Master Thesis Double Degree Program: Master of Arts in
Master of Science in International Economic and Business (University of Groningen)
The Impact of
on the Skill Structure of Em
A Global Supply Chain Perspective
Supervisors: Dr. Bart Los (University of Groningen) Prof. Inmaculada Martínez
Affiliation: University of Groningen, Faculty of Economics and Business University of Göttingen, Faculty of Economic Sciences
Deadline: 8th July 2013
Student number: 2418509 (University of Groningen) 21146917 (University of Göttingen)
Email: L.R.M.Gerling@student.rug.nl
Master Thesis Double Degree Program:
in International Economics (University of Göttingen)
Master of Science in International Economic and Business (University of Groningen)
The Impact of International Production Fragmentation
the Skill Structure of Employment in Emerging Economies:
A Global Supply Chain Perspective
Presented by Lena Gerling
Dr. Bart Los (University of Groningen)
Prof. Inmaculada Martínez-Zarzoso, Ph.D. (University of Göttingen)
University of Groningen, Faculty of Economics and Business University of Göttingen, Faculty of Economic Sciences
July 2013
2418509 (University of Groningen) 21146917 (University of Göttingen)
L.R.M.Gerling@student.rug.nl
International Economics (University of Göttingen)
Master of Science in International Economic and Business (University of Groningen)
Production Fragmentation
ployment in Emerging Economies:
Zarzoso, Ph.D. (University of Göttingen)
I
Abstract
This paper analyzes the impact of international production fragmentation on employment changes in 17 emerging economies over the period from 1995 to 2008 by applying a structural decomposition technique on international input-output tables. The results show that emerging economies have become strongly integrated into international production networks, gaining mainly in terms of medium-skilled employment. In addition, the analysis investigates the de-terminants of skill-specific employment relocations within global supply chains towards emerging economies. While shifts in low-skilled labor are found to be predominantly deter-mined by factor endowments, for high-skilled intensive activities additional factors like spa-tial proximity, market size, political institutions and the physical infrastructure of a country of employment must be taken into account. In contrast, the relocation of medium-skilled tasks is driven in a more balanced way by the different determinants.
Keywords
II
Table of Content
List of Tables ... III List of Figures ... IV List of Abbreviations ... IV
1. Introduction ... 1
2. Theoretical Background ... 2
2.1 A Global Supply Chain Perspective on Employment Changes in Emerging Economies 2 2.2 Determinants of the Distribution of Skills along Global Supply Chains ... 9
3. The Impact of Production Fragmentation on Employment in Emerging Economies – Input-Output Analysis ... 14
3.1 Empirical Approach and Assumptions ... 14
3.2 Analytical Framework ... 16
3.3 Data from the World Input-Output Database ... 21
3.4 Stylized Facts ... 22
4. Determinants of Employment Relocations to Emerging Economies along Global Supply Chains – Regression Analysis ... 31
4.1 Model Specification ... 31
4.2 Variable Operationalization and Data ... 35
4.3 Analysis and Results ... 37
4.3.1 OLS Results ... 37
4.3.2 SUR Results ... 41
4.3.3 Results for Country of Employment Subsamples ... 45
4.3.4 Results for Manufacturing Activities Only ... 50
4.4 Discussion ... 52
5. Conclusion ... 56
References ... 60
III
List of Tables
Table 1 Employment distribution along the U.S. automotive GSC, by country group and skill type.
Table 2 Distribution of employment shares along advanced countries’ manufacturing GSCs.
Table 3 Employment changes caused by relocations along advanced countries' manu-facturing GSCs, 1995-2008.
Table 4 Employment changes caused by relocations along advanced countries' manu-facturing GSCs, by GSC type, 1995-2008.
Table 5 Descriptive statistics, full sample.
Table 6 OLS results for absolute and relative employment changes. Table 7 SUR results, full sample.
Table 8 Test for equal coefficients across skill-specific equations, full sample. Table 9 Descriptive statistics for CoE-subsamples.
Table 10 SUR results by CoE-subsamples.
Table 11 Test for equal coefficients across skill-specific equations, CoE-subsamples. Table 12 SUR results for employment changes in manufacturing sectors only.
Table 13 SUR result for standardized variables. Table A1 Sample Countries.
Table A2 Industry Classification in the WIOD.
Table A3 Decomposition of employment changes due to participation in advanced coun-tries manufacturing GSCs, 1995-2008.
Table A4 Variable Descriptions and Data Sources. Table A5 Correlation matrix, full sample.
Table A6 Variance inflation factors.
IV
List of Figures
Figure 1 Employment distribution within the U.S. automotive GSC, by country group. Figure 2 Stylized word input-output table.
Figure 3 Decomposition of employment changes in emerging economies due to partici-pation in advanced countries’ manufacturing GSCs, 1995-2008.
Figure 4 Decomposition of employment changes in emerging countries due to participa-tion in advanced countries’ manufacturing GSCs, 1995-2008, selected coun-tries.
List of Abbreviations
CEPII Center d’études prospectives et d’informations internationales CEEC Central and Eastern European countries
CoC Country of completion
CoE Country of employment
FDI Foreign direct investment GDP Gross domestic product GSC Global supply chain GVC Global value chain
HS High skilled
ICT Information and communication technology IO table Input-output table
ISCED International Standard Classification of Education
LS Low skilled
MS Medium skilled
NACE Statistical Classification of Economic Activities in the European Community OLS Ordinary Least Squares
SDA Structural decomposition analysis VIF Variance inflation factor
1
1. Introduction
Economic activities in the contemporary era of globalization are increasingly structured around global supply chains (GSCs) that account for a rising share of international trade, global production and employment. Firms no longer simply make products and export them, but participate in highly complex cross-border arrangements that involve a wide array of part-ners, customers and suppliers (Gereffi, 2006). This “unbundling” (Baldwin 2006, p. 7) of ac-tivities into vertically fragmented production processes has significant implications on how developing country firms, producers and workers are integrated in the global economy (Kaplinsky, 2000; Gereffi & Fernandez-Stark, 2011). However, empirical evidence on the effects of international production fragmentation on the employment structure in less devel-oped countries is mixed and often contradictory (Amiti & Freund, 2007; Campos-Vázquez et al., 2011). While trade statistics reveal that emerging countries become increasingly involved in exports of high-skilled labor intensive goods, recent studies have demonstrated that such countries often specialize in low-skilled labor intensive activities within a broad range of sec-tors (Dedrick et al., 2009). Hence, a natural starting point for the analysis of the trade-employment nexus in emerging countries is to look at the spatial distribution of skill types within international production networks. Therefore, this paper aims at examining the effects of international production fragmentation in the context of GSCs for the evolution of em-ployment structures in emerging countries. Two main questions are tackled: First, how has the job distribution within GSCs changed over time, both in terms of geography (with a focus on employment effects in emerging economies) and in terms of skill levels? Second, which fac-tors determine the prospects of emerging economies to upgrade into production stages associ-ated with high-skilled labor-intensive activities within such chains? More precisely, do mac-roeconomic variables like wages, trade costs and infrastructure differently impact on the ex-tent to which low-skilled, medium-skilled and high-skilled jobs are captured by emerging countries?
