Vertical Specialization of Trade and Its Changes in the OECD Countries: an Input-Output Analysis

(1)

Vertical Specialization of Trade and Its Changes in the OECD

Countries: an Input-Output Analysis

Final version

International Economics & Business

Bing Zhao

August 30th, 2008

Faculty of Economics

University of Groningen

Supervisor: Prof.dr. Erik Dietzenbacher

Abstract

(2)

1 Introduction

The pattern and nature of international trade have been changed significantly during the last several decades. Due to the rise of global production value chains in recent years, international trade is not just trade in final goods but more and more also in intermediate goods. Each country engages and specializes in different production processes in a value chain and increasingly exports to other countries its intermediate and final goods that it specializes in. Therefore, vertical specialization or fragmentation of industries across countries has become an important element of the changing pattern and nature of international trade. Moreover, the measurement of vertical specialization has been proved to be a useful way to proxy globalization. Generally speaking, vertical specialization of trade exhibits an increasing interconnectedness of production processes in a vertical trading chain, with each country specializing in certain stages of the production chain.

Despite the vast amount of theoretical models on vertical specialization, there is a lack of empirical evidence, other than case studies. One exception is Hummels et al. (2001). They used the so-called the imported input content of exports to quantify vertical specialization across countries by using national input-output tables. They measure vertical specialization of trade by calculating direct and indirect use of imported inputs to produce the exports of a country. Their empirical results show a clear upward trend of vertical specialization of trade in OECD and some emerging economies.

The method of Hummels et al. (2001) is similar to the backward linkages in the input-output literature where backward multipliers obtained from input coefficients are used to measure the imported inputs embodied in production of the exports of a country1. The backward linkages method has been widely applied in many other studies. For example, Dietzenbacher and Los (2002) use backward linkages to analyze by how many dollars R&D expenditures of one industry would increase if the final output of the other industry were to grow by one dollar. Meanwhile, there is also a strand of input-output literature focusing on forward linkages, which use forward multipliers obtained from output

1

(4)

coefficients (For example, see Dietzenbacher and Los, 2002). Therefore, it seems logical for us to analyze also vertical specialization of trade by using forward linkages method. Building on Hummels et al. (2001), we will develop formulas to calculate vertical specialization of trade by using the forward linkage approach in this paper.

After analyzing vertical specialization of trade, we employ a structural decomposition method to separate the changes in vertical specialization to investigate what factors are behind these changes. We divide the changes in vertical specialization of trade into the effects caused by changes in information technology (IT) industries and the effects caused by changes in non-IT industries. Moreover, it is well known that the results of a structural decomposition are not unique2. Following Dietzenbacher and Los (1998), we will use the mean of the two polar decomposition results to cope with potential bias.

The starting point of our empirical research is to investigate the trend of vertical specialization of trade in recent years. We will use output tables from OECD input-output tables (1995, 2006) for 9 OECD countries and two emerging economies3. We calculate vertical specialization (VS) shares for 9 OECD countries from 1970 to 2000, and for Brazil and China from 1995 to 2000 since there are no early data available for Brazil and China. Further, we determine decomposition results based on constant price tables of OECD input-output tables (1995) because we have to control for price changes of IT products over the years4.

Our empirical results show that VS shares of exports were always increasing for almost all countries except Japan and Denmark where these two countries exhibited fluctuation in their VS shares of exports between 1970 and 1990. Between 1990 and 2000, the VS shares increased in France, Germany, Japan, the Netherlands, and the US. On the

2

If n is the number of variables in the expression that is to be decomposed, there exists n! decomposition formulas. For a detailed discussion of non-uniqueness of structural decomposition, see Dietzenbacher and Los (1998).

3

The 9 OECD countries are Australia, Canada, Denmark, France, Germany, Japan, the Netherlands, the United Kingdom, and the United States. The emerging economies are Brazil and China. We exclude countries that have only one table (for example, Italy has only one table available before 1990).

4

(5)

contrary, the VS shares of Canada, Denmark, and the United Kingdom experienced small negative growth rate. Large emerging countries Brazil and China exhibited relatively high VS shares, especially China’s VS shares are as high as those of France and the United Kingdom. The growth rate of VS shares in Brazil reached around 16% which was the third fastest growing country in our dataset.

Finally, our decomposition results show that the changes of VS shares are mainly caused by the changes from IT industries. In most cases, the contribution of IT industries accounts for most changes of VS shares between 1970 and 1990. On the other hand, non-IT industries display negative impacts on the changes of VS shares in most of cases between 1970 and 1990.

We make three contributions in this paper. First, we use newly published OECD input-output tables (2006) to analyze the recent trend of vertical specialization of trade for a group of OECD countries and two emerging countries. Second, we develop formulas to measure vertical specialization of trade by using forward linkages method. Third, we use structural decomposition method to investigate what industries contribute the most to the changes of vertical specialization of trade.

(6)

2 Literature review

In this literature review we will first discuss the concepts of international trade and production value chain as well as the relationship between these two. International trade has changed considerably in the late 20th century. Countries are connected by trading not only finished goods but also intermediate inputs and they preside in different stages of production chain. For example highly developed countries specialize in high value-added and high technology stages of production, whereas developing countries specialize in low value-added and labor intensive stages. Therefore, the process of globalization exhibits specialization in different stages of the production value chain. This leads to our discussion of vertical specialization in trade.

Theoretical insights from a range of theories related to vertical trade specialization are introduced. Several empirical studies of vertical specialization, especially in OECD and large emerging countries, will be discussed. The aim is to provide a foundation for the discussion of the extent of vertical specialization in OECD countries.

Finally, in order to discover the sources of vertical specialization, we will introduce a so-called structural decomposition technique. It allows us to identify what contributes the most to the changes of vertical specialization.

2.1 International trade and production value chain

2.1.1 International trade patterns and models

(7)

Rodrigue et al. (2006) explain that significant technical changes in the transport sector shortened the transport time of large amounts of goods at similar or lower costs after 1960s. It has become increasingly possible to trade between parts of the world that previously had limited contact with each other. Similarly, Jesse et al. (2008) argues that GATT/WTO membership contributes positively to international trade by weakening high market-protecting institutions and therefore lowering transaction costs. As a result, international trade grew dramatically after 1960. Husted and Melvin (2001) estimate that the world total exports, measured in 1950 real volume terms, were 2200 percent higher in 2002 than in 1950. Further, the division and the fragmentation of production that went along with technological changes in the transport sector and the development of trade treaties also expanded trade. Kleinert (2003) estimates that imported inputs account for roughly one half of total imports of developed countries and this share proved to be remarkably constant over the 1980s and 1990s. Despite the increasing share of trade in intermediate goods, it is surprising that theoretical and empirical analyses on trade have almost exclusively been dealing with trade in final goods. Only recently, with the increasing international division of labor through fragmentation of the production process, research in trade moved its focus from trade in final goods to trade in intermediate goods (For an overview, see Krugman, 1995, or Jones and Kierzkowski, 2000.).

