 2: 223-47.

Chaney, T. (2008) 'Distorted Gravity: The Intensive and Extensive Margins of International Trade', American Economic Review, 4: 1707-21.

Das, S., M. Roberts and J. Tybout (2007) 'Market Entry Cost, Producer Heterogeneity and Export Dynamics', Econometrica, 75(3): 837-73.

Deardorff, A. (1999) Economic Implications of Europe-Maghreb Trade Agreements, Working Papers 442, Research Seminar in International Economics, University of Michigan.

Deardorff, A., D. Brown and R. Stern (1996) Some Economic Effects of the Free Trade Agreement between Tunisia and the European Union, Working Papers 385, Research Seminar in International Economics, University of Michigan.

Eaton, J., M. Eslava, M. Kugler and J. Tybout (2007) The Margins of Entry into Export Markets: Evidence from Colombia, NBER Working Paper 13531.

Eichengreen, B. and J.A. Frankel (1995) 'Economic Regionalism: Evidence from Two 20th Century Episodes', The North American Journal of Economics and Finance, 6: 89-106.

European Commission (2009) Progress Reports on the Neighbourhood Policy Morocco 2008 SEC(2009) 520/2, Brussels, Belgium.

European Commission (2010) COM 2010 172 Final, Brussels, Belgium.

European Commission (2010) COM 2010 485 Final, Brussels, Belgium.

Francois, J.F., M. McQueen and G. Wignaraja (2005) 'European Union-Developing Country FTAs: Overview and Analysis', World Development, 10: 1545-65.

Frankel, J., E. Stein and S.-J. Wei (1996) 'Regional Trading Arrangements: Natural or Supernatural', American Economic Review, 2: 52 -56.

Frankel, J., E. Stein and S.-J. Wei (1998) 'Open Regionalism in a World of Continental Trade Blocs', IMF Staff Papers, 45: 440-53.

Gamberoni, E. (2007) Do Unilateral Trade Preferences Help Export Diversification? HEI Working Paper No 17/2007, Graduate Institute of International Studies, Geneva.

Gradeva, K. and I. Martínez-Zarzoso (2009) Trade as Aid: The Role of the EBA-Trade Preferences Regime in the Development Strategy, Ibero America Institute for Economic Research (IAI) Discussion Papers 197. Göttingen, Germany: Ibero-America Institute for Economic Research.

Gutierrez de Pineres, S. and M. Ferrantino (1997) 'Export Diversification and Structural Dynamics in the Growth Process: the case of Chile', Journal of Development Economics, 52: 375 -91.

Hillberry, R. and D. Hummels (2008) 'Trade Responses to Geographical Frictions: A Decomposition Using Micro-Data', European Economic Review, 52: 527-50.

Hoekman, B. (1998) 'Free Trade Agreements in the Mediterranean: A Regional Path towards Liberalisation', The Journal of North African Studies, 2: 89-104.

Hoekman, B. and D.E. Konan (1999) Deep Integration, Nondiscrimination, and Euro-Mediterranean Free Trade, Policy Research Working Paper Series 2130, The World Bank.

Hoekman, B. and D.E. Konan (2005) Deepening Egypt-US Trade Integration: Economic Implications of Alternative Options. Department of Economics, University of Hawaii at Manoa. Hümmels, D. and A. Skiba (2004) 'Shipping the Good Apples Out? An Empirical Confirmation of the Alchian-Allen Conjecture', Journal of Political Economy, 6: 1384-402.

Jbili, A., K. Enders and V. Treichel (1997) Financial Sector Reforms in

Algeria, Morocco and Tunisia: A Preliminary Assessment, IMF Working Paper °

97/81. l

Karray, B. (2003) 'The rules of origin in the Euro-Mediterranean economic space', e

in Peter G. Xuereb (ed.), Euro-Med Integration and the 'Ring of Friends': The ^

Mediterranean's European Challenge, vol. IV. Msida, Malta: European 3

Documentation and Research Centre. p

Magee, C.S.P. (2008) 'New Measures of Trade Creation and Trade Diversion', 00

Journal of International Economics, 75: 349-62. x

Martínez-Zarzoso, I. (2009) 'On transport costs and sectoral trade: further evi- d

dence for LA imports from the European Union', in G. Tondl (ed.), The g'

