The Impact of a Customs Union between Turkey and the EU on Turkey's Exports to the EU

Abstract

This article investigates Turkey's sectoral trade flows to the EU based on panel data from the period 1988 to 2002. Turkey's 16 most important export sectors are analysed. Emphasis is placed on the role of price competition, EU protection and transport costs in the export trade between Turkey and the EU. The empirical model used is an extended version of the gravity model. This study is also a contribution to the current discussion of whether Turkey should be granted full EU membership or a privileged partnership with the EU, which for Turkey would mean improved access to the EU market for its products, among other benefits. Our investigation focuses on the latter policy outcome: the impact of deepening the customs union between Turkey and the EU and applying the common agricultural policy (CAP) to Turkish agricultural exports. To this end, the impact of the 1996 customs union covering most industrial goods and processed agricultural goods, is evaluated on a sectoral level. We also perform simulations to quantify the impact of the potential inclusion of agricultural goods, as well as iron and steel and products thereof, into the full customs union between Turkey and the EU which is still to come. Copyright (c) 2007 The Author(s); Journal compilation (c) 2007 Blackwell Publishing Ltd.

Full text

Page 1

The Impact of a Customs Union between Turkey
and the EU on Turkey’s Exports to the EU
FELICITAS NOWAK-LEHMANN
Georg-August University, Goettingen
DIERK HERZER
Georg-August University, Goettingen
INMACULADA MARTINEZ-ZARZOSO
University Jaume I, Castellón
SEBASTIAN VOLLMER
Georg-August University, Goettingen
Abstract
This article investigates Turkey’s sectoral trade flows to the EU based on panel data
from the period 1988 to 2002. Turkey’s 16 most important export sectors are analy-
sed. Emphasis is placed on the role of price competition, EU protection and transport
costs in the export trade between Turkey and the EU. The empirical model used is an
extended version of the gravity model. This study is also a contribution to the current
discussion of whether Turkey should be granted full EU membership or a privileged
partnership with the EU, which for Turkey would mean improved access to the EU
market for its products, among other benefits. Our investigation focuses on the latter
policy outcome: the impact of deepening the customs union between Turkey and
the EU and applying the common agricultural policy (CAP) to Turkish agricultural
exports. To this end, the impact of the 1996 customs union covering most industrial
goods and processed agricultural goods, is evaluated on a sectoral level. We also
perform simulations to quantify the impact of the potential inclusion of agricultural
goods, as well as iron and steel and products thereof, into the full customs union
between Turkey and the EU which is still to come.
Introduction
On 1 May 2004, a further round of EU enlargement became a reality, expand-
ing the Union to include a total of 25 Member States. At that point in time,
Turkey did not yet qualify for EU accession even though promises of a
customs union (CU) and common market between the EU and Turkey had
JCMS 2007 Volume 45. Number 3. pp. 719–743
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148,
USA

Page 2

been made as long ago as 1963 in the Ankara Treaty. On 17 December 2004,
the EU heads of state finally decided to begin accession negotiations with
Turkey based on the recommendation of the Commission, which has given a
conditional ‘yes’. Turkey’s chances of EU accession depend more heavily, at
the moment, on political factors (fulfilment of Copenhagen criteria1) than on
economic factors2according to EU Commissioner of Enlargement Günter
Verheugen’s Progress Report (see Presidency Conclusions of Brussels Euro-
pean Council 16–17 December 2004, 1 February 2005). Furthermore, the
decision by the European Council of 17 December 2004 to open negotiations
with Turkey on 3 October 2005 is conditional on the enlargement of the
customs union to include Cyprus.
Nonetheless, the EU and Turkey already have well-integrated economies
as far as a large part of the trade in goods is concerned. An incomplete CU
between the EU-15 and Turkey was created on 1 January 1996, guaranteeing
free circulation of industrial goods and processed agricultural products.
Quotas were prohibited in the CU with the EU. In addition, voluntary restraint
agreements (VRA) concerning trade in textiles were abolished. Turkey’s
commercial and competition policies had to be harmonized with those of the
EU and a level of intellectual property protection similar to that in the EU was
agreed upon.
The CU with the EU-15 does not deal with agriculture or services. Exemp-
tions do apply for iron and steel and products thereof and textile trade is
impeded by EU’s antidumping actions and safeguard measures. Nonetheless,
there is a commitment on the part of both the EU and Turkey to expand and
strengthen the CU.Agriculture will be included through ongoing negotiations
on mutual concessions, with the objective of establishing a free trade area
(FTA). Turkey and the EU are pushing ahead to extend the CU to cover new
areas such as services and public procurement.
In preparing for EU accession, Turkey has concluded free trade agree-
ments with the majority of the countries that joined in the 2004 enlargement:
Czech Republic, Slovakia, Hungary, Lithuania, Estonia, Latvia, Slovenia and
Poland (Ülgen and Zahariadis, 2004).
Nevertheless, it is reasonable to expect Turkey’s entrance into the EU no
earlier than 10 to 15 years from now according to the EU Commissioner for
Enlargement Günter Verheugen and Germany’s former Foreign Minister,
Joschka Fischer (Fischer, 2004).
1The ‘Copenhagen criteria’ have three components: a) political stability: democracy, human rights,
protection of minorities; b) economic criteria: a market economy and competition with the older Member
States in the single market; c) the acquis criteria: adoption of EU law and acceptance of the objectives of
the Union.
2See Pre-Accession Economic Programme (2003) at: http://ekutup.dpt.tr/ab/kep/pep2003.pdf.
720
FELICITAS NOWAK-LEHMANN ET AL.
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 3