Alongside the changing pattern and volume of international trade, also theoretical trade models have changed to accommodate recent trends. We will first give an overview of the classical trade models, and then we will discuss their recent extensions5.

The classical trade models, i.e. the Ricardian model and the Heckscher-Ohlin model, use comparative advantages as the basic motive for trade between countries that employ either only labor or labor and capital as factors of production. For example, in the Heckscher-Ohlin model, each country only produces the same two products, one of which uses labor intensively and the other uses capital intensively. The country that has relatively abundant labors will specialize in producing the labor intensive product (in which it has comparative advantage), and vice versa. The model predicts that -under free

(8)

trade- each country will completely specialize in the production of the good it has a comparative advantage, this is the so-called law of comparative advantage.

The extensions of the Heckscher-Ohlin model have been done in different respects in order to cope with the real world situation. For instance, Dornbusch et al. (1977) derive the law of comparative advantage by introducing a continuum of goods instead of only two in the Heckscher-Ohlin model. Later, Dornbusch et al. (1980) show that a definite pattern of specialization with a continuum of goods is not always obtained if there are more goods than factors. Only when endowments differ sufficiently, complete specialization will take place.

Recently, extensions of the classical trade models shift their focus from trade in final goods to trade in intermediate goods due to the growing division of labor and fragmentation of the production process. Intermediate goods are goods used as inputs into other production processes ranging from raw materials, metals and oil, to machine tools and computer chips. Feenstra and Hanson (1996) among others, introduce intermediate goods in the Heckscher-Ohlin model as inputs for the production of a final good. Their empirical research shows that the law of comparative advantage prevails for such intermediate goods. In a similar manner, Campa and Goldberg (1997) confirm a growing dependence on imported inputs in the production of nearly all manufacturing industries in Canada, the United Kingdom and the United States. The exception is Japan where almost no sector relies so heavily on imported intermediate inputs. Further, contrary to the other analyzed countries, the imported intermediate goods’ share of total imports of Japan declined over the last 25 years.

(9)

2.1.2 Global production value chain

We now proceed to discuss what behind the changing process of globalization and the growth in intermediate products. It has been argued by Lawton and Michaels (2000), among others, that a fundamental change of production structure differentiates the globalization process of the late 19th and late 20th centuries. The growing importance of international production sharing has been identified by them as one of the most important driving forces of globalization. Porter (1985) was one of the first to introduce the concept of production value chain, he defines it as: “The value chain is a collection of activities that are performed by the firm to design, produce, market, deliver, and support a product or service.” (Porter, 1985, p. 36)

From figure 1, we can see that a firm’s value chain is consisted of all primary and secondary activities which again are part of a large value system.

Figure 1 Porter’s Value Chain and Value System

(10)

As the international division and specialization of labor and capital develops, a

multinational’s value chain may locate design, production, distribution and other value-added activities across different countries. Modern companies operate with production plants, subsidiaries or through arm-length relationships, in several countries to exploit their competitive advantages. For example, Nike designs its shoes in the US; produces them in Vietnam and China; sets up customer service centers in India; and distributes them back into the US. Consequently, the global value chain covers production activities that are geographically dispersed across different countries. It includes coordination and re-integration of different production activities where each country specializes in one or more production activities. The development of global value chains enables modern companies to exploit the advantages of different locations and facilitates vertical specialization in trade.

2.2 Vertical specialization in trade

In this section, we will discuss the concept of vertical specialization and its application in empirical research.

2.2.1 Vertical specialization

In recent years, production processes are divided into many stages as a result of the development of global value chains. Different production stages are carried out in different locations. Thus, the production of a finished good involves the participation of many countries, with each of them specializing in different stages of the vertical production chain. International trade is then increasingly dominated by trade in intermediate goods. In the international trade literature, this phenomenon is called vertical specialization in trade.

(11)

subsets of US import data to construct measures of outsourcing and found that outsourcing in the US is 10% to 15% of total imports. Miller (2001) looks at cross-border outsourcing by US firms. By using a computable general equilibrium model, he found that the growth of income inequality between skilled and unskilled workers in the US was mainly due to changes in the structure of production not the outsourcing activities. Egger and Egger (2001) examine outsourcing by European Union (EU) firms in non-EU countries. Their results show that outsourcing is more prevalent in import-competing industries (the industries compete with other countries’ imports) in EU which use low-skilled labor intensively. However, it is interesting to note that outsourcing usually happens between developed countries and developing countries. There is no or very little outsourcing going back from developing countries to developed countries. Thus, another strand of literatures emerges in order to investigate the involvement of developing countries in the global production value chain.

More recently, vertical specialization in trade focuses on measuring the imported input content of exports. Different from outsourcing, vertical specialization of trade takes into account of the involvement of developing countries in the globalization process. Hummels, Ishii, and Yi (2001, hereafter HIY) was among the first to use the imported input content of exports to measure vertical specialization in trade by using a national input-output framework. They used national input-output tables to measure vertical specialization because it is impossible to obtain data on the production process and direction of trade flow for every stage of each good that is traded. They found that vertical specialization was increasing for a group of OECD and emerging countries from 1968 to 1990, which implies a spread of globalization in terms of interconnectedness of industries in these countries. Also, smaller countries have higher vertical specialization (VS) shares, the VS shares have increased for all countries but Japan between 1968 and 1990, and the variation in industry VS shares accounts for almost all of the overall VS share variation over time and across countries.

(12)

(2007) and Koopman et al. (2008) estimate the share of imported input of exports for China by refining the original HIY model. Both of them identify that the HIY formula is not appropriate for countries that engage actively in tariff/tax-favored processing exports such as China and Vietnam. They adapt the HIY formula to capture a large amount of processing trade in China. According to the estimates by Koopman et al. (2008), the share of import content of exports in China is about 50%, almost twice the estimate given by the HIY formula. Further, Yi (2003) shows that a dramatic increase in vertical

specialization after the Second World War is likely to be responsible for a faster growth of world trade relative to world GDP over the last five decades.