European Union and Emerging Markets, Vienna, New York: Springer. |

Martínez-Zarzoso, I. and G. Wilmsmeier (2010) 'Determinants of Maritime r

Transport Costs - A Panel Data Analysis', Transportation Planning and ^

Technology, 1: 117-36. n

Martínez-Zarzoso, I., F. Nowak-Lehmann D. and N. Horsewood (2009) 'Are r Regional Trading Agreements Beneficial? Static and Dynamic Panel Gravity

Models', North American Journal of Economics and Finance, 20: 46-65. g

Mayer, T. and G.I.P. Ottaviano (2007) The Happy Few: New Facts on the S Internationalization of European Firms, Bruegel-CEPR EFIM 2007 Report,

Bruegel Blueprint Series. o

Melitz, M.J. (2003) 'The Impact of Trade on Intra-Industry Reallocations and 3

Aggregate Industry Productivity', Econometrica, 6: 1695-725. S

Mundlak, Y. (1978) 'On the Pooling of Time Series and Cross Section Data,' i.

Econometrica, 46: 69-85. z

Persson, M. and F. Wilhelmsson (2006) Assessing the Effects of EU Trade g

Preferences for Developing Countries, Mimeo, Lund University. o

Rajan, R. and L. Zingales (1998) 'Financial Dependence and Growth', American 2

Economic Review, 88: 559-86. 2

Soloaga, I. and L.A. Winters (2001) 'Regionalism in the Nineties: What Effect on 4

Trade?', North American Journal of Economics and Finance, 1: 1 -29. Swann, G.P. (2010) International Standards and Trade: A Review of the Empirical Literature, OECD Trade Policy Working Papers, No. 97, OECD Publishing.

Tahari, A., P. Brenner, E. DeVrijer, M. Moretti, A. Senhadji, G. Sensenbrenner and J. Sole (2007) Financial Sector Reforms and Prospects for Financial

Integration in Maghreb Countries, International Monetary Fund, Working Paper WP/07125, May 2007. Viner, J. (1950) The Customs Union Issue. New York: Carnegie Endowment for International Peace.

Appendix

Table A1: Main Exports Commodities and Main Partners

Algeria

Commodities Partners

Morocco Commodities

Partners

Tunisia

Commodities

Partners Egypt Commodities

Partners

Petroleum, natural gas and petroleum products 97%

USA 23.9%, Italy 14.9%, Spain 11.1%, Canada 9.6%, France 8.6%, Netherlands 4.5% (2008)

Clothing and textiles, electric components, inorganic chemicals, transistors, crude minerals, fertilisers (including phosphates), petroleum products, citrus fruits, vegetables, fish

Spain 18.7%, France 17.1%, Brazil 6.9%, USA 4.4%, Belgium 4.3%, Italy 4.2% (2008)

Clothing, semi-finished goods and textiles, agricultural products, mechanical goods, phosphates and chemicals, hydrocarbons, electrical equipment

France 28.4%, Italy 18%, Germany 9.6%, Libya 5.8%, Spain 5% (2008)

Crude oil and petroleum products, cotton, textiles, metal products, chemicals, processed food

Italy 9.5%, USA 7.1%, Spain 6.2%, India 6%, Syria 4.7%, Saudi Arabia 4.6%, Japan 4.5%, Germany 4.5% (2008)

Table A2: SITC 2 Classification

5 - Chemicals and related products, n.e.s.

51 - Organic chemicals

52 - Inorganic chemicals

53 - Dyeing, tanning and colouring materials 54- Medicinal and pharmaceutical products

55 - Essential oils and resinoids and perfume materials; toilet, polishing and cleansing preparations

56 - Fertilisers (other than those of group 272)

57 - Plastics in primary forms

58 - Plastics in non-primary forms

59 - Chemical materials and products, n.e.s.

6 - Manufactured goods classified chiefly by material

61 - Leather, leather manufactures, n.e.s., and dressed furskins

62 - Rubber manufactures, n.e.s.

63 - Cork and wood manufactures (excluding furniture)

64 - Paper, paperboard and articles of paper pulp, of paper or of paperboard

65 - Textile yarn, fabrics, made-up articles, n.e.s., and related products

66 - Non-metallic mineral manufactures, n.e.s.