In order to address these questions, the analysis takes a GSC perspective that is based on a two-step approach: First, a structural decomposition analysis is applied in order to compute the contributions of changes in trade patterns on the spatial organization of skill-specific tasks within GSCs over 14 years. The analysis is based on data provided by the recently released World Input Output Database (WIOD)1 which allows to trace back the skill-specific labor inputs of 17 emerging economies in the production process of final manufacturing products completed in 21 advanced countries. In this way, the study aims to analyze the integration of
1
2
emerging countries’ workers into international production networks, focusing on the dynam-ics in North-South trade in the presence of vertical specialization. Moreover, the decomposi-tion analysis focuses on the effects of changes in intermediate trade patterns and separates these effects from the forces of final demand and technological change that impact on the skill distribution within international production networks. Thereby, the motivations of production relocations along GSCs can be reflected more adequately.
Second, the results obtained in the first part of the analysis are integrated into a regres-sion framework in order to investigate the underlying determinants of employment changes in emerging economies that are caused by relocation decisions within GSCs. Based on the litera-ture on international trade, economic geography and politico-economic models, an explorato-ry framework is developed emphasizing the role of distance, wages as well as countexplorato-ry- country-specific characteristics of emerging countries. The strategy aims to shed some light on the underlying forces of the skill-upgrading prospects of emerging economies participating in GSCs in a quantitative manner that so far has not been discussed extensively in the literature.
The remainder of this paper is organized as follows: Section 2 gives an intuitive intro-duction into the chosen GSC perspective on employment changes as well as the decomposi-tion method applied and discusses the relevance of different variables for the employment structure of emerging economies in the context of international production networks. Section 3 introduces the methodology and data of the structural decomposition analysis and presents some stylized facts regarding the evolution of skill-level specific employment relocations within advanced countries’ GSCs. Subsequently, section 4 develops the empirical model on the determinants of employment changes, presents the regression results and discusses policy implications and limitations. Finally, section 5 provides concluding remarks.
2. Theoretical Background
2.1 A Global Supply Chain Perspective on Employment Changes in Emerging
Economies
3
In the literature, many different terms are used to describe the internationalization of production that can be summarized under the term international production fragmentation. Examples include ‘vertical specialization’, ‘slicing up the value chain’, ‘off-shoring’ or ‘in-ternational outsourcing’ (European Commission, 2012). The related literature covers a varie-ty of perspectives on international production fragmentation, ranging from the inter-firm link-ages and governance issues within production networks (Sturgeon et al., 2008), the role of institutions and policies for the structure of such networks (Gereffi, 1999), to the impact of cross-border fragmentation on trade patterns and the welfare prospects of participating coun-tries, both theoretically (Grossman & Rossi-Hansberg, 2008; Baldwin, 2011; Costinot et al., 2011) and empirically (Kimura et al., 2007; Martinez-Zarzoso et al., 2011; Johnson & Noguera 2012a). In the present analysis, attention is focused on the geographic dimension of internationally fragmented production processes by looking at the movement of production stages and related jobs between countries. As proposed by Jones and Kierzkowski (1990), international production fragmentation is therefore defined as the splitting of production pro-cesses into parts and activities that can be done in different countries, irrespective whether these tasks are relocated within or across the boundaries of a firm (which is referred to as ‘in-ternational outsourcing’ and ‘offshoring’, respectively). Akin to this definition is the concept of vertical specialization that describes a “vertical trading chain that stretches across many countries, with each country specializing in particular stages of a good’s production se-quence” (Hummels et al. 2001, p. 76). Importantly though, the present analysis does not ac-count for the ownership structure and distribution of power within international production networks. Yet, in the presence of substantial FDI flows between countries, part of the capital in a country’s territory is owned by firms headquartered in other nations. In other words, giv-en that the analysis takes a domestic rather than a national perspective, conclusions about where the returns to capital finally end up cannot be drawn (Timmer et al., 2012a).2
To emphasize the spatial distribution of a production process from the sourcing of raw materials over the manufacturing of parts and the provision of supporting services to the com-pletion of the final product and the delivery to the end consumer, the present paper makes use of the concept of global supply chains (Costinot et al., 2011). Even though the chain metaphor is purposely simplistic, it focuses attention on the “location of work and linkages between tasks as a single product or service makes its way from conception to end use” (Sturgeon et
4
al., 2008, p. 299) and thus emphasizes the patterns of vertical specialization of countries and industries in specific skill-type activities.
The logic behind the GSC perspective as applied in this study is illustrated using an example of the U.S. automotive industry (Sturgeon et al., 2008). The output of the U.S. transport equipment industry relates to all final products that are sold to foreign and domestic end users.3 Accordingly the U.S. automotive GSC relates to the whole range of activities in-volved in the design, production and distribution of a final vehicle, which, for brevity, will be referred to as “U.S. cars” henceforth. However, this labeling does not necessarily imply that a “U.S. car” predominantly consists of U.S. capital and labor inputs, neither that it is associated with a U.S. brand. Rather, the labeling is chosen to emphasize that the last stage of production of the final vehicle is geographically located on U.S. territory. Hence, a U.S. car could be a Chevrolet as well as a Toyota as long as the vehicle is finished in the U.S. before being sold to end users at home or abroad. Along the U.S. automotive chain, some activities will be carried out by the domestic transport equipment industry itself, inducing the employment of workers of different skill types in that industry. However, the production of U.S. cars also requires the supply of various parts and components such as engines and braking systems as well as relat-ed services like R&D, design, logistics and financial services that are providrelat-ed by upstream industries. As these industries also use labor inputs, changes in the demand for U.S. cars will induce employment changes in various interdependent industries that can be at home or abroad through the import of intermediate inputs.