2.2.2 The idea of vertical specialization

We will follow HIY’s idea of verticality in trade which basically consists of the following elements (HIY, p. 77):

(1) A good is produced in two or more sequential stages,

(2) Two or more countries provide value-added during the production of the good, (3) At least one country uses imported inputs in its stage of the production process, and some of the resulting output is exported.

Figure 2 demonstrates the concept of vertical specialization of trade. Country 2 is the target country; its gross output uses both domestic and foreign intermediate goods. So, the issue is to investigate by how much country 2’s export consists of imported

(13)

Figure 2 Illustration of imported intermediate goods as part of country 2’s export

Source: Adapted from Hummels et al. (2001)

2.3 Structural decomposition

In order to disentangle what is behind the changes of vertical specialization of trade, we will employstructural decomposition techniques in our research. Structural

decomposition techniques are a common and useful analytical tool for unraveling the growth over time in certain variables. Separating the changes of a variable into its constituent parts can identify what factors contribute the most to the change of a variable (for an overview of the early literatures on input-output structural decomposition analysis and its relationship to other methodologies, see Rose and Casler 1996). The application of structural decomposition is widely used in many areas of economic analyses.

(14)

The methodology of structural decomposition analysis is similar to that of growth

accounting, where the objective is to break down the growth rate of aggregate output into contributions from the growth of the inputs and the change in technology. It is well known that structural decomposition does not yield a unique result. Depending on the choice of the decomposition formula, the contribution of a certain source appears to exhibit considerable variability. Dietzenbacher and Los (1998) examine 24 equivalent decomposition forms in analyzing the changes in sectoral labor costs and imports by using Dutch input-out tables from 1986 to 1995. Their empirical results show that the mean of the two so-called polar decompositions appears to be considerably close to the mean of the full set of 24 decompositions. Therefore, we will use the mean of the two polar decompositions to investigate which factors contribute to the changes of vertical specialization and to what extent. Our strategy is to decompose the changes of vertical specialization of trade into two parts, namely the contribution from IT and non-IT industries. The next section will provide our intuition behind this strategy.

2.4 Information technology and output

(15)

Italy, Japan, the United Kingdom, and the US. Their results show that in the 1980s and 1990s, ICT contributed between 0.2 and 0.5 percentage point per year to economic growth, depending on the country. This figure rose to 0.3 to 0.9 percentage point per year in the second half of the 1990s.

The contribution of information technology equipment to economic growth is not only observed in developed countries but also in emerging economies. Piatkowski (2003) shows that the contribution of investments in information technology (IT) hardware, software and telecommunication equipment to output growth and labor productivity in 8 East European countries between 1995 and 2000 was much higher than what might have been expected on the basis of the level of their GDP per capita. Thus, it suggests that the transition economies, through the use of ICT, are benefiting from the technological upgrade to increase the growth rates in output and labor productivity and hence accelerate the process of catching-up.

Different classification of IT producing and using sectors has been used in different context6. In this paper, we will follow the classification of IT sectors used by Jorgenson et al. (2005). In their book, they defined IT sectors and non-IT sectors according to whether an industry produces IT products, or uses more than 15% IT products as the capital inputs. They grouped IT producing and using industries as IT sectors and the rest as non-IT sectors. Appendix 1 shows a detailed classification of IT and non-IT industries.

One problem arises when analyzing the role of ICT investment in economic growth and international trade because the traditional price index is no longer able to capture adequately quality-price changes of ICT products. It is well known that the computing power of semiconductors and computers have improved tremendously over the past few decades, while at the same time the prices have decreased. Taking quality improvements into full account, the same computer would have become much cheaper. Traditional price indices for these goods will certainly underestimate the rate of price decline and, because of that, the rate of productivity growth (see Nordhaus, 2001, for a long-term perspective

(16)

on the increase in computing power). Thus, one of the most difficult issues in analyzing the effects of investments in IT equipment and software to economic growth is how to measure constant quality prices. For example, a normal computer with a very low power of computing might have cost $1000 in the 1990s, while a powerful computer can be bought for just $500 nowadays. Therefore, as pointed out by Jorgenson et al. (2005), it is important to separate the observed price changes into changes in ICT equipment

performance and changes in prices that would have applied if performance were held constant.

The development of a constant quality price index first began in the U.S. by the Bureau of Labor Statistics (BLS). In 1997, BLS incorporated a matched model price index for semiconductors in the Producer Price Index (PPI) and after that the national accounts have been composed by using data from PPI7. Identified by Wyckoff (1995), because of discrepancies among official price indexes for IT equipment and software in OECD countries, an international price index is much more difficult to construct. Schreyer (2000) provides the first set of “international harmonized” deflators for G7 countries, which controls for some of the differences in methodology of constructing price indexes in different countries. However, an important limitation of the study is that calculations could only be carried out for the years up to 1996 for all G7 countries due to internationally incomparable data. Thus, there is no claim that the “harmonized” deflator is necessarily the correct price index for a given country.

Regarding our empirical analyses in structural decomposition, it is the best for us to construct input-output tables in constant quality prices. However, this data construction process is tremendously time consuming and beyond the scope of this paper. Hence, we have to use the constant price tables instead of the current price tables in structural decomposition because of the following reasons: Constant prices ensure that the overall effect of price inflation is eliminated, whereas current prices assume that price inflation

7

(17)

affects the prices of the inputs and outputs in a country to the same extent8. Therefore, we will use constant price tables because we want to eliminate the price inflation of IT related products on the changes of VS shares.

8

(18)

3 Research questions & hypotheses

3.1 The extent of vertical specialization

Empirical evidence of vertical specialization in trade was provided by Hummels et.al (2001) and in general focused only on early years (before 1995). The publication of recent input-output tables by OECD (2002, 2006) offers the opportunity to analyze recent development of verticality of trade.

Therefore, in order to study the recent development of vertical specialization, this paper first gives an overview of the extent of vertical specialization in OECD countries. (Research question 1)

Varies sources including case studies and empirical researches suggest that vertical specialization has increased over the years9. Thus, we expect that vertical specialization has also increased after 1995. In general, we expect that small countries such as the Netherlands and Belgium in our dataset may exhibit large vertical specialization because they need more foreign resources to fulfill domestic and foreign demand of their products. Large countries, for example the US and Japan, which produce more output and have more resources are expected to be more self-sufficient than smaller countries, thus vertical specialization in large countries will be smaller. Next to this thought, export oriented countries may exhibit large vertical specialization due to the interconnectedness of their domestic industries and foreign industries. Driven by large foreign demand of domestic products, export oriented countries need more raw materials and intermediate goods from other countries to produce final products. Thus, their vertical specialization is high.