It is the purpose of this article to analyse Turkey’s most important export
sectors, to evaluate the impact of the CU of 1996 on industrial goods (such as
plastics and rubber, textiles and clothing, machinery and furniture) and to
simulate the impact of a strengthened and expanded CU on Turkey’s dis-
criminated sectors (such as agriculture and iron and steel). For a forecast of
future trade flows and Turkey’s chances on the EU market, it is necessary to
assess underlying trade structures and the determinants of current trade flows.
In this study, emphasis will be placed on the role of price competition,
protection and transport costs in the export trade between Turkey and the EU.
Relying mainly on EUROSTAT’s trade database COMEXT (Commission,
2003), we discarded countries with incomplete data (Austria, Belgium,
Finland, Luxemburg, Sweden) and concentrated instead on Denmark, France,
Germany, Greece, Ireland, Italy, the Netherlands, Portugal, Spain and the UK.
We analysed Turkish exports on a two-digit level, based on the harmonized
systems (HS) classification.
I. Turkey’s Exports to the EU
According to the Global Trade Negotiations homepage ‘Turkey Summary’
(2004), Turkey’s principal exports are textiles and clothing, followed by
agricultural products, iron, steel and machinery. Its largest trading partner
worldwide is Germany, followed by Italy. Turkey’s agricultural sector is the
largest3of all the OECD countries, accounting for about 17 per cent of GDP,
20 per cent of exports and 40 per cent of the labour force. Its production
includes tobacco, cotton, grain, olives, sugar beets, pulses, citrus and live-
stock. Cotton, fruit and vegetable production has increased dramatically in
recent years due to irrigation efforts and government support. The govern-
ment employs multiple incentives to promote exports, including output and
input subsidies, tax credits, guarantees and insurance programmes.
As far as agricultural products are concerned, competition comes mainly
from the EU. Greece, Spain and Italy are serious competitors with Turkey in
the field of edible vegetables (olives, pulses), edible fruit (citrus) and pro-
cessed agricultural products. Greece has proved to be one of the most sig-
nificant competitors with Turkey both in terms of agricultural and industrial
products (ICAP, 2004).
Steel and iron are produced by a variety of countries, among them China,
India, Russia, Ukraine, Brazil and Australia. Hence Turkey – being a smaller
producer – has to deal with stiff competition in the production of iron and
steel and products thereof.
3In percentages of GDP, exports and the labour force employed in the agricultural sector.
THE IMPACT OF A EU–TURKEY CUSTOMS UNION
721
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 4

With respect to textiles and clothing, a major concern for Turkey is the
expiration of quotas on textiles and clothing on 1 January 2005. Abolition of
quotas will mainly benefit low-cost producers such as China. Chinese textile
exports constitute a third of global trade in textiles and clothing.
When looking at machines, mechanical appliances and vehicles, Turkey’s
principal competitors outside the EU are the central and eastern European
countries (CEECs). Turkey faces stiff competition from Poland in the trade
with vehicles.
In Table 1 we list the 16 largest sectors in which Turkey is exporting to the
EU.
We consider averages of sectoral export values over the period 1988 to
2002 in order to smooth out peaks and valleys. As far as agriculture is
concerned, we selected sectors with an export value of more than €100
million (yearly average 1988–2002). Concerning industrial sectors, the
minimum export value was set in most cases at €200 million (yearly average
1988–2002). Pre-selection of the 16 sectors was based on the 30 largest
sectors in 2002.
Table 1 highlights the fact that agricultural production and food process-
ing are not particularly dynamic sectors given their low growth rates. With
respect to cotton, the literature suggests a sharp increase resulting from
irrigation and government programmes, such that the figure of 5.2 per cent
export growth presented in Table 1 probably underestimates the future devel-
opment. Sectors 72 and 73 (iron and steel) and sectors 84–94 (machinery,
vehicles and furniture) can be considered the most dynamic export sectors. In
terms of export shares, the most important sectors are articles of apparel,
motor vehicles, electrical machinery, machinery and mechanical appliances,
whereas agriculture’s export share was surprisingly low4(also compare
TÜSIAD (2004).
II. Factors Influencing Trade According to the Gravity Model
One of the most established models for empirical studies in international
trade is the gravity model. In recent decades, the gravity model has performed
remarkably well as an empirical framework for explaining bilateral trade.
There exist a huge number of empirical applications of the gravity model that
have contributed to the improvement of performance of the gravity equation.
Some of them are closely related to our work. First, in recent papers, Mátyás
(1997), Chen and Wall (1999), Breuss and Egger (1999) and Egger (2000)
have improved the econometric specification of the gravity equation and
4In 2002 agriculture’s export share was 6 per cent, whereas in 1970 it still reached 75 per cent!
722
FELICITAS NOWAK-LEHMANN ET AL.
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 5