67 - Iron and steel

68 - Non-ferrous metals

69 - Manufactures of metals, n.e.s.

7 - Machinery and transport equipment

71 - Power-generating machinery and equipment q

72 - Machinery specialised for particular industries n

73 - Metalworking machinery 33

74 - General industrial machinery and equipment, n.e.s., and machine parts, n.e.s. d

75 - Office machines and automatic data-processing machines &

76 - Telecommunications and sound-recording and reproducing apparatus and 3 equipment p

77 - Electrical machinery, apparatus and appliances, n.e.s., and electrical parts thereof j (including non-electrical counterparts, n.e.s., of electrical household-type equipment) &

78 - Road vehicles (including air-cushion vehicles) &

79 - Other transport equipment OO

8 - Miscellaneous manufactured articles |

81 - Prefabricated buildings; sanitary, plumbing, heating and lighting fixtures and fit- S

tings, n.e.s. r

82 - Furniture, and parts thereof; bedding, mattresses, mattress supports, cushions and t similar stuffed furnishings n

83 - Travel goods, handbags and similar containers e

84 - Articles of apparel and clothing accessories t

85 - Footwear d

87 - Professional, scientific and controlling instruments and apparatus, n.e.s. i

88 - Photographic apparatus, equipment and supplies and optical goods, n.e.s.; watches U and clocks 3

89 - Miscellaneous manufactured articles, n.e.s. o

Source: United Nations, 2009. 3

Note: n.e.s. stands for not elsewhere specified. 3

Table A3: Variable Descriptions and Sources of Data 3

Description Source 3

Dependent variables ,

Xj exports from i to j Nominal X Eurostat 0

Nij\ extensive margin Number of type of products exported Eurostat 4 from i to j

AV,j-: intensive margin Average value of the products exported Eurostat

from i to j

AQj average Average quantity of the products exported Eurostat

quantity from i to j

Table A3: Continued

Description Source

AP-: average price Average price of the products exported from i to j Eurostat

Independent variables

GDP,-: exporter's Exporter's GDP, PPP (current €) WDI

income

GDP,-: importer's Importer's GDP, PPP (current €) WDI

income

FTA dummy,- Dummy variable = 1 if the trading partners European

have an FTA, 0 otherwise Commission

D_Cumulation,j Dummy variable = 1 if the RoO allow diag- European

onal cumulation with the other MENA Commission

countries

Pa n_EuroMed_RoO ,, Dummy variable = 1 if the countries have European

adopted Pan EuroMed RoO Commission

Input_EU,- Import value of machinery from five European Economies (current €) OECD

Input_RoW,- Import value of machinery from the main producers of the RoW (current €) OECD

Dist-: distance Distances between country capitals of trading partners (km) CEPII

Colony - Dummy variable = 1 if the trading partners had colonial links in the past, 0 otherwise CEPII

Table A4: Cumulation Rules (Mediterranean Countries)

Preferential arrangement RoO/cumulation

Algeria (1 September 2005) Protocol No. 6

Euro-Mediterranean Association OJ L 297 of 15 November 2007

Agreement, OJ L 265, 10 October 2005 Bilateral, diagonal and full cumulation

Tunisia (1 March 1998) Protocol No. 4

Euro-Mediterranean Association OJ L 260 of 21 September 2006

Agreement, OJ L 97,30 March 1998, p. 2 Bilateral, diagonal and full cumulation

Morocco (1 March 2000) Protocol No. 4

Euro-Mediterranean Association OJ L 336 of 21 December 2005

Agreement, OJ L 70,18 March 2000, p. 2 Bilateral, diagonal and full cumulation

Israel (1 June 2000) Protocol No. 4

Euro-Mediterranean Association OJ L 20 of 24 January 2006

Agreement, OJ L 147, 21June 2000, p. 3 Bilateral and diagonal cumulation

Table A4: Continued

Preferential arrangement

RoO/cumulation

Palestinian Authority of the West Bank and the Gaza Strip (1 July 1997)

Euro-Mediterranean Interim Association Agreement, OJ L 187, 16 July 1997, p. 3

Egypt (1 June 2004)

Mediterranean Association Agreement, OJ L304 of 30 September 2004, p. 39

Jordan (1 May 2002)