3.2 An alternative measure of vertical specialization

(19)

measurement uses forward linkages to provide a different perspective on vertical specialization.

Consequently, we will develop new formulas to measure vertical specialization by using the forward linkages method, and we will compare new formulas to HIY’s formula. (Research question 2)

The input-output literature indicates that both measurements of vertical specialization are essentially the same thing; the only difference is their interpretation10. We will provide a mathematical proof to show that HIY’s formula and our formula are mathematically equivalent in the following section.

3.3 The contribution of IT products to the change of vertical specialization

Technology advances and technical changes in production processes are widely observed both in developed and developing countries. The intensive use of IT equipment and software due to their ever decreasing constant quality prices revolutionizes not only production processes but also new product innovation processes. Our intuition suggests that the intensive use of IT products in some industries may contribute the most to the growth of vertical specialization compared to other industries. We will separate all industries into IT industries and non-IT industries. As a result, we decompose the change of vertical specialization into several constituent parts in order to investigate what sources contribute the most to the change of vertical specialization.

Therefore, this study will analyze whether the contribution of IT products to the change of vertical specialization is more than that of non-IT products. (Research question 3)

The IT revolution has a profound influence on every corner of our society. In particular, the resurgence of economic growth in developed countries is largely due to the

10

(20)

(21)

4. Methodology

4.1 Vertical specialization: Backward linkages

Hummels et al. (2001) use the so-called imported input content of exports to quantify vertical specialization across countries by using national input-output tables. Their formulas to measure the import content of export correspond to the backward linkages in input-output literature (see for example Miller & Blair, 1985). Essentially, they measured by how much the export of industry i in a country consists of imported intermediates. They used the following equation:

i i i i )*export output gross tes intermedia Imported ( VS1 = (1)

The interpretation of equation (1) is the following. For industry i, it multiplies the value of exported products with the percentage of imported inputs per dollar of production. Simply put, it represents the foreign value-added embodied in the exports.

The calculation of (1) is straightforward for each industry and for a country as a whole,

we have

∑

=             = n i1 i i i M i *e X Z

VS1 , where ZMi stands for the imported intermediate goods

of industry i in a country,X is the total output of industry i, and i e is the export of i

industry i, n is the total number of industries. Summing over all industries gives total vertical specialization of a country and dividing by the aggregated exports of all industries represents an export-weighted average of the industry VS export shares.

∑

= = = _n i i n i i e VS 1 1 1 E VS1 , (2)

(22)

Hummels et al. (2001) use national input-output tables to measure this vertical specialization as it is impossible and tedious to obtain data on the production process and the direction of trade flows for every stage of each good that is traded except on a case-by-case basis. Since input-output tables provide a relative easy way to empirically analyze vertical specialization at the macro level, we will first introduce the input-output framework. Typically the input-output framework consists of the following equations:

e C AX + + = X or X ( ) 1( ) e C A I− + = − , (3) M F MX + = Y , (4)

where X is a n×1 vector of gross outputs , Y is a n×1 vector of imports, and e is a n×1 vector of exports. The elementsa_ij and m_ijof matrices Aand M represent the domestic and imported input of goods i per dollar of outputs in industry j. The vector C represents investment and final demand for domestic goods and the vector M

F is the final demand for imported goods.

The coefficient matrices A and M are determined as follows,

1 ) ˆ ( − =Z x A D _and M_{( )}_ˆ 1

M ₌Z x − _{, where a “hat” is used to indicate a diagonal matrix,}

D

Z and M

Z are domestic and imported intermediate inputs (n×n matrices). Equation (3) represents a domestic input-output relationship, whereas MX in equation (4) shows the requirement of imported inputs in the production of domestic outputs.

By using the techniques in input-output analysis, we can rewrite equation (2) into the following equation, MS u e u Me u ' ' ' DVS= = , (5) where u is1× vector of 1’s , n e u e S '

(23)

$100 imported steel and $100 domestic machinery. Suppose that the production of $100 machinery uses $10 imported steel, so the direct and indirect share of steel in the exports are 0.1 and 0.01.

We can compute total VS by using the following equation,

S A I M u e u e A I M u 1 1 ) ( ' ' ) ( ' TVS − − − = − = , (6) The term₍ ₎−1 − A

I is usually called the Leontief inverse matrix and its elements (i,j) captures the imported input of good i to be embodied in a dollar final demand of good j. Since exports are part of final demand, it measures the amount of import of good i needed for the exports of good j. In general, equation 6 uses backward linkages and multipliers as obtained from the input coefficients A and M .

Hummels et al. (2001) use equation (6) as their main measure of vertical specialization. As a result, they found that vertical specialization was increasing for a group of OECD and emerging countries from 1968 to 1990, which implies a spread of globalization in terms of interconnectedness of industries in these countries. Moreover, smaller countries have higher VS shares; the VS share has increased for all countries but Japan between 1968 and 1990. Further, their decomposition results show that the variation in industry-level VS shares accounts for almost all of the overall VS share variation over time for every country in their study.

4.2 An alternative method: forward linkages

(24)

countries, i.e. we want to measure, for a country, the dollar amount of exports goods j due to the dollar amount of imports good i .

We again start with the input-output framework. First, we rewrite equation (3) into the following equation,

' '

'

X'=u ZD +uZM +v , (7)

where 'v is the row vector of value added terms.

In order to build up a forward linkage, we can writeX'= X'H+(u'ZM +v')or

1 ) )( ' ' ( X' − − + = uZM v I H _where D Z x H (ˆ)−1

= . A typical element h_i_,_j of H denotes the share of the output of industry i that is sold to industry j. Output coefficients and the inverse matrix ₍ ₎−1

−

= I H

G have been widely used for measuring forward linkages in the framework of the supply-driven Ghosh (1958) model (see Dietzenbacher, 1997, for an overview) and G is usually called the Ghosh inverse. A typical element gi,j denotes the additional output value in industry j when the primary costs in industry i are increased by one dollar. Then, we define the export shares of total outputsK ₌(xˆ)−1e_{. Thus, an} element (i, j) of (I₋H)−1K

represents additional value of export good j per extra dollar of import by industry i.

Export destinations of the imports for country r can be measured in the following equation,

VS2 u'ZM(I H)−1K −

= , (8)

An element (i, j) of VS2 measures how much of actual imports of goods i ends up (is embodied) in exports of industry j

(25)

u Z u K H I Z u u Z u M M M ' ) ( ' ' VS2 −1 − = , (9)

The term (I−H)−1K measures the total export destination of the imports in the sense that it determines the export of good j due to import of good i at after good i embodied in good j that is exported.

