International Fragmentation of Production
(dedicated to my dearest parents)
International Fragmentation of Productio...
14 downloads
501 Views
2MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
International Fragmentation of Production
(dedicated to my dearest parents)
International Fragmentation of Production The Impact of Outsourcing on the Japanese Economy
Nobuaki Yamashita La Trobe University, Australia
Edward Elgar Cheltenham, UK • Northampton, MA, USA
© Nobuaki Yamashita 2010 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA
A catalogue record for this book is available from the British Library Library of Congress Control Number: 2009941411
ISBN 978 1 84844 637 3
04
Printed and bound by MPG Books Group, UK
Contents Abbreviations and acronyms Preface Acknowledgements
vii ix xi
1
Introduction 1.1 The issues 1.2 Structure and preview
2
International fragmentation of production: a survey of theory and the measurement issue 2.1 Introduction 2.2 International fragmentation of production: an introduction 2.3 A survey of theory 2.4 The measurement issues 2.5 Concluding remarks
5 5 6 9 21 28
Production fragmentation and trade patterns in Japanese manufacturing 3.1 Introduction 3.2 An overview of trade patterns 3.3 Trends and patterns of fragmentation trade 3.4 Direction of fragmentation trade 3.5 Concluding remarks
31 31 31 34 43 48
Determinants of fragmentation trade 4.1 Introduction 4.2 Empirical evidence 4.3 The model specification 4.4 Measurement, data and estimation procedure 4.5 Results 4.6 Conclusion Appendix 4.1 Data used and variables construction
50 50 51 53 55 56 61 62
3
4
1 1 2
v
vi
5
International fragmentation of production
Structural transformation and labour market adjustment in Japanese manufacturing 5.1 Introduction 5.2 Development of Japanese manufacturing and labour market 5.3 The trends and patterns of skill upgrading in Japanese and US manufacturing 5.4 Conclusion Appendix 5.1 Measure of worker skills intensity Appendix 5.2 The Census of Manufactures (CM) and the Basic Survey on Wage Structure (BSWS)
6
The impact of production fragmentation on skill upgrading 6.1 Introduction 6.2 The skill upgrading effect of production fragmentation 6.3 Econometrics analysis 6.4 Data and econometric methodology 6.5 Results 6.6 Conclusion Appendix 6.1 Country composition Appendix 6.2 Japan industrial productivity 2006 database Appendix 6.3 Re-estimation of Feenstra and Hanson (1999)
7
Overseas operations and home employment of Japanese multinational enterprises 7.1 Introduction 7.2 Patterns and trends of the home and overseas operations of MNEs 7.3 The effect of the overseas operations of MNEs on domestic operations 7.4 The analytical framework 7.5 Estimation method 7.6 Results 7.7 Concluding remarks Appendix 7.1 Construction of the matched panel data Appendix 7.2 METI surveys
8
Conclusion 8.1 Findings 8.2 Policy implications
References Index
70 70 71 75 80 80 84 86 86 87 92 96 98 102 103 104 105
110 110 111 118 122 125 127 138 139 141 147 148 149 151 163
Abbreviations and acronyms AFTA ANU ASEAN ASM BEA BEC BOJ BSWS CES CIF CM CPI CRS CV FDI FIATA
FOB GDP GLS GM GMM HS ILO INSEE I-O IPT IRS ISCO ISIC IVs JIP JSIC
ASEAN Free Trade Area The Australian National University Association of South East Asian Nations Annual Survey of Manufactures Bureau of Economic Analysis Broad Economic Category Bank of Japan Basic Survey on Wage Structure constant elasticity of substitution cost, insurance and freight Census of Manufactures consumer price index constant return to scale coefficient of variation foreign direct investment Fédération Internationale des Associations de Transitaires et Assimilés (International Federation of Freight Forwarders Association) free on board gross domestic product generalized least squares General Motors generalized method of moments harmonized system International Labour Organization Institut National de la Statistique et des Études Économiques input-output inward processing trade increasing returns to scale International Standard Classification of Occupations International Standard Industry Classification instrument variables Japan Industrial Productivity Japan Standard Industrial Classification vii
viii
LSDV METI MNE NACE
International fragmentation of production
least square dummy variable Ministry of Economy, Trade and Industry multinational enterprise Nomenclature statistique des Activites economiques dans la Communaute Europeenne NAFTA North American Free Trade Agreement NBER National Bureau of Economic Research NIEs newly industrialized economies NTBs non-tariff barriers OAP offshore assembly programme OECD Organisation for Economic Co-operation and Development OLS ordinary least squares OPT outward processing trade R&D research and development RER real exchange rate RIETI Research Institute of Economy, Trade and Industry SITC Standard International Trade Classification TRAINS trade analysis and information system ULC unit labour cost UNCTAD United Nations Conference on Trade and Development UNIDO United Nations Industrial Development Organization USITC US International Trade Commission WLS weighted least squares WPI wholesale price index
Preface This book examines the patterns, determinants and labour market implications of international fragmentation of production – the crossborder splitting of the production process within vertically integrated manufacturing industries – focusing on the experience of Japanese manufacturing. Despite the rapidly growing importance of production fragmentation in Japanese manufacturing, there is little systematic empirical analyses of this phenomenon and its implications for labour market performance and adjustments. The few available studies have focused narrowly on the emerging patterns and intensity of fragmentation-based trade. In the absence of such systematic analyses of the Japanese case, debates on the implications of production fragmentation for manufacturing performance and labour market adjustments rely largely on studies of the US manufacturing experience, though it is well known that there are important differences between Japanese and US manufacturing industries. As in other developed economies, there is a widespread perception in Japan that production fragmentation leads to losses in Japanese manufacturing employment. This book aims to explore the Japanese case in depth, compare it with the US experience and to draw out the main trade and labour market implications of production fragmentation from this comparative perspective. The book begins with a comprehensive interpretative survey of the theory of production fragmentation in order to place the empirical analysis in context. The empirical analysis, which forms the core of the book, has three main components. First, it examines patterns and determinants of cross-border trade in parts and components (‘fragmentation trade’) by using trade data compiled on the basis of a new commodity list of parts and components of manufacturing trade over the period 1988–2005. The second part comprises an analysis of the effects of fragmentation trade on the skills structure of manufacturing employment in Japan using a panel data set covering 52 Japanese manufacturing industries over the period 1980–2000. The analysis makes use of a newly constructed measure of the intensity of fragmentation trade which enables a richer analysis to be conducted. It also allows for differential effects of the geographical location of fragmentation trade on the skills mix of employment, an important ix
x
International fragmentation of production
aspect of fragmentation-based international specialization which has been overlooked so far in the literature. The third part is a firm-level data analysis of the implications of production fragmentation for employment in the home country operations of Japanese manufacturing multinational enterprises (MNEs). This analysis is based on new panel data compiled from the unpublished returns of two firm-level surveys; ‘The Basic Survey of Business Structure and Activity’ and ‘The Basic Survey of Overseas Japanese Business Activity’, both conducted by the Japanese Ministry of Economy, Trade and Industry (METI), for the period 1991–2002. It estimates directly the effects of expanded overseas operations of foreign affiliates driven by production fragmentation on home employment in Japan. One of the striking differences between Japan and the USA is that for Japan unit labour costs and transportation costs (proxied by geographical distance) are relatively more important determinants of inter-country differences in fragmentation trade. The greater importance of labour costs in the Japanese case suggests that its fragmentation trade is more strongly driven by potential cost reductions from relocating labour-intensive production processes to neighbouring countries. Japan’s global production management strategy, ‘just-in-time’ logistics and the importance of physical proximity for quality maintenance may explain why transport costs (distance) play a stronger role. Turning to the labour market implications, our findings indicate that fragmentation trade has brought about a shift in favour of skilled workers in the employment structure of Japanese manufacturing. In particular, fragmentation trade with developing countries in East Asia has had a significant skill upgrading effect on manufacturing employment in Japan. In contrast, trade with OECD countries has had a skill downgrading effect. The firm-level analysis finds no evidence of employment losses in Japanese manufacturing at home due to product fragmentation. In fact, there is some evidence of a complementary relationship between overseas employment of MNEs and their home employment. In conjunction with the fact that increased profitability of Japanese firms would have a demand induced indirect employment expansion effect in other sectors of the economy, our empirical findings lead us to strongly reject the view that production fragmentation and associated overseas investment by Japanese firms has a negative impact on overall employment in Japan.
Acknowledgements This book grew out of my PhD thesis that I completed at The Australian National University (ANU). Most importantly, I would like to thank Professor Prema-chandra Athukorala for guiding me throughout my PhD study. His insights and constructive criticisms have improved the quality of this research immensely. He also financially helped my study by offering me a research assistantship when I was a student. Beyond supervision, he also became one of my best friends in Canberra. I will miss the discussions we had under the big gum trees in the courtyard of the Coombs tea room at the usual tea time. I would also like thank everyone who helped me write this book. I extend my profound gratitude to Professor Martin Richardson (ANU) for making comments on Chapter 2. He also taught me International Trade Theory and Industrial Organisation in my course works. What I have learned in his excellent courses helped me greatly in developing the analytical framework of this book. I am indebted to Professor Sisira Jayasuriya who has been acting as my mentor since I came to Melbourne. He gave me a lot of confidence in this project and kindly offered to help me revise the entire book manuscript and devoted his time in editing Chapters 1 and 8. I am grateful to Professor Gary B. Magee for his advice and would like to extend appreciations to my PhD thesis examiners, Professor Henryk Kiezkowski and Professor Fukunari Kimura for valuable comments and constructive criticism which led to significant improvement of the book. Parts of Chapters 6 and 7 were written while I was visiting the Institute of Economic Research, Hitsotsubashi University, Tokyo, Japan. My exceptional gratitude goes to Professor Kyoji Fukao for his kind invitation to visit, for making the arrangement for office space and sharing his insights on the topic. I am also grateful for the kind support from Ms Yumiko Moriyama during my stay in Kunitachi. I wish to express my gratitude to Carol Kavanagh and Sue Holzknecht for helping me improve my English writing while I was at ANU. I am grateful to Alberto Posso, Ryo Ochiai, Russell Thomson, Rodrigo Taborda and Shiro Armstrong for their valuable discussions, assistance and friendships. I would also like to thank Sreeni Jayasuriya for drawing the figures, and Maimi Hinata for supporting me in the final stage of completing the book manuscript. xi
xii
International fragmentation of production
I would like to extend my special gratitude to Alex Pettifer for reading and accepting my book proposal at Edward Elgar Publishing. I am also grateful to Bob Pickens and Laura Seward for editorial assistance. Last, but certainly not least, I would like to express my profound appreciation to my parents, Jiro and Suga Yamashita, for their love, spiritual and financial assistance for my study, without which this task would have been impossible to complete. This book is dedicated to them.
1
Introduction
1.1
THE ISSUES
International fragmentation of production – the cross-border splitting of the production process within vertically integrated manufacturing industries – has been a key facet of economic globalization over the past few decades. This relocation of production processes has been the main driver of the rapidly growing trade in parts and components (‘fragmentation trade’) between developed and developing countries in recent years.1 International vertical specialization implies two important forms of structural adjustment in the manufacturing processes of developed countries. First, it brings about a notable change in the patterns of the manufacturing trade by increasing fragmentation trade. Second, it has implications for both the level and composition of domestic employment. These are important and policy relevant issues. Over the past two decades international fragmentation of production has been reshaping the structure and performance of Japanese manufacturing, providing an excellent laboratory to investigate these issues. Despite a proliferation of research on this topic, the few existing Japanese studies have focused narrowly on the growth of international specialization without placing this in the context of the related and ongoing process of structural transformation. In addition, the findings and policy implications reported in the existing literature have been mostly derived from studies of the US manufacturing experience. However, the impact of globalization processes such as production fragmentation on industry can be quite heterogeneous across countries, depending on factors such as the specific features of industrial and labour market structures. This important point has been largely ignored in existing Japanese studies that have not addressed some unique aspects of Japanese fragmentation trade and Japanese labour market structures. This book examines the implications of the ongoing process of international fragmentation of production for structural adjustment and performance of Japanese manufacturing, paying attention to some peculiar aspects of industrial structure and labour markets in Japan. We place the Japanese experience in a comparative perspective, comparing it with
1
2
International fragmentation of production
the US manufacturing experience to gain a broader understanding of production fragmentation, which is itself a part of the wider process of globalization. This book makes three main contributions. First, it reports on the first rigorous comparative analysis of trends and patterns of fragmentation trade between Japan and the USA. Previous studies were unable to make a proper comparison of patterns of fragmentation trade across countries because of the lack of a suitable measure of the intensity of fragmentation trade. This problem was overcome in our analysis by the development of a new and improved measure. Second, we undertake a deeper empirical examination of the effects of fragmentation trade on the skills structure of manufacturing employment in Japan, recognizing that the geographical location of fragmentation trade may have differential effects on demand for skilled workers. This is an important aspect of fragmentation trade which had been largely overlooked in the previous literature. Third, the study informs the policy debate on the employment effects at home of increased overseas operations by multinational firms (MNEs), which has become the subject of heated debate in industrial countries because of concerns that these reduce home employment by ‘exporting jobs’.
1.2
STRUCTURE AND PREVIEW
The book consists of eight chapters. Chapter 2 provides a comprehensive interpretative survey of the theoretical literature and a discussion of measurement issues central to this study. The major focus of the survey is to place the ensuing empirical analysis in the appropriate context, and to justify the methodological approach chosen to conduct the empirical analysis. The discussion on measurement issues focuses on the selected indicators of fragmentation trade and their strengths and limitations with respect to indicators used in previous studies. Chapters 3 and 4 are concerned with structural changes in trade patterns brought about by the emergence of international fragmentation of production. Chapter 3 surveys the major structural shifts in the manufacturing trade patterns of Japan in the context of global trends and the direction and commodity composition of trade. Against this background, Chapter 4 examines the determinants of fragmentation trade in a Japan– USA comparison, using a newly compiled three-dimensional panel data set over the period 1988–2005. The analytical framework employed is the gravity model of trade flows. The basic gravity model is enriched by
Introduction
3
incorporating additional variables suggested by the theory of production fragmentation. Chapters 5 and 6 examine structural adjustments in the labour market in relation to the rapid growth of production fragmentation. Chapter 5 provides an overview of the changing nature of the labour market with special emphasis on the changing skill composition of employment and wages and the process of skill upgrading. Chapter 6 undertakes an econometric analysis of the skill upgrading effects of fragmentation trade, using 52 cross-industry panel data over the period 1980–2000. Chapter 7 describes a firm-level data investigation of the implications of production fragmentation for employment adjustment in manufacturing MNEs. The ‘working hypothesis’ of the analysis is that expansion of the overseas operations of MNEs, driven by international fragmentation of production, has occurred at the expense of their domestic (home) employment. This hypothesis is examined using newly constructed firm-level panel data drawn from unpublished returns in two firm-level surveys, ‘The Basic Survey of Business Structure and Activity’ and ‘The Basic Survey of Overseas Japanese Business Activity’, conducted by the Japanese Ministry of Economy, Trade and Industry (METI). Chapter 8 summarizes the major findings of the study, discusses the implications for better management of modern globalization in developed countries and draws policy implication for developing countries. The findings can be summarized as follows. First, the most striking finding in a USA–Japan comparison is that unit labour costs and geographical proximity are the most important factors explaining the modality of Japan’s fragmentation trade. This finding suggests that Japan’s fragmentation trade is driven mainly by the availability of lower labour costs and geographical proximity. This is consistent with the dominant position of East Asian countries, especially China, in Japan’s fragmentation trade. This also suggests that Japanese firms’ outsourcing decisions are driven mainly by the desire to achieve overall cost reductions by relocating labour-intensive production processes into low-wage countries. Second, it is found that the impact of fragmentation trade on skill upgrading varies depending on the factor endowment profiles of trade partners. The expansion of fragmentation trade with developing East Asian countries has had a significant positive impact on the skills composition of Japanese manufacturing employment. In contrast, trade with high-income countries (Organisation for Economic Co-operation and Development countries) has had a skill downgrading effect. Third, the findings do not support the hypothesis that expanded overseas operations has had any adverse effects on home employment levels
4
International fragmentation of production
of these MNEs. Instead, there is some evidence of a complementary relationship between overseas operations of MNEs and their home employment. Thus, the concerns expressed in public debates about the negative employment effects of foreign investments by MNEs appear misplaced. Overall, the evidence indicates that the impact of international fragmentation of production is to alter the skill composition of the manufacturing labour force in Japan, increasing its skill intensity, rather than to reduce the level of home employment in MNEs. Combined with the findings of US studies, the main argument of this book challenges the popular perception that production fragmentation, a key facet of ongoing globalization, has been growing at the cost of domestic manufacturing jobs in developed countries. On the contrary, production fragmentation strengthens the international competitiveness of the domestic manufacturing bases of the country by reinforcing the tendency towards specialization in skillintensive processes. While it cannot be entirely ruled out that this may have adverse effects on wages and employment of unskilled workers in manufacturing, it must be recognized that there are offsetting demand effects associated with the increased profitability of manufacturing MNEs. Hence, more empirical research is needed for any firm conclusions to be drawn on the effects of product fragmentation on low skilled workers.
NOTE 1. There is a significant and growing fragmentation trade among developed countries as well.
2
International fragmentation of production: a survey of theory and the measurement issue
2.1
INTRODUCTION
Until recently, international trade theory was dominated by the traditional notion of a horizontal specialization where goods were produced entirely from start to finish in one country and only completed goods were exchanged between countries. This notion has become increasingly inappropriate due to the increasing importance of fragmentation trade under which countries can specialize in different stages of the production process of a given product. While the doctrine of comparative advantage still offers a useful conceptual framework, the increasingly complex nature of fragmentation trade warrants a new analytical paradigm that goes beyond comparative advantage (Grossman and Rossi-Hansberg, 2008). In response, a sizable body of theoretical literature has emerged that puts forward the viewpoint of the neoclassical trade theory and industrial organization arising from the pioneering work of Jones and Kierzkowski (1990). The objective of this chapter is two-fold. The first is to provide a comprehensive interpretative survey of the theoretical literature, with emphasis on the determinants and the labour market implications of production fragmentation. This is done in order to place the ensuing empirical approach into the appropriate context and justify the methodological approach. The discussions are also drawn from the business management literature in order to obtain some sense of reality. The second objective is to discuss alternative approaches to the measurement of fragmentation trade. Despite the proliferation of theoretical research, empirical research has lagged behind due to the absence of readily available data relating to fragmentation trade. In applied work three different measures of fragmentation trade are generally used: statistics on the trade in parts and components, trade statistics collected under the Offshore Assembly Programme (OAP) and the Input-Output (I-O) Table. This chapter is organized as follows. Section 2.2 defines and presents 5
6
International fragmentation of production
an overview of international fragmentation of production. Section 2.3 undertakes a survey of the theory and is organized into three themes: neoclassical trade theory, industrial organizations and international production networks and the theory of MNEs. Section 2.4 delineates the issues that arise in the measurement of fragmentation trade and describes the approach taken in this study. Section 2.5 concludes.
2.2 2.2.1
INTERNATIONAL FRAGMENTATION OF PRODUCTION: AN INTRODUCTION Definition
International fragmentation of production is given different terms in the literature – ‘outsourcing’, ‘vertical specialization’, ‘production sharing’, ‘intra-product specialization’, ‘slice up the value chain’ and ‘intra-mediate trade’ (Krugman, 1995; Feenstra, 1998; Hummels et al., 2001; Bhagwati et al., 2004; Spencer, 2005; Helpman, 2006). In the eighteenth century the term ‘veredelungsverkehr’ (‘upgrading trade’) in German was already used for the same purpose (Grunwald and Flamm, 1985). All these terms broadly describe the process of breaking up the vertically integrated production process into finer stages and relocating each stage to the most suitable location across borders. The term ‘production fragmentation’ is used throughout this study to avoid unnecessary confusion. It is also useful to give a precise definition of the term ‘production fragmentation’ since its meaning is slightly different in the literature. In this study international fragmentation of production means both intra-firm transactions of parts and components between parent firms of MNEs and their foreign affiliates as well as international arm’s-length subcontracting transactions (trade with unaffiliated suppliers) in these items.1 This definition entails the physical transportation of parts and components across national borders. This automatically excludes subcontracting arrangements between foreign affiliates and the indigenous firms in the given host countries. Domestic subcontracting arrangements in home countries are also ignored. A distinction between parts and components and the traditional intermediate inputs trade is also made (for example, car engines versus basic metals). As will be explained in Section 2.4, this distinction is important for the accurate measurement of fragmentation trade. Additionally, the main focus of this study is on the physical separation of production stages in the manufacturing production process across international borders. Fragmentation trade in services (subcontracting) is beyond the scope of this study.2 There are many practical examples of production fragmentation, from
A survey of theory and the measurement issue
7
Mattell’s Barbie dolls (Feenstra, 1998), Texas Instruments’ high-speed telecommunication chip and sewing machines (Watanabe, 1972). An illustrative example also comes from the semiconductor industry (Grunwald and Flamm 1985; Brown and Linden, 2005). One of the most important semiconductor products is an ‘integrated circuit’ or ‘chip’, which is a network of tiny wires fabricated on a surface connecting transistors that switch processing data in a binary code on and off. The manufacturing process of the chip consists of three primary discrete value-chain activities: design, wafer fabrication and test and assembly (Brown and Linden, 2005). In this process of specialization in a value chain, the design is the most skill-intensive, requiring a very high standard of sophisticated technology and highly skilled labour.3 The next step, wafer fabrication, needs to be performed in an extremely clean location, but requires relatively lower skills than the design process. The fabrication stage also entails a huge fixed investment to build a plant (called a fab) that holds a wide variety of expensive equipment. Finally, assembly is typically the process of cutting the wafer into delicate individual chips (or dyes) and packaging them, with the intensive use of manual labour. Among these three value-added activities, assembly is likely to be relocated first to benefit from cheaper labour costs overseas, and then fabrication is likely to move next. Design activities are likely to remain inside the home country. For example, in 2002 the world’s leading chip maker, the US Intel Corporation, found assembly locations and testing facilities mostly in developing countries such as Malaysia, the Philippines, China and Costa Rica. The other sophisticated and high-end value processes, such as wafer fabrication, design and manufacture of the chips are still concentrated in the USA (UNCTAD, 2002). 2.2.2
Overview
Production fragmentation has been an important source of competitive advantage for multinational enterprises (MNEs) since the late 1960s (Helleiner, 1973; Grunwald and Flamm, 1985). Hence, the development of the fragmentation process has principally been driven by the international production of MNEs from industrial countries. MNEs based in the USA initially engaged in international fragmentation of production in order to gain cost competitiveness in the world market. This practice was encouraged in the US electronics and garments industries by the OAP, a special government scheme where tax exemption was granted to reimported products, comprising US value-added after offshore assembly (Finger, 1975).4 This practice subsequently spread to other heavy industry such as the automobile industry (Watanabe, 1972; Helleiner, 1973; Finger, 1975;
8
International fragmentation of production
Sharpton, 1975). In the late 1970s European MNEs began to become involved with the process of international fragmentation of production. From the late 1980s Japanese MNEs began to establish assembly operations, mainly in Southeast Asia. More recently, MNEs from the newly industrialized economies (NIEs) in East Asia have begun to contribute to further development of the process. A number of factors have contributed to the recent surge in worldwide production fragmentation. Product-specific technology advancement has increased the separability of the production process into finer degrees and segments depending on the factor intensity used (the technical divisibility of the production process) (Jones, 2000). This has facilitated a process once trapped within the domestic trade to move across international borders (Krugman, 1995). For instance, engineering activities, such as the manufacture of automobiles and electronics, have increasingly been separated into discrete production stages – manufacture of components, assembly, testing and packaging – with different skill requirements, scales and factor inputs (Lall et al., 2004). In contrast, the continuous production process of the chemical industry creates technical difficulty in separating the production segment into discrete steps. The communication revolution (such as the broadband Internet) has also led to significant cost reductions in making it easier to coordinate a separated production process across international borders, called the service link costs in Jones and Kierzkowski (1990). The continuous decline in transportation costs, especially air freight costs and improved containerization methods, has made it less costly and faster to move parts and components from one location to another (Hummels, 2007). The reduction in transportation costs has also facilitated the international separation of products which comprise higher values relative to their bulk (for example, computer chips) (Helleiner, 1973; Lall et al., 2004). Multilateral trade liberalization has added to the rapid growth of fragmentation trade across national borders. Yi (2003) makes the point that even a small tariff reduction has a so-called ‘magnification effect’ on fragmentation trade. This is simply because, unlike finished products, components and unfinished products cross international borders multiple times before reaching the final stage of the production process. Therefore, any marginal reduction in the protection scheme can significantly lower trade costs. The development of modular technology has had the effect of significantly expanding the possibility of global-scale fragmentation networks. In general, modular technology has allowed some components to be standardized for the use of multiple final-products across different sectors. Examples include computer chips and long-lasting batteries: computer
A survey of theory and the measurement issue
9
chips are now used in the manufacture of computers and automobiles to toasters, and long-lasting batteries originally designed for mobile phones are now widely used in electronic organizers, transmitters, radios, laptop computers and missiles (Jones and Kierzkowski, 2001b; Athukorala, 2005; Nishimura, 2005). The modular international production network in electronics has also been facilitated by the emergence of contract manufacturers (see Section 2.3.3 for more details). They provide ‘standardized’ components and services on a contract basis to lead firms, for example, purchasing of parts, testing and packaging, supply of chain management and services (Sturgeon, 2003). On the other hand, integral technology makes components that are either firm-specific or product-specific (Nishimura, 2005). Hence, unlike the case of modular technology, it is involved with confidential technological and knowledge assets. For this reason, firms have greater incentive to maintain the production of components with integral technology within their own boundaries (intra-firm trade) or to acquire them from suppliers with which they have a long history of business partnership. In summary, the initial stage of international fragmentation of production was motivated simply by lower foreign production costs to maintain the international competitiveness of major MNEs from industrial countries. Since then several important factors, including technological progress and the continuous reduction in transportation and communication costs, made the option of production fragmentation more attractive to achieve further cost reduction. While intra-firm trade by MNEs still dominates the form of fragmentation trade, the emergence of contract manufacturers facilitates the international arm’s-length transaction of parts and components (this development will be explained in detail in Section 2.3.3). Overall, international production networks have gradually begun to spread, involving many diverse developing countries working at different production stages and tasks.
2.3
A SURVEY OF THEORY
This review is divided into three sub-sections: the neoclassical trade model, the industrial organization model and the international production network. 2.3.1
Neoclassical Trade Model
A general framework for understanding the international fragmentation of production was first demonstrated by Jones and Kierzkowski (1990)
10
International fragmentation of production
and subsequently refined and developed in Arndt (1997), Venables (1999), Jones and Kierzkowski (2001a) and Kohler (2001) as an extension to the standard trade theory. More recently, further developments and generalizations have been done by Grossman and Rossi-Hansberg (2008) and Baldwin and Robert-Nicoud (2007). The standard approach usually begins with the canonical model of neoclassical international trade such as the Ricardian, Heckscher-Ohlin and Ricardo-Viner models, and then goes on to allow for production fragmentation. Vertically integrated production at each stage may require inputs (capital and labour) in different proportions given the technology and this can potentially match up with the presence of international factor price differences. This suggests that relatively labour-intensive components or assembly is located in labour-abundant countries and more capitalintensive ones in capital-abundant countries. This framework allows for an analysis of how factor prices, prices of goods, production and trade patterns adjust to international fragmentation of production. Jones and Kierzkowski (1990, 2001a) and Arndt (1997) provide a diagrammatic exposition of the effects of international fragmentation of production on changes in factor prices and the endowment of trading countries. These effects crucially depend on the relationship between the relative factor intensities of the industries, the factor intensity of the function that is relocated and the relative endowments. Hence, it is not generally possible to determine an unambiguous direction of the effects of production fragmentation on factor prices. Figure 2.1 depicts the familiar Lerner-Pearce diagram for two goods (X and Y) and two factors of production: high-skilled (H) and low-skilled (L) labour (Arndt, 1997).5 The modification to the standard trade model is to allow each of the final goods of production (X and Y) to contain two tasks of the production process: X production comprises x1 and x2 and Y production contains y1 and y2. The specification of production technology therefore contains two separable sub-production technologies; F (K, L) 5 f (g1 (H, L) , g2 (H, L)) where the output is produced using highskilled (H) and low-skilled (L) workers. Each sub-production technology also contains high-skilled (H ) and low-skilled (L) labour. The production function is assumed to be an increasing, concave and homogeneous degree of one in the inputs (H, L). This means there are constant returns to scale (CRS). The assumption of perfect competition in product and factor markets and the ‘small’ country are both retained (the price taker). The factors of production can also be freely mobile between sectors, but not between countries. The factor-price for high-skilled labour (H) is denoted wh, with wL for low-skilled labour. The factor-price ratio (wH / wL) is given a straight line and two unit-isoquants represent X and Y, respectively. As
A survey of theory and the measurement issue
y1
y
11
y2
High skills
Y H
x1
x
x2 X 0 Figure 2.1
H/L
(wL/wH)
L
Low skills
Fragmentation-based specialization
shown in Figure 2.1, production of good Y is relatively high-skilled intensive (for example, electrical machinery), whereas good X is relatively lowskilled intensive (for example, textiles). The isoquants cut through once and the factor-price ratio is tangent to both isoqaunts. This implies there is no factor intensity reversal. If the factor endowment lies within ‘the cone of diversification’ between 0x and 0y, both goods are produced. To keep the analysis simple, Figure 2.2 focuses only on the possibility of international fragmentation of production of X whose production process is broken into two tasks with the expansion path of the factor-intensity, 0x1 and 0x2. As shown in Figure 2.1, 0x1 is the relatively high-skillsintensive component of production (for example, design, the engineering activity and research and development) and 0x2 is the relatively low-skills intensive component (for example, assembly, simple processing and packaging). The original expansion path 0x is the weighted average of the factor intensities of the two component stages. Subsequently, international fragmentation becomes an attractive option due to the significant decline in transportation and communication costs. Production of component x2, the low-skilled worker’s intensive stage of commodity X, is relocated to another country where the cost of labour
12
International fragmentation of production
H’’
y1
y
High skills
H
x’1
Y
x1
H’
x x2 X
X’ 0
Figure 2.2
(wL/wH)
1
L’’
L’
L (wL/wH)
Low skills
The labour market effect of fragmentation on industry X
is relatively low (Figure 2.2). Clearly, the production process is only relocated if benefits exceed costs. In this analysis the additional costs of acquiring 0x2 are ignored. This is indeed the restrictive assumption made, but it is inevitable in order to maintain a manageable analysis. After allowing for international fragmentation, production of commodity X solely specializes in production stage x1. This shifts the expansion path to 0x1 (Figure 2.2). The overall cost of commodity X should be lower to reflect the lower costs of components of a production process, x2. This cost saving is represented by the inward shift of the X-isoquant to X’ on the new expansion path of 0x1. As asserted by Jones and Kierzkowski (1990, 2001a) this inward shift of isoquant is akin to a technological progress. The new isoquant, denoted X’ is tangent to a line H’L’, parallel to the original factor-price ratio (wH / wL). This indicates the lower cost of producing commodity X at unchanged factor and product prices. In other words, at existing factor prices the unit production of commodity X is sustained at the same high:low skilled labour ratio in equilibrium as the original one.
A survey of theory and the measurement issue
13
It then follows that production of X, which is relatively less-skilled labour-intensive, increases because the overall cost of production has fallen at the existing relative price. This raises the demand for low-skilled workers relative to high-skilled workers and pushes up the relative wage of less-skilled (L) to high-skilled workers (H). The resultant relative factorprice ratio (wH / wL)1 is now tangent to the Y-isoquant and the new isoquant X1. This further encourages both industries to substitute high-skilled labour for low-skilled labour. This shifts the factor endowment employed in both industries, from 0x1 and 0y to 0x’1 and 0y1 with the increased high:low skilled labour worker ratio. In the end, the factor-price ratio has shifted in favour of low-skilled workers (L). In short, international fragmentation of production in low-skills-intensive components results in an increased relative wage for low-skilled workers due to relocation of the low-skills-intensive component, x2. However, the result obtained in Figure 2.2 is not robust. If the least skill-intensive component of Y, which is y2, is less skill intensive than the most skill-intensive component of X (that is, y2 lies below x1), fragmentation of y2 would lead to the declining relative price of low-skilled workers. Hence, the outcome of production fragmentation in low-skill-intensive components depends critically on the factor-intensity of the sector and the factor endowment of the economy. This ambiguous outcome in the theory makes the empirical analysis important, as will be demonstrated in Chapter 6. One of the key contributions in a series of papers by Jones and Kierzkowski, (1990, 2001a) is to highlight the role of the service link costs. The cost saving effect of the international fragmentation of production is only realized if the factor-price savings outweigh the service link costs (the disintegration of costs of production). The service link costs have the composite structure incorporating transportation costs, costs involved in quality control, communications, management coordination and other logistics among the geographically dispersed production blocks (Jones and Kierzkowski, 1990, 2001a; Jones 2000). When production is geographically proximate, the service link is inexpensive. Jones and Kierzkowski (1990, 2001a) analytically demonstrate how the benefits from production fragmentation might interplay with the existence of the service link costs. They characterize the service link costs exhibiting strong increasing returns to scale (IRS), which is intuitive, given modern technology. For example, building up broadband networking might require a larger fixed cost, but the marginal costs are substantially lower. This creates a trade-off between fixed costs and lower production costs, which is demonstrated in Figure 2.3. Two types of costs are introduced in Figure 2.3. The fixed cost of the
14
International fragmentation of production
Costs TC1
TC2 TC3
B A
O Figure 2.3
Output The role of service links costs and lower marginal costs of production
service link is shown on y-axis and the marginal cost is shown by the slope of each cost schedule, denoted by TC. TC1 from the origin is a benchmark case where there is no fragmentation of production involved. Subsequently, TC2 involves a break up of the production process into two, which incurs the fixed cost of the distance OA. However, the marginal cost, shown by the slope of TC2, is lower before a break up of the production process (notice that the slope of TC2 is flatter than that of TC1). Further break up of the production process creates even lower marginal costs (shown by the slope of TC3), but it entails higher fixed costs (shown as the distance of OB). Finally, the optimal total cost schedule is shown by the heavy broken cost line. Thus, owing to the IRS nature of service link costs, the benefits of fragmentation of production become greater as output increases. More recently, Grossman and Rossi-Hansberg (2008) and Baldwin and Robert-Nicoud (2007) have developed a simple analytical framework using a mathematical approach. Grossman and Rossi-Hansberg (2008) decompose the impact of international fragmentation (‘offshoring’ in their terminology) on wages into three components: a labour-supply effect, a relative price effect and a productivity effect. It is shown that the productivity effect dominates the other two effects by raising the wages of relocated
A survey of theory and the measurement issue
15
low-skilled (and high-skilled) workers in a familiar trade model. This wage effect implies that international fragmentation of production essentially boosts the factor return of domestic workers with skill levels similar to those used in the relocated production process overseas. This inference is similar to the one obtained in Arndt (1997) and Jones and Kierzkowski (2001a). Baldwin and Robert-Nicoud (2007) revisit the four fundamental theorems of trade theory, the Factor Price Equalization (FPE) theorem, Heckscher-Ohlin theorem, the Stolper-Samuelson theorem and the Rybczynski theorem to examine how these theorems are altered to allow for international fragmentation of production. The literature employing the neoclassical trade model highlights the important potential gains from international fragmentation of production. However, two-fold criticisms are in order. First, treating production fragmentation merely as intermediate input trade is overly simplistic and misses the substance of the true dynamics of the phenomenon (Hummels, 2002). The national endowment pattern largely pins down the production and specialization pattern for the traditional intermediate inputs trade. On the other hand, fragmentation trade is determined by various other factors including transportation costs, labour costs and the level of infrastructure (Jones and Kierzkowski, 1990; Venables, 1999; Yeats, 2001). Trade in industrial inputs, generally in raw materials, is an age-old tradition. In fact, the early contributions have generalized the traditional HecksherOhlin model by incorporating trade in intermediate goods (Batra and Casas, 1973; Schweinberger, 1975; Riedel, 1976). Second, the information intensity between the ordinary intermediate inputs trade and fragmentation trade is vastly different, and the simple trade model is incapable of dealing with information-related issues (Hummels, 2002; Spencer, 2005). Fragmenattion trade requires information flows between buyers and sellers because of the complex product specifications, the quality maintenance and the intellectual property contents (information-intensive inputs). On the other hand, trade in traditional intermediate goods does not require a high level of information flow between buyers and sellers. This information flow is crucial for the organizational structure of international fragmentation. MNEs tend to undertake intra-firm trade for information-intensive inputs rather than the alternative arms’-length trade. The issue of the firm’s boundary and international organization structure is considered next. 2.3.2
Industrial Organization Model
The more recent literature combining trade theory with industrial organization models provides an analytical framework from the viewpoint of
16
International fragmentation of production
the firms (McLaren, 2000; Grossman and Helpman, 2002, 2005; Antràs, 2003; Spencer, 2005; Helpman, 2006). The key decision on the organizational form of international production is whether to manufacture parts and components in-house or to acquire them from unaffiliated firms. An equally important decision is whether to manufacture parts and components at home or in a foreign country. In this view an international labour costs differential stressed by neoclassical trade theory might not always be the primal determinant of firms’ decisions about international fragmentation. Other factors such as the presence of local supply capabilities of goods, good infrastructure and a well organized distribution network significantly shape the patterns and extent of international fragmentation, although labour costs are still important. Other considerations include firm-specific property rights and ‘market thickness’. ‘Thickness of market’ refers to the number of suppliers available to serve the needs of manufacturers. This notion was initially developed by McLaren (2000) and refined in Grossman and Helpman (2005). The first issue concerns the ownership structure determining the boundaries of firms (McLaren, 2000; Grossman and Helpman, 2002; Antràs, 2003; Antràs and Helpman, 2006). When carrying out international fragmentation, a firm must choose whether or not to keep production of components within its own boundary or acquire them from unaffiliated suppliers.6 In other words, a firm must decide to engage in intra-firm or arm’s-length transactions. The arm’s-length transaction might incur significant production costs, and usually entails extra search and negotiation costs related to contract agreements (transaction costs), and the risk of uncertainty (Helleiner, 1973). The option of internalizing can stabilize input costs by reducing uncertainties as to timing and quality of input delivery and generally increase control over the firm’s economic environment (Helleiner, 1981). However, internalization might also incur higher costs in setting up foreign affiliates in a host country. One specific consideration in the arm’s-length transaction is the possibility of an incomplete contract and the hold-up problem. An incomplete contract arises as a result of failing to cover for all possible future contingencies such as sudden unexpected changes in demand or input prices. In particular, it often happens when the transaction of goods and services between buyers and sellers involves complex contracts with a high degree of specification. This makes it difficult for a third party or a court to easily verify the contents of a contract, leading to the creation of transaction costs – the costs associated with negotiating, subsequent renegotiation, enforcing agreement and finding the alternative suppliers. If these transaction costs are too high, a firm wanting to purchase components may decide to vertically integrate an upstream supplier. As a consequence, the
A survey of theory and the measurement issue
17
presence of incomplete contracts and transaction costs creates the hold-up problem which leads to underinvestment from both or either party, failing to maximize the joint supply of the gains from the transaction. Generally, the more complex the transaction or the more uncertain the future, the greater the transaction costs and the risk of the hold-up problem associated with making the contract by the arm’s-length transaction more appealing. Hence, this consideration creates a trade-off between arm’slength and intra-firm transaction.7 The choice of internalization might be a less preferred choice due to the cost considerations of setting up one’s foreign affiliates. However, the intra-firm transaction alleviates the hold-up problem and transaction costs. It also protects technological secrecy by reducing the risk of leaking information on valuable technological assets to rival firms (Helleiner, 1973). Firms tend to engage in intra-firm transaction when components are particularly specialized and technologically advanced. Antràs (2003) combines the Grossman-Hart property-rights model of the boundaries of firms with the monopolistic competition model of international trade. His model demonstrates that imports from capital-abundant countries tend to be intra-firm trade, whereas imports from capital-scarce countries are likely to be traded at arm’s-length. Antràs (2003) provides theoretically consistent evidence that the share of intra-firm trade is positively correlated with the capital:labour ratio across US industries (for example, high intensity of intra-firm imports is found in the chemical and electrical machinery industry) and he also finds positive correlation of the share of intra-firm trade with the capital:labour ratio across the US trading partner countries. Nunn and Trefler (2008) provide the updated econometrics evidence of the findings of Antràs (2003). When there are no concerns about the transaction costs and technological secrecy, firms might opt for international subcontracting8 of parts and components with unaffiliated firms. This might have the merit of economizing on the costs of purchasing components which are provided in the competitive upstream market. This practice has recently been facilitated by the emergence of contract manufacturers and modular production networks (discussed in the next section). However, the various costs of international subcontracting involve the search for suitable external independent suppliers, negotiating with potential suppliers and upgrading and monitoring suppliers in order to maintain quality levels (Grossman and Helpman, 2005). A second issue raised in the literature on industrial organization is concerned with the location choice of international fragmentation of production. Grossman and Helpman (2002, 2005) developed a series of papers in which the number of firms is endogenous and each firm faces the
18
International fragmentation of production
choice between domestic and international subcontracting.9 According to Grossman and Helpman (2002, 2005), firms must search for suppliers who can produce the particular customized components with the appropriate expertise. Countries might differ in their degrees of contract incompleteness. This introduces the important role of market thickness, the legal system and other institutional qualities of government to influence the location choice of international fragmentation. Firms described in Grossman and Helpman (2002, 2005) conduct a search for a suitable partner either in technologically advanced developed countries in the North with relatively higher assembly costs or in the lowwage countries with less technology in the South. Grossman and Helpman (2005) consider that there is an infinite number of component specifications, located in the unit length of a circle where the buyers of components with different requirements of specification are uniformly distributed. International trade is introduced between North and South, in which final-good producers enter only in the North and component suppliers are resident in both the North and the South. In their model the location decision involves two considerations. First, labour costs are different across countries, making it attractive to search for component suppliers in the South (lower wage countries). Second, the market thickness differs across countries, making it attractive to search in the North where the probability of a successful match is higher. Searching for ideal partners is complicated by random elements involved in matching the particular technological expertises and the specific requirements between suppliers and final-good producers. In this setting ‘thickness of market’ becomes crucial for determining the probability of a successful search between final-good producers and specialized suppliers. The ‘thicker’ the market, the more suppliers exist with the expertise to serve the final-good producers’ requirements. Similarly, suppliers might also find it more profitable to enter the larger market of potential finalgood producers (Grossman and Helpman, 2005, p. 145). Total profits from a contractual relationship are equally divided between final-good producers and upstream suppliers. For this reason, final-good producers have more incentives to find suppliers with expertise closest to their requirements in order to minimize investments in customization. An important implication derived from the firm-specific model is that lower wage costs alone do not determine the location decision. In particular, Grossman and Helpman (2005) allude to the possibility that highwage countries with technological advancements and better infrastructure support in the North might outweigh the costs saving accruing if the South is chosen as the assembly location. The empirical analysis in Chapter 4, therefore, takes into account the quality of infrastructure, market size and
A survey of theory and the measurement issue
19
a legal environment that secures an enforceable contract relationship, in addition to the labour costs and productivity.10 2.3.3
International Production Network
It has been observed that there are two kinds of international production networks, the modular and the relational production network (Sturgeon, 2003). Interestingly, they are closely associated with the national characteristics of parent MNEs from industrial countries (Borrus, 1997). For example, it has been found that the Japanese international production network in the electronics industry has been a relatively closed system with a tightly controlled buyers-suppliers linkage (Froot, 1991; Belderbos, 1997; Hackett and Srinivasan, 1998). On the other hand, the US international production network in the same industry is often characterized by its full integration of modularity and the heavy use of contract manufacturers (Sturgeon, 2003). The latter development has been one of the most notable changes in the US electronics machinery industry over the past 15 years. The modular production networks originated in the USA in the late 1980s and gained importance in the 1990s. The modular production network is driven by contract manufacturers who provide traditional and standardized manufacturing functions, product (re)design, component processing and purchasing, inventory management, routine tests, as well as after-sales services and repairs. It is also facilitated by highly standardized inter-firm linkage requiring less frequent and intense interactions. These functions of contract manufacturers are highly modular in nature, being accessed and shared by a wide array of ‘lead firms’. The use of contract manufacturers may bring cost and flexibility advantages to ‘lead firms’ (Borrus et al., 2000; Sturgeon, 2003). As a result of the widespread use of the modular technology major firms such as Hewlett Packard and Ericsson in the electronics industry have been able to sell most of their worldwide manufacturing infrastructure to contract manufacturers, Solectron and Flextronics (Sturgeon, 2003). The modular production network has also spread into semiconductor and other heavy industry in the USA. In the US automotive industry Ford and General Motors (GM) have retained vehicle design and final assembly and rely on an increasing volume of components such as entire automotive interior systems, headlights, carpets, cockpits, and interior panels and module design from Leair, Johnson Contrils, Magna and TRW.11 The lower operating costs of contract manufacturers come from lower wage costs, and component prices, shared production capacity, better utilization of plant and equipment and economies of scale due to the presence
20
International fragmentation of production
of a variety of ‘lead firms’. The use of temporary workers comprising 10–50 per cent of their workforce also contributes to a reduction in labour costs. The provision of standardized goods and services to a wide pool of lead firms creates ease of entry for contract manufacturers and results in flexible networks where neither lead firm nor contract manufacturer is locked into a specific relationship beyond the current contractual arrangement. In contrast, the relational production network developed by Japanese MNEs is based on the social relationship of ‘trust’ and ‘reputation’. Product and process specification remain relatively tacit and are involved with intensive information flows between firms and suppliers which lead to greater asset specificity and relation-specific investment. In general, this form of production network relies heavily on technology-intensive components (sound display, memory chips, microprocessors, power and mechanical components or advanced design and development) supplied from related Japanese suppliers and sourcing simpler and non-strategic components from unaffiliated suppliers, usually for the previous generation model (Borrus et al., 2000). This procurement arrangement essentially blocks outside vendors from becoming involved with this international production network. The advantages of the relational production network are the steady technological upgrading in the supplier base, close coordination of just-intime deliveries and tight quality controls (Sturgeon, 2003). In particular, it can adapt well to a volatile market, as suppliers can respond quickly to changing market conditions by allowing for the replacement of workers and suppliers at short notice. The disadvantage is that it can build up excessively relation-specific investment and bilateral dependency between firms and suppliers. The way in which the relational production network has been organized seems to reflect the traditional industrial structure in Japan. In Japan firms traditionally rely on long-term subcontracting, known as vertical Keiretsu (Asanuma, 1989).12 This vertical structure is associated with a relational investment between some large downstream firms (for example, Toyota and Nissan in the automobile industry, Hitachi and Toshiba in the electronics industry) and a substantial number of small and mediumscale component manufacturers and assemblers (Shitauke in Japanese). It means more than just the standard hierarchical upstream-downstream vertical linkage involving an intensive interaction, dedicated investment efforts in R&D, reciprocal stock holdings, information flows, sharing of employees, managers, directors and even technology across the corporate groups (Asanuma, 1989). Given the close links with domestic component suppliers, Japanese manufacturing firms seem to find it difficult to switch to unaffiliated firms for provision of necessary parts and components.13
A survey of theory and the measurement issue
21
However, a change in the Japanese type of international production networks seems to be underway due to pressures to maintain international competitiveness. This cost pressure has contributed to the increasing use of outside purchasing (Kimura and Ando, 2003; Paprzycki, 2004; Ando and Kimura, 2005; Amano, 2008). Two large Japanese electronics firms, NEC and Sony, initiated a deal with US contract manufactures, Solectron and Clestina, for the management of supply chains, assembly and testing (Sturgeon, 2003). In summary, this section has identified some key features of Japanese and US MNE production networks. Japanese production networks, while changing, continue to rely on a large share of production being carried out in-house and/or within a tightly controlled group network. In contrast, the US production network is an open system that increasingly relies on the global operations of independent contract manufacturers, creating an international modular production network. This contrasting structure of production network might contribute to the distinctive specialization and patterns of fragmentation trade. Therefore, it is important to undertake an empirical analysis separately for Japan and the USA in order to highlight the similarities and differences of the determinants of fragmentation trade (Chapter 4).