Table 1: Turkey’s 16 Most Important Export Products in the EU Market in 2002a
HS code
Sector
Av. export value in mill.
current ECU (1988–2002)
Annual percentage
change (1988–2002)
Export shareb
in 2002
Serious extra-EU-15
competitorc
07
Edible vegetables
126.4
5.5%
0.5%
–
08
Edible fruit and nuts
689.0
6.3%
2.4%
–
20
Preparations of vegetables, fruit, nuts
288.6
10.3%
1.1%
–
39
Plastics and plastic products
100.8
22.8%
0.7%
Brazil
40
Rubber and articles thereof
160.8
23.6%
1.1%
Brazil
52
Cotton
365.7
5.2%
2.2%
China
55
Man-made staple fibres
211.5
7.4%
0.8%
China
61
Articles of apparel and clothing;
knitted or crocheted
2050.6
14.7%
12.4%
China
62
Articles of apparel and clothing; not
knitted or crocheted
1405.2
12.5%
9.0%
China
63
Other made up textile articles
367.9
16.1%
3.5%
China
72
Iron and steel
281.2
34.0%
5.9%
China
73
Articles of iron and steel
214.3
22.0%
3.5%
China
84
Machinery and mechanical appliance
429.0
30.0%
5.9%
CEEC
Poland
85
Electrical machinery and equipment
771.6
29.5%
7.9%
CEEC
Poland
87
Vehicles other than railway or tramway
rolling stock
538.2
44.6%
6.2%
CEEC
Poland
94
Furniture, medical and surgical
furniture, bedding, mattresses
106.0
29.8%
0.9%
CEEC
Poland
Sources: Commission (2003) EUROSTAT’s COMEXT CD ROM, ‘Intra- and Extra-EU trade, Annual Data, Combined Nomenclature’; Authors’ own
calculations and TÜSIAD (2004).
Notes:aIn the EU-15 market. Commission (2003) treats the trade flows of the EU-15 countries with all the other countries as extra-EU trade. Therefore, trade
with the CEEC countries is considered extra-EU trade;bShare of EU exports of sector k in total exports to the EU-15;cBased on TradeCAN (Competitiveness
Analysis of Nations) 2002 CD-ROM (World Bank, 2002).
THE IMPACT OF A EU–TURKEY CUSTOMS UNION
723
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 6

highlight the advantages of using panel data methodology. Second,
Bergstrand (1985), Helpman (1987), Wei (1996), Soloaga and Winters
(1999), Limao and Venables (2001), Bougheas et al. (1999) and Anderson
and Wincoop (2003) among others, have contributed to the refinement of
the explanatory variables considered in the analysis and to the addition of
new variables.
According to the generalized gravity model of trade, the volume of exports
between pairs of countries, Xij, is a function of their incomes (GDPs), their
populations, their geographical distance and a set of dummies,
XY Y N N D A u
iji
0
ij
j ij ij
ij
=β
ββββββ
123456
(1)
where Yi(Yj) indicates GDPs of the exporter (importer), Ni(Nj) are popula-
tions of the exporter (importer), Dijmeasures the distance between the two
countries’ capitals (or economic centres) and Aijrepresents any other factors
aiding or preventing trade between pairs of countries. The error term is uij.An
alternative formulation of equation (1) uses per capita income instead of
population,
X Y Y YH YH D A u
iji
0
ij
j ijij
ij
= γ
γγγγγγ
123456
(2)
where YHi(YHj) are the exporter (importer) GDP per capita. The two models
above are equivalent and the coefficients are expressed as: b3= -g3; b4= -g4;
b1= g1+ g3; b2= g2+ g4. The second specification is usually chosen when
the gravity model is used to estimate bilateral exports for specific sectors
(Bergstrand, 1989), whereas the specification given by equation (1) is often
used to estimate aggregated exports (Endoh, 2000).
For estimation purposes, model (2) in log-linear form for a single year, is
expressed as,
lX lY
1
lYlYH lYHlDPu
ijijijijhijhij
h
=+++++++
∑
γγγγγγδ
02345
(3)
where l denotes variables in natural logs.
trade dummy variables. Pijhtakes the value one when a certain condition is
satisfied (e.g. belonging to a trade bloc, being part of a customs union), zero
otherwise. Dummy variables for trading partners sharing a common language
and common border, as well as trade bloc dummy variables evaluating the
effects of preferential trading agreements, are usually considered. The coef-
ficients of all these trade variables (dh) are expected to be positive.
A high level of income in the exporting country indicates a high level of
production, which increases the availability of goods for exports. Therefore
we expect g1 to be positive. The coefficient of Yj, g2 is also expected to be
δhijh
h
P
∑
is a sum of preferential
724
FELICITAS NOWAK-LEHMANN ET AL.
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 7