Backward and forward linkages and multipliers are generally viewed as indicators of two entirely different things. The backward linkages method provides a measure of the imported inputs ‘needed’ directly for an industry i to produce exports and, indirectly, for the inputs necessary to produce i and for the inputs required to produce the inputs, and so on. Forward linkages method is rooted in an opposite direction. It measures where exports go to by using output multipliers. However, it turns out thatZM(I₋H)−1K ₌M(I₋A)−1e

. The following proof shows this result: e A I M e x H x I M e x H I x M K H I ZM( )−1 ˆ( )−1(ˆ)−1 ( ˆ (ˆ)−1)−1 ( )−1 − = − = − = − . Since D Z H xˆ = and D Z x

A =ˆ , we have xˆH(xˆ)−1 = A and the rest follows. Therefore, the two expressions are mathematically equivalent and the numerators of equation (6) and equation (9) are the same11.

4.3 Decomposition of vertical specialization

Vertical specialization increased considerably in the past few decades through close interconnectedness of production value chains. In order to disentangle what industries contribute the most of a country’s vertical specialization, we will decompose the changes of vertical specialization into two components, namely IT industries and non-IT

industries. Jorgenson et al. (2005) propose a method to distinguish between IT industries and non-IT industries according to whether an industry produces IT related products,

11

(26)

whether an industry uses IT capital inputs more than 15% of total capital inputs12. Our intuition suggests that the tremendous growth in IT-industries may contribute the most to vertical specialization of a country because the use of IT products as inputs influences total outputs and therefore exports. Our intuition is based on the special features of IT products: they are capital intensive to innovate and design but labor intensive to produce and assemble; their transportation cost per volume is low (high value but lightweight) which provides a motive for international trade. Therefore, the international trade of IT products may especially exhibit a vertical linkage between developed and developing countries based on competitive advantages. Since IT using sectors use a relatively large amount of IT products as inputs, vertical specialization of trade in IT using sectors are also evident. Thus, we group IT using and producing sectors as IT industries.

Since forward and backward linkages are mathematically equivalent, we will only decompose vertical specialization of equation (6) in this section. Before we proceed our decomposition of the change of VS share, we will first provide a simple example of structural decomposition to illustrate how this method works. Suppose that X =YZ , where X , Y and Z can be scalars, vectors or matrices. The change in X between two points in time is ) 1 ( ) 2 ( X X X = −

∆ , can be decomposed as follows:

) )( 1 ( ) 2 ( ) ( Y Z Y Z X = ∆ − ∆ ∆ (10) =(∆Y)Z(1)−Y(2)(∆Z) (11)

Equations (9) and (10) are called polar decompositions. In this simple example, there are only two components on the right hand side of each equation. However, as the number of components increases to n, there are n ! ways of decomposition (Dietzenbacher and Los, 1998).

(27)

polar decompositions appears to be considerably close to the mean of the full set of 24 decompositions13. Thus, we derive two polar decompositions for the changes of vertical specialization.

Based on equation (6), we can decompose ∆TVS into the following equations,

∆(TVS)=TVS(2)−TVS(1) ' ( ) 1(2) (2) ' (1)* ( ) 1 (2) S A I M u S A I M u∆ − − + ∆ − − = +uM I−A − ∆S ) 1 ( ) )( 1 ( ' 1 , （12）

where (1) and (2) stand for the first sub-period and the last sub-period.

Rewriting equations 9, we can derive the following equations: ) 1 ( ) 2 ( TVS TVS TVS = − ∆ ' ( ) 1(2) (2) ' (1)

(

( (2)) 1 ( (1)) 1

)

(2) S A I A A I M u S A I M u∆ − − + − − ∆ − − = +u'M(1)(I−A)−1(1)∆S_{, (13)} where ( )−1

(

( (2))−1 ( (1))−1

)

− ∆ − = −

∆ I A I A A I A has been applied14.

The other possibility to decompose the change of VS is the following,

) 1 ( ) 2 ( TVS TVS TVS = − ∆ ' *( ) 1(1) (1) ' (2)* ( ) 1 (1) S A I M u S A I M u∆ − − + ∆ − − = +uM I−A − ∆S ) 2 ( ) ( * ) 2 ( ' 1 ' ( ) 1(1) (1) ' (2)

(

( (2)) 1 ( (1)) 1

)

(1) S A I A A I M u S A I M u∆ − − + − − ∆ − − = 13

Dietzenbacher and Los (1998) analyze an equation with 4 variables on the right hand side and therefore there are 24 (4!) different decomposition results.

14

(28)

+u'M(2)(I−A)−1(2)∆S_{, (14}_）

where equation (14) is the mirror image of equation (13) in the sense that it takes the other polar case into consideration.

Finally, the mean of two-polar decomposition is the best approximation for the mean value of all decomposition possibilities according to Dietzenbacher and Los (1998). We therefore derive the mean value of two-polar decomposition,

TVS ∆ =0.5 equation (13) + 0.5 equation (14)

(

)

(

u ∆M I−A S +u M I−A ∆A I−A S +u M I−A ∆S

)

= ' ( )− (2) (2) ' (1)( (2))− ( (1))− (2) ' (1)( )− (1) 2 1 1 1 1 1 +

(

)

(

u∆M I−A − S +u M I−A − ∆A I−A − S +u M I −A − ∆S

)

) 2 ( ) )( 2 ( ' ) 1 ( )) 1 ( ( )) 2 ( ( ) 2 ( ' ) 1 ( ) 1 ( ) ( ' 2 1 1 1 1 1 , （15）

In order to compare the difference in the contribution of IT industries and of non-IT industries to the changes of vertical specialization, we decompose the change of vertical specialization into two parts: the change of non-IT industries and the change of IT industries. Then, we can write∆A=∆A1+∆A2 , where ∆A1is the input coefficient changes

in non-IT industries and ∆A₂is the input coefficient changes in the IT industries. Thus, we can calculate the contribution of the IT and non-IT industries to vertical specialization by determining by how much ∆A₁ and ∆A₂contribute to the total change of vertical specialization.

One possible way of decomposing A∆ is to write∆A=

[

∆A₁|0

] [

+ 0|∆A₂

]

; the other way

is to write       ∆ +       ∆ = ∆ 2 1 0 0 A A

(29)

all industries. The later assumes that the use of IT products has changed which means that new IT products have been invented.We feel that both ways of decomposition have their pros and cons because the IT revolution has changed not only production processes of IT industries but also IT products. In our research, we want to build up a relationship between the use of new IT products in production processes and the changes of vertical specialization of trade. Moreover, the use of new IT products may change the quantity and the quality of a country’s output which in turn may influence the country’s export structure. Consequently, the changing export structure will influence the share of vertical specialization in the country. Thus, we only consider the impact of new IT products on export structures in this research. This implies that we will only take into account of row decomposition.