2.4
THE MEASUREMENT ISSUES
There is no unique way to measure the degree of the fragmentation process in manufacturing trade. This section discusses the limitations of the available measures of production fragmentation: the trade data on parts and components, the offshore assembly programme (OAP) data and InputOutput (I-O) tables. 2.4.1
Trade Data in Parts and Components
The first approach and the one taken in this study relies on published international trade statistics on parts and components identified at the most highly disaggregated commodity level (that is, five-digit). This was pioneered by Yeats (2001) who used a list of commodity classifications based on Standard International Trade Classification (SITC) Revision 2.14 However, the product coverage of parts and components presented in Yeats (2001) suffers from two limitations (Lall et al., 2004; Kaminski and Ng, 2005). First, only items specifically labelled as ‘parts and accessories’ in the United Nations (UN) trade commodity classification system are covered in the Yeats (2001) list. This method of reclassification excludes
22
International fragmentation of production
important components such as semiconductors and transistors, which are used in the production of computers, telecommunications and consumer electronics products. Printed circuits are also excluded since they are entered as discrete products under the SITC commodity description (Lall et al., 2004).15 Second, the classification based on the SITC Rev. 2 system has become overwhelmingly out of date, given the spectacular information technology (IT) revolution since the early 1990s. Against this backdrop, Athukorala (2005) used a more comprehensive list of parts and components on the basis of SITC Rev. 3 which was introduced in the late 1980s. Compared to SITC Rev. 2, the Rev. 3 system provides a wider coverage of parts and components within machinery and transport equipment (SITC 7) by readdressing some overlapping issues in Rev. 2. For instance, what Rev. 2 classified as ‘electrical application such as switchgear’ (SITC 772) is more finely reclassified by separating ‘Parts of Switchgear’ (SITC 77282) from other final products in Rev. 3. Moreover, Rev. 3 identifies parts and components for a number of products belonging to the miscellaneous manufacturing category (SITC 8) (for example, parts of furniture). There are three key advantages for using trade data for components. First, it avoids mixing traditional intermediate inputs into the estimates by making a distinction between fragmentation trade and the ordinary intermediate inputs (raw materials). This separation is critical because the former represents the rapidly growing production fragmentation, while the latter is not a new type of trade flow (Yeats, 2001; Athukorala, 2005). While raw material imports are mainly driven by resource endowments, fragmentation trade is influenced by totally different factors. Trade in intermediate inputs is not part of the rapidly growing production fragmentation in world trade. Second, trade data capture both the export and import sides of production fragmentation. Fragmentation trade is not confined to importing labour and low-skills-intensive components from low-wage countries. It has mainly been evolving by exporting domestically produced highly technological and capital-intensive components to developing countries for further processing and assembling. In particular, Japanese and US MNEs are heavily involved in the export orientation of the fragmentation process for further processing offshore. Third, trade data can examine the direction of fragmentation trade. This is important because fragmentation trade with developed and developing countries has entirely different implications. Of course, the trade data approach is not entirely free from shortcomings. The main one is the limited industry coverage, since a detailed separation of the parts and components trade is mainly for machinery and
A survey of theory and the measurement issue
23
transport equipment (SITC 7) and miscellaneous manufacturing (SITC 8) by the available trade commodity classification system. This automatically ignores the intensity of fragmentation trade in other industries. There is ample evidence from the case studies to suggest the fragmentation process has increasingly been prevalent in other sectors such as pharmaceutical and chemical products (falling under SITC 5) and machine tools and various metal products (SITC 6) (Macher and Mowey, 2004). The process for personal computer (PC) software manufacturing has been expanding at a rapid pace, but it is still lumped together with ‘special transactions’ under SITC 9. Moreover, components related to some products in SITC 8 are included under other SITC categories. However, a focus on the above listed industries (SITC 7 and 8) is justified because the available case studies confirm the bulk of production fragmentation is highly concentrated in the machinery industry (Brown and Linden, 2005).16 In addition to the incomplete coverage of other industries, the trade data suffer from some measurement issues. First, the recorded trade data might contain a significant portion of double-counting, because components might travel across international borders several times. This is particularly likely to overestimate the actual values of trade flows. Second, some degree of arbitrariness in the classification of commodities cannot be avoided, when it comes to distinguishing between ‘goods-in-process’ and finished goods. For instance, while some components can be intermediate inputs if they are integrated into the production process, they can also be final consumer goods. This separation becomes important because the focus here is on components which are incorporated in the production of manufactured goods. Nevertheless, it is not possible to know to what degree this measurement error has been introduced into the data. Third, there is a concern for the treatment of entrepôt trade. In particular, this problem is severe when assessing China’s recent integration to global production fragmentation due to the presence of Hong Kong entrepôt trade. The international trade data are based on the value reported by reporting (exporters) countries (that is, export values reported by exporting countries). There is evidence to suggest the reported records from exporters do not necessarily match the values quoted from partner (importing) countries (Yeats, 1978, 1995). This statistical discrepancy can be very large, not only because of the cost, insurance and freight (CIF) and free on board (FOB) differences, but also due to factors such as transportation costs, different goods classification between the reporter and the partner country, time lags in recording and multiple exchange rates (Yeats, 1978, 1995). However, there is a general belief that importer records are more accurate, since the importing country has more incentive to record trade statistics accurately for tax revenue collection purposes (Feenstra et al.,
24
International fragmentation of production
1999). It is highly probable that recording errors are less susceptible in the case of industrial economies due to their more effective customs systems.17 Therefore, this study focuses on countries’ export records, rather than just using importers’ records. 2.4.2
Offshore Assembly Programme (OAP) Data
The second approach is to utilize the data collected on the special operations of foreign processing and assembly, such as the US Offshore Assembly Programme (OAP) or the Inward Processing Trade/Outward Processing Trade (IPT/OPT) scheme of the European Union (EU) (Helleiner, 1973; Sharpton, 1975; Egger and Egger, 2005; Swenson, 2005, 2007). Under the US OAP schemes tax exemption is granted to re-entry of domestically produced components assembled abroad (Finger, 1975). In other words, US firms can export components or materials and have them processed or assembled using relatively cheaper foreign labour. When those processed goods return to the USA tax duties only apply to foreign value-added, not the US value-added. EU member countries also have similar tariff provisions on IPT/OPT.18 Japan also has special tariff arrangements of a similar nature, but their use is more restricted than those of the US OAP or the EU’s IPT/OPT. Using the processing trade data it is possible to distinguish between re-entry of dutiable imports, representing the dutiable value-added associated with foreign assembly production and non-dutiable parts of valueadded. In this sense, the data precisely contain the accurate information on the operation of overseas assembly (Grunwald and Flamm, 1985; Feenstra et al., 2000). The US OAP imports were a provision of the US Tariff Act of 1930 (Grunwald and Flamm, 1985; Hanson, 1997; Feenstra et al., 2000). The original intention of creating this special scheme was to facilitate the manufacturing practice of US steel firms, many of which maintained production plants in Canada and engaged in extensive cross-border shipment of intermediate inputs. The US International Trade Commission (USITC) is responsible for collecting OAP imports statistics. A set of OAP imports was initially classified under the US tariff code 806/807. Code 806 permits reimport of ‘fabricated’ but unfinished metals production in the USA for further processing, and code 807 permits re-entry of foreign ‘assembly’ of finished goods for final sale in the US market. OAP imports under provisions 807 are by far the most important items due to its relatively wide coverage. They were later renamed the 9802 provisions of the Harmonized System code (Finger, 1975; Feenstra et al., 2000). The data are collected at the four-digit level of SIC.
A survey of theory and the measurement issue
25
However, these processing trade data suffer from two major limitations. First, the coverage of these schemes has a somewhat limited focus, since only the items under those special schemes are recorded. In order to qualify for OAP imports the goods finally assembled abroad need to be returned to the USA. However, production fragmentation is not only confined to goods that have been processed abroad returning home. It might be the case that foreign assembly goods containing US produced components and parts are shipped to other third countries from the assembly locations, instead of coming back to the USA for sale. The OAP statistics do not trace such trade flows (Grunwald and Flamm, 1985). Second, the benefit of tax exemption is disappearing due to the ongoing process of multilateral tariff reductions that make the OAP arrangement redundant (Hijzen et al., 2005). The importance of OAP imports in total US imports has been declining over the years, dropping to 8 per cent in 2000 from 12 per cent in 1990 (Swenson, 2005). 2.4.3
Input-Output (I-O) Table
The third approach is to use an Input-Output (I-O) table technique to compute imported intermediate inputs used in domestic production of a given country or industry (Feenstra and Hanson, 1996, 1999; Campa and Goldberg, 1997; Hummels et al., 2001). A national I-O table generally includes inter-industry flows of intermediate goods, generating an accounting framework for the circulation of the whole economy at the finer disaggregated industry level (sector-level gross output, value-added, exports and imports and classified into intermediate use and final demand). The purpose is simply to measure the overall degree of dependence on imported intermediate inputs, as an indication of the fragmentation process for a given industry. Feenstra and Hanson (1996, 1999) proposed a measure based on an I-O table and applied it to US manufacturing for computing the ratio of imported intermediate inputs. The data on imported intermediate inputs are extracted from the Annual Survey of Manufactures, the US Bureau of Census, which provides the raw data for an I-O table. The FeenstraHanson approach has been very popular as an indicator of fragmentation trade (Hansson, 2000; Strauss-Kahn, 2004; Hijzen et al., 2005; Hsieh and Woo, 2005; Ito and Fulao 2005; Ekholm and Hakkala, 2006). There are two types of I-O tables, depending on the way import transactions are compiled in the accounting framework (Bulmer-Thomas, 1982). A competitive type I-O table includes all types of imports without making a distinction between a product imported as intermediate input or final demand imported. In a non-competitive (or complementary) I-O table
26
International fragmentation of production
the independent import matrix table is prepared consisting of the interindustry use of imported intermediate inputs. If an independent imported input matrix is not available, imported intermediate inputs for each industry i have to be estimated by the following formula (Feenstra and Hanson, 1999): importsj Imported Intermediate Inputsi 5 a [ inputs from industry j to i ] * c d domestic absorption j
j
"" inter-industry intermediate inputs flows
import penetration ratio
(2.1) where subscript i is purchasing industry and j denotes the supplying industry with intermediate inputs. Equation (2.1) corresponds to ‘broad outsourcing’ in the Feenstra-Hanson terminology. It aims to capture interindustry intermediate imported input purchases by purchasing industry i from industry j. The first term in equation (2.1) comprises the total volume of intermediate input purchases of industry i from other manufacturing industries j. The first term is multiplied by the second, which comprises the import penetration ratio of producing industry j. Domestic absorption is usually defined as gross output plus imports and minus exports. ‘Broad outsourcing’ is then defined as the ratio of imported intermediate inputs to the total expenditure on intermediate inputs. Alternatively, ‘narrow outsourcing’ is defined to capture intra-industry flows of intermediate goods purchases of industry i from the same industry classification at the two-digit level. Of these two measures, the broad outsourcing measure is often the second preference, because its broad definition includes inter-industry flows of intermediate inputs. Narrow outsourcing is preferred because it only includes, for instance, intra-industry purchase of automobile parts (parts of automobiles) by downstream firms in the automobile industry. Hijzen et al. (2005) and Ekholm and Hakkala (2006) employ a noncompetitive (or complementary) import type of I-O table where the source of the intermediate inputs is distinguished between domestic (the domestic intermediate use matrix) and foreign suppliers (the import intermediate use matrix). The computation of outsourcing intensity corresponds to summing up each column in the import matrixes of a non-competitive type of I-O table. It essentially refers to diagonal elements in an intermediate import use matrix. The use of the import use matrix is superior to the competitive type of I-O table because the intensity of fragmentation trade is not driven by increased import penetration of final goods and the imported intermediate inputs are clearly separated from the domestic source (Hijzen et al., 2005).
A survey of theory and the measurement issue
27
There are several reasons why these indicators based on the I-O table do not appropriately capture the true dynamics of the fragmentation process in a meaningful way. First, the use of the total imports penetration ratio, the last term in equation (2.1), limits the precision of measurement of the imports dependency on intermediate inputs. The ratio of imported intermediate inputs to the consumption of total intermediate inputs might be very different from the ratio of the imported final goods to final goods consumption (Strauss-Kahn, 2004). If trade in intermediate inputs grows faster than trade in final goods, it can induce a significant downward bias into the measure. In fact, there is ample evidence to suggest trade in parts and components has been growing at a faster rate than trade in final goods in recent years (Yeats, 2001; Athukorala, 2005). The direct use of the imports use matrix alleviates the mixing up problem by directly computing the import penetration ratio of the intermediate inputs (Hijzen et al., 2005). Second, the I-O table by construction does not permit separating imported intermediate inputs (raw materials) such as steel, metals, plastics and chemical products and parts and components, as discussed in Section 2.4.1. This separation is crucial because the latter represent a rapidly growing category of international trade flows, whereas the former is not a new type of trade flow. This separation is particularly important to a resource-poor country such as Japan due to its high dependency on imported raw materials. Therefore, failing to distinguish between ordinary intermediate inputs and fragmentation trade flows might lead to misleading inferences. This will be demonstrated in Chapter 6. Third, by its very nature, the I-O table only focuses on the import side. However, as discussed previously, the export side is also an important consideration to appropriately measure involvement of fragmentation trade for given industries. Related to this problem, the I-O table does not permit an analysis of bilateral trade patterns. For this it requires matching the I-O table with trade statistics for specifying the direction of trade (a very cumbersome exercise). Hummels et al. (2001) attempted this matching, but it is far from perfect. 2.4.4
The Method of Data Compilation
Mindful of these limitations, in this study trade data on parts and components are used to measure the intensity of fragmentation trade. The data are tabulated from the UN Comtrade database using a list of parts and components prepared building on the previous studies by Yeats (2001) and Athukorala (2005). Identification of trade in parts and components takes a more systematic
28
International fragmentation of production
approach following the commodity classification system provided by the UN’s Broad Economic Category (BEC), whereas Yeats (2001) and Athukorala (2005) simply identify a list of components by focusing on the product description. The BEC classification system is intended to categorize SITC-based trade statistics into a large economic class of items and to supplement the summary data compiled on the basis of the SITC system.19 The original BEC was published in 1971, and Revision 1 was issued in 1976 and Revision 2 in 1986. The BEC was developed in such a way that it would provide the elements which enable the construction of aggregates of trade goods approximately comparable to those for the three basic classes of goods in the 1968 Social National Account (SNA). A number of sub-categories were established to supplement these main categories. The sub-categories reflect the various end-uses of commodities. Among seven major commodity categories, industrial supplies (BEC 2), capital goods (BEC 4) and transport equipment (BEC 5) include a subcategory for ‘parts and accessories’. The corresponding sub-categories are BEC 22, BEC 42 and BEC 53. However, not all of the items classified under BEC 22, 42 and 53 correspond to parts and components. Only the items under these three BEC sub-categories which at the same time correspond to SITC 7 (machinery and transport equipment) and SITC 8 (miscellaneous manufacturing) are identified as parts and components in this study. Limiting items to SITC 7 and 8 prevents the inclusion of some components traded as ‘products in their own right’ under specific trade names (for example, automobile tyres). The final list prepared through this procedure contains a total of 264 items.
2.5
CONCLUDING REMARKS
This chapter has surveyed both the theory and measurement of the international fragmentation of production. The survey of the theory covers three main strands of literature: the neoclassical trade model, the combination of industrial organization and trade model and the international production network. The analysis based on the neoclassical trade model is useful for harnessing the basic workings of the effects of international fragmentation of production on the factor prices, production and endowment pattern. However, this model is less satisfactory in practice, when international fragmentation is driven not only by lower labour costs in a foreign country, but also by other cost considerations and the supplyside capacity such as the quality of the infrastructure, political stability, the level of technological capacity and the good endowment of workers. The new strand of trade literature combined with industrial organization
A survey of theory and the measurement issue
29
provides these accounts. In the empirical exercise undertaken in Chapter 4, the infrastructure, the quality of government and labour productivity in addition to labour costs will be explicitly taken into account. The measurement issue also remains one of the unsolved issues among applied trade economists. Three widely used approaches were reviewed in this chapter: trade statistics in parts and components, trade statistics collected under the Offshore Assembly Programme (OAP) and the InputOutput (I-O) table. It is argued, inter alia, that the trade data for parts and components well capture the dynamic aspects of fragmentation trade. This is especially the case for measuring the intensity of international fragmentation in Japanese industries because Japan traditionally has a high intensity of importing industrial raw materials.
NOTES 1.
2. 3.
4. 5. 6. 7.
8. 9. 10.
This broad definition is similar to the definitions used in Grossman and Rossi-Hansberg (2006) and Helpman (2006). These authors make a distinction between ‘offshoring’ and ‘outsourcing’. Offshoring simply means fragmentation of previously vertically integrated industries and relocation overseas whereas outsourcing suggests production processes formerly performed within-firm are now being purchased by arm’s-length transactions. Despite this slight difference in definition, offshoring and outsourcing are used interchangeably in the literature. On subcontracting in services trade, see Amiti and Wei (2005), Blinder (2006) and the works cited therein. The design of a chip consists of a hierarchical procedure of specification, logic design and physical design before reaching the prototype stage (Brown and Linden, 2005). The first stage is to generally specify how the chip behaves within the system. The next stage, logic design, describes how signals will be processed within the chip, first at a register level and then at a gate level. Physical design translates the symbolic abstract version into actual wires and devices interconnecting multiple layers on the silicon layers. It also requires expensive electronic design automation EDA software with high fixed investment. A complex chip design such as Intel’s Pentium 4, with 42 million transistors on a 180pmm line-width process, requires hundreds of engineers for the full length of a five year project. See also Section 2.4.2 for description of the US OAP data. Alternatively, the two factors of production can be capital (K) and labour (L). This issue is directly related to the old economic question, raised by Coase (1937), of whether firms should ‘make-or-buy’ these intermediate inputs. Some papers have examined ‘make-or-buy’ decisions within the context of imperfect competition without an incomplete contracts framework (Shy and Stenbacka, 2003; Chen et al., 2004). Firms strategically make decisions on internalization, given their rivals’ choice. The Subgame Nash equilibrium (SPNE) is influenced by the large fixed cost of the vertical integration option and relative lower production costs. The term ‘international subcontracting’ is used to mean arm’s-length transaction. As described in Section 2.2, this is a subset of international fragmentation of production. In order to avoid repetition, the possibility of internalization and arm’s-length transaction is ignored. Grossman and Helpman (2005) exclusively focus on the location choice of arm’s-length transactions (international subcontracting). The exact measurement for these variables is detailed in Chapter 4.
30 11.
12.
13.
14. 15. 16.
17. 18.
19.
International fragmentation of production A set of smaller contract manufacturers also emerged in Taiwan in the 1990s. Compared to their US counterparts, Taiwanese contract manufacturers have a much narrower product focus, specializing in design services, and low- to mid-range personal computers. Another type of Keiretsu is horizontal Keiretsu linking companies from different industries at a horizontal level (Lawrence, 1991; Saxonhouse, 1993). Member firms (usually big corporations) are attached to each other in a close linked business and tight partnership through cross-share holdings and usually share a common ‘main bank’. There is a continuous flow of information and exchange of personnel between members. Typical examples of horizontal Keiretsu are Mitsubishi and Mitsui (Lawrence, 1991). This description is perhaps too broad a definition of Keiretsu relationships. However, as Lawrence (1991) noted, it is not easy to be more specific, because Keiretsu relationships, which usually occur under implicit contracts, are often ambiguous and fluid, especially as there is sometimes an overlap in inter-firm linkages between horizontal and vertical Keiretsu. It is also important to note that Japanese MNEs have a tendency to replicate the formation of a vertical Keiretsu link in a host country. The evidence suggests vertical Keiretsu suppliers tend to follow the final-good assemblers by conducting foreign direct investment (FDI) in the same host country (Head et al., 1999). This indicates a very tight vertical contractual relationship between core firms and vertical Keiretsu suppliers. The SITC Rev. 2 which was introduced in the late 1970s was a significant departure from the original Rev. 1 classification in that is separated a list of parts and components from final goods within ‘the machinery and transport sector’ (SITC 7) (Yeats, 2001). Yeats (2001) strangely includes all products under ‘Telecommunication Equipment and Parts’ (SITC 764) which do not necessarily include components (for example, telephone sets). The finer disaggregated product classification (six-digit level) is possible by the Harmonized System (HS), which started in 1988, the same year as the publication of SITC Rev. 3. However, there are two practical problems associated with using HS for this purpose. First, the country coverage is relatively small compared with the SITC system, since most developing countries are still slow to report by HS. Second, as far as component items are concerned, there is not much difference between the five-digit level of SITC Rev. 3 and the six-digit level of HS. That is, parts and components identified under HS are more disaggregated from component items under five-digit level of SITC classification. Therefore, the separation of components from final assembled goods is virtually the same between HS and the SITC system at this level of disaggregation. See Yeats (1978) for a sceptical view. The data are available from Eurostat, Comtex (CD-Rom). See more details at http:// www.iue.it/LIB/Guides/Economics/Statistics/Descriptions/comext.shtml. OPT allows goods to be temporarily exported from the EU territory for the purpose of offshore processing, and the resultant assembled goods to be returned to the EU with total or partial exemption from import duties (Görg, 2000). IPT permits duty free imports of components for processing inside the EU territory and subsequently exports of the assembled goods without payment of duty. See more details on the description of the BEC at http://unstats.un.org/unsd/cr/family2. asp?Cl=10.
3
Production fragmentation and trade patterns in Japanese manufacturing
3.1
INTRODUCTION
Over the past two decades international fragmentation of production has transformed the fundamental nature of global trade. Firms in the US electronics industry pioneered this process of economic globalization. Since the late 1980s Japanese firms have also become increasingly involved in this process. However, despite the increasing importance, only a few studies have explicitly elucidated the modality of fragmentation trade for Japan (Fukao et al., 2003; Kimura and Ando, 2003; Ando and Kimura, 2005; Tomiura, 2005). This gap in the literature has been mainly due to the absence of the readily available measure of fragmentation trade. Against this backdrop, this chapter examines Japan’s emerging trade pattern using data on parts and components as the indicator of fragmentation trade (see Chapter 2 for the data compilation method). In particular, the patterns and trends of fragmentation trade are examined in a wider global perspective with the emphasis on Japan and other industrial countries. The chapter is organized as follows. Section 3.2 depicts the overall patterns and composition of Japanese manufacturing trade from 1962 to 2005.1 Section 3.3 provides an analysis of the patterns of fragmentation trade in Japan’s machinery trade. Section 3.4 examines the two key machinery sectors of electronics and the automotive industry. The final section presents a summary of the major findings.
3.2
AN OVERVIEW OF TRADE PATTERNS
In the 1930s the share of Japan’s manufacturing exports was high at around 75–80 per cent (Ohkawa and Rosovsky, 1973). The predominant share of manufactures on the exports side has also continued in the post-war period, accounting for over 90 per cent of total merchandised exports since 1962. Manufacturing exports also achieved double-digit growth in the post-war industrialization period of the 1960s and 1970s. 31
32
International fragmentation of production
As a consequence, Japanese manufacturing exports established a strong position in the 1970s and 1980s, accounting for close to 18 per cent of the world share on manufacturing exports (Table 3.1a). Hence, many studies have suggested that Japan’s post-war economic growth was driven by strong manufacturing exports performance (Ohkawa and Rosovsky, 1973; Boltho, 1975; Patrick and Rosovsky, 1976; Minami, 1986). General explanations for this exports success during this period have been based on the availability of well-educated, well-disciplined and low-wage workers, combined with rapid adaptation of foreign technologies (Ohkawa and Rosvosly, 1973). However, the competitive position of Japanese manufacturing exports began to decline from the early 1990s because of a deterioration in international competitiveness arising from domestic cost pressure and the persistent appreciation of the Japanese yen. After achieving the highest world share of 19 per cent in manufacturing exports in 1986, Japan’s share continuously declined to 14.2 per cent in 1990 and then to 9.8 per cent in 2000 (Table 3.1a). In the 1960s and 1970s the share of primary products including raw materials in total imports accounted for about 80 per cent (Table 3.1b). Raw materials and fuel imports were needed to sustain the growth of manufacturing in post-war industrialization. In the 1970s Japan dominated world imports of raw materials. The world share of primary product imports amounted to around 17–19 per cent in the 1960s and 1970s. However, Japan’s dependency on imports of primary products began to decline in the late 1980s. The share of primary products in total imports was reported at 55 per cent in 1990 and 42.5 per cent in 2000. This decline was largely due to the fall in crude oil imports, partly as a result of changes in the resource usage as well as changes in production techniques to minimize resource imports dependency in Japan. Since the mid-1980s manufactured imports have begun to replace the dominant share of primary products in Japan’s total imports. The share of manufacturing imports accounted for 28 per cent in 1985, substantially increasing to 45 per cent in 1990 and reaching close to 60 per cent in the late 1990s and early 2000s (Table 3.1b). This dominant position of manufacturing characterizes Japan’s trade patterns for the past 40 years. Accordingly, the product composition of Japan’s exports bundle has shifted from light- and labour-intensive towards more capital- and technology-intensive manufacturing such as electrical machinery products. In recent years, China had become the most important source country for Japan’s imports, shifted from the USA. Given this changing nature of manufacturing trade, the following sections focus on the evolution of fragmentation trade in the global context.
Production fragmentation and trade patterns in Japan
Table 3.1 (a)
Key indicators of Japan’s foreign trade, 1962–2005
Exports
Year
Total Merchandise Exports (US$ billion)
1962–67 1968–73 1974–79 1980–85 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
(b)
33
7.6 22.8 75.8 150.7 207.1 226.8 261.3 271.2 282.3 309.5 334.3 354.8 387.9 433.5 400.6 409.3 375.7 403.7 461.6 386.6 398.8 451.5 540.9 567.3
Share in Total Exports of
Exports Growth of:
Share in World Exports of:
Primary Productsa (%)
Mfgb
Primary Products (%)
Mfg
Primary Products (%)
Mfg
9.2 6.2 4.3 3.3 2.5 2.5 2.5 2.5 2.6 2.4 2.5 2.6 2.6 2.8 2.8 2.9 2.8 2.7 2.7 3.3 3.0 2.9 3.2 3.8
90.8 93.8 95.7 96.7 97.5 97.5 97.5 97.5 97.4 97.6 97.5 97.4 97.4 97.2 97.2 97.1 97.2 97.3 97.3 96.7 97.0 97.1 96.8 96.2
5.7 17.9 16.6 4.1 10.0 10.3 12.3 7.6 5.9 3.3 11.2 7.7 12.7 19.2 −8.6 5.1 −9.9 4.3 11.7 2.7 −6.2 11.3 30.1 25.3
17.5 23.7 19.8 10.2 19.2 9.5 15.3 3.7 4.1 9.8 7.9 6.1 9.3 11.5 −7.6 2.1 −8.2 7.5 14.4 −16.8 3.5 13.3 19.5 4.2
1.5 1.6 1.0 0.9 1.2 1.2 1.2 1.1 1.0 1.0 1.1 1.2 1.3 1.3 1.0 1.1 1.1 1.1 0.9 1.0 1.0 0.9 0.9 1.0
12.3 15.9 17.9 17.7 18.7 16.9 16.6 15.6 14.2 12.7 12.4 13.2 12.5 11.7 10.3 10.0 9.0 9.4 9.8 8.4 8.3 8.1 8.1 7.6
Imports
Year
1962–67 1968–73 1974–79 1980–85
Total Merchandise Imports (US$ billion)
8.3 21.3 73.5 131.8
Share in Total Imports of:
Growth of:
Primary Productsa (%)
Mfgb
Primary Products (%)
78.5 75.7 80.1 78.2
21.5 24.3 19.9 21.9
16.5 22.0 22.9 2.4
Mfg
14.0 27.3 17.4 5.5
Share in World Imports of: Primary Products (%)
Mfg
12.2 16.7 18.7 16.7
2.8 3.6 3.4 3.3
34
International fragmentation of production
Table 3.1 Year
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
(continued) Total Merchandise Imports (US$ billion)
117.7 144.1 181.1 205.1 228.6 231.1 227.7 235.6 268.4 328.6 342.8 333.1 275.3 304.2 373.2 342.8 330.5 376.8 446.9 506.4
Share in Total Imports of:
Growth of:
Share in World Imports of:
Primary Productsa (%)
Mfgb
Primary Products (%)
Mfg
Primary Products (%)
Mfg
65.7 61.9 57.8 55.8 55.2 54.4 54.0 51.9 48.8 45.1 44.0 44.4 41.6 40.9 42.5 41.7 40.8 41.5 42.1 44.8
34.3 38.1 42.2 44.2 44.8 45.6 46.0 48.1 51.2 54.9 56.0 55.6 58.4 59.1 57.5 58.3 59.2 58.5 57.9 55.2
−17.3 15.4 17.4 9.3 10.2 −0.3 −2.2 −0.6 7.1 13.1 1.8 −1.9 −22.6 8.7 27.3 −9.7 −5.7 16.0 20.3 20.6
24.8 35.9 39.1 18.6 13.1 2.8 −0.6 8.2 21.3 31.3 6.4 −3.6 −13.2 11.8 19.5 −7.0 −2.1 12.7 17.3 8.0
15.2 15.3 16.7 16.2 15.9 14.5 13.7 14.5 14.2 13.5 12.7 12.3 10.7 11.2 11.3 10.6 9.9 9.5 9.1 8.9
3.5 3.9 4.6 5.0 5.0 4.4 4.0 4.4 4.6 5.1 5.1 4.7 4.0 4.2 4.6 4.4 4.1 4.0 4.0 3.9
Notes: a Primary products cover SITC 0 (Food and live animals), SITC 1 (beverages and tobacco), SITC 2 (crude materials). SITC 3 (mineral fuels, lubricates and other materials), SITC 4 (animal and vegetable oil, and fats), SITC 68 (non-ferrous metal). b Mfg. (manufacturing) is defined to be the sum of SITC 5 (chemicals and related products), SITC 6 (manufactured goods by materials) less SITC 68 (non-ferrous metals), SITC 7 (Machinery and transport equipment) and SITC 8 (Miscellaneous manufactured). Source:
3.3
Compiled from the UN Comtrade database.
TRENDS AND PATTERNS OF FRAGMENTATION TRADE
World exports of components have grown from US$269 billion in 1988–89 to over US$1600 billion in 2004–05 (Table 3.2). The corresponding shares in the world manufacturing exports increased from 20 per cent in 1988–89 to 24 per cent in 2004–05. Component exports have also been growing
Production fragmentation and trade patterns in Japan
35
faster than total manufacturing exports. All the evidence underlines the point that fragmentation trade has become an ever more important element in global trade, and that trade flow analysis can therefore easily be misguided without taking account of the production fragmentation process in the data.2 Developed countries still export the bulk of components (over US$1000 billion in 2004–05), accounting for around 60 per cent of world components exports. However, the developed countries’ share has continuously declined from 92 per cent in 1988–89. In contrast, the corresponding share of developing countries in world component exports has increased sharply from less than 9 per cent in 1988–89 to 37 per cent by 2004–05. The total value of exports from developing countries amounted to US$600 billion in 2004–05. In line with the world trend, Japan’s component exports also dramatically increased from US$62 billion in 1988–89 to US$134 billion in 2000–01 and US$233 billion in 2004–05. The share of components in Japan’s manufacturing exports also rose from 24 per cent in 1988–89 to 33 per cent in 2000–01 and 31 per cent in 2004–05. It appears that the expansion of Japan’s component exports experienced a slowdown in the 2000s. The slowing down of dynamism in Japan’s component exports in the 2000s has been made more explicit by its share in the world component exports. The position of Japan in the world component exports dramatically declined from 24 per cent in 1988–89 to 11.7 per cent in 2000–01 and then to 10.3 per cent in 2004–05. Like Japan, the USA still exports the bulk of components, but its dynamism has slowed down in the 2000s. For example, the USA’s exports [was US$233 billion] in 2004–05, accounting for 14 per cent of the world’s component exports. However, this share has persistently declined from 16 per cent in 1988–89 and 20 per cent in 2000–01. In terms of the share of components in US manufacturing exports in 2004–05, it has remained around 33 per cent since 1988–89. On the import side, a similar pattern can be observed. The position of developed countries has been declining in world component imports, as developing countries have grown rapidly (Table 3.2b). In 2004–05 developed countries were still important, importing US$887 billion. However, the share of developed countries in world component imports has declined sharply from 86 per cent in 1988–89 to 54 per cent in 2004–05. In contrast, the share of developing countries rose substantially from 13 per cent in 1988–89 to 44 per cent in 2004–05. Japan’s position in world component imports has been relatively small, accounting for about 4.2 per cent for the period of 1988–2005. However, the evidence suggests the increasing importance of components in Japan’s manufacturing imports. For instance, the share of components
36
Exports Share of Parts and Components (PCs) in Mfg Exports (%)
World PCs Exports Share (%)
Growth of PCs Exports (%)
Growth of Mfg Exports (%)
134 231 56 106 57 15 255 34 48 39 55 62
62 42 26 76 26 7 19
0 1
8 5 5
69 90 73
113 96
168 233 74 180 63 22 464
13.8 30.3 34.3
1.6 9.3
24.0 33.4 20.5 21.3 21.7 17.5 20.1
27.0 49.5 42.5
15.0 25.4
32.5 36.8 22.8 22.3 24.6 23.3 29.4
28.3 51.5 42.0
18.1 36.2
31.4 33.1 21.5 22.4 22.5 22.2 29.8
2.9 1.8 1.8
0.0 0.0
23.1 15.7 9.7 0.0 9.6 2.8 6.9
3.4 4.8 5.4
3.0 4.1
11.7 20.2 4.8 9.3 4.9 1.3 22.3
4.2 5.5 4.5
6.9 5.8
10.3 14.2 4.5 10.9 3.9 1.4 28.3
12.8 17.7 16.3
63.0 45.0
5.7 9.9 6.0 4.9 5.1 6.3 19.5
8.4 14.3 15.0
25.0 19.0
4.1 10.0 5.7 4.6 4.9 4.9 17.0
1988–89 2000–01 2004–05 1988–89 2000–01 2004–05 1988–89 2000–01 2004–05 1988–2005 1988–2005
Values of Parts and Components (PCs) Exports, US$ billion
International comparison of fragmentation trade, 1988/89–2004/05
Japan USA France Germanya United Kingdom Sweden Developing East Asiab China China, Hong Kong SAR Rep. of Korea Singapore AFTAc
(a)
Table 3.2
37
South Asia Oceania European Union (EU) Europe East South America Developed Countries Developing Countries World
2 4 376
25 7 787
337
1146
1 2 173
1 1 247
23
269
1641
607
62 11 1002
5 5 553
20.3
18.1
4.8 15.3 20.7
5.5 19.8 18.1
25.4
25.7
24.9 12.8 26.1
5.6 21.4 20.7
24.1
26.0
26.5 12.4 24.0
6.8 17.6 19.8
100
8.4
0.2 0.5 91.5
0.2 0.6 64.3
100
29.4
2.2 0.6 68.6
0.2 0.4 32.8
100
37.0
3.8 0.7 61.0
0.3 0.3 33.7
10.6
20.0
29.4 12.0 8.1
12.2 6.6 6.7
9.5
17.6
17.7 13.3 7.2
10.9 7.3 6.1
38
Imports Share of Parts and Components (PCs) in Mfg Imports (%)
World PCs Imports Share (%)
Growth of PCs Imports (%)
Growth of Mfg Imports (%)
52 211 53 87 62 14 272
59 53
35 52 56
11 42 23 51 30 7 27
2 4
10 6 7
48 78 79
179 98
69 235 72 132 70 19 506
27.6 32.9 32.8
15.9 20.0
13.4 23.0 17.4 18.0 20.3 19.7 29.4
37.7 51.2 46.4
32.6 28.2
25.2 22.5 23.1 25.0 22.6 28.4 37.4
32.4 54.1 51.3
38.8 37.3
25.9 19.8 21.1 25.0 19.2 24.5 40.7
3.9 2.3 2.6
0.0 0.0
4.2 16.2 9.0 0.0 11.6 2.8 10.5
3.0 4.5 4.9
5.1 4.6
4.5 18.3 4.6 7.6 5.4 1.2 23.6
2.9 4.7 4.8
10.8 5.9
4.2 14.2 4.4 7.9 4.2 1.2 30.5
9.1 15.4 14.5
45.0 35.0
10.8 10.0 6.5 5.4 4.8 5.5 17.6
8.1 12.2 11.7
21.0 15.0
6.8 10.9 5.3 3.5 5.1 4.2 15.5
1988–89 2000–01 2004–05 1988–89 2000–01 2004–05 1988–89 2000–01 2004–05 1988–2005 1988–2005
Values of Parts and Components (PCs) Imports, US$ billion
(continued)
Japan USA France Germanya United Kingdom Sweden Developing East Asiab China China, Hong Kong SAR Rep. of Korea Singapore AFTAc
(b)
Table 3.2
39
5 11 369
30 23 702
419
1149
3 6 160
0 2 224
34
260
1658
731
61 29 887
11 15 507
19.8
26.3
6.5 28.5 19.2
22.2 16.2 18.3
25.0
30.1
24.7 20.8 23.0
15.5 17.4 22.4
24.1
31.4
24.8 20.6 20.8
15.3 13.8 20.3
100
13.1
0.2 0.6 86.2
1.1 2.3 61.4
100
36.5
2.6 2.0 61.1
0.4 1.0 32.1
100
44.1
3.7 1.7 53.5
0.7 0.9 30.6
10.8
18.6
30.9 17.4 7.9
7.8 5.2 6.6
Source:
Compiled from the UN Comtrade database.