positive, since a high level of income in the importing country suggests higher
imports. The coefficient estimate for exporter’s per capita income, g3, is
interpreted by Bergstrand (1989) as a proxy for the exporter’s K/L ratio. It
may carry a positive or negative sign, depending on whether the gravity
equation is estimated for a capital- or labour-intensive industry. The coeffi-
cient of the importer per capita income, g4, also has an ambiguous sign: it may
be negative when the products imported are necessities and positive when
they are luxuries (Bergstrand, 1989). The distance coefficient is expected to
be negative since it is a proxy of all possible trade cost sources. Traditionally,
the gravity model uses distance to model transport costs. However, recently
Bougheas et al. (1999) showed that transport costs are a function not only of
distance but also of public infrastructure. They augmented the gravity model
by introducing additional infrastructure variables (stock of public capital and
length of motorway network). Their model predicts a positive relationship
between the level of infrastructure and the volume of trade, which is sup-
ported using data from European countries.
III. Empirical Application of the Gravity Model to Turkey–EU Trade
Augmented Gravity Model and Estimation Techniques
A variant of the gravity equation, see equations (4) and (5) below, is used to
model bilateral export flows from Turkey to the EU (see Martínez-Zarzoso
and Nowak-Lehmann, 2003, 2004). Due to missing data, we consider only
Turkey’s exports to Germany (D), Denmark (DK), Spain (E), France (F), the
United Kingdom (UK), Greece (EL), Ireland (IRL), Italy (I), the Netherlands
(NL) and Portugal (P). Export data, described in section II, covers 16 sectors
at the two-digit HS chapters. Sources of the data are outlined in theAppendix.
The period covered goes from 1988 to 2002. We have a maximum of ten
cross-sectional5trade flows and 15 years, resulting in a maximum of 150
observations per sector. The number of observations varies depending on the
product studied. A log-linear specification was selected.
We deviate from the gravity model presented in section II, equation (3), in
several respects. First, we do not focus on infrastructure and in particular not
on terrestrial infrastructure (i.e. the circumstances of arriving at the domestic
port and departing from the foreign port), but on maritime transport costs
when measuring distance. For this purpose, we scaled geographical distance
(actual nautical miles) by using the freight cost index to construct a
5But not in all sectors! For example, we have a large amount of missing data as far as Portugal’s imports
in sectors 07 and 20 are concerned.
THE IMPACT OF A EU–TURKEY CUSTOMS UNION
725
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 8

