Economic Upgrading and Social Upgrading in Global Value Chains

(1)

__________________________________________________________

Master Thesis

Economic Upgrading and Social Upgrading

in Global Value Chains

- Insights from the Manufacturing Sector of Emerging Economies

_________________________________________________________________________

Marcel Kühne

July 7, 2014 Student Information:

Student Number: s1943324 (Groningen) Email address: m.kuehne@student.rug.nl

Thesis Supervisor: Bart Los, Ph.D.

Co-Assessor: Prof. Inmaculada Martínez-Zarzoso, Ph.D.

Degrees: Master of Science, International Economics and Business (Groningen) Master of Arts, International Economics (Göttingen)

(2)

Abstract

Purpose – This study provides a review of recent trends in international fragmentation of

production identified by the academic literature and assesses quantitatively to what extent economic upgrading within Global Value Chains (GVCs) can be associated with improvements in the social realm of working conditions.

Methodology – Using a GVC as analytical device, the final sample compiles cross-sectional data

of 17 emerging economies involved in manufacturing supply chains finalized in 23 advanced economies. The analysis is accompanied by means of a structural decomposition technique and a regression analysis in order to measure both quality and quantity of the upgrading relationship.

Findings – Three key results stand out. First, the decomposition analysis suggests a widespread

trend of successful economic upgrading, especially for CEE and BRIC economies. Second, even though a significantly positive channel between improvements in the economic and social sphere is detected, social downgrading might occur on the individual country level. Third, initial drivers, such as the level of education or unionism in a country, play an important role in enhancing the potential for social upgrading.

Originality – The paper is of value to both academics and practitioners working in the field of

international trade and production networks. The research of social upgrading is in its early stages and the paper introduces further work in this area by introducing new measures of quantity and quality of work.

(3)

ii

List of Acronyms

BICS Brazil, India, China and South Africa BRIC Brazil, Russia, India and China CEE Central and Eastern Europe FDI Foreign Direct Investment

GGDC Groningen Growth & Development Centre GDP Gross Domestic Product

GPN Global production network

GVC Global value chain

ICFTU International Confederation of Free Trade Unions ICT Information and Communication Technology ILO International Labour Organisation

OECD Organisation for Economic Cooperation and Development OEM Original Equipment Manufacturing

ODM Original Design Manufacturing OBM Original Brandname Manufacturing SDA Structural Decomposition Analysis

List of Tables and Figures

Tables

Table 1 – Export performance of Emerging Economies (1980-2008)

Table 2 – Descriptive Statistics

Table 3 – Summary of Hypotheses

Table 4 - Decomposition of Value Added Advanced Country GVCs

Table 5a – Economic Upgrading & Downgrading (overall economies, 1995-2007)

Table 5b – Social Upgrading & Downgrading (overall economies, 1995-2007)

Table 6 – Regression Output

Table 7 – Test for Multicollinearity

Table 8 – Test for Heteroskedasticity

(5)

iv

Figures

Figure 1a - Stylized Representation of an Internationally Fragmentated Value Chain Figure 1b – Specified Representation of GVC Relationship

Figure 2 – The Upgrading Framework

(6)

1

G

1. Introduction

lobalization describes the pervasive decline in barriers to the global flow of information, ideas, factors of production (labour and capital), technology and goods (Kaplinsky & Morris, 2001). Over the last two decades, the global economy has been increasingly structured around Global Value Chains (GVCs) that account for a rising share of international trade, national income and employment. Essentially, a value chain describes the full range of activities that firms and workers perform to bring a product from its conception to its end use and beyond. This includes activities, such as design, production, marketing, distribution and support to the final consumer. Thereby, the ‘global’ notion of value chains refers to a network of multiple firms across national borders and large geographic distances that split up the set of activities1. GVCs connect workers and consumers around the world and provide the stepping stone for supplier firms from developing countries to integrate in the global economy. Unleashed by significant declines in transport and communication costs, production networks have seen a massive wave of globalization over time and space2, recently, that generated remarkable economic and social outcomes (Gereffi et al., 2005). As a consequence, policy-makers in emerging economies have considered upgrading strategies as a way to gain access and increase the benefits from participation in GVCs (Bair & Gereffi, 2003). Thereby, the term ‘economic upgrading’ has become synonymous with economic development and describes the movement of firms into higher productivity activities in value chains measured by rises in value-added content of production.

Three motives have inspired this thesis. First, while much of the academic attention has exclusively centred around the relationship of international trade performance and economic upgrading, few surveys have highlighted the meaning of such upgrading for living standards, including wages, working conditions and rights, gender equality and economic security. In the following, I refer to improvements in per capita labour compensation and work standards over time thus as ‘social upgrading’. Moreover, evidence from primarily case studies hint to the

1

For an elaborate review of value chain concepts and research, see globalvaluechains.org.

2_{This process of disintegration of production across plant- or country-borders has been referred to as ‘international}

(7)

2

presence of substantial sectorial and industry-specific differences in upgrading experiences and provide a good starting point to whether or not the effect of fundamental factors on upgrading, such as education and trade union strength, varies across countries and industries. Third, economic upgrading is often simultaneously associated with improvements in the social sphere. Yet, this ‘conventional wisdom’3

is disapproved by an augmenting number of studies that describe the conditions of successful upgrading, while stressing the weak support for causality running from economic to social upgrading still persistent in the literature.

