A partnership with academia

Building knowledge for trade and development

Vi Digital Library - Text Preview

Structural Changes in Exports of an Emerging Economy : Case of Turkey

Working paper by Saygılı, Hülya, Saygılı, Mesut, 2011

Download original document

This study examines structural changes in Turkish export supply and demand functions, which can be used as a good reference in understanding the determinants of trade performance of an emerging market economy. Results show that as the export shares of new non-traditional commodities that have not only higher import and income sensitivity, but also lower real exchange rate elasticity than the traditional commodities increases, coefficients of the total export functions changes accordingly. Process of transformation accelerates during the major reform periods and economic crisis.







New York and Geneva, 2011


Structural Changes in Exports of an Emerging Economy: Case of

Hülya Saygılı
Central Bank of the Republic of Turkey, Ankara

Mesut Saygılı

United Nations Conference on Trade and Development, Geneva


This study examines structural changes in Turkish export supply and demand functions, which
can be used as a good reference in understanding the determinants of trade performance of an
emerging market economy. Results show that as the export shares of new non-traditional
commodities that have not only higher import and income sensitivity, but also lower real
exchange rate elasticity than the traditional commodities increases, coefficients of the total export
functions changes accordingly. Process of transformation accelerates during the major reform
periods and economic crisis.

Keywords: Turkish Exports, Structural Change, Kalman Filter, Commodity Based Exports, Panel

JEL Classifications: F1, F4

* The opinions expressed in this study are those of the authors and should not necessarily reflect the views
of the UNCTAD secretariat or the Central Bank of the Republic of Turkey.


1. Introduction

The concept of structural change in international trade pattern is not a new phenomenon

in the international trade literature. Improvements in information and communication

technologies, decreases in transportations costs, reductions in barriers to trade and advances in

production technologies allow development of new global production networks, such that

production can be divided into different stages and performed at different locations. Division of

production process across countries, meanwhile, not only causes structural change but also

increases the import dependence (Krugman, 1995; Feenstra, 1998; Hummels et. al., 2001; Ando,

2006; Jones et. al., 2005; Nordas, 2007). However, some development economists consider

deverticalization of production structures in developing countries as rather undesirable by product

of trade liberalization. Besides, market-oriented reforms reduce domestic content of production

by forcing small to medium size domestic subcontractors of large exporting companies to exit the

market (Cimoli and Katz, 2003). These policies may actually lead developing countries to

specialize on commodities or production processes that they have static comparative advantages

but require transferring R&D and engineering activities to headquarters of multinational

companies in developed economies.

Transformation of trade pattern in developing countries is rather rapid compared to

developed countries. Reallocation of multinationals’ production processes towards emerging

countries and development of global supply chains stimulate structural changes in emerging

markets. Yet, developing countries are also trying to cope with difficulties of domestic and

international turbulences while integrating their economy to the world markets via trade reforms.

During the transformation process it is crucial to design prudential trade reforms as it determines

long-term trade and growth performance of a country. Policy makers should have clear idea about

growth potential of their sectors in order to guide domestic and foreign investments efficiently.

Cimoli and Katz (2003) argues that Latin American (LA) experience is an evidence against the

internationally integrated production system that pushes these countries into a low development


trap by integrating them to the world production process via allocating low domestic knowledge

generation and value-added stages or production of traditional commodities to these economies.

Balaguer and Cantavella-Jorda (2004) and Montobbio and Rampa (2005), meanwhile, suggest

that countries investing in non-traditional manufactured and semi-manufactured commodities,

which are considered as high technology intensive sectors, could boost their economic growth

and exports. In this respect, analyzing the Turkish experiment, which depicts an increasing trend

in exports in the same internationally integrated production system, will allow us to understand

potential changes in a typical emerging market economy’s export behavior.

There are numerous empirical studies in the literature documenting the changes in the

nature and structure of world trade. A study by Stern et. al. (1979) reports evidence of structural

change in US imports but not in US exports by using data for the 1956-1976 period. In a similar

fashion Hall et. al. (1996) report shifts in UK export demand function by using sequential testing

procedure. Ben-David and Papell (1997) show that export to GDP and import to GDP ratios

increased substantially exhibiting a structural break during the postwar period in 48 countries.

Balaguer and Cantavella-Jorda (2004) state that changes in the composition of exports in favor of

non-traditional manufactured and semi-manufactured commodities explain the Spain’s higher

export growth after the implementation of outward-looking trade regime in 1959. Montobbio and

Rampa (2005) also link the export performance of nine developing countries to their

technological activities. They argue that a country, which expands its economic activity in high

technology sectors with increasing technological opportunities and avoid expansion of medium

technology sectors with low opportunity for technical change, can achieve higher average export


Empirical studies on the structural changes in trade pattern of developing countries is

rather limited and only few studies in the literature actually analyze the changes in the

coefficients of trade functions and their implications for the policymakers. Yet, effectiveness of

public policies to promote international competitiveness depends crucially on these trade


elasticities. In a highly integrated world economy, traditional export promoting strategies, such as

high protective tariffs and exchange rate depreciations, may actually backfire if the process of

export growth is not well understood. Acceleration of global competition and trade liberalization

increased division of labor among nations resulting in import dependence of national production

via vertical specialization of production processes. Deverticalization of production processes and

increased import dependence may decrease exchange rate sensitivity of trade flows and weaken

domestic competition policies based on exchange rate depreciation.

Main contributions of this study to the literature are twofold. First, it analyzes how

sensitivities of exports to competitiveness indicators, income and imports change as the structure

of trade alters over time. Majority of studies in the literature consider the structural change as

one time jump in export or import functions due to external shocks. This study, estimate

coefficients of export function as time varying parameters without imposing any predetermined

breaking point in time by applying Kalman filter approach. Second, having shown that

parameter values are instable over time the impact of the changes in composition of exports in

favor of non-traditional commodities is examined as a potential source of structural change in

total exports. A panel co-integration analysis is conducted across the group of non-traditional and

traditional commodities to examine group-wise differences in export function parameters.

Application of the panel co-integration techniques gives more reliable long-run coefficient

estimates than the single equation methods, especially when time period is not long enough.

Lastly, the study estimates export supply and demand functions separately that was also largely

ignored in this literature.

The rest of the paper is organized as follows. Section two briefly discusses the recent

trends in Turkish exports. Section three, after explaining the model and econometric methods

used in the analysis presents empirical results while section four concludes.