Sturgeon et al. (2008) emphasize the distinctive structure of the global automotive in-dustry that has witnessed profound changes during the last two decades shifting from a series of discrete national industries to a more integrated global industry. The economic geography of the global automotive chain is characterized by a dispersion of final assembly activities close to end markets that corresponds to political pressures to ‘build where they sell’ (Sturgeon et al., 2008, p. 303), while trade in parts and components is organized mainly re-gionally, with bulky and model-specific parts production concentrated close to final assembly plants and lighter, more generic components relocated to more distant locations to take ad-vantage of scale economies and low labor costs. To illustrate the dynamics observed in the economic geography of the U.S. automotive chain, figure 1 and table 1 display the develop-ment of labor input shares in the production process of U.S. cars between 1995 and 2008 by different country groups and three skill levels, namely low-skilled, medium-skilled and
skilled labor (in the following LS, MS, and HS).
in this example almost doubled their labor inputs from 23 to 42 percent of total hours worked along the U.S. automotive GSC emphasizing the accentuated process of international produ tion fragmentation in manufactu
by advanced countries decreased to less than 50 percent. However, the decomposition of labor inputs into different skill levels reveals that emerging countries mostly gained in terms LS and MS intensive production stages (table 1)
for two third of all LS labor inputs employed along the U.S. automotive GSC, almost one third of MS hours worked, but only 15 percent of HS labor inputs. In contrast, a
tries (including the U.S.) specialized in HS intensive production stages and lost substantial LS labor inputs shares. Hence, even though the globalization of the U.S. automotive industry has induced the rapid integration of less developed count
tion seems to be related to a “magnification of comparative advantages”
2012b, p. 38). On the other hand, the uneven distribution of labor shares between Latin ican countries and China supports th
Figure 1: Employment distribution within the U.S. automotive GSC, by country group.
Note: ADV refers to the advanced countries included in WIOD, including the U.S (see table A1 for a list of countries).
refers to China and LA relates to Brazil and Mexico.
Hungary, Indonesia, India, Latvia, Lithuania, Mexico, Poland, Romania, Russia and Turkey which are all covered by the WIOD. ROW relates to the rest of the world that is not explicitly modeled in the WIOD plus Malta and Cyprus.
Source: WIOD, own calculations.
4 As stated before, the labor shares presented for the U.S. automotive GSC in table 1 and labor inputs of those products for which the last stage of produc
5
The numbers presented in this section were o
in more detail in section 3.2. Intuitively, the labor inputs that are required final U.S. vehicle were traced backed to their country
6 Skill groups are classified based on educational attainment levels according to ISCED (see section 3.3 for a more detailed description of the data used in this section).
0 0.2 0.4 0.6 0.8 1 % o f to ta l h o u rs w o rk ed ADV
skilled labor (in the following LS, MS, and HS).4 Overall, the emerging economies included in this example almost doubled their labor inputs from 23 to 42 percent of total hours worked along the U.S. automotive GSC emphasizing the accentuated process of international produ tion fragmentation in manufactures (fig. 1). 5 In contrast, the share of labor inputs provided by advanced countries decreased to less than 50 percent. However, the decomposition of labor inputs into different skill levels reveals that emerging countries mostly gained in terms LS and S intensive production stages (table 1)6. Comprehensively, emerging economies accounted for two third of all LS labor inputs employed along the U.S. automotive GSC, almost one third of MS hours worked, but only 15 percent of HS labor inputs. In contrast, a
tries (including the U.S.) specialized in HS intensive production stages and lost substantial LS labor inputs shares. Hence, even though the globalization of the U.S. automotive industry has induced the rapid integration of less developed countries, international production fragment tion seems to be related to a “magnification of comparative advantages”
. On the other hand, the uneven distribution of labor shares between Latin ican countries and China supports the analysis of Sturgeon et al. (2008). Car manufactures are
Employment distribution within the U.S. automotive GSC, by country group.
refers to the advanced countries included in WIOD, including the U.S (see table A1 for a list of countries). relates to Brazil and Mexico. Other emerging countries include Bulgaria, Czech Republic, Estonia, ndia, Latvia, Lithuania, Mexico, Poland, Romania, Russia and Turkey which are all covered by the relates to the rest of the world that is not explicitly modeled in the WIOD plus Malta and Cyprus.
stated before, the labor shares presented for the U.S. automotive GSC in table 1 and figure 1 reflect only the inputs of those products for which the last stage of production is located within the U.S.
presented in this section were obtained by applying an input-output analysis that will be described in more detail in section 3.2. Intuitively, the labor inputs that are required from conception to the completion of a final U.S. vehicle were traced backed to their country-origin using international input-output tables.
Skill groups are classified based on educational attainment levels according to ISCED (see section 3.3 for a more detailed description of the data used in this section).
1995 2008
ADV CHN LA Other emerging ROW
5
Overall, the emerging economies included in this example almost doubled their labor inputs from 23 to 42 percent of total hours worked along the U.S. automotive GSC emphasizing the accentuated process of international
produc-In contrast, the share of labor inputs provided by advanced countries decreased to less than 50 percent. However, the decomposition of labor inputs into different skill levels reveals that emerging countries mostly gained in terms LS and . Comprehensively, emerging economies accounted for two third of all LS labor inputs employed along the U.S. automotive GSC, almost one third of MS hours worked, but only 15 percent of HS labor inputs. In contrast, advanced coun-tries (including the U.S.) specialized in HS intensive production stages and lost substantial LS labor inputs shares. Hence, even though the globalization of the U.S. automotive industry has ries, international production fragmenta-tion seems to be related to a “magnificafragmenta-tion of comparative advantages” (Timmer et al.,
. On the other hand, the uneven distribution of labor shares between Latin Amer-e analysis of SturgAmer-eon Amer-et al. (2008). Car manufacturAmer-es arAmer-e
Employment distribution within the U.S. automotive GSC, by country group.
refers to the advanced countries included in WIOD, including the U.S (see table A1 for a list of countries). CHN countries include Bulgaria, Czech Republic, Estonia, ndia, Latvia, Lithuania, Mexico, Poland, Romania, Russia and Turkey which are all covered by the relates to the rest of the world that is not explicitly modeled in the WIOD plus Malta and Cyprus.
figure 1 reflect only the tion is located within the U.S..
output analysis that will be described conception to the completion of a
output tables.
6
politically forced to provide vehicles that have a high local content and therefore se-lect their first tier suppliers mainly at a na-tional level. This explains why producers in the U.S. have not concentrated their pro-duction in Mexico, despite lower operation-al costs and the existence of a free trade agreement. In contrast, suppliers with a global presence have an incentive to source specific parts and components that can be easily traded across long distances from low-cost countries and ship them to plants where modules and sub-systems are built up. This tendency in turn can explain the rapid integration of countries with a huge labor surplus like China within the U.S. automotive chain. Moreover, the strong increase of hours worked by highly-skilled Chinese workers suggests that the reloca-tion of generic parts and component pro-duction was at least in part accompanied by more sophisticated supportive activities.