9.6
17.4
21.4 19.5 7.5
10.0 6.2 6.0
Notes: a 1991–92 data used for 1988–89 for Germany. b Developing East Asia is defined to include South Korea, Singapore, Taiwan, Hong Kong, China, Indonesia, Malaysia, the Philippines, Thailand and Vietnam. c ASEAN Free Trade Area.
South Asia Oceania European Union 15 Europe East South America Developed Countries Developing Countries World
40
International fragmentation of production
in manufacturing imports has substantially increased from 13.4 per cent in 1988–89 to 25 per cent in 2000–01 and to 26 per cent in 2004–05 (Table 3.2b). During the same period annual growth rates of component imports have been recorded as 10.8 per cent, although total manufacturing imports have grown only by about 7 per cent. In contrast to Japan, the importance of components in US manufacturing imports has been declining since 1988. This contrasting pattern between Japan and the USA is highlighted in Figure 3.1. The share of components in Japan’s manufacturing trade increased sharply from the mid-1980s, although the 2000s experienced stagnation. On the other hand, the share of components in manufacturing exports and imports in US manufacturing trade has remained almost constant over the same period. The share of components in US manufacturing imports has only increased slightly, by around 35 per cent, whereas the share of components in US manufacturing exports has increased marginally since 1978. The importance of components in US manufacturing exports and imports has declined in the last five years (Figure 3.1). 3.3.1
Composition of Fragmentation Trade
Table 3.3 presents data on the composition of fragmentation trade at the SITC two-digit level in 2005. The world components trade is predominantly concentrated in four major product groups: power-generating (SITC 71), office machines (SITC 75), electrical machinery (SITC 77) and road vehicles (SITC 78). Among them, electrical machinery (STC 77) alone accounted for close to 35 per cent of components exports and over 35 per cent of components imports in world trade in 2005. These four product groups are important for both developed and developing countries. However, some variations can be observed. For example, developing countries have more concentration of electronic machinery (SITC 77), accounting for 44 per cent in exports and 52 per cent in imports. In contrast, the product composition for developed countries is much more balanced. While electrical machinery occupies the largest share in this composition, accounting for 29 per cent, road vehicles occupy 17 per cent of the share, and power generating and office machinery accounts for more than 10 per cent. Japan’s product composition is broadly similar to that of the USA. However, there are some notable differences between Japan and the USA. For instance, on the imports side, electrical machinery (SITC 77) is by far the most important item, reaching 45 per cent of the share for Japan. The same share is significantly lower for the US components trade (27 per cent). On the other hand, the share of imports of components for road
Production fragmentation and trade patterns in Japan
41
35 Share of PCs in Mfg Imports Share of PCs in Mfg Exports
30
25
20
15
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
10
40
35
30
25
20
15
Share of PCs in Mfg Imports Share of PCs in Mfg Exports
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
10
Year
Figure 3.1
Share of manufacturing (Mfg) in total exports and total imports, 1962–2005
vehicles (SITC 78) is 18 per cent for the USA and only 5.7 per cent for Japan. These findings suggest a different degree of importance of fragmentation trade between electronics machinery and the automobile industry for Japan and the USA. It is consistent with the view expressed in Chapter 2 that industry profile is deeply associated with the degree of development of fragmentation trade.
42
International fragmentation of production
Table 3.3
Composition of fragmentation trade (%) in 2005 Exports
SITC Description 7 Machinery and transport 71 Power generating 72 Specialized industry machinery 73 Metalworking machinery 74 General industrial machinery 75 Office machinery 76 Telecom and sound recording 77 Electrical machinery 78 Road vehicles 79 Other transport equipment 8 Miscellaneous manufactured 82 Furniture 87 Professional scientific 88 Photographic apparatus Total
Japan USA Developed Developing World Countries Countries 94.7 8.6 2.7
95.5 14.3 5.6
95.6 13.9 5.5
95.9 5.7 2.2
95.7 10.9 4.4
1.1 0.9 8.6 7.7 10.1 8.4 7.8 4.8 39.7 32.3 15.4 14.0 0.8 7.4 5.3 4.5 0.4 1.0 2.8 2.7 1.9 0.5 100 100
1.4 11.5 7.8 5.6 28.8 17.2 4.0 4.4 1.0 2.2 0.8 100
0.5 5.4 16.2 12.3 44.1 8.4 1.2 4.1 1.4 1.2 1.1 100
1.1 9.5 10.7 7.9 34.0 14.3 3.0 4.3 1.2 1.8 0.9 100
Imports SITC Description 7 Machinery and transport 71 Power generating 72 Specialized industry machinery 73 Metalworking machinery 74 General industrial machinery 75 Office machinery 76 Telecom and sound recording 77 Electrical machinery 78 Road vehicles 79 Other transport equipment 8 Miscellaneous manufactured 82 Furniture 84 Apparel 87 Professional scientific 88 Photographic apparatus Total Source:
Japan USA Developed Developing World Countries Countries 93.8 7.5 3.0
94.6 12.1 3.4
95.3 12.7 4.2
96.7 6.1 2.9
95.9 9.9 3.7
0.8 0.9 7.0 10.3 13.5 13.5 9.9 6.9 44.6 26.8 5.7 18.1 1.8 2.5 6.2 5.4 1.1 2.7 0.0 0.0 3.1 1.5 1.6 0.7 100 100
1.0 11 11.5 6.1 26.6 18.5 3.6 4.7 1.8 0.0 1.7 0.8 100
0.8 6.7 10.9 8.7 52.1 7.2 1.3 3.3 0.5 0.0 1.6 1.0 100
0.9 9.3 11.1 7.2 37.2 14.0 2.5 4.1 1.2 0.0 1.6 0.8 100
Compiled from the UN Comtrade database.
Production fragmentation and trade patterns in Japan
43
In summary, this section has highlighted the involvement of fragmentation trade for Japan in the wider global context. In line with other developed countries, components exports and imports have been becoming important elements for Japan’s manufacturing trade. In comparison with the USA, fragmentation trade for Japan has exhibited a much more dynamic pattern. In particular, components imports for Japan have increased substantially over the last two decades, seen in its impressive annual growth rate.
3.4
DIRECTION OF FRAGMENTATION TRADE
This section looks at the direction of fragmentation trade with the emphasis on two major machinery product groups, electronics machinery and automotive. Table 3.4 shows the geographical distribution of components and final goods trade for both Japan and the USA for the period 1988/89– 2004/05. Geographical location has been divided into six country groups: East Asia, China, North America, EU 15, developed and developing countries. Final products are defined as differences between total products and components. For instance, within automotive (SITC 78) motor car engines, engines and gear boxes are treated as components. Passenger cars, diesel buses, motorcycles and trucks are defined as final goods. Electronics products are further disaggregated into three items: office machines (SITC 75), telecommunications and sound recording (SITC 76), and electrical machinery (SITC 77) and components and final goods are also identified by each of two-digit items. 3.4.1
Components
The bulk of Japan’s component exports go to East Asian countries, accounting for about half of the exports share in 2004–05 (Table 3.4a). In particular, the rise of China as a destination for final assembly during this period has been remarkable. In 1988–89 China only accounted for a 1.9 per cent share of Japan’s component exports, but this increased substantially to around 15 per cent in 2004–05. Over the same period Japan’s component exports have shifted way from North America. In 1988–89 North America accounted for 57 per cent of Japan’s component exports, but the share declined to 22 per cent in 2004–05. The same declining share can be observed for the EU, but the magnitude is less than that of the USA from 17 per cent to 14 per cent over the same period. The shift in the direction of fragmentation trade from North America to East Asian countries, particularly to China, is consistent with the location choices of Japanese MNEs. As will be shown in Chapter 7, the overseas
44
Parts and components Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles Final goods Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles (2) Imports Parts and components Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles Final goods Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles
53.7 40.9 51.7 74.0 30.2 42.0 24.9 20.9 56.3 13.1 65.5 82.7 86.1 75.4 47.2 57.6 90.1 87.1 71.9 50.1
24.6 29.9 56.4 25.0 18.9 33.7 74.1 67.2 33.4 14.4
2004– 05
32.3 11.0 34.4 47.2 19.9 25.9 9.3 13.1 32.9 12.6
1988– 89
East Asia
0.6 0.1 0.7 0.2 0.1 7.0 6.6 4.2 1.6 0.3
1.9 0.2 1.6 3.8 0.7 3.4 1.3 1.9 4.5 1.5
1988– 89
24.8 45.0 45.5 18.0 18.6 36.7 63.7 55.1 46.3 28.1
15.3 15.4 21.8 18.2 9.8 11.7 5.9 5.0 16.5 0.7
2004– 05
China
66.1 60.4 39.3 59.2 28.0 28.5 14.6 27.3 40.3 42.5
56.9 54.9 26.8 27.8 48.1 25.9 46.0 42.6 28.6 37.5
1988– 89
15.5 11.9 9.4 16.7 15.6 31.2 2.7 7.3 14.9 22.8
21.9 29.1 17.1 11.4 40.1 16.0 30.2 35.1 18.7 23.7
2004– 05
North America
16.1 6.7 2.5 14.3 46.0 29.4 10.0 4.4 23.7 42.1
17.3 29.5 27.6 15.9 13.3 20.2 32.1 32.1 22.4 17.7
1988– 89
10.6 3.4 2.6 6.2 29.5 19.7 6.2 4.2 10.8 25.8
13.7 27.8 15.5 8.9 14.0 14.3 32.6 31.2 14.4 19.0
2004– 05
EU
67.8 66.0 41.5 69.3 57.3 49.7 20.3 31.6 53.2 56.3
54.0 77.6 47.5 39.9 64.4 55.8 74.0 69.6 48.7 61.5
1988– 89
32.2 16.0 12.2 22.9 45.6 38.6 9.5 12.2 26.8 49.2
37.6 57.5 32.5 20.2 55.8 43.7 63.1 69.4 34.0 55.1
2004– 05
Developed Countries
Geographical distribution of parts and components and final machinery trade, 1988/89–2004/05
Japan
(1) Exports
(a)
Table 3.4
25.2 32.2 57.8 25.2 18.6 37.0 74.1 67.5 33.5 14.7
37.5 13.1 41.4 53.0 26.2 34.7 13.9 19.5 42.7 31.7
1988– 89
66.0 83.8 86.8 75.0 49.9 58.5 88.5 85.0 70.3 50.6
59.5 41.0 65.9 77.5 39.7 53.4 33.8 29.8 63.4 40.7
2004– 05
Developing Countries
45
USA
28.0 28.2 24.6 48.5 6.0 20.1 11.8 19.1 25.0 4.0 45.4 84.0 58.8 54.8 34.5 44.8 82.1 81.9 52.5 18.6
46.8 70.8 55.6 64.8 38.2 50.1 84.6 91.2 52.6 35.5
2004– 05
24.2 26.8 22.5 40.9 5.1 25.2 15.3 28.7 23.7 6.4
1988– 89
East Asia
0.4 0.1 1.3 0.3 0.1 3.9 1.7 5.9 5.7 0.0
1.0 0.2 0.7 0.4 0.1 1.7 0.2 1.3 1.0 0.8
1988– 89
11.8 33.6 16.2 10.5 5.5 22.2 49.5 43.8 32.0 3.6
4.0 5.0 4.2 5.8 1.2 3.9 2.0 4.2 4.5 0.7
2004– 05
China
27.3 14.1 8.2 8.1 36.2 19.0 1.2 2.5 5.9 59.8
31.9 10.8 10.0 17.9 59.2 24.4 19.0 10.0 15.0 63.4
1988– 89
29.4 2.5 9.7 6.2 27.8 22.6 5.7 3.4 6.0 43.2
37.7 12.0 19.6 13.3 57.5 36.8 27.0 20.3 19.9 63.0
2004– 05
North America
21.8 10.5 7.0 12.1 15.9 11.5 12.6 2.7 21.3 2.2
27.5 46.7 18.9 19.3 9.7 29.2 40.2 28.3 30.4 9.1
Source:
Compiled from the UN Comtrade database.
19.3 6.7 7.5 12.3 14.5 23.3 6.9 4.5 15.9 7.1
18.9 23.5 21.5 10.9 9.6 22.5 19.0 17.0 22.1 5.7
2004– 05
EU 1988– 89
Note: The sum of developed and developing countries does not necessarily add up to 100 per cent.
Parts and components Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles Final goods Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles (2) - Imports Parts and components Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles Final goods Electronics: Office machines Telecommunications Electrical machinery Automobile, road vehicles
(1) - Exports
(b)
65.7 68.4 46.5 47.1 81.8 58.0 71.9 57.5 43.0 96.8
56.8 64.4 43.3 44.1 72.1 59.2 67.1 52.9 52.4 76.9
1988– 89
50.2 26.6 29.8 32.6 65.7 46.2 35.3 18.5 32.4 64.0
48.6 41.1 52.4 29.3 70.8 56.0 51.1 45.7 52.6 73.2
2004– 05
Developed Countries
27.4 29.4 51.8 48.1 12.3 34.6 26.1 41.9 47.0 2.7
36.7 24.1 52.3 51.1 24.6 33.0 27.0 41.4 38.6 18.8
1988– 89
49.0 73.1 68.8 66.3 34.0 52.0 63.9 79.9 66.9 35.6
50.1 58.0 46.4 70.1 28.6 40.6 47.6 52.8 45.7 24.2
2004– 05
Developing Countries
46
International fragmentation of production
operations of Japanese manufacturing MNEs have become increasingly concentrated in East Asian countries. Despite this similarity in overall trends, there has been some notable difference in the patterns of fragmentation trade between electronics machinery and automotive (Table 3.4a). While electronic components have declined significantly over the period 1988–2005, around 40 per cent of Japan’s automotive component exports still continued to be directed towards North America in 2004–05. In addition, the share of the EU in Japan’s component exports has increased slightly from 13.3 per cent in 1988–89 to 14 per cent in 2004–05. This pattern is possibly underpinned by the long period of operations by Japan’s auto producers in North America (mainly the USA) and the EU. Perhaps Japan’s automotive parts exports are mainly used to facilitate the production of Japanese automobiles in North America and the EU. On the import side, a similar pattern can be observed. That is, there has been a notable shift of Japan’s component imports from North America and the EU towards East Asia. Integration of Japan with East Asian countries appears to be much stronger on the imports side than for exports. The share of East Asia in Japan’s component imports has grown from 25 per cent in 1988–89 to 66 per cent in 2004–05 (Table 3.4a). More strikingly, China accounted for 25 per cent in Japan’s component imports in 2004–05, up from less than 1 per cent in 1988–89. As in the case of exports, the share of North America in Japan’s component imports declined substantially from 66 per cent in 1988–89 to 16 per cent in 2004–05. In contrast to the export side, Japan’s automotive component imports have also shifted from North America and the EU to East Asian countries. For instance, the share of the EU in Japan’s component imports declined from 46 per cent in 1988–89 down to 30 per cent in 2004–05. The same share for North America also went down from 28 per cent in 1988–89 to 16 per cent in 2004–05. East Asia has thus become the primal source of Japan’s component imports. 3.4.2
Final Goods
The overall pattern of the final goods trade for Japan is quite similar to what has been found for the direction of fragmentation trade. However, there are some key differences between electronics and automotive products. Japan’s final goods exports in automotives (for example, passenger motors, motor cycles and trucks) are still largely influenced by demand in the large markets. Japan’s automotive final goods exports to East Asian countries remained at a constant 13 per cent between 1988 and 2005 (Table 3.4a). Japan’s automotive final goods exports to China even
Production fragmentation and trade patterns in Japan
47
declined slightly, to 0.7 per cent in 2004–05 from 1.5 per cent in 1988–89. Interestingly, the North American share in Japan’s final goods exports in automotives has also declined substantially, from 38 per cent in 1988–89 to 24 per cent in 2004–05, but the EU’s share has risen marginally from 18 to 19 per cent during the same period. On the import side, final electronics goods have predominantly come from East Asia accounting for 74 and 67 per cent of Japan’s final goods imports in office machines and telecommunications, respectively, in 1988–89, increasing further to 90 per cent and 87 per cent by 2004–05 (Table 3.4a). China’s share in Japan’s final goods imports in automotives increased significantly from less than 1 per cent in 1988–89 to 28 per cent in 2004–05. The bulk of this increase from China is most probably due to an increase in imports of motorcycles to Japan. 3.4.3
A Comparison with the USA
The direction of fragmentation trade for the USA is markedly different from that of Japan. While the fragmentation trade is predominantly concentrated in East Asia, in contrast with the Japanese experience, there has not been much of an increase since 1988–89 (Table 3.4b). US component exports to East Asia increased only a little, from 24 per cent in 1988–89 to 28 per cent in 2004–05. Over the same period the share of countries in North America (Canada and Mexico) in US component exports has marginally increased from 32 per cent in 1988–89 to 38 per cent in 2004–05 (Table 3.4b). In particular, the USA’s component exports in office and telecom machinery to Canada and Mexico increased from 11 per cent and 10 per cent in 1988–89 to 12 per cent and 20 per cent in 2004–05, respectively. Even before the North American Free Trade Agreement (NAFTA) became effective in 1994, there was already a high proportion of US automotive components exported towards Canada and Mexico. This location choice is consistent with the value-toweight explanation (Lall et al., 2004 and Chapter 2). The export and import of automotive parts are more constrained than for electronics, because these items (for example, body parts, vehicle bumpers and vehicle engines) entail relatively higher transportation costs due to the heavy weight. As a result, the assembly and processing locations of automotive parts have naturally become geographically proximate to the USA in order to minimize transportation costs. However, it is interesting to stress the point that the valueto-weight explanation does not seem to explain a higher share of Japan’s automotive components to North America (Table 3.4a). On the imports side, the position of East Asia in total US components has changed very little, accounting for around 46 per cent since 1988 (Table
48
International fragmentation of production
3.4b). However, the product disaggregation reveals a change of components trade patterns. For example, in 2004–05 the share of East Asia in office machine component imports increased from 70 per cent in 1988–89 to 84 per cent in 2004–05 and the same share for components of telecom remained around 58 per cent for the same period. The share of East Asia in US electrical machinery declined from 65 per cent in 1988–89 to 55 per cent in 2004–05. The rise of China as an electronics exporter to the USA has been spectacular. In 1988–89 China’s share in US parts imports in office machines and telecom accounted for only 0.1 per cent and 1.3 per cent, respectively, but this had increased substantially to 34 per cent and 16 per cent by 2004–05. With regard to the USA final goods trade, the share of East Asia in US exports has actually declined from 50 per cent to 45 per cent over the period 1988–2005. The share of East Asia for the USA’s final goods exports in office machines and telecom accounted for 15 per cent and 29 per cent, respectively, in 1988–89, but had declined to 12 per cent and 19 per cent by 2004–05. On the import side, the rise of China as the main source of US final goods imports has been very impressive. China only accounted for around 3 per cent in 1988–89 in US final goods imports, but the share had increased to 22 per cent in 2004–05. In summary, Table 3.4 presents an interesting pattern of the direction of fragmentation trade for Japan and the USA. Consistent with total manufacturing trade, the direction of Japan’s fragmentation trade has been shifting towards East Asian countries. In particular, the recent rise of China as the location of final assembly has been remarkable through importing the bulk of components from Japan, especially electronic components. One of the key findings of this chapter is that while the direction of fragmentation trade in electronics products has mainly been shifting towards East Asian countries, those of automotive products are still directed towards North America and the EU. The pattern of direction of the fragmentation trade might be closely linked with the geographic concentration of the overseas operation of Japanese MNEs. As suggested in Chapter 2, the characteristics of industry and products might also be largely influencing the direction pattern of the fragmentation trade. The next chapter provides a systematic empirical examination of the underlying factors determining the extent of fragmentation trade for Japan and the USA, using the gravity model of trade flows.
3.5
CONCLUDING REMARKS
This chapter has surveyed the trends and patterns of fragmentation trade in the context of overall development of Japan’s manufacturing trade.
Production fragmentation and trade patterns in Japan
49
There has been a noticeable ongoing shift of location of the production processes from developed countries to developing countries, although developed countries are still the major players in fragmentation trade. The evidence points to a rapid increase in the fragmentation trade within Japanese manufacturing since the mid-1980s. In particular, the increasing importance of components in Japan’s manufacturing imports has been noteworthy. This contrasts with the US experience where there have been virtually no rapid changes in terms of the share of components in total manufacturing trade. Japan’s components trade with East Asian countries has also increased dramatically. In particular, the involvement of China in Japan’s fragmentation trade has been a remarkable phenomenon. The direction of fragmentation trade follows some interesting patterns, especially in relation to the two major machinery products, electronics machinery and automotives. It is found that while Japan’s fragmentation trade in electronic machinery has been shifting towards East Asia, automotive components are still largely shipped to North America and the EU where the bigger markets exist. This pattern is perhaps influenced by the overseas operations of Japanese MNEs’ foreign affiliates. These findings suggest the importance of controlling for the location of MNEs’ foreign affiliates, industry and country characteristics such as relative wage rates, market size and geographical distance in the empirical analysis of the determinants of inter-country differences in fragmentation trade.
NOTES 1. The trade data in the United Nation (UN) Comtrade online database (http://comtrade. un.org/db/) are available from the year 1962. The year 2005 is the latest end point available at the time of writing. 2. This point was forcefully demonstrated in Athukorala (2005) and Athukorala and Yamashita (2006) by focusing on the conventional method of computing trade integration.
4
Determinants of fragmentation trade
4.1
INTRODUCTION
This chapter undertakes an econometric analysis of the determinants of fragmentation trade in Japan, using panel data covering the period from 1988 to 2005. To the best of the author’s knowledge this is the first systematic analysis of the Japanese experience.1 Existing studies, including Baldone et al. (2001) and Egger and Egger (2005), focus on specific country groups such as the EU. Jones et al. (2004) and Athukorala and Yamashita (2008) examine production fragmentation at the global level, while Egger and Egger (2003) focus on the Austrian experience and Görg (2000) and Swenson (2005, 2007) examine the case of the USA. This chapter extends the existing literature in a number of aspects. First, an econometric analysis uses a comparative approach between Japan and the USA. As noted in Section 3.3, there are some visible differences in the patterns of fragmentation trade between Japan and the USA. In addition, the business management literature indicates that cross-border production networks appear to be different between Japanese and US MNEs (see Section 2.3.3). Hence, the relative importance of factors affecting the extent of fragmentation trade between Japan and the USA can be different. Second, few econometric studies have been done so far to exploit trade data relating to parts and components as a measure of fragmentation trade apart from Jones et al. (2004) and Athukorala and Yamashita (2006), as discussed in Section 2.4. This chapter extends their approach and constructs three dimensional panel data sets (industry, country and year). Third, the analytical framework extends the standard gravity model of trade flows by incorporating a set of industry-specific variables (labour costs and productivity) as well as country characteristics (market size, the level of infrastructure and the quality of governance). While these factors are important in the theoretical analysis, they have not yet fully been incorporated in the empirical examinations. This chapter is organized as follows. The next section reviews the existing empirical literature and then Section 4.3 depicts the model specification, providing the theoretical reasoning and a brief discussion on
50
Determinants of fragmentation trade
51
variables used for regression analysis. Data and the econometrics procedure is discussed in Section 4.4. Finally, the results and highlights of the main findings are drawn together in Sections 4.5 and 4.6.
4.2
EMPIRICAL EVIDENCE
The existing empirical studies have employed two alternative data sources for examining the determinants of production fragmentation: trade data in parts and components and trade data collected under the special tax exemption scheme for overseas assembling (for example, the US Offshore Assembly Programme (OAP) or Processing Trade Scheme of the EU).2 Jones et al. (2004) and Athukorala and Yamashita (2006) examined fragmentation trade in a wider global context. The former study examined the importance of changes in service link costs for fragmentation trade whereas the latter examined the implications of production fragmentation for the regional and global trade pattern with an emphasis on East Asian countries. These studies use trade in parts and components as a proxy of fragmentation trade. Görg (2000), Baldone et al. (2001) and Egger and Egger (2005) extracted the data from EU official records on inward processing trade/outward processing trade (IPT/OPT) statistics (see Chapter 2 for further details on the IPT/OPT data). Görg (2000) provides one of the early examinations of fragmentation trade using the US inward processing trade in the EU. Eurostats record trade transactions, which are imported for processing purposes in the EU territory from the USA and then re-exported to the USA for sale and final assembly. The main focus of Görg (2000) is to examine the development of the US inward processing trade, IPT, in 12 EU countries for the period 1988–94. This type of US processing trade has recently grown in EU periphery countries. Ireland especially has become one of the major locations for US offshore assembly in the EU territory. Görg (2000) examines the relative importance of three principal variables determining US processing imports in a cross-country and crossindustry panel data set for the 12 EU countries, using the IPT data. They are the US FDI stock data, manufacturing wage costs and the measure of comparative advantage. The main finding is consistent with the trade theory: the pattern of US inward processing imports mostly accords with the comparative advantages suggested by the standard theory. In EU periphery countries the stock of US FDI was found to be strongly related to the US inward processing trade. However, odd results are found for relative wage variables: it seems that higher wages are positively associated with the US inward processing imports. In other words, higher wage
52
International fragmentation of production
rates make industry more attractive for US inward processing imports in EU countries. This finding is counter-intuitive at first sight. However, it is interpreted as higher relative wage rates being an indication of better-endowed skilled workers. Overall, Görg (2000) concluded that patterns of US processing imports in the EU territory are mainly driven by comparative advantage and the location of US MNEs. Building upon Görg (2000), Egger and Egger (2005) make use of the EU processing trade (IPT/OPT) from the Eurostat database for the period 1988 to 1999 to examine the determinants of bilateral processing trade flows of the 12 EU countries with over 40 countries. In the 12 EU country panel data sets there are four groups of explanatory variables considered: market size, relative factor endowments, relative price differentials and infrastructure variables. The general finding of Egger and Egger (2005) is that the relative cost variables and the level of infrastructure are the key determinants for explaining the variation in the bilateral EU processing trade, while market size is found to be of little importance. Swenson (2005, 2007) examines the determinants of the US fragmentation trade using data relating to the US Offshore Assembly Programme (OAP) for the period 1980 to 2000. Despite its importance in the initial period around the 1970s, the share of OAP imports in the total US imports has been declining over time and has gone down from 12 per cent over the period 1980 and 1990 to 8 per cent for 1991 and 2000 (Swenson, 2005). The focus of Swenson (2005, 2007) is slightly different from that of other studies in that she investigates the location choice of processing and assembly for the US OAP import data. The dependent variable in her regression is the country share in the overall US OAP imports. In other words, it examines the country distribution in total US OAP imports. One of the key contributions of this study is the construction of a country’s unit production costs as well as competitor costs, accounting for production techniques, transportation costs and tariff levels. Swenson (2005) tests how a country’s share in US OAP imports is influenced by changes in these cost measures, after controlling for market size, stage of development and proximity. The findings suggest that that country share in the US OAP imports, ceteris paribus, increases when its own costs fall or when its competitors’ costs rise. In addition, the degree of cost sensitivity is greatly different between developed and developing countries (Swenson, 2005). To sum up, many empirical studies find that fragmentation trade is largely explained by inter-country comparative cost differences as suggested by the standard international trade theory. The main limitation of existing studies is a lack of comparative study. As discussed in Chapter 3, there are various reasons why international outsourcing strategies can be different across countries.
Determinants of fragmentation trade
4.3
53
THE MODEL SPECIFICATION
The empirical analysis of this chapter is based on the gravity model of bilateral trade flows which has been the workhorse for the empirical research of international trade for many years (see Harrigan, 2003; Anderson and van Wincoop, 2004; Feenstra, 2004; Armstrong, 2007 for a survey of the gravity model). The dependent variable is trade flows in parts and components of machinery trade (see Section 2.4.4 for the data compilation method). The standard gravity model is then extended by incorporating the following variables informed by the theory of production fragmentation. Appendix 4.1 delineates the data source and variable construction. 4.3.1
Unit Labour Cost
Differences in wage rates between domestic and foreign country are frequently cited as a major factor driving the recent surge of fragmentation trade: the greater the difference in relative labour costs, the more incentive for industry to relocate the parts of the production stages that use this factor relatively intensively. Therefore, it is important to control for the level of foreign industry wages relative to home in order to gauge its deterministic role in trade in machinery parts. Rather than using the simple relative wage rates, unit labour costs (ULC) is employed. ULC is defined as the cost of labour required to produce one unit of output relative to that of Japan or the USA. By this construction, wages are adjusted for the level of productivity. The use of ULC is appealing in the present context because of its connection to the theory of comparative advantage. In its simple way, the Ricardian trade model states countries will specialize in products in which they have relatively lower unit labour requirements. This implies the role of both wage level and labour productivity. While overall differences in productivity determine the wage level, the level of education, skill endowments and efficient use of production workers influence the overall productivity performance. In relation to production fragmentation, lower wage countries become attractive for relocation of the production process, but the benefits of lower wages could be offset by lower labour productivity. In this respect, high-wage countries might become more conducive locations for production fragmentation due to higher labour productivity. 4.3.2
Institution and Infrastructure
The importance of the level of infrastructure (such as better ports and communication systems) might also become important determinants of
54
International fragmentation of production
production fragmentation (Jones and Kierzkowski, 1990; Jones, 2000; Grossman and Helpman, 2005). A country with a better infrastructure set-up is a preferred location of production fragmentation due to the reduction in service link costs (see Chapter 2 for the underlying theory). The model also incorporates the quality of the legal and institutional arrangement of countries. The model incorporates these accounts with two sets of variables. First, the model includes an indicator of the efficiency of trade facilitation at the port (denoted as Infra).3 For this purpose, the data on the trade fasciculation, ‘time for processing trade’ are employed. They purports to capture the overall quality of the physical infrastructure and efficiency in administrative procedures for processing trade. It is expected that better ports and a higher level of infrastructure quality requires less time for processing trade and can facilitate larger trade flows between countries. Hence, the expected sign is negative. In fact, the low infrastructure level becomes a major obstruction for some developing countries participating in global production sharing (Limao and Venables, 2001). While there are many developing countries with significantly lower wage levels, the international production network is less likely to be firmly developed if the low level of infrastructure quality significantly outweighs the benefits of access to cheaper labour. Second, a proxy for contract enforcement, property rights, investor protections, the political system, the legal system and the quality of governance, is included to represent institutional quality (denoted Contract) in the model (Anderson and Marcouiller, 2002; Nunn, 2007; Levchenko, 2008). It is hypothesized that better institutional arrangements lead to greater trade flow, especially technology-intensive trade in parts and components. Hence, the expected sign is positive. The institutional quality is relevant to the process of production fragmentation which involves establishing a complex contract relationship between two parties engaging in specific investment relationships, as compared to spot market transactions. In essence, expansion of production fragmentation will be limited if contracts are more incomplete due to the lower quality of contract governance. 4.3.3
Other Variables
Additionally, the real exchange rate (RER), a measure of the overall competitiveness of traded-goods production, is included. Once pooling data over time, cost competitiveness vis-à-vis real exchange rates (RER) become important (Soloaga and Winters, 2001; Elliot and Ikemoto, 2004). The higher RER index (that is, depreciation) is the cheaper are exports for the source country (that is, exporters). Hence, the expected sign is positive for the exports equation and negative for the imports equation.
Determinants of fragmentation trade
55
Both industry- and time-specific dummies are incorporated in order to guard against omitted variable accounts for the variation of fragmentation trade in the respective dimensions. The industry fixed effect controls for any unmeasurable (or unobserved) industry-specific time-invariant heterogeneity across industries. Time-specific effects are also included in a similar spirit for controlling a homogeneous form of technological change across industries, but varying across time and capturing other business cycle effects. Based on the discussion of the variables above, the full specification of the model can be written as follows: ln FRGj,k,t 5 b0 1 b1 ln GDPk,t 1 b2 ln GDPPk,t 1 b3 ln Distk 1 b4 ln ULCj,k,t 1 b5Contractk 1 b6Infrak 1 b7RERk,t 1 gj 1 tt 1 ej,k,t
(4.1)
where subscripts j, k, and t symbolise industry, country and time, respectively. A symbol ln before a variable denotes the natural logarithms. The variables are listed below with the postulated sign of the regression coefficient of each explanatory variable in brackets: FRG GDP GDPP Dist ULC Contract Infra RER g t e
4.4
Fragmentation trade (real components exports or imports) Real gross domestic products (+) Real GDP per capita (+) The geographical distance (–) The relative unit labour costs with respect to a foreign country (+) The quality of contract enforcement (+) The quality of trade facilitation at port (+) Real exchange rate (+ for exports, – for imports) Industry fixed effects Time-fixed effect Stochastic error term.
MEASUREMENT, DATA AND ESTIMATION PROCEDURE
One novelty of the empirical analysis in this chapter is that it makes use of a three-dimensional (country, industry and year) panel data set. It covers the 41 trade partner countries of Japan/the USA, which are selected based on the world share of machinery exports accounting for 0.1 per cent or more in 2000, and they represent over a 95 per cent share of Japan’s
56
International fragmentation of production
machinery exports in 2000 (see Table 4A.1 in Appendix 4.1 for a list of countries). The data set covers five machinery industries at three-digit International Standard Industry Classification (ISIC) (fabricated metal, industrial machinery, electronics machinery, transport equipment, and professional and scientific equipment) over the period 1988–2005. The data set was assembled from various data sources (see Table 4A.2 in Appendix 4.2 for a summary). The initial period is 1988, because this is the first year that the UN Comtrade database started reporting under SITC Revision 3, on which the commodity listing of parts and components in this study is based (see Chapter 2). The end point is 2005, since this was the latest year for which data for most of variables are available. The model is estimated using the random-effect technique. The alternative fixed-effect model is not appropriate because our model includes some time-invariant variables. Hence, the specification does not include country-specific fixed effects, which Anderson and Marcouiller (2002) suggest may be important. Inclusion of a country-specific time-invariant variable, namely distance may partially remedy the problem.4 There is a possible two-way causation between the trade flows and trade facilitation variable (Infra), because higher trade flows may lead to better trade facilitation (Wilson et al., 2003; Djankov et al., 2006). The HausamWu specification test is conducted to judge whether this causation is a problem in the data compiled in this chapter. The test result rejects the null hypothesis that there is causality running from trade flows to improvement in trade facilitation. In other words, the reverse causation from trade flows to better port infrastructure is not a concern in the data set. Indeed, this is quite consistent with observations made by an economic historian about port development leading into exports growth (Yasuba, 1978).5 Another relevant estimation issue is the potential truncation bias by the ‘zero’ records of bilateral trade flows. However, ‘zero’ bilateral trade in the data are less than 3 per cent out of the full sample. Therefore, the possible truncation bias is likely to be minimal, if not absent. Moreover, there is evidence that the results based on the sample selection bias correction do not change the basic results (Eichengreen and Irwin, 1995; Soloaga and Winters, 2001; Estevadeordal et al., 2003). Finally, to guard against possible violation of the homothetic residuals, the heteroscedasticity-consistent standard errors (that is, White correction) are used.
4.5
RESULTS
Table 4.1 presents the results for regressions of Japan and the USA. One of the most striking differences between Japan and the US regression is
Determinants of fragmentation trade
Table 4.1
57
Determinants of fragmentation trade in Japan and the US, 1988–2005 Japan Exports (1)
Imports (2)
USA Exports (3)
Dependent variable =real exports/imports of parts and components GDP 1.10*** 1.25*** 0.82*** (0.18) (0.15) (0.15) GDPP 0.21 −0.28 0.56 (0.49) (0.62) (0.37) Dist −1.63*** −2.08*** −0.30 (0.54) (0.62) (0.56) ULC −0.33*** −0.54** −0.06 (0.12) (0.24) (0.12) Contract −0.28 0.87 −0.58 (0.44) (0.68) (0.43) Infra −0.58 −1.06 −0.48 (0.68) (0.84) (0.54) RER −0.04 −0.43** 0.04 (0.09) (0.18) (0.06) Constant −7.44 −10.39 −11.66* (5.63) (8.73) (6.95) R-squared 0.69 0.68 0.66 Observations 1510 1486 1571
Imports (4) 1.04*** (0.14) −0.05 (0.39) −0.65 (0.50) −0.39** (0.19) 0.40 (0.55) −0.57 (0.57) −0.26*** (0.09) −18.12** (7.89) 0.68 1571
Note: Time and industry dummies are included, but the results are suppressed here. Standard errors based on White’s heteroscadasticity correction cluster by country are given in parentheses, with statistical significance (two-tailed test) denoted as: *** 1%, ** 5% and * 10%.
the estimated elasticity of Dist. Its estimated coefficient was found to be the expected negative sign with strong statistical significance for Japan, whereas it is hardly significant in the US regression. This finding indicates that, on average, trade costs related to geographical proximity are less of a concern for US firms. As discussed in Chapter 2, geographical closeness is one of the most important factors in maintaining the level of quality of international production. Hence, the finding may be a reflection of the predominant preference of Japanese firms for maintaining the quality level of international production. In addition, interestingly, the estimated coefficient of Dist for the imports equation is larger than that of the exports equation for Japan (equations (4.2) and (4.3) in Appendix 4.1). The estimated elasticity for distance is reported to be about 2 per cent or more for import regressions compared to less than 1.6 per cent in
58
International fragmentation of production
export regressions. This evidence indicates the relatively high sensitivity of machinery parts imports to changes in transportation costs. On the other hand, exports of machinery parts for the purpose of final assembly might require relatively less quality control. The other striking difference is that labour costs (proxied by ULC) are very important in terms of statistical significance for Japan but not for the USA. In Japanese regression a 10 per cent increase of unit labour costs would lead to around a 3–5 per cent decline of trade in parts and components. This implies that Japanese firms tend to locate to countries with relatively lower wage costs for the operations of international production. Combined together with the finding of Dist, it is found that the extent of fragmentation trade for Japan is largely driven by lower costs and geographical proximity. This is consistent with the finding of the rapid growth of Japan’s trade in parts and components with neighbouring East Asian countries, as found in Chapter 3. Contract variable enters the unexpected negative sign for the exports equation and the expected positive sign for the imports equation for both Japan and the USA, although it is not statistically significant. As will be discussed later, the results are improved once taking into account the sample division into developing and developed countries. Infra variable shows the expected negative sign both for Japan and the USA, although it is not statistically significant. Perhaps, these findings are driven by a high correction between per capita GDP (GDPP): a correlation coefficient between GDPP and Contract is reported as 0.83 and between GDPP and Infra is 0.77. To address this concern, we have run a separate regression retaining Contract and Infra but not GDPP. However, it turns out that there are virtually no changes in the statistical significance of variables (Contract and Infra) across different specifications. Table 4.2 presents results for the sample divided into developed and developing countries separately. The rationale for such a division of the sample is that in principle we can expect the strategy of international fragmentation to be different between operations in developed and developing countries. One of the interesting findings in Table 4.2 is that Dist variable for US regression restored statistical significance. In particular, the geographical proximity is very important for trade in parts and components with developed countries, whereas it is not the case for Japanese regression. Instead, Dist variable is only found to be statistically significant with the expected negative sign in Japanese regression for trade with developing countries. On the other hand, US firms seem to take the geographical factor into account when they only trade with developed countries. Contract variable shows some statistical significance once we divide the sample into developed and developing countries. In particular, for
59
Infra
Contract
ULC
Dist
GDPP
GDP
Table 4.2
0.98** (0.42) −0.62* (0.35) −0.91** (0.39) −0.90*** (0.24) 0.77 (0.48) −1.85** (0.94)
Developing countries (1)
Developed countries (6)
1.21*** (0.19) 0.59 (0.80) −1.63*** (0.15) −1.17*** (0.43) 0.41 (0.29) 0.13 (0.48)
Developing countries (5)
Imports
Dependent variable =real exports/imports of parts and components 1.39*** 0.47 1.23*** 1.16*** 1.09*** 0.95*** (0.09) (0.36) (0.18) (0.30) (0.16) (0.19) −1.05*** −0.86** 1.25 −0.11 −0.12 −0.57 (0.40) (0.35) (0.94) (0.28) (0.54) (0.42) 0.16 −2.01*** −1.05 −0.73* −1.69*** −1.14** (1.00) (0.50) (1.43) (0.44) (0.40) (0.53) −1.24*** −0.84* −1.47** −0.31 −1.11*** −0.51 (0.25) (0.49) (0.69) (0.24) (0.43) (0.35) 1.26*** 0.59 0.32 0.58* 1.09** 1.03* (0.33) (0.55) (0.33) (0.33) (0.52) (0.57) −0.31 −3.14*** −0.11 −1.44* 0.25 −2.34*** (0.32) (0.75) (0.54) (0.81) (0.29) (0.77)
Developed countries (4)
Exports
Developed countries (8)
Developing countries (3)
Imports
USA
Developing countries (7)
Developed countries (2)
Exports
Japan
Determinants of fragmentation trade with developed and developing countries in Japan and the USA, 1988–2005
60 0.76 (0.57) −44.05*** (11.88) 0.91 686
0.45* (0.24) 11.33 (18.89) 0.67 800
0.68 (0.73) −57.91** (22.90) 0.85 686
0.09 (0.12) −14.23 (16.44) 0.72 850
Developing countries (5) 1.21** (0.55) −31.20** (12.20) 0.89 721
Developed countries (6)
Exports
0.08 (0.20) −5.26 (12.65) 0.65 850
0.37 (0.45) −41.90*** (9.68) 0.90 721
Developed countries (8)
Imports Developing countries (7)
USA
Note: Time and industry dummies are included, but the results are suppressed here. Standard errors based on White’s heteroscadasticity correction cluster by country are given in parentheses, with statistical significance (two-tailed test) denoted as: *** 1%, ** 5% and * 10%.