We can further decompose equation (12) into the changes in IT product using and producing industries and the changes in non-IT industries. Rewrite equation (12) into the following equation, ) 1 ( ) 2 ( TVS=TVS −TVS ∆ =∆ML(2)S(2)+M(1)

(

L2∆AL(1)

)

S(2) +M(1)L(1)∆S , (16) where we use the fact that ₌( ₋ )−1

A I L . For IT industries,                          ∆             +                  ∆ = − − − − − − nonIT IT IT non IT IT IT non IT IT non IT IT non IT IT non IT IT S S L L A L L M M S S L L M ) 2 ( ) 2 ( ) 1 ( ) 1 ( 0 ) 2 ( ) 2 ( ) 1 ( ) 1 ( ) 2 ( ) 2 ( ) 2 ( ) 2 ( 0 _     ∆             + − − 0 ) 2 ( ) 1 ( ) 1 ( ) 1 ( ) 1 ( _IT IT non IT IT non IT S L L M M , (17)

where IT and non-IT stand for IT industries and non-IT industries. Similarly, the same calculation can be applied to non-IT industries except that we use 

(30)

and       −IT non S 0

. Therefore, we can transform the change of VS shares into six parts with three parts from IT industries and three parts from non-IT industries. Finally, we can do the same analyses on equation (14) and apply equation (15) to calculate the average of two polar decompositions.

(31)

5. Data description

Our methodology in the previous section is based on a set of national input-output tables compiled by OECD (1995, 2006). A national input-output table is usually consisted of a n×n (n is the number of industries in a country) domestic intermediate input matrix and a n×n imported intermediate input matrix. OECD (1995) contains input–output tables for 10 developed countries, using a 35-industry classification15. There are three to five tables available for each country for the period 1968–1990. However, the years for which tables are compiled do not exactly coincide and we have to follow the suggestion made in OECD (1995) to assign each table to a sub-period. Table 1 presents this grouping of tables. In addition, these input-out tables were compiled by using current and constant prices for each country except the US where only current price tables are available.

Different from OECD 1995, OECD 2006 uses a more detailed 48-industry classification for a group of OECD countries and some emerging markets, such as China, and Brazil16. Again, the years of compilation of these tables are not exactly the same which forces us to assign them to a sub-period (see table 1). Further, this dataset only contains current price tables.

We finally decided to use 9 countries’ tables from the OECD 1995 and their tables from the OECD 2006. Further, including large emerging countries may give us a clear indication of whether vertical specialization occurs in countries other than highly developed ones. Therefore, we include China and Brazil in our dataset. In order to analyze whether IT industries contribute more to the changes to vertical specialization than other industries, it is ideal to use an internationally harmonized price index to account for changes in constant quality price of IT products. However, this requires a large amount of time to adjust our dataset, which is out of the scope of this research. Thus,

15_{These countries are Australia (AU), Canada (CA), Denmark (DK), France (FR), Germany (GE), Japan (JP),The}

Netherlands (NL), the United Kingdom (UK) and the United States of America (US). The OECD 1995 also contains one table for Italy but we will not use it since we want to estimate VS changes

16

(32)

we decided to use constant price tables to proxy the constant quality price. Since there are no constant tables available after 1990, we can only decompose the changes of vertical specialization before 199017.

Table 1: Availability and grouping of OECD input-output tables

1971 AU(68) CA(71) DK(72) FR(72) JP(70) NL(72) UK(68) US(72) 1976 AU(74) CA(76) DK(77) FR(77) GE(78) JP(75) NL(77) US(76) 1980 CA(81) DK(80) FR(80) JP(80) NL(81) UK(79) US(80) 1985 AU(86) CA(86) DK(85) FR(85) GE(86) JP(85) NL(86) UK(84) US(85) 1990 AU(89) CA(90) DK(90) FR(90) GE(90) JP(90) UK(90) US(90)

1995 CA(95) DK(95) FR(95) GE(95) JP(95) NL(95) UK(95) US(95) BR(95) CH(95) 2000 AU(99) CA(00) DK(00) FR(00) GE(00) JP(00) NL(00) UK(00) US(00) BR(00) CH(00)

Note: First column denotes labels for sub-periods. Values in the parentheses refer to years.

Table 2 shows the export/import ratios (x100) per country and its annual percentage changes between the first and last period for all countries in our dataset. These values give a quick overview of the relative size of exports over imports in each individual country. Table 2 shows that most countries have the export/import ratios around 100. One exception is the Netherlands which has this ratio always higher than 120. This implies Dutch economy was export oriented and its trade surplus was always positive during this period. Finally, one surprising result is that the United Kingdom had a large deterioration of the current account in 1990. One possible explanation is that the prices of U.K. exports rose much faster than the price of its imports, the actual trend in trade performance of the U.K. was much better than might be concluded from this finding. As explained by Dietzenbacher et al. (2007), the current account position of the U.K. worsened much less when using current price input-output tables which imply that the change in terms of constant prices can be attributed to a favorable terms-of-trade effect.

17

(33)

Table 2: Export/import ratios(x100) and its annual % changes* Country 1971 Export/Import 1976 Export/Import 1980 Export/Import 1985 Export/Import 1990 Export/Import Annual Percentage Change AU 84.13 93.55 93.60 102.07 86.08 0.12% CA 92.89 83,79 101.14 103.20 86.50 -0.34% DE 73.68 72.7 85.78 89.97 98.96 1.72% FR 82.09 92.86 93.91 103.33 91.68 0.58% GE 89.04 96.79 94.07 102.26 94.52 -0.16% JP 66.82 83.91 95.01 126.53 81.54 1.1% NL 120.32 115.82 125.56 127.50 129.41 0.3% UK 95.11 90.66 99.85 93.66 60.02 -1.84% US 88.81 98.94 107.72 88.71 88.64 -0.19%

(34)

6 Results and discussion

In this section, we will discuss empirical results to the three research questions posed in the early section. Thus, this section is divided into three parts correspond to our research questions.

6.1 The overview of vertical specialization of trade

6.1.1 VS shares between 1970 and 1990

The first step in our calculation is to show the trend of vertical specialization by using HIY’s formulae for 9 OECD countries. For the VS shares between 1970 and 1990, we use constant price tables. In order to investigate the recent trend of vertical specialization, we also include OECD (2006) input-output tables in our calculation. Further, the more comprehensive OECD (2006) contains tables from some large emerging countries, so we include Brazil and China in our calculation.

