Comparative advantage and patterns of specialization in East Asia and the United States

(1)

Comparative advantage and patterns of specialization

in East Asia and the United States

- with a focus on international fragmentation -

Final thesis

International Economics & Business

Maaike Corinne Bouwmeester August 31st, 2007

Faculty of Economics University of Groningen

Supervisor: Prof.dr. Jan Oosterhaven Co-assessor: Prof.dr. Erik Dietzenbacher

Abstract

(2)

1 Introduction

Starting in the 1960s four East Asian countries began displaying increasing growth rates that persisted for decades. The impressive growth rates of Hong Kong, South Korea, Singapore, and Taiwan did not go unnoticed.1 In articles they were alternately named the ‘Newly Industrialized Countries’ or – more imaginative – the ‘East Asian Tigers’ or the ‘Four Little Dragons’. Due to the changes in the structure of the economy and the increasing growth rates the economies started to converge towards the income levels of Europe and the United States. This growth experience was considered to be a miracle by most2, but some noted some less miraculous reasons for the impressive growth rates3. The explanation of this latter group is based on the direct relationship between the high growth rates of inputs used in production and the high growth rates of the economy. Later on other Asian countries also started to realize higher growth rates. A substantial number of countries in East Asia has become associated with high growth rates and catching up to the industrialized countries. Figure 1 shows the increase of the GDP per capita levels in a number of East Asian countries and the United States.

1

Hong Kong is actually not an independent country. In 1997 Hong Kong became a special administrative region of China, after it had been under British rule for more than 150 years. Taiwan has a separate government although it is not formally recognized by China to be independent. It is also referred to as Chinese Taipei and Republic of China. Mainland China’s full name is People’s Republic of China.

2

The World Bank (1993) uses this word to describe the development in East Asia. Other authors share this view, see for example: Lucas Jr. (1993).

3

See Krugman (1994) for this view and his references to other critics.

Figure 1: GDP per capita – PPP in 2000 constant prices

(4)

In order to take advantage of the economic growth and industrialization an increasing number of companies decided to move part of their production process to these countries either by subcontracting or by foreign direct investment. Even though the countries started to industrialize rapidly wages of industrial workers were, and still are, lower than in Europe and the United States. The lower wage rates, declining transport cost, and the combination of industrialization and export-promoting policies of Asia were generally indicated as the main reasons that made this a very attractive way of organizing the production chain. This shift of economic activity attracted much attention in the news due to serious concerns of massive job loss of low-wage workers in the industrialized countries.

The East Asian countries that started their period of high growth before the rest attracted most of this increased demand and investment. This increase in the demand for intermediate and final goods produced in these countries led to an increase in the demand for labor, which in turn led to an increase in wages. The comparative advantage of low wages that attracted the international buyers and investors decreased for these countries. Other East Asian countries were able to take the role of the first group over and started producing manufactured products that only require little skills and knowledge turning their peasants into a large unskilled manufacturing labor force. The initial growth of a few countries has initiated a restructuring of the economies of all countries in the region. Production chains of intermediate and final goods are now linked across borders forming intricate production networks to serve the European and American markets and increasingly their own region.

The relocation of production activities to other countries in order to exploit factor cost differentials has received much attention from the academic literature. Several academic studies focused on establishing models that identify the important factors and trade-offs in the decision to fragment the production process. Another group of researchers has investigated the effects of fragmentation on relative wages and the welfare gains and losses of the developed countries that fragmented their production chains. International Business researchers were increasingly interested in the organization of the industries in these countries and have thoroughly discussed the different networks and clusters present. This study tries to build a bridge between the notions of comparative advantage and specialization that are widely incorporated in economic literature and the industrial structure and production chain analyses performed in business literature.

(5)

standard trade theories. In light of this theory international fragmentation should follow the patterns of comparative advantage. A common method to indicate whether a country has a comparative advantage in a specific sector is to examine the export specialization of its sectors. These two types of specialization are compared in this study.

A third type of specialization is introduced to establish the extent of within-sector specialization. It represents the extent to which the product of an industry differs from the product produced by the same sector in other countries. When an industry produces goods that are more different from the ‘average’ produced good it can be said that this industry is specialized. Following this reasoning product differentiation is also a type of specialization. It is investigated whether this type of specialization is also in agreement with the comparative advantage of a country.

(6)

2 Literature review

In this literature review first the two standard trade theories are discussed; the Ricardian model, and the Heckscher-Ohlin model. Both models are based on the notion of comparative advantage and both models implicate that each country specializes in the production of the good they have this advantage in. Then extensions that introduce many goods are discussed. As variants of the many goods assumption, the roles of intermediate goods and product differentiation have also gained attention in the literature. In empirical work on trade theory a measure that calculates the export specialization of a country has been used as proxy for comparative advantage.

Next, the theoretical insights from a range of theories that can be related to international fragmentation are introduced. Several empirical studies of fragmentation, of which some are specifically on East Asia, will be discussed to set the context. The objective is to provide, by the combination of the theoretical framework and the empirical results, a basis for the discussion of the extent of fragmentation in East Asia.

2.1 International trade theory and comparative advantage

2.1.1 International trade models

There are two standard models of international trade.4 Both models use the comparative advantages of the countries as mechanism that determines the direction of trade. Basically a country has a comparative advantage in the production of a good if the product has the lowest relative price before any trade takes place. The Ricardian model is based on two countries that produce two products with one factor of production; labor. It is assumed that the two countries differ in their production technologies. Therefore each country has a comparative advantage in one of the products, in terms of the relative price of units of labor used in production. This comparative advantage dictates in which good each country should specialize – which good to produce. The production in excess of domestic demand will then be exported. The country obtains a higher price for its export product than before trade in the home market en it can import the other good for a lower price. In this setting total specialization will occur.