With a primary focus on emerging economies, this thesis provides a GVC perspective to the relationship between economic and social upgrading. Considering the gaps in the literature on upgrading in GVCs, it queries the role of global value chain participation for employees engaged in production networks with respect to changes value added and work conditions: i) How do cross-country differences in international fragmentation in production help to explain changing conditions in national labour markets? Furthermore, it is not clear how economic and social upgrading affect different groups of workers according to income and skills. Accordingly, ii) who are the winning and losing workers of upgrading skill-wise in the respective countries? And iii) what are the conditions under which economic upgrading translates into social upgrading? In order to start answering these questions, the analytical approach consists of two parts. First, I will employ a structural decomposition technique in order to determine economic and social upgrading. induced by re-allocation of production activities within GVCs. In this context, an indicator of value added is used as proxy for economic gains, while changes in labour compensation measure gains in the social sphere. Thereby, particular attention is paid to value chains of manufactures that have seen an unprecedented process of fragmentation of production processes compared to primary goods and services. For notation, country-sectors that generate intermediate inputs will be denoted Countries-of-Supply (CoS), while Countries-of-Completion (CoC) denote locations for the last stage of the production process (akin to Los et al., 2012). For means of illustration, Figure 1 provides a simple depiction of the way CoS and CoC are interconnected4. Each country commands over capital and labour endowments that are used to add value via domestic activities to an intermediate or final good in the value chain. While countries 1 and 2 countries produce and process intermediate goods, country 3 bundles all

3_{Term dubbed by Kucera (2006) with respect to the tendency of scholars to assume a parallel occurrence of}

economic and social improvements.

(8)

3

intermediate goods and own factor endowments in order to assemble a final product that is ready for consumption domestically or abroad. In this study, emerging countries, e.g. China and Mexico, are assumed to represent the former two countries, whereas advanced countries, like the US or Germany, take on the role of the last stage of production.

Figure 1a Stylized Representation of an Internationally Fragmented Value Chain

Using a structural decomposition technique, I decompose changes in value added and labour compensation in emerging economies, characterized by skill type and industry of employment, into changes induced by re-allocation of activities particularly. Building upon a world input-output model, industry-level data are retrieved from the recently released World Input-Output Database (WIOD) that covers 27 EU countries plus 13 other major emerging economies around the globe for the time horizon 1995 to 2011.

The remainder of this report is structured as follows: the second section surveys relevant academic literature comprising theoretical and empirical research and methods used to measure economic and social upgrading, while section 3 to 5 cover the methodology, theoretical model and data employed in this study, respectively. Section 6 then presents the main findings, before robustness checks verify the validity of the results. Finally, section 8 summarizes the key lessons learnt, discusses research limitations and gives suggestions for future research.

Country 1

Country-of-Supply

Country 2

Country-of-Supply

Country 3

Country-of-Completion Domestic Capital & Labour

Domestic intermediate production for export

Domestic Capital & Labour

Domestic intermediate production

Domestic intermediate production Final good for domestic and foreign consumption Intermediate good

for export

(9)

4

2. Literature Review and Hypotheses Development

Concepts and theories are necessary to organize scientific facts. This section provides therefore a survey of theoretical and empirical literature studying the economic and social outcomes for countries’ participation in GVCs. Thereby, sub-section 2.1 highlights the rise of international trade since the 1980s between North and South5. Subsequently, the emergence of trade in intermediate goods and service and the connected emergence of Global Value Chains is discussed in sub-section 2.2. Finally, I present a framework for upgrading that illustrates the ways emerging economies can potentially gain from participation in GVCs of advanced countries without the need to master the entire production process themselves.

2.1 Recent trends in International Trade

Beginning in the 1980s, international trade has seen an unprecedented surge of export growth. Over this period, world trade expanded more than eightfold in real terms, recording almost US$16 trillion in 2008. Ongoing trade integration across North and South also brought about structural changes. Table 1 shows the export performance of selected emerging economies (and the United States as benchmark) for the pre-crisis period 1980-2008. While developing countries accounted for about one third of global exports in 1980, the table shows that their chunk in international exchange of goods and services escalated quickly to more than 40 percent, on the costs of developed countries’ share. This trend is reflected by the average annual growth rate of developing countries in the 1990s and 2000s, where export growth outperformed developed country figures attaining staggering rates of respectively 10.1 and 15.9 percent – most notably, in China, Turkey and India. Recently, other nations have joined this distinct club of rapidly growing markets, commonly known as ‘emerging economies’ ( IMF, 2012)6

. Besides the prominent cases of BRIC7 economies, this group also includes countries from Central and Eastern Europe (CEE), South East Asia as well as Latin America. While the story of exports is well known, the import side gives another important indication about the involvement of developing countries in international supply chains (Cattaneo et al., 2010). Constituting for just about 14 percent of parts and component trade in 1990, developing countries have steadily

5_{‘North’ and ‘South’ are used as synonyms for developed and developing countries, respectively.} 6

According to the IMF definition, the following countries are labelled ‘developing’ or ‘emerging economies’: Bulgaria, Brazil, China, Czech Republic, Estonia, Hungary, India, Indonesia, Latvia, Lithuania, Mexico, Poland, Romania, Russia, Slovak Republic, Slovenia and Turkey.

(10)

5

increased their share in world imports to a third in 2010. At the same time, the share of OECD

countries in imports fell from 86 to 64 percent (OECD, 2012). Furthermore, trade from participation in value chains has brought about wide-ranging benefits, such as employment creation, and larger scale production, more efficient geographical allocation of industrial activities, as well as a higher variety of intermediate products (Gereffi, 2006). Therefore, the next sub-section continues with more insights to the way workers and consumers across the globe become increasingly integrated and how participation in these trade networks can benefit emerging economies.