new transport cost variable. We assumed that merchants would use sea trans-
port whenever possible, given the fact that a certain quantity transported by
ship (40-foot containers) costs about one-fifth of the same quantity trans-
ported by road (13.6 m trailer). In 2003, maritime transportation was the
leading transportation method for Turkish exports, followed by road trans-
port.6We do not consider land transport costs here since they are the same for
all exporting countries and independent of the export port (Turkey, Bulgaria,
Ukraine) once the destination (foreign) port (e.g. Hamburg) has been reached.
But still it has to be noted that land transport costs of the exporting country
(e.g. Turkey, from Ankara to Istanbul) will differ from exporting country to
exporting country (Turkey, Ukraine, Georgia) and should therefore be con-
sidered. However, they are partly incorporated into the income variable of the
exporting country. A country with higher GDP will also have better public
infrastructure.
Second, concerning economic distance, we use differences in incomes
between trading countries, a variable similar to that used in Arnon et al.
(1996) and in McPherson et al. (2000). Our variable is constructed as the
absolute difference in per capita incomes in purchasing power parities (PPP).
We can identify two conflicting effects of this variable on trade. On the one
hand, when the trading countries have very different per capita incomes,
lower economic distance might foster trade, on the basis of the Linder (1961)
model.According to this effect, countries tend to increase their bilateral trade
in similar products when their per capita incomes are more similar. We
therefore expect more trade to be intra-industry trade (countries should both
export and import the same goods) when per capita incomes converge.
On the other hand, higher economic distance might foster inter-industry
trade (countries import and export different goods) if we consider the
Heckscher-Ohlin (H-O) model. H-O centres on expected trade patterns when
countries have different factor endowments, but similar tastes. Per capita
income differences can represent inter-country differences in factor scarcity.
We expect present trading patterns to be affected by both factors. For some
commodities, the Linder effect will dominate the H-O effect and economic
distance will have a negative effect on trade, whereas for others the opposite
might occur, in which case economic distance will have a positive effect on
trade.
Finally, a real exchange rate variable is added to our specification
(Bergstrand,1985,1989;SoloagaandWinters,1999).WecalculatedTurkey’s
6Maritime transportation was used for 49.2 per cent of Turkish exports (by value) and road transportation
for 43.0 per cent of Turkish exports (by value) in 2003, with a steady increase in the importance of sea
transport in the last decade (IGEME – Export Promotion Center of Turkey, 2004).
726
FELICITAS NOWAK-LEHMANN ET AL.
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 9

and its competitors’ bilateral real effective7exchange rates (price quotation
system) taking into account protection. Average tariffs imposed by the EU
and EU subsidies enter the formula (WTO, 2000, p. 101).All the calculations
are shown in the Appendix.
Exports from country i to country j in period t of commodity k are then
modelled as:
lx lytlydiff
1
lreerltcindex
ijkt ijkijt ijtijkt ijt
=++++αββββ
023
+ +μijkt
(4)8or
lx lytlydiff
1
lreerlreer
ijktijk ijtijt ijktijkt
*
=+++++αββββ
023
β βμ
4ldtcijtijkt
* +
(5)9
where lxijktis the natural logarithm of exports of sector k from country i to
country j in period t. The total income of the trading countries (in purchasing
power parities, PPP) is lytijt. This summarizes the impact of the income of
trading pairs on exports. The natural logarithm of differences in per capita
income in absolute terms and in PPP between the trading countries is lydiffijt,
while lreerijkt is the real effective exchange rate (price quotation system),
taking into account sector-specific protection.Accordingly, lreerikjt* is the real
effective exchange rate of Turkey’s extra-EU competitors. We assume the
competitors’ (extra-EU price competition) real effective exchange rate to be
especially relevant in textiles and clothing (sectors 52–63) and in iron and
steel (sectors 72 and 73), where China is a serious competitor. Concerning
plastics and rubber and products thereof (sectors 39 and 40), we treat Brazil
as the main competitor and with respect to machinery, vehicles and furniture
(sectors 84, 85, 87 and 94) we presume that Poland is in competition with
Turkey. We have information suggesting that extra-EU competition is not
very influential in agriculture (sectors 07, 08, 20), but of course intra-EU
competition is (ICAP, 2004). ltcindexijt stands for the natural logarithm of
transport costs between countries i and j and ldtcijt* is used in equation (5) to
signal the difference in transport costs between Turkey and its main extra-EU
competitor.
The construction of the variables is described in the Appendix. aijkstands
for the specific country-pair effects for sector k and allows us to control for all
omitted variables that are cross-sectionally specific but remain constant over
time, such as contiguity, language and cultural ties.
7Effective implies that EU import tariffs and subsidies are taken into account. This definition differs from
the IMF definition, which understands real effective exchange rates as multilateral trade-weighted real
exchange rates.
8Partial adjustment model:
lxlyt lydifflreerltcindex
ijktijkijtijtijtijt
=++++αββββ
0123
9Partial adjustment model:
lxlytlydifflreer lreer
ijkt ijkijtijtijtijt
=+++++αββββ
0123
*
+λ λμ⋅+
−
lxijkt ijkt
1
(4)′
β
4
l ldtclx
ijt
* + ⋅
ijktijkt
+
−
λμ
1
(5)′
THE IMPACT OF A EU–TURKEY CUSTOMS UNION
727
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 10