In the Heckscher-Ohlin model there are two factors of production; capital and labor. Each country has a certain amount of each of these factors which are referred to as endowments. One of the products uses capital intensively in its production process while the

4

(7)

other is labor intensive. Each country has a relative abundance of one of its endowments. The good in which the country has a comparative advantage is the one that uses the factor with which the country is relatively well endowed intensively in its production. The Heckscher-Ohlin model, like the Ricardian model stipulates that each country should specialize in the production of the good it has a comparative advantage in. In contrast to the Ricardian model specialization does not have to be complete, however factor price equalization (FPE) will take place.

Although the source of the comparative advantage is different in the two models, under free trade each country will specialize in the production of the good it has a comparative advantage in. This trade pattern is sometimes even referred to as the law of comparative advantage. Dornbusch et al. (1977) also derive this basic result when a continuum of goods is introduced instead of two in the Ricardo model. However, Dornbusch

et al. (1980) show that in the Heckscher-Ohlin model with a continuum of goods a determinate pattern of specialization is only obtained when endowments differ sufficiently to induce complete specialization. The indeterminacy of the latter model is a result of the assumption that there are more goods than factors, which is rather realistic in the real world. In the same setting Deardorff (1980) argues that even though there is a problem with the law of comparative advantage in case of individual or pairs of commodities, it is still valid when the law is reformulated in terms of averages across all commodities. In addition, he shows that although transport cost and tariffs may reverse the direction of trade for a particular good, they will not reverse the average relationship that must hold between comparative advantage and trade.

(8)

2.1.2 Intermediate goods

Feenstra and Hanson (1996) introduce intermediate goods that are used in the production of one single final good in a Heckscher-Ohlin model and find that law of comparative advantage prevails concerning these intermediate goods. Deardorff (2005a) argues that in terms of capturing comparative advantage in a Ricardian model with intermediates a new definition of comparative advantage should be adopted. He indicates a problem with the use of autarky prices in the definition. This practice misses the fact that a country may export goods with a relative high autarky price when it substitutes expensive inputs with cheaper imported inputs. His proposed solution is to compare labor per unit of value added, where value added is based on actual trading prices. However, he acknowledges that these actual prices depend crucially on barriers to trade which may cause distortions to an extent that nothing can be said about the pattern of trade even in the simplest model. His conclusion is that only an average relationship between comparative advantage and trade, identified by the correlation of trade patterns with autarky prices, seems to be robust.

Deardorff (2005b) extents the analysis of his generalization of the law of comparative advantage that is based on the weaker average relationship between comparative advantage and trade. He provides a list of assumptions that are consistent with the correlation interpretation of the law of comparative advantage. This includes, among others, multiple goods and countries, transport costs, tariffs, intermediate inputs, and the inclusion of services. Problems arise when there are domestic distortions or when increasing returns are present. Deardorff (2005b, p. 1014) concludes with the remark that ‘the concept of comparative advantage tells us something fundamentally important about the effects and determinants of international trade.’

2.1.3 Product differentiation

In theories of international trade the assumption is often made that an industry produces one standardized product. In some articles quality difference between the products produced in the same industry in the two stylized countries may be a factor in the determination of the specialization pattern. Deardorff (2005b) includes the fact that goods may be differentiated in his list of assumptions that are consistent with the correlation of comparative advantage and trade. He indicates that the appropriate way to handle these goods is by simply defining differentiated products as different goods. This is a perspective often taken in the theoretical literature. However, in the empirical literature problems arise due to the use of industry-level data. The assumption that each industry only produces one good becomes problematic. In addition the same industry in different countries is supposed to produce the same variety.

(9)

larger countries in order to investigate whether large countries export larger quantities of each good, a wider set of goods, or higher-quality goods. They find that more than 60 percent of these exports are accounted for by a wider set of goods. Although this result indicates that product variety is important in trade it does not address the issue of product variety within a sector.

Schott (2004) observes that unit values of traded products vary widely even if the product classification is very detailed. Capital and skill abundant countries and countries with more capital intensive production technologies can demand higher unit values for their products. The data he uses is consistent with specialization of countries within products. He finds that capital and skill abundant countries use their endowment advantage to produce vertically superior varieties. Basically these different varieties can be seen as different goods to the higher level of quality embodied in them. Tests of trade theory that use industry-level data are obviously problematic because much of specialization due to difference in endowments occurs at a level that was previously not incorporated in these tests. Schott (2003) finds support for the relationship between industry participation in international trade and the endowments of a country when the industry-level data are adjusted to allow for intra-industry product heterogeneity.

2.2 Export specialization and comparative advantage

According to the definition of comparative advantage, a country possesses an advantage when its good has the lowest relative price in autarky. However, in the real world these autarky prices cannot be observed. Due to the problems with estimating the autarky prices, Balassa (1965) introduced the concept of ‘revealed’ comparative advantage. The idea is that when a country has a comparative advantage in the production of a good it will specialize in its production and hence export the good – as stated by the law of comparative advantage. Establishing whether a country is specialized in the export of a specific good is relatively easy. The extent to which a country is specialized in the export of a good can then be used as a measure of comparative advantage. As the structure of exports of a country is the outcome of the specialization process due to the comparative advantage related to relative autarky prices it is referred to as the ‘revealed’ comparative advantage. The measure as introduced by Balassa is shown in equation (1).

(10)

Here X represent value of exports, the subscript j indicates the industry, the superscript A and REF refer to respectively the country for which the measure is calculated and the group of reference countries. The measure reflects the share of exports of industry j in total exports of country A relative to the share the exports of this industry have in total exports of the group of reference countries. Basically it is a measure of export specialization of a particular industry in a particular country compared to the average export specialization.