The changes observed in the employment structure along the U.S. automotive chain might have different sources though (Los et al., 2012). First, changes in the overall employ-ment within the U.S. automobile GSC might be induced by changes in the final demand for U.S. cars. Variations in the final demand levels for U.S. automobiles lead to proportional changes of employment at all stages of production and hence do not affect the relative distri-bution of skill-type jobs across country-industries associated with that specific GSC (so-called “size effects”).7 Second, technological progress is likely to affect the labor requirements per unit of output. Improvements in technology will therefore reduce skill-specific labor inputs at all stages of production. However, technological progress might not be neutral across skill
7 Yet, to the extent to which final demand shifts from one GSC to another within the same industry (e.g. from cars completed in the U.S. to cars completed in Germany), the skill-structure of a country that is engaged in both GSCs might change in a not-neutral manner given that different GSCs employ different skill intensities (so called mixed effects, see Los et al. (2012).
Table 1: Employment distribution along the U.S. au-tomotive GSC, by country group and skill type.
skill
type country group
Share in skill-specific hours worked % change in hours worked 1995 2008 LS ADV 0.36 0.17 -45% CHN 0.27 0.49 119% LA 0.09 0.07 -6% Other Emerging 0.11 0.10 6% ROW 0.16 0.17 24% MS ADV 0.83 0.64 -28% CHN 0.07 0.20 160% LA 0.04 0.07 64% Other Emerging 0.04 0.06 29% ROW 0.02 0.04 92% HS ADV 0.92 0.79 -14% CHN 0.02 0.10 437% LA 0.03 0.04 30% Other Emerging 0.02 0.05 120% ROW 0.02 0.03 85%
Note: Country groups defined as in figure 1.
Source: WIOD, own calculations, based on the input-output
7
types to the extent that the reduction in jobs can be greater for one skill group than for anoth-er, thereby producing shifts of the skill distribution within that GSC and impacting on the skill content provided by individual countries. For instance, assuming that Brazil contributes main-ly MS labor intensive inputs to the U.S. automotive chain and that MS labor requirements within the chain would decline due to computerization, the relative skill content contributed by Brazil to U.S. cars would change as employment of MS workers would be reduced com-pared to HS and LS labor inputs. Similarly, the skill content of the GSC as a whole would be affected. Third, changes in the employment shares captured by various country-industries can also be affected by changes in country-specific labor productivities (with a productivity catch-up reducing the labor requirements in that country).
Finally, the relocation of employment generating activities across countries will change the labor requirements per unit of total output provided by a specific country-industry. Such ‘within-effects’ are not necessarily neutral across countries, industries and skill groups and appear from changes in the specialization patterns of different countries within a GSC. For example, the entrance of a new country, say China, in the production process of U.S. ve-hicles is likely to reduce the employment shares of other participants within that GSC. On the other hand, under the assumption that China captures the production of seat upholstery which was done in Mexico in the first place, relocating these LS labor-intensive activities from Mex-ico to China might induce the former to upgrade into more sophisticated stages such as sup-portive services or more capital-intensive parts production, thereby increasing the overall skill content of employment that is provided by Mexico (Gereffi, 1999). On the other hand, pre-suming that seat production in China requires a larger amount of LS labor per unit of output than in Mexico, the relocation of these activities reduces the overall skill intensity of the final vehicle. Explaining how such trade-related effects of international production fragmentation have changed the employment structure of participating countries is the main aim of this pa-per.
ad-8
vanced countries. However, this notion is not always correct. In some industries, especially electronics, firms frequently keep intermediate production stages like software development at home while off-shoring final assembly activities to low-cost countries which then go directly to the end user (Dedrick et al., 2009). Unfortunately, these aspects of international fragmenta-tion cannot be captured based on the informafragmenta-tion given in internafragmenta-tional input-output tables. As stated before, input-output tables provide information about the location of production inputs in different country-industries and hence do not inform about the ownership structure of pro-duction (i.e. to which firms and countries value added is actually ascribed). Therefore, it is not possible to disentangle whether the assembly activities of a country-industry, say the Chinese electronical equipment sector, are carried out for domestic electronic firms or for foreign firms due to off-shoring. In both cases, the assembled final product will be attributed to the Chinese electronic industry. In this sense it is not possible to isolate those parts of the produc-tion of an industry that are the result of off-shoring from advanced to emerging countries. However, in many industries the ultimate production stages are still attributed to advanced countries and thus can be analyzed using input-output analysis.
Furthermore, the analysis will concentrate on the employment changes in the produc-tion process of final manufacturing goods (Los et al., 2013). The reason is that the producproduc-tion processes of manufacturing products are much more fragmented internationally than the sup-ply chains for agricultural goods and services (Timmer et al., 2012b). However, changes in manufacturing GSCs will not only affect labor demands within the manufacturing sectors, but also in service industries and raw materials production. The labor inputs associated with these indirect activities will be explicitly taken into account through the modeling of input-output linkages across sectors.
9
2.2 Determinants of the Distribution of Skills along Global Supply Chains
A theoretical framework of international production fragmentation was first proposed by Jones and Kierzkowski (1990). In their framework, production processes are geographically dispersed into different blocks connected by service links. Such vertical specialization can be explained based on the Heckscher-Ohlin model of international trade, according to which more labor intensive stages of production are located in labor abundant, low wage countries, while more capital intensive stages take place in capital abundant countries. Alternatively, producitivity differences according to Ricardian comparative advantage can favor the geographical separation of production, according to which labor skills in some countries will be more appropriate for one stage of production, while labor skills of other countries are more suited with respect to another fragment of the production process. In both cases, the dispersion of activities is aided by the possible existence of economies of scale wihtin the production blocks. In this way, a country does not necessarily have a comparative advantage in every stage of production, and different country-specific advantages can be combined by firms through vertical specialization (Zeddis, 2011). However, vertical specialization is not without costs. Rather, the production in different locations requires substantial coordination needs that are associated with transportation, communication and insurance costs. Hence, fragmentation alters the trade-off between such fixed ‘service link’ costs (Jones & Kierzkowski, 1990, p. 31) necessary to join production blocks and lower marginal costs of output that are obtained by relocating production stages according to the principles of comparative advantage. In the same line of argument, theoretical models of trade in tasks emphasize the relevance of reductions in transportation and coordination costs for the spatial despersion of production stages (Grossman & Rossi-Hansberg, 2008; Robert-Nicoud, 2008).