2.2 Global Value Chains (GVCs)

(11)

6

services that later serve for final consumption or exports. In this context, GVC analysis provides a framework of how production of goods and services disintegrates increasingly into global supply chains across countries that work and transfer intermediate inputs. In other words, production processes may occur as a result of input channelling within plants and unbundled movements between plants in different countries. Baldwin and Venables (2010) denote these types of production structures ‘spiders’ and ‘snakes’, respectively. In Sturgeon (2001, p.10), the latter is referred to as the ‘vertical sequence of events leading to the delivery, consumption and maintenance of goods and services’, where value is added via productive activity. This paper restricts its focus to the observation of sequential trade flows between country-industries due to limited firm-level data availability that would provide information about the internal plant production. In a study of international trade flows, Hummels and others (1998) show that Asian electronics trade increased by 900 percent between 1986 and 1995 due to ‘vertical specialization’, where countries specialize in particular stages of a good’s production. Moreover, the ongoing disintegration of production is also re-enforced by the fact that trade flows in final goods and services have been passed by trade in intermediates that are not initially consumed but further used in production of other goods and services. A recent OECD study concludes that trade in intermediate inputs takes place mostly among developed countries and represents 56 percent and 73 percent of overall trade flows in goods and services, respectively (Miroudot et al., 2009).

(12)

7

productivity effect of moving to low-cost locations, a relative-price effect that occurs due to a change in a country’s terms of trade, and a labour-supply effect that drives down wages for workers whose jobs are outsourced. Thereby, especially the last effect is of interest for this study, as it reflects changes for working conditions of labourers. Before turning to the social consequences of value chain participation, however, sub-section 2.3 reviews potential economic gains.

2.3 Economic Upgrading

(13)

8

upgrading from assembly towards original equipment manufacturing (OEM)8 represents the most commonly observable transformation. Indeed, most case study work in this field has paid special attention to functional upgrading with focus on chains in agricultural and manufacturing sectors, such as horticulture in Brazil (Selwyn, 2008), apparel and textile chains in Asia (Gereffi, 1999) and North America (Bair and Gereffi, 2003), as well as automotive chains in developing (Humphrey, 2003) and developed countries (Sturgeon et al., 2008). This strand of research delivers interesting insights to country-, industry- and chain-specific aspects of trade integration. At the same time, however, it does not allow us to compare the extent of upgrading between different industries in a country or even across countries. Another drawback of case studies is their focus on mostly success stories of upgrading, which indicates a selection bias that skews the possibility to generalize findings up to a cross-country setting (Milberg and Winkler, 2011). One of the reasons why case studies represent the main body of upgrading literature is linked to the serious lack of appropriate data for a large number of countries on the industry- and firm-level. Therefore, the thesis on-hand attempts to address the current niche in the research literature of upgrading by a compromise between industry- and country-level data in order to establish a cross-country chain perspective on recent developments in the economic and social spheres. Via structurally decomposing changes in value added and labour compensation accompanied by a regression analysis, I aim to answer questions connected to the winners and losers of the fragmentation process as well as the fundamental conditions under which economic upgrading translates into social upgrading. The next sub-section reviews these social consequences of GVC integration.

2.4 Social Upgrading

The fact that supply chains exert commercial pressure on management practices and employment standards is a relatively recent phenomenon. At the same though, this development has major consequences for our traditional understanding of employment relations and regulatory institutions (Marchington et al., 2005). So far, there has been relatively less focus on how national and local standards and institutions are responding to the fragmentation of production systems and supply chains, even though the national markets remain the main locus of institutional and regulatory labour market activity. While economic development and

(14)

9

participation in GVCs have typically been associated with ‘economic upgrading’, Milberg and Winkler (2011) stress the relevance of ‘social upgrading’ to this discussion – the amelioration of entitlements and rights of workers as social actors, which enhances the quality of their employment9. More specifically, the social sphere covers aspects relevant to workers and employees, such as improvements in wages, working conditions, rights, gender equality and economic security (Sen, 2000). The complex nature of interlinkages running between both concepts of upgrading makes it difficult to determine whether social upgrading automatically follows its economic counterpart. Essentially, there are two competing economic theories of wage determination with rather different implications for the link between economic and social gains. On one hand, proponents of neoclassical theory argue that labour demand, and thus wages are mainly driven by technology. Here, employees are considered solely as factor of production. Regarding the circumstances under which workers gain from GVC participation, Barrientos and others (2011) argue that access to better work might directly result from economic upgrading. However, in modern supply chains, quality standards and decent working conditions represent another crucial commercial driver for corporations. Besides, fellows of institutionalist theory argue that the relative bargaining power of employer and employee organs determines wage fixing. Here, the social realm of upgrading is mediated by labour market institutions, including trade unions and collective bargaining arrangements. Compared to neo-classical theory, social upgrading is thus detached from technological change and instead associated with institutional drivers. In order to account for this duality, Barrientos et al. propose a new framework that links

9_{At the same time, they consider employment growth not as an appropriate measure of social upgrading, since it}

(15)

10

economic and social upgrading in production networks. This theoretical framework is illustrated by figure 2 that shows four potential combinations of changes in the economic and social realms. Newly, the framework also suggests the possibility of downgrading or losses of GVC participation, besides upgrading outcomes. Supposing international competitiveness is captured by production costs, we can derive two paths of raising competitiveness: the so-called ‘high road’ and ‘low road’ of development10