0.33** (0.13) −14.17 (16.11) 0.68 824
Developed countries (4)
Imports Developing countries (3)
Japan
Developed countries (2)
Exports
Developing countries (1)
(continued)
R-squared Observations
Constant
RER
Table 4.2
Determinants of fragmentation trade
61
both Japan and the USA exports of parts and components to developed countries indicate that the contract environment can be a very important factor. This makes sense when the bulk of exports into developed countries are technology-intensive components. Hence, this indicates that both Japanese and US firms are very concerned about conducting the assembly operation of technologically-intensive components from Japan or the USA in a better governance environment in developed countries. Infra variable is also illustrated in Table 4.2. The result indicates that the poor level of infrastructure for processing trade can have a significant adverse effect on trade in parts and components, particularly trade with developing countries. A 1 per cent in delay of time for processing trade can depress about 2–3 per cent of trade with the USA and Japan (equations (4.3) and (4.7) in Appendix 4.2). In sum, the estimation results reveal some interesting contracts between Japan and the USA in relation to the relative importance of deterministic factors for fragmentation trade. The most striking difference is that geographical closeness is very important for explaining Japan’s fragmentation trade, especially trade with developing countries. Unit labour costs are also found to be very important, whereas that of US trade is found to be relatively less relevant. These findings reinforce the predominant position of East Asia as a location of production fragmentation for Japanese firms. The similarity is found to be their location choice partly driven by the level of governance quality and the quality of infrastructure, particularly when they trade with developed countries.
4.6
CONCLUSION
This chapter has examined the deterministic factors of fragmentation trade by a comparative analysis of Japan and the USA, using newly compiled three-dimensional panel data sets covering the period 1988–2005. The analytical tool used is the gravity model of trade flows; it is systematically extended by following the guidance of the theory of production fragmentation. It was found that Japanese and US firms’ decisions of trade in parts and components exhibits some differences, especially relating to cost variables such as transportation and labour costs. On the other hand, both firms consider the level of governance quality and the infrastructure level equally important. In particular, we found strong evidence once the sample is divided into developed and developing countries. To some extent, these findings imply a distinctive feature of the patterns of production fragmentation by Japanese firms. The patterns of Japan’s fragmentation trade are
62
International fragmentation of production
mainly driven by availability of lower wage countries and the geographical proximity, while that of USA’s trade does not follow a similar pattern. The underlying reason for such a distinction may be related to a Japanese firm’s outsourcing decision primarily based on comparative advantage motives. As discussed in Chapter 2, Japanese firms tend to relocate only labour-intensive production processes while retaining capital- and technology-intensive production processes in Japan. Hence, Japanese firms primarily use offshore production in order to reduce the overall cost of production. US firms, on the other hand, appear to care less about exploiting lower labour cost countries.
APPENDIX 4.1
DATA USED AND VARIABLES CONSTRUCTION
Trade in Parts and Components The dependent variable comprises trade in parts and components in the five listed machinery industries. Trade data are sourced from the UN Comtrade database. The values of trade flows from the UN Comtrade are originally expressed in nominal US$, and they are converted into the real series by using the export and import price indices. The export and import price indices (the base year is 2000) are collected from the Bank of Japan.6 They are genuine price indices, rather than unit value series which are frequently used as rough indicators for price proxies. Since these price indices are quoted on the basis of the Japanese yen, trade flow values (reported in US$) are first converted into Japanese yen using nominal exchange rates from the IMF (2006). Nominal trade flows in Japanese yen are then transformed into real terms using the exports and import price index. To compile the trade data at the industry level it is necessary to match the trade flow data reported by the SITC system with the International Standard Industry Classification (ISIC). However, both classification systems (that is, SITC and ISIC) are designed for a distinct purpose and are not easily compatible at any level of aggregation. Maskus (1991) developed the concordance table between the two-digit level of SITC Rev. 2 and the three-digit level of ISIC Rev. 2 (table 1.6, p. 42 in Maskus (1991)). The concordance table by Maskus (1991) includes the code mapping between SITC and ISIC as well as the corresponding weights used. Maskus’s table (1991) is employed because it was developed at a reasonably high level of aggregation (that is, the three-digit level). This level of aggregation makes it easy to link between the SITC and ISIC systems. However, a particular complication arises because Maskus’ (1991) concordance refers
Determinants of fragmentation trade
63
to SITC Rev. 2, while a list of parts and components is developed based on SITC Rev. 3. Hence, the five-digit level of SITC Rev. 3 is first mapped into SITC Rev. 2 by using the concordance table available at the classification registry at the Statistical Division of the United Nations.7 Then those five-digit items at SITC Rev. 2 are aggregated into a two-digit level to be consistent with Maskus (1991). Based on these procedures, data on trade in parts and components in SITC 7 (machinery and transport equipment) and SITC 8 (miscellaneous) are mapped into five three-digit ISIC Rev. 2 categories: ISIC 381 (fabricated metal products), ISIC 382 (machinery, except electrical), ISIC 383 (electric machinery), ISIC 384 (transport equipment) and ISIC 385 (professional and scientific equipment). Gravity Variables GDP and GDP per capita are expressed in constant US$ for year 2000. They are obtained from the online database of World Bank Development Indicators.8 The geographical distance is the straight distance (in kilometres) between the capital cities of the respective countries. The data series is derived from the online database of Joe Haveman’s International Trade Data Source.9 Unit Labour Costs (ULC) The construction of unit labour costs (ULC) requires data on employee compensation and output (value-added) measured at industry level. This measure of competitiveness is the defined cost of labour per unit of production. More specifically, ULC for the j-th industry, k-th country can be formulated as: ULCjk 5 (Wjk /Ejk) / (Qjk /Ejk)
(4.2)
where W is employee compensation (including non-wage costs), E is the number of employees and Q is output measured by value-added. Relative ULC for Japan is then computed by dividing Japan’s ULC by that of the corresponding trade partner country. The same formulation is applied for the USA. The data for ULC at the industry level are compiled from the electronic data files of the Annual Survey of US Direct Investment Abroad conducted by the Bureau of Economic Analysis (BEA), US Department of Commerce (henceforth called ‘the BEA data’ for short).10 This is the most comprehensive and consistent source of data available on international production by US MNEs (Lipsey, 2003). The BEA data include wage
64
International fragmentation of production
bills paid as well as valued-added produced by majority-owned11 foreign affiliates of US MNEs operating in various host countries. The biggest advantage of using the BEA wage data is the international compatibility as the data are collected though a mandatory survey conducted using the same questionnaire and methodology in all countries.12 While the BEA data feature the desirable property for computing ULC in a consistent manner, they do not reflect the true national level wages, salaries and output. In fact, it is most likely that these wage bills in the BEA data overestimate the actual ones in a host country. Many studies have found that foreign affiliates of MNEs pay much more than local firms (Lipsey, 2003). However, presumably this bias becomes less important when undertaking a comparison of wage levels across countries and over time. What is more important for measuring ULC is the consistency of measure across countries and over time in a given country. The BEA data meet this requirement. One unavoidable limitation is that there are no data for the USA. Therefore, the wage data for the US machinery industry is extracted from the online database maintained by the Bureau of Labour Statistics, US Department of Labour.13 Institutional Quality Variable The ‘Rule of Law’ indicator (denoted as Contract) is employed in order to capture the governance quality of trade partner countries (Kaufmann et al., 2007). This indicator captures agents’ confidence in the rules of a society such as the quality of contract enforcement, the police and the courts, as well as the likelihood of crime and violence.14 Data for 42 sample countries for the year 2005 were extracted from Kaufmann et al. (2007). Even though several years of data are available in the data set (1996, 1998, 2002 and 2003–05), the focus on a single year is justified based on the reasonable assumption that there is little variation in the time series for this nature of variable (especially, the small year-toyear variation). Hence, Contract variable essentially measures the crosscountry difference in governance quality. Infrastructure Variables The level of infrastructure is measured by the average time measured by days taken for moving a dry-cargo, 20-foot long full container of homogeneous goods from a factory in the largest business city to a ship in the most accessible port (Djankov et al., 2006). In terms of SITC clustering, these goods are SITC 65 (textile yarn), SITC 84 (articles of apparel and clothing accessories) and SITC 07 (coffee, tea, cocoa, spices and manufactures
Determinants of fragmentation trade
65
thereof). The data are drawn from the World Bank’s Doing Business Survey.15 The survey data are based on a detailed World Bank questionnaire sent to freight forwarding companies in 2005. Freight forwarders provide business services such as organizing the safe, efficient movement of goods for exporters and importers, and sometime dealing with packing and storage (Djankov et al., 2006). Taking into account the type of goods and the customers’ delivery requirements, freight forwarders arrange the best means of transport, using the services of shipping lines, airlines or road and rail freight operators. They are extremely knowledgeable about logistics and the surrounding infrastructure environment for international trade and handle approximately 85 per cent of world trade (Djankov et al., 2006). In order to ensure a homogeneous feature of transactions across countries, benchmark transactions of processing trade are defined by the World Bank’s Doing Business survey as follows. First, exporters are confined to a local business 100 per cent owned by an indigenous company with 200 or more employees and located in the country’s most populous city. From its sale, more than 10 per cent goes to foreign trade. Second, a cargo is dry and 20-foot wide, with three categories of goods: textile; yarn and fabrics (SITC 65); articles of apparel and clothing accessories (SITC 84); and coffee, tea, cocoa, spices and manufactures thereof (SITC 07). Real Exchange Rate (RER) The real exchange rate (RER) is computed by the following conventional formula: RER 5
ePw Pd
(4.3)
where e denotes the nominal exchange rate measured in terms of foreign currency, Pw is an index of foreign price and Pd is an index of domestic prices. The producer (wholesale) price index is used as a proxy for Pw and the GDP deflator is a proxy for Pd (Athukorala and Rajapatirana, 2003). The data are derived from online World Development Indicators. RER based on price index is not appealing in a pure cross-sectional context because the data only reflect changes relative to the base year of the index used, with no meaningful indication of overvaluation or undervaluation of a given currency. To eliminate spurious cross-sectional effects, RER is measured as a deviation from its period average of RER (Soloaga and Winters, 2001). It is assumed that countries are in exchange rate equilibrium at the means and RER effects are identified by changes through time relative to those means.
66
Table 4A.1
International fragmentation of production
List of 42 countries used in regression analysis
Country Costa Rica* Argentina* Slovenia* India* Slovakia* Russian Federation* South Africa* Turkey* Norway Australia Portugal Indonesia* Israel* Poland* Denmark Czech Rep. Brazil Hungary Finland Hong Kong SAR, China Austria Ireland Switzerland Thailand* The Philippines* Belgium Sweden Spain Netherlands Singapore* Malaysia* Italy Mexico* Rep. of Korea* Taiwan * United Kingdom Canada France China* Germany Japan USA
Share of machinery exports in world (%) 0.1 0.1 0.1 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.3 0.4 0.4 0.4 0.4 0.5 0.6 0.7 0.7 0.8 0.8 0.9 1.0 1.2 1.2 1.6 1.7 1.7 2.2 2.4 2.9 3.1 3.7 3.7 4.0 4.4 4.5 4.9 5.4 10.6 13.1 16.4
Note: Countries with asterisk are defined as developing countries in regression analysis.
Determinants of fragmentation trade
Table 4A.2
67
Summary of variable definition and data source for gravity model
Label
Definition
Data Source
FRG
Trade flows of parts and component of machinery industry Deflators for FRG
UN Comtrade (online database, available at http://comtrade.un.org/db/)
Export and import price index
GDP
GDPP (GDP per capita) Exchange rates Infra
Contract
Distance
ULC
GDP in constant US$ (base year = 2000) GDP per capita in constant US$ (base year = 2000) Period average exchange rates vis-à-vis US$ Time for processing trade (days) Contract enforcement and law regulations The geographical distance in kilometres Employee compensation and value added by US MNEs’ foreign affiliates
For Japan, Bank of Japan (BOJ), online database at http://www.boj.or.jp/theme/ research/stat/pi/cgpi/index.htm. For the USA, they are derived from the online database maintained by the Bureau of Labour Statistics, US Department of Labour, available at http://www.bls.gov/ home.htm World Bank Development Indicators online database, World Bank, available at http:// devdata.worldbank.org/dataonline/ As above
As above
As above
Kaufmann et al. (2007)
Joe Haveman’s International Trade Data Source, available at http://www.macalester. edu/research/economics/ PAGE/HAVEMAN/Trade.Resources/ TradeData.html. For countries except for the USA, the data are compiled from the electronic data files of the Annual Survey of US Direct Investment Abroad, Bureau of Economic Analysis, US Department of Commerce, available
68
International fragmentation of production
Table 4A.2 Label
(continued) Definition
Data Source at http://www.bea.gov/bea/ai/iidguide. htm#USDIA1. The US wage rates are obtained from the Bureau of Labour Statistics, US Department of Labour, available at http://www.bls.gov/ home.htm
NOTES 1. 2. 3. 4. 5.
6. 7. 8. 9. 10. 11. 12.
Fukao et al. (2003) focus on the determinants of Japan’s vertical intra-industry trade with East Asian countries. See Section 2.4 for an for explanation of the data. Trade facilitation is generally defined as an improvement of efficiency in logistics and related trade-enhancing infrastructure at ports and trade customs for the movement of goods in international trade (Wilson et al., 2003). In addition, adding on country-specific dummy requires dropping time-invariant variables which are the key to the analysis. The Hasuman-Wu specification test requires the instrument variables (IV) for the port infrastructure variable. Following Djankov et al. (2006), the instrument used here is the number of signatures required from customs officials for processing trade transactions. The survey respondents (freight forwarders) of the ‘World Bank Doing Business’ are asked to describe what documents are needed, where these documents are submitted and whose signature is necessary in order to process trade. The number of signatures required for customs clearance is a measure of excessive bureaucracy that slows down trade facilitation. It is qualified as a valid instrument variable by directly affecting only an improvement in trade facilitation, but is less likely to directly affect trade flows. These price indices are available at the Bank of Japan’s website, http://www.boj.or.jp/ type/stat/dlong/price/cgpi/index.htm. The concordance table is available at http://unstats.un.org/unsd/cr/registry/default. asp?Lg=1. Available at http://devdata.worldbank.org/dataonline/. Available at http://www.macalester.edu/research/economics/PAGE/HAVEMAN/ Trade.Resources/Data/Gravity/dist.txt. The electronic files are available http://www.bea.gov/bea/ai/iidguide.htm#USDIA1. That is, more than a 50 per cent share of control. An alternative data source for cross-country wage and employment might be derived from the United Nations Industrial Development Organization data set (UNIDO, 2006). This is a frequently used data set due to its easy accessibility. However, the UNIDO data set is a less preferred choice due to the lack of international consistency of the data and the limited country and time coverage. The UNIDO data rely on the data reported by individual country governments. As pointed out by Maskus (1991), there might be substantial differences in the data collection methods across countries (for example, sampling versus full enumeration, mail versus in-person interviews, frequency of revision and different industry classification). In addition, the data for many countries (especially developing countries) are not reported, and even if they are reported,
Determinants of fragmentation trade
13. 14. 15.
69
the data end at 2002 for most countries in the sample. For these reasons, the wage data from the UNIDO database are not considered in this study. Available at http://www.bls.gov/ces/home.htm. See more details at http://www.doingbusiness.org/. Available at http://www.doingbusiness.org/.
5
Structural transformation and labour market adjustment in Japanese manufacturing
5.1
INTRODUCTION
The subject of this chapter is the labour market implications of production fragmentation: how the relocation of the production process inevitably changes the nature of labour demand in the home economy. Before the formal empirical investigation this chapter documents the background to the changing nature of Japanese manufacturing and its labour market performance. In particular, the primary focus is to highlight the trends and patterns of the skill compositional change and the wage structure by skills in Japanese manufacturing over the last two decades (known as skill upgrading in the literature). The analysis in this chapter draws on the basic labour market indicators and makes a comparison with US manufacturing. The deteriorating position of unskilled workers is clearly one of the key policy concerns for developed countries in recent years. Wage inequality between skilled and unskilled workers has widened substantially since 1980 in many developed countries, particularly in the UK and the USA (Katz and Revenga, 1989; Katz and Murphy, 1992; Berman et al., 1994). At the same time, the manufacturing employment structure has also been changing in favour of skilled workers in most developed countries (Berman et al., 1994; OECD, 1997). In order to place the main analysis in perspective, this chapter begins with a discussion of the overall manufacturing experiences and structural change in the Japanese economy over the last 40 years. Section 5.3 then examines the ongoing skill compositional changes in manufacturing employment and the wage structure by skills and Section 5.4 summarizes the key findings.
70
Structural transformation and labour market adjustment in Japan
5.2
71
DEVELOPMENT OF JAPANESE MANUFACTURING AND LABOUR MARKET
The miraculous post-war economic growth in Japan was mainly driven by rapid growth of the manufacturing sector (Minami, 1986). The average annual growth rate of value-added in manufacturing was at 9 per cent in 1960–65 and 15 per cent between 1966 and 1970.1 The high labour productivity during this period was also accompanied by rapid growth in real wage rates (Figure 5.1). This combination offset any deterioration in the international competitiveness of Japanese manufacturing. As a result, the unit labour cost remained almost unchanged despite substantial real wage growth (Minami, 1986; Athukorala and Manning, 1999). However, as the Japanese economy achieved maturity in the 1980s, the relative importance of manufacturing in the domestic economy began to gradually decline (Figure 5.2). During this period the manufacturing share of GDP declined to 30 per cent and then to around 20 per cent by the early 2000s. Similarly, the manufacturing employment share declined from 26 per cent in the early 1970s to less than 20 per cent in the early 2000s. The growth rate of real manufacturing wages also began to stagnate in the late 1990s (Figure 5.1). This downsize in manufacturing is also observed for US manufacturing (Figure 5.2). The shrinking size of US manufacturing seems to have 120 110 100
per cent
90 80 70 60 50
Real labour productivity (2000 = 100) Real wages per employee (2000 = 100) ULC (2000 = 100)
40 30
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1972
1970
1968
1966
1964
1962
1960
20
Year
Figure 5.1
Selected indices of the performance of Japanese manufacturing (%), 1960–2004
72
International fragmentation of production 36 31
per cent
26 21 16 Mfg employment share (Japan) Mfg GDP share (Japan) Mfg employment share (USA) Mfg GDP share (USA)
11
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
1962
1960
6
Year
Figure 5.2
Output and employment share in Japanese and US manufacturing (%), 1960–2004
occurred at an earlier time period and its speed has been much faster than that of Japan. Both output and employment in US manufacturing began to fall in the late 1960s and has continued to fall very sharply. By 2004 the employment share of US manufacturing accounted for only 10 per cent of total employment, whereas the same share for Japanese manufacturing was 18.2 per cent (Figure 5.2). Generally speaking, three main reinforcing forces have been generating this trend of dwindling manufacturing in developed countries, commonly known as the ‘deindustrialization’ process (Katz and Revenga 1989). First, rising productivity and technological progress in the manufacturing sector have required fewer workers to work with machines. The brisk improvement in labour productivity in Japanese manufacturing in the 1970s and 1980s reinforces this point (Figure 5.1). Second, the expenditure patterns of consumers have changed as income per capita has risen. Consumers are prepared to spend a lot more on services. Third, imports of manufacturing goods have increased substantially, providing direct substitutes for domestically manufactured goods. Data on the sectoral composition of manufacturing output and employment in Japan over the period of 1960 to 2004 are presented in Table 5.1. There are some noticeable changes in manufacturing composition, moving from the traditional labour-intensive sectors to relatively more capital-intensive sectors. In the 1950s and 1960s the light manufacturing
Structural transformation and labour market adjustment in Japan
Table 5.1
73
Sectoral composition of output and employment in Japanese manufacturing (%), 1960–2004 Output
Employment
1960 1970 1980 1990 2004 1960 1970 1980 1990 2004 Food Beverage and tobacco Apparel and other finished products Textile mill products Leather tanning and leather Lumber and wood products Furniture and fixtures Pulp and paper Publishing and printing Ceramic, stone and clay products Chemical and allied products Petroleum and coal products Rubber products Iron and steel Fabricated metal products Non-ferrous metals and products General machinery Electrical machinery equipment Precision instruments and machinery
11.9 –
9.5 –
9.8 –
8.0 3.2
8.0 4.1
10.3 –
9.8 10.6 – –
9.8 13.7 1.2 1.3
1.1
1.1
0.9
1.4
0.6
2.6
3.5
4.8
5.2
3.7
13.9
5.7
2.9
2.4
0.7
18.1 10.8
6.7
4.8
1.9
0.5
0.3
0.2
0.4
0.1
0.6
0.7
0.8
0.7
0.4
3.4
2.3
1.4
1.4
0.6
6.0
4.6
3.5
2.3
1.7
0.9
1.1
0.9
1.3
0.6
2.1
2.6
2.5
2.1
1.7
3.9 2.5
3.3 2.6
3.2 2.8
2.7 3.9
2.6 2.3
3.4 4.2
2.9 4.0
2.7 4.6
2.5 5.0
2.7 4.4
3.4
3.4
3.2
3.3
2.0
5.2
4.9
4.9
4.1
3.9
9.5
8.9
9.7
7.3
9.2
6.2
4.2
4.0
3.6
4.3
2.4
2.9
8.5
2.6
3.9
0.5
0.3
0.4
0.3
0.3
1.5
1.1
1.2
1.1
1.1
1.7
1.5
1.5
1.5
1.5
10.8 10.5 3.9 4.5
9.2 3.5
5.6 5.8
4.2 3.9
5.5 5.1
4.7 7.2
4.2 7.2
3.0 7.6
2.5 8.0
4.2
2.4
2.2
1.9
1.8
1.8
1.5
1.6
8.3 10.5
9.1
8.5 10.1 10
10.7 11.3
8.6 11.7 11.6 16.9 22.5
5.9 11.5 13
17.4 17.5
1.1
1.9
4.4
4.8
6.7 10.3
1.3
1.7
1.6
1.3
2.1
2.6
2.2
1.9
74
Table 5.1
International fragmentation of production
(continued) Output
Employment
1960 1970 1980 1990 2004 1960 1970 1980 1990 2004 Transportation equipment Other manufacturing Total mfg
8.7 11.7 13.6 14.5 20.0
6.6
7.5
8.6
8.4 10.2
2.4
3.6
5.1
5.5
2.3
3.1
3.2
1.6
100 100 100 100
1.5 100
100
100 100
100
2.2 100
Source: Compiled from the electronics data files of the Census of Manufactures, METI, available at http://www.meti.go.jp/statistics/tyo/kougyo/index.html.
sectors of textiles and food accounted for the largest output and employment share. For instance, the output and employment shares of the textile industry in 1960 were 14 and 18 per cent, respectively (Table 5.1). Indeed, the performance of the textiles industry was one of the main vehicles for the post-war industrialization in Japan (Minami, 1986). The success of labour-intensive manufacturing during this period was a result of successful import substitution and the introduction of modern technology in the nineteenth century. However, the declining share of the textile and milling industry became increasingly evident in the 1970s and 1980s. This, in turn, gave rising importance to heavy- and capital-intensive manufacturing industries. More importantly, capital- and technology-intensive sectors, such as general machinery and the electrical and transportation equipment sectors, began to take a larger share of output and employment in Japanese manufacturing from around the mid-1970s. By 2004 the output and employment share of the electrical machinery industry had become the largest sector in Japanese manufacturing, with 22.5 per cent in the output share and 17.5 per cent in employment share (Table 5.1). This was followed by the transport equipment industry with a 20 per cent output share and 10.2 per cent employment share in 2004. In summary, this section has highlighted the changing nature of Japanese manufacturing in its historical context. The manufacturing industry was the main driver of the post-war growth in the 1960s and 1970s, but its importance in the Japanese economy has gradually declined over time. Over the last 40 years the composition of Japanese manufacturing has transformed from the dominance of light manufacturing towards high-tech and capital-intensive industry.
Structural transformation and labour market adjustment in Japan
5.3
75
THE TRENDS AND PATTERNS OF SKILL UPGRADING IN JAPANESE AND US MANUFACTURING
Against the background of Japanese manufacturing presented above, this section closely examines the trends and patterns of the labour market structure by skilled and unskilled workers in the last two decades (skill upgrading). Skill upgrading is generally measured either by an increase in wages of skilled workers relative to unskilled workers or by an increase in the relative employment of skilled workers. Many studies have documented the modality of skill upgrading in the context of US manufacturing (Katz and Revenga, 1989; Katz and Murphy, 1992; Lawrence and Slaughter, 1993; Berman et al., 1994; Sachs and Shatz, 1994). One explanation for skill upgrading postulates that changes in the wage structure of developed countries have been driven by the process of deindustrialization, reducing the demand for unskilled workers (Katz and Revenga, 1989). Another popular explanation is that technological change (mainly due to the computer revolution) has required relatively more educated and skilled workers in the workforce. Hence, the increased production efficiency brought about by technological progress has inevitably reduced the demand for physical labour. A third type of explanation is that the increased manufacturing trade and FDI with developing countries has effectively reduced the demand for less-skilled workers in developed countries by importing labour-intensive products or shifting the production process overseas (Sachs and Shatz, 1994). This last explanation is closely linked with the growing importance of production fragmentation, which is the focus of this volume. This section starts with a discussion of the definition and measurement of skill intensity of workers. This is followed by an analysis of the data on the trends and patterns of skill upgrading in Japanese as well as US manufacturing. The simple labour demand-supply framework will be used to make an inference of the data analysis. 5.3.1
Relative Manufacturing Employment and Wage Changes
Figure 5.3 provides key evidence of the trends of skill upgrading in Japanese manufacturing over the period 1960–2004. The figure plots the wage of non-production workers relative to production workers, measured on the left axis and relative employment on the right axis for the period 1960–2004. This figure shows a key aspect of the trend towards skill upgrading is summarized into wages and employment of skilled relative to unskilled workers (non-production/production workers). The
76
International fragmentation of production
2.00
0.90
1.80
0.80
1.60
0.70
1.40 0.60 1.20 0.50 1.00 0.40
0.80
Relative wage (left scale) Relative employment (right scale)
0.60
0.30
0.40 2004
2002
2000
1998
1994
1996
1992
1990
1988
1986
1982
1976
1974
1972
1970
1968
1966
1964
1962
1960
0.20
Year
Figure 5.3
Relative wage and employment of non-production workers to production workers in Japanese manufacturing, 1960–2004
level of workers’ skills is measured by the occupational data of workers (see Appendix 5.1 for a discussion of the measurement of workers’ skills). The data are a combination of the Census of Manufactures (CM) and the Basic Survey on Wage Structure (BSWS) (Appendix 5.2 describes the two surveys).2 Figure 5.3 clearly shows relative employment has moved in favour of skilled workers in Japanese manufacturing. In particular, there has been a sharp rise in the relative employment of skilled workers since the early 1960s in Japanese manufacturing. On the other hand, the relative wages of skilled workers actually fell in the post-war industrialization period, in the 1960s and early 1970s. This might be related to downward pressure on the relative wages of skilled workers due to a rapid increase in labour supply. However, since the mid-1970s the relative wage of skilled workers in Japan has remained almost constant. Similarly, Katz and Revenga (1989) and Sakurai (2001) found an almost constant relative wage rate for skilled workers in Japan since the mid-1970s.3,4 This suggests that there has been increasing relative employment of skilled workers in Japan since the mid-1970s, despite the lack of decrease in relative wages. This observed fact sits uncomfortably with the theory of supply and demand for labour. In theory, the uninterrupted massive increase of relative supply of skilled workers should have somehow lowered the relative wage for
Structural transformation and labour market adjustment in Japan
Relative Wages (RW)
Relative demand for skilled workers (skilled/unskilled)
RS1
77
RS2 Relative supply of skilled workers (skilled/unskilled)
RD2 RD1
RW1 = RW3
RW2
RE1
Figure 5.4
RE2
Relative Employment (RE)
Relative labour demand-supply explanations for Japanese manufacturing
skilled workers. However, instead, the strong rise of the relative demand for skilled workers might have worked against the relative supply effect, stopping the rise of the relative wage during this period. Sasaki and Sakura (2005) and Ahn et al. (2008) make a similar point on this account. The simple supply-demand framework is used for a more systematic interpretation of the data in Figure 5.3. The relative supply schedule of workers is assumed to be perfectly inelastic. This implicitly assumes that all workers are sufficiently willing to work regardless of ongoing relative wages. Based on this assumption, Figure 5.4 draws the relative demand curve for more-skilled workers (RD) relative to less-skilled workers, against the perfectly inelastic relative supply (RS) of more-skilled relative to less-skilled workers. As observed in Figure 5.3, an increase in the relative supply from RS1 to RS2 in the 1960s, shifted along the initial equilibrium relative demand curve (RD1) and reduced the relative wage of more-skilled workers from RW1 to RW2 in Figure 5.3 (that is, a shift along the relative demand curve). In the 1980s and 1990s the employment of non-production/production workers continued to rise in Japanese manufacturing, but in the context of stable relative wages between the two types of worker. The only consistent
78
International fragmentation of production
1.80
0.50
Relative wage (left scale) Relative employment (right scale)
1.75
0.45 0.40 0.35
1.70
0.30 1.65
0.25 0.20
1.60
0.15 0.10
1.55
0.05 1.50 2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
1962
1960
1958
0.00
Year
Figure 5.5
Relative wage and employment of non-production to production workers in US manufacturing, 1958–2004
explanation is that there must have been an outward shift of the relative demand curve from RD1 to RS2, as demonstrated in Figure 5.4. The new equilibrium is found to be RW3 which is the same level as the initial equilibrium RW1, but the relative employment is located at RE2. In short, the movement of relative wages and employment for skilled workers over the period 1980–1990 is characterized by the simultaneous shifts of relative demand and supply. In particular, a shift of RD is relatively less in magnitude than a shift in RS (this makes sure there is no change in equilibrium wage). More importantly, what distinguishes this from the earlier period is an outward shift of relative demand curves in the 1980s and 1990s. The experience of Japanese manufacturing is compared with that of US manufacturing (Figure 5.5). Interestingly, the movement of employment and real wage rate for production and non-production workers in US manufacturing is seen to be quite different from that of Japanese manufacturing. While numbers of non-production workers seemed to be growing in the 1980s, production workers have experienced some slowdowns since the early 1980s. Over this period a substantial wage inequality has arisen between skilled and less-skilled workers (Figure 5.5). The data on US manufacturing presented in Figure 5.5 can also be interpreted by using the simple demand-supply framework. The upward trends
Structural transformation and labour market adjustment in Japan
Relative Wages (RW)
Relative demand for skilled workers (skilled/unskilled)
RS1
RS2 Relative supply of skilled workers (skilled/unskilled)
RE1
RE2
RD2
79
RD1
RW2 RW1
Figure 5.6
Relative Employment (RE)
Relative labour demand-supply explanations for US manufacturing
of relative employment and in wages for non-production workers towards the same direction in the 1980s and 1990s suggest an outward shift in demand for skilled workers during those periods, leading to an increase in relative employment and wages. As in the case of Japanese manufacturing, there have been simultaneous shifts of relative demand and supply towards the same outward direction (Figure 5.6). The exact same shift of RD and RS is exercised as in the case of Japanese manufacturing. However, the only difference is that the relative wage rate of skilled workers is found to be greater at RW2 than the initial level of RW1 in the state of equilibrium. This must mean the magnitude of the relative demand shift is relatively bigger than that of the relative supply curve. In the end, the final equilibrium relative wage in RW2 and employment at RE2 are found to be at the higher level in the case of US manufacturing. In summary, skill upgrading is clearly evident in both Japanese and US manufacturing in terms of changes in relative wages and employment in favour of skilled workers over the last two decades. However, there are some differences in terms of the adjustment mechanism between the two countries. In particular, the relative wage adjustment for skilled workers
80
International fragmentation of production
does not seem to be taking place in the Japanese labour market, compared to US manufacturing. The simple supply-demand analysis suggests that a change in relative labour demand is the key to understanding this difference in relative wage behaviour, rather than the supply effect. This demand effect in the last two decades is closely related to production fragmentation, and this possible link is formally examined in Chapter 6.
5.4
CONCLUSION
This chapter has surveyed the overall structural transformation of Japanese manufacturing, with emphasis on the skill composition of employment and wages. The key inference of this chapter is that a change in employment composition and relative wages in favour of skilled workers has been occurring in both Japanese and US manufacturing over the last two decades. It was also evident that an interesting contrast exists in the labour market adjustment between Japan and the USA: it seems that skill upgrading in Japanese manufacturing has taken the form of change in the employment of skilled workers relative to unskilled workers, while their relative wages have not changed. On the other hand, the wage gap between skilled and unskilled workers has been widening in US manufacturing since the mid-1980s. The simple relative demand-supply framework suggested the magnitude of a shift in relative labour demand is a primary factor in the shift in the employment and wage composition rather than the supply factor for both Japanese and the US manufacturing. This key demand factor is formally examined in the next chapter. In particular, the hypothesis that industries engaged in production fragmentation experience greater changes in the demand for skilled workers in Japanese manufacturing data is examined. As discussed in Chapter 2, changing trade patterns by production fragmentation have direct implications for changing the nature of labour demand in the domestic manufacturing process.
APPENDIX 5.1
MEASURE OF WORKER SKILLS INTENSITY
A proper measurement of skill intensity must account for the level of educational attainment, on-the-job training and work experience (Hamermesh, 1993). However, there is no single measure of capturing these accounts. Additionally, the concept of workers’ skills is quite vague. This is especially so in Japan because Japanese workers are less
Structural transformation and labour market adjustment in Japan
81
occupation-conscious owing to their company orientation compared with their Western counterparts (Galenson and Odaka, 1976). Against this backdrop, either the educational attainment level of workers or the occupational data according to the tasks performed by workers is usually used as a proxy for the skill intensity of workers. This study adopts the latter approach to measure the skill intensity of workers (Berman et al., 1994; Feenstra and Hanson, 1999; Ito and Fukao, 2005). This approach essentially divides workers into production and non-production workers. ‘Non-production’ (white-collar or skilled) workers comprise technical workers (systems engineers and computer programmers), managers, administrative, advertising and sales workers. ‘Production’ (blue-collar or less-skilled) workers refer to manual workers, assemblers and operational workers. There are two main reasons for using this indicator. First, this measure of workers’ skills fits well conceptually with production fragmentation. The primary interest of this study rests on what types of jobs and tasks in manufacturing are relocated to overseas production and what types are retained at the domestic level. Second, in practice, the occupational distinction for workers closely matches the educational attainment level. On the whole, the data reasonably support the view that non-production workers in Japanese manufacturing have a higher education level than production workers. For example, the employment share of university graduates in total non-production workers in Japanese manufacturing was around 50 per cent in 2004, up from 39 per cent in 1985 (Table 5A.1). The employment share of high school graduates in total non-production workers has continued to decline from the mid-1980s, reaching 37 per cent in 2004. The share of junior high school graduates accounted for only 3.4 per cent in 2004.5 Of course, occupational segregation is not entirely a satisfactory proxy of workers’ skill levels, for a number of reasons. First, there is the misclassification of jobs between non-production and production workers (Leamer 1994). For instance, according to the International Standard Classification of Occupations (ISCO) provided by the International Labour Organization (ILO),6 line supervisors and product development personnel are included among production workers whereas delivery truck drivers and cafeteria workers are included with non-production workers.7 Another reason why occupational segregation is not a good proxy of workers’ skill levels is that the concern arises in the specific context of Japanese employment practice. Under the seniority wage system workers with a long period of service receive a high salary regardless of educational attainment levels and type of jobs (Ahn et al., 2008). As a consequence, it is possible to observe inexperienced workers
82
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
276 265 249 258 240 231 227 208 200 172 171 164
1482 1541 1532 1564 1682 1812 1928 2001 2017 2006 1970 2071
Real cash earning (1000 yen) 11.6 11.1 10.6 10.3 9.7 9.0 8.6 8.1 7.6 6.7 6.3 6.3
Share in non-prod. (%)
Junior High School Graduates
1057 1064 1017 1084 1076 1121 1154 1112 1115 1067 1119 1048
No. of workers (units = 1000) 1469 1528 1520 1563 1668 1779 1901 1960 1961 1996 1938 1972
Real cash earning (1000 yen) 44.5 44.5 43.5 43.4 43.4 43.7 43.9 43.1 42.4 41.9 41.2 40.1
Share in non-prod. workers (%)
High School Graduates
929 939 953 1025 1028 1063 1089 1094 1144 1138 1230 1212
No. of workers (units = 1000)
1773 1818 1811 1818 1953 2088 2230 2294 2334 2295 2234 2290
Real cash earning (1000 yen)
39.1 39.3 40.7 41 41.4 41.5 41.4 42.4 43.5 44.7 45.3 46.4
Share in non-prod. workers (%)
University Graduates
The education attainment of non-production workers in Japanese manufacturing (male workers), 1985–2004
No. of workers (units = 1000)
Table 5A.1
83
2095 2149 2016 1949 1984 1809 1816 1746
5.8 5.4 4.7 4.6 4.0 3.7 3.6 3.4
1022 984 923 897 884 801 780 809
2028 2037 1950 1865 1919 1839 1746 1733
Real cash earning includes bonus. The label ‘non-prod.’ refers to non-production.
149 135 113 107 94 81 78 73
39.9 39.5 38.5 38.4 37.9 36.4 36.5 37.2
1196 1181 1176 1142 1157 1130 1094 1099
2349 2420 2251 2171 2229 2253 2092 2056
46.7 47.4 49.1 48.9 49.7 51.4 51.1 50.6
Source: Compiled from the online database of the Basic Survey of Wage Structure (BSWS), Japan Institute for Labour Policy and Training, available at http://stat.jil.go.jp/.
Note:
1997 1998 1999 2000 2001 2002 2003 2004
84
International fragmentation of production
with skilled jobs receiving lower wages, and experienced workers with less-skilled jobs. These considerations make it difficult to retain the assumption of factor substitutability between skilled and less-skilled workers.
APPENDIX 5.2
THE CENSUS OF MANUFACTURES (CM) AND THE BASIC SURVEY ON WAGE STRUCTURE (BSWS)
The Census of Manufactures (CM) is administered by the Japanese Ministry of Economy, Trade and Industry (METI) of the government of Japan, and was first conducted in 1909, 1914 and 1919. Since 1920 it has become an annual survey. This manufacturing survey covers all establishments, in calendar years ending with 0, 3, 5 and 8. In other calendar years only establishments with four or more workers are covered by the survey. The survey items include the number and cash earnings of production and non-production workers other than the basic manufacturing indicators at the detailed four-digit industry classification level. However, the information for non-production/production workers is only available for every three years and, since 1990, the data for these two types of workers has become unavailable in the CM survey. The Basic Survey on Wage Structure (BSWS), managed by Japanese Ministry of Health, Labour and Welfare, began in 1948. It is designed to present a clear picture of the employment and wage structure in the Japanese labour market. Therefore, it contains the labour market indicators for various types of workers, such as occupation, sex, age, school career, length of service, and the size of firms for the regular workers, part-time workers and so on. This survey covers establishments with ten or more regular workers. Employment and compensation for production and non-production workers are mainly available for establishments with five or more workers in total manufacturing over the period 1985–2004. However, most of the data are not available at the disaggregated two-digit industry level.
NOTES 1. The figures are from the Census of Manufacturers, available at http://www.meti.go.jp/ statistics/tyo/kougyo/index.html. 2. The wage and employment data for production and non-production workers for Japanese manufacturing were compiled from the CM for the period 1960–1985. The series were then spliced by the data from the BSWS from 1985 onwards. Looking just at
Structural transformation and labour market adjustment in Japan
3. 4.
5.
6. 7.