Expanding the CU between Turkey and the EU is expected to have a
noticeable impact on Turkish exports facing high or very high protection in
the EU, such as agricultural products, iron and steel. Turkey’s price competi-
tiveness is expected to be decisive for export success in all sectors under
investigation. Expectations about the role of transport costs, differences in
transport costs and differences in per capita income in Turkey’s export trade
are less conclusive. The importance of those factors is believed to vary from
sector to sector.
Panel data methodology is used to estimate equations (4) and (5). We
mainly apply the seemingly unrelated regression (SUR) technique, thus con-
trolling for correlation between cross-sections. The generalized method of
moments (GMM) is the method of choice for the partial adjustment version of
the models. However, in some cases, in which we utilize pooled least squares
(PLS), neither the SUR technique nor the GMM technique can be applied,
due either to an insufficient number of observations or to the lack of accept-
able instruments. The use of panel data methodology has several advantages
over cross-section analysis. First, panels make it possible to capture the
relevant relationships among variables over time. Second, a major advantage
of using panel data is the ability to monitor the possibly unobservable trading-
partner pairs’ individual effects. When individual effects are omitted, OLS
estimates will be biased if individual effects are correlated with the regres-
sors. Mátyás (1997), Chen and Wall (1999) and Egger (2000) present a
discussion of the advantages of using this methodology to estimate the gravity
equation of trade.
Panel unit-root tests are conducted for imports in real terms (aggregated),
for the real exchange rate, total income, per capita income differences and
transport costs. Stochastic trends that express themselves as autocorrelation
of the error terms10are found to prevail in all series analysed.
Due to missing data and possibly an insufficient number of observations,
period SUR11cannot be performed. However, we control for autocorrelation
of the disturbances by plugging in AR-terms whenever they prove to be
significant.
Partial adjustment models are used mainly in agricultural sectors to iden-
tify slower reactions in this sector. When running the regressions for the
sectors already participating in the CU in 1996, a step dummy variable is
plugged in to capture a possible upward shift in exports caused by the CU
betweenTurkey and the EU.This time dummy takes a value of 0 in the period
10Non-stationary, integrated series can be corrected in two ways: (1) by taking first, second or third
differences of the series or (2) correcting for autocorrelation.This is due to the fact that autocorrelation and
non-stationarity are inter-linked.
11Which controls for correlation between periods.
728
FELICITAS NOWAK-LEHMANN ET AL.
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 11

of 1988–95 and of 1 in the period of 1996–2002.The step dummy was left out
of the analysis whenever it was insignificant.
Simulations are based on 1988–95 data in sectors that became part of the
CU in 1996 to derive the effect of a CU based on pre-CU coefficients.12In
contrast, simulations rely on 1988–2002 data if the sectors were not yet
integrated into a CU by 2002 (or up to now). We assume that a change in
tariffs has the same effect on exports as a change in subsidies according to the
construction of the real effective exchange rate variable.The coefficients used
in simulating agricultural exports (Table 2) are based on the fixed effects (FE)
model (sector 07, 08, 20). The coefficients entering the simulations concern-
ing industrial products (Tables 3 to 6) stem from the long-run model, which
does not include a lagged endogenous variable and works with a common
intercept to simplify the simulations and alleviate the computations. In the
latter model, the real effective exchange rate elasticities differ slightly from
the ones computed via the FE-model. All our simulations are based on
multiple-regression equations derived from the models described above.
Nonetheless, the impact of a change in protection could also be computed by
means of standardized real effective exchange rate coefficients,13thus con-
sidering each variable’s contribution to changes in exports. To make our
simulation results comprehensible, a separate line with the standardized real
effective exchange rate coefficients is added in the simulation segment.
Some Caveats
We hope to contribute to the EU–Turkey CU debate by providing the EU
demand elasticities for Turkish exports, which enter the simulations per-
formed. Nevertheless, it must be admitted that the simulation results hinge
very strongly on the EU tariff and subsidy rates chosen. Simple statements on
the ‘true’ extent of prevailing sectoral tariffs or tariff-like duties are rather
difficult. According to Grethe (2004b), there still exist some types of market
barriers against Turkish products, even though almost all ad valorem tariffs
have been abolished in the agricultural sector. Seasonal tariffs apply to four
kinds of fruit and nine vegetables, thus complicating computations of tariffs.
High specific duties are imposed on core products of the CAP and specific
duties apply to many processed products. Tariff statements are further
12It is well known that forecast errors (simulation errors) can be two-fold: (1) regression coefficients might
(slightly) change under a CU, (2) the magnitude and distribution of the disturbances under a CU are
unknown. We circumvented the second problem by computing regression line values for both actual
exports/imports and simulated exports/imports.
13In a bivariate regression model with only one independent variable (reer) the impact of a change in reer
on exports could be calculated by multiplying the reer-elasticity with the change in reer. In the multiple
regression model, one must consider reer’s relative contribution to a change in exports.
THE IMPACT OF A EU–TURKEY CUSTOMS UNION
729
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 12