The usefulness of the measure as a proxy for the comparative advantage of a country is investigated by several authors. Vollrath (1991) discusses the efforts that have appeared in the literature to quantify a country’s comparative advantage. The measures are analyzed based on their merits and pitfalls. He concludes that one of the most satisfying measures is the Balassa index as represented in equation (1) as it represents an actual-to-expected trade ratio. The expected trade refers to a situation in which no country has a comparative advantage. A deviation of the actual trade from the expected trade can be interpreted to indicate that the country has a relative comparative advantage (ratio larger than one) or disadvantage (ratio smaller than one).

2.3 International fragmentation of production chains

2.3.1 Theoretical framework of international fragmentation

When specifically allowing for the fragmentation of a production process, it is possible that part of the production process is relocated to be performed by another company or in another country. The main reason for this relocation is the exploitation of factor cost differentials. The firm fragments its production process because the individual parts can be cheaper or more efficiently produced somewhere else. Large production costs differentials across countries provide a strong force in favor of fragmentation. However, trade and transport cost also increase significantly due to the larger distance, tariffs, and exchange rate fluctuations. Negotiating contracts may involve more problems due to cultural and institutional differences that may lead to miscomprehension and a larger extent of asymmetry in the information available to the parties. The costs of international contracts, communication, transport and trade costs that are incurred by establishing an international production chain are grouped together under the name linkage costs. The increase in these costs due to the increased trade in intermediate goods will have to be outweighed by the decrease in production cost that is expected.

(11)

one particular location. The splitting of a production process involves introducing another kind of linkage than the proximity (location and ownership) link in place before. Kuroiwa (2006, p. 89) gives the following definition of a spatial linkage: “a spatial linkage represents interdependency of industries over the space (or borders) which are linked through transactions of intermediate goods”.

Table 1 shows these two dimensions of fragmentation. The factors that determine the choice for a particular mode of fragmentation are not independent. Many issues of importance in regional fragmentation are magnified in the international setting. The mode of fragmentation chosen by a firm is the result of weighing the trade-offs of factors. The different modes of fragmentation in turn include different modes of organization. Trade relations with a second company may include long term contracts, strategic alliances or licensing. Foreign direct investment may be undertaken to acquire a minority or majority stake in a company, to enter a joint venture, or to set up new production facilities.

One of the previously integrated stages of the production process may be outsourced to another firm. Outsourcing means that an activity that was first performed in-house will now be undertaken by another company that supplies to the first firm. Alternatively the activity can also stay integrated within the firm. In this case the relocation can be both within the national borders as well as in a foreign country. For regional economics the relocation of parts of firms, or the shift of parts of production to a partner firm within the national borders can be of interest, but this study focuses on national economies and therefore the interest lies primarily with the crossing of borders. This is referred to as international fragmentation.

Table 1 – Modes of fragmentation Location →

Ownership ↓

Within national borders - Regional fragmentation -

Across national borders

- International fragmentation - Integrated in firm Domestic investment Foreign direct investment Outside of firm National outsourcing International outsourcing

(12)

transactions may rise and the entrepreneur may fail to allocate the resources most efficiently. Coase indicates that the cost of organizing and the efficiency losses will increase with an increase in the spatial distribution of the transactions organized. The resulting trade-off is the cost of using the price mechanism (contract negotiation cost) versus the increase in cost and decrease in efficiency of allocation of resources by the entrepreneur.

The theory of transaction costs as a reason for the existence of firms was elaborated by Williamson (1975). He introduced the concepts of asset specificity and the hold-up problem. When a buyer requires specific inputs that cannot be used in the production of other goods, the production of these assets becomes subject to the holdup problem. This problem involves the fact that a buyer can renegotiate the contract after the specific inputs have been produced because they can only be sold to other buyers at a lower cost due to their specificity. Due to the limited options the producer has in selling to other buyers he might not be willing to produce according to the wishes of the buyer. The incompleteness of the contracts can then lead to an inefficient outcome. To reduce these inefficiencies the buyer can decide to produce the specific inputs within the firm.

At the international level the issue of transaction cost may also result in the decision to keep activities within the firm. In order to still profit from production cost differentials there has been a large increase the number of multinational enterprises (MNEs), which are firms that have a presence in multiple countries. The framework to analyze MNEs by Dunning (1977) focuses on three characteristics of these firms: the ownership advantage (O), the location advantage (L) and the internalization decision (I). The first relates to the presence of an identifiable competitive advantage owned by the MNE. Next, there has to be a location where this ownership advantage can be exploited (different from the location of the firm and by physically being present there). The third characteristic is the decision to internalize the fragmented activity instead of exporting or licensing it.

Fragmentation may occur due to the fact that it is recognized that another company may have a larger ownership advantage and may produce the desired input more efficiently. However, the company itself has an ownership advantage in the production of a certain good it may fragment in order to exploit production cost differentials in another country by licensing its production technology or establishing a foreign subsidiary. The location advantage in the OLI framework is the characteristic that is directly related to the comparative advantage of nations as considered in trade theory. The location advantage results in fragmentation some activity (in the broadest sense of the word) can be undertaken somewhere else less costly than in present situation. In case of an MNE the activity that is relocated will be kept internal to the firm.

(13)

across borders. Standardization of a product makes it possible to shift its production to a low-wage foreign location. Due to the product cycle production will first be fragmented abroad within firm boundaries, and only later to independent foreign firms. Incomplete contracts create hold-up problems which give rise to suboptimal relationship-specific investments. The assumption is made that in the North contracts can be better enforced. The product development manager faces a trade-off between the lower costs of manufacturing in the South and the higher incomplete-contracting distortions. Contracting issues are also investigated by Antràs and Helpman (2006). In their research better contracting institutions in the developing countries raise the prevalence of off-shoring, but may reduce the relative prevalence of FDI or foreign outsourcing. Improvements in the contractibility of inputs controlled by final good producers have different effects than improvements in the contractibility of inputs controlled by suppliers.