(35)

table 2). The share of imported inputs over total exports is therefore much higher than before18.

Table 3: the VS share of total export between 1970 and 1990*

1970

1976

1980

1985

1990

Percentage

change between first and last period

AU

0.095

0.112

0.090 0.101 6.3%

CA

0.223 0.234 0.209

0.251 0.267 20.7%

DK

0.329

0.324

0.311

0.302 0.291 -11.3%

FR

0.222

0.212

0.210 0.235 5.9%

GE

0.206

0.219 0.234 13.6%

JP

0.219

0.193

0.151

0.113 0.136 -37.9%

NL

0.294

0.347

0.351

0.354 20.4%

UK

0.213

0.227

0.216 0.466 118.7%

US

0.040

0.057

0.054

0.052 0.089 122.5%

*Calculated from OECD input-output 1995 edition with constant price except the US

6.1.2 VS shares of exports after 1990

The publication of recent input-output tables gives us an opportunity to calculate the VS shares after 1990. Table 4 shows the level and growth of vertical specialization of 9 OECD countries and two emerging economies by using the current price tables19.

As we can see from table 4, the VS shares increased between 1995 and 2000 in France, Germany, Japan, the Netherlands, and the US. In particular, Japanese VS shares increased in this period compared to the period before 1990, and large economies such as Germany and the US kept their growth in VS shares. On the contrary, the VS shares of Canada, Denmark, and the United Kingdom experienced small negative growth rate. This fact partly reflected that the VS shares in these countries had already been relatively high

18

Table 2 shows that British export over import is 60 in the last period which is much lower than previous periods.

19

(36)

before 1995. Finally, large emerging countries Brazil and China exhibited relatively high VS shares, especially China’s VS shares are as high as those of France and the United Kingdom. The growth rate of VS shares in Brazil reached around 16% which was the third fastest growing country in our dataset.

Table 4: the VS share of total export after 1990*

1995

2000

Percentage

change between first and last period

AU

0.152 --

CA

0.315 0.308

-2.2%

DK

0.229 0.210

-8.3%

FR

0.199 0.205

3.1%

GE

0.207 0.261

26.1%

JP

0.094 0.108

14.9%

NL

0.337 0.373

10.7%

UK

0.225 0.205

-8.9%

US

0.095 0.116

22.1%

BR

0.109 0.127

16.5%

CH

0.198 0.200

1%

*Calculated from OECD input-output 2006 edition with current price.

6.2. Decomposing the changes of VS shares

After discussing the trend of VS shares of imports and exports, we will analyze our results from decomposing the changes of VS shares in this section. As shown by equation 14, there are three parts of contribution of IT industries to the changes of vertical specialization and three parts from non-IT industries. In order to offset the bias caused by not using the constant quality price of IT products, we use the constant price tables to estimate the results because they are the only available data that can proxy the constant quality price of IT products.

(37)

IT-using, and non-IT using industries according to whether an industry produces IT related products, whether an industry uses IT capital inputs more than 15% of total capital inputs. However, this industry classification in our research is difficult to implement because our dataset has no detailed industry level data on different category of capital inputs. We have to define IT and non-IT industries by using the classification of Jorgenson et al. (2005). The classification of IT and non-IT industries is in appendix 1.

For expositional convenience, we add up all the results from two-polar decomposition for both IT and non-IT industries a total ‘‘IT’’ and a total ‘‘non-IT’’ contribution20. These results are given in table 5. We see that changes in IT sectors are almost always positive and account for most of the growth in overall VS shares. Even when the changes of VS shares are negative, changes in IT sectors offsets the negative growth of VS shares. Changes in non-IT sectors are negative in almost all countries with the exception of the Netherlands where changes in non-IT sectors explain most of the increase in vertical specialization. A close look at the Dutch economy between 1970 and 1990 shows that the Dutch economy is traditionally strong in agriculture, mining and refining industries, and it was even more so before 1990. As can be seen in appendix 1, these traditional industries belong to non-IT industries and therefore they influence the change of VS shares in the Netherlands considerably21. As the fast development of service sectors (IT industries) in recent years, we conjecture that IT products must take a great share in the changes of VS shares in the Netherlands. Finally, the trend of changes in non-IT sectors is relatively close to that of total changes in VS.

Table 5 also shows that Germany and the United Kingdom experienced sharp increases in the changes of their VS shares in the 1980s. Most other countries, such as Australia, Canada, and France, exhibited fluctuating changes of their VS shares. To sum up, our results confirm that IT industries contribute the most to the changes of VS shares.

20

A detailed table containing all results is in appendix 2.

21

(38)

Table 5: Decomposition results of the changes of VS shares

Country IT Non-IT Total changes

Sub-period AU 1+2+3 1+2+3 71-76 0.0095 0.0075 0.017 76-85 0.0095 -0.0308 -0.0213 85-90 0.0004 0.0104 0.0108 71-90 0.018 -0.0115 0.0065 CA 1+2+3 1+2+3 71-76 0.0257 -0.0149 0.0108 76-80 0.0052 -0.0296 -0.0245 80-85 0.0561 -0.014 0.0422 85-90 0.0026 0.0133 0.0159 71-90 0.0999 -0.0554 0.0444 DK 1+2+3 1+2+3 76-80 -0.0102 -0.0034 -0.0136 80-85 -0.0013 -0.0076 -0.0089 85-90 0.0037 -0.0149 -0.0112 71-90 -0.0031 -0.0347 -0.0378 Fr 1+2+3 1+2+3 71-76 0.0062 -0.0061 0.00015 76-80 0.00019 -0.0043 -0.0041 80-85 0.0048 -0.0125 -0.0077 85-90 0.0134 0.0117 0.0251 71-90 0.0251 -0.0116 0.0135 GE 1+2+3 1+2+3 76-85 0.0158 -0.003 0.0128 85-90 0.0087 0.0064 0.0151 76-90 0.0245 0.0033 0.0279 JP 1+2+3 1+2+3 71-76 0.0123 -0.0383 -0.026 76-80 0.0148 -0.0574 -0.0427 80-85 0.011 -0.0481 -0.0371 85-90 0.0128 0.0097 0.0225 71-90 0.0803 -0.1636 -0.0833 NL 1+2+3 1+2+3 71-76 0.003 0.0496 0.0526 76-80 0.0062 -0.0017 0.0044 80-85 -0.0015 0.0037 0.0022 71-85 0.0078 0.0514 0.0592 UK 1+2+3 1+2+3 71-80 -0.005 0.0185 0.0136 80-85 0.0085 -0.019 -0.0105 85-90 0.1396 0.1097 0.2493 71-90 0.1401 0.1123 0.2524

(39)

7 Conclusions

The changing nature of international trade has shifted trade economists’ focus from analyzing trade in final goods to trade in intermediate goods. This paper attempts to investigate the interconnectedness of countries by measuring the verticality of international trade. The main focus has been to analyze the trend of vertical specialization of trade in a group of OECD and emerging countries as well as to examine in which industries the changes contribute the most to the changes of vertical specialization of trade.