Table 2: Estimation and Simulation Results for Agricultural Products and Processed Agricultural Productsa
Sector 07
Edible vegetables
Sector 08
Edible fruit and nuts
Sector 20
Preparations of vegetables, fruit, nuts
EU protection in this sector (T, S)b
T = 0.12
S = 0.05
T = 0.12
S = 0.05
T = 0.25
S = 0.05
CU
No
No
No
Regression results based on eq. (4)′ or eq. (4)
Estimation technique
SUR
SUR
SUR
Fixed effects
Yes
Yes
Yes
AR-termc
No
No
Yes (0.64***)
Partial adjustmentd
l
Yes
0.66*** (10.48)
Yes
0.15*** (3.80)
No
–
Lyt
-6.37*** (-8.31)
0.36 (1.72)
0.07 (0.12)
Lydiff
2.35*** (4.62)
-0.92*** (-8.89)
1.46*** (5.41)
Lreer
1.16*** (4.81)
1.04*** (14.80)
1.46*** (12.44)
Ltcindex
-1.28*** (-3.49)
-0.69*** (-3.83)
1.03*** (3.15)
S.E. of regression
1.05
1.05
1.03
R-squared
0.91
0.99
0.97
DW
1.75
1.70
2.07
Obs.
70 (5 EU countries)
140 (10 EU countries)
130 (9 EU countries)
Simulation results based on 1988–2002 data
Standardized reer elasticity (base period)
1.17
0.94
1.53
Impact of CU (abolition of tariffs)
+14.0%
+12.5%
+38.5%
Impact of trade (CAP) integratione
+21.0%
+18.7%
+49.3%
Source: Authors’ own data.
Note: t-values are stated in brackets; *** signal the tolerated error-level and stand for a = 1%;aA very thorough discussion of the CU on Turkish agriculture
can be found in Grethe (2004a). This dissertation contains computations of changes in prices and output, in producer and consumer surplus and net budgeteffects due to a CU between Turkey and the EU. In contrast to Grethe, we concentrate on the trade effects of a CU between Turkey and the EU;bT = tariff rate
(WTO Trade Policy Review EU, 1995, 2000; S = subsidy rate (qualitative information to be found in Supper (2001), converted into a very rough subsidy
equivalent);cAnAR-Term has been included whenever it turned out to be significant, thus correcting for autocorrelation of the disturbances and non-stationarity
of the series;dA partial adjustment model has been used whenever the adjustment coefficient was significant, thus modeling the lagged adjustment of exports
with respect to changes in transport costs, the real effective exchange rate etc;eTrade integration could imply that Turkish exports are freed from tariffs and
are given a support (subsidy) corresponding to the subsidies prevailing in this sector in the EU.
730
FELICITAS NOWAK-LEHMANN ET AL.
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 13

complicated by the entry price system of the EU, which acts like a tariff on
Turkish vegetables and fruit,14thus erasing preferential tariffs granted to
Turkey (CONSLEG, 1984). Besides, some preferential tariffs for Turkish
agricultural and industrial goods have also been annulled by EU safeguard
measures taking the form of temporary tariffs.
14‘[T]he introduction of a countervailing charge on imports of fruit and vegetables originating in Turkey
is equivalent to finding that the condition provided for in Article 1 (2) of Regulation (EEC) No 562/81 is
not fulfilled; whereas application of the preferential tariff should at the same time be suspended for the
products in question’ (CONSLEG, 1984, 2).
Table 3: Estimation and Simulation Results for Plastics and Rubber Trade
Sector 39
Plastics and plastic
products
Sector 40
Rubber and articles
thereof
EU protection in this sector
(T, S)a
CU
T = 0.07
S = 0.00 (low protection)
Yes, since 1996
Regression results based on eq. (5)
SUR
Yes
No
No
-0.56 (-1.08)
1.59*** (6.23)
0.72*** (6.09)
Not significant
-2.20*** (-11.61)
Not significant
Not significant
1.05
0.77
1.94
130
Simulation results based on 1988–95 data
0.31
T = 0.02
S = 0.00 (low protection)
Yes, since 1996
Estimation technique
Fixed effects
AR-term
Partial adjustment
Lyt
Lydiff
Lreer
Lreer*
Ltcindex
Ldtc*
Time-dummy
S.E. of regression
R-squared
DW
Obs.
SUR
Yes
Yes
No
-0.83 (-1.39)
-0.15 (-0.44)
1.63*** (7.76)
Not significant
-3.71*** (-8.87)
Not significant
Not significant
1.04
0.94
1.98
140
Standardized reer elasticity
(base period)
Impact of CU
(abolition of tariffs)
0.65
+2.13% increase in
export level
+1.31% increase in
export level
Source: Authors’ own data.
Notes: t-values are stated in brackets. *** signal the tolerated error-level and stand for a = 1%;aThe tariff
rates are taken from WTO Trade Policy Review EU, 2000. The degree of subsidization seems to be low
according to the information collected in Supper (2001).
THE IMPACT OF A EU–TURKEY CUSTOMS UNION
731
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 14