Jones and Kierzkowski (1990) stress the role of services in the fragmentation of a business. The production of goods by the fragmented production blocks is coordinated by the service links which include, among coordination, administration, transportation, and financial services. They show that the growth of the output of a firm, the presence of increasing returns and the advantages of specialization of factors within the firm are forces towards fragmentation of the production process. Due to technological advances in the provision of services the relative costs of international coordination and communication have fallen significantly. Widespread implementation of the so-called just-in-time delivery method has contributed to holding down production cost. Improvements in technology have led to a larger degree of reliability of deliveries, more efficient information management, and synchronization of output streams of intermediates. This fall in the relative prices of many services combined with the presence of the fragmentation forces explains the increase in fragmentation. The specific locations of the fragmented activities can be explained by the international comparison of factor prices and productivities.

(14)

2.3.2 Empirical evidence of international fragmentation

Over the past few years much research has been done to describe and analyze the importance of production sharing in the world, and more specific in East Asia. This section discusses the main results and conclusions of the empirical literature on fragmentation.

Yeats (1998) concludes on the basis of trade data of parts and components combined with additional data that global production sharing comprises at least 30 percent of world trade in manufactured products5. He also shows that trade in components and parts is growing faster than trade in other products, indicating a growing importance of production sharing. In the same line Ng and Yeats (1999) continue to investigate the importance of production sharing focusing but now East Asia. They find that components trade constitutes 20 percent of East Asian exports of manufacturers. Imports of components are expanding faster in East Asia than in Europe or the United States and global exports of components grew faster than any other product group in East Asia. Next to presenting evidence of production sharing, the authors also investigate why it expanded. Revealed comparative advantage indices are used to show whether a country has a comparative advantage in the production or in the assembly of a product where the latter is calculated using import statistics for components. Ng and Yeats (1999) find that Japan, Singapore, and Taiwan increased their specialization in the manufacture of components whereas the labor-intensive assembly operations tended to relocate to lower-wage East Asian countries. Indonesia, Malaysia, and Thailand are the most important assembly countries in East Asia, however none of the countries has developed its domestic assembly operations to the extent Mexico has.

Ng and Yeats (2003) analyze the major trade trends in East Asia in order to assess the economic integration in the region. During the last quarter of the 20th century East Asia’s share of global exports increased more than three-fold and intra-regional exports as share of world trade expanded more than six-fold. Of the total intra-regional exports, 80 percent is exported by the five largest exporters – China, Korea, Malaysia, Singapore and Taiwan. The authors also find that seven out of the 14 considered East Asian countries depend for 50 percent or more of their total imports on regional suppliers. Almost half of all goods traded within the region belong to the machinery and transport equipment group. Of this electronic products and parts and components are of major importance. Trade in parts and components accounts for 20 percent of all intra-regional trade in manufactured goods. Japan is an important node in the production sharing network of East-Asia. The strategic trade policies of several countries focus on trade in parts and components as means of entering markets for high technology or high-skilled products. It is found that the production of parts and

5

(15)

components is situated in higher-income countries as it uses capital intensively. The lower-wage countries, like Indonesia and Thailand, are specialized in assembly which is more labor intensive.

Hummels et al. (2001) investigate vertical specialization using a narrowly defined concept. Their measure of vertical specialization (VS) is defined as the share of imported intermediates in production multiplied by exports. It represents the foreign value-added embodied in exports. They aggregate all the sector VS values and divide by the aggregated exports of all sectors to arrive at an export-weighted average of the sector VS export shares, see equation (2).

VS share of total exports k i ki _i ki ki

k _i ki k ki

VS

X

VS

X





 





≡

=





 









 









∑

_∑

∑

Hummels et al. (2001) use national input-output tables for 12 OECD countries and Taiwan. The domestic input coefficient matrix of the input-output table can be integrated in the formula to account for all imported inputs that first circulate in the domestic economy before they are embodied in exports. The resulting formula in matrix notation is given in equation (3).

VS share of total exports VSk /Xk

[

]

/Xk

1 X A I uAM D − − = ≡

Here u is a 1×nunit vector, AMis the

n ×

n

imported inputs coefficient matrix, I is the identity matrix, AD is the

n ×

n

domestic input coefficient matrix, X is a n×1vector of exports, X_kis total country exports, and n is the number of sectors. Hummels et al. (2001) use this formula as their main measure of vertical specialization.

They find that smaller countries have higher VS shares and that the VS share has increased for all countries but Japan between 1968 and 1990. The growth of the VS share is then decomposed into changes in sector VS intensity and changes in the sector composition of overall exports. The main findings are that variation in sector VS intensity accounts for almost all of the overall VS share variation over time and across countries, and that in most of the countries the chemicals and machinery sectors are the most important contributors to the growth of the VS share over time.

Kimura and Ando (2005) identify two dimensions along which fragmentation takes place. The first dimension concerns the geographical distance, or location component of fragmentation. Fragmentation can occur both in a country as well as across the border. The second dimension is based on the controllability of a firm, which indicates whether the firm (2)

(16)

remains the owner of the part that is relocated, or whether it is outsourced to another company. Along both dimension a cost is involved of splitting up the production process which is called service link cost. The advantage of fragmentation is the reduction in production costs due to locational advantages and ownership advantages.

Their empirical evidence is based on trade data and FDI data for East Asia, and on micro-data of Japanese corporate firms. They show that intra-regional trade in East Asia has increased and that there has been an increase in machinery parts and components trade. The combination of these two findings leads them to conclude that vertical production networks have been developed in East Asia. Their micro-data analysis confirms that Japanese corporate firms have contributed to the development of production networks in the region.