85
the series, the data adjustment did not seem to distort the overall picture of the wage rates and employment movement of non-production and production workers. While Sakurai (2001) makes a distinction between non-production and production workers, Katz and Revenga (1989) use education attainment data to measure the skill intensity of workers. Head and Ries (2002) observed a continuous increase in the share of skilled workers in total wage bills since the 1970s (see figure 2, p. 92 in their paper). This is consistent with the finding in this volume that the massive increase in relative employment of skilled workers contributes to the rising share in the total wages bill, despite the relatively stable wage rate. It was also observed that since 1970 there has been a declining share in the wage bill for production workers in Japanese manufacturing. The figures reported in Head and Ries (2002) are more revealing. In 1980, 34 per cent of non-production workers in Japanese manufacturing had some college education, while only 14 per cent of them had less than high school diplomas. On the other hand, college graduate accounted only for 5.7 per cent of production workers and 60 per cent of them had not graduated from high school. ISCO is available at http://laborsta.ilo.org/. In relation to the misclassification of jobs into skilled/unskilled workers, Lawrence and Slaughter (1993) consider the following example: consider an experienced machine-tool technician with a university degree in computer science who programmes computers driving the tools, and a recent high school dropout who files reports and run mails. If both work for a manufacturing firm, the non-production/production distinction will clarify the technician as unskilled and the office runner as skilled.
6
The impact of production fragmentation on skill upgrading
6.1
INTRODUCTION
Chapter 5 presented some evidence of a shift over the last two decades in the skill composition of employment in favour of skilled workers in Japanese manufacturing. The present chapter empirically examines the role that production fragmentation has played in this labour market adjustment. The rise of production fragmentation has the effect of shifting labour demand away from less-skilled labour towards skilled labour within the manufacturing industry (or within the firm), since domestic production increasingly specializes in higher skilled and technology-intensive tasks. As a result, demand for skilled workers is pushed up, consequently raising the relative wage of skilled workers while suppressing demand and wages for less-skilled workers. Feenstra and Hanson (1996, 1999, 2003) have demonstrated that fragmentation trade contributed 15–24 per cent of the total increase in the wage share of skilled workers in US manufacturing during the 1980s. Following these studies, similar analyses have been undertaken for a range of other developed countries: Strauss-Kahn (2004) for France, Hijzen et al. (2005) for the UK, Helg and Tajoli (2005) for Germany and Italy, Hsieh and Woo (2005) for Hong Kong, Egger and Egger (2003) for Austria and Hansson (2000) for Sweden. Broadly speaking, the findings of these studies are consistent with the Feenstra-Hanson results for US manufacturing. However, the findings of the few available studies on Japanese manufacturing are inconclusive about the skill upgrading effects of production fragmentation (Sakurai, 2000; Ito and Fukao, 2005; Sasaki and Sakura, 2005). This is rather surprising, given the active involvement of production fragmentation in Japanese manufacturing over the last two decades (see Chapter 3). The exploration in the present chapter is thus motivated by this inconclusiveness in the findings of existing studies on Japan. It argues that the failure to find a robust relationship between the increased fragmentation intensity and industry skill upgrading in Japanese manufacturing might be associated with methodological shortcomings. The empirical analysis in this chapter is based on a panel data set covering 52 Japanese manufacturing industries over the period 1980–2000. 86
The impact of fragmentation on skill upgrading
87
The major novelty is the use of a new measure of fragmentation intensity based on trade data on parts and components through regression analysis. Based on this new measure it is found that the expansion of fragmentation trade with developing East Asian countries has had a significant impact on the skills composition of Japanese manufacturing employment. This chapter also finds that the impact of fragmentation trade on skill upgrading varies depending on the factor endowment characteristics of trading partners; trade with high-income countries (OECD countries) has had a skill downgrading effect in contrast to the skill upgrading effect of trade with developing countries.1 The export intensity of fragmentation trade is also considered. While existing studies only focus on the import side of production fragmentation, it is not confined to purchase of foreign intermediates inputs. Rather, it has mainly evolved due to the outward orientation of the fragmentation process by exporting parts and components manufactured in Japan to developing East Asian countries for the purpose of final assembly. Failure to capture the export orientation of production fragmentation might result in underestimating the actual impact of fragmentation on skill upgrading in Japan. More importantly, the existing studies might suffer from omitted variable bias through excluding the export side of fragmentation.2 Finally, the complied data set has a wider coverage in terms of the period and number of industries compared to previous studies. The updated time coverage is particularly important because fragmentation activities in Japanese manufacturing began to grow rapidly from the late 1980s. The organization of this chapter is as follows: the next section conceptually describes how an accelerated growth of fragmentation trade has implications for skill upgrading of domestic manufacturing, followed by a survey of the relevant empirical studies. Section 6.3 discusses model specification. Section 6.4 describes the data and econometrics methodology, followed by the interpretation of the results in Section 6.5. Section 6.6 concludes by summarizing the key findings.
6.2
THE SKILL UPGRADING EFFECT OF PRODUCTION FRAGMENTATION
Production fragmentation either takes the form of importing parts and components or exporting the domestically produced components for further processing and final assembly. The former case involves the lowerskill contents of the intermediate processing stage being performed in low-wage countries and then imported by a developed country for the further higher valued-added processing. In the latter case the relatively
88
International fragmentation of production
high skills-intensive components are exported by developed countries for the purpose of further labour-intensive processing and final assembly in developing countries. In both cases an upgrading of the skills content of the remaining production process is implied due to greater specialization in developed countries (called the skill upgrading effect). This is the key hypothesis to be examined in this chapter. While the labour market implication on the skill upgrading effect of production fragmentation is straightforward, the theory provides less clear-cut guidance to the impact of production fragmentation on different types of workers, as demonstrated in Chapter 2. It is concluded in Chapter 2 that the labour market effect of production fragmentation crucially depends on the complex interaction of factor endowment, factor intensity and the output pattern of a country. There is also an extensive empirical literature examining the skill upgrading effects of production fragmentation for developed countries. Feenstra and Hanson (1996, 1999) develop the seminal work. Their measurement of outsourcing intensity basically involves calculation of the imported intermediate inputs from the Annual Survey of Manufactures, US Bureau of the Census (see Chapter 2 for further details on the outsourcing intensity measures developed by Feenstra and Hanson). Feensta and Hanson’s (1999) data sets cover 447 industries based on the US Standard Industrial Classification (SIC) over 1979–90. In these regressions the dependent variable is the change of non-production (skilled) workers’ shares in total wage bills over the period. The estimation framework is based on a translog cost function, first employed in this literature by Berman et al. (1994). The results in Feenstra and Hanson (1996, 1999) support the hypothesis that increased outsourcing has had a positive impact on the non-production share of total wage bills, alongside technological change indicators. The Feenstra and Hanson (1996, 1999) calculations suggest foreign outsourcing contributed 15–24 per cent of the total change to the non-production wage shares associated with a shift in total demand for labour towards more skilled workers over the period 1979–90. Following Feenstra and Hanson (1996, 1999), a similar analysis has been undertaken for some other industrial countries. These include Anderton and Brenton (1999) and Hijzen et al. (2005) for the UK, Strauss-Kahn (2004) for France, Hansson (2000) for Sweden, Helg and Tajoli (2005) for Germany and Italy and Hsieh and Woo (2005) for Hong Kong. Overall, the results suggest that increased fragmentation trade has a sizable impact on shifting labour demand towards more skilled workers, although the estimated magnitude of the impact varies across countries. Sakurai (2000), Ito and Fukao (2005) and Sasaki and Sakura (2005) examined the skill upgrading effects of the fragmentation intensity in
The impact of fragmentation on skill upgrading
89
Japanese manufacturing using a similar methodology.3 However, unlike other country studies, the studies on Japanese manufacturing have not been able to present clear-cut results. Sakurai (2000) used wage data for production and non-production workers for 39 manufacturing industries, drawn from the Census of Manufactures, Japanese Ministry of Trade, Economy and Industry, over the period 1987–90. He constructed measures of outsourcing intensity following Feenstra and Hanson (1996) and tested for its statistical significance of change in non-production workers’ share of total wage bills in Japanese manufacturing. He found no statistical relationship between the intensity of imported intermediate inputs and skills upgrading. The relatively short time period was considered to be the reason for this insignificant result. Ito and Fukao (2005) extended the analysis to cover 35 manufacturing industries over a longer time period (1988–2000). Unlike Sakurai (2000), they examined the employment effect of production fragmentation rather than wages due to the data constraint. In their various regression runs the indices of intensity of production fragmentation based on the I-O table exhibited the expected positive sign, indicating the skill upgrading effect, but they found no statistical significance. Sasaki and Sakura (2005) examined the possible impact on industry skill upgrading, based on education attainment levels (higher or lower educated) for a panel of 14 Japanese manufacturing industries during the period 1988–2003. This study was motivated by a concern that the inconclusive evidence of the previous studies was due to the failure to specifically allow for Japan’s growing imports from countries in East Asia. They used the manufactured imports penetration ratio from East Asian countries as an indicator of the fragmentation intensity. They found that increased imports penetration from developing East Asian countries contributed to around a 10–13 per cent increase in the highly educated workers’ wage bill share across industries over the period 1988–2003. However, their analysis only focuses on the impact of increased manufacturing imports from East Asian countries. Additionally, their measure of fragmentation intensity by imports penetration is a crude proxy. There are two main shortcomings in the existing studies of the effects of increasing the fragmentation intensity on skill upgrading in Japanese manufacturing. First, as demonstrated in Chapter 2, the imported intermediate inputs from the I-O table as a proxy for the intensity of production fragmentation might not be appropriate in the context of Japan. Second, apart from Sasaki and Sakura (2005), the previous studies failed to take into account a geographic-specific effect of production fragmentation. A survey of empirical studies on the skill upgrading effects of fragmentation trade is presented in Table 6.1. The following section addresses these issues.
90
USA
Feenstra and Hanson (1996, 1999) Strauss-Kahn (2004)
Germany and Italy
Hong Kong
UK
UK
Helg and Tajoli (2005)
Hsieh and Woo (2005)
Hijzen et al. (2005)
Anderton and Breton (1999)
France
Country
1982–96, 50 manufacturing industries, 1971–86, 11 ISIC (textile and non-electrical machinery) industries
1971–96, 54 manufacturing industries
20 mfg sector, twodigit ISIC (Rev. 3)
1979–90, 447 SIC manufacturing industries 1977–93, 50 threedigit manufacturing industries (INSEE)
Data Description
Import penetration ratio in manufacturing
Imported intermediate input from China constructed from industry census Imported intermediate input from I-O table
OAP (Offshore Assembly Programme) data
Imported intermediate inputs from the Annual Survey of Manufactures Imported intermediate input from the I-O table
Indicator of Fragmentation Trade Intensity
Reduces the demand for less-skilled workers, but not for the semi-skilled and skilled workers Low-wage imports accounts for 40% of decline in unskilled worker wage share and 33% of decline in employment share in textiles
Fragmentation explains 11–15% change in decline in the less-skilled workers’ employment share for the period 1977– 85, and over 25% for the period 1985–93 Fragmentation increases relative employment of skilled workers in Italy, but not Germany Outsourcing to China accounts for about 40% increase of the wage share of skilled workers
Fragmentation accounts for 15–24% change for skilled workers’ wage share
Main Results
Survey of empirical studies on the skill upgrading effects of fragmentation trade
Study
Table 6.1
91
Japan
Japan
Sakurai (2000)
Sasaki and Sakura (2005)
Sweden
Hansson (2000)
Japan
Austria
Egger and Egger (2003)
Ito and Fukao (2005)
Denmark
Skaksen and Søresen (2002)
1990–98, 20 manufacturing industries (NACE two-digit) 1986–95, 34 (19) manufacturing industries 39 manufacturing industries, 1987– 1990 (Census of Manufactures) 35 manufacturing industries 1988– 2000 (2002) (JIP 2003 database) 14 manufacturing industries, 1988–2003 (mainly from Census of Manufactures, and Basic Structure of Wage)
1981–98, 50 manufacturing industries (ISIC Rev. 3)
While the level of outsourcing measure accounts for around 45% of change in the skilled workers’ wage share, there is no impact when it is measured in changes Positive association between outsourcing measure and change in the skilled workers’ employment share, but not statistically significant Import penetration from East Asian countries accounts for 10–13% of increase in the skilled workers’ wage share
Imported intermediate input I-O table
Import penetration in manufacturing from East Asian countries
Imported intermediate input I-O table
Fragmentation accounts for 5.4% change in relative demand for skilled workers
Outsourcing measure decreases relative demand for unskilled workers and increases the relative demand for skilled workers. No impact of relative demand on the semi-skilled workers Fragmentation accounts for 25% increase in relative skilled employment ratio
Imported intermediate input I-O table
Imported intermediate input from I-O table
Imported intermediate inputs from I-O table
92
6.3
International fragmentation of production
ECONOMETRICS ANALYSIS
As documented in the previous section, the existing empirical studies on Japan have found less clear-cut evidence of the importance of the fragmentation intensity of trade on skill upgrading at the industry level. This section re-examines this by conducting an econometric study of the panel data of 52 manufacturing industries over the period 1980–2000. The innovation of this analysis is the incorporation of a better measure of the fragmentation intensity of trade for a given industry, namely trade in parts and components. Using this index allows examination of the impacts of both the imports and exports side of fragmentation intensity on skills upgrading, in contrast to the previous studies only considering the imported intermediate inputs. It also permits investigation of any differential impact on skill upgrading by disaggregating the geographical orientation of fragmentation intensity. Skilled and less-skilled workers are defined in Chapter 5: skilled workers correspond to ‘non-production’ workers, consisting of technical and professional workers, whereas ‘production workers’ are a proxy for less-skilled workers with manual, assembling and operational jobs. The estimation is based on a reduced form of labour demand function, which is widely used in this literature (Berman et al., 1994; Feenstra and Hanson, 1996, 1999; Strauss-Kahn, 2004; Ito and Fukao, 2005). First, the cost minimization framework is described in a concise manner. Industry minimizes a quasi-fixed (short-run) cost function, C (w, y) in which output (y) and w are a vector of factors of production such as capital (k) as a fixed factor (as exogenous) and more-skilled and less-skilled labour as variable factors. The cost function takes a translog form, which is the second order Taylor series approximation linearly homogeneous function with a concave in factor prices, à la Chiristensen et al. (1973). The translog shortrun cost function (c) with a subscript z denoting industry is then written as follows (a time subscription is dropped for convenience); M
K
ln cz 5 a 90 1 a 9i ln wz,i 1 a :k ln xz,k 0a 0b i51 i1
k 51 1
K K 1 M M 1 g d a; a ?i,j ln wz,i ln wz,j 1 a
(6.1)
K
1 Fi,k ln wz,i ln xz,k a 0a 0f i51 k 51
where wi refers to the optimally chosen variable factor prices with subscripts denoting i, j 5 1, cM and xk denotes either the quantities of fixed inputs (capital), outputs or other structural parameters with subscripts k, l 5 1, cK.
The impact of fragmentation on skill upgrading
93
Equation (6.1) requires the following linear parameter restrictions to satisfy the property of linearly homogeneous with respect to variable factor costs (wi); M
M
M
i1 i51
i1 i51
i1 i51
g?j,i, di,k 5 0 g?i,j 5 0a 0 g?i,j 5 0 Differentiating equation (6.1) with respect to ln wi yields the cost share of variable factor i: (0 ln Cz) / (0 ln wi) 5 (0C/0w) (wi/Cz) where (0C/0w) refers to factor demand for input i by Shephard’s lemma. It follows that (0 ln Cz) / (0 ln wi) 5 Eiwi/Cz) 5 Sz,i is equal to the share of factor i in total costs, denoted by Sz,i (where E is a factor i employment). In the end, it yields a cost share equation of variable factor of input i; M
K
M
j1 j51
k 1 51
i1 i51
Sz,i 59 ai 1 0 1 0 i,k ln xz,k and 0 51 a ln wz,j a> a Sz,i
(6.2)
Equation (6.2) relates to the cost share of variable factor i to factor prices and the output level and fixed input capital. A cost share for variable factor j can be similarly derived. By assuming the coefficients of independent variables are equal across all industries, equation (6.2) can be pooled cross-industry and by time. The most important variable included in the right-hand side of equation (6.2) is the measure of the production fragmentation intensity (FRG) across industries (see Chapter 2 for a description of the method of data compilation for trade in parts and components). In terms of the model formulation, the imports values of parts and components are taken as a ratio to total industry expenditure on the intermediate inputs uses. The export values of parts and components are taken as a ratio to the industry gross output. FRGimport 5
Imports of Parts and Components Intermediate Inputs Uses
FRGexport 5
Exports of Parts and Components Gross Outputs
(6.3)
As discussed in Chapter 2, there are three added advantages of this approach compared to the alternative measure of fragmentation intensity. First, it avoids mixing traditional intermediate inputs into the estimates by making a distinction between trade in parts and components and ordinary intermediate inputs. Second, trade data capture both export and import orientation of fragmentation trade. Third, controlling for the direction of trade in parts and components makes it possible to differentiate the possible heterogeneity effects of fragmentation trade on skill upgrading.
94
International fragmentation of production
For example, the possible impact on skill upgrading might be different depending on whether an increase in parts and components imports is from developing countries or developed countries due to the difference in skills content. The former might be expected to have skill upgrading effects in domestic manufacturing whereas the latter might be expected to have skill downgrading effects. This distinction is particularly crucial because recent years have witnessed a rapid increase in components imports from developing East Asian countries (see Chapter 3). This division of labour with East Asian countries through the fragmentation process might be expected to result in a significant impact on skill upgrading. Therefore, the direction of the sign is expected to differ according to the geographical locations of trade in parts and components. An increasing component trade with developing East Asian countries is hypothesized to be positively related to change in employment of skilled workers (see Table 6A.1 for country composition). On the other hand, a skill downgrading effect with a negative sign is expected with higher intensity of fragmentation activity with OECD countries. As highlighted in Section 2.4.1, the main limitation of this measurement approach is the limited industry coverage, since a detailed separation of parts and components trade is possible for machinery and transport equipment (Standard International Trade Classification, SITC 7) and miscellaneous manufacturing (SITC 8). For a robustness check, the findings from the fragmentation intensity by trade data in parts and components will be compared with a regression with the conventional indicator based on the imported intermediate inputs from the I-O table. The dependent variable is the employment share of skilled workers in total manufacturing employment as examined in Ito and Fukao (2005). In analysing Japanese manufacturing, the employment share for a proxy of skill upgrading seems to be a more interesting variable. Chapter 5 described strong indications of a compositional shift of employment in favour of skilled workers, while the relative wage of skilled workers has been almost constant since about the mid-1980s (see Figure 5.5). An alternative measure is the share of skilled workers in the total wage bills. However, this cannot be computed for the time period covered in this chapter because of the unavailability of wage data for non-production workers at the disaggregated industry level.4 Two potential candidates to represent the industry scale of production (Y) are value-added and gross output.5 Value-added is used to represent the industry output measure, rather than gross output, because gross output might be too inclusive to serve as a clear indicator of industry output scale (Maskus, 1991).6 The sign of this variable depends, ceteris paribus, on whether the expansion of the industry output scale would
The impact of fragmentation on skill upgrading
95
require more skilled workers. If the estimate coefficients are zero, the hypothesis that the underlying production function is homothetic cannot be rejected. Otherwise, it implies non-homothetic, suggesting the ratio of the optimal inputs demands depend on the level of outputs. The ratio of capital stock to value-added (K:Y) is used to measure capital intensity of production. The sign of this variable can be either positive or negative depending on whether skilled workers are complementary (the positive sign) or substitutes (the negative sign) to physical capital stock in the production process. R&D intensity (the R&D expenditure ratio to value-added) is included in the model to capture any effect of skills-biased technological change introduced into working practices in association with change in production technologies, new capital investment and the use of computers. The expected sign of the coefficient on this variable is positive. Alternatively, the stock of patent data can be used, but is not considered here due to data constraints.7 Finally, both the industry fixed effect and time-specific effect are incorporated in order to guard against omitted variables for explaining the variation in the employment share of skilled workers in the respective dimensions: the former is needed to control for any unmeasurable (or unobserved) time-invariant heterogeneity, such as persistent technological differences, an industry-specific effect or difference in the average management quality. Time-specific effects are also introduced to control for a homogeneous form of technological change across industries, but varying across time and capturing other macroeconomic shocks. Based on the above discussion, the stochastic form of equation (6.2) can be written as: Shz,t 5 0 1 1Yz,t 1 2Kz,t 1 3R&Dz,t 1 4FRGmz,t 1 5FRGz,tx 1 aZ 1 gt 1 ez,t
(6.4)
where Sh is non-production (skilled) workers share in total employment, and subscripts z and t denote industry and time. Superscripts m and x represent imports and exports. The explanatory variables are listed below with the expected sign of the regression coefficient of each variable given in parentheses. Y K R&D FRG
Output (1 or 2) Capital intensity (1 or 2) Research and development intensity (1) Fragmentation intensity of trade (exports and imports) (+)
96
International fragmentation of production
aZ gt e
Industry-specific fixed effect Time-specific fixed effect (base year = 1980) Random error term, representing other omitted influences.
6.4
DATA AND ECONOMETRIC METHODOLOGY
In this section we describe a regression analysis using a panel data set for 52 Japanese manufacturing industries at five-year intervals over the period 1980 to 2000 (namely, 1980, 1985, 1990, 1995 and 2000). The data are mainly sourced from the Japan Industrial Productivity database 2006 (JIP 2006) (see Appendix 6.2 for the details on the JIP 2006 database).8 The data series on intensity of fragmentation trade (FRG) was compiled from the UN Comtrade database. Trade data based on a commodity list of five-digit product level of parts and components in machinery and transport equipment (SITC 7) and miscellaneous manufacturing (SITC 8) categories were extracted.9 These identified five-digit level of products are then mapped into the corresponding JIP 2006 industries. This procedure mainly covers trade in parts and components in the industry such as tobacco (JIP 14), furniture (JIP 17), general machinery (JIP 42–44), office machinery (JIP 45), electronic machinery (JIP 46–53), motor vehicles (JIP 54–55), precision machinery (JIP 57) and plastic products (JIP 58). Gross outputs, value-added, intermediate inputs, employments, capital stocks and R&D expenditures are extracted from the JIP 2006 database. The most desirable feature of this data set is that it gives the employment proportion of non-production and production workers in each industry. The original employment data on the skill composition of employment in the JIP 2006 database are derived from the survey data of the Population Census and the Basic Survey on Employment Structure. These two surveys are conducted every 5 years by the Statistics Bureau, Japanese Ministry of Internal Affairs and Communication, and contain information on employment across detailed industries (three-digit) and employment in more than 300 different occupations. The detailed occupational employment information was then reclassified into 52 manufacturing industries of the JIP 2006. As discussed in Section 5.3.1, skilled workers are measured by nonproduction workers with the occupation of professional and technical, managers and administrators, clerical and secretarial, sales and services. Less-skilled workers are measured as production workers, defined as plant and machine operators and also those who engage in craft and related occupations. Skilled workers are also defined narrowly by focusing only on professional and technical occupations, denoted as Tech. Gross outputs (in Japanese yen) are measured as the sum of industry
The impact of fragmentation on skill upgrading
97
shipment, revenues from repairing and fixing services, and revenues from performing subcontracting works. Intermediate inputs (in Japanese yen) are defined as the sum of raw materials, fuels, electricity and subcontracting expenditure. Both gross outputs and intermediate inputs are available in the year 1995 as constant prices in the JIP 2006 database. Real valueadded is then defined as the difference between real gross outputs and real intermediate inputs. Real capital stock (Japanese yen at 1995 prices) refers to the real book value of tangible fixed assets including buildings, machinery, tools and transport equipment. Nominal R&D expenditures (in Japanese yen) have been updated to JIP 2006 industries from the previous version of the JIP database (JIP 2003).10 For the estimation procedure, both fixed effects and random effect estimators are used in order to exploit the panel feature of the data set. The choice between the fixed- and random-effect models depends on whether the time-invariant industry heterogeneity are fixed or random (Wooldridge, 2000). If they are random, then an ordinary least squares (OLS) estimator will understate the standard error. Therefore, this calls for the use of a generalized least squares (GLS) estimator (that is, the randomeffects model). If they are fixed and correlated with any of the explanatory variables, this leads to the problem of the omitted variable bias. In this case, the fixed-effect models need to be employed to remove such biases. The standard Hausman test was implemented to determine which estimator, random or fixed effect, is appropriate. The test results were mixed, but the estimation results based on fixed-effect and random-effect estimators were comparable. In the case of the fixed-effect model, there are three alternative estimation techniques available to purge the industry-specific effects: the timedemeaning (that is, within-transformation), least square dummy variables (LSDV) and the first-differencing. The within-transformation of fixed effect is preferred, because the first-differencing frequently used in the literature might not be suitable for the current context due to the nature of the data set.11 When the number of time-periods exceeds two as in this case, two other estimators become more efficient under the assumption of no serial correlation in the error term (Wooldridge, 2000). Otherwise, the first-differencing method is preferred. However, the data are less likely to be prone to the problem of serial correlation for a panel of five-year intervals of records. Moreover, the first-differencing data approach can exacerbate any potential problems arising from measurement errors in the data (Griliches and Hausman, 1986; Hijzen et al., 2005). The choice of using the within-transformation or the LSDV is more subtle, since both estimators should give identical estimated coefficients and test-statistics under normal circumstances. The former is preferred
98
International fragmentation of production
because the alternative method is not appropriate due to the problem of degree of freedom by the inclusion of the slope dummy for all 52 industries. Following standard practice in the literature, the model is then estimated by the weighted least squares (WLS) method, in which the weights are the manufacturing employment share. In this procedure more ‘weight’ is placed on relatively large industries. In order to demonstrate the superiority of the proposed measures in equation (6.3), regression is performed with the alternative measure of the intensity of production fragmentation based on the Feenstra-Hanson approach (Table 6.3). As discussed in Chapter 2, the Feenstra-Hanson approach computes the dependency of the imported intermediate inputs across industries based on the I-O table.12 Table 6A.2 provides the same comparison using the data for US manufacturing available from Feenstra and Hanson (1999). In order to guard against possible heteroscedasticity, White’s robust standard errors clustered by industry have been used in calculating t-ratios. All variables, other than time-dummy variables, were used in natural logarithms, and hence the estimated coefficients can be interpreted as elasticities.
6.5
RESULTS
Reg. 1 of Table 6.2 presents the benchmark results with the newly constructed measure of intensity of fragmentation trade as defined in equation (6.3). Reg. 1 of Table 6.3, on the other hand, presents the results obtained by using both narrow measures (denoted as narrow outsourcing) and the difference between broad and narrow measures (denoted as outsourcing difference). Both indicators fail to detect the skill upgrading effects of the fragmentation intensity with any statistical significance, although these variables mostly show the expected signs. This does not change even with the narrower definition of skilled workers (Tech). These findings suggest there is no statistical relationship between fragmentation intensity and skill upgrading. This is an unexpected result, although it is consistent with the results of the existing studies on Japan (Sakurai, 2000; Ito and Fukao, 2005). This comparison also implies that the results obtained in Table 6.3 are not dictated by the limited industry coverage of the alternative measures used as opposed to the conventional measure. More importantly, total component trade might be masking some heterogeneity skill upgrading effects of production fragmentation. The baseline specification, Reg. 1 of Table 6.2, is then re-estimated by disaggregating the component trade into source and destination country groups: developing East Asia countries, and OECD countries. The
The impact of fragmentation on skill upgrading
Table 6.2
99
Evidence of skill upgrading effects in Japanese manufacturing, 1980–2000, weighted fixed effect (within-transformation) estimates
Explanatory Variables:
Dependent variable= the share of skilled workers in total employment Reg. 1 Reg. 2 Reg. 3 Reg. 4 Skill=non- Skill=non- Skill=tech Skill=tech prod. prod.
0 Y K R&D
FRGimport:
East Asia OECD FRGexports:
East Asia OECD Diagnostic Test Stats
Constant term
0.01 0.02 −0.05 −0.04 (0.02) (0.02) (0.08) (0.09) Output scale −0.10 −0.11 −0.25 −0.25 (0.01)*** (0.01)*** (0.04)*** (0.04)*** Capital intensity 0.02 0.04 0.13 0.17 (0.08) (0.09) (0.28) (0.29) R&D 1.36 1.38 2.90 2.94 expenditure (0.32)*** (0.31)*** (1.18)** (1.17)** intensity Intensity of 1.13 0.99 fragmentation (1.15) (2.50) trade (total imports); from East Asia 5.71 10.90 (2.24)** (5.53)* from OECD −2.85 −6.53 (1.43)* (3.98) Intensity of −0.79 −1.28 fragmentation (0.50) (1.05) trade (total exports); to East Asia 1.50 0.66 (1.08) (1.83) to OECD −0.39 0.82 (0.91) (1.23) 0.73 0.75 0.67 0.67 R2
F-Statistic Number of observations
29.6*** 260
43.4*** 260
23.9*** 260
18.3*** 260
Note: All variables are in natural logarithms. For Reg. 1 and 2, the dependent variable is the share of non-production workers, while the employment share for Reg. 3 and 4 is confined to professional and technical workers (denoted as tech). Time-dummy variables are included for all estimations, but the results are suppressed here. Weighted least squares (WLS), weights equal to the industries’ employment share in total manufacturing. Standard errors based on White’s heteroscedasticity correction are given in parentheses, with statistical significance (two-tailed test) denoted as: *** 1%, ** 5%, and * 10%. East Asian countries and OECD countries are defined in Table 6A.1.
100
International fragmentation of production
Table 6.3
Evidence of skill upgrading effects in Japanese manufacturing, 1980–2000, alternative measure of the fragmentation intensity
Dependent variable= the share of skilled workers in total employment Explanatory Variables:
Reg. 1
Reg. 2
Reg. 3
Reg. 4
Skill=nonprod. Skill=tech Skill=nonprod.Skill=tech Within-transformation 0
Constant term Y Output scale K Capital intensity R&D R&D expenditure intensity Outsourcing Narrow (narrow) outsourcing intensity Outsourcing Difference (difference) between broad and narrow outsourcing Diagnostic R2 Test Stats F-Statistic Root MSE Number of observations
0.02 (0.02) −0.11 (0.01)*** 0.03 (0.10) 1.32 (0.41)***
−0.05 (0.08) −0.26 (0.04)*** 0.11 (0.29) 2.53 (1.17)**
First-difference −0.00 0.01 (0.00) (0.00)* −0.10 −0.26 (0.01)*** (0.03)*** −0.00 −0.10 (0.04) (0.15) 0.21 0.46 (0.19) (0.51)
0.40 (2.64)
5.08 (7.71)
0.31 (1.35)
0.51 (4.19)
1.51 (1.01)
4.13 (4.09)
1.01 (0.53)*
2.97 (2.35)
0.74
0.68
0.67
0.68
26.43*** 0.01 260
23.72*** 0.01 260
39.10*** 0.01 208
31.31*** 0.04 208
Note: All variables are in natural logarithms. The within-transformation estimator is implemented for Reg. 1 and 2. The first-difference estimator is used for Reg. 3 and 4. For Regressions 1 and 3, the dependent variable is the share of non-production workers, while the employment share for Regressions 2 and 4 is only confined to professional and technical workers (denoted as tech). Time-dummy variables are included for all estimations, but the results are suppressed here. Weighted least squares (WLS), weights equal to the industries’ employment share in total manufacturing. Standard errors based on White’s heteroscedasticity correction are given in parentheses, with statistical significance (twotailed test) denoted as: *** 1%, ** 5% and * 10%. East Asian countries and OECD countries are defined in Table 6A.1.
coefficient of component imports intensity of fragmentation trade from developing East Asian countries is now positive and statistically significant at the 5 per cent level, suggesting significant skill upgrading effects on overall change in the employment share of non-production workers
The impact of fragmentation on skill upgrading
101
(Reg. 2 of Table 6.2). In particular, the estimated coefficient suggests that on average a 1 per cent increase in the component imports ratios from developing East Asian countries would lead to about a 6 per cent increase in skilled workers’ employment share. This accounts for about 12.3 per cent of the increase in the share of skilled workers during this period.13 In other words, increasing component imports from developing East Asian countries would involve a substantial increase of the employment share of skilled workers in Japanese manufacturing. Interestingly, the economic significance is similar to the impact of manufacturing import penetration from East Asian countries, computed in Sasaki and Sakura (2005). In the narrow definition of skilled workers, the estimated coefficient of the same explanatory variable is larger, while it becomes less statistically significant (Reg. 4, Table 6.2). It appears that export intensity of fragmentation trade has no link with skill upgrading in Japanese manufacturing. In all regressions in Table 6.3 the exports intensity variable does not have any statistical significance. However, this result could be driven by a high correlation between imports and exports intensity of fragmentation trade (r = 0.86). As expected, an increase in component imports intensity from OECD countries seems to have skill downgrading effects (Reg. 2, Table 6.2), but is only significant at the 10 per cent level. This suggests increased component imports from OECD countries require more unskilled workers for further processing. Quantitatively, the import intensity of fragmentation trade with OECD countries explains a marginal 1.02 per cent decline in skilled workers’ share. This is indeed consistent with the argument put forward in the previous section that component imports from high-income countries, presumably highly capital- and technology-intensive contents, might substitute for the domestic skilled worker. Overall, the findings presented in Table 6.3 add a new dimension to the literature on industry upgrading through international production fragmentation in Japanese manufacturing. All regressions in Table 6.2 show a negative industry output scale effect (Y) on the demand for skilled workers. The negative scale effect suggests Japanese manufacturing industries would require, ceteris paribus, fewer skilled workers as output expands. The estimated coefficient on capital intensity (K) suggests capital utilization has a positive relationship for skilled workers (that is, the complementary relationship between skilled workers and capital investments), but is found to be statistically insignificant. In fact, the capital:output ratio accounts, on average, for very little of the variation in the employment change of skilled workers. This finding is markedly different from the commonly found robust complementary relationship between capital utilization and skilled workers in US manufacturing (for example, Berman et al., 1994). However, this is quite
102
International fragmentation of production
consistent with a previous study of Japanese manufacturing (for example, Sakurai, 2000). The result for the R&D intensity variable suggests a positive and statistically significant effect on skill upgrading on average. This finding is consistent with the general findings supporting that skillsbiased technological change is strongly associated with skill upgrading in Japanese manufacturing (Sakurai, 2000; Ito and Fukao, 2005). It should be noted that the size of the estimated coefficient for the R&D variable is somewhat smaller than the FRG variable in some of the specifications in Table 6.3. To sum up, the results suggest a significant effect of increasing fragmentation trade on skill upgrading across industries in Japanese manufacturing over the period 1980–2000. In particular, the main skill upgrading effects come from the increased import intensity of fragmentation trade with developing East Asian countries. On the other hand, the evidence also points to skill downgrading effects from increasing component imports from OECD countries. These findings are in contrast to those of Sakurai (2000) and Ito and Fukao (2005) who failed to find any evidence that increasing practices of production fragmentation contribute to skill upgrading in Japanese manufacturing.
6.6
CONCLUSION
This chapter examined the hypothesis that industries engaged in international fragmentation of production experience greater skill upgrading using a panel data set of 52 Japanese manufacturing industries over the period 1980–2000. Previous studies have failed to find a significant effect of fragmentation trade intensity on skill upgrading for the Japanese industry-level data (Sakurai, 2000; Ito and Fukao, 2005). In particular, these studies have not been able to replicate the commonly found results for the USA and other OECD countries. The present chapter improves upon the existing empirical framework by incorporating a better measure of the fragmentation trade intensity. It also explicitly allows for the possible differential impact of fragmentation trade intensity with developing East Asian countries and high-income countries. It was found that increased fragmentation trade with developing East Asian countries significantly contributed to change in skilled worker employment in Japanese manufacturing over the period 1980–2000. At the same time, component imports from OECD countries had skills downgrading effects. The findings in this chapter hence suggest that production fragmentation is one of the most significant demand shifters for the skills composition of Japanese manufacturing. This is clearly related to changes
The impact of fragmentation on skill upgrading
103
in the employment pattern observed in Chapter 5. More fragmentation trade with East Asian countries would be expected to lead to increasing demand for skilled workers, at the same time putting downward pressure on unskilled workers. The next chapter continues to examine labour market implications of production fragmentation, but the focus is slightly different from the present chapter. Adjustment for home (domestic) employment of Japanese manufacturing MNEs is examined relative to expanded overseas operations induced by the expanded production fragmentation.
APPENDIX 6.1 Table 6A.1
COUNTRY COMPOSITION
Classification of countries
Developing East Asian Countries
OECD Countries
(10 countries) China Hong Kong Indonesia Korea, Republic of Malaysia Philippines Singapore Taiwan Thailand Vietnam
(21 countries) Australia Austria Belgium Canada Denmark Finland France Germany Greece Ireland Italy Mexico Netherlands New Zealand Norway Portugal Spain Sweden Switzerland UK USA
104
International fragmentation of production
APPENDIX 6.2
JAPAN INDUSTRIAL PRODUCTIVITY 2006 DATABASE
The Japan Industrial Productivity 2006 database (JIP 2006) is the outcome of a research collaboration between the Research Institute of Economy, Trade and Industry (RIETI) as a part of its research project ‘Study on industry-level and firm-level productivity in Japan’ and the Institute of Economic Research, Hitotsubashi University as a part of the ‘21st-century COE program, research unit for statistical analysis in the social sciences (Hi-Stat) project’. The JIP 2006 can be accessed at http://www.rieti.go.jp/ en/database/d05.html. The original version of the JIP database (JIP 2003) was compiled by the Economic and Social Research Institute (ESRI), Cabinet Office, Government of Japan as part of its research project on ‘Japan’s potential growth’ and the Hi-Stat project of Hitotsubashi University. A brief description of the variables used in the regression analysis is given below. Value-added is derived from gross output (100 millions in Japanese yen) and intermediate input use (100 millions in Japanese yen). Gross output is measured as the sum of industry shipment, revenue from repairing and fixing services and revenue from performing subcontracting work. Intermediate inputs are defined as the sum of raw materials, fuel, electricity and subcontracting expenditure. The notable feature of the JIP database is that a price index of intermediate input use is constructed, making it possible to convert the nominal values into the real series. Therefore, real value-added is approximated for a given industry by subtracting real intermediate input from real gross output. Capital stock (100 millions in Japanese yen) refers to the nominal book value of tangible fixed assets including buildings, machinery, tools and transport equipment. Nominal R&D expenditures (100 millions of Japanese yen) are not available in JIP 2006, but are available in JIP 2003. R&D expenditure is reported in the industry classification of JIP 2003 and this series is updated to JIP 2006. A close inspection of the concordance table between JIP 2003 and JIP 2006 industry classification reveals that some JIP 2003 industries is further disaggregated and others are aggregated in JIP 2006. In the case of the disaggregation of industry from JIP 2003, it is assumed that R&D expenditure does not vary across the corresponding JIP 2006 industries. On the other hand, in the case of aggregation, the average value of R&D expenditure in JIP 2003 is used for the corresponding JIP 2006 industry. Data on the employment share of non-production and production workers in JIP 2006 are originally from the Population Census of Japan, produced by the Statistics Bureau, Japan Ministry of Internal Affairs and Communication. This is conducted every five years, covering
The impact of fragmentation on skill upgrading
105
detailed occupational categories (three digit, close to 300 different occupations) and industries. Non-production workers are defined as those with the occupation of professional and technical, managers and administrators, clerical and secretarial, sales and services. Production workers are plant and machine operators and also engage in craft and related occupations.
APPENDIX 6.3
RE-ESTIMATION OF FEENSTRA AND HANSON (1999)
This appendix examines the sensitivity of the results of Feenstra and Hanson (1999) to the use of the alternative measure of fragmentation used in this chapter in place of their measure. The data set used in Feenstra and Hanson (1999) is drawn from Robert Feenstra’s website, available at http://www.econ.ucdavis.edu/faculty/fzfeens/textbook.html. This data set covers 447 four-digit SIC industry data over the period 1979–90. The dependent variable is the change in the share of skilled (nonproduction) workers in US manufacturing over the period 1979–90. Feenstra and Hanson (1999) constructed two measures of foreign outsourcing intensity at each industry level – broad and narrow outsourcing (see Chapter 2). The broad measure covers imports of intermediate inputs from all other industries whereas narrow measure is limited to imported intermediate inputs from the same two-digit SIC industry. The latter measure is preferred since production fragmentation essentially takes place within industry rather than across industries. Narrow outsourcing and the difference between broad and narrow outsourcing are included in regression analysis, as in Feenstra and Hanson (1999). Difference outsourcing is obtained by subtracting narrow outsourcing from broad outsourcing. Hence, it represents the imported intermediate inputs from the outside of the two-digit industry. The other explanatory variables include the shipment of each industry (a proxy for output), the capital:output ratio (K:Y), the share of computers and other high-tech capital in total capital stock and the share of the computer expenditure in total investment share. The last two variables are meant to explicitly capture skill-biased technological changes (for more detail, see Feenstra and Hanson, 1999, pp. 925–7). All explanatory variables are weighted by industry share of the total manufacturing wage bills. The construction of the fragmentation intensity measures based on component exports and imports follows several steps. The trade data were first compiled from the UN Comtrade database. The five-digit SITC of parts and components were then mapped into the four-digit SIC industry level using the concordance to match with the Feenstra-Hanson data set.14
106
Table 6A.2
International fragmentation of production
Comparison of the two measures of the fragmentation intensity for US manufacturing, 1979–90
Outsourcing Intensity Measure (weighted average) The Feenstra-Hanson measure of: Outsourcing (narrow) Outsourcing (broad) The new measure of: Imports intensity of fragmentation trade Imports from East Asian countries Imports from OECD countries Exports intensity of fragmentation trade Exports to East Asian countries Exports to OECD countries
1979 (%)
1990 (%)
Annual Average Change (%)
3.12 7.83
4.86 11.84
0.16 0.37
4.2
8.92
0.43
0.69 3.39 2.13
1.48 7.37 3.25
0.07 0.36 0.10
0.23 1.37
0.44 2.49
0.02 0.10
Note: All variables are weighted by the industry share of total manufacturing shipments. They are computed over 447 four-digit SIC industries. Variable definitions: Outsourcing (broad) = (imported intermediate inputs)/(total non-energy intermediates)*100; Outsourcing (narrow) = (imported intermediate inputs in the same two-digit industry as a buyer)/(total non-energy intermediates)*100; Imports intensity of fragmentation trade = (imports of parts and components)/(total nonenergy intermediates)*100; Exports intensity of fragmentation trade = (exports of parts and components)/(total shipments)*100. Source: The data set used in Feenstra and Hanson (1999) is available at http://www. econ.ucdavis.edu/faculty/fzfeens/textbook.html. The new measure of fragmentation trade intensity is compiled from the UN Comtrade database.