2. Trends and Structure of the Turkish Exports

Turkey changed its trade policy from import substitution to export-led growth in 1980s,

and since then, the share of Turkish exports in the world trade had increased from 0.36 percent in

1980-1990 to 0.44 percent in 1991-2000 and then to 0.60 percent in 2001-2005 (Aydın, et al.,

2007). Performance of the Turkish exports after 1980 can be classified in four sub-periods

(Figure 1.a). In the first period, early 1980s, a rapid fall in prices and a gradual rise in real exports

were observed due to the export promoting policies based on depreciation of Turkish lira (TL)

and export subsidies. The second period covers 1987-1996 and represents the period of gradual

capital account liberalization of Turkey, in which TL appreciated in real terms, as a result of an

increase in capital inflows. Appreciation of TL, in turn, slowed down the growth rate of exports

during this period. Very extensive and important Turkey-EU customs union (CU) in 1996, so

called “era of crisis” (1994-2001) put pressure on and initiated the IMF sponsored reform

programs in Turkey. Sharply falling export prices were affected mostly from the 1996 CU, 1997

Asian and 1998 Russian crises and resulted in disappearance of positive relationship between

export prices and real effective exchange rates in the third period (1997-2001). As a response to

the 2001 crisis, Turkish policymakers initiated another extensive reform program under the

supervision of IMF aiming to reduce public deficit, reforming the banking sector, implementing

floating exchange rate regime and decreasing inflation rate to single digits in the fourth period of

post-2001. As the Turkish economy struggle to cope with the post-2001 crisis era, average yearly

growth rate of Turkish exports reached to 20.9 percent, which is registered as the highest in the

recent Turkish history. High export performance was mainly attributed to the success of domestic

firms in adopting non-price competitive strategies. Effect of the CU on exports was evident with

lags after the impact of Asian and Russian crises disappeared. Under these conditions, it would be

rather unrealistic to expect stable export function coefficients. In fact, the swift changes in


relative importance of determinants of exports are reported in various studies (Şahinbeyoğlu and
Ulaşan, 1998; Saygılı et. al., 1998; Aydın et. al., 2004; Sarıkaya, 2004).

Figure 1: Turkish Exports, Imports, Prices and REER (2003=100)
a b







1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006





Export unit value index (right axis)
Export quantity index










1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007

intrm. & capital goods imports/exports (right axis)
pm/px(right axis)


As real exports increase, a parallel sharp rise has been observed in imports since 2001 as

well. According to Aydın, et al. (2007) and Yükseler and Türkan (2006) Turkish exports

performance becomes highly dependent on imports during this period, as it is also the case in the

new EU members. Sönmez (2005) reaches to the same conclusion by examining the

implementation of inward processing regime in Turkey. Figure 1.b shows that intermediate and

capital goods imports per unit of exports had an increasing trend even before the 1996 CU. CU

with EU in 1996 seems to decrease imports of intermediate and capital goods per unit of exports

until the financial and currency crises in 2001, probably due to real appreciation of TL during the

1997-1999 crises period. With the reform and stabilization policies the ratio starts to gain an

increasing trend again. This is the period analyzed by Aydın, et al. (2007), Yükseler and Türkan

(2006) and Sönmez (2005). It is worthy noting that though imports to exports ratio is greater than

the 1980s figure, it has not reached its pre crises level yet. During this period, even though there

was a slight increase in ratio of import prices to export prices, the change was not that significant

to affect imports of intermediate and capital goods per unit of exports. Albeit, it appears that there


is a positive relationship between REER and ratio of imports to exports.

Country composition of Turkish exports also showed interesting pattern. Europe has been

the major trade partner of Turkey since 1960s. Turkey and EU (formerly known as European

Community) signed a trade treaty in 1963 aiming to gradually integrate Turkish economy to the

union. As part of the agreement, member states eliminated tariffs on imports from Turkey in 1973

on agreed group of commodities but a long transition period was granted to Turkey’s complience

to the agreement. In the early 1980s Turkey begun its trade liberalization era by eliminating

restrictions on its foreign trade. There is a gradual increase in the share of European economies in

the Turkey’s trade volume during this period. When CU came into force in 1996, tariffs on

bilateral trade was alrady low and thus CU seem to have no significant increasing effect on

Turkey and EU trade volume(Figure 2). Another interesting trend is the increasing development

in trade share of Asian economies after 2001 that was falling before that year. Detailed analysis

shows that, while imports from EU27 and North America have been falling since the CU, the

deficiency is replaced by imports from European emerging markets and Asia (other than Near and

Middle East). On the other hand, growth rate of Turkish exports to Near and Middle East as well

as European emerging markets accelerated apparently after 2001, while there was no significant

change in exports to EU27.

Figure 2: Regional Export and Import Shares

Exports Imports






1980 1983 1986 1989 1992 1995 1998 2001 2004 2007

Europe Africa America Asia






1980 1983 1986 1989 1992 1995 1998 2001 2004 2007

Europe Africa America Asia



If export as well as import shares of Europe are not changing significantly, then it might

be the composition of trade that is influenced from CU and the recent global developments. Table

1 reports the list of 10 commodities that have the highest share in Turkish exports from 1982 to

2006. The content of the list changes throughout this period. Mid 1990s marks a swift change.

Among the commodity groups that constituted the list of 2006, only few were in the same list of

1980s and the first half of 1990s. Top-10 list for 1982, 1985, and 1990 included only 3, 4, and 5

commodities from the top-10 of 2006, respectively. However, 1995 and 2000 lists included 8 and

9 commodities from the 2006 top-10 list, respectively. Those commodities that were in the top of

the list before the financial turmoil period (1994-2001) were mostly replaced by the new ones

afterwards. Thus, it might be concluded that the crisis and structural adjustment period worked as

a selection process among the export sectors of Turkey. There are only three commodity groups

recorded in the top 10 list of all selected years: (08) Edible fruits and nuts, (61) Articles of

apparel and clothing accessories knitted, (73) Articles of Iron and steel, of which are relatively

labor intensive and Turkey traditionally has static comparative advantage. After the crises and

structural adjustments process Turkey starts to export less of relatively labor intensive traditional

export commodities such as (07) Edible vegetables, (24) Tobacco and manufactured tobacco, (25)

Salt, sulphur, earths and stone plastering materials, (27) Mineral fuels and oils and production of

their distillation. Instead, exports of relatively capital intensive commodities such as (63) Other

make up textile articles, (84) Nuclear Reactors, boilers, machinery and mechanical textile articles,

(87) Vehicles other than railway or tramway rolling stock are increasing in the same period. Also

note that increasing trend in share of exports of (72) Iron and steel and (85) Electrical Machinery

and equipment which are considered as capital intensive commodities started before the



Table 1: Top 10 Export Items of Turkey (1982-2006)

1982 1985 1990 1995 2000 2006
1 55 73 61 61 61 87
2 08 55 72 62 62 61
3 73 08 62 72 85 84
4 25 61 08 08 72 85
5 01 42 42 85 87 72
6 24 84 52 84 84 62
7 27 27 85 87 08 73
8 07 24 24 55 63 39
9 61 60 07 63 52 08
10 58 07 25 20 73 63
01: Live animals.
07: Edible vegetables.
08: Edible fruits and nuts.
20: Prep. of vegetables, fruits, nuts and other parts of
24: Tobacco and manufactured tobacco substitutes.
25: Salt, sulphur, earths and stone plastering materials.
27: Mineral fuels and oils and production of their
39: Plastic and articles thereof.
42: Articles of leather.
52: Cotton, cotton yarn and cotton fabrics.
55: Man-made staple fibers.

58: Special woven fabrics.
60: Knitted or crocheted fabrics.
61: Articles of apparel and clothing accessories knitted.
62: Articles of apparel and clothing accessories not
63: Other make up textile articles.
72: Iron and steel.
73: Articles of Iron and steel.
84: Nuclear Reactors, boilers, machinery and mechanical

85: Electrical Machinery and equipment.
87: Vehicles other than railway or tramway rolling stock.

Note: Table is taken from Aydın, et. al. (2007). Items that were also in the top 10 of 2006 were written in bold.

Figure 3: Concentration Measures (1982-2006)

Weighted Spread of Exports by Commodity Groups Shares of Top 10 and 20 Sectors in Total Exports







1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006














1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006













Top20 (left scale)

Source: Aydın, et. al. (2007).