. The former stands for upgrading by means of improving productivity, income, stronger labour relations and greater state-supported social protection rather than squeezing wages and profit margins, as well as weak protective measures, such as in the latter case. Due to a focus on workers involved in GVCs in the present paper, capital costs are left aside. Hence, high road growth can be simplified to an issue of both increasing wages, in the social realm, and raising labour productivity, in the economic realm. Thereby high-road or

low-road refers to gains or losses in the social sphere, while growth or decline represents

economic expansion or recession, respectively. High and low road of development vary with respect to the extent of capability building. Thereby, innovation is considered a key capability that ensures continuous improvement in product and process development. Placing emphasis of the ability to learn in the production process has thus implications for the innovation system as a whole. However, innovation in itself may not be adequate. Besides, the speed of innovation relative to competitors decides over the distribution of value added and market share. In this context, upgrading to new sectors and activities provides incentives for emerging country

suppliers to innovate, in the form of accruing extra-normal rents in international markets as well

10_{The high road / low road denotation is adapted from Gordon (1996), whereas the first distinction of strategies by}

(16)

11

as the different dynamic learning capabilities of industries and management functions. However economic upgrading may also be associated with social downgrading. That is, a low road upgrading strategy might lead to ‘immiserizing growth’ for the respective firm, sector or economy, where economic growth results in a country being worse off regarding standards of living than before upgrading (Kaplinsky & Morris, 2001). In theory, it is also possible that economic downgrading accompanies improvements in the social sphere, as well simultaneous downgrading of both realms. While social upgrading may be hard to quantify without any agreed-upon measures in empirical research, finding a proximate indicator, such as changes in wages and labour standards over time, still helps to determine how much upgrading has occurred across industries and countries and to design policies intended to create social upgrading relative to economic gains. Further information on the definition of social upgrading is provided in section 4. In an attempt to link both concepts of upgrading, economic upgrading is defined by measures of trade performance and social upgrading is denoted in terms of work characteristics. 2.4 Hypotheses – Linking economic and social upgrading

From the preceding discussion, the following three hypotheses are derived to link economic with social upgrading in cross-country supply chains. The first two hypotheses thereby concern the impact of economic gains via changes in value-added output, while the final hypothesis relates to other determinants of social upgrading and the improvement of work conditions.

(17)

12

investigation of how social upgrading can be attained for workers in GVCs is provided by Rossi (2013) who examines benefits for employees perceived from process, product and functional upgrading in the fast fashion model of the Moroccan garment industry. Interestingly, Rossi’s study reveals that functional upgrading generates social up- and downgrading simultaneously when accounting for different types of workers. The causal link, as suggested by neoclassical theory in section 2.4 seems thus to be less clear-cut than expected and turns out to be more conditional on the dimensions on analysis we examine. Another strand of research suggests that the relationship between economic and social upgrading may be negative. A rather tragic example for low-road decline where economic upgrading did not translate into social benefits represents the case of multiple suicides at a Foxconn factory in China, the world’s largest producer and exporter of electronical equipment since 2005 (Reuters, 2010). Foxconn, a Taiwanese contract manufacturer and with one million employees China’s largest manufacturer conducts assembly activities for western lead firms, such as Apple Microsoft and Google (Verité, 2010). After first allegations of ‘sweatshop’-type working conditions with forced overtime, insufficient safety measures and military-style management practices, 17 cases of suicide have been reported between 2007 and 2010 at Foxconn facilities (Fair Labour Association, 2010). In order to test whether the forecast by neo-classical theory of productivity growth is still valid, I expect that economic upgrading, denoted in changes of value added, has simultaneously generated social upgrading, measured in rising labour compensation and work standards in emerging economies between 1995 and 2007. Particular focus is thereby paid to production systems of manufactures, which are highly prone to international fragmentation due to a high degree of adoptability for activities (Timmer et al., 2013). That is, manufacturing activities can be undertaken in any country with relatively little variation in quality11. Moreover, value chains differ substantially with respect to the way they are governed by large organizations, so-called ‘lead firms’, that drive the restructuring of production by moving towards greater specialization in activities performed (Gereffi and Fernandez-Stark, 2011). Lead firms are considered the primary sources of material inputs, technology transfer, and knowledge in these organizational networks. Depending on their position in a network, a critical distinction is made between ‘producer-driven’ and ‘buyer-driven’ supply chains. The former describes chains with large

11_{However, this does not imply that manufacturing goods consist exclusively of intermediate manufacturers.}

(18)

13

producers that possess the leading role due to their capital and knowledge of techniques essential to the production process, such as General Motors and Toyota. On the contrary, buyer-driven chains are characterized by increased linkages between production, distribution, retailing and consumption by large corporate buyers who own no production – think of Tesco and Nike for instance. Due to an absence of appropriate firm- and chain-level data, however, a clear distinction of producer- and buyer-led chains is beyond the scope of this paper. Therefore, we will assume that lead firms, located in advanced countries, govern manufacturers value chains including supplier firms from emerging economies. Indeed, Humphrey and Schmitz (2002) argue that firms in developing countries have been aiming increasingly at participation of GVCs governed by lead firms situated in developed countries. For example, China became a world-leading exporter in the 1990s, most notably by mastering the dynamics of buyer-driven value chains and supplying a wide range of labour-intensive consumer products, such as apparel, footwear, toys, and sporting goods for advanced country firms (Gereffi, 1999). Accordingly, we can formulate the first hypothesis:

Hypothesis 1: Economic upgrading, measured as a rise in value added induced by demand of manufacturing GVCs of advanced countries-of-completion (CoCs), has translated into social upgrading by means of gains in labour compensation and work standards in emerging economies from 1995-2007.