Then the imports and exports intensity of fragmentation trade were computed by taking the ratio of parts and components to gross outputs for each industry (see equation (6.3)). The average imports intensity of fragmentation trade has grown from 4.20 per cent in 1979 to 8.92 per cent in 1990 (Table 6A.2). This increase is similar to the increase in intensity of imported intermediate inputs during the same period. However, this is not surprising since imports intensity by parts and components is a strict subset of imported intermediate inputs (broad outsourcing). Most of the increase in import intensity by parts and components comes from trade with OECD countries during this period.15 The exports intensity of fragmentation trade in US manufacturing also rose from 2.13 per cent in 1979 to 3.25 per cent in 1990. Reg. 1 of Table 6A.3 is based on the same specification as in Feenstra
The impact of fragmentation on skill upgrading
Table 6A.3
107
Evidence of skill upgrading effects in US manufacturing, 1979–90
Dependant variable=change in the wage share of skilled (non-production) workers Explanatory Variables:
Mean
Δln (K/Y)
0.706
Δln (Y)
1.541
Outsourcing (narrow)
0.223
Outsourcing (difference) FRGimport:
0.200
from East Asian countries from OECD countries FRGexports: to East Asian countries to OECD countries Computer share in capital High-tech share in capital Constant R2 F-Statistic Root MSE Number of observations
Reg. 1 0.042 (0.014)*** 0.017 (0.008)** 0.245 (0.169) 0.121 (0.046)**
0.360
Reg. 2 0.049 (0.011)*** 0.027 (0.007)***
0.325 (0.048)*** −0.028 (0.011)**
0.296 0.001
−0.009 (0.019)
0.007 0.046
0.144
0.044 (0.012)*** 0.023 (0.007)***
0.021 (0.014)
0.083
0.251
Reg. 3
0.206 (0.102)* −0.039 (0.128) 0.207 (0.045)*** 0.163 7.00*** 0.387 447
0.220 (0.109)* −0.041 (0.144) 0.256 (0.044)*** 0.125 6.76*** 0.397 447
−0.092 (0.083) 0.011 (0.045) 0.199 (0.103)* −0.027 (0.126) 0.257 (0.042)*** 0.156 44.47*** 0.390 447
Note: The mean of the dependent variable equals 0.389. The first column shows mean values of the independent variables for 1979–90. All regressions and means are computed over 447 four-digit SIC industries and weighted by the average industry share of the manufacturing wage bill. Δln (K/Y) is the average change in the log capital-shipment ratio, and Δln (Y) is the average annual change in log real shipments. The outsourcing variables and the computer and high-tech shares are in annual change. Standard errors based on White’s heteroscedasticity correction clustered by the two-digit SIC industry are given in parentheses, with statistical significance (two-tailed test) denoted as: *** 1%, ** 5%, and * 10%.
108
International fragmentation of production
and Hanson (1999) (with their measure of outsourcing intensity).16 The estimated coefficient for outsourcing (narrow) is exactly 0.246, as reported in Feenstra and Hanson (1999). The implied contribution to the average change of the skilled workers’ wage bill share is 14 per cent. While not highlighted by Feenstra and Hanson (1999), outsourcing (narrow) is not statistically significant at all. On the other hand, outsourcing (difference) is statistically significant at the 5 per cent level. Reg. 2 is obtained by replacing the Feenstra-Hanson outsourcing measures with the imports and exports intensity of fragmentation trade, as proposed in this chapter. Both fragmentation intensity measures are now found to be of much smaller magnitude, with little statistical significance. The contribution of imports intensity to the change in the skilled workers’ wage share turns out to be only 1.9 per cent. Again, a comparison of these results supports the argument that the limited industry coverage by the new measure is not a serious limitation. Reg. 3 of Table 6A.3 is the result based on the geographical separation of total fragmentation trade intensity into East Asian and OECD countries. It is strikingly consistent with the finding for Japanese manufacturing. The estimated coefficient of the imports intensity of fragmentation trade with East Asian countries is statistically significant at the 1 per cent level. This accounts for 7 per cent of the average change in the skilled workers’ wage share. The estimated coefficient for the imports intensity of fragmentation trade with OECD countries indicates the skill downgrading effects with the negative sign. This explains about 2.2 per cent of the shift away from skilled workers in the total wage bills in US manufacturing. Thus, it seems that the geographical distinction of the fragmentation intensity also matters in appropriately capturing the effect of production fragmentation on skill upgrading in US manufacturing.
NOTES 1.
2. 3. 4.
In the previous Japanese study, Sasaki and Sakura (2005) also considered distinguishing the sources of Japan’s manufacturing imports, but only focused on total manufacturing imports from East Asian countries without making a distinction on fragmentation trade. As will be argued, the simple total import penetration ratio lacks precision on measuring the intensity of fragmentation trade. Strauss-Kahn (2004) also distinguishes imports of intermediate inputs between OECD and non-OECD countries in her study on French manufacturing. I would like to thank Professor John Ries, the editor of the Journal of the Japanese and International Economies, for pointing this out. While not directly dealing with the issue of skill upgrading, Fukao et al. (2003) and Tomiura (2005) also examine the pattern of Japan’s foreign outsourcing. Two data sources are generally available for compiling the wage bills of nonproduction/production workers at industry level of Japanese manufacturing, the
The impact of fragmentation on skill upgrading
5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.
109
Census of Manufactures (CM) and the Basic Survey on Wage Structure (BSWS) (see Appendix 5.2 for descriptions of these data sources). The CM includes cash earnings of production and non-production workers at the detailed four-digit level of Japan Statistical Industry Classification (JSIC). However, since 1990 this information has become unavailable in the published data of the CM. The BSWS provides wage earnings data for non-production and production workers, but they are available only for the total manufacturing industry and from 1985 onwards. Berman et al. (1994) and Feenstra and Hanson (1999) each prefer the use of valueadded, but in the empirical application this alternates with the value of industry shipment (that is, gross outputs) due to the absence of reliable price deflators. The estimation results are, however, less sensitive to the use of gross output. Feenstra and Hanson (1996, 1999) alternated more specific high-tech capital variables such as computer investment whereas other studies have used the employee computer usage (Autor et al., 1998). Available at http://www.rieti.go.jp/en/database/d05.html also, see Fukao et al (2006). Chapter 2 describes the procedure for trade data compilation in greater details. Available at http://www.esri.go.jp/en/archive/bun/abstract/bun170index-e.html. Berman et al. (1994), Feenstra and Hansen (1996, 1999), Ito and Fukao (2005) and others use the first-differencing estimator. I am grateful to Dr Keiko Ito for providing the data of these outsourcing measures for JIP 2006 industries. This is computed by multiplying the estimates coefficient by the weighted average of change in the import intensity of fragmentation, divided by the weighted average of change in the dependent variable. Available at http://www.macalester.edu/research/economics/page/haveman/Trade. Resources/tradeconcordances.html#FromSITC. Note that Japan is included in OECD countries. More specifically, it refers to Regression (2) in table 3, p. 927 in Feenstra and Hanson (1999).
7
Overseas operations and home employment of Japanese multinational enterprises
7.1
INTRODUCTION
This chapter considers the labour market implications of production fragmentation from a different perspective. It undertakes a firm-level econometric analysis of the effects of the expanded overseas operations of Japanese manufacturing MNEs on home (domestic) employment. As will be shown in Section 7.2, overseas operations of Japanese manufacturing MNEs during this period have mainly been driven by international fragmentation of production. The controversy over the possible adverse effects of overseas production by MNEs on employment in the home economy first arose in the USA in the late 1960s and has gained increased attention in policy circles of industrial countries in recent years with the growing importance of international fragmentation of production (Lipsey, 1995; Harrison and McMillan, 2006). This phenomenon, the possible substitution of home employment of MNEs with increased overseas production, is known in the literature as the ‘exporting jobs’ hypothesis (Kravis and Lipsey, 1988). It also became the subject of heated policy debate in Japan under the label of ‘manufacturing hollowing-out’ (sangō kudouka) following a surge of Japanese FDI outflow associated with the spread of production networks to low-cost countries in East Asia from the mid-1980s. Numerous journalistic reports on possible job losses to overseas largely based on anecdotal evidence have been published (for example, Nikkei newspaper). However, only a few systematic empirical studies are available and most of them have largely focused on the overall trends and patterns based on readily available FDI data at the aggregated industry level (Fukao, 1995; Fukao and Amano, 1998; Fukao and Yuan, 2001). There is virtually no direct evidence of how Japanese MNEs adjust home employment in response to changes in the production capacity of foreign affiliates. This is certainly an area where studies on Japanese MNEs lag behind those of the USA and Sweden-based MNEs (Lipsey, 1995; Brainard and 110
Overseas operations and home employment of Japanese MNEs
111
Riker, 1997a, 1997b; Braconier and Ekholm, 2000; Desai et al., 2005; Fors and Kokko, 2000; Harrison and McMillan, 2006).1 This chapter aims to fill this gap. The analysis is based on a data set compiled from the unpublished returns to two firm-level surveys; the Basic Survey of Business Structure and Activity and the Basic Survey of Overseas Japanese Business Activity, collected by the Japanese Ministry of Economy, Trade and Industry (METI) over the period 1991– 2002.2 Fortunately, Japan is one of the few countries, besides the USA and Sweden, where detailed information on the overseas operations of national firms has been collected systematically over a long period of time. Recently, these firm-level surveys containing direct measures of Japanese MNEs’ performance have become increasingly available to researchers (Kimura and Ando, 2003, 2005; Ando and Kimura, 2005; Hijzen et al., 2006; Kimura and Kiyota, 2006; Shimizutani and Todo, 2007; Todo and Shimizutani, 2008). Appendix 7.1 describes the data used in this chapter. Section 7.2 discusses the patterns and trends of home and overseas operations of MNEs. Section 7.3 undertakes a survey of the existing empirical evidence for the relationship between the overseas and domestic operations of MNEs. Section 7.4 depicts the empirical framework and explains variable construction and the estimation methodology is presented in Section 7.5. Section 7.6 interprets the results and Section 7.7 concludes the chapter.
7.2 7.2.1
PATTERNS AND TRENDS OF THE HOME AND OVERSEAS OPERATIONS OF MNEs Home Operations
Selected key indicators of home (domestic) operations of MNEs, using the METI firm survey data for the period 1991–2002 are summarized in Table 7.1. Total domestic sales by Japanese parent firms rose from 128 trillion yen in 1991 to 136 trillion yen in 2002 (Table 7.1). Accordingly, the number of parent firms increased from 616 in 1991 to 1114 firms in 2002. On the other hand, the employment figure contracted from about 2.2 million in 1991 to 2 million in 2002. This indicates that about 180 000 jobs were shed in the home employment of MNEs over the period. However, this loss of home employment by MNEs’ parent firms could be considered relatively small, compared with the 3 million jobs lost in total Japanese manufacturing during the same period (see Chapter 5). The share of parent firms of MNEs in total manufacturing accounted for an average of about 6.6 per cent over the period 1991–2002. While
112
Source:
Note:
616 863 782 902 950 914 989 926 984 1144 907.0
128.8 124.2 128.5 143.2 142.7 131.4 138.6 144.8 139.3 136.8 135.8
(trillions of yen) 2245 2275 2267 2328 2292 2188 2261 2215 2121 2066 2225.8
(in 1000) 4.5 6.3 5.4 6.3 6.7 6.5 7.1 7.6 7.3 8.7 6.6
48.2 49.6 49.0 51.7 52.8 52.0 54.4 57.2 55.8 54.6 52.5
Number Output of of
37.2 38.3 38.0 39.4 40.1 39.2 41.1 44.0 41.4 42.1 40.1
Employment of
43.9 45.2 44.6 46.5 47.6 46.7 48.6 51.9 49.5 50.8 47.5
Workers earnings
47.1 49.3 48.7 50.5 51.9 50.4 52.4 55.7 53.5 52.9 51.2
Capital stock of
– 86.2 80.8 82.5 82.5 82.8 84.8 86.5 83.2 86.1 83.9
Exports of
Share in Total Japanese Manufacturing of:
Based on the METI database, which is explained in Appendix 7.1.
See the definition of MNEs in the main text. The survey data are not available for the years 1992 and 1993.
1991 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
(unit)
MNE parent firms
Employment of:
Number of:
Year
Sales of:
Selected indicators of parent firms of Japanese manufacturing MNEs, 1991–2002
Table 7.1
– 60.6 60.7 58.5 60.3 60.9 62.2 69.1 65.0 64.8 62.5
Imports of
73.5 74.6 75.7 77.2 77.8 77.8 79.0 81.2 78.4 80.6 77.6
R&D of MNEs (%)
Overseas operations and home employment of Japanese MNEs
113
this seems small, these parent firms of MNEs contributed the majority of economic activity to total manufacturing over 1991–2002. In 2002 parent firms of MNEs accounted for close to 55 per cent of aggregate manufacturing outputs and over 40 per cent of aggregate manufacturing employment as well as more than half of aggregate capital stock. Almost half of manufacturing workers’ compensation was also paid by MNEs. Not surprisingly, parent firms conducted the major proportion of international trade, accounting for over 80 per cent and 60 per cent of exports and imports, respectively, and contributed over a 75 per cent share of the research and development (R&D) expenditure in total manufacturing over the same period. These figures suggest that any effects on the operations of MNEs are likely to be deeply felt in the home economy. The MNE dominance in domestic manufacturing is not unique to Japan. For example, MNEs in US manufacturing accounted for over 60 per cent of total manufacturing sales, over 70 per cent of total exports, almost 60 per cent of manufacturing employment and 82 per cent of domestic R&D expenditure during 1982–99, although the number of US MNEs also looked small (Harrison and McMillan, 2006). Similar figures are reported for Swedish MNEs (Fors and Kokko, 2000). 7.2.2
Overseas Operations
Table 7.2 summarizes the data on the key performance indicators of foreign affiliates of manufacturing MNEs over the period 1981–2002. The number of foreign affiliates steadily increased from 2656 in 1989 to over 10 000 in 2000, but the following two years (2001 and 2002) saw some decline in the number of foreign affiliates. Similarly, sales of foreign affiliates achieved a five-fold increase between 1989 and 2000, but they significantly dropped between 2001 and 2002. Employment of foreign affiliates continuously expanded since 1989, and reached close to 3 million in 2002. This indicates that the number of workers employed in foreign affiliates is higher than that of workers employed by parent firms of MNEs (Table 7.1). The data in Table 7.2 also indicate an increase in the size of foreign affiliates in terms of average employment and output over the period under study. On average, sales ratios both to Japan and other countries have also been increasing since 1989, while the local sales ratio has remained stable at around 65–70 per cent over the period. There is also some indication of upgrading in the technological capacity of the foreign affiliates of Japanese MNEs. This is consistent with the finding from Odagiri and Yasuda (1996) that overseas R&D activity has been rising rapidly despite a slow beginning. Foreign affiliates of Japanese MNEs are heavily concentrated in general
114
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Year
2656 3407 3535 3040 4548 7992 7345 7626 9279 9069
(unit)
914 1242 1261 986 1516 1972 1986 2258 2540 2339
(in 1000)
22.4 26.2 25.4 25.1 29.2 85.2 87.1 106.7 110.8 109.6
8.4 7.7 7.7 9.0 6.9 11.7 13.2 15.2 13.0 13.3
344.1 364.4 369.6 371.0 357.0 267.4 310.6 296.6 274.5 294.3
(unit)
2.7 – – 3.0 – 3.1 4.2 3.7 4.2 4.1
24.5 21.1 20.8 24.3 19.4 43.6 42.6 51.2 47.4 45.2
(millions (output of yen) per worker) 1.2 – – 3.7 – 0.6 1.0 3.0 1.1 1.1
(%)
74.8 69.5 67.0 75.1 65.8 68.7 74.5 70.8 68.7 71.6
(%)
foreign affiliate of manufacturing MNEs
17.9 15.3 17.8 25.0 18.1 15.6 29.3 21.8 22.7 24.6
(%)
18.2 17.7 18.7 27.3 17.3 15.8 30.5 24.6 22.9 34.0
(%)
52.9 48.9 50.3 55.1 50.9 41.0 58.8 48.8 47.1 51.3
(%)
44.9 41.2 40.9 49.9 38.8 43.0 53.7 44.4 45.5 47.6
(%)
13.4 11.1 12.5 22.9 11.4 16.3 21.1 25.9 23.7 32.1
(%)
Average Average Average Average Average Average Average Average Average Average Average Sales Employ- Wage labour R&D local sales sales purpurpurment rates produc- intensity sales ratio to ratio to chase chase chase tivity ratio Japan other ratio ratio ratio counfrom from from tries local Japan other market country
(trillions (billons of yen) of yen)
Sales of:
Key indicators of foreign affiliates of Japanese manufacturing MNEs, 1989–2002
Number Employof: ment of:
Table 7.2
115
211.2
2812 3049 2645 2844 1930
189.7
102.9 112.8 64.2 64.9 42.5 16.7
11.7 12.3 10.4 9.8 1.41 19.2
289.8 289.9 398.8 410.3 66.2 25.9
3.6 4.1 3.7 3.4 0.7 −2.0
40.2 42.4 26.1 24.0 −0.5 41.7
1.2 4.2 3.3 1.7 0.5 −11.2
72.2 70.7 68.6 66.4 −8.4 41.9
23.6 21.8 28.5 25.4 7.5 36.8
25.1 23.2 37.9 24.9 6.7 6.2
50.7 49.6 58.2 56.2 3.3
−12.7
46.9 46.0 43.4 39.2 −5.7
49.3
25.8 23.3 32.8 20.0 6.6
Source:
Based on the METI data set, which is explained in Appendix 7.1.
Note: Average labour productivity is measured by output per worker. R&D intensity refers to the ratio of R&D expenditures to output. Worker compensation and R&D expenditures for 1990, 1991 and 1993 are not contained in the original returns to the METI foreign affiliate survey. The METI foreign affiliate survey is available from 1989.
1999 9828 2000 10549 2001 7068 2002 8006 Level 5350 change Percent201.4 age change
116
International fragmentation of production
Table 7.3
Industry distributions of foreign affiliates of Japanese manufacturing MNEs (%), 1989–2002 Firm
Employment
Sales
1989 1995 2002 1989 1995 2002 1989 1995 2002 Textile Chemicals Primary metals Metal goods General machinery Electronics machinery Information and communications Transport equipment Scientific equipment Other manufacturing Total manufacturing Unit
Source:
6.9 13.6 6.5 3.2 11.0
8.5 13.6 6.4 3.0 10.6
6.0 14.4 5.4 2.9 10.6
7.8 7.7 7.5 1.2 6.3
8.2 7.5 5.8 1.2 6.4
5.5 6.3 3.9 1.2 5.5
2.1 8.9 6.7 0.7 9.7
1.9 9.3 5.8 0.5 6.3
1.3 10.2 3.9 0.5 5.5
7.2
6.3
7.0
8.8
6.8
7.3
5.6
4.6
4.6
15.6
16.9
16.3
23.2
26.2
27.1
21.3
28.5
24.2
12.1
12.1
15.9
17.8
19.9
23.7
30.5
28.6
36.7
3.8
3.4
3.6
2.6
2.4
3.1
1.9
1.7
2.2
20.1
19.4
17.9
17.3
15.7
16.3
12.8
12.6
10.9
100
100
100
100
100
100
100
100
100
Number of firms
Employment in Sales in billions of 1000 yen 5834 8637 8014 1485 2618 3251 32816 50591 72295 Based on the METI database, which is explained in Appendix 7.1.
machinery, electronics, information and communications, and transport equipment industries (Table 7.3). There is an increasing share of sales in the transport equipment industry, growing from about 30 per cent in 1989 to 37 per cent in 2002. The similar expansion of sales can be seen for the information and communications industry. While the electronics machinery industry has been one of the most important components in Japanese outward FDI, its sale share stagnated over the whole period. There was even a slight decline in the employment and sales share from 1989 to 2002. Foreign affiliates of Japanese MNEs were overwhelmingly concentrated in East Asian countries in the period under review (Table 7.4). About 60 per cent of the foreign affiliates of Japanese MNEs were located in East Asia in 2002, up from 49 per cent in 1989, with a corresponding employment share
Overseas operations and home employment of Japanese MNEs
117
Table 7.4 Geographical distribution of foreign affiliates of Japanese MNE (%), 1989–2002 Number of:
Employment of:
Sales of:
foreign affiliates of manufacturing MNEs 1989 East Asia China South Asia Oceania North America USA South America European Union Eastern and Central Europe Africa World Unit
Source:
1995
2002
1989
1995
2002
1989
1995
2002
49.2 3.6 1.5 3.3 24.7
56.8 16.4 1.2 2.9 20.2
59.0 19.3 1.7 2.4 18.4
48.5 1.8 2.2 3.1 25.3
58.8 12.7 2.0 1.5 20.5
64.6 22.2 1.9 1.0 17.3
30.4 0.4 0.8 4.5 46.0
33.0 2.1 0.8 2.4 41.9
32.6 6.1 1.1 2.4 41.9
22.5 6.4
18.6 4.7
17.2 3.7
23.8 10.2
19.4 6.1
16.5 3.7
43.2 3.5
39.7 4.3
38.9 3.4
12.7
12.0
11.6
8.9
9.7
8.5
13.5
16.0
16.5
0.6
0.8
1.1
0.3
0.4
0.8
0.4
0.7
0.9
0.7 0.4 0.4 0.6 0.1 0.5 0.2 0.1 0.4 100 100 100 100 100 100 100 100 100 6197
9042
8265
1533
2603 3285 (in 1000)
35886 52429 73886 (billions of yen)
Based on the METI database, which is explained in Appendix 7.1.
growth. Within East Asia the rise of China as a destination for foreign affiliates is impressive. Only 3.6 per cent of Japanese foreign affiliates were operating in China in 1989, but that figure had jumped sharply to over 19 per cent by 2002. Accordingly, 1.8 per cent of the employment share had grown to 22 per cent in 2002. This geographical shift of overseas operations has facilitated the creation of an international production network by Japanese MNEs in East Asia (Kimura and Ando, 2005). Foreign affiliates of Japanese MNEs began to turn away from North America and the EU over the years. For example, the employment share of foreign affiliates in the USA fell from 24 per cent in 1989 to 16.5 per cent in 2002; 22.5 per cent of the foreign affiliates of Japanese MNEs were located in the USA in 1989, but by 2002 this had fallen to 17 per cent. Other developed country regions such as the EU countries experienced a
118
International fragmentation of production
slight decline or no change in foreign affiliate and employment share. This shift in the location of overseas operations of Japanese manufacturing MNEs coincides with patterns of Japan’s fragmentation trade, discussed in Chapter 3. Interestingly, while there is strong evidence of increased concentration of the operations of foreign affiliates in East Asian countries over time, the geographical composition of overseas sales are still concentrated in the USA. The share of foreign affiliate sales in East Asian countries remained at around 30 per cent, although with a notable increase of sales share in the Chinese market. The sales share in North America also remained at virtually the same level between 1995 and 2002, at about 42 per cent. For the EU, the share of foreign affiliate sales actually grew from 13.5 per cent in 1989 to 16.5 per cent in 2002, despite a decline in the share of employment and foreign affiliates. Overall, in this period there was a strong indication that the overseas operations of Japanese manufacturing MNEs were mostly driven by international fragmentation of production. Their overseas operations have been shifting towards East Asia, which is consistent with patterns of fragmentation trade, reviewed in Chapter 3. By looking at the aggregated data, overseas operations appeared to have grown at the cost of home operations of MNEs over the period 1991–2002.
7.3
THE EFFECT OF THE OVERSEAS OPERATIONS OF MNEs ON DOMESTIC OPERATIONS
International fragmentation of production is generally facilitated by a vertical type of MNE separating a vertically integrated production process between parent MNEs and their foreign affiliates. Under operations of vertical MNEs overseas and domestic employment can be substitutes, since some domestic operations are relocated to overseas operations (the substitution effect). However, this simplistic substitution ignores the positive effects of overseas operations on the home operations of MNEs. It is equally possible that increased overseas operations might enhance the scale of home economic activities, including sales, employment and R&D spending, due to better resource allocation and the expanded market overseas (the scale effect). In this case, the overseas and domestic employment can be complementary. Therefore, the net impact of increased overseas operations on home economic activity can be either positive or negative, depending on the magnitude of the scale and the substitution effects (Hanson et al., 2003). Different types of MNEs can also complicate the net effect of overseas
Overseas operations and home employment of Japanese MNEs
119
operations (Caves 1996). The horizontal type of MNE overseas operation is motivated by the objective of expanding overall sales. In this sense, expanded overseas operations may have little effect on the scale of the domestic operation of MNEs. However, it is equally possible that the domestic operations of these MNEs might be facilitated by the expanded worldwide scale of production.3 To date, the theory of MNEs does not provide clear-cut predictions about the possible effects of foreign production on home operations. For this reason, the empirical literature has gained importance on this subject. A large amount of the empirical research on the effects of overseas operations on home operations is based on US MNEs (Kravis and Lipsey, 1988; Lipsey, 1995; Brainard and Riker, 1997a, 1997b; Hanson et al., 2003; Desai et al., 2005; Harrison and McMillan, 2006). These studies make use of firm-level survey data, conducted by the Bureau of Economics Analysis (BEA), US Department of Commerce. The BEA data are a comprehensive and integrated data set for tracking the operations of US MNEs’ non-bank foreign affiliates in host countries and the operation of parent firms in their home countries.4 The survey format includes information on the classification of industry, sales, trade in goods and services, employment, wages, assets, expenditure for plant and equipment, and R&D expenditure. Benchmark surveys are conducted every five years covering an extensive range of information on firms’ activities. While the BEA must maintain strict confidentiality over the information contained in the survey, a special research programme allows researchers to access the original return data. Among such researchers, Kravis and Lipsey (1988) and Lipsey (1995) make initial attempts to examine the impact of foreign production on the home employment of US MNEs. Due to insufficient coverage of variables, their regression analysis does not take a systematic approach. Nevertheless, a higher level of foreign affiliate production in developing countries is found to be associated with lower home employment for a given level of home production. Brainard and Riker (1997a) develop a more systematic analysis and estimated the foreign affiliate cross-wage elasticities of parent firms’ labour demand for the period 1983–92. The framework they developed includes key factors influencing the labour demand of parent firms and changes in foreign affiliates’ wages. They find evidence of a substitution relationship between foreign and domestic employment, although the degree of substitution is low (Brainard and Riker, 1997a). On the other hand, a strong substitution relationship was found among the various foreign affiliates of MNEs, operating in developing countries. The evidence indicates any employment substitution effect predominantly takes place between the
120
International fragmentation of production
foreign affiliates of MNEs operating in overseas locations rather than between parents and their foreign affiliates. If anything, parent firms adjusted employment very little in response to changes in foreign affiliate wages. Using the same BEA data, Hanson et al. (2003) extended the analysis of Brainard and Riker (1997a) in two important ways: first, they cover the more recent time period 1989–99. They had previously argued in Hanson et al. (2001) that the broad patterns of US MNEs’ activity were quite different between the 1980s and 1990s. The second extension focus on the relationship between the scale and scope effects of the overseas operations of MNEs. Expansion in the overseas operations of MNEs might affect the scale of domestic economic activity, and at the same time change the composition of the parents’ home economic activities (the scope effect). For example, as the overseas operations of MNEs concentrate on the labourintensive segment of vertical production, their remaining home operations may become more specialized in R&D, management, headquarters’ services, design and other skill-intensive services. For this purpose their labour demand regressions consider two dependent variables, total and R&D employment. Although the quantitative effect is small, Hanson et al. (2003) find that expansion in the sales of foreign affiliates of US MNEs raises the labour demand for their home operations. This finding supports a hypothesis of a mild complementary relationship between increased overseas sales and parent employment. Their second main finding is that the relationship between the parent and its foreign affiliates appears to depend on the skilled/unskilled labour costs of foreign affiliates. When the cost of skilled labour is lower in foreign affiliates, the demand for home labour appears to increase. This result suggests that changes in the prices of high-skilled employment in foreign affiliates tend to increase overall employment, both in foreign affiliates and parent firms. On the other hand, where the cost of unskilled labour for foreign affiliates is lower, the US parent firms decrease the demand for home employment. Desai et al. (2005) use a broader analytical framework than the preceding studies by directly testing the effects of foreign affiliate employment, property, plant and equipment investment, assets, worker compensation and sales on these home operations. The BEA data in Desai et al. (2005) cover the period 1982–99. They carefully take into account the problem of endogeneity employing GDP per capita of host country as instrumental, because any decisions on the overseas and domestic operations of MNEs are jointly rather than independently determined. Desai et al. (2005) find evidence of increased overseas operations of MNEs enhancing the scale of home operations. A 10 per cent greater
Overseas operations and home employment of Japanese MNEs
121
accumulation of foreign property plant and equipment is associated with a 2.2 per cent increase in domestic net property plant and equipment. Similarly, a 10 per cent rise in foreign employee compensation is associated with a 4 per cent greater domestic employee compensation, and a 10 per cent higher number of foreign employees with a 2.5 per cent higher number of domestic employees. In sum, the results amply support the hypothesis that expanded operations of US MNEs’ foreign affiliates have stimulated the domestic activity of US parent firms over the last two decades. Harrison and McMillan (2006) explore the BEA data sets, but cast the data series back to 1977. In addition to dealing with the problems of the endogenity bias, the sample selection bias problem is also taken into account using a two-stage Heckman sample selection approach and modelling the survival of parent firms. The sample in the BEA data set is highly unbalanced due to the significant amount of parent firm entry and exit from 1977 to 1999. Harrison and McMillan (2006) compare the estimates of labour demand elasticities by parent firms across three specifications, the factor demand function with factor quantities, factor costs and the translog cost function approach. They find strong evidence that the employment of foreign affiliates in developing countries substitutes for the home employment of parent firms in US manufacturing. However, the effect is quantitatively small. On the other hand, home employment in the USA and the employment of foreign affiliates in developed countries are found to be complementary, characterized by a decline in employment both at home and in developed countries. In other words, any decline in employment of foreign affiliates in developed countries leads to some contraction in employment of the parent firm in the USA. By and large, the finding of Harrison and McMillan (2006) is consistent with that of Brainard and Riker (1997a). A number of studies have been undertaken using firm-level data on Swedish MNEs. The Research Institute of Industrial Economics (IUI) in Stockholm has been conducting detailed surveys on the foreign operations of Swedish firms every four years since the mid-1960s (Fors and Kokko, 2000; Lipsey, 2003). One of the latest studies (Braconier and Ekholm, 2000) using this firm-level survey data follows the previous US MNE studies by estimating the cross-wage elasticities of labour demand by Swedish MNEs for the period 1970–94. There is a slight difference in the results between US and Swedish MNEs. Braconier and Ekholm (2000) uncover some evidence of a mild substitution relationship between home and foreign affiliate employment in developed countries. However, they do not find any evidence of substitution between home and foreign affiliate employment in low-income countries. The difference in this finding compared with US
122
International fragmentation of production
MNE studies is explained by the difference in the nature of the strategy adopted by the Swedish MNEs. Of the available studies on Japanese MNEs, a disproportionately large number of studies have focused on the relationship between expanded overseas production and exports of home countries in Japan (Fukao and Amano, 1998; Lipsey et al., 1999; Head and Ries, 2001; Kimura and Kiyota, 2006). Fukao (1995) makes an early attempt to examine the possible impacts of foreign affiliate production on domestic employment. Fukao and Yuan (2001) develop a three-digit level of cross-industry data, concerning the impact of FDI on the employment growth rate over the period 1989 to 1998. The unique feature of their study is the differentiation of FDI by investment motivation and region of the host country. They find that Japanese FDI in East Asia led to shedding around 600 000 workers in home country employment. They also find that marketoriented FDI in East Asia seemed to increase the amount of home country employment.
7.4
THE ANALYTICAL FRAMEWORK
The regression analysis is based on a reduced form of labour demand equation widely used in this strand of literature (see Navaretti and Venables (2004) for a survey). Following Hamermesh (1993), the standard labour demand can simply be written as follows: ln Liht 5 a 1 b1 ln wiht 1 b2 ln Qiht 1 b3 ln rzht
(7.1)
where subscripts i, h and t denote parent firm, home country and time. The dependent variable (L) is the quantity of home employment; w, Q and r represent own wage rate, output and the price of capital; a proxies the unobserved features such as the parent’s level of technology and firm-specific capital; ln indicates natural logarithm. Hence, the log-linear specification offers the direct interpretation of elasticity between factors, holding the output constant (that is, own-wage elasticity and cross-factor elasticity). The labour demand equation (7.1) is expanded to incorporate a variable capturing overseas operations of foreign affiliates of MNEs (denoted as MNE) and other relevant variables influencing the demand of labour by parent firms. The estimated coefficient of MNE should provide a direct test of the effect of overseas operations on home employment of parent firms. The ‘exporting job’ hypothesis suggests the negative coefficient on MNE. On the other hand, the positive coefficient indicates the scale
Overseas operations and home employment of Japanese MNEs
123
effects dominate the substitution effects of overseas operations on home employment. The own-wage rate of home employment is expected to be negatively related with home employment, given a downward sloping labour demand curve (Hamermesh, 1993). This would suggest that as the cost of home country workers rises, profit-maximizing firms substitute other production inputs. Product demand shocks both at home and in host countries are included in the model (Brainard and Riker, 1997a; Braconier and Ekholm, 2000; Harrison and McMillan, 2006). These variables are expressed by (home) output (Q), time-specific dummy (gt) and GDP per capita of host countries (GDPP). Any shocks to product demand are likely to move labour demand in the same direction (Hasan et al., 2007). Positive shocks on product demand are likely to raise the demand for home employment under the assumption of constant returns to scale. The inclusion of the output scale of parent firms (Q) also controls for the size of parent firms constant when estimating the labour demand equation (Kravis and Lipsey, 1988). Time-specific dummies (gt) capture pure random shocks to the labour demand equation common to all firms, but varying over time. Similarly, foreign demand is proxied by GDP per capita of host countries. The positive impact of the product market in host countries should translate positively into an increase in home employment (the market expansion effect), while the negative demand shocks depress home employment. Labour demand for a given level of output also depends on the cost of capital service (r). The sign of the cross-factor price indicates the nature of the relationship between labour and capital. A positive sign is expected if they are substitutes, and a negative sign if complementary. The level of technology is proxied by the intensity of R&D (denoted as R&D) as well as by unobserved firm- and industry-specific characteristics ( f and f). The sign of R&D depends on the nature of technological progress. It can substitute for employment of parent firms since the new technology may require fewer operational workers. At the same time, technological progress increases demand for skilled workers, engineers and IT-related personnel. Therefore, a priori, the expected sign for R&D is ambiguous. The unobserved heterogeneity across firms can arise from differences in organization, the ageing of capital equipment, the extent of unionization, the quality of output produced or the quality of management inputs. Failing to take them into account might lead to permanent observable differences in output, employment and wages (Westbrook and Tybout, 1993).5 Another factor influencing labour demand is the force of international
124
International fragmentation of production
competition. Tomiura (2004) and Bernard et al. (2006) confirm that manufacturing employment growth in developed countries is negatively related to a rapid increase of imports from low-wage countries. To control for this effect import penetration (IMP) is included in the model. The expected sign of IMP is negative. However, a rapid increase of component imports within manufacturing imports, as documented in Yamashita (Chapter 3, this volume), may raise the demand for home employment. Hence, the estimates sign of IMP could go either way. Based on the discussion above, the econometrics specification takes the following form: ln Liht 5 a0 1 b1 ln wiht 1 b2 ln Qiht 1 b3 ln rzht 1 b4 ln R&Diht 1 b5 ln IMPzht 1 b6 ln MNEiht 1 b7GDPPft 1 fi 1 gt 1 ei,t (7.2) where subscripts z, i and f represent industry, foreign affiliate and host country. The explanatory variables are listed below with the expected sign of each regression coefficient given in the bracket: w Q r R&D IMP MNE GDPP f g e 7.4.1
Home wages rate (2) Gross output (+) The user cost of capital (+ or 2) Research and development intensity (+ or 2) Import penetration (+ or 2) Employment or outputs of foreign affiliates (+ or –) Host country GDP per capita (+) Firm-specific fixed effect Time-specific fixed effect Random error term representing other omitted influences. Variable Construction
We use two different measures of MNE: employment and output of foreign affiliates (MNEQ and MNEL). They are expressed as the weighted average with the weight being the share of worldwide employment and outputs of foreign affiliates. More specifically, the following formula is applied to compute MNEQ and MNEL (a subscript t is suppressed for brevity): m
MNELi,h 5 a wgtj,iLj,f
(7.3)
j51 m
MNEQi,h 5 a wgtj,iQj,f j51
(7.4)
Overseas operations and home employment of Japanese MNEs
125
The weight (wgt) is the share of foreign affiliate j in the worldwide (aggregate) foreign affiliate sales of the corresponding parent firm i.6 GDP per capita of host country is computed in a similar fashion.7 Other variables The dependant variable (L) is measured by the average number of regular employees.8 Unfortunately, the skill composition of home employment is not available in the original METI data. Hence, there is no distinction made between skilled or unskilled labour. Output (Q) is the reported total sales by parent firms. The nominal gross outputs are deflated by the wholesale price index (WPI) at industry level taken from the Bank of Japan.9 The home wage rate is computed by dividing the annual wages and salaries by the annual number of regular workers. Wages and salaries include bonus payments as well as non-wage compensations. The nominal wage series is deflated by the total consumer price index (CPI) taken from the Bank of Japan. The user cost of capital (r) is proxied by the wholesales index of investment goods obtained from the same online database of the Bank of Japan.10 R&D expenditure refers to average values of R&D expenditure spent on knowledge creation and technological upgrading activity by firms, excluding R&D activities done by other firms. R&D intensity is then computed by taking the share of R&D expenditure of the total sales of parent firms. The import penetration ratio (IMP) is computed taking the ratio of imports to apparent domestic absorption, which is defined as (Outputs + Imports) – Exports, and is constructed at the three-digit industry level.
7.5
ESTIMATION METHOD
The most important estimation issue is the possible endogeneity of some explanatory variables in Equation (7.2). MNEs might make a decision on the overseas and domestic operations in terms of employment and outputs simultaneously rather than independently. Therefore, the common factor, which is excluded from the model, could influence either the positive or negative correlation of the OLS regression in the conditional labour demand equation (Desai et al., 2009). In this regard, a generalized method of moments (GMM) instrumental variable (IV) procedure is employed (Griliches and Hausman, 1986; Arellano and Bond, 1991). This procedure essentially applies instrumental variables to the first-differenced data using the moment conditions. It is often shown in the literature that the lagged values of the potentially endogenous variables in level are
126
International fragmentation of production
potentially useful instruments for the time-differenced variables (Griliches and Hausman, 1986; Hasan et al., 2007). Instrument variables for employment and output of foreign affiliates (MNE) in a host country are the lagged employment output and wage rates of a foreign affiliate, the percentage of the manufacturing labour force and the percentage of national income spent on education. The last two exogenous variables are considered to determine the supply side of labour in the host country, and should only affect home labour market outcomes through their impact on the choice of employment in the host country. These variables are taken from the online version of the World Bank Development Indicators for each host country.11 There is also concern about possible correlation between the output variable (Q) of the parent firm and the error term in equation (7.2). The use of time-dummies, industry- and firm-specific fixed effects to some extent alleviates the potential endogeneity problem. However, it is still possible that the output variable (Q) is correlated with some parts of the error term which are not covered by the fixed effects. In this case, the instrument variables (IV) approach is employed to deal with this potential endogeneity problem on domestic output. Instruments include the lagged capital stock, the lagged intermediate inputs and lagged output. There might also be concern about the endogeneity problem of home wages in estimating the conditional labour demand in Equation (7.2). However, the firm-level data are less prone to this problem, because wages are exogenously determined with perfect elastic labour supply (Griliches and Hausman, 1986; Hamermesh, 1993). Both labour supply and demand depend on wages observed. However, when labour supply is perfectly elastic, the position of the labour demand is determined solely by nonlabour factor prices and output or product demand shock (Hamermesh, 1993). Both the within-transformation and first-difference estimators of the fixed-effect model are employed to eliminate the firm-specific effects and the estimation results are compared between the two estimators. The heteroscedasticity-robust standard errors clustering for each firm is used to compute the standard errors. The OLS estimator is also performed to provide a benchmark comparison for results based on the other estimators. The first-difference estimator provides the better treatment for the endogenity problem, which is common to firm-level data, compared with the within-transformation estimator. However, this method may suffer from the potential selectivity bias because it excludes firms not present in both periods t and t–1. It is also known that the first-difference estimator can exacerbate the bias due to measurement errors by reducing the
Overseas operations and home employment of Japanese MNEs
127
amount of systematic variations in the data. Therefore, the first-difference and within-transformation estimators are treated as complementary estimation procedures.