Analysis of concentration ratio, further, points out the self-selection process in favor of

non-traditional sectors in Turkey. Figure 3 reports the evolution of two different export


concentration ratio indicators: Weighted spread1 of exports and shares of top 10 and 20

commodities in total exports. Note that the first indicator gets values between zero and plus

infinity and increase in its value indicates deterioration of distribution of exports among

commodity groups. According to the first indicator, except for the 1993-2001 period, exports

tended to concentrate on particular commodity groups. The pre-1994 phase can be explained by

the influence of export led growth strategy implemented at the beginning of 1980s. Yet,

concentration phase halted at the beginning of the economic and financial turmoil periods

(1993-2001), and at the end of this turmoil period, it climbed up again, as the new commodities

becoming the driving force of the export growth. Figure 3 also shows that the share of top 10

commodities, which remained mostly the same in 1980s and 1990s around 58 percent, begun to

increase in 2001, and reached to 69 percent in 2006. The share of top 20 commodities in total

exports, which fell during 1980s and most of 1990s, has shown an increasing trend since 1998.

Further evidence for the change in the composition of Turkish exports can be

documented by analyzing the best end worst performing commodity exports. Following Pineres

and Ferrantino (1997), commodity groups are separated into two different categories: traditional

and non-traditional. Traditional commodity is the one in which its export experience is

concentrated at the earlier years of the period. Yet, export experience function of the

non-traditional commodity is concentrated at the later years of the period. In other words,

traditional commodities experience faster but non-traditional commodities slower growth at the

earlier stage of the period. After ranking total of 96 commodities according to their traditionality

index, top 20 and low 20 goods were pooled into two groups.2

1 Weighted Spread of Exports is the sum of squared deviation from the overall mean of commodity-based

exports across goods corrected for mean:
 2 / 1


x Nti ti


  . tix , t , and N are exports of goods i in time t,

average commodity exports in time t, and total number of commodities, respectively.

2These commodities roughly cover the first and last 20% of the commodities of the total group. Though a 20% cut-off
criterion for the groups is rather arbitrary, the data and econometric method impose limitations on the cut-off
percentages. Simulation exercises on the panel cointegration method conclude that a minimum of 10 or more


Table 2 shows the share of each group of commodities in total exports, the list of the

commodity groups can be found in appendix B. Majority of non-traditional commodities consist

of capital and high tech intensive commodities such as vehicles other than railway or tramway

rolling stocks, furniture, electrical machinery and equipment, ships, boats and floating structures

etc. Share of non-traditional commodity exports increases from 14 percents at the end of 1980 to

41 percents in 2000s. Share of traditional commodity exports which mainly consist of agricultural

and labor intensive commodities such as live animal, fertilizers, leather, tobacco, edible

vegetables etc., decreases from 27 percents to 5 percents in the same period. In this respect, the

Turkish experience is rather different from the LA case. As opposed to the LA example the share

of traditional processing sectors lost their share after the trade liberalization of 1980s and 1990s.

Yet, sectors such as transportation vehicles and consumer electronics thrived in both LA and

Turkey. Though they are not considered as traditional sectors by general classification for both of

them, these sectors increased the economies’ import dependence of exported commodities.

Table 2: Share of Traditional and Non-traditional Commodity Exports

Traditional Non-traditional
1987-91 26.71 14.01
1992-96 14.38 18.20
1997-01 9.77 29.43
2002-07 5.05 40.60

Source: TURKSTAT and our calculations.

The preceding analysis demonstrates structural change in the Turkish exports in which

shares of capital and high-tech intensive commodities in total exports increases. Aydın et al.

(2007) state that European emerging markets are also in a similar transformation process. Yet,

increase in the share of capital and high-tech industries does not necessarily mean that Turkey is

cross-sections in the panel improve the panel results considerably. On the other hand, too large cut-off shares may add
additional noise to our estimates. Commodities which had less than 15 million USD export value in all of the 20 years
of observations were excluded from the analysis. Moreover, the export growth rates of the commodities between the
best and worst performing ones are very close to the overall average export growth. Thus, their inclusion in the analysis
will blur our efforts to identify common properties of fast growing and slowly growing export commodities.


specializing on the high-tech process of the production. It may simply indicate increase in the

dependence of domestic production on imports of high value added inputs. Across the

manufacturing sectors, the rate is relatively higher in electrical machinery and apparatus followed

by motor vehicles and textile and wearing apparels. The first two sectors are considered as

non-traditional sectors that exhibited rapid growth path after the crises period (1994-2001). Thus,

transformation mainly takes place in manufacturing industries that have import dependency rate

higher than that of the overall economy (Aydın, et al. 2007).

What derives this sectoral selection and sorting process? Trade liberalization and

market deregulations may initiate such a process. As it is noted in Cimoli and Katz (2003) firms

and sectors differ not only because they produce different commodities but also because they

respond differently to the changes in macro-micro settings and regulations. After the

macro-economic stabilization and trade liberalization programs of LA economies during the

1970s and 1980s, local firms faced contraction of domestic markets on the one hand and immense

arrival of imports on the other. At the end, some incumbent firms were able to adjust to the new

environment by lowering production costs-expelling labor- but many firms -mainly small and

medium sized enterprises- had to exit the market. Thus, ECLAC (2000 and 2002), and Cimoli

and Katz (2003) conclude that export–oriented sectors and firms closer to the static comparative

advantages of the country react swiftly and positively to the new micro-macro environment.

These sectors include unskilled labor intensive and natural resource-processing industries in the

case of LA economies. Contrary to the LA case, the share of highly unskilled labor intensive

agricultural sectors decreased as a response to macro-micro regulations. Knowledge-intensive

sectors producing for the domestic market under the protection of tariffs react negatively to the

new economic environment. Among the sectors that forge ahead during the trade liberalization

and deregulation era of the LA economies, non-tradable service sectors, ‘in bond’ assembly

industries producing electronic equipment, televisions, etc and the vehicle industry, showed

strong success in increasing its share in national income and exports. A similar pattern is also


observed in Turkey during the transformation process.

According to Cimoli and Katz (2003) macro-economic stabilization and trade

liberalization programs increased business concentration ratio as well. As the small and medium

sized companies forced to exit the market, large firms that have accumulated technological

capability and access to long-term financing and technology markets survived. Moreover, these

large firms, in order to survive in the new environment, adapted a new production process. The

new regime turned the local production into assembly activity that uses imported parts and

components as well as foreign technology and engineering services. Thus, the industries and

firms that have been successful in recent decades are the ones that deverticalize their production

structure and specialize on particular stage of the production process.

3. The Model and Main Results

A standard export demand equation can be specified as a function of competitiveness

indicators and foreign income. Ala Rao and Singh (2007) a typical export function of a county

can be specified as follows:


d Y

PX lnlnln 210  


 (1)

where X , fY , dP , fP and E are exports, foreign income, domestic prices, foreign prices and

exchange rate, respectively. The term f

 represents the competitiveness indicator, real

exchange rate. When more than two country is competing in a single market the real exchange

rate formulation above may not be sufficient to measure how competitive is the price of a single

country. Then a trade weighted competitiveness indicator such as REER would be the better

indictor to be used in estimating export demand function. CPI based REER assess price

competitiveness of a country relative to its principal competitors in international markets.