The importance of vertical international production sharing in East Asia is confirmed by Ando (2006). He finds that vertical intra-industry trade, which is defined as intra-industry trade where the goods have differences in unit-prices, has increased sharply. The share of vertical intra-industry trade in machinery parts and components has increased even more rapidly than that of trade in machinery products. This suggests the rapid development of cross-border production sharing in East Asia.

Athukorala and Yamashita (2006) show that the trade in parts and components is more important in East Asia than in North America and Europe. However, they also indicate that this has not reduced the dependence of East Asia on the global economy. They explicitly discuss the fact that using reported trade data leads to double counting. This can imply making wrong inferences about the relative importance of the region relative to the rest of the world. After augmenting a gravity equation with relevant variables Athukorala and Yamashita conclude that production sharing is more important in East Asia because of the more favorable policy setting with regards to foreign direct investment, the agglomeration benefits arising from an early start in production sharing, and wage differentials between the countries.

The only paper that investigates the relationship between the extent and the determinants of fragmentation is written by Jones, Kierzkowski and Lurong (2005). They find support for their hypotheses that growth of the world economy increases the degree of fragmentation in East Asia and that trade in parts and components increases as services become cheaper and more readily available.

2.3.3 Input-output empirical evidence of East Asian fragmentation

(17)

consumption goods are much lower. These dependency ratios imply that the trade in intermediate goods was the main purpose of trade.

Next Kuroiwa investigates the level of intra-industry trade by calculating the Grubel-Lloyd index. The results indicate that there has been a rise in intra-industry trade. Furthermore the analysis of backward and forward linkages indicates the presence of both linkages in the electric machinery sector indicating the organization of production in a network. Simple cluster analysis of the machinery industry shows several independent clusters for 1975 and 1985 whereas the clusters in 1995 show more interaction, especially between electric machinery and precision machinery.

In the discussion paper of Meng et al. (2006) the structure of the economy is analyzed in light of the development of the countries and the alleged change in the structure from a primary to secondary to tertiary based economy along the development path of a country. Japan and the United States both feature a service sector with a large share in the economy. Korea, Taiwan, and Singapore are found to specialize both in heavy industry and chemical industry as well as the service sector. Indonesia, China, The Philippines, and Thailand continue to have a relatively important agricultural sector.

Hasebe and Shrestha (2006a) analyze the 1985, 1990, and 1995 input-output tables of East Asia and an estimated table for 2000. In order to study economic integration in the region they extend the intermediate input method, which includes both endogenous and exogenous countries as well as the Leontief inverse, to include the size effect of final demand and exports. They find that there exists a diverse interdependence structure, that the average self-dependence has declined although dependence on other countries in the region has increased, and that the dependence on Japan, the United States and the rest of the world is relatively high. In a working paper Hasebe and Shrestha (2006b) extend the analysis to include the 2000 input-output table and now use the table with maximum disaggregated production sectors. They conclude from the interdependence analysis that the progress of economic integration in the region is justified.

(18)

3 Research questions & hypotheses

3.1 The extent of international fragmentation

Empirical evidence of international fragmentation is not widespread and in general focuses only on direct linkages. The use of international input-output tables offers the opportunity to include domestic as well as international indirect linkages for the countries included in the table. In order to study international fragmentation this study first gives an overview of the extent of international fragmentation in East Asia. (Research question 1)

Case studies and empirical evidence suggest that there is a truth to the public sentiment that international fragmentation has increased over the years. Therefore it is expected that the international fragmentation measure used in this study will also indicate that international fragmentation has increased over the period 1990 to 2000. Next to the overall extent of international fragmentation two possible sources of variations are investigated. In general it is thought that the extent of international fragmentation is related to the economic size and development level of the country. The countries can be ranked according to the total value of output they produce. Countries that produce more output are expected to be more self-sufficient than smaller countries, thus the extent of international fragmentation will be negatively related to the total output of a country. In addition, the different countries in East Asia can be divided into groups that have different levels of development. Countries are allocated to the same group when each of them started to experience high growth rates at approximately the same point in time. The first countries to display high growth rates will have developed a substantial home market and be more self-sufficient. The countries that have started growing rapidly only recently are expected to still have limited international fragmentation. The relationship between development and international fragmentation is expected to be characterized by an inverse U shape. Finally, the contribution of specific sectors to the overall extent of fragmentation will be studied. As Asian is know for its production of electronic devices it is expected that especially this sector will be one of the major contributors to the extent of fragmentation.

3.2 International fragmentation and export specialization

(19)

to trade theory countries gain by specializing in the production of goods in which they have a comparative advantage. When fragmented production processes are considered there are even more opportunities for a country to specialize in the production of a certain input, or in a certain part of the production chain. This study will investigate the relationship between the international fragmentation of a sector and the revealed comparative advantage or

disadvantage of the sector. (Research question 2)

Theory indicates that both types of specialization; the vertical specialization represented by international fragmentation and the export specialization as measured by the revealed comparative advantage, occur in line with existing comparative advantages of the countries. The comparative advantages of a country can be relatively better exploited in sectors where the factor that enjoys that comparative advantage, for example labour, is relatively important in the production process. This sector will have a production cost advantage compared to the same sector in other countries without the comparative advantage. Exports of this particular sector will thus be relatively higher than the exports of the same sector in other countries. The same production cost advantage that leads to export specialization is also the force behind a larger extent of international fragmentation. The cost advantage will focus attention of the sector to producing that particular product. The sector will specialize in the production of this product and will outsource the required inputs to countries that enjoy a comparative advantage in the production of these inputs. The hypothesis accompanying research question 2 is that sectors that display high levels of international fragmentation are expected to be the same sectors that are characterized by export specialization.

(20)

important there is a larger chance that more stages in different countries are also involved in the production chain.