10
“because of less skilled workers, worse infrastructure, or inferior contractual enforcement, both costly defects and delays in production are more likely in some countries than in others” (Costinot et al., 2011, p.2). In this framework, a higher probability of mistakes results in a lower overall factor productivity of countries, leading them to produce and export at earlier stages where the inputs are less capital-intensive and mistakes are less costly.8 Following the-se theoretical considerations, the investigation on the determinants of the skill distribution along a GSC concentrates on four broad categories of variables that are derived from different theoretical models:
i. Trade and communication costs between the CoE and the CoC (gravity models). ii. Factor endowments of the CoE (Heckscher-Ohlin model of comparative advantage). iii. The broad infrastructure of the CoE (Economic geography and politico-economic
mod-els).
iv. GSC-specific characteristics and controls.
Trade and Communication Costs
Following the insights from the gravity literature, the effect of distance is expected to influ-ence the relocation of different types of jobs depending on the characteristics of these tasks. According to the distinction of activities into routine and non-routine tasks, the literature on economic geography suggests that routine tasks can be relocated easily, whereas activities requiring face-to-face contact and complex interactions are likely to be clustered into a single geographic location (Storper & Venables, 2004; Grossman & Rossi-Hansberg, 2008). There-fore, geographic distance is likely to be a greater obstacle for the relocation of HS employ-ment generating activities that are associated with complex interactions than for shifting rou-tine tasks within a GSC. Similarly, improvements in information and communication technol-ogies (ICT) are expected to reduce the coordination costs for tasks associated with non-codifiable information by creating substitutes for spatial proximity and hence leading to the geographic separation of these activities (Leamer & Storper, 2001).9 In addition, trade agree-ments are also expected to play a stimulating role for the spatial organization of employment within a GSC as has been described for the global apparel chain by Gereffi (1999) and found
11
by Johnson and Noguera (2012b) for bilateral trade flows in value added. In fact, free trade agreements are likely to have a more pronounced influence in international production networks than for trade in final products because of the multiple border-crossing that takes place along a GSC (Athukorala & Nobuaki, 2006).
However, when considering the impact of distance for the distribution of labor along GSCs, one has to distinguish between the bilateral distance of the CoE and the country where the final product is completed and the multilateral geographic mapping of the whole chain. In internationally fragmented production networks, intermediates might cross multiple borders before they end up in the final product. Therefore, production relocations might take place among first, second and further tiers of suppliers as well. In this sense, the geographic posi-tion of a potential CoE relative to all other economic actors within the network is of crucial importance to understand the extent to which certain countries capture productive activities10. However, since an intermediate input provided by a given CoE will eventually end up in a certain country-industry of completion, the geographic proximity between these two countries is assumed to be of special relevance for employment shifts along GSCs.
Factor Endowments of the Country of Employment
One of the most important factors stressed by international trade theories to explain vertical specialization relates to differences in country-specific factor endowments. In the presence of international production fragmentation, the Heckscher-Ohlin theorem predicts that countries will specialize in those activities which local value added is relatively intensive in their rela-tively abundant factor (Timmer et al., 2012b). Accordingly, firms in advanced economies have an incentive to relocate production stages to countries that are different with respect to their relative factor endowments and hence differ in factor prices. Usually, HS abundant coun-tries will re-locate LS labor intensive production stages to emerging councoun-tries with a relative abundance of LS workers to exploit factor price differences and reduce overall production costs (Feenstra & Hanson, 1997). Therefore, skill-specific wages and low prices of natural resources are considered as major determinants for employment shifts within international production networks. However, as the wage gap between typical assembly hubs like China and industrial countries is narrowing while the competition among developing countries in-creases through the emergence of even lower-cost countries such as Vietnam, other location-related variables become increasingly relevant. Therefore, the host country environment can be argued to be an additional factor determining the decision were to locate production stages.
12
The Broad Infrastructure of the Country of Employment
Key aspects of potential countries of employment are their economic structure and market size. The literature on economic geography suggests that monopolistic competition and econ-omies of scale motivate intra-industry trade (Krugman, 1991). This reasoning also relates to GSCs. With increasing economies of scale, higher output levels of suppliers are likely to re-duce unit costs in international production networks. In addition, fragmentation is always connected to the search for potential partners abroad. Thus, firms seeking for production relo-cations along GSCs should concentrate their search on larger markets since it is more likely to find suppliers with the appropriate competencies and skills that meet the final producers’ standards (Zeddis, 2011). Taken together, a dense industrial structure and larger market size may result in agglomeration effects and network externalities that are particularly important for the provision of HS intensive production inputs.
Besides, studies with a politico-economic focus propose that the quality of the physical infrastructure of a potential CoE is an important determinant for the spatial distribution of labor demands along GSCs (Egger & Egger, 2005). Egger and Falkinger (2006) analyze the role of public infrastructural expenditures on the attraction of international outsourcing and find that “in the context of vertical fragmentation, governments can use public infrastructure provision as a policy instrument to attract a higher number of intermediate input producers and […] to increase their attractiveness as a target for foreign outsourcing” (p.1994). While investments in the physical infrastructure reduce transportation and fixed costs associated with production relocations in general, improvements in the educational system are expected to favor more strongly the relocation of more sophisticated activities as such progress increas-es the general competencincreas-es and skills of the workforce of a potential CoE.
off-13
shoring firms might have an incentive to favor countries with low levels of contract enforce-ment and weak labor protection, providing better opportunities to realize cost reductions through exploitative agreements.
GSC-characteristics
Finally, the effects of changes in the reallocation of productive activities along GSCs on the employment shares captured by participating countries can be expected to be specific for dif-ferent final product categories. Importantly, final product categories differ in their degree of vertical fragmentation. While the production processes of transport equipment and electronic products are most internationally fragmented, other final products like food or non-metallic minerals display a much lower degree of vertical specialization in their production process (Los et al., 2013). As mentioned before, such GSC-specific differences in the degree of frag-mentation can be due to protectionist policies (e.g. in the automobile industry national policy-makers force car manufacturers to include high levels of local content, thereby reducing the opportunities for international production sharing) or to different prospects for the realization of economies of scale in the production of certain inputs (Jones & Kierzkowski, 1990). More-over, the degree of international fragmentation of a final product (that in turn drives the cross-border division of labor) depends on the relevance of transport costs at each production stage. Products with low weight-to-value ratios like electronics can be shipped at lower relative costs than bulky goods like coal or cement (Hummels, 2007). At the same time, products with a low weight-to-value ratio can be expected to embody a higher skill sophistication since HS workers usually contribute more value to a good or service than LS workers. Yet, it is essen-tial to understand that within international production networks, transport costs do not (only) manifest at the level of the final product, but also at the level of parts and components. Hence, even for bulky final products like cars or machinery there might be generic parts and compo-nents that have a low weight and/or high skill intensity like the electronic equipment of a car. In this way, low weight-to-value ratios of specific components create the incentives for firms to relocate these activities to the most attractive production locations. Still, the overall skill intensity of a final product and its respective value roughly indicate to which extent the whole production process could be potentially split up.