Standard export supply equation, on the other hand, can be specified as a function of

competitiveness indicators such as relative prices, unit labor costs, effective exchange rate and

scale variables such as domestic output and output gap, as well as some form of import constraint

variables such as total imports, imported raw materials and capital goods. Muscatelli et. al. (1995)

defines export supply as a function of export prices, variable costs (includes imported raw

material and wage costs) and stock of fixed capital. The prices and costs are in common currency

terms so they also include changes in exchange rates. However there are two issues need to be

solved related to the equation. Firstly, if firms are mark-up pricing then export prices and variable

costs would be collinear and by including only the variable cost one may consistently estimate the

export supply function. Besides, Muscatelli et. al. (1995) did not estimate export supply function

directly but use the equation to estimate export supply prices in order to estimate integrated

supply and demand equations. In our study, we used unit labor cost index (ULC) since it is the

most reliable measure of costs in Turkey. Domestic currency dominated unit labor costs are

converted into foreign currency terms by computing ULC based real effective exchange rate.3

Secondly, Muscatelli et. al. (1995) include capital stock to estimate the effect of product quality

on export supply. However, there is no official capital stock statistics in Turkey and its alternative

measures are usually considered as rather problematic. Since the focus of our paper is not on the

effect of product quality and there are significant data problems we did not include capital stock

in our analysis. Yet, the export supply specified by Muscatelli et. al. (1995) excludes a very

significant variable that we believe derives the recent trend in international trade. Increase in

import dependence of production in developing countries necessitates the use of imported

intermediate goods for exported commodities. Since the current deverticalized mode of

production led developing countries to specialize on a particular process of production, they may

have neither means nor technology to produce the other parts and components. Thus, imports of

3 Our analysis suggests that among the price competitiveness indictors the use of REER_ulc in the export supply
function reveals empirically better results.


parts and components impose constraints on domestic production and exports.

One may also include domestic production capacity as a variable determining export supply

of a country. Though, the actual and potential outputs are alternative measures of physical

productive capacity we found poor evidence for their inclusion in the long-run supply function.

Also, we avoid using import and output variables in the same equation due to the endogeneity


Following the literature, supply and demand functions for the overall exports in Turkey are

analyzed for the period of 1987q1-2008q1. Exports and imports USD values and respective unit

price values, unit labor costs and domestic output are taken from the Central Bank of the

Republic of Turkey (CBRT) website while both unit labor costs based (REER_ulc) and consumer

price index based real effective exchange rates (REER_cpi) are taken from Eurostat.4 Lastly,

foreign income in current prices and PPP for the OECD countries is taken from the OECD


Application of standard augmented Dickey-Fuller unit root tests suggests that all variables are

integrated of order one, I(1). Table A in appendix reports the results for the variables chosen for

the rest of the analysis. Then, following the literature, co-integration tests over the different

vector of non-stationary variables are applied to find out well-defined long-run supply and

demand equations. In addition, USD value of the sum of capital goods and raw material imports,

which are used as a measure of import dependency of exports, are found to be significant factors

in determining the long-run supply function.5 For the export demand equation, the standard set up

is preserved and in addition to REER_cpi, foreign income is included, specifically OECD income,

as a demand factor in estimations.

4 Currencies of 34 countries are used to calculate real effective exchange rate indexes, Belgium, Germany, Greece,
Spain, France, Ireland, Italia, Luxembourg, Netherlands, Austria, Portugal, Finland, Denmark, Sweden, United
Kingdom, Czech Republic, Estonia, Cyprus, Latvia, Lithuania, Hungary, Malta, Poland, Slovenia, Slovakia, Australia,
Canada, United States, Japan, Norway, New Zealand, Mexico, Switzerland, and Turkey.
5 Using total imports in the export supply function did not make any significant change in our results. The export
supply function is estimated by using quantity indices, as well. The results were quite similar. However, since quantity
indexes are not available at commodity level, which we need in the panel co-integration analysis in section 5, we use


Accordingly, the following model is specified to analyze export supply and demand functions,

respectively for Turkey.

0 1 2 3 1 4 2 5 3 2_t t t txs m reer ulc s s s u            (2)
0 1 2 3 1 4 2 5 3 1_

t t t txd y reer cpi s s s u            (3)

Here, txs and txd are the log of exports;

ty is the log of foreign income; tm is the log of

imports, _ treer cpi and _ treer ulc are the log of REER_cpi and REER_ulc, respectively, ukt

 iidN (0,1) k=1,2 and js (j=1,2,3) are the seasonal dummies. Since quarterly data is used in the
analysis seasonal dummies are added exogenously to deal with seasonality.

3.1 Kalman Filter Approach and Results

Kalman filter approach or state space models, which are developed by Kalman (1960, 1963),

have been used extensively in economics. The Kalman filter is a recursive algorithm for

expressing dynamic systems that involve unobserved state variables tz , conditional on observed

vector t
y (Kim and Nelson, 2000). A state space model consists of two equations of which a

general form of a linear state space system representation is written down bellow:

1=t t t t tz z G w   (4)
=t t t ty H z  (5)


tz R is an ( 1)n state vector, mty R is the vector of observation, t , tH , tG

are known matrices that are also allowed to vary with time, kw , and t are vectors of normally
distributed i.i.d shocks. Equation (4) is called a transition equation that describes the dynamics

of the state variables. Equation (5) is the measurement equation points out a relationship

USD values in all analyses in order to have comparable coefficient estimates.


between observed variables and unobserved state variables. The model satisfies the following


  0tE   ,   0tE w  (6)
t j t tjE R      , t j t tjE w w Q     (7)

0t jE w     ,  0 0E z z (8)
 0 0 0 0 0( )( )E z z z z P   (9)
 0 0tE z w  ,  0 0tE z    (10)

Under these assumptions, ˆtz can be determined by the Kalman filter:

 1 1ˆ ˆ ˆt t t tt t t tz z K y H z    (11)
0ˆ(0)z z

Here, tK is the Kalman gain, which determines the weight assigned to new information

about tz and calculated by

  11 1t t t t tt t t tK P H H P H R     (12)
where tP is the ( )n n covariance matrix of tz conditional on information up to ( 1)t  and
calculated as follows

1 1 1 1 1 11 t t t t t tt tP P G Q G           (13)
  1t t t t tP I K H P   (14)

1 11ˆ ˆt tt tz z    (15)
As it is clear from equation (14) the success of the estimation depends on the representation of the


dynamics of the system. If the best Kalman gain is used then

1ˆt t t t tg y H z   (16)

The residual vector tg satisfies all white noise properties and its covariance matrix can be

calculated as

 0, 1t t t t t tt tC E g g H P H R    . (17)
Kalman filter approach explained above can be modified to our export demand and supply

models. Since this paper is interested in analyzing how model parameters change over time, it is

assumed that all parameters of the equation 2 and 3 follow a random walk process. Then the

transition equations for the supply and demand functions are:

1 2it it t     , i=0,1...5 (18)
1 1it it t     , i=0,1…5 (19)

where t is normal white noise processes. Then, the measurement equation can be written as
1* 't t t txs F B u  , (20)

2* 't t t txd H A u  , (21)

where,  1 2 31 _t t t t t tF m reer ulc s s s , 1 2 31 _ft t t t t tH y reer cpi s s s    ,
 0 1 2 3 4 5t t t t t t tB       ,  0 1 2 3 4 5t t t t t t tA       , u1t  iidN(0,1)

and u2t  iidN(0,1). Kalman Filter approach is a recursive process that updates estimated
coefficients over time as new information arrives. In estimations fixed-point Kalman smoother,

which gives the estimated value of the state variable at time t based on all the available

information up to time T, where T t is applied. The idea is that as new data are made available,
we can improve our estimation results from the Kalman filter by taking into account the

additional information.