The place of the sector in the international production chain relative to the place of the same sector in the other countries gives an indication whether this sector is situated more towards the beginning or the end of the international production chain. If this position is relatively far from the average position of the same sector in all countries it can be said that due to the different position in the international production chain the industry is likely to produce different goods than the same sector in the other countries. The further away the position is, the more differentiated the goods are, and the more specialized this particular sector is. This type of specialization, which is based on differentiation, is also likely to be related to the comparative advantage of a country. This study will analyze to what extent product differentiation is related to the extent of fragmentation and the extent of (export)

specialization. (Research question 3)

(21)

4 Methods

4.1 The extent of international fragmentation

In this section the measure of international fragmentation used in this study is introduced. First the concept of international fragmentation is explained and a corresponding simple measure is presented. This measure is extended by combining it with the input-output model. The availability of bilateral linkages in the international input-output table that is used in this study gives room to extent the measure even further.

4.1.1 The concept of international fragmentation

International fragmentation has been studied in several ways. Most studies look at trade in intermediate goods. These data imply that there are at least two countries involved in the production of a good. Whether these international linkages are just bilateral or involve more countries cannot be examined on the basis of macro-economic trade data. Hummels et al. (2001) introduce a more specific definition of what they call vertical specialization. Their concept of vertical specialization is equal to our concept of international fragmentation as the vertical specialization is considered at the country level. International fragmentation will be used in the rest of the study to refer to the phenomenon that countries, instead of managing a complete production chain domestically, are more and more part of international production chains that span multiple countries. Instead of considering bilateral trade of intermediate goods, the concept of international fragmentation also requires that part of the production of a sector, which is produced using imported inputs, is exported to other countries. The minimum requirements to consider a production chain as international fragmented are: 1) a sector imports intermediate goods that are used in production, and 2) some of the products produced by this sector are exported.

The measure of international fragmentation applied in this study is closely related to the measure of vertical specialization as introduced by Hummels et al. (2001). In other studies it is also referred to as the import content of exports. Basically it represents the value of imports that are embodied in a product that is exported. As there is no separate account of the inputs that go solely into export goods in input-output tables (or trade data) it is assumed that the share of imports in total output is equal to the share of imports in exports. This is reasonable as long as the same technology is used to produce for domestic demand as well as for exports purposes. See equation (4).

(22)

Where IF stands for international fragmentation, M is imports, X is total inputs (which equals total outputs), and E is exports. The indices s and i refer to the country and sector respectively. The total value of imports that is embodied in a country’s total exports is a simple summation of the sector IF values.

s s

i i

IF

=

∑

IF

In order to compare these values across countries they are normalized with respect to the total value of exports. This measure is referred to as the international fragmentation share of a country; IFSs. See equation (6).

s i s i s i i

IF

IFS

E

=

∑

Returning to the sector level, the value of imports that is embodied in exports, IFi s

, is not divided by the value of sector exports as this would divide out the export value altogether and only leave the import-to-output ratio. First aggregating the embodied values and then dividing by the total exports is actually equal to an export-weighted summation of import-to-output ratios. See equation (7).

s s s i s i i i s s s i i i i i i

IF

E

M

IFS

E

X









=

_

×

_









∑

The part between brackets represents the IF share of a single sector. Here again, as was the case with the international fragmentation value in equations (4) and (5), the country’s international fragmentation share is a summation of the sector’s international fragmentation shares. From the part between brackets of (7) it can be seen that this summation represents an export weighted average of the import share in output per sector. At the sector level the measure represents per unit of total exports of the country what the value of imports is that go into producing exports. This can also be interpreted as the value of imports that is required for producing the exports of the sector per unit of total exported products.

4.1.2 Introducing the input-output model

In the input-output framework next to direct linkages, all indirect linkages can also be taken into account. For example, extra demand for products of industry A will be met by an increase in A’s production. However, A uses inputs from industry B, of which the required output will (5)

(6)

(23)

therefore also increase. If B then uses products of A to produce the inputs there is an indirect effect on industry A next to the direct increase in production to meet the new final demand. Shown in a matrix equation this system is represented by equation (8).

x = Ax + f

In this equation x is the vector of total outputs. The elements aij of matrix A represent the

input shares of each industry i in the output of each industry j. The vector f represents final demand. Solving this system leads for x results in equation (9).

x = (I – A)-1_f

Where the identity matrix I is of the same size as A. The final demand vector or matrix is multiplied by (I – A)-1 to obtain the required output. The matrix (I – A)-1 is often called the Leontief inverse. Each element lij of this matrix represents the extra output that is needed from

industry i in order to fulfil one extra unit of final demand for the product of industry j.

Going back to the international fragmentation share of a country equation (10) shows it in matrix definition making use of the Leontief inverse.

s

IFS = uAM(I-AD)-1e

In this equation u represents a summation row vector, AM represents the import matrix of which each element

a

_ijrs represents the share of imported inputs produced by industry i in country r used in the goods produced by industry j in country s. The Leontief inverse (I-AD)-1 represents the total, direct and indirect linkages, between the domestic industries. The vector e is a column vector of exports shares, which represents the value shares of products that are exported to satisfy intermediate and final foreign demand. The result of this multiplication is a single scalar that is similar to the result of equation (6), although (10) will result in larger values due to the inclusion of domestic indirect linkages between industries.

In order to obtain IF shares per sector the e vector has to be transformed to a diagonal matrix Ê. This matrix has the export shares of the sectors on the main diagonal. The result is that the sector IF shares are not directly summed up through the matrix calculations, but that the outcome is a row vector of 63 values that represents all the individual sectors.

(8)

(9)

(24)

4.1.3 International input-output tables

The above matrix calculations are all based on national or country input-output tables of domestic and imported intermediate inputs. In case of an international input-output table, indirect linkages between each of the endogenous countries are taken into account in the calculation of the international fragmentation measure. The following equation shows the measure based on an international input-output model. This measure is related to the measure used by Shrestha (2007). The difference in calculation originates from a broader definition of international fragmentation that we employ. The structure of the inter-country IF equation (11) resembles the structure of the national equation as given in (10).