7.6
RESULTS
The regression results for the labour demand equation (7.2) are reported in Table 7.5. In this table Regressions (7.1) and (7.2) report the estimation results based on OLS, and Regressions (7.3) and (7.4) by withintransformation, Regressions (7.5) and (7.6) by first-difference and Regressions (7.7) and (7.8) by the instrument variables (IV) approach. Table 7.6 presents results for each of the four regions – East Asia, North America, the European Union and South America. There is some evidence of a positive complementary relationship between overseas operations (MNE) and home employment, but the magnitude of the estimated coefficient is very small (Table 7.5). Model 3 (within-transformation) suggests that a 10 per cent increase of foreign affiliate employment leads to a 0.18 per cent increase of home employment. MNEQ also indicates a statistically significant positive effect on home employment with a similar magnitude (Model 4). Further, foreign demand shocks, captured by GDP per capita, have no statistical relationship with change in home employment, apart from OLS results. The first-difference estimator (Models 5 and 6) in Table 7.5 also suggests a complementary relationship between overseas operations and home employment. However, the magnitude of the estimated coefficients for MNEQ and MNEL are significantly lower than reported for Models 3 and 4. The IV procedure in Models 7 and 8 improves the results for foreign affiliate employment, but the correction of endogeneity for foreign affiliate sales loses the statistical significance of this variable.12 The OLS result in Model 1 in Table 7.5 indicates a positive complementary relationship between foreign affiliates and home employment and the negative impact of foreign affiliate output on home employment. The evidence also indicates a positive impact of foreign market demand shock (GDPP) on home employment. However, comparing the estimation results between OLS and the alternative fixed-effect models points to the importance of controlling for the firm-fixed effects. The OLS results that did not account for firm-fixed effects largely overestimate the statistical significance of labour demand variables. Table 7.6 presents results for each region, East Asia (Table 7.6a), North America (Table 7.6b), the European Union (Table 7.6c) and South America (Table 7.6d). Even though Japanese MNEs have been actively operating
128
Log Import penetration
Log R&D intensity
Log Output
Log Capital prices
Log Wage rate
(0.014)
−0.266*** (0.048) 1.040*** (0.228) 0.669*** (0.013) 0.151*** (0.013) −0.040***
(0.013)
Reg. 7.2
0.014 (0.011) −0.286*** (0.049) 1.098*** (0.227) 0.692*** (0.014) 0.151*** (0.013) −0.030**
0.059*** (0.011)
OLS
Regression 7.1
Reg. 7.4
(0.006)
−0.116*** (0.019) 0.365*** (0.137) 0.138*** (0.022) 0.022*** (0.005) 0.001
0.018*** (0.005)
(0.006)
0.016*** (0.004) −0.117*** (0.019) 0.375*** (0.137) 0.136*** (0.022) 0.022*** (0.005) 0.001
Within-transformation (WT)
Reg. 7.3
Reg. 7.6
(0.003)
−0.123*** (0.015) 0.081 (0.103) 0.045*** (0.014) 0.009*** (0.003) 0.007**
0.006* (0.003)
(0.003)
0.007*** (0.003) −0.123*** (0.015) 0.087 (0.103) 0.043*** (0.014) 0.009*** (0.003) 0.007**
1st diff.
Reg. 7.5
Dependent variable = log (home employment)
Labour demand by parent firms of MNEs, 1991–2002
COEFFICIENT Log MNE Employment Log MNE Sales
Table 7.5
Reg. 7.8
(0.004)
−0.120*** (0.017) 0.106 (0.134) 0.062 (0.064) 0.009* (0.005) 0.008**
0.022* (0.013)
(0.004)
0.003 (0.008) −0.121*** (0.016) 0.106 (0.136) 0.064 (0.064) 0.009* (0.005) 0.008**
1st diff.-IVS
Reg. 7.7
129 0.503 1294
0.496 1290
0.114 1290
6170 0.296
4.296*** (0.662)
0.007 (0.005)
0.114 1294
6220 0.292
4.187*** (0.654)
−0.002 (0.005)
0.102 1023
4289 0.0917
−0.047 (0.046)
−0.001 (0.003)
0.101 1026
4335 0.0921
−0.055 (0.045)
−0.004 (0.003)
0.102 952
3691 0.0807
−0.085* (0.051)
0.001 (0.004)
0.102 953
3700 0.0876
−0.084* (0.050)
−0.001 (0.005)
Note: Time- and industry-dummy variables (three-digit level) are included for all estimations, but the results are suppressed here. Standard errors based on White’s heteroscedasticity correction clustered by individual firm are given in parentheses, with statistical significance (two-tailed test) denoted as: *** 1%, ** 5% and * 10%. The instrument variables for output, foreign affiliates output and employment used in estimating Models 7 and 8 are discussed in the main text. The overidentifying test statistic for instruments used is 3.69, which does not reject the null hypothesis that all instruments are uncorrelated with the error term at the 5% significant level (c2q 54=9.49).
6220 0.852
−2.548** (1.048)
0.034*** (0.011)
6170 0.855
−2.513** (1.053)
Constant
Observations Adjusted R-squared RMSE No. of parent firms
0.055*** (0.011)
Log GDPP
130
East Asia
Log R&D intensity
Log Output
Log Capital prices
Log Wage rate
Log MNE Sales
(0.013)
−0.267*** (0.049) 1.081*** (0.247) 0.714*** (0.011) 0.170***
(0.013)
(0.006)
−0.122*** (0.021) 0.302** (0.147) 0.122*** (0.025) 0.028***
(0.004)
(0.008) −0.020** (0.009) −0.275*** (0.049) 1.050*** (0.244) 0.727*** (0.012) 0.167***
Reg. 7.4
(0.006)
0.006** (0.003) −0.123*** (0.021) 0.297** (0.147) 0.121*** (0.025) 0.028***
Within-transformation (WT)
Reg. 7.3
0.008**
Reg. 7.2
0.007
OLS
Regression 7.1
(0.004)
−0.128*** (0.017) 0.042 (0.100) 0.037** (0.015) 0.011***
(0.002)
0.002
Reg. 7.6
(0.004)
0.003 (0.002) −0.128*** (0.017) 0.044 (0.100) 0.037** (0.015) 0.011***
1st diff.
Reg. 7.5
Dependent variable = log (home employment)
Labour demand by parent firms of MNEs by region, 1991–2002
COEFFICIENT Log MNE Employment
(a)
Table 7.6
Reg. 7.8
(0.006)
−0.126*** (0.020) 0.131 (0.146) 0.114 (0.070) 0.016***
(0.006)
0.012*
(0.006)
0.002 (0.005) −0.127*** (0.020) 0.123 (0.145) 0.113 (0.070) 0.016**
1st diff.-IVS
Reg. 7.7
131
Observations Adjusted R-squared RMSE No. of parent firms
Constant
Log GDPP
Log Import penetration
4986 0.875 0.475 1061
4947 0.874
0.475 1058
−1.943* (1.153)
(0.015) −0.007 (0.011)
(0.015) −0.027*** (0.009)
−1.902 (1.161)
−0.027*
Reg. 7.2
−0.032**
OLS
Regression 7.1
Reg. 7.4
0.109 1058
4947 0.324
5.594*** (0.731)
(0.006) −0.004 (0.004)
−0.000
0.109 1061
4986 0.320
5.633*** (0.731)
(0.006) −0.006 (0.005)
0.000
Within-transformation (WT)
Reg. 7.3
0.0994 829
3426 0.0986
−0.004 (0.009)
(0.003) −0.001 (0.003)
0.002
0.0991 834
3464 0.100
−0.005 (0.010)
(0.003) −0.002 (0.003)
0.002
Reg. 7.6
1st diff.
Reg. 7.5
Dependent variable = log (home employment) Reg. 7.8
0.102 767
2898 0.0730
−0.012 (0.009)
(0.004) −0.004 (0.005)
0.005
0.101 768
2907 0.0775
−0.020* (0.010)
(0.004) 0.000 (0.005)
0.005
1st diff.-IVS
Reg. 7.7
132
Log R&D intensity
Log Output
OLS
−0.286*** (0.063) 0.486 (0.309) 0.664*** (0.017) 0.154*** (0.020)
0.068*** (0.012)
Model 1
North America
Log Capital prices
Log Wage rate
COEFFICIENT Log MNE Employment Log MNE Sales
Table 7.6 (b)
0.031** (0.015) −0.313*** (0.064) 0.729** (0.301) 0.685*** (0.019) 0.158*** (0.020)
Model 2
Model 4
−0.119*** (0.022) 0.185 (0.173) 0.105*** (0.026) 0.016** (0.007)
0.014* (0.008) 0.023*** (0.008) −0.121*** (0.021) 0.184 (0.174) 0.101*** (0.026) 0.015** (0.007)
Within-transformation (WT)
Model 3
Model 6
−0.120*** (0.018) 0.079 (0.170) 0.032** (0.015) 0.004 (0.003)
0.005 (0.005) 0.006 (0.004) −0.120*** (0.017) 0.075 (0.171) 0.031** (0.015) 0.004 (0.003)
1st diff.
Model 5
Dependent variable = log (home employment) Model 8
−0.104*** (0.022) 0.043 (0.213) 0.049 (0.071) 0.004 (0.006)
−0.003 (0.023) 0.003 (0.015) −0.102*** (0.021) 0.068 (0.214) 0.068 (0.067) 0.005 (0.005)
1st diff.-IVS
Model 7
133
Observations Adjusted R-squared RMSE No. of parent firms
Constant
Log GDPP
Log Import penetration
4049 0.837 0.511 815
3996 0.841
0.503 812
−0.128 (1.396)
(0.018) −0.031 (0.031)
(0.017) −0.090*** (0.025)
1.480 (1.421)
−0.015
OLS
Model 2
−0.022
Model 1
Model 4
0.108 812
3996 0.247
5.953*** (0.837)
(0.007) −0.018 (0.013)
0.001
0.108 815
4049 0.252
6.019*** (0.840)
(0.007) −0.037** (0.017)
0.002
Within-transformation (WT)
Model 3
0.0943 662
2785 0.0836
−0.003 (0.009)
(0.004) −0.010 (0.007)
0.002
0.0947 665
2840 0.0816
−0.004 (0.009)
(0.004) −0.014 (0.009)
0.002
Model 6
1st diff.
Model 5
Dependent variable = log (home employment) Model 8
0.0934 589
2198 0.0651
−0.012 (0.015)
(0.005) −0.000 (0.030)
0.003
0.0937 590
2203 0.0584
−0.016 (0.013)
(0.005) −0.008 (0.029)
0.004
1st diff.-IVS
Model 7
134
Log R&D intensity
Log Output
Log Capital prices
Log Wage rate
COEFFICIENT Log MNE Employment Log MNE Sales
Table 7.6 (c)
OLS
(0.357) 0.689*** (0.026) 0.212*** (0.024)
−0.206*** (0.062) 1.199***
(0.347) 0.672*** (0.018) 0.208***
(0.023)
Model 2
0.023 (0.023) −0.219*** (0.064) 1.428***
0.059*** (0.013)
Model 1
European Union
Model 4
(0.008)
(0.215) 0.097*** (0.033) 0.012
−0.128*** (0.027) 0.142
0.011 (0.007)
(0.007)
(0.213) 0.100*** (0.032) 0.012
0.030** (0.015) −0.117*** (0.027) 0.238
Within-transformation (WT)
Model 3
Model 6
(0.005)
(0.192) 0.023 (0.021) 0.004
−0.125*** (0.022) 0.067
0.005 (0.004)
(0.005)
(0.185) 0.014 (0.020) 0.007
0.021** (0.009) −0.130*** (0.022) 0.063
1st diff.
Model 5
Dependent variable = log (home employment) Model 8
(0.007)
(0.264) 0.030 (0.087) 0.003
−0.104*** (0.023) 0.068
0.031 (0.019)
(0.007)
(0.260) −0.002 (0.087) 0.004
0.006 (0.028) −0.104*** (0.023) −0.037
1st diff.-IVS
Model 7
135
Observations Adjusted R-squared RMSE No. of parent firms
Constant
Log GDPP
Log Import penetration
2473 0.857 0.475 495
2432 0.862
0.466 493
−2.830* (1.654)
(0.022) −0.041 (0.039)
(0.020) −0.087*** (0.024)
−1.358 (1.613)
−0.019
OLS
Model 2
−0.021
Model 1
Model 4
0.106 493
2432 0.277
6.376*** (1.117)
(0.008) −0.018 (0.012)
0.005
0.106 495
2473 0.285
6.279*** (1.099)
(0.008) −0.061** (0.029)
0.006
Within-transformation (WT)
Model 3
0.0956 399
1715 0.0814
0.003 (0.012)
(0.005) −0.013* (0.008)
0.006
0.0978 400
1761 0.0883
−0.060 (0.065)
(0.005) −0.045*** (0.017)
0.007
Model 6
1st diff.
Model 5
Dependent variable = log (home employment) Model 8
0.0946 342
1271 0.0412
0.002 (0.008)
(0.007) −0.040* (0.023)
0.010
0.0948 345
1285 0.0660
0.002 (0.009)
(0.007) −0.013 (0.052)
0.012
1st diff.-IVS
Model 7
136
Log R&D intensity
Log Output
Log Capital prices
Log Wage rate
Log MNE Sales
COEFFICIENT Log MNE Employment
Model 2
0.060 (0.041) −0.488*** (0.133) 0.739 (0.715) 0.731*** (0.043) 0.154*** (0.031)
OLS
−0.470*** (0.130) 0.744 (0.719) 0.761*** (0.030) 0.157*** (0.033)
0.040 (0.032)
Model 1
Table 7.6 (d) South America
Model 4
−0.241*** (0.066) 0.821** (0.347) 0.223*** (0.079) 0.003 (0.015)
0.021* (0.013) 0.050*** (0.019) −0.244*** (0.066) 0.764** (0.349) 0.196*** (0.075) 0.003 (0.015)
Within-transformation (WT)
Model 3
−0.252*** (0.072) 0.226 (0.261) 0.071 (0.074) −0.000 (0.007)
0.010 (0.008)
Model 6
0.010 (0.009) −0.253*** (0.069) 0.223 (0.250) 0.057 (0.072) −0.001 (0.007)
1st diff.
Model 5
Dependent variable = log (home employment) Model 8
−0.174* (0.102) −0.233 (0.495) 0.191 (0.213) −0.001 (0.009)
−0.003 (0.030) 0.018 (0.027) −0.177* (0.102) −0.177 (0.462) 0.178 (0.194) −0.001 (0.009)
1st diff.-IVS
Model 7
137
(0.043) −0.110 (0.076)
(0.043) −0.071 (0.054)
0.465 156
764 0.882
0.467 154
Observations Adjusted R-squared RMSE No. of parent firms
Model 4
0.108 154
764 0.420
2.392 (1.782)
(0.013) −0.001 (0.018)
0.039***
0.107 156
780 0.426
2.820 (1.808)
(0.013) −0.064* (0.034)
0.039***
Within-transformation (WT)
Model 3
0.0923 129
546 0.227
0.031*** (0.012)
(0.010) −0.001 (0.013)
0.003
Model 6
0.0929 131
563 0.217
0.031*** (0.012)
(0.009) −0.008 (0.017)
0.003
1st diff.
Model 5
Model 8
0.0904 96
320 0.267
0.101 (0.130)
(0.012) 0.013 (0.044)
0.012
0.0899 97
321 0.273
−0.039 (0.040)
(0.011) −0.023 (0.050)
0.010
1st diff.-IVS
Model 7
Note: Time- and industry-dummy variables (three-digit level) are included for all estimations, but the results are suppressed here. Standard errors based on White’s heteroscedasticity correction clustered by individual firm are given in parentheses, with statistical significance (two-tailed test) denoted as: *** 1%, ** 5% * 10%. The instrumental variables for output, foreign affiliates output and employment used in estimating Model 4 are discussed in the main text.
780 0.882
−1.179 (3.291)
−0.722 (3.337)
−0.019
−0.015
OLS
Model 2
Constant
Log GDPP
Log Import penetration
Model 1
Dependent variable = log (home employment)
138
International fragmentation of production
in East Asia since the mid-1980s, their expansion in terms of employment and sales do not seem to negatively affect the level of home employment. In fact, foreign operations in East Asia seem to have little impacts on home employment.13 In North America foreign affiliates employment and sales have a positive impact (Models 3 and 4 in Table 7.6b). However, the findings are sensitive to the estimation method. Similar inferences can be made for the European Union (Table 7.6c). Overall, there is no clear-cut evidence of ‘exporting jobs’ by Japanese MNEs, despite the concerns expressed in the public debates. In fact, there is some weak evidence to suggest that expanded overseas operations may have actually helped to maintain the level of home employment. Other determinants of labour demand by parent firms can be summarized as follows. Wage elasticity of labour demand consistently has the expected negative sign, indicating a downward sloping of labour demand. The own-wage elasticity is consistently reported in the range of –0.1 to –0.2. The output elasticity is statistically significant both in the withintransformation and the first-difference estimators (Model 3–6). However, this result changes once corrected for the endogeneity problem in Models 7 and 8. The estimated coefficient of r (the user cost of capital) shows mixed results, making it impossible to infer whether capital and home employment are substitutes for or complementary to each other. Interestingly, there is a strong effect of R&D intensity of foreign affiliates in East Asia on home employment. Similar results are obtained for North America (Table 7.6b), but the results are sensitive to the estimation method used.
7.7
CONCLUDING REMARKS
This chapter has examined the hypothesis that expansion of overseas operations of Japanese manufacturing MNEs reduces home employment within the MNEs’ operations. A standard labour demand equation is estimated by allowing the effects of foreign affiliate employment and sales on home employment. In addition, geographic locations of foreign affiliates is controlled in order to control for the specific regional characteristics of MNEs. The empirical exercise is based on the newly constructed panel data set, covering information for both home and foreign affiliates’ operations within matched manufacturing Japanese MNEs for the period 1991–2002. Despite concerns expressed about the adverse effects of FDI, the evidence does not support the view that overseas operations expand at the cost of home employment in Japan. On the contrary, the findings suggest
Overseas operations and home employment of Japanese MNEs
139
that overseas operations have somewhat helped to maintain the level of home employment in Japanese manufacturing. However, the results are sensitive to the estimation method used and whether the estimation based on the panel data set is balanced or unbalanced.
APPENDIX 7.1
CONSTRUCTION OF THE MATCHED PANEL DATA14
The data set was constructed by matching the information on parent firms extracted from the Basic Survey of Business Structure and Activity (Kigyo Katsudou Kihou Chosa in Japanese) and the information on their corresponding foreign affiliates from the Basic Survey of Overseas Japanese Business Activity (Kaigai Gigyou Katsudou Kihon Chosa in Japanese). Both surveys are conducted by the Ministry of Economy, Trade and Industry (METI) (Appendix A7.2 discusses each survey in detail). For brevity, the former will henceforth be called the ‘METI Firm survey’ and the latter the ‘METI foreign affiliates survey’ in this chapter. The starting point of the panel data is 1991, when the first METI firm survey was conducted. The second survey was undertaken in 1994, and it has been continued since then. The most recent data for both METI surveys available for this project are 2002 (note that 1992 and 1993 are missing, since the METI firm survey was not conducted in these two years). The matched panel data set only includes parent firms that have both more than 50 employees and capital of more than 30 million yen. The industry classification is available at the two-digit level of the Japan Standard Industrial classification (JSIC). The matched panel set is unbalanced due to the ‘entry/exit’ of parent firms. Creating the panel data using these two METI surveys involved the following steps. First, information from both surveys was restricted to manufacturing industry by excluding non-manufacturing industry data. This necessarily removed information on any foreign affiliates whose industry classification is not manufacturing. Indeed, it is possible that this process somewhat underestimates the overseas operations of Japanese MNEs, since some parent manufacturing firms set up foreign affiliates in non-manufacturing industries. However, such downward bias is considered to be minimal. After limiting the data to the manufacturing sector, the consistent three-digit level of the manufacturing industry classification was assigned to each parent firm. This is important because there were some changes in the three-digit manufacturing industry classification over the time period 1991–2002. Second, merging the two surveys was achieved by linking the permanent
140
International fragmentation of production
identifier assigned to each individual parent firm of the METI firm survey to the same code reported by each individual foreign affiliate from the METI foreign affiliate survey. To ensure successful matching, a careful cross-checking was implemented by examining the name, address and ownership structure. As a result, this procedure systematically combined information on the overseas operations of Japanese MNEs with domestic economic activity, based on the unique parent firm code. Third, following Hanson et al. (2003) and Harrison and McMillan (2006), sales weighted averages of foreign affiliate variables were constructed (see Section 7.4.1 for the construction of foreign affiliate variables).15 This is essential to make the panel data estimation operational, because Japanese parent firms often own more than one foreign affiliate operating in multiple locations. For example, Toyota has foreign affiliates in Thailand as well as in England. Some parent firms in the METI firm survey had reported abnormally large or small values (Nishimura et al., 2005). Any parent firms were dropped if at least one of the values of employment, sales, industry classification and identification code was missing. As a consequence, about 1 per cent of the data was excluded from the main data set. The pooled data was also disaggregated into four regions of host countries, East Asia, North America, the European Union and South America, and the labour demand is estimated for each region (again for the balanced and unbalanced panel data). The main motivation of the regional disaggregation was to control for the level of the host country’s development, the geographic proximity to Japan and the possible characteristics of foreign affiliate production. Foreign affiliates of Japanese MNEs operating in developing countries (East Asia and South America) are most likely to be the vertical type of MNEs, whereas those in developed countries (North America and the European Union) are most likely motivated by horizontal MNEs. Therefore, the postulated employment relationship between home and abroad critically depends on the location of foreign affiliates (Brainard and Riker, 1997a; Harrison and McMillan, 2006). In addition to these considerations, the firm-level data were aggregated up to the three-digit industry level to collaborate with the firm-level findings. While the firm-level investigation brings about more advantages, the industry level also has several merits. First, the industry-level data capture not only the within-firm employment changes, but also the across-firm employment changes. It is possible that some parent firms reduce home employment whereas other parent firms expand foreign affiliate employment. The firm-level data do not track the across-firm employment substitutions (Harrison et al., 2007). Second, the estimated labour demand
Overseas operations and home employment of Japanese MNEs
141
at industry level is likely to reflect changes in employment resulting from the entry and exit of firms to the industry (Roberts and Skoufias, 1997). Third, the industry-level aggregation mitigates the potential problem of small employment changes that are relatively symmetrically distributed around the origins (see Figure 7.1). The industry-level data only show the net effects of employment changes between home and foreign affiliate employment. One limitation of the matched panel set is that some parent firms disappear in one year in the data coverage and reappear in another presumably because of varying sample restrictions imposed. This means that the entry and exit of firms in this survey do not necessarily correspond to the standard definitions of origin and termination of firms (Nishimura et al., 2005). The matched panel data set also excludes small-scale Japanese firms, which do not meet the sample selection criteria of the METI firm survey even if they do have foreign affiliates. However, their omission does not affect the capture of the overall trends of MNEs’ operations.
APPENDIX 7.2
METI SURVEYS
METI Firm Survey (the Basic Survey of Business Structure and Activity) This survey, first conducted in 1991, has become an annual survey since 1994. It covers all firms in both manufacturing and non-manufacturing including mining, wholesale, agriculture, retail, and construction as well as the service sector that have both more than 50 employees and capital of more than 30 million yen. It collects sufficient information to quantify details on the domestic operations of Japanese firms, including total sales, total purchases, employment, workers’ compensation, fixed tangible and non-tangible assets, capital, number of establishments, R&D expenditure, year of establishment, exports and imports. Most key variables have been reported continuously since 1991 except for the years in 1992 and 1993. This survey also covers limited information about the operation of foreign affiliates such as the number of affiliates, employment and value of sales, if the parent firm engages in FDI. Transactions are recorded in millions of Japanese yen and measure the amounts paid or received by individual firms. All individual firms are assigned unique identifiers, making it possible to track operations of the same firms over time. The survey is mandatory16 and hence the response ratio is very high (around 90 per cent). The well-known limitation of the METI firm survey is that the entry and exit of firms in this survey do not necessarily correspond to the standard
142
International fragmentation of production
definitions of origin and termination of firms due to the sample selection criteria (Nishimura et al., 2005). METI Foreign Affiliates Survey (the Basic Survey of Overseas Japanese Business Activity) The METI foreign affiliates survey is designed to trace the scale and functions of foreign affiliates of Japanese MNEs operating overseas. The survey is sent out to their parent firms located in Japan. There has been a relatively long history of conducting this survey which commenced in 1971, a detailed survey every three years since 1981 and a standard one each year in other years. However, data are available by electronic means for this project only since 1989. Most importantly, each individual affiliate is assigned its own unique code as well as the parent firm identifier. This makes it possible to link between the METI firm survey and the METI foreign affiliates survey. The METI foreign affiliates survey contains the main variables, for example, sales output distinguished by destinations such as local market, Japan or other countries, total purchase distinguished by sources, wages and salaries, employment, fixed tangible assets, capital and R&D spending. However, not all have been reported consistently since 1989. For instance, wage and salaries only appear continuously from 1994, and fixed tangible assets are only available for the years 1989, 1992, 1995, 1998 and 2001. The METI foreign affiliates survey also reports limited information about the operations of parent firms, such as sales, purchases, employment and capital. While the METI foreign affiliates survey has been a very useful and valuable data source for evaluating the overseas operations of Japanese MNEs, its quality has been questioned from time to time (Ramstetter, 1996).17 These problems can be summarized as follows. Unlike the METI firm survey, responding to this survey is not a mandatory requirement. This yields a wide fluctuation in sample coverage from year to year (Ramstetter, 1996). The response rate varied from 33 per cent in 1980 to 51 per cent during 1983–92, but has increased somewhat in more recent years. In 2005 the questionnaire was sent to 4564 Japanese firms, 3176 completed the questionnaire and the corresponding return rate accounts for 69.6 per cent. Information on foreign affiliates operating in developing host countries is far less satisfactory than for those operating in developed host countries. There is also a wide variation in the reported coverage of variables from year to year, making it difficult to track the same variable over time (Matsuura, 2004).18 However, the key variables, including sales,
Overseas operations and home employment of Japanese MNEs
143
employment and the year when foreign affiliates were established are available for each year. Other items, such as intermediate inputs expenditure and capital stock, have not been reported on a consistent basis. In addition, the fluctuation in the survey response rate also significantly influences the stability of variables over time (Matsuura, 2004). Some key variables, such as sales and employment, are found to follow a smooth time-series pattern, while variables such as worker compensation and R&D expenditure behave less consistently over time. These data quality problems are highlighted in Table 7A.1 where a comparison is made between the adjusted foreign affiliate data provided by the Research Institute of Economy and Trade (RIETI) FDI database,19 and the data tabulation extracted from the original returns from the METI foreign affiliates survey used in this chapter. There have been some recent efforts to improve the quality of the METI foreign affiliates survey by estimating and supplementing the missing raw data for the number of foreign affiliates, sales outputs and number of employees by three-digit level industry and country/region (see Matsuura (2004) for a detailed explanation of the adjustment method). The first three columns in Table 7A.1 present the aggregated data based on the original (unadjusted) METI foreign affiliates survey whereas the following three columns display the adjusted data by the RIETI. The last three columns present a ratio between two. A few key inferences can be made by comparing the adjusted and unadjusted METI foreign affiliates data. They reveal wide discrepancies, especially for firm numbers and employment prior to 1997, but after that the figures from the unadjusted data are quite close to those reported in the adjusted data. In addition, the number of foreign affiliates, their employment and sales output in the unadjusted data all appear to have grown, although they indicate sudden drops and increases in some years. For example, the employment figure dropped from over 1.2 million workers in 1991 to 986 000 in 1992. The following year (1993) the same employment figure rose to over 1.5 million workers. Similarly, sales output was recorded at 29 trillion yen in 1993, but sharply increased to 85 trillion yen in 1994. There was also a large dip in reported sales from 2000 to 2001. Overall the better coverage of the unadjusted METI data over time appears to have increased the quality. However, despite this improvement, the figures for sales output after 1997 indicated unexplained fluctuations. It might be preferable to use the adjusted data series of the METI data to remove any biases introduced by the low quality of the survey, particularly for earlier years. Unfortunately, the adjusted METI data are only available for the listed three main variables at the aggregated industry level.
144
1989 1990 1991 1992 1993 1994 1995 1996 1997
Year
Table 7A.1
2656 3407 3535 3040 4548 7992 7345 7626 9279
(unit)
Employment of:
Sales of:
914 1242 1261 986 1516 1972 1986 2258 2540
(1000) 22 26 25 25 29 85 87 107 111
(trillion yen) 6197 6805 7147 7335 7756 8346 9042 9423 9605
(unit) 1533 1820 1938 2037 2233 2427 2663 2926 3000
(1000) 36 38 39 40 42 46 52 61 66
(trillion yen)
Number Employment Sales of: of: of:
METI Adjusted Data
foreign affiliates of Japanese manufacturing MNEs
Number of:
METI Unadjusted (Raw) Data
0.4 0.5 0.5 0.4 0.6 1.0 0.8 0.8 1.0
0.6 0.7 0.7 0.5 0.7 0.8 0.7 0.8 0.8
Number Employment
0.6 0.7 0.6 0.6 0.7 1.8 1.7 1.8 1.7
Sales
Ratio (Unadjusted/Adjusted)
Comparison between METI survey on foreign affiliates for unadjusted and adjusted data in manufacturing industry
145
9069 9828 10549 7068 8006 5350 101.4
2339 2812 3049 2645 2844 1930 211.2
110 103 113 64 65 43 189.7
9470 9271 9091 8793 8265 1118 15.6
3052 3204 3256 3216 3286 1348 69.5
67 66 69 72 74 35 91.3
1.0 1.1 1.2 0.8 1.0
0.8 0.9 0.9 0.8 0.9
1.6 1.6 1.6 0.9 0.9
Source: Based on the METI database, which is explained in Appendix 7.1 and the Research Institute of Economy, Trade and Industry (RIETI) FDI database, available at http://www.rieti.go.jp/en/database/d08.html.
1998 1999 2000 2001 2002 Level difference % Change
146
International fragmentation of production
NOTES 1.
2. 3. 4. 5. 6. 7. 8.
9. 10. 11. 12.
13. 14. 15.
16. 17. 18. 19.
This could be partly due to the stringent Japanese government policy on access to the original returns of firm-level information, which has been eased to some extent in recent years. However, access to the original METI survey database is still limited to protect private firm information. I am grateful to Professor Kyoji Fukao from Hitotsubashi University for making this METI database available to this project. Complex integration is another type of MNE (UNCTAD, 1998, 2002; Yeaple, 2003). This type shares certain features of both the vertical and horizontal type of MNE. The survey began in 1929, but its scope was limited to one question – the value of foreign commercial assets controlled by US companies (see Mataloni (1995) for more details). In our data set industry classification of parent firms changes over time. To capture this regressions include industry-specific dummy variables. However, the main results are resilient to exclusion of industry dummy variables. In the experimental stage an alternative weighting scheme was attempted using the employment share, but the results were similar. Therefore, the results reported below are based on the sales share of foreign affiliates. GDP per capita is taken from the World Bank Development Indicators. The METI Firm survey only collects information on the number of workers, not on hours worked. While fluctuations in hours per worker are crucial for understanding short-run labour demand, in the long run variation in the number of workers is the primary adjustment method (Hamermesh, 1993). Therefore, a focus on employment, rather than hours worked, is consistent with the objective of explaining long-run labour demand differences at the firm level. Available at http://www.boj.or.jp/type/stat/dlong/price/cgpi/index.htm. They are available for the following industries: textile products, iron and steel, nonferrous metals, metal products, general machinery, electrical machinery, transport equipment, precision instruments and other manufacturing industry products. Available at http://devdata.worldbank.org/dataonline/. The overidentifying test statistic for instruments amount to 3.69, which does not reject the null hypothesis that all instruments are uncorrelated with the error term at the 5 per cent significant level (c2q 54 = 9.49). In other words, the selected instruments are valid instruments with no direct correlation with the error term in equation (7.2). The first stage regression also finds a strong correlation between the selected instruments and the endogenous variables (the results are suppressed for brevity). However, the increased international production in East Asian countries has changed the skill composition of home employment in Japanese manufacturing (Head and Ries, 2002; Yamashita, Chapter 6 this volume). During work on this data set I have extensively referred to Matsuura and Kiytoa (2004) and the resources available from the RIETI website, available at http://www.rieti.go.jp/ jp/database/d02.html#01. In principle, it would be possible to include variables for each host country where foreign affiliates potentially operate without aggregating foreign affiliate variables. However, this creates the problem of repeating the same information for the corresponding parent firms, making it difficult to interpret the estimated results (Brainard and Riker, 1997a). It would be particularly daunting to repeat the same home employment in the dependant variable. This means firms failing to return the survey to the METI face heavy fines. An alternative data source is available from a private publishing company, Toyo Keizai. However, Ramstetter (1996) concluded that the data from Toyo Keizai have even more serious coverage and quality problems for the sample collection and variables. See also http://www.rieti.go.jp/jp/database/d02.html#01. Available at http://www.rieti.go.jp/jp/database/d08.html.
8
Conclusion
In the 1970s and 1980s production fragmentation – the cross-border splitting of the production process within vertically integrated manufacturing industries – was happening on a small scale, involving a limited number of developed and developing countries. Since then, transfers of manufacturing processes from developed into developing countries have been growing at a remarkable speed. This development was mainly due to the further reduction of trade and transportation costs, service-link costs, investment liberalization and technological advancements (see Chapter 2). In line with this global trend, international fragmentation of production has contributed to the further acceleration of the globalization process of Japanese manufacturing. The purpose of this book has been to examine the patterns, determinants and implications for host country trade and labour market performance in light of the Japanese experience, placed within a comparative perspective, drawing particularly on the US experience. The book began with a comprehensive interpretative survey of the theory of production fragmentation in order to place the empirical analysis in context. The empirical analysis was carried out in three stages. First, patterns and determinants of fragmentation trade by a comparative analysis of Japan and the USA were examined by using trade data compiled on the basis of the new commodity list of parts and components of manufacturing trade over the period 1988–2005. The focus of the next stage of analysis was on effects of fragmentation trade on the skills structure of manufacturing employment in Japan. This analysis was based on a panel data set covering 52 Japanese manufacturing industries over the period 1980–2000. The two key innovations of the analytical methodology were the use of newly constructed measure of the intensity of fragmentation trade and allowing for the possible differential effects of the geographical pattern of fragmentation trade on the skill composition of manufacturing employment. At the third stage, a firm-level analysis of the implications of production fragmentation for employment was undertaken using a new panel data set compiled from the unpublished returns from two firm-level surveys: the Basic Survey of Business Structure and Activity and the Basic Survey of Overseas Japanese Business Activity, both conducted by the Japanese Ministry of Economy, Trade and Industry (METI).
147
148
8.1
International fragmentation of production
FINDINGS
Since the mid-1980s fragmentation trade as well as Japanese MNE operations have expanded dramatically. At first, assembly operations were located mostly in lower-wage countries in Asian NIEs, but more recently these have been shifting to China at a rapid pace. The general trend of fragmentation trade follows this geographic profile. However, there have been some notable differences across products suggesting the importance of factors other than wage costs and proximity. For instance, the bulk of automotive components (motor car engines, engines and gear boxes) are still shipped from Japan to North America and the EU. The gravity model estimated in Chapter 4 confirmed the importance of labour costs and geographical distance as determinants of fragmentation trade. These were stronger factors in the case of Japan compared to the USA. This is probably closely related to comparative cost considerations driving Japanese MNEs and to Japan’s production management strategy, ‘just-in-time’ logistics and the importance of physical proximity for quality maintenance. On the other hand, some determinants such as the level of governance quality and the level of infrastructure were equally important for both Japanese and US trade in parts and components. Chapter 5 examined key aspects of labour market adjustments in both Japanese and US manufacturing using the standard indicators. The analysis identified a shift in labour demand in favour of skilled workers (skill upgrading) in both countries over the last two decades, confirming findings of other studies. However, the actual form of skill upgrading was different between Japan and the USA: skill upgrading in Japanese manufacturing took the form of an increase in the relative employment of skilled workers with stability in relative wages in contrast to the USA case where the wage gap between skilled and unskilled workers has been accelerating in the USA since the mid-1980s. Two key aspects were investigated in relation to the role of production fragmentation: the skill composition of employment and the impacts of overseas operations on home employment of MNEs. These aspects were further explored in Chapters 6 and 7. Previous studies of Japanese manufacturing industries had failed to find a robust relationship between the intensity of fragmentation trade and changes in skill composition. However, our analysis, reported in Chapter 6, showed that the expansion of fragmentation trade with developing East Asian countries has had a significant impact on the skills composition of Japanese manufacturing employment. This impact varied with the factor endowment profiles of trading partners. In contrast to the skill upgrading effect of fragmentation trade with developing countries, fragmentation trade with OECD countries had skill downgrading effects. The analysis of
Conclusion
149
the USA manufacturing data also obtained similar results to this Japanese experience using a new measure (Chapter 6). All in all, the findings in Chapter 6 imply that increased intensity of fragmentation trade with developing countries would shift up the labour demand for skilled workers in domestic manufacturing whereas fragmentation trade with developed countries tends to increase demand for unskilled workers. Chapter 7 focused on the effects of the expanded overseas operations of Japanese manufacturing MNEs driven by the process of production fragmentation on home (domestic) employment in manufacturing over the period 1991–2002. Despite earlier concerns and the heated policy debates in the late 1980s and the early 1990s, Chapter 7 found no concrete evidence to suggest that increased employment in foreign MNE affiliates substitutes for employment of Japanese manufacturing MNEs at home. This inference was robust even after controlling for the geographical locations of foreign affiliates. There is in fact some weak evidence that overseas operations may have even contributed to maintaining the level of home employment.
8.2
POLICY IMPLICATIONS
Production fragmentation is likely to expand over time, shaping trade patterns and the industrial structure of the global economy. Hence, the findings reported in this book have policy implications for both Japan and other developed countries, as well as developing countries. Production fragmentation is a means to transforming and upgrading the industrial structure by adoption of new, improved production methods and relocating obsolete production processes that erode international competitiveness. However, there is a perception in developed economies that there is an efficiency-employment trade-off involved in this process, driven by concerns about a negative impact on domestic employment. The findings of this book challenge the popular perception that production fragmentation has been growing at the cost of domestic manufacturing jobs in Japan. On the contrary, this process helps to strengthen the domestic manufacturing base of the country by facilitating the ability of domestic firms to effectively and profitably compete in the international arena. In general, manufacturing firms that are now more profitable because they have been able to take advantage of international fragmentation of production employ as many or more workers than before even in home operations. While the skill upgrading in manufacturing can exert some downward pressure on wages and employment of unskilled workers, the demand
150
International fragmentation of production
effects of higher profits of MNE firms can have an offsetting effect. In any case, the appropriate policy response is clearly not trade protection which, as both theory and history demonstrate, would adversely affect outcomes. The challenge for government is to create appropriate incentives and policy schemes to help unskilled workers to respond to the new employment opportunities being created in skilled manufacturing occupations by providing appropriate adjustment assistance to enhance their skills. This also raises important policy questions: are market-based incentives powerful enough to encourage the needed changes in skills formation? How should government support this process? Answers to these questions are critical if policy makers in developed countries are to assist in maximizing efficiency gains from production fragmentation and better manage the process of globalization. Production fragmentation also opens up new opportunities for developing and emerging countries to participate in a new form of international vertical specialization by specializing in the different stages (tasks) of production processes depending on their stages of development and comparative cost advantages in international production. Thus, the process of industrialization can start from manufacturing simple components with labour-intensive assembly of components. Over time, these countries can develop their own competency in the manufacturing of complex components and gradually move into more capital-intensive activities. As the economy proceeds towards further industrial development, specialized industrial skills will be developed and consolidated in order to build up the basic supporting industries and industry clusters. However, the process of industrialization through production fragmentation is not automatic. As discussed in Chapter 4, lower labour costs alone are not attractive enough for a country to become an assembly location from the investors’ point of view. Governments in developing countries have a critical role to play in establishing an appropriate investment environment by upgrading the physical infrastructure – roads, ports, water and electricity supply, and telecommunications – by developing non-physical infrastructures such as R&D expertise and human resource development, and by establishing sound institutions, good governance and policy stability.