Estimated coefficients for the export function are presented in Figure 4. 1994, 1997-1998,

and 2001 crises and 1996 CU are marked on these graphs as shaded areas. Smoothed Kalman

filter estimates show that import elasticity of export supply fluctuates around 0.74 during the

1987-1996 period. It rises steadily after 1996. The trend is disrupted three times in 1994, 1999

and towards the end of 2000. However, these shocks are not sufficient to change the path of the

import elasticity. It seems that the CU agreement in 1996 and economic reforms of post-2001

have significant positive impact on Turkish exports, such that the elasticity gains an increasing

trend. Indeed, import elasticity, which is about 0.73 at the beginning of the period, reaches 0.81 at

the end of the period, showing about 11 percent increase. Meanwhile, the responsiveness of the

export supply to the changes in REER_ulc steadily decreases from about -0.24 in 1987 to around

-0.18 in 2007. That counts roughly 22.8 percent decrease in elasticity. As in the case of import

elasticity, domestic economy originated shocks in these years do not have long lasting effects on

the falling REER_ulc elasticity of exports.

Increase in import and decrease in ULC based exchange rate elasticity of exports can be

due to increased import dependence of Turkish exports. As industries use greater share of

imported inputs in production, their imported inputs possess greater influence on export

performance. Moreover, profits, and so export supply of firms would be less sensitive to

exchange rate movements since not only revenues but greater share of total costs will be in the

form of foreign currencies. Thus, when import dependence increase, fluctuations in exchange

rates have less effect on profits and so export supply decisions of exporting firms. Increase in

import dependence may emerge as a result of economy wide increase in use of imported inputs or

increase in the share of highly imported dependent industries in the total exports or a combination

of both. We will address this issue in the next section.


Figure 4: Time Varying Export Supply and Demand Coefficients
Export Supply Equation Export Demand Equation

a) Import Elasticity a) Income Elasticity








































1 1.45













































b) REER_ULC Elasticity b) REER_CPI Elasticity

















































































* Sum of squared residuals for the supply and demand functions are 0.1066 and 0.0929 respectively.

The column on the right side of Figure 4 reports estimated parameters of the export

demand equation. Income elasticity of exports, which was decreasing during the 1987-1994

period and roughly constant during the 1994-2000 period, gains an increasing trend after the

financial crises in 2001. It appears that 2001 financial crises have a path breaking impact on the

income elasticity of export demand. Thereby, towards the end of the period, the income elasticity

shows almost 2.8 percent increase and goes up from about 1.56 in 1987 to 1.61 in

2006.REER_cpi elasticity demonstrates a decreasing path during the 1987-2008q1 period, too. It

appears that the sensitivity of export demand to REER_cpi decreases from about –0.30 in 1987 to


around –0.16 at the end of the period, demonstrating about 47 percent fall. The speed of decrease

increases after the 2001 crises.

The coefficient changes in the export demand function can be due to various factors.

Increase in income elasticity of exports may indicate change in the composition of Turkish

exports towards commodities with high-income sensitivity. Decrease in export demand exchange

rate elasticity, on the other hand, may indicate decrease in substitution between average consumer

good in Turkey’s main export destinations (mostly developed countries) and Turkish

commodities. Turkish producers may failed to lead their production towards commodities that

compete with domestic producers of the developed economies. Instead, they are mainly

competing with exports of other developing countries in the world markets. This is known as

“fallacy of composition”. Similarly Razmi and Blecker (2008) by using time series and panel

data for 21 developing countries argued that low-technology intensive products exporting

developing countries, including Turkey, compete with other developing countries rather

than industrialized countries.

The analysis in the following section performs commodity based export supply

and demand analysis to examine how changes in commodity composition of exports may

affect model parameters

3.2 Commodity Based Panel Co-integration Analysis

The behavior of the best-performed commodity groups and low-performed commodity

groups are mentioned above in section 2. Commodities were grouped following Pineres and

Ferrantino (1997), in traditional and non-traditional categories. Traditional commodity is the one

in which its export experience is concentrated at the earlier years of the period. Yet, export

experience function of the non-traditional commodity is concentrated at the later years of the

period. Here in this section a panel cointegration technique is applied separately over these two


groups of commodities to analyze first if long-run demand and supply functions differs across

these groups and second how the aggregate demand and supply functions changed with respect to

the changes in the composition of exports in favor of non-traditional commodity groups.

Multivariate panel co-integration technique developed by Pedroni (1999) and Pedroni

(2001) is employed to test long run properties of the commodity based export demand and supply

functions. Panel co-integration technique is a powerful method to investigate existence of

co-integration, since it combines both time series and cross sectional information. Pedroni uses

the following standard panel regression to develop test statistics for panel cointegration:

it i i i it ity t x u      i=1,...,N; t=1,...,T. (22)

where yit and xit are panels of observations over the members of the panel and assumed to be

integrated of order one (I(1)) for each panel member i. Under the null of no cointegration,

residual uit is assumed to be I(1). Parameters αi and δi capture any fixed effects and deterministic
trends that are specific to each member of the panel, respectively and βi is a vector of parameters
that are allowed to vary across members of the panel.

Pedroni (2000) and Pedroni (2001) propose use of FMOLS methods for estimating and

testing hypothesis for cointegrating vectors in dynamic time series panels. The method modifies

least squares to account for serial correlation and the endogeneity in the regressors that results

from the existence of a cointegrating relationship. Based on equation (22), Pedroni suggests two
sets of statistics that uses fully modified OLS (FMOLS) for testing the null hypothesis H0: “All

of the individuals of the panel are not cointegrated, uit  I(1)” against the alternative H1: “A
significant portion of the individuals are cointegrated, uit  I(0)”. Thus, under the alternative
hypothesis Pedroni permits individual members of the panel to differ whether they are

cointegrated or not. Use of FMOLS principles not only accommodates considerable heterogeneity

across individual members of the panel, but also produces asymptotically unbiased estimators.

Kao and Chiang (2000) suggest the use of panel within-dimension DOLS estimator

based on including leads and lags of the first differences of the regressors in the estimated

equations. Pedroni (2001) demonstrates that FMOLS and OLS estimators have minor size

distortions. Between-dimension FMOLS estimator has advantage over the within-dimension

DOLS estimator by setting the null hypothesis in a way to allow parameters to vary across the


panel members. In other words, between-dimension estimators allow to test H0: βi = β0 versus H1:
βi ≠ β0 for all i. In addition, when the true co-integrating vectors are heterogeneous,
between-dimension estimator provides mean value of the co-integrating vectors that may reveal

some information on the behavior of sample group. Due to these advantages, and also to the space

limitation, only between-dimension estimators are used in this paper.