14
these relocations. For instance, the introduction of minimum wages in the U.S. is likely to increase the fragmentation of U.S. GSCs in general but will not determine whether the re-leased activities are relocated to Mexico rather than to China.
The empirical analysis that is be carried out in section 4 investigates the impact of the different categories of factors discussed on skill-specific employment changes in emerging economies that can be attributed to relocations within advanced countries’ manufacturing GSCs. Though it is not a rigorous test of the underlying theoretical models, the analysis aims to give first insights into the relevance of different groups of variables for employment chang-es along GSCs and to provide valid policy recommendations for emerging countrichang-es’ govern-ments. Yet, before moving to the analysis of the determinants of employment changes, such changes need to be identified first. For this purpose, the next section introduces an input-output framework that allows to trace back the input content of final products to their country-industry origin and separates the effects of intermediate trade from the forces of final demand and technological change within a GSCs by applying a structural decomposition analysis.
3. The Impact of Production Fragmentation on Employment in Emerging
Economies – Input-Output Analysis
In order to analyze the consequences of relocating employment generating activities within advanced countries’ manufacturing GSCs on workers in emerging economies, this section provides a structural decomposition analysis of the effects of intermediate trade within global production networks based on the assessment of international input-output (IO) tables. To this end, the benefits of this approach are shortly discussed next, followed by a description of the terminology and analytical framework. Then, some trends concerning the participation of emerging economies in advanced countries’ GSCs will be presented.
3.1 Empirical Approach and Assumptions
15
industries that join a GSC use domestic upstream products, thereby creating labor demand in related industries. Moreover, the approach is supply-sided, whereas the focus here is to trace final products to their origin in terms of employment, which is essentially demand-side ori-ented. An alternative approach is to analyze the sophistication of exports from developing countries. For many emerging countries it reveals that the skill content of exports has substan-tially increased (Krugman, 2008; Hanson, 2012). Yet, gross trade statistics are misleading if one does not account for the imported inputs that are embodied in these exports (Hummels et al., 2001). A closer look shows that most of the exports of developing countries are concen-trated in unskilled labor-intensive activities within a wide range of industries while the major part of value that is added to these products takes place outside the country (Johnson & Noguera, 2012a). In contrast, the IO approach chosen in this paper makes the link between international fragmentation and employment explicit by tracing the intermediate input content of final products. Moreover, by examining the consequences of production fragmentation for the entire economy, the analysis provides a more comprehensive picture about how current global trade patterns affect the industrialization prospects of emerging countries. In contrast to single country studies, the application of structural decomposition analysis on world IO tables allows for cross-country comparisons of inter-temporal developments.
Before moving to the description of the analytical framework, some assumptions and limitations of the analysis are outlined. IO models are demand-driven since the activities car-ried out within a GSC are assumed to be determined by the final demand levels directed to the chain. For a given IO table production is assumed to be proportional in the sense that the technical input coefficients are constant. Hence, economies of scale in production or factor substitutions are not taken into account. However, the comparison of annual IO tables allows for highly flexible cost shares in the implied production functions capturing inter-temporal trends in country-industry linkages (Timmer et al., 2012b). Yet, when comparing IO tables over time, it is important that the values used are expressed in constant prices in order to ob-tain reliable results that are not driven by relative prices changes, for instance due to inflation, and to isolate the real quantity effects of interest.
16
Figure 2: Stylized word input-output table.
Intermediate use Final use Total
(M columns per country) (C columns per country)
1 … N 1 … N Intermediate deliveries M industries, country 1 . . … . .. . . .. . . M industries, country N . . Primary inputs Capital HS labor . MS labor LS labor Output ′ . ′ ′
Source: Adaptation of Los et al. (2013).
3.2 Analytical Framework
Assume that there are countries, sectors, final demand categories per country and primary production inputs, including capital and labor inputs distinguished by skill type. Time subscripts are left out at this stage for the ease of exposition and will be introduced later. Each country-sector produces one good and hence there are products. The output in each coun-try-sector is produced using domestic production factors and intermediate inputs that can be sourced domestically or from foreign suppliers. For a particular trade flow, is defined as the source country and as the destination country. Similarly, is the source sector and the destination sector. Moreover, output may be used to satisfy final demand or as an intermediate input in production. Final demand consists of household and government consumption and investment both at home and abroad. By definition, the value of a product produced in a par-ticular country-sector must equal the value of this product used domestically and abroad, as product market clearing is assumed (Miller & Blair, 2009). Using matrix algebra, the market clearing conditions for each of the goods can be combined to form a global input-output system capturing the interrelations between different country-industries and demand catego-ries. A graphical representation of a simplified world input-output table is given in figure 2. In matrix algebra, the relationship can be written as:11
= + (1)
In this expression, is the production vector of dimension 1 containing the output levels in each country-sector. is a 1 vector that contains aggregate final demand levels of the output from each country-sector. Moreover, is defined as the global intermediate input
17
efficient matrix of dimension . The typical element of describes the intermediate input share of sector in country necessary to produce one unit of output in sector in country . Note that the use of goods can be at home (in case = ) or abroad ( ≠ ), and intermediate deliveries might take place within sectors ( = ) or across sectors ( ≠ ). Re-arranging (1) yields:
= − (2)
is an identity matrix containing ones on the diagonal and zeros elsewhere. − is known as the Leontief inverse and will be referred to in the following as . The typical element !"#$ of this matrix gives the total production value of sector in country occurring at all stages of production to satisfy one unit of final demand for industry in coun-try . The intuition behind the Leontief inverse is that the production of intermediate inputs often requires intermediate inputs itself. Hence, captures all direct and indirect production inputs of a country-sector in the production process of a final product, or in other words, ac-counts for an infinite number of production rounds. Referring back to the example of the U.S. automotive GSC, this could be all intermediate inputs that flow from the Mexican machinery sector to the U.S. transport equipment industry, both as direct production inputs (e.g. for car construction at the final assembly line) or indirectly by providing machines to other sectors that in turn produce parts and components of the final vehicle. Since the purpose of this paper is to identify the labor employed in emerging economies that is directly and indirectly in-volved in the global production chain of a final good, expression (2) is extended according to Los et al. (2012):12
μ&' = ()′*+
, (3)
The diagonal matrix *+, contains the direct labor coefficients in the diagonal cells, defined as the labor requirements of skill type - per unit of gross output in each of the in-dustries in each of the countries. ()′ is a 1 selection-vector that has ones in the cells associated with the focal CoE and zeros otherwise. The scalar ./# therefore reflects the labor requirements of skill type - employed in country to satisfy the final demand levels in .