Table 4: Estimation and Simulation Results for Textiles and Clothing Trade
Sector 52
Cotton
Sector 55
Man-made
staple fibres
Sector 61
Articles of
apparel and
clothing;
knitted or
crocheted
Sector 62
Articles of
apparel and
clothing; not
knitted or
crocheted
Sector 63
Other made up
textile articles
EU protection
in this sector
(T, Quotas)
T = 0.08
Yes, quotas
Very high
protection
Yes, since 1996
T = 0.09
Yes, quotas
Very high
protection
Yes, since 1996
T = 0.13
Yes, quotas
Very high
protection
Yes, since 1996
T = 0.13
Yes, quotas
Very high
protection
Yes, since 1996
T = 0.10
Yes, quotas
Very high
protection
Yes, since 1996 CU
Regression results based on eq. (5)
SUR Estimation
technique
Fixed effects
AR-term
Partial
adjustment
Lyt
SURSURSURSUR
Yes
No
No
Yes
No
No
Yes
Yes
Yes
Yes
No
No
Yes
Yes
No
2.55***
(4.18)
-0.97***
(-2.50)
0.78**
(3.05)
-0.87***
(-3.73)
-0.09
(-0.30)
Not significant
8.79***
(15.56)
-1.49***
(-3.66)
3.76***
(17.32)
-0.37**
(-2.34)
2.87***
(10.52)
Not significant
1.10***
(2.88)
0.26
(1.10)
1.21***
(7.86)
0.14
(0.92)
-0.16
(-0.64)
Not significant
6.33***
(20.32)
-0.14
(-0.60)
1.96***
(18.37)
-0.21**
(-2.41)
0.56**
(2.04)
Significant, but
negative
1.04
0.23
(0.34)
1.44***
(3.39)
1.55***
(11.03)
-0.43***
(-5.91)
-2.46***
(-9.98)
Significant and
positive
1.05
Lydiff
Lreer
Lreer*
Ldtc*
Time-dummy
S.E. of
regression
R-squared
DW
Obs.
1.031.041.05
0.87
1.77
150
0.78
1.73
140
0.98
2.05
130
0.94
1.80
150
0.95
2.06
140
Simulation results based on 1988–95 data
0.71Standardized
reer elast.
Impact of
abolition of
tariffs and
quotas
0.60 0.090.18 0.72
+4.8% increase
in export level
+6.5% increase
in export level
+1.2% increase
in export level
+2.4% increase
in export level
+7.2% increase
in export level
Source: Authors’ own data.
Note: t-values are stated in brackets. *** and ** signal the tolerated error-level and stand for a = 1% and
5% respectively.
732
FELICITAS NOWAK-LEHMANN ET AL.
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd

Page 15

Table 5: Estimation and Simulation Results for the Iron and Steel Trade
Sector 72
Iron and steel
Sector 73
Iron and steel products
EU protection in this sector
(T, S)a
CU
T = 0.03 (Tmaxb= 0.07)
S = 0.10 (high protection)
No, excluded
T = 0.03 (Tmax= 0.05)
S = 0.10 (high protection)
No, excluded
Regression results based on eq. (5)
PLS
Yes
Yes
No
6.62
(1.59)
6.89***
(3.92)
5.19***
(3.77)
-2.75***
(-2.60)
3.78
(1.40)
No
1.11
0.85
2.18
128
Estimation technique
Fixed effects
AR-term
Partial adjustment
Lyt
SUR
Yes
Yes
No
5.51***
(8.61)
1.60***
(5.06)
1.57***
(9.83)
0.02
(0.18)
-0.55
(-1.07)
No
1.04
0.96
2.22
120
Lydiff
Lreer
Lreer*
Ldtc*
Time-dummy
S.E. of regression
R-squared
DW
Obs.
Simulation results based on 1988 to 2002 data
Standardized reer elasticity
(base period)
Impact of CU
(abolition of tariffs)
0.500.82
+1.5% increase in export level +2.5% increase in export level
Source: Authors’ own data.
Notes: t-values are stated in brackets. *** signal the tolerated error-level and stand for a = 1%;aThe tariff
rates are taken from WTO Trade Policy Review EU, 2000. The degree of subsidization seems to be low
according to the information collected in Supper (2001);
information, this tariff is applied to Chinese iron and steel imports.
bTmax= maximum tariff. According to our
THE IMPACT OF A EU–TURKEY CUSTOMS UNION
733
© 2007 The Author(s)
Journal compilation © 2007 Blackwell Publishing Ltd