Pedroni suggests two sets of statistics for the panel co-integration tests. The first set

consists of three panel statistics; ‘panel variance ratio statistics’, ‘panel rho statistics’ and ‘panel

t-statistics’ that are based on pooling the residual of the regression along the within dimension of

the panel. The second set of statistics consists of two statistics; ‘group rho statistics’ and ‘group

t-statistics’, which are based on pooling the data along with between dimensions of the panel. As

noted in Pedroni (2004), the first set of statistics is constructed by summing the numerator and

denominator terms separately for the analogous time series statistics. The second set of statistics,

as opposed to the first set, is constructed by first calculating the ratio corresponding to the time

series statistics and then computing the standardized sum of the ratio over the cross section of the

panel. In fact, the second set of statistics is the group mean of the respective individual time series

statistics. Given the general form of the equation 1 and 2 and the assumptions on the parameters,

the null hypothesis is set as ‘all of the individuals of the panel are not co-integrated’ and the

alternative hypothesis is set as ‘significant portion of the individuals of the panel are


Panel cointegration analysis may suffer from cross-sectional dependency that must be

treated cautiously. Demeaning the series by adding a time dummy helps to control the common

aggregate shocks that are likely to generate cross sectional dependency. However, here in our

analysis REER as well as OECD income are the same for each individual commodity, implying

that common time effect and both REER and OECD income have only t dimension, which makes

them perfectly collinear. Thus, it is redundant to include both common time effect and REER as

well as OECD income.


Before proceeding to panel co-integration analysis panel unit root tests of the Im et. al.

(2003), which is based on the Dickey-Fuller t-statistics and Maddala and Wu (1997)’s Fisher-test,

which is based on p-values of individual unit root test are conducted. Results suggest that all of

the panel variables are I(1). Next, as a standard procedure, a multivariate panel co-integration

analysis is conducted to examine existence of a co-integrating vector for supply and demand

equation of each group. Results are presented with and without heterogeneous trends, in Table 3.

The first four columns of the table report the panel statistics and the next three columns display

the group statistics. The parametric ADF version of these statistics is added next to the each set of

statistics for comparison purpose. All these test statistics unanimously reject the null hypothesis

of no co-integration and suggest that significant portion of the individuals of the panel are


Table 3: Panel Co-integration Test
Panel Group # of Rej.
Var.ratio-stat Rho-stat. T-stat ADF-stat Rho-stat T-stat ADF-stat %5

Non -traditional
Supply Function
with trend 4.977 -6.406 -10.798 -8.698 -7.029 -13.693 -10.283 7
without trend 4.201 -7.178 -14.988 -8.862 -8.362 -17.95 -8.445 7
Demand Function
with trend 0.929 -7.613 -12.121 -4.977 -7.66 -14.29 -4.926 6
without trend -1.379 -9.129 -16.606 -6.745 -8.343 -17.221 -5.604 6
Supply Function
with trend 2.325 -14.288 -13.244 -10.085 -13.576 -14.197 -9.168 7
without trend -1.009 -11.82 -14.264 -9.367 -10.301 -13.841 -8.337 6
Demand Function
with trend 0.907 -14.091 -13.548 -9.808 -12.26 -13.607 -8.505 6
without trend -1.845 -13.721 -15.362 -11.439 -10.626 -13.748 -9.354 6
Note: Under the alternative hypothesis the variance ratio statistics converges to +∞ while the other statistics converges
to -∞. Therefore, the right tail of the normal distribution is used to reject the null hypothesis for the variance ratio test;
where as the left tail of the normal distribution is used for the other statistics. Lags are selected by using Schwarz
Information Criteria.

After determining co-integration properties of each panel, individual and panel group

FMOLS results are reported in Table 4 and 5 below. Estimated coefficients of the group rather

than the coefficients of individual commodities are relevant for the purpose of the study. The

panel group FMOLS results are presented at the lower panel of each table. First of all, all


parameters are statistically significant for both commodity groups. Secondly, the average import

elasticity is higher for the group of non-traditional commodities (1.57) compared to that of the

group of traditional commodities (0.08). Besides, the group of non-traditional commodities has

average REER_ulc elasticity of -0.05, which is lower than that of the group of traditional

commodities (-1.12) in absolute terms. These are clear indications of high use of imported inputs

and low REER sensitivity of the non-traditional commodity exports. In turn, this may also

explain how producers of this commodity group might survive the turbulences and currency

crises of the 1994-2001 period. Thirdly, for the group of non-traditional commodities, the average

export income elasticity (4.09) is higher and average REER elasticity (-0.18) is lower than the

respective elasticities (0.52 and –1.78) for the group of traditional commodities. 6

Table 4: Individual and Panel Group FMOLS Results for Non-Traditional

Import Elasticity REER_ulc Income elasticity REER_cpi
Panel Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat

1 1.06 (34.04) 0.02 (0.14) 2.34 (16.65) 0.76 (2.42)
2 1.23 (13.82) -1.18 (-4.37) 2.95 (11.52) -0.3 (-0.52)
3 0.88 (9.01) 0.08 (0.24) 2.12 (7.39) 0.34 (0.59)
4 1.41 (11.34) 1.17 (2.64) 3.26 (12.16) 0.82 (1.38)
5 1.7 (7.18) 2.87 (3.40) 3.81 (6.45) 2.08 (1.58)
6 1.43 (22.87) -0.19 (-0.85) 3.44 (32.40) -0.03 (-0.13)
7 1.7 (6.76) 2.28 (2.63) 4.28 (7.32) 0.97 (0.73)
8 0.52 (0.95) 3.85 (1.66) 6.09 (4.02) -2.72 (-1.77)
9 1.23 (19.6) -1.2 (-5.39) 3.14 (14.99) -0.99 (-2.12)

10 1.8 (12.26) -0.32 (-0.62) 4.4 (14.06) -1.28 (-1.83)
11 1.8 (22.61) 0.01 (0.03) 4.34 (34.14) 0.33 (1.18)
12 2.3 (21.46) -0.96 (-2.51) 5.81 (28.66) -0.81 (-1.79)
13 1.61 (20.45) -1.4 (-4.99) 4.04 (15.17) -0.85 (-1.43)
14 1.44 (16.32) -0.37 (-1.18) 3.38 (14.88) 0.06 (0.12)
15 2 (7.12) 1.65 (1.60) 5.02 (7.43) 1.88 (1.34)
16 1.22 (20.53) -0.87 (-4.10) 2.95 (12.97) -0.03 (-0.06)
17 1.83 (12.71) -2.46 (-4.80) 4.71 (9.23) -1.59 (-1.39)
18 3.01 (14.22) -0.49 (-0.66) 7.64 (29.96) -1.6 (-2.81)
19 1.61 (27.73) -0.29 (-1.40) 3.89 (32.79) 0.23 (0.85)
20 1.63 (20.58) -1.16 (-4.12) 4.1 (17.69) -0.82 (-1.58)

Panel Group FMOLS Results
1.57 (71.91) -0.05 (-5.06) 4.09 (73.77) -0.18 (-1.17)

Note: t-stats for H0:

 

are in parenthesis. The list of commodities in the non-traditional is in Appendix B. Lags
are selected by using Schwarz Information Criteria.

6 Similarly, Gönenç and Yılmaz (2007) classify Turkish manufacturing industries based on their competitiveness as
highly competitive, intermediary and declining sectors. The more profitable industries produce, export and create
employment more than the others.