Next, four determinants of inter-temporal changes in ./# will be specified following Los et al. (2012) that relate to changes in final demand levels, productivity and technological progress as well as trade-induced reallocations of labor inputs. First, changes in the
12
18
gy of a global supply chain will affect the distribution of labor across countries. GSC
technol-ogy is defined as the quantities of labor of a particular skill type required per unit of final
de-mand for the GSC’s final product. This includes all labor involved, irrespective of the country of location or industry of employment. However, since productivity levels of workers are like-ly to differ across countries, an 1 productivity vector 1 is introduced which elements contain country-industry-specific labor productivities relative to U.S. levels.13 For example, while the labor productivity of Chinese workers is only 18 percent of U.S. workers, productiv-ity in Slovenia is about 60 percent of the U.S. levels. In other words, to produce one unit of output, China employs three times as much labor as Slovenia, holding all else equal. There-fore, the comparison of production processes across countries would produce a misleading picture with regard to overall labor requirements if one does not account for these productivi-ty differences. Consequently, GSC technology needs to be expressed in terms of labor meas-ured in efficiency units as follows:14
*,∗= 3 ∘ *, ′ (4)
The 1 vector *,∗ gives the worldwide inputs of skill category - necessary to produce one unit of each of the final products. If, for example, improvements in ICT lead to a substi-tution of LS labor by computers within a GSC, less workers of the LS group will be employed at all production stages of the chain. In the literature, this effect is known as skill-biased tech-nological change (Autor et al., 2003). In addition, the second driver of changes in the em-ployment structure along a GSC relates to improvements in country-specific labor
productivi-ty levels, 5#. If a country catches up in terms of productivity compared to the U.S., this will lead to relatively lower labor demands of all skill types in that country, ceteris paribus (Los et al., 2012).
Third, Los et al. (2012) stress that the values in the cells of matrix are not only de-termined by the technical production requirements in terms of intermediate input coefficients, but also by the shares of these intermediate inputs delivered by each of the potential country-sectors of origin. The allocation of intermediate inputs sourced from different country-country-sectors is affected by discrete firm choices. As a consequence, some country-industries will employ more labor of a given skill-type than expected on the basis of *,∗, while others will employ less. Therefore, the authors construct a labor share matrix 6, , reflecting the
13
While the elements of 3 change over time, it is assumed that they are identical across industries and skill-types within a country. This assumption had to be made due to the low quality of productivity data at the sectoral level for many emerging countries.
14 The symbol " ∘ " stands for cell-by-cell multiplication (i.e. W = Y ∘ X means that w
19
shares of each of the country-industries in the total employment of skill type - per unit of final demand for a GSC. Rows of 6, represent industries of employment, while columns indi-cate the GSC to which labor of skill-type - contributes:
6,= @3A*B C*+, ,∗ (5)
The typical element of 6,, /#$DE, reflects all direct and indirect labor inputs of skill-type - that are involved in sector in country in order to produce one unit of final demand for products of sector in country , relative to the total labor inputs of skill-type - required at all stages of production.15 This could, for example, relate to the total hours worked by LS workers in the Mexican machinery sector necessary to provide the corresponding parts to the U.S. transport equipment industry, relative to the total LS labor inputs required to produce one unit of final demand for U.S. vehicles.
Writing *+, = 3A 6, *+,∗ and substituting this into (3), the employment of skill-type - in period 0 in the focal country can be expressed as:
µ&G' = ()H 3AI 6,I *+)I∗ I (6)
However, not all GSCs as captured by are important in this exercise. Therefore, a special demand matrix +I∗ = +ĴL is chosen by multiplying + with a selection matrix ĴL that contains positive values in the diagonal cells associated with the manufacturing industries of advanced countries of completion, and zeros otherwise.
In order to introduce a dynamic perspective on employment, a 1 employment vector ∆NOP is finally obtained for each CoE and skill type - for which the typical element, ∆./Q#$, is the employment change caused by changes in a particular GSC for which a certain advanced country has the last stage of production of product category R. To calculate inter-temporal changes in NOP, the total changes in employment are decomposed into the changes of its various components. Following the structural decomposition analysis applied by Los et al. (2012), this yields:
15
20 ∆S,) = S,) − S,I) = ()H 3A 6, *+∗) +∗− ()H 3AI 6,I *+)I∗ +I∗= (7) ()′ T3A − 3A I U6, *+)∗ +∗+ (7a) ()H 3A I 6, − 6,I *+)∗ +∗+ (7b) ()H 3A I 6,IT*+)∗ − *+)I∗ U +∗+ (7c) ()H 3A I 6,I *+)I∗ +∗− +I∗ (7d)
Equations (7a-c) identify the three within determinants of employment changes as discussed in section 2, namely technological change, productivity improvements and production reloca-tions, while equation (7d) reflects the effect of changes in the demand for final products (size and mixed effects). The partial effects of each of determinant can be isolated by assuming that the other three partial effects are zero. In particular, equation (7b) indicates the employment effects on skill-group - in the focal country if only the shares of GSC production captured by the country had changed. For the purpose of this paper, this partial effect, i.e. the employment changes caused by a reorganization of activities along a certain GSC, is the core variable of interest and labeled as:
∆VWX,)= ()H 3AI 6, − 6,I *+)∗ +∗ (8)
The typical element of this 1 vector, ∆Y Z/Q#$, measures the absolute change in
employ-ment of skill type - in country contributed to GSC R for which country has the last stage of production caused by changes in the relocation of intermediate production shares.
Employ-ment changes are measured in million hours worked. For illustrative purposes, consider the example of the U.S. automotive chain again: If China captures some LS manufacturing activi-ties related to seat upholstery that previously were carried out in Mexico, this will affect LS labor inputs of China as well as of Mexico within the U.S. transport equipment GSC. Howev-er, the relocation of these activities might induce Mexico to upgrade into more sophisticated stages such as production services, thereby potentially causing employment changes for MS and HS workers as well.