(19)

14

change related to computerization of work. They conclude that there is surprisingly little evidence linking both economy-wide phenomena. Instead, Card and DiNardo suspect other factors, notably re-allocation of labour, to be driving education-related wage gaps. Moreover, research of international fragmentation of production has found similarly specific outcomes for different skill levels, although few studies have employed a structural decomposition approach as employed in this paper. Feenstra and Hanson (2001) argue that trade in intermediate inputs is a potentially important explanation for the increase in the wage gap between skilled and unskilled workers. They show that trade in inputs affects not only technological change but also labour demand. In both cases, demand shifts away from low-skilled activities, while high-skilled activities experience rising relative demand and wages. Thus, distinguishing whether the change in wages is due to international trade or technological change is fundamentally an empirical rather than a theoretical question.

In order to single out the effect of re-allocation of production on upgrading from effects of technological change, productivity growth and changes in demand patterns, I apply a structural decomposition technique, as used in Timmer and others (2013). Studying the effect of international fragmentation, Timmer et al. derive the foreign value-added content of production by means of Structural Decomposition Analysis (SDA) in order to estimate factor income distributions from the perspective of a product in international production networks. They find a strong shift towards value being added by both capital and high-skilled labour, and away from less-skilled labour. Moreover, in manufactures GVCs the income shares of capital and high-skilled workers have been increasing, while those of other labour skill-levels have been declining recently. Their findings suggest that the rising presence of production fragmentation has drastic consequences for the factor income distribution both across and within countries.

To investigate the outcome of fragmentation of production on skill-specific wage levels, the most commonly observable type of gains in the economic realm, functional upgrading, is employed12. To recap, this type of upgrading captures changes in activities performed by a supplier firm towards a higher value added combination of manufacture and services tasks. In case of the apparel sector, this shift from mere assembly tasks to supplying, for example, OEM has been described extensively in the literature (see for instance Bair & Gereffi, 2003). The base for functional upgrading constitutes a higher level of skill in the labour force, which in turn takes up

(20)

15

substantial investment in training as well as provision of higher wages for qualified employees. Rossi (2013) argues that market pressures force firms to maintain a dual workforce consisting of permanent, skilled workers and temporary, unskilled workers at the same time. This way, market participants are able to react more dynamically to changing demand patterns and to the dyad of providing high quality and labour standard compliance together with low cost production (Barrientos et al., 2011). Especially low-skilled workers are thereby employed on unsound terms, such as repetitions of short-term contracts. Accordingly, hypothesis 2 suspects an impact of functional upgrading, induced by re-allocating production activities, on changes in work conditions that depends on workers’ skill-levels:

Hypothesis 2: Functional upgrading, measured by changes in value-added output by

manufactures GVC in emerging economies, has induced relatively more benefits to high-skilled workers than other skill groups in emerging economies, with respect to labour compensation and measurable standards of work between 1995 and 2007.

Before moving on to a description of methods used to analyse GVC participation in section 3, the final hypothesis suspects social upgrading to be the outcome of an institutional bargaining process affected by the national environment. Industrial relations play an important role for the effective channelling of economic into social gains. The relevance of institutional arrangements that commit capital to benefit labour is highlighted by the capital-labour struggle (Navarro, 2014). This process describes the natural tendency of capital owners to reduce activities that represent a cost or that limit valorisations by undersupplying workers, in the absence of strong institutions. According to Selwyn (2012), workers’ ability to transform their structural power into associational power with the goal of extracting benefits from capital represents a core aspect of the link between economic and social upgrading. Thus, I expect the extent of trade union representation to be positively associated with cases of successful social upgrading:

Hypotheses 3: The strength of trade unions in emerging economies, as share of trade union

(21)

16

3. Upgrading in GVCs - Methodology

This section introduces the methodological approach employed for our investigation of upgrading in international supply chains. Before proceeding to the decomposition technique, however, a discussion of basic concepts and the identification approach deployed is in order. 3.1 Identification strategy

(22)

17

The remaining arrows reflect flows of inputs primarily, despite the orange arrow denoting the impact of re-allocating production activities from advanced CoC to an emerging CoS.

Social upgrading, on the other hand, can be divided into several sub-components, such as measurable standards and enabling rights. While the former refers to those aspects of worker well-being that are more easily

observable and

quantifiable, including job opportunities, wage levels, social protection and working hours, the latter is less easily quantified (Elliot and Freeman, 2003). Common measures of enabling rights capture freedom of association, the right to collective bargaining, non-discrimination, voice and empowerment. A lack of access to enabling rights thus undermines the ability of workers to negotiate improvements in their working conditions that can enhance their well-being. Accordingly, I expect social upgrading in a given sector when the following two requirements are met: i) a rise in real wages, denoted by the share of labour compensation in valued added, or ii) a hike in our index of labour standards that compiles information on work conditions. Now that we have clarified the rules to identify the two spheres of upgrading, the next section will present technical means of variable derivation.