Table 5: Individual and Panel Group FMOLS Results for Traditional Commodities

Import Elasticity REER_ulc Income elasticity REER_cpi
Panel Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat

1 -0.13 (-2.13) -0.24 (-1.12) -0.21 (-1.40) -0.79 (-2.37)
2 -2.66 (-9.65) -0.39 (-0.47) -5.49 (-9.83) -2.69 (-2.11)
3 1.11 (2.09) -1.77 (-0.96) 3.59 (2.78) -5.6 (-2.13)
4 -0.25 (-4.66) 0.27 (1.38) -0.81 (-7.47) 0.89 (3.69)
5 1.26 (4.18) -5.58 (-5.21) 4.19 (4.28) -4.95 (-2.26)
6 0.06 (0.55) -1.45 (-3.76) 0.3 (0.98) -1.82 (-2.65)
7 0.00 (-0.01) -1.03 (-2.77) -0.39 (-1.24) -0.49 (-0.67)
8 0.17 (0.82) -2.38 (-2.73) 1.84 (2.82) -3.43 (-5.15)
9 -1.26 (-5.54) -1.71 (-2.12) -2.54 (-4.23) -2.78 (-2.07)

10 0.26 (2.91) 0.30 (0.92) 0.43 (1.77) 0.45 (0.82)
11 -1.31 (-3.78) 0.81 (0.66) -2.52 (-2.63) -3.6 (-1.68)
12 0.31 (5.76) -0.41 (-2.12) 0.91 (6.93) -0.71 (-2.42)
13 0.15 (2.34) 0.25 (1.14) 0.23 (1.41) 0.41 (1.14)
14 0.11 (1.20) -0.63 (-2.00) 0.54 (2.28) -0.85 (-1.61)
15 1.03 (7.54) -1.78 (-3.55) 2.53 (5.41) -2.16 (-2.22)
16 0.24 (1.67) 0.43 (0.85) 0.71 (1.94) -0.25 (-0.30)
17 0.24 (2.12) -0.01 (-0.03) 0.67 (2.37) -0.69 (-1.09)
18 1.30 (6.23) -4.35 (-5.87) 3.68 (5.08) -4.16 (-2.58)
19 0.68 (7.40) -2.56 (-7.77) 2.16 (6.41) -3.03 (-4.03)
20 0.31 (5.16) -0.10 (-0.45) 0.54 (3.08) 0.55 (1.41)

Panel Group FMOLS Results
0.08 (5.41) -1.12 (-8.04) 0.52 (4.64) -1.78 (6.33)

Note: t-stats for H0:

 

are in parenthesis. The list of commodities in the traditional is in Appendix B. Lags are
selected by using Schwarz Information Criteria.

4. Conclusion

This study examines the time and the sources of the structural changes in the Turkish

exports. Results suggest a continuous increase in imports as well as income elasticities, but

persistent decrease in the real effective exchange rate elasticity supporting the increased import

dependency argument. Structural change in the composition of exports seems to determine the

recent changes in export function in Turkey. Those sectors that have high income and import, but

low exchange rate elasticities of exports tend to increase their shares in total Turkish exports. It is

noted that composition thereby exports coefficients of Turkey changes rather rapidly during the

major reform periods (customs union with EU in 1996) and economic crisis (2001 financial crisis)

that usually takes long periods for a matured economy. This interesting finding is also

documented in De Pineres and Ferrantino (1998), who argued that export diversification and


structural change accelerated in six LA economies during the debt crisis of early 1980s and policy

reform periods. Their previous study on Chile also reached to similar conclusions (De Pineres and

Ferrantino, 1997).

Important implications can be driven from these finding, not only for the sources of the

structural change in total exports but also for the policy issues. Increase in the share of

non-traditional commodities in total exports raises not only the overall income elasticity of total

Turkish exports but also its import elasticity, which explains the recent surge in the import

dependence of overall exports. It is evident that exchange rate elasticity of non-traditional

commodities is smaller than that of the traditional goods, which also pulls down the overall

exchange rate elasticity of exports over time especially after 1996. While required intermediate

and capital goods for production of non-traditional goods are imported from abroad, sale of these

commodities abroad by using the same currency makes non-traditional commodity exports and

imports less responsive to the exchange rate fluctuations. The change in the composition of

Turkish exports in favor of low exchange rate elastic non-traditional commodities may explain

the seemingly puzzling coincidence of high growth of total exports and real appreciation of the


Global trends towards the vertical specialization of production may also contribute to the

recent structural changes in the Turkish exports. Nordas (2007) states that in some industries such

as automotive and electronics, technological advances allowed countries to specialize in

particular stage of a production process of a good. In this respect, economic reforms and financial

crises may actually speed up the process of integration to the world economy via vertical

specialization since only those sectors which are fit and quick enough to cope with the global

trend in international production sharing were able to survive the turbulence period. In a similar

vein, Cimoli and Katz (2003) argued that trade liberalization and market deregulations may

initiate the process. Export–oriented sectors and firms closer to the static comparative advantages

of a country may react swiftly and positively to the new micro-macro environment while other


local firms face difficulties due to contraction of domestic market and massive arrival of imports.

As large firms in some sectors survive in the new regime mostly small and medium sized firms

forced to exit. Large firms, in order to survive in the new environment, adapt a new production

process which turn local production into assembly activity that uses imported parts and

components as well as foreign technology and engineering services. This new global production

sharing activities may possess some dangers for the long-term growth prospects of a developing

country as well. If the new production process based on deverticalization of production induces

substitution of domestically produced capital goods by imports, then the process may lower

utilization of highly skilled labor and contraction of R&D activities. This, in turn, may increase

structural unemployment of skilled labor and decrease domestic technological capability of a



5. References

Ando, Mitsuyo (2006). ‘Fragmentation and vertical intra-industry trade in East-Asia’, North
American Journal of Economics and Finance, 17, pp. 257-281.

Aydın, F., Çıplak, U. and Yücel, E. (2004). ‘Export supply and import demand models for the
Turkish economy’, The Central Bank of Republic of Turkey Working Paper No. 04/09.

Aydın, F., Saygılı, M. and Saygılı, H. (2007). ‘Empirical analysis of structural change in Turkish
exports’, Central Bank of the Republic of Turkey, Working Paper no. 07/08.

Balaguer, J. and Cantavella-Jorda, M. (2004). ‘Structural change in exports and economic growth:
Co-integration and causality analysis for Spain (1961-2000)’, Applied Economics, 36, pp.

Ben-David, D. and Papell, D. H. (1997). ‘International Trade and Structural Change,’ Journal of
International Economics , 43 (3-4), pp.513-523.

Cimoli, M. and Katz J. (2003), ‘Structural reforms, technological gaps and economic
development: a Latin American perspective’, Industrial and Corporate Change, 12 (2), pp.

De Pineres S. A. G. and Ferrantino, M. (1997). ‘Export diversification and structural dynamics in
the growth process: The case of Chile’, Journal of Development Economics, 52(2), pp.

De Pineres S. A. G. and Ferrantino, M. (1998). ‘Export diversification trends: Some comparisons
for Latin America’, The International Executive, 39(4), pp. 465-477.

ECLAC (2000), Equity, development and citizenship, ECLAC-UN, Santiago.
ECLAC (2002), Globalization and development, ECLAC-UN, Santiago.
Feenstra, R. C. (1998). ‘Integration of Trade and Disintegration of Production in the Global

Economy’, Journal of Economic Perspectives. 12, pp. 31-50
Gönenç, R. and Yılmaz, G. (2007). ‘The evolution and determinants of profitability in Turkish

manufacturing industry, 1997-2006’, The Central Bank of the Republic of Turkey
Working Paper No:07/01.

Hall, S. G., Urga, G. and Whitley, J. (1996). ‘Structural Change and Economic Behavior: The
Case of UK Exports’, Economic Issues, 1(1), pp. 39-50.

Hummels, D., Ishii, J. and Yi, K-M (2001). ‘The nature and growth of vertical specialization in
world trade’, Journal of International Economics, 54, pp. 75-96.