21
effect of (7b), namely changes in the organization of a GSC, this average is computed as fol-lows leading to the final specification of employment changes in emerging countries:
∆[\]OP = ^_∗ `PH 1AI aO − aOI b+P∗ cd∗ − `PH 1A aO − aOI b+PI∗ cdI∗ (9)
3.3 Data from the World Input-Output Database
To implement the decomposition framework introduced above, data for each country is need-ed on gross output by industry ( )) and final goods shipments ( )) as well the global input-coefficient matrix ( ). Moreover, data on labor inputs by skill-type and industry is required (*)). This type of data is available from the recently released World Input-Output Database.16 Basically, a world input output table (WIOT) is a combination of national input-output tables and international trade statistics in which the use of products is broken down according to their origin. In particular, imported products are disaggregated by country-industry origin and are allocated to a use category, applying different import shares of products with regard to intermediate use, final consumption or investment (Timmer, 2012). The tables provide a time-series of world input-output tables from 1995 to 2009 containing data for 40 countries, includ-ing all EU 27 countries and 13 other major advanced and emerginclud-ing economies. It covers more than 85 per cent of the world GDP in 2008 (Timmer, 2012). The remaining non-covered part of the world is estimated as well, such that the decomposition of final output as given in equa-tion (1) is complete. The WIOD distinguishes 35 industries according to the NACE classifica-tion and 59 product groups.
The present decomposition analysis is applied to the period 1995-2008. The pre-crisis period has been chosen due to the possibility of structural breaks in the patterns of interna-tional fragmentation following the financial crisis (Bems et al., 2011; Los et al., 2013). The in-ternational IO tables provided by the WIOD are expressed in current prices. However, Wolff (1985) argues that structural decomposition analysis must distinguish between price and real quantity effects. This also applies to the present context since employment changes can be due to either changes in the quantity of hours worked that are relocated or to price changes affect-ing the labor requirements per US$ of output when the analysis based on current prices. Even though the labor data is expressed in quantitative terms (million hours worked by person en-gaged), the labor coefficients are calculated by dividing employment by total output in million US$. Therefore, potential price effects like an increase in relative prices due to inflation will bias the labor coefficient vector downwards. Similarly, all components of the SDA will be
22
affected by relative price changes. To isolate changes in quantities from price changes, Los et al. (2012) kindly provided a WIOT for 2008 that had been transformed into 1995 price levels. The countries included in the present analysis are grouped in two categories: countries
of completion (CoCs), that are characterized by high levels of development and countries of employment (CoEs) that are characterized by lower developmental levels. Countries of
com-pletions include the EU-15 countries that constituted the European Union prior to 200417 plus 6 further advanced economies, namely Canada, the United States, Japan, South Korea, Aus-tralia and Taiwan. On the other hand, countries of employment relate to the 10 Central and Eastern European member states (CEECs)18 plus the BRIC states (Brazil, Russia, India and China) as well as Mexico, Turkey and Indonesia. Appendix table A1 lists the countries in-cluded in each category and reports per capita GDP levels as an indicator of their develop-mental status. Moreover, the analysis will concentrate on the employment changes in the pro-duction process of final manufacturing goods. For this purpose, only the GSCs for final manu-facturing products defined as industries 15t16 to 36t37 (based on NACE rev 1) are included. Table A2 lists the industry classifications and indicates the manufacturing industries under consideration.
Finally, the complementary Socio-economic Accounts (SEA) of the WIOD contain data on the total hours worked by persons engaged and the shares of hours worked by differ-ent skill categories at the industry level that have been taken to compute the labor input coef-ficients bO. Labor skill types are classified on the basis of educational attainment levels (de-fined by ISCED). Low-skilled workers are those with primary education and/or lower second-ary education (ISCED 1 and 2), MS workers relate to those with secondsecond-ary education and/or post-secondary non-tertiary education completed (ISCED 3 and 4), while high-skilled workers are defined as having completed tertiary education (ISCED 5 and 6).
3.4 Stylized Facts
The analysis of international IO tables allows to trace back the inputs provided by emerging economies in the production process of final goods completed by advanced countries. Hence, the participation of developing countries’ workers in international production networks can be analyzed, also with regard to the skill content associated with the relocations within such net-works. In particular, the SDA applied in this paper allows to separate employment changes
17
The EU-15 countries include Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden and United Kingdom.
23
along GSCs that are induced by trade in intermediates from the forces of final demand and technological change. In this way, the focus of analysis is directed to the discrete choices within advanced countries’ manufacturing GSCs to relocate employment generating activities to the most attractive production locations.
Trade in intermediates has boomed in recent decades between advanced nations and emerging economies. Various studies have confirmed that Asian countries are strongly partic-ipating in global production networks, the same holding for Latin American and other less developed countries (Athukorala & Nobuaki, 2006; Johnson & Noguera, 2012a). Central and Eastern European transition countries are another interesting example for studying the conse-quences of international production fragmentation as these countries have become integrated into global, mostly EU-based networks of production quickly after the fall of the iron curtain (Kaminiski & Ng, 2005; Martinez-Zarzoso et al., 2011). Therefore, the present analysis looks at employment changes in this broad set of countries that can be defined as “emerging” or “transition” economies for the period from 1995-2008. Even though there might be differ-ences in the developmental status of the CoEs considered, they all share the common pattern to be intensively involved within advanced countries’ production networks (Kimura et al., 2007). This pattern is highlighted by table 2 that display the distribution of employment shares by GSC type. GSC type employment is obtained by aggregating industries of comple-tion over countries of complecomple-tion. Hence, the table aggregates all labor inputs that occur along the production process of a certain final product category, irrespective of the advanced coun-try in which its induscoun-try of completion is located and presents the shares of different councoun-try groups in the total amount of hours worked in the production process of that final good.
The table confirms that emerging economies have become increasingly integrated into GSCs over the last two decades. Advanced countries have lost employment shares in all man-ufacturing industries, whereas emerging economies and in particularly China gained substan-tially from off-shoring activities. The largest shifts in employment shares occurred in chemi-cals, machinery, electronic products and transport equipment. Here, China and the other emerging countries included in WIOD accounted for roughly a third of all labor inputs in 2008 and more than 40 percent in the case of electronics where labor shares almost doubled compared to 1995. These sectors are also important sources of employment generation due to their large final output values (European Commission, 2012).