References Ahn, S., K. Fukao and K. Ito (2008), ‘Outsourcing in East Asia and its impact of the Japanese and South Korean labour markets’, paper presented at the Ninth Global Development Network Conference Research Workshop, Emerging Trends and Patterns of Trade and Investment in Asia, February Brisbane, Australia. Amano, T. (2008), ‘Competitive strategies of Japanese/U.S. multinationals in global production networks and clusters: the case of HDD industry’, paper presented at the Ninth Global Development Network Conference Research Workshop, Emerging Trends and Patterns of Trade and Investment in Asia, February Brisbane, Australia. Amiti, M. and S.-J. Wei (2005), ‘Fear of service outsourcing: is it justified?’, Economic Policy, 20(42), 308–47. Anderson, J.E. and D. Marcouiller (2002), ‘Insecurity and the pattern of trade: an empirical investigation’, Review of Economics and Statistics, 84(2), 342–52. Anderson, J.E. and E. van Wincoop (2004), ‘Trade costs’, Journal of Economic Literature, 42(3), 691–751. Anderton, B. and P. Brenton (1999), ‘Outsourcing and low-skilled workers in the UK’, Bulletin of Economic Research, 51(4), 267–85. Ando, M. and F. Kimura (2005), ‘The formation of international production and distribution networks in East Asia’, in. T. Ito and A. Rose (eds), International Trade, NBER-East Asia Seminar on Economics, no. 14, Chicago, IL: University of Chicago Press, 177–213. Antràs, P. (2003), ‘Firms, contract and trade structure’, Quarterly Journal of Economics, 118(4), 1375–418. Antràs, P. and E. Helpman (2006), ‘Global outsourcing’, Journal of Political Economy, 112(3), 552–80. Arellano, M. and S. Bond (1991), ‘Some test of specification for panel data: Monte Carlo evidence and an application to employment equation’, Review of Economic Studies, 58(2), 277–97. Armstrong, S. (2007), ‘Measuring trade and trade potential: a survey’, Australia–Japan Research Centre (AJRC) Asia Pacific economic papers, no. 368, Canberra, The Australian National University. Arndt, S.W. (1997), ‘Globalization and the open economy’, North American Journal of Economics and Finance, 8(1), 71–9. 151
152
International fragmentation of production
Asanuma, B. (1989), ‘Manufacturer–supplier relationships in Japan and the concept of relation-specific skill’, Journal of the Japanese and International Economies, 3(1), 1–30. Athukorala, P. (2005), ‘Product fragmentation and trade patterns in East Asia’, Asian Economic Papers, 4(3), 1–27. Athukorala, P. and C. Manning (1999), Structural Change and International Migration in East Asia: Adjusting to Labour Scarcity, Oxford and New York: Oxford University Press. Athukorala, P. and N. Yamashita (2006), ‘Production fragmentation and trade integration in a global context’, North American Journal of Economics and Finance, 17(4), 233–56. Athukorala, P. and N. Yamashita (2008), ‘Patterns and determinants of production fragmentation in world manufacturing trade’, in. F. Mauro, W. McKibbin and S. Dees (eds), Globalisation, Regionalism and Economic Interdependence, Cambridge, MA: Cambridge University Press, pp. 45–72. Autor, D.H., L.F. Katz and A.B. Krueger (1998), ‘Computing inequality: have computers changed the labour market’, Quarterly Journal of Economics, 113(7), 1169–214. Baldone, S., F. Sdogati and L. Tajoli (2001), ‘Patterns and determinants of international fragmentation of production: evidence from outward processing trade between the EU and Central Eastern European countries’, Weltwirtschaftliches Archiv, 137(1), 80–104. Baldwin, R. and F. Robert-Nicoud (2007), ‘Offshoring: General equilibrium effects on wages, production and trade’, National Bureau of Economic Research working paper series, no. 12991, Cambridge, MA. Batra, R.N. and F.R. Casas (1973), ‘Intermediate products and the pure theory of international trade: a neo-Heckscher-Ohlin framework’, American Economic Review, 63(3), 297–311. Belderbos, R.A. (1997), Japanese Electronics Multinationals and Strategic Trade Policy, Oxford: Oxford University Press. Berman, E., J. Bound and Z. Griliches (1994), ‘Changes in the demand for skilled labour within US manufacturing: evidence from the Annual Survey of Manufactures’, Quarterly Journal of Economics, 109(2), 367–97. Bernard, A.B., B. Jensen and P.K. Schott (2006), ‘Survival of the best fit: exposure to low wage countries and the (uneven) growth of US manufacturing plants’, Journal of International Economics, 68(1), 219–37. Bhagwati, J., A. Panagariya and T.N. Srinivasan (2004), ‘The muddles over outsourcing’, Journal of Economic Perspectives, 18(4), 93–114. Boltho, A. (1975), Japan – An Economic Survey: 1953–1973, Oxford: Oxford University Press.
References
153
Borrus, M. (1997), ‘Left for dead: Asian production networks and the revival of US electronics’, in. B. Naughton (ed.), The China Circle: Economics and Technology in the PRC, Taiwan, and Hong Kong, Washington, DC: Brookings Institution Press, pp. 139–63. Borrus, M., D. Ernst and S. Haggard (2000), International Production Networks in Asia: Rivalry or Riches, London: Routledge. Braconier, H. and K. Ekholm (2000), ‘Swedish multinationals and competition from high and low-wage locations’, Review of International Economics, 8(3), 448–61 Brainard, L. and D. Riker (1997a), ‘Are US multinationals exporting US jobs?’, National Bureau of Economic Research working paper series, no. 5958, Cambridge, MA. Brainard, L. and D. Riker (1997b), ‘US multinationals and competition from low wage countries’, National Bureau of Economic Research working paper series, no. 5959, Cambridge, MA. Brown, C. and G. Linden (2005), Offshoring in the semiconductor industry: a historical perspective’, in. L. Brainard and S.M. Collins (eds), Brookings Trade Forum: Offshoring White-collar Work, Washington, DC: Brookings Institution Press, pp. 279–322. Bulmer-Thomas, V. (1982), Input-Output Analysis in Developing Countries: Source, Methods and Application, New York: John Wiley and Sons. Campa, J. and L.S. Goldberg (1997), ‘The evolving external orientation of manufacturing industries: evidence from Four Countries’, National Bureau of Economic Research working paper series, no. 5919, Cambridge, MA.. Caves, R.E. (1996), Multinational Enterprises and Economic Analysis, 2nd edn, Cambridge: Cambridge University Press. Chen, Y., J. Ishikawa and Z. Yu (2004), ‘Trade liberalization and strategic outsourcing’, Journal of International Economics, 63(2), 419–36. Chiristensen, L., D.W. Jorgenson and L.J. Lau (1973), ‘Transcendental logarithmic production function’, Review of Economics and Statistics, 55(1), 28–45. Coase, R. (1937), ‘The nature of the firm’, Economica, 4(16), 386–405. Desai, M., F. Foley and J. Hines (2005), ‘Foreign direct investment and domestic economic activity’, National Bureau of Economic Research working paper series, no.11717, Cambridge, MA. Desai, M., F. Foley and J. Hines (2009), ‘Domestic effects of the foreign activities of US multinationals’, American Economic Journal: Economic Policy, 1(1), 181–203. Djankov, S., C. Freund and C.S. Pham (2006), ‘Trading on time’, World Bank policy research working paper, no. 3909, Washington, DC.
154
International fragmentation of production
Egger, H. and P. Egger (2003), ‘Outsourcing and skill-specific employment in a small economy: Austria after the fall of the Iron Curtain’, Oxford Economic Papers, 55(4), 625–43. Egger, H. and P. Egger (2005), ‘The Determinants of EU processing trade’, The World Economy, 28(2), 147–68. Eichengreen., B. and D.A. Irwin (1995), ‘Trade blocs, currency blocs and the reorientation of world trade in the 1930s’, Journal of International Economics, 38(1–2), 1–24. Ekholm, K. and K. Hakkala (2006), ‘The effect of offshoring on labour demand: evidence from Sweden’, Centre for Economic Policy Research working paper, no. 5648, London. Estevadeordal, A., B. Frantz and A.M. Taylor (2003), ‘The rise and fall of world trade, 1870–1939’, Quarterly Journal of Economics, 118(2), 359–407. Feenstra, R.C. (1998), ‘Integration of trade and disintegration of production in the global economy’, Journal of Economic Perspectives, 12(4), 31–50. Feenstra, R.C. (2004), Advanced International Trade: Theory and Evidence, Princeton, NJ: Princeton University Press. Feenstra, R.C. and G.H. Hanson (1996), ‘Globalization, outsourcing, and wage inequality’, American Economic Review, 86(2), 240–45. Feenstra, R.C. and G.H. Hanson (1999), ‘The impact of outsourcing and high-technology capital on wages: estimates for the United States, 1979–1990’, Quarterly Journal of Economics, 114(3), 907–40. Feenstra, R.C. and G.H. Hanson (2003), ‘Global production sharing and rising inequality: a survey of trade and wages’, in. E.K. Choi and J. Harrigan (eds), Handbook of International Trade, vol. 1, Oxford: Blackwell Publishing, pp. 147–85. Feenstra, R.C., G.H. Hanson and D.L. Swenson (2000), ‘Offshore assembly from the United States: production characteristics of the 9802 program’, in. R.C. Feenstra (ed.), The Impact of International Trade on Wages, Chicago, IL: University of Chicago Press, pp. 85–128. Feenstra, R.C., W. Hai, W.T. Woo and S. Yao (1999), ‘Discrepancies in international data: an application to China–Hong Kong entrepôt trade’, American Economic Review, 89(2), 338–43. Finger, J.M. (1975), ‘Tariff provisions for offshore assembly and exports of developing countries’, Economic Journal, 85(338), 365–71. Fors, G. and A. Kokko (2000), ‘Home country effects of FDI: foreign production and structural change in home country operations’, in M. Blomstrom and L. Goldberg (eds), Topics in Empirical International Economics: A Festschrift in Honour of Bob Lipsey, Chicago, IL: University of Chicago Press, pp. 137–62.
References
155
Froot, K.A. (1991), ‘Japanese foreign direct investment’, National Beau of Economic Research working paper series, no. 3737, Cambridge, MA. Fukao, K. (1995), ‘Outward direct investment and jobs in Japan’, Monthly Journal of the Japan Institute of Labour, 37(7), 2–12 (in Japanese). Fukao, K. and T. Amano (1998), ‘Outward foreign direct investment and manufacturing hollowing-out’, Keizai Kenkyu, 49(3), 256–76 (in Japanese). Fukao, K. and T. Yuan (2001), ‘Japanese outward FDI and hollowing out’, Research Institute of Economy, Trade and Industry discussion paper series, no. 003, Tokyo. Fukao, K., H. Ishido and K. Ito (2003), ‘Vertical intra-industry trade and foreign direct investment in East Asia’, Journal of the Japanese and International Economies, 17(4), 468–506. Fukao, K., Y.G. Kim and H.U. Kwon (2006), ‘Plant turnover and TFP dynamics in Japanese manufacturing’, Hitotsubashi University Institute of Economic Research Hi-Sat discussion paper series no. 180, Tokyo. Galenson, W. and K. Odaka (1976), The Japanese labour market, in. H. Patrick and H. Rosovsky (eds), Asia’s New Giant: How the Japanese Economy Works, Washington, DC: Brookings Institution, pp. 587–671. Görg, H. (2000), ‘Fragmentation and trade: US inward processing trade in the EU’, Weltwirtschaftliches Archiv, 136(3), 403–22. Griliches, Z. and J. Hausman (1986), ‘Errors in variables in panel data’, Journal of Econometrics, 31(1), 93–118. Grossman, G.M. and E. Helpman (2002), ‘Integration versus outsourcing in industry equilibrium’, Quarterly Journal of Economics, 117(1), 85–120. Grossman, G.M. and E. Helpman (2005), ‘Outsourcing in a global economy’, Review of Economic Studies, 72(1), 135–59. Grossman, G. and E. Rossi-Hansberg (2008), ‘Trading tasks: a simple theory of offshoring’, American Economic Review, 98(5), 1978–97. Grunwald, J. and K. Flamm (1985), The Global Factory: Foreign Assembly in International Trade, Washington, DC: Brookings Institution. Hackett, S.C. and K. Srinivasan (1998), ‘Do supplier switching costs differ across Japanese and US multinational firms’, Japan and the World Economy, 10, 13–32. Hamermesh, D.S. (1993), Labour Demand, Princeton, NJ: Princeton University Press. Hanson, G.H. (1997), ‘The effects of offshore assembly on industry location: evidence from US border cities’, in R.C. Feenstra (ed.), Effects of
156
International fragmentation of production
US Trade Protection and Promotion Policies, Chicago, IL: University of Chicago Press, pp. 297–322. Hanson, G.H., R.J. Mataloni and M.J. Slaughter (2001), ‘Expansion strategies of US multinational firms’, in Susan M. Colins and Dani Rodrick (eds), Brookings Trade Forum 2001, Washington, DC: Brookings Institution, pp. 245–82. Hanson, G.H., J. Raymond J. Mataloni and M.J. Slaughter (2003), ‘Expansion abroad and the domestic operations of US multinational firms’, accessed 15 April, 2010 http://mba.tuck.dartmouth.edu/pages/ faculty/matthew.slaughter/. Hansson, P. (2000), ‘Relative demand for skills in Swedish manufacturing: technology or trade?’, Review of International Economics, 8(3), 533–55. Harrigan, J. (2003), ‘Specialization and the volume of trade: Do the data obey the laws?’, in E.K. Choi and J. Harrigan (eds), Handbook of International Trade, vol. 1 Malden, MA, and Oxford: Basil Blackwell, pp. 85–118. Harrison, A.E. and M.S. McMillan (2006), ‘Outsouring jobs? Multinationals and US employment’, National Bureau of Economic Research working paper series, no. 12372, Cambridge, MA. Harrison, A.E., M.S. McMillan and C. Null (2007), ‘US multinational activity abroad and US jobs: substitutions and complements?’, Industrial Relations, 46(2), 347–65. Hasan, R., D. Mitra and K.V. Ramaswamy (2007), ‘Trade reforms, labour regulations and labour demand elasticities: empirical evidence from India’, Review of Economics and Statistics, 89(3), 466–81. Head, K. and J. Ries (2001), ‘Overseas investment and firm exports’, Review of International Economics, 9(1), 108–22. Head, K. and J. Ries (2002), ‘Offshore production and skill upgrading by Japanese manufacturing firms’, Journal of International Economics, 58(1), 81–105. Head, K., J. Ries and D. Swenson (1999), ‘The attraction of foreign manufacturing investments: investment promotion and agglomeration economies’, Regional Science and Urban Economics, 29(2), 197–218. Helg, R. and L. Tajoli (2005), ‘Patterns of international fragmentation of production and implications for the labour markets’, North American Journal of Economics and Finance, 16(2), 235–54. Helleiner, G.K. (1973), ‘Manufacturing exports from less developed countries and multinational firms’, Economic Journal, 83(329), 21–47. Helpman, E. (2006), ‘Trade, FDI, and the organization of firms’, Journal of Economic Literature, 44(3), 589–630. Hijzen, A., H. Görg and R.C. Hine (2005), ‘International outsourcing and the skill structure of labour demand in the United Kingdom’, Economic Journal, 115(506), 860–78.
References
157
Hijzen, A., T. Inui and Y. Todo (2006), ‘The effects of multinational production on domestic performance: evidence from Japanese firms’, Research Institute of Economy, Trade and Industry discussion paper series, no. 006, Tokyo. Hsieh, C. and K.T. Woo (2005), ‘The impact of outsourcing to China on Hong Kong’s labour market’, American Economic Review, 95(5), 1673–87. Hummels, D. (2002), ‘Book reviews for Fragmentation: New Production Patterns in the Global Economy, in S.W. Arndt and H. Kierzkowski (eds), Oxford and New York: Oxford University Press’, Journal of Economic Geography 2(3), 368–9. Hummels, D. (2007), ‘Transportation costs and international trade in the second era of globalization’, Journal of Economic Perspectives, 21(3), 131–54. Hummels, D., J. Ishii and K.-M. Yi (2001), ‘The nature and growth of vertical specialization in world trade’, Journal of International Economics, 54(1), 75–96. International Monetary Fund (IMF) (2006), International Financial Statistics, (CD-rom), Washington, DC: IMF. Ito, K. and K. Fukao (2005), ‘Physical and human capital deepening and new trade patterns in Japan’, in T. Ito and A.K. Rose (eds), International Trade in East Asia, NBER-East Asia Seminar on Economics, no. 14, Chicago, IL: University of Chicago Press, 14, pp. 7–49. Jones, R.W. (2000), Globalization and the Theory of Input Trade, Cambridge, MA: MIT Press. Jones, R.W. and H. Kierzkowski (1990), ‘The role of services in production and international trade: a theoretical framework’, in. R.W. Jones and A.O. Krueger (eds), The Political Economy of International Trade: Essays in Honour of Robert A. Mundell, Cambridge, MA: MIT Press, pp. 31–48. Jones, R.W. and H. Kierzkowski (2001a), ‘A framework for fragmentation’, in S.W. Arndt and H. Kierzkowski (eds), Fragmentation: New Production Patterns in the World Economy, New York: Oxford University Press, pp. 17–34. Jones, R.W. and H. Kierzkowski (2001b), ‘Horizontal aspects of vertical fragmentation’, in L.K. Cheng and H. Kierzkowski (eds), Global Production and Trade in East Asia, Boston, MA: Kluwer Academic Publishers, pp. 33–51. Jones, R., H. Kierzkowski and L. Chen (2004), ‘What does evidence tell us about fragmentation and outsourcing?’, International Review of Economics and Finance, 14(3), 305–16. Kaminski, B. and F. Ng (2005), ‘Production fragmentation and
158
International fragmentation of production
integration of Central Europe into global markets’, International Review of Economics and Finance, 14(3), 377–90. Katz, L. and K. Murphy (1992), ‘Changes in relative wages, 1963–1987: supply and demand factors’, The Quarterly Journal of Economics, 107(1), 35–78. Katz, L. and A. Revenga (1989), ‘Changes in the structure of wages: the United States vs. Japan’, Journal of the Japanese and International Economies, 3(4), 522–53. Kaufmann, D., A. Kraay and M. Mastruzzi (2007), ‘Governance matters VI: aggregate and individual governance indicators, 1996–2006’, World Bank policy research working paper, no. 4280, Washington, DC. Kimura, F. and M. Ando (2003), ‘Fragmentation and agglomeration matter: Japanese multinationals in Latin America and East Asia’, North American Journal of Economics and Finance, 14(3), 287–317. Kimura, F. and M. Ando (2005), ‘Two-dimensional fragmentation in East Asia: Conceptual framework and empirics’, International Review of Economics and Finance, 14(3), 317–48. Kimura, F. and K. Kiyota (2006), ‘Exports, FDI, and productivity: dynamic evidence from Japanese firms’, Weltwirtschaftliches Archiv (Review of World Economics), 142(4), 695–719. Kohler, W. (2001), ‘A specific-factors view on outsourcing’, North American Journal of Economics and Finance, 12(1), 31–53. Kravis, I.B. and R.E. Lipsey (1988), ‘The effect of multinational firms’ foreign operations on their domestic employment’, National Bureau of Economic Research working paper series, no. 2760, Cambridge, MA. Krugman, P. (1995), ‘Growing world trade: causes and consequences’, Brookings Papers on Economic Activity, 56, 269–93. Lall, S., M. Albaladejo and J. Zhang (2004), ‘Mapping fragmentation: electronics and automobiles in East Asia and Latin America’, Oxford Development Studies, 32(3), 407–32. Lawrence, R.Z. (1991), ‘Efficient or exclusionist? The import behaviour of Japanese corporate groups’, Brooking Papers on Economic Activity, 1, 311–41. Lawrence, R.Z. and M. Slaughter (1993), International trade and American wages in the 1980s: giant sucking sound or small hiccup’, in M.N. Baily and C. Winston (eds), Brookings Papers on Economic Activity: Microeconomics, 161–226. Leamer, E.E. (1994), ‘Trade, wages and revolving door ideas’, National Bureau of Economic Research working paper series, no. 4716, Cambridge, MA. Levchenko, A.A. (2008), ‘Institutional quality and international trade’, Review of Economic Studies, 74(3), 791–819.
References
159
Limao, N. and A. Venables (2001), ‘Infrastructure, geographical disadvantage, transport costs and trade’, World Bank Economic Review, 15, 451–79. Lipsey, R.E. (1995), ‘Outward direct investment and the US economy’, in M. Feldstein, J.R. Hines and G. Hubbard (eds), The Effects of Taxation on Multinational Corporations, Chicago, IL: University of Chicago Press, pp. 7–33. Lipsey, R.E. (2003), ‘Foreign direct investment and the operations of multinational firms: concepts, history, and data’, in. E.K. Choi and J. Harrigan (eds), Handbook of International Trade, vol. 1, Malden, MA and Oxford: Basil Blackwell, pp. 287–319. Lipsey, R., E. Ramstetter and M. Blomström (1999), ‘Parent exports and affiliate activity in Japanese multinational companies, 1986, 1989, 1992’, in Institute for International Trade and Investment (ed.), Analytical Research Based on Data from the Survey of Overseas Business Activities, Survey Research on Harmonizing Globalization Based on the 1997 Survey of Overseas Business Activities, Tokyo: Institute for International Trade and Investment, pp. 50–97. Macher, J.T. and D.C. Mowey (2004), ‘Vertical specialisation and industry structure in high technologies industries’, in J.A.C. Baum and A.M. McGahm (eds), Business Strategy Over the Industry Life Cycle, no. 21, London: Elsevier, pp. 317–55. Maskus, K.E. (1991), ‘Comparing international trade and products and national characteristics data for the analysis of trade models’, in P. Hooper and D.J. Richardson (eds), International Economic Transactions: Issues in Measurement and Empirical Research, no. 55, Chicago, IL: University of Chicago Press, pp. 17–60. Mataloni, R. (1995), ‘A guide to BEA statistics on US multinational companies’, Survey of Current Business, 38, 38–55. Matsuura, T. (2004), ‘Estimation of economic activity of Japanese multinationals’, Keizai Kenkyu, 32(4), 1–16 (in Japanese). Matsuura, T. and K. Kiyota (2004), ‘On the construction and utilisation of panel data based on the “Basic Survey of Business Structure and Activity”: application to economic analysis and the data management issue’, Research Institute of Economy, Trade and Industry policy discussion paper, no.004, Tokyo (in Japanese). McLaren, J. (2000), ‘“Globalization” and vertical structure’, American Economic Review, 90(5), 1239–54. METI (Ministry of Economy, Trade and Industry) (various years), The Census of Manufactures, Tokyo: METI, government of Japan. Minami, R. (1986), The Economic Development of Japan: A Quantitative Study, Basingstoke and London: Macmillan.
160
International fragmentation of production
Navaretti, G.B. and A.J. Venables (2004), Multinational Firms in the World Economy, Princeton, NJ: Princeton University Press. Nishimura, K.G. (2005), ‘Comments on “Product fragmentation and trade patterns in East Asia” Athukorala, P’, Asian Economic Papers, 4(3), 28–9. Nishimura, K.G., T. Nakajima and K. Kiyota (2005), ‘Does the natural selection mechanism still work in severe recessions? Examination of the Japanese economy in the 1990s’, Journal of Economic Behaviour and Organization, 58, 53–78. Nunn, N. (2007), ‘Relationship-specificity, incomplete contracts, and the pattern of trade’, Quarterly Journal of Economics, 122(2), 569–600. Nunn, N. and D. Trefler (2008), ‘The boundaries of the multinational firm: an empirical analysis’, in E. Helpman, D. Marin and T. Verdier (eds), The Organization of Firms in a Global Economy, Cambridge: Cambridge University Press, pp. 55–83. Odagiri, H. and H. Yasuda (1996), ‘The determinants of overseas R&D by Japanese firms: an empirical study at the industry and company levels’, Research Policy, 25(4), 1059–79. OECD (Organisation for Economic Co-operation and Development) (1997), OECD Employment Outlook, Paris: OECD. Ohkawa, K. and H. Rosovsky (1973), Japanese Economic Growth: Trend Acceleration in the Twentieth Century, Stanford, CA: Stanford University Press. Paprzycki, R. (2004), Interfirm Networks in the Japanese Electronics Industry, London and New York: Routledge Curzon. Patrick, H. and H. Rosovsky (1976), ‘Japan’s economic performance: an overview’, in H. Patrick and H. Rosovsky (eds), Asia’s New Giant: How the Japanese Economy Works, Washington, DC: Brookings Institution, pp. 1–61. Ramstetter, E.D. (1996), ‘Estimating economic activity by Japanese transnational corporations: how to make sense of the data?’, Transnational Corporations, 5(2), 107–43. Riedel, J. (1976), ‘Intermediate products and the theory of international trade: a generalization of the pure intermediate good case’, American Economic Review 66(3), 441–7. Roberts, M.J. and E. Skoufias (1997), ‘The long-run demand for skilled and unskilled labor in Colombian manufacturing industries’, Review of Economics and Statistics, 79(2), 330–4. Sachs, J. and H. Shatz (1994), ‘Trade and jobs in US manufacturing’, Brookings Papers on Economic Activity, 1, 1–84. Sakurai, K. (2000), ‘Globalisation and labour market: The case of Japanese
References
161
manufacturing’, in Keizai Keiei Kenkyu [Economics Today], no. 2102, Tokyo: Research Institute of Capital Formation, Development Bank of Japan. Sakurai, K. (2001), ‘Biased technological change and Japanese manufacturing employment’, Journal of the Japanese and International Economies, 15(3), 298–322. Sasaki, H. and H. Sakura (2005), ‘Changes in the demand for skilled labour within Japan’s manufacturing sector: Effects of skill-biased technological change and globalization’, Bank of Japan working paper series, no. 012, Tokyo. Saxonhouse, G.R. (1993), ‘What does Japanese trade structure tell us about Japanese trade policy?’, Journal of Economic Perspectives, 7(3), 21–43. Schweinberger, A. (1975), ‘Pure traded intermediate products and the Heckscher-Ohlin theorem’, American Economic Review, 65(4), 634–43. Sharpton, M. (1975), ‘International subcontracting’, Oxford Economic Papers, 27(1), 94–135. Shimizutani, S. and Y. Todo (2007), ‘What determines overseas R&D activities? The case of Japanese multinational firms’, Research Institute of Economy, Trade and Industry discussion paper series, no. 010, Tokyo. Shy, O. and R. Stenbacka (2003), ‘Strategic outsourcing’, Journal of Economic Behaviour and Organization, 50(2), 203–24. Skaksen, J.R. and A. Sørensen (2002), ‘Skill upgrading and rigid relative wages: the case of Danish manufacturing’, Institute for the Study of Labour (IZA), IZA discussion paper series no. 664, Bonn. Soloaga, I. and A. Winters (2001), ‘Regionalism in the nineties: what effect on trade?’ North American Journal of Economics and Finance, 12, 1–29. Spencer, B.J. (2005), ‘International outsourcing and incomplete contracts’, Canadian Journal of Economics, 58(1), 49–79. Strauss-Kahn, V. (2004), ‘The role of globalization in the within-industry shift away from unskilled workers in France’, in R.E. Baldwin and L.A. Winters (eds), Challenges to Globalization: Analysing the Economics, Chicago, IL: Chicago University Press, pp. 201–31. Sturgeon, T.J. (2003), ‘What really goes on in Silicon Valley? Spatial clustering and dispersal in modular production networks’, Journal of Economic Geography, 3, 199–225. Swenson, D.L. (2005), ‘Overseas assembly and country sourcing choices’, Journal of International Economics, 66(1), 107–30. Swenson, D.L. (2007), ‘Competition and the location of overseas assembly’, Canadian Journal of Economics, 40(1), 155–75.
162
International fragmentation of production
Todo, Y. and S. Shimizutani (2008), ‘Overseas R&D activities and home productivity growth: evidence from Japanese firm-level data’, Journal of Industrial Economics, 56(4), 752–77. Tomiura, E. (2004), ‘Import competition and employment in Japan: Plant start-up, shutdown and product changes’, Japanese Economic Review, 55(2), 141–52. Tomiura, E. (2005), ‘Foreign outsourcing and firm-level characteristics: evidence from Japanese manufactures’, Journal of the Japanese and International Economies, 19(2), 255–71. UNCTAD (United Nations Conferences on Trade and Development) (1998), World Investment Report: The Financial Crisis in Asia and Foreign Direct Investment: An Assessment, Geneva: United Nations. UNCTAD (2002), World Investment Report: Transnational Corporations and Export Competitiveness, Geneva: United Nations. UNIDO (United Nations Industrial Development Organization) (2006), Industrial Statistics (CD-Rom), Rome: United Nations. Venables, A.J. (1999), ‘Fragmentation and multinational production’, European Economic Review, 43(4–6), 935–45. Watanabe, S. (1972), ‘International subcontracting, employment and skill promotion’, International Labour Review, 105(5), 425–49. Wilson, J.S., C.L. Mann and T. Otsuki (2003), ‘Trade facilitation and economic development: measuring the impact’, World Bank working paper series, no. 2988, Washington, DC. Wooldridge, J.M. (2000), Introductory Econometrics: A Modern Approach, Cincinnati, OH: South-Western Educational Publishing. Yasuba, Y. (1978), ‘Freight rates and productivity in ocean transportation for Japan, 1875–1943’, Explorations in Economic History, 15(1), 11–39. Yeaple, S.R. (2003), ‘The complex integration strategies of multinationals and cross country dependencies in the structure of foreign direct investment’, Journal of International Economics, 60(2), 293–314. Yeats, A.J. (1978), ‘On the accuracy of partner country trade statistics’, Oxford Bulletin of Economics and Statistics, 40(4), 341–61. Yeats, A.J. (1995), ‘Are partner-country statistics useful for estimating “missing” trade data?’, World Bank policy research working papers, no. 1501, Washington, DC. Yeats, A.J. (2001), ‘Just how big is global production sharing’, in. S.W. Arndt and H. Kierzkowski (eds), Fragmentation: New Production Patterns in the World Economy, Oxford: Oxford University Press, pp. 63–109. Yi, K.-M. (2003), ‘Can vertical specialization explain the growth of world trade?’, Journal of Political economy, 111(1), 52–102.
Index AFTA 36–7, 38–9 age Japan rewards seniority 81–2 apparel 42 see also garments; textiles Asia East x, 3, 8, 43, 44–7, 58, 61, 87, 89, 94, 98–102, 103, 108, 118, 127, 130–31, 138, 140 SE 8 see also China Argentina 66 arms-length transaction 6, 9, 16–17 assembly jobs 11, 18 Australia 66, 103 Austria 66, 86, 91, 103 automotive industry 6, 7, 9, 19, 20, 26, 28, 31, 41, 42, 43–8, 96 components 40–41 Barbie doll 7 batteries, long-life 8 BEA (Bureau of Economics Analysis) data 119 see also pay BEC 28 Belgium 66, 103 bonus payments 125 boundaries of firms 16, 17 Brazil 66 broadband 8, 13 Broad Economic Category (BEC) 28 Canada 24, 66, 103 categories see classification Census of Manufactures see surveys chemicals industry 8, 17, 23 China 3, 32, 36–7, 38–9, 43, 44–5, 66, 103, 148 CIF 23
circuits, printed 22 see also semi-conductors classification systems 119 Broad Economic Category (BEC) 28 Japan Statistical Industry Classification 109 Standard Industrial Classification (SIC) 24, 88, 90, 105, 106, 107 see also BEA; ISCO; ISIC; standards; SITC Clestina 21 CM see surveys communications 8 see also costs, service link, broadband ‘comparative advantage’ theory see theories components 150 automotive 148 econometric analysis 5–68 electrical machinery 40–48 exports 35, 36, 43 imports 38–9 office machines 40–48 power-generating 40 road vehicles 40 Shitauke (component manufacturers and assemblers) see also parts computer 9 chips 8, 20 see also semi-conductors laptops 9 organisers 9 PCs exports 36–7 imports 38–9 software 23 confidential information 17
163
164
International fragmentation of production
contract difficulties 16 manufacturers 9 Costa Rica 7, 66 costs 13, 14 fixed 13, 14 marginal 14 service link 8, 13, 14, 51, 54, 147 transaction 16–17 unit labour costs 3, 53, 55, 58, 61, 63, 71 see also labour market; transportation crude oil 32 currency exchange rates 23, 62 yen appreciation 1990s 32 Czech Republic 66 data, trade 5–29, 62–8 see also statistics Denmark 66, 91, 103 design 11 see also IP; R&D developing countries 35, 36–7, 38–9, 54, 103 see also Asia, East; China, Hong Kong, Korea, Malaysia, Phillipines, Singapore domestic workforce 110–45 see also employment; skills econometric analysis 5–68, 92–6, 110–45 education 82–3, 89 see also skills electrical machinery 11, 17, 41, 42, 74, 96 electronics 7, 9, 20, 21, 22, 31, 41, 43–8 see also semi-conductors employment effect on Japan’s employment of ‘exporting jobs’ 2, 110 globalization in manufacturing 3–4 production fragmentation 110–45, 149 manual labour 7, 75 see also skills, production/ non-production substitution effect 118, 119, 123
temporary workers 20 see also labour market; skills engineering 11, 123 entrepot trade 23 see also ports Ericsson 19 estimation method 55–61, 125–7 first difference 100 generalized least squares 97 generalized method of moments 125 LSDV (least square dummy variable) 97 weighted least squares (WLS) 98, 99, 100 within-transformation 97, 99, 100, 126, 127, 128, 130, 131, 132, 133, 134, 135, 136, 137, 138 EU 127–8, 134–5, 138, 140 ‘EU 15’ 43 exports 36–7, 43, 44–6 imports 38–9, 43, 44–6 processing schemes 51 IPT (Inward Processing Trade) 24, 51 OPT (Outward Processing Trade) 24 exchange rates RER (real exchange rates) 54, 65–6 see also currency exports out of Japan 31–2, 34–9, 42, 44–5 Factor Price Equalization 15 FDI (foreign direct investment) 75, 110, 116, 138, 143 ‘final goods’ 46–7 Finland 66, 103 Flextronics 19 FOB (free on board) 23 food sector 74 Ford 19 France 36–7, 38–9, 66, 86, 88, 90, 103 freight forwarders 65 see also CIF; infrastructure; ports furniture 42, 96 garment industry 7, 42 see also apparel; textiles GDP deflator 65 consumer price index 125
Index General Motors (GM) 19 geographic proximity 50–68 Germany 36–7, 38–9, 66, 86, 88, 90, 103 global trends 31–49 gravity model’ 2–3, 50–68, 148 variables 63 Greece 103 Grossman-Hart see models Hausmam-Wu specification test 56 Heckscher-Ohlin theory 10, 15 heteroscedasticity 56, 57, 60, 98, 99, 100, 107, 126, 129, 137 Hewlett Packard 19 Hitachi 20 Hong Kong 23, 36–7, 38–9, 66, 86, 88, 90, 103 Hungary 66 import penetration 26, 27, 89, 90, 91, 101, 124, 125, 128, 131, 133, 135, 137 imports into Japan 32–4, 36–9, 42 indigenous workforce 110–45, 149 see also employment; skills India 66 Indonesia 66, 103 industrial organization see model infrastructure 53–4, 61, 64–5, 150 input-output see I-O Tables institutional structure 53–4, 64 see also legal system insurance see CIF Intel Corp 7 ‘intermediate inputs’ 27 international fragmentation of production 6 international production networks 19–21 modular production network 8, 9, 17, 19, 21 modular technology 8, 9, 19 relational production network 19, 20 ‘I-O’ Tables 25–6, 27, 29, 89 imported intermediate inputs 25–7, 88, 89, 90–92, 94, 105, 106 IPT see EU Ireland 51, 66, 103
165
ISCO (International Standard Classification of Occupations) 81 ISIC (International Standard Industry Classification) 62 Israel 66 Italy 66, 86, 88, 90, 103 ITC (information/technology communications sector) 116, 123 see also computers; telecoms JIP 96–7, 104–5 (Japan Industrial Productivity) jobs see employment Korea, Republic of 36–7, 38–9 labour market 110–45 adjustment 86–108, 148 costs x, 15, 18, 5, 58, 61, 150 unit costs (ULCs) 53, 61, 63–4 structural adjustments 3, 70–85, 86–108 temporary workers 20 lead firms 9, 19, 20 legal system 54 see also institutional structure Lerner-Pearce theory 10, 11 literature business 5–29 econometric analysis 51–2 theoretical 5–29 machine tools 23 machinery 31, 42, 44–5, 53, 55–6, 62, 73–4, 96, 116 see also automotive industry; electronics Malaysia 7, 66, 103 manual labour 7, 75 ‘market thickness’ 16, 18 materials imported 32 raw 15, 29 measurement 21–8, 55–6 metals 27, 116 methodology estimation 125–7 METI surveys Basic Survey of Business Structure and Activity
166
International fragmentation of production
(1991–2002) x, 3, 96, 110–45, 147 Overseas Japanese Business Activity (1991–2002) x, 3, 110–45, 147 see also surveys Mexico 66, 103 milling 74 Ministry of Economy, Trade and Industry (METI) 3, 84, 111, 125, 139, 140–46 missiles 9 MNEs x, 2, 4, 7, 9, 19, 22, 50, 64, 138, 148, 150 domestic employment harmed by international fragmentation of production? 4, 111–45 employment adjustment in 110–45 European 8 models (fixed v random effects) 97 gravity model of bilateral trade flows 53–5 Grossman-Hart property rights model 17 industrial organization 15–19 monopolistic competition 17 Ricardian theory 10, 53 motor industry see automotive industry NEC 21 neo-classical trade model 5, 9–15, 5–29 Netherlands, The 66, 103 New Zealand 103 Nissan 20 Norway 66, 103 OAP see statistics; US OECD countries 87, 94, 98–102, 103 use of OECD workforce having skills downgrading effect? x, 3, 87 office machinery 42, 96 see also electronics OLS (ordinary least squares) 126, 127 OPY see EU outsourcing ‘broad’ 26 ‘narrow’ 26
packing jobs 11 parts and components 5, 6, 8, 16, 27, 28, 50–68 firm-specific 9 product-specific 9 trade data 5, 21–4 pay 14–15, 51–2, 78–9 BEA wage data 64 increase for low-skilled workers/fall of production costs 13, 14–15 US 68 see also labour market, costs; skills pharmaceuticals 23 Phillippines 7, 66, 103 photographic apparatus 42 plastic products 96 Poland 66 policy 1, 2, 149–50 see also legal system ports 54, 64 see also entrepot trade; infrastructure Portugal 66, 103 processing jobs 11 producer price index 65 wholesale price index 65, 125 product fragmentation 6 production fragmentation 86–108 jobs 75–6, 81–2, 84 see also skills (production/nonproduction workers) professional scientific exports/imports 42 quality control 20 R&D 11, 20, 95, 99, 104, 119, 120, 123, 125 regression analysis 66, 87, 96, 98, 104, 122, 127 regulatory environment see institutional structure; legal system RER see exchange rates research and development see R&D Research Institute of Economy, Trade and Industry (RIETI) 104, 143, 145
Index Research Institute of Industrial Economics (IUI) 121 Ricardian theory 10, 53 Ricardo-Viner theory 10 Russian Federation 66 Rybczynski theory 15 scale effect 101, 118, 122–3 scientific equipment 116 semi-conductors 7, 19 service link costs see costs sewing machines 7 shipping see freight; infrastructure; ports Singapore 36–7, 38–9, 66, 103 SITC (Standard International Trade Classification) 21–3, 28, 40, 42, 43, 56, 62–3, 64 skills 32, 52, 53, 70–84, 86–108 affected by deindustrialization 75 geographical fragmentation of trade? ix–x East Asian workforce x highly-skilled 10, 11, 77, 123 design 11 engineering 11, 123 IT 123 non-production 75–6, 81–2, 84, 105 see also R&D low-skilled 10, 11, 77, 149–50 assembly 11, 18 packing 11 processing 11 production 75–6, 81–2, 84, 105 measure of worker skills’ intensity 80–81 OECD countries’ workforce x, 3, 87, 103 see also OECD relative pay 77–80 supply and demand 77–80 upgrading 75–84, 86–108, 148 unskilled workers 4 Slovakia 66 Slovenia 66 Solectron 19, 21 Sony 21
167
sound recording apparatus 42, 43–8 see also electronics South Africa 66 South America 136–7, 140 Spain 66, 103 standards see Broad Economic Category (BEC); classification systems; ISCO; SITC statistics 5 Input-Output Table (I-O Table) 5, 21, 25–7, 29, 89 Offshore Assembly Programme (OAP) 5, 21, 24–5, 29, 51 see also US UN Comtrade 27, 34, 39, 42, 45, 49, 56, 62, 67, 96, 105, 106 trade data see parts and components steel industry 24, 27 see also metals Stolper-Samuelson theorem 15 subcontracting 6, 17–8 suppliers see ‘market thickness’ surveys Annual Survey of Manufacturers 25, 88, 90 Basic Survey on Wage Structure 76, 83, 84 Census of Manufacturers (CM) 76, 84 see also METI surveys Sweden 36–7, 38–9, 66, 86, 88, 91, 103, 110, 111, 121–2 switchgear 22 Switzerland 66 Taiwan 66, 103 tariffs 8, 24, 52 multilateral tariff reductions 25 see also CIF; FOB tax exemptions 24, 51 incentives 7 see also tariffs technology see computers; ITC telecoms 42, 43–8 see also electronics; ITC temporary workers 20 Texas Instruments 7 textiles 11, 73–4, 116 see also apparel; garments
168
International fragmentation of production
Thailand 66, 103 theoretical literature 9–21 theories (theorem) 5–29 ‘comparative advantage’ 5 Factor Price Equalization 15 Heckscher-Ohlin 10, 15 Lerner-Pearce 10, 11 Ricardian 10 Ricardo-Viner 10 Rybczynski 15 Stolper-Samuelson 15 toasters 9 tobacco 96 Toshiba 20 Toyota 20 toys Mattell Barbie doll 7 trade data see EU, IPT/OPT; statistics fragmentation trade 1–3, 8, 9, 15, 21, 22, 23, 25, 26, 27, 29, 31, 32, 34–6, 40–43, 46, 48, 50–69, 86, 87, 88, 89, 90, 93, 96, 98–102, 106, 108, 118, 147, 148, 149 theory see theories trade facilitation 54, 55, 56, 68 transportation 74, 116 costs x, 8, 13, 15, 23, 147 air freight 8 containers 8 equipment exports/imports 42 see also costs, service link; infrastruture Turkey 66 UK 36–7, 38–9, 66, 70, 86, 88, 90, 103 ULCs see labour market costs
UN see Broad Economic Category (BEC) unionisation 123 United Kingdom see UK unskilled workers see skills US 66, 103, 106, 132–3 experience ix, x, 1, 2, 3, 50–68 exports 35, 36–7, 43–5 imports 32, 38–9, 40, 43–5 industry sectors 7 inward processing trade (IPT) 51 Japanese firms in 117–18 loss of domestic jobs 110, 121 manufacturing 35–8 OAP (Offshore Assembly Programme) 7, 52 outsourcing/effect on nonproduction wages 88 skills upgrading 90, 107 Tariff Act 1930 24 US–Japan comparison 19, 47–8, 50–68, 110–11, 113, 147, 148 value-added 7, 24, 25, 63, 71, 94, 95, 96, 104 vertical structure (Keiretsu) 20 Vietnam 103 wages wage rates 68 see also pay; labour market workforce, indigenous to Japan 32 World Bank Development Indicators 126 Doing Business Survey 65 yen see currency; exchange rates