Im, K. S., Peseran, M. H. and Shin, Y. (2003). ‘Testing for unit roots in heterogenous panels’,
Journal of Econometrics, 115, pp. 53-74.

Jones, R., Kierzkowski, H. and Lurong, C. (2005). ‘What does evidence tell us about
fragmentation and outsourcing?’, International Review of Economics and Finance, 14, pp.

Kalman, R.E. (1960). ‘A new approach to linear filtering and prediction problems’, Transactions
of the ASMA-Journal of Basic Engineering, 82, pp. 35-45.

Kalman, R.E. (1963). ‘New methods in Weiner filtering theory problems’, in Proceedings of the
First Symposium of Engineering Applications of Random Function Theory and
Probability (eds) J.L. Bogdanoff and F. Kozin, , New York, pp. 270-388.

Kao, C. and Chiang M.H. (2000), “On the estimation and inference of a cointegrated regression in
panel data.” Advances in Econometrics, 15, pp. 179-222.

Kim, C. J. and Nelson, C. R (2000). State space models with regime switching: Classical and
Gibbs sampling approaches with applications, MIT Press, Cambridge, MA.

Krugman, P. (1995). ‘Growing world trade: causes and consequences’, Brookings Papers on
Economic Activity, 1, pp. 327-362.

Maddala, G.S. and Wu, S. (1999). ‘A comparative study of unit root tests with panel data and a
new simple test’, Oxford Bulletin of Economics and Statistics, 61, pp. 631-652.

Montobbio F. and Rampa, F. (2005). ‘The Impact of Technology and Structural Change on


Export Performance in Nine Developing Countries’, World Development, 33(4), pp.

Muscatelli, V.A., Stevenson, A.A., and Montagna, C. (1995). ‘Modeling aggregate manufactured
exports for some Asian newly industrialized economies’, The Review of Economics and
Statistics, 77, pp. 147-155.

Nordas, H. K. (2007). ‘International production sharing: a case for a coherent policy framework’,
WTO Discussion Paper No:11.

Pedroni, P. (1999). ‘Critical values for co-integration test in heterogenous panels with multiple
regressors’, Oxford Bulletin of Economics and Statistics, 61, pp. 653-70.

Pedroni, P. (2001). ‘Purchasing power parity tests in co-integrated panels’, The Review of
Economics and Statistics, 83, pp. 727-31.

Pedroni, P. (2004). ‘Panel co-integration, asymptotic and finite sample properties of pooled time
series test, with an application to PPP hypothesis’, Econometric Theory, 20, pp. 576-625.

Rao, B.B. and Singh, R. (2007), “estimating export equations”, Applied Economic Letters, 14, pp.

Razmi, A. and Blecker, R.A. (2008), “Developing country exports of manufactures: Moving up
the ledder to escape the fallacy of composition?,” Journal of Development Studies, 44 (1),
pp. 21-48.

Sarıkaya, Ç. (2004). ‘Export dynamics in Turkey’, Central Bank Review, 2, 41-64.
Saygılı, M., Şahinbeyoğlu, G.; and Özbay, P.(1998). ‘Competitiveness indicators and the

equilibrium real exchange rates dynamics in Turkey’, in Macroeconomic Analysis of
Turkey: Essays on Current Issues (ed.) E.M. Üçer, Central Bank of Turkey, pp.51-100.

Stern, R.M., Baum, C.F. and Greene, M. N. (1979). ‘Evidence on Structural Change in the
Demand for Aggregate US Imports and Exports’, Journal of Political Economy, 87(1) pp.

Şahinbeyoğlu, G. and Ulaşan, B. (1999). ‘An empirical examination of the structural stability of
export function’, Central Bank of Turkey Discussion Paper No. 9907.

Sönmez, Mustafa (2005), “Türkiye İhrcatının İthalata Bağımlılığı: 2000-2004”, unpublished

Yükseler, Z. and Türkan, E. (2006). The transformation in the production and foreign trade
structure of Turkey: Global tendencies and reflections, TÜSİAD-Koç University
Economic Research Form, İstanbul, http://eaf.ku.edu.tr/working_papers.asp.


6. Appendix

Table A: Unit Root Test Results(1)

Level First Difference
Exports 2.79 (1.00) 0.81 (0.99) -3.50 (0.00) -13.87 (0.00)
Imports 0.73 (0.99) 0.20 (0.97) -5.20 (0.00) -13.32 (0.00)
OECD Income -1.21 (0.66) -1.71 (0.42) -4.94 (0.00) -4.84 (0.00)
REER_CPI(2) -2.68 (0.08) -2.69 (0.08) -8.20 (0.00) -8.88 (0.00)
REER_ULC -2.36 (0.15) -2.36 (0.15) -7.39 (0.00) -7.30 (0.00)

(1) ADF is for the Augmented Dickey-Fuller Unit Root Test and PP is for the Phillips-Perron Unit Root Tests. Test equations
include intercepts. Values in parentheses are the probabilities. Schwarz Information Criteria and Newey-West using Bartlett
kernel are used to calculate lag length for ADF and bandwidth for PP tests respectively.

(2) Though standard tests reject unit root in CPI based real effective exchange rate in level form, when we allow for deterministic
trend both ADF and PP tests do not reject unit root at 5 percent level. ADF and PP test statistics in this case are –2.99 and –3.03

Table B: List of commodities made up non-traditional and traditional commodity groups
# Non-traditional (Non-Traditional

Index # Traditional (Traditional Goods) Index

12 Vehicles other than railway or
tramway rolling stock

0.175 2 Live animals 0.740

18 Natural or cultured pearls 0.184 9 Fertilizers 0.724
17 Arms and ammunition 0.204 4 Articles of leather 0.574
11 Furniture 0.219 7 Meat and edible meat offal 0.574
19 Stone, plaster, asbestos 0.231 11 Cereals 0.560
20 Essential oils and resinoids,

perfumery, cosmetics
0.241 8 Residues and waste from the food industries 0.542

13 Nuclear reactors, boilers, machinery
and mechanical appliances

0.243 6 Prep. of meat, of fish or of crustaceans, mulls
of other aquatic invertebrates.


7 Ships, boats and floating structures 0.249 5 Other vegetable textile fibres, paper yarn and
woven fab


2 Albuminoidal substances 0.257 1 Products of animal origin; not elsewhere


6 Electrical machinery and equipment 0.267 17 Tobacco and manufactured tobacco


8 Aircraft, spacecraft and parts thereof 0.271 18 Wadding, felt and nonwovens 0.490
10 Cocoa and cocoa preparations 0.274 13 Coffee, tea, mate and spices 0.487
3 Aluminium and articles thereof 0.277 15 Knitted or crocheted fabrics 0.477
14 Optical instruments and apparatus 0.280 14 Organic chemicals 0.476
1 Miscellaneous articles of base metal 0.281 20 Edible vegetables 0.475
15 Toys, games and sports 0.283 3 Railway and tramway locomotives, rolling

stocks and parts thereof

16 Plastics and articles thereof 0.285 16 Vegetable plaiting materials 0.456
9 Paper and paperboard 0.294 10 Animal or vegetable fats and oils 0.456
5 Other made up textile articles 0.296 19 Wool, fine or coarse animal hair 0.437
4 Miscellaneous manufactured articles 0.298 12 Inorganic chemicals 0.432

Note: Rankings are according to their order in the panel co-integration estimates.