IFS = AT_(I-AR₎-1_Ê

However, AT now represents all inputs produced in the endogenous and exogenous countries including domestically produced inputs. The number of rows of this matrix is r times the number of sectors, where r runs from 1 to 12 in case of the 1990 and 1995 tables and from 1 to 13 in the 2000 table.6 The number of columns of AT is s times the number of sectors, where

s runs from 1 to 10 – representing the endogenous countries. AT is a rectangular matrix of size 756 by 630 for the 1990 and 1995 tables and 819 by 630 for the 2000 Asian input-output table.

The matrix (I-AR)-1 includes all endogenous countries and instead of the domestic AD matrix it represents the A matrix of the region, which includes all the indirect linkages between any of the industries in any of the endogenous countries. The export vector is replaced by the diagonal export matrix Ê. This matrix contains the diagonalized column vectors representing the export shares of the sectors of each of the individual countries. The export shares cover both the exports to the exogenous countries and the exports to the endogenous countries. This presumes the exports to the endogenous countries to be exogenous, which is inconsistent with the model, but justified when regarding the IFS measure as a picture taken at a particular point in time.

The elements of the resulting IFS matrix are accompanied by four indices;

ifs

_ijrs. This matrix has the exact same dimensions as the AT matrix. Each element gives the value of inputs produced by i in country r that are embodied in the exports of industry j in country s per unit of total exports of country s. This measure inherently results in some double counting, because part of the endogenous part of exports is double-counted in the imported

6

In the 1990 and 1995 table there are two exogenous regions: Hong Kong and the ‘Rest of the World’. In the 2000 table Europe is added as the third exogenous region.

(25)

inputs values of IFS. Also there is an implicit assumption that time is not a relevant dimension. The (I-AR)-1 matrix represents all extra production required when all effects of extra demand have rippled through the system. Pre-multiplying this matrix by the input coefficient matrix AT is done to obtain the input that are needed for the extra production as represented by the elements of (I-AR)-1. However, different products have production chains of different lengths. The measure as introduced in (11) does not take the multiple rounds of production in account but assumes instantaneous extra production as given by the (I-AR)-1 matrix.

To arrive at the IFSs of a particular country s, the IFS matrix is split into the separate country block vectors (resulting in 10 block vectors) and each is aggregated over all countries

r. This aggregated matrix for country s includes the matrix of purely domestic inputs, which are represented as AssLssÊs. This particular sub-matrix is part of the IFSss sub-matrix, which also includes the AsrLrsÊs sub-matrices, where r ≠ s. These other sub-matrices represent inputs that are imported from country r to be used in the production process by s. The inputs require in turn inputs which are sourced in country s. In our opinion these also represent a fragmented production chain. Shrestha (2007) deducts the complete IFSss sub-matrix from the IFSs_{matrix. In this study only the A}ss_Lss_Ês_{is deducted from the IFS}s_{matrix. The resulting} matrix with elements

ifs

_ijs_{, where ◦ represents the summation of all countries excluding s, is} aggregated into one row to arrive at

ifs

_•s_j_{. Here the • is a summation of all sectors including j.} This vector represents, as in the national case, the imported inputs embodied in the exports of sector j in country s. This measure results in larger values than the measure based on the national table, because the international linkages are included.

The sector disaggregation of the international fragmentation values gives the opportunity to analyze which sectors contribute most to the country IFSs values, in terms of the relative size of the

ifs

_•s_j_{of a particular sector. Additionally the} s

j

ifs

_• _{values will be} compared per sector over the countries.

4.2 Export specialization

(26)

1 1 1 1 s n n s A A A REF REF i i i i i i REF REF i

ARCA

X

= = = =

=

∑

−

∑

∑ ∑

In this equation s equals the total number of sectors, and n the total number of endogenous countries. Country A has a revealed comparative advantage in industry i if the export share of industry i in total exports is larger than the average share the sector has in the total exports of the reference countries. In this case it can be said that A is specialized in industry i. If an industry does not have a comparative advantage or disadvantage the value of ARCA is zero. A comparative disadvantage results in values between –1 and 0, a comparative advantage results in values between 0 and 1. The country the measure is calculated for is also included in the group of reference countries due to the limited amount of countries. Comparability is better as the value that is deducted from a certain sector’s export share is constant for each country. In addition, in case of a limited amount of countries the individual country values can severely influence the average.

The revealed comparative advantage is measured as the deviation of the sector’s export share from the average export share of the same sector in the reference countries. For the IF shares a corresponding calculation will be made in order to be able to interpret the magnitude of the sector IF shares relatively to the same sector in the reference countries. See equation (13). Here RIFS refers to relative international fragmentation share.

1 1 n n s s REF REF i i i i REF REF

RIFS

IFS

IF

E

= =

=

−

∑

The correlation between the ARCA and the sector IF shares will be analyzed in order to see whether there is agreement between the two different kinds of specialization. It is expected that they are related because trade theory indicates that each country will specialize in the good it has a relative production cost advantage in.

4.3 Product differentiation

In an input-output model an element lij of the Leontief inverse represents the extra output that

is needed from industry i in order to fulfil one extra unit of final demand for the product of industry j. (See also equations (8) and (9) and their accompanying texts.) These values are also referred to as backward linkage, because they show the additional production of inputs, which are earlier or backward stages in the production chain. The elements are also often referred to as multipliers, because they show all the extra output necessary for one more unit (12)

(27)

of final demand, including the production of the good itself and all the intermediate inputs included in each of the inputs used in the production chain.

Aggregating these values by column into one row, an element l•j of this row vector

shows the total extra output from all industries that is needed to fulfil one extra unit of final demand for the product of industry j. The magnitude of the backward multiplier indicates the generation of extra output in earlier stages of the production chain and can be interpreted as an indication of the length of the production chain before the inputs reach sector j.