3.2 Input-Output Analysis

(23)

18

States does not only generate factor income in those countries where the last stage of production takes place, but also in related industries abroad that supplied intermediate inputs. A tool that allows us to determine those inter- and intra-industry linkages is the Input-Output (IO) model that was originally developed by Leontief (1936). Essentially, IO analysis seeks to explain how one sector affects others in the same nation or region. According to the fundamental input-output identity, the output of one sector must be either consumed or become an intermediate input for another sector, in turn. This technique allows us to trace the amount of factor inputs needed to produce output of manufactures GVCs for worldwide final demand. Timmer et al. (2012b) label this approach ‘Global Value Chain perspective’. Compared to Leontief’s research (1936), however, which covered the United States mainly, the research horizon in this paper is extended towards a multi-country setting, in line with the international input-output modelling in Johnson and Noguera (2011). For example, the aircraft manufacturer Airbus might final assemble its A350 in Toulouse (France), but most of the input components originate from supplier firms abroad (Airbus Group). In order to account for these changes, ‘Structural Decomposition Analysis’(SDA) is widely used by the IO literature (for a review, refer to Miller and Blair, 2009). It is implicitly assumed that each country-industry only produces one homogenous product, implying ij different products. Even though products might be highly heterogeneous in reality, the lack of detailed data at this level does not allow further investigation. As illustrated in figure 1, output is produced employing factors of production as well as intermediate inputs. In CoS this output serves as intermediate input in domestic or foreign production, whereas domestic or foreign final demand is satisfied in CoC. Final demand f, in turn, can be separate in various categories, such as household consumption, government expenditures or investment. In this analysis, I abstain from a differentiation of final demand elements and use aggregated final demand worldwide.

As point of departure, matrix algebra proves handy to model the market clearing condition for each (ij) product13. This way, y represents a stacked mn×1 vector of output and f an equally stacked vector of finale demand for manufactures in each of the n industries in each of the m countries. Furthermore, A depicts a mn×mn matrix of input coefficients that determine the input from one country-industry required for one unit of output by another country-sector, say CoShk

and CoCij respectively. This set of equations can be written as:

(24)

19

, where the typical element yi denotes the output of industry j in country i, such as electrical and optimal equipment in China. The market clearing condition can also be expressed in matrix notation:

y = Ay + f (1) , where y denotes a column vector of (gross) output in each of the n industries in each of the m countries, corresponding to the set of equations. Similarly, f is depicted by a mn-vector of final consumption and A compiles a matrix of input coefficients that determine how much intermediates are required to produce a unit of output for a particular product. In turn, Ay combined describes the total amount of intermediates used in production. By the addition of a square identity matrix I with 1’s on its northwest-southeast diagonal axis to equation 1, we obtain the following relationship:

y = (I – A)-1f (2) , where (I-A)-1 constitutes the well-established Leontief inverse that tells us how much gross output value is created for goods and services in all stages of the production process of one unit of consumption, essentially. For reasons of simplification, B replaces (I-A)-1 and represents the Leontief inverse, in what follows. Now that we know how to derive the gross output generated in the production of a particular product, value added by factors that are directly and indirectly involved in this process will be determined next. In order to do so, the square matrix lˆ depicts factor requirements per unit of gross output in each industry n and country m. For illustration, imagine a particular industry in an emerging economy - say Mexican automobiles, that construct vehicles for General Motors and Ford for example. Thereby, l determines the value added per US$ of output generated by labour and capital in this specific location. Rearranging equation 2 yields:14

xi = uk lˆ Bf (3) _i

14_{The ‘^’ symbol (e.g. as in}

(25)

20

, where a typical element of xi denotes the exports of value added created in industry i and

country j to satisfy final demand levels in f. Johnson and Noguera (2011) define exports of value added as value added produced in a country and absorbed by other countries. This technique of distributing value added over country-industries is very useful as it prevents ‘double counting’. That is, conventional gross trade statistics data overstate the domestic, or value added content of exports at each border crossing rather than covering the net value added between border crossings. They give the example of China producing goods with intermediaries from Japan in order to satisfy final demand in the USA. The variable exports of value added only captures the value added in China, rather than total gross output values which would include Japanese value added. uk is a column vector used for selecting country-industries of interest. While it contains ones in the cells associated with the industries in the focal country, the remaining elements are zero.

After having determined what factor inputs are required to satisfy demand for production of a particular GVC h, we need to identify what changes in value added and labour compensation have been induced by re-allocation of production across countries and industries. In the spirit of Los et al. (2012), the origin of factor input changes in an industry or a specific skill group within a production network can be attributed to shifts in worldwide demand for output in the chain – so-called size effects - and changes in labour requirements per unit of output of the respective value chain – namely, within effects. The latter can be divided in three sub-effects. First,

technological change in a GVC affects the demand for jobs. It is measured by the total quantity

of labour required per unit of output from all country-industries involved in the supply chain. Roughly speaking, increases in technology cause on average a lower demand for jobs, everything else held constant. A prominent example for this is the automation of activities where workers are replaced by capital goods. In order to control for labour productivity differentials that might prevail across country-industries, a stacked mn×1 vector π consisting of country- and industry-specific productivity levels is deployed. Not accounting for this differentials would generate misleading results with respect to overall factor requirements. Accordingly, equation 4 expresses supply chain technology as:15

*'

i

l

=(π○l_i)’B (4)

15_{The symbol ‘}_○_{’ represents the Hadamard product, which is obtained from cell-by-cell-multiplication (e.g. Z=X○Y}

(26)

21

, where

l

*_i is depicted by a mn×1 column vector and stands for the worldwide factor inputs necessary to produce one unit of each of the final products. Productivity catch-up represents the second within-effect and refers to increases of labour productivity relative to the United States. In turn, this implies that demand for jobs will decline characterized by a surge in labour productivity levels. Efficiency levels of labour of a skill type across countries (π) are taken as productivity levels in country i relative to United States, based on output PPPs from the GGDC Productivity Level Database (Inklaar and Timmer, 2012). The measure of relative prices is thereby obtained via dividing purchasing power parity (PPP) by the exchange rate and is assumed to be identical across skill groups.