Euro-Mediterranean Association

Agreement, OJ L 129, 15 May 2002, p. 3

Lebanon (1 March 2003 Interim Agreement)

Euro-Mediterranean Association

Agreement, OJ L 143, 30 May 2006, p. 2

Syria (1 July 1977)

Cooperation Agreement, OJ L 269, 27 September 1978, p. 2

Protocol No. 3

OJ L 187 of 16 July 1997

Bilateral cumulation

Protocol No. 4

OJ L 73 of 13 March 2006

Bilateral and diagonal cumulation

Protocol No. 3

OJ L 209 of 31 July 2006

Bilateral and diagonal cumulation

Protocol No. 4

OJ L 143, 30 May 2006, p. 73 Bilateral cumulation

Protocol No. 2 Bilateral cumulation

Source: http://ec.europa.eu/taxation_customs/customs/customs_duties/rules_origin/ preferential/article_779_en.htmpaneuro.

Table A5: Results for All Countries and Sectors with Dyadic-Sectoral Fixed Effects

Xj N„ AVj AQj APj

FTA (Barcelona 0.315*** 0.138*** 0 217*** 0.385*** _0 212***

Process) (6.464) (9.859) (4.927) (7.372) (-6.380)

Ln GDPi 0.067 -0.099*** 0.151 0.252** -0.001

(0.636) (-3.271) (1.593) (2.231) (-0.020.)

Ln GDPj 0.252** 0 112*** 0.152 -0.212* 0.275***

(2.114) (3.306) (1.413) (-1.651) (3.377)

Constant term 5.097*** 1.801*** 3.357*** 3.094*** -0.211

(29.172) (19.660) (20.574) (17.996) (-1.469)

R-squared 0.081 0.016 0.090 0.013 0.108

Number of 7,820 7,829 7,820 7,524 7,524

observations

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of export

value (current €). Incomes (GDPs) are also in natural logarithms. An Ar(1) term was added

to correct for autocorrelation.

*Significance at 10%.

"Significance at 5%.

***Significance at 1%.

Table A6: Results for All Countries and Sectors with Exporter-Year and Importer-Year and Dyadic Fixed Effects

Xj NU AVj AQj APj

FTA (Barcelona 0.666*** 0.182*** 0.481*** 0.029 (0.269) -0.045

Process) (5.989) (5.514) (4.702) (- 0.644)

R-squared 0.148 0.128 0.110 0.095 0.030

Number of 8,517 8,526 8,517 8,214 8,214

observations

Log-likelihood -13,558.750 - 3,568.870 -12,760.890 - 13,444.960 -9,960.060

RMSE 1.196 0.370 1.089 1.251 0.819

AIC 27,325.490 7,345.739 25,729.780 27,097.920 20,128.120

BIC 28,058.670 8,079.030 26,462.970 27,827.330 20,857.530

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of export value (current €). The estimation uses White's heteroscedasticity-consistent standard errors.

*Significance at 10%. "Significance at 5%. ***Significance at 1%.

Table A7: Instrumental Variables Estimation in First Differences

Nj AVij AQij APij

D.Ln GDP 1.975* 0.56 (1.374) 1.413 (1.377) - 0.087 1.800**

(1.768) (-0.071) (2.049)

D.Ln GDPy 0.519 (1.333) -0.176 0.693* 0.157 (0.373) 0.421 (1.439)

(-1.264) (1.937)

D.FTA 0.192*** 0.026(1.143) 0.166*** 0.259*** -0.130**

(2.956) (2.735) (3.577) (-2.409)

Constant term -0.028 0.014(0.759) - 0.041 0.041 (0.741) - 0.089**

(- 0.547) (-0.87) (-2.224)

R-squared 0.004 - 0.005 0.002 0.002 0.001

N 5,801 5,807 5,801 5,624 5,624

Endogeneity 0.153 8.833 0.927 4.534 6.149

Probability 0.926 0.012 0.629 0.104 0.046

Notes: t-Statistics are in brackets. The dependent variable is the natural logarithm of export

value (current €). Incomes (GDPs) are also in natural logarithms. The estimation uses

White's heteroscedasticity-consistent standard errors.

*Significance at 10%.

"Significance at 5%.

***Significance at 1%.