We developed formulas to measure vertical specialization of trade through the techniques of forward and backward linkages. Our empirical results showed that VS shares of exports and export destination of imports were always increasing for almost all countries except Japan where it exhibited a fluctuation in their VS shares of exports between 1970 and 1990. Between 1990 and 2000, the VS shares of exports increased in France, Germany, Japan, the Netherlands, and the US. The VS shares of exports in Canada, Denmark, and the United Kingdom experienced small negative growth rate. Large emerging countries Brazil and China exhibited relatively high VS shares, especially China’s VS shares are as high as those of France and the United Kingdom. The growth rate of VS shares in Brazil reached around 16% which was the third fastest growing country in our dataset.

(40)

interconnectedness between countries. A possible explanation is that the separation of design and production of IT products in different countries has driven the increase in vertical specialization of IT-industries, especially in recent years. This is caused by special nature of IT industries: IT products need specialized knowledge in their design phase and the production process is rather labor intensive. Therefore, IT industries contribute the most to the changes of vertical specialization of trade.

(41)

References

Campa, J. and Goldberg, Linda S. (1997). "The Evolving External Orientation of

Manufacturing Industries: Evidence from Four Countries," NBER Working Papers 5919, National Bureau of Economic Research, Inc.

Colecchia, A. and P. Schreyer (2002), “ICT Investment and Economic Growth in the 1990s: Is the United States a Unique Case?: A Comparative Study of Nine OECD Countries”, Review of Economic Dynamics 5, 408–442.

Dietzenbacher, E. (1997) “In Vindication of the Ghosh Model: A Reinterpretation as a Price model”, Journal of Regional Science, 37, pp. 629-651.

Dietzenbacher, E. and Los, B. (1998) “Structural Decomposition Techniques: Sense and Sensitivity”, Economic Systems Research, 10, pp. 307 – 324.

-- (2002) “Externalities of R&D expenditures”, Economic Systems Research 14 , pp. 407-425.

Dietzenbacher, E. De Groot, Olaf J. and Los, B. ( 2007) “Consumption Growth Accounting”, Review of Income and Wealth, Vol. 53, No. 3, pp. 422-439.

Dornbusch R., S. Fischer, and P.A. Samuelson (1977), ‘Comparative Advantage, Trade, and Payments in a Ricardian Model with a Continuum of Goods’, The American

Economic Review, Vol 67, pp. 823 – 839.

-- (1980), ‘Heckscher-Ohlin Trade Theory with a Continuum of Goods’, The Quarterly Journal of Economics, Vol. 95, pp. 203 – 224.

Dulberger, Ellen. (1993) “Sources of Price Decline in Computer Processors: Selected Electronic Components.” In Murray Foss, Marilyn Manser, and Allan Young, eds., Price Measurements and Their Uses. Chicago: University of Chicago Press for the National Bureau of Economic Research.

Egger, H and Egger P. (2001) “Cross-Border Sourcing and Outward Processing in EU Manufacturing”, North American Journal of Economics and Finance; 12; pp. 243-256. Feenstra R.C. (2004) Advanced International Trade, Princeton University Press. Feenstra, Robert C. and Hanson, Gordon H. (1996) “Foreign Investment, Outsourcing and Relative Wages.” In R.C. Feenstra, G.M. Grossman and D.A. Irwin, eds., The Political Economy of Trade Policy: Papers in Honor of Jagdish Bhagwati. MIT Press: Cambridge, MA, pp. 89-127.

(42)

Ghosh, A., (1958) “Input-Output Approach to an Allocative System”, Economica, Vol. XXV: 58-64.

Haan, M. de (2001) “A Structural Decomposition Analysis of Pollution in the Netherlands”, Economic Systems Research, 13, pp. 181- 196.

Hummels, D. Ishii, J., and Yi, K-M. (2001) “The Nature and Growth of Vertical

Specialization in World Trade”, Journal of International Economics, Vol. 54, pp. 75 – 96. Husted S. and Melvin, M. (2001), International Economics, fifth ed., Addison Wesley Longman.

Jesse, J. Smith, D. and Souva, M. (2008), "Market Protecting Institutions and the WTO's Ability to Promote Trade" Paper presented at the annual meeting of the Southern

Political Science Association, Hotel Intercontinental, New Orleans

Jones, R. and Kierzkowski, W., Henryk (2000) “A Framework for Fragmentation”, Tinbergen Institute Discussion Paper, TI 2000-056/2, Amsterdam

Jorgenson , Dale W. Ho, Mun S. and Stiroh, Kevin J. (2005) Information Technology and the American Growth Resurgence, Productivity, Vol. 3, Cambridge, MIT Press.

Judith, D., Fung, M., K.C., and Wang, Z,, (2007) “Measuring the Vertical Specialization in Chinese Trade,” Office of Economics Working Paper No. 2007-01-A, U.S.

International Trade Commission.

Kleinert, J., (2003) “Growing Trade in Intermediate Goods: Outsourcing, Global Sourcing, or Increasing Importance of MNE Networks?” Review of International Economics, Vol. 11 Issue 3 pp. 435-575

Koopman, R., Wang Z., and Wei Shang-jin, (2008) “How Much Chinese Exports is Really Made in China – Assessing Foreign and Domestic Value-Added in Gross

Exports,” Office of Economics Working Paper No. 2008-03-B, U.S. International Trade Commission.

Krugman, P.R. (1995). “Growing World Trade: Causes and Consequences”, Brookings Papers on Economic Activity 1: pp. 327–377.

Krugman, P., Obstfeld, M., (1997) International Economics: Theory and Policy, 4th edition, W.W. Norton.

Lawton, Thomas C. & Michaels, Kevin P. (2000) “Exploding the Global Value Chain: The Changing Nature of the Global Production Structure and Implications for

Vertical Specialization of Trade and Its Changes in the OECD Countries: an Input-Output Analysis