The third sub-effect of changes in labour requirements is referred to as offshoring of activities. In a nutshell, it describes changes in shares of intermediate inputs from different countries. Importantly, Los and others (2012) note that the values in the cells of matrix B determined by both intermediate inputs and the shares of these intermediate inputs delivered by each of the countries-of-supply. This insight is essential as l*i incorrectly predicts the factor inputs employed

by a particular country-industries. For this reason, we can estimate a mn×mn matrix R that covers shares of each of industries supplying factor inputs per unit of final demand produced by a particular GSC. Accordingly, equation 5 specifies a factor share matrix:

Ri = [πˆ lˆ B] _i l*_i 1 (5) In the last step,

Reorganizing equation 5 to lˆ B = _i _π_ˆ-1

Ri l*i and substituting into equation 3 gives the factor

inputs required in CoS j in period t:

it

x

= u’

k πˆt-1 Rit

*

it

l ft

(6)

In order to measure changes in particular factor inputs between 1995 and 2007, a dynamic perspective is introduced that can be written as:

1 i x - x_i₀= ' u_k πˆ₁-1 R_i₁ˆl*_i₁ * 1 fˆ

-

u'_k _π-1 0 ˆ R_i₀ˆl*_i₀ fˆ₀* (7)

Furthermore, the change in factor requirements can be decomposed into four elements:

(27)

22 * 1 -1 1 ' f l ) R -(R π u_k _i₁ _i₀ ˆ*_i₁ˆ (7b) * 1 * -1 1 ' f ) l -l ( R π u_k _i₁ ˆ*_i₁ ˆ_i₀ ˆ (7c) ) f f ( l R π u'_k -1₁ _i_{1 i}ˆ*₁ ˆ₁* ˆ₀* (7d)

Holding everything else constant, the above equations represent four partial effects of overall factor input changes into three within effects (equation 7a-c) due to productivity catch-up, production re-allocation and technological change. Moreover, equation 7d indicates the effect of differential rates of consumption growth for final output in countries-of-completion. Besides all other effects, the impact on changes in factor requirements due to re-allocation of production, or off-shoring, represents our key measure. In order to proceed further with the estimation of economic and social upgrading of emerging-country industries within advanced-country GSCs, equation 8a and 8b define our core variables of interest:

i

VA u'_kπ₁-1(R_i₁-R_i₀)ˆl_i*₁fˆ₁*

(8a) , defines a column vector that measures the absolute change in value added of intermediate goods output by CoS j for final assembly in a particular supply chain of CoC k generated by re-allocation of production activities. Positive changes in value added will be considered as economic upgrading. In a similar vein, changes in the social sphere are depicted by equation 8b:

is

LC u_k'π₁-1(R_i₁-R_i₀)ˆl*_i₁fˆ₁* (8b) , where a typical element ΔLCijs measures the absolute change in labour compensation for

workers in CoS j within GSC h of country k due to off-shoring of activities abroad. With respect to quantification, both economic upgrading and social upgrading are be measured in millions of US$. Importantly, these US$-values ought to be expressed in terms of the prices of the same period. That is, measuring the data at constant prices allows to remove potentially misleading effects of prices changes over time for our SDA estimates. Not controlling for inflationary price changes would cause our data for 2007 upwards biased, this way. By means of coherence, both trade data and factor values are hence converted to constant prices of 1995.

(28)

23

4. Model Specifications

Most empirical research on the determinants of social upgrading has been approached by case studies that concentrate on particular countries, industries or value-chains. At the same time, relatively little is the stock of cross-country studies that would allow us to compare experiences of upgrading across country-industries and would enable us to design development policy. One of these rare examples is provided by Milberg and Winkler (2011) who employ an ordinary least squares (OLS) regression analysis to link trade with different forms of upgrading. They find export growth to be positively associated with value added per worker, on average. Interestingly, trade does not seem to possess any association social upgrading, measured by employment growth, while economic and social upgrading are adversely aligned. Regarding the latter finding, Milberg and Winkler admit that employment growth might present a somewhat deceptive measure of social upgrading, since it fails to account for the quality and conditions of work. By extending our knowledge of the upgrading relationship with information on more quantitative and qualitative measures of work across value chains, this studies seeks to contribute to the existing stock of research.

While the structural decomposition analysis in the previous section allowed us to identify changes in upgrading due to international production fragmentation, this section aims to specify what factors affect patterns of upgrading in global supply chains. For this purpose, changes in labour compensation, denoted as ΔLCis, represent the dependent variable in the empirical model.

LCis refers to total pre-tax wages paid by employers to employees for work accomplished in a

year. Based on the Worker Rights Consortium (WRC) code16, it is considered an appropriate proxy for social upgrading: 'Wages are essential to meeting employees’ basic needs [...] and to

establish a dignified life, such as housing, energy, nutrition, clothing, health care, education, potable water, childcare, transportation and savings'. Recalling section 3.2, ΔLCis typifies

changes in million US$ of compensation to workers employed in country j to satisfy demand for intermediate inputs in an advanced country manufacturing chain. In order to visualize the relationship we are interested in, figure 3 illustrates both the predictors and the response variable for social upgrading (red bubble). The arrows represent thereby potential effects on the dependent variable, the regression analysis aims to identify with respect to direction and magnitude. The predictors can be roughly divided into four groups that all potentially affect

(29)

24

changes in labour compensation, as suggested by theory: changes in productivity (economic upgrading), country-specific fundamentals, value chain characteristics and other determinants, this research does not cover and that are captured by the error term. Among the explanatory variables, this study also pays attention to the mediating effect of labour market institutions, like trade union strength, on the translation of economic gains into the social sphere.