Comparably the Gosh forward linkages can be calculated. The Gosh supply model and its solution are given in equation (14).

x = xB + v’ ⇒ x = (I – B)-1v’

The B matrix is analogous to the A matrix, except that the elements represent output shares, i.e. all values in a row of the inter-sector transaction flows are divided by total output instead of all the values in a column as is the case for the A matrix.

The interpretation of the elements gij of the Gosh inverse matrix has been subject of a

discussion on the implied causality. Early interpretations assert that these elements show the extra production in sector j if there is one unit of extra value added available (one unit extra spend on wages) in sector i. In other words, due to the increase in the value added in sector i the sector can produce more and output sold to each sector j therefore increases by the ratio of outputs sold to sector j by i over total outputs sold by i. In turn, sector j increases its output by the same ratio, because due to more inputs of i that have become available sector j also increases the procurement of all other sectors of which it needs output to produce its product. Aggregating the Gosh inverse matrix over the rows into one single column, the values gi•

allegedly show the total output in all sectors that is generated by increasing value added in sector i by one unit (Miller and Blair, 1985). There are problems with the implied causality and its economic interpretation in this quantity version of the Gosh model. However, Oosterhaven (1988) argues that usage of the Gosh model for descriptive purposes is acceptable. In this study the forward linkage is used as an indicator of the length of the production chain after a product is sold by sector i.

The total length of the production chain in which industry j participates is calculated as the summation of the l•j and gj• elements of the backward and forward linkage vectors. The

backward linkage is then divided by the total length of the production chain to get an indication of the position of the sector. If the backward linkage is exactly as large as the forward linkage the result of the calculation is ½. If the backward linkage is larger than the forward linkage the measure will be larger than ½, and if the forward linkage is larger the measure will be smaller than ½.

(28)

(15) In correspondence to the export specialization measure, the average position of the same sector in all ten countries will be deducted from the country specific value of that sector in order to find the deviation of the country from the average position. See equation (15). In this measure • refers to all countries including the country that the measure is calculated for.

DIFF is used to reflect the fact that the measure gives an indication of the product differentiation within a sector.

(

)

1 1

S _n _n

j

S REF REF REF

i S S j j i REF REF j i

l

DIFF

l

g

l

g

• • • • • • • • • • = = • •

=

−

+

∑

(29)

5 Data description

The analysis in this study is performed based on the 1990, 1995, and 2000 Asian international input-output tables (IDE 1998, 2001, 2006). These tables contain sector specific information on the inputs and outputs of more than seventy sectors over ten endogenous countries; Indonesia, Malaysia, Philippines, Singapore, Thailand, China, Taiwan, Korea, Japan, and the United States. The complete group of countries is referred to as region. In order to be able to compare individual industries over years, sector detail is reduced to 63 separate sectors.

The unclassified sector has been taken out of the endogenous part of the table for two reasons. The practical reason is the problem with this sector in the 2000 table for Malaysia. Total inputs of the unclassified sector in Malaysia are ascribed to only itself, leading to the computational problem that the matrix cannot be inverted. The conceptual reason for leaving out the unclassified sector is the lack of real information. Each product of unknown origin or destination is allocated to the unclassified sector implying a structural relationship between these products that does not exist.

(30)

Table 2: Total output of country in billion US dollars and % of total output of the region

1990 1995 2000

output % output % output %

United States 9926 54.2 13456 47.9 17945 54.6 Japan 5897 32.2 9746 34.7 8682 26.4 China 883 4.8 1875 6.7 3111 9.5 Korea 576 3.1 1058 3.8 1200 3.7 Taiwan 363 2.0 566 2.0 658 2.0 Indonesia 202 1.1 442 1.6 321 1.0 Thailand 179 1.0 356 1.3 301 0.9 Singapore 96 0.5 223 0.8 253 0.8 Malaysia 93 0.5 205 0.7 241 0.7 Philippines 86 0.5 142 0.5 153 0.5 Region* 18300 100.0 28069 100.0 32864 100.0 - without U.S. 8374 45.8 14613 52.1 14919 45.4

- without U.S. & Japan 2477 13.5 4867 17.3 6237 19.0

* Region refers to the ten countries combined

Calculations on basis of the 64 sector tables, including the unclassified sector Source: Asian input-output tables of 1990, 1995, and 2000 + author’s calculations

In Table 3 the growth rates of output are represented. Over the total period 1990 – 2000 all countries show positive growth rates. When investigating the separate periods some negative growth rates appear. For Indonesia and Thailand these can be attributed to the Asian crises of 1997 that has hit these countries hard. Japan also experiences a negative growth rate for this period, which can be ascribed to a prolonged recession of the economy of the country.

Table 3: Total output growth rates growth % 1990–1995 growth % 1995–2000 growth % 1990–2000 United States 35.6 33.4 80.8 Japan 65.3 -10.9 47.2 China 112.4 66.0 252.5 Korea 83.7 13.4 108.3 Taiwan 55.9 16.2 81.2 Indonesia 119.2 -27.4 59.1 Thailand 98.7 -15.4 68.0 Singapore 132.8 13.5 164.2 Malaysia 120.3 17.4 158.7 Philippines 65.1 7.5 77.6

Comparative advantage and patterns of specialization in East Asia and the United States

Comparative advantage and patterns of specialization

in East Asia and the United States

Table of contents

1

Introduction

2

Literature review

VS

VS

X

VS

X

X

X

X





 





≡

=

=





 









 









∑

∑

∑

[

]

n ×

n

n ×

n

3

Research questions & hypotheses

4

Methods

IF

=

∑

IF

IF

IFS

E

=

∑

∑

IF

E

M

IFS

E

E

X









=

=



×











∑

_∑

_

_