As key predictor variable, economic upgrading is measured by the value added content in output. Compared to using gross exports as indicator of economic upgrading and competitiveness, value added has the advantage that it avoids ‘double counting’ of intermediate goods value (Koopman et al., 2012). Essentially, value added equals the value paid to the factors of production in the exporting country. Thereby, ΔVAi refers to absolute changes in

value added embodied in intermediate goods exports from emerging countries caused by off-shoring of production activities within advanced-country GSC g. In the modern production process, many intermediate inputs cross borders multiple times. Thus traditional trade statistics that measure in gross terms including both intermediate inputs and final products become unreliable. That is, the value of intermediate goods that cross international borders more than once is counted twice. Value added thus provides a robust measure of economic upgrading by accounting for trade in intermediates. Since setting up offshore activities abroad is considered a time-consuming process (Los et al., 2012), changes are estimated over the long-term. For this reason, the model is specified for a twelve year time horizon. By estimating changes during the 1995-2007 time period, the estimation model for labour’s share of value added in CoS-industry i = 1,...,I can be expressed as follows:

(30)

25

, where ΔLCijsk = Proxy for social upgrading measured by growth of labour compensation in millions

of US$

ß0 = Intercept of the regression equation

ß1-ß10 = Parameters to be estimated

ΔVAij

g = Value added in millions of US$

Unioni 95 = Share of trade union membership in total workforce in 1995

Educi95 = Share of public spending on education of country i’s GDP

GDPi95 = Gross Domestic product of country j

Interneti95 = Internet users per 1,000 citizen

Prodi95 = Labour productivity, defined as goods and services produced by one hour of labor

and measured in millions of US$

RoLi = Rule of Law Index ranging between a minimum of -2.5 and a maximum of +2.5

Wforcei95 = Total labour force of country i in 1995

GVC_Sizej95k = Total employment induced in order to produce final goods within GVC k with its

last stage of production in country j

Skillj95 = Share of high-skilled labour involved in production of output in GVC k led by

country j ∑J

j=2 Dj = Sum of dummy variables created for each country-of-completion εij

k = Residual or error term, which assumed to be well-behaved with mean zero and

independent of explanatory variables

Importantly, national-level variables might ignore heterogeneity of production locations, think of special economic zones (EPZs) as prominent example. An ILO report (2007) finds that in most but not all of the major EPZ operating countries the national labour and industrial-relations legislations are applicable in the zones. Therefore the assumption seems to be reasonable. More information on this issue is presented in section 5.d. Despite changes in value added, growth in labour compensation in our model is thus determined by explanatory variables’ initial levels. This assumption is justified by the idea that the distribution of factor income in global value chains might be driven by fundamental factors whose effects differ significantly across countries. Moreover, our cross-country sample might suffer from potential unobserved heterogeneity, which is a form of omitted variables bias. It occurs when omitted variables are fixed for an individual country over a long time horizon. For example, a country’s history, geographic conditions and cultural background will not change tremendously. For reason, dummy variables for each country-of-completion are introduced that control for country-specific factors that might have driven the decision to offshore production activities in the first place. A detailed description of data sources is provided in section 5.

(31)

26

are likely to be driven by country size and country-specific factor endowments. For example, a country like China possesses a larger capacity to generate value added and to distribute large amounts of labour compensation due the extensive size of its market compared to the smaller Baltic states. Before proceeding to the next section, our variables of upgrading are adjusted to control for country-size effects via division by national workforce levels. Exemplary for this adjustment: _i 1,000,000 t i gst i st Wforce LC LCP

Thus, the response variable ΔLCPisg measures the per capita change in millions of US$ in labour

compensation induced by re-allocation of production activities over the time horizon of 1995 to 2007.Augmenting the original model by the new specifications, we obtain:

ij g j J j j g j g i i i i i i i ij ij sk D Skill Size GVC Wforce RoL od Internet GDP Educ Union VAP LCP 2 95 8 95 7 95 6 5 95 4 95 4 95 3 95 3 95 2 1 0 _ Pr (10) , where both measures of social upgrading and economic upgrading are adjusted to account for per capita changes in million US$ per capita. The specifications in equation 10 are similar to those in Feenstra and Hanson (2001). In particular, when the wage share of labour by skill type is increasing I am interested in determining how much of that change is due to changes in value added or initial levels of fundamental drivers, such as factor endowments or proxies for the developmental state of a country-of-supply.

5. Data – Sources and final Sample

(32)

27

classifies 150 countries as ‘emerging markets’ The classification is based on the their per capita income level, export diversification and level of integration into the global financial system. Repeated cross-sectional data allows us to control for country-specific variables that might be hard to capture, such as like cultural factors but also variables that change over time but not across entities like as national policies, federal regulations and international agreements. This way, the analytic framework accounts for individual heterogeneity in the data, implying that unobserved time-invariant variables are controlled for. On the downside, panel data users might face issues of data collection issues linked to sampling design and coverage, or cross-country correlation in the case of macro panels.

The time frame was chosen due to reasons data availability and in order to exclude the potential impact the recent worldwide financial crisis, which has been demonstrated to alter the structure of production networks (Gereffi and Frederick, 2012). In presence of ‘zero’ or missing values, the observations have been dropped as they cause an unintended bias in the remaining data. This is the case for the coke and refined petroleum value chain (NACE 23) in Cyprus or the leather and footwear chain (NACE 19) finalized in Luxembourg, for instance17. Table 2 summarizes descriptive statistics of dependent and independent variables used in this study. The final sample covers 5474 observations and does not differentiate by skill level. The remainder of this section will discuss the data collection and sources.

a. The World Input-Output Database (WIOD)

In order to implement the decomposition technique introduced in the previous section, information on gross output and value added at the industry level need to be collected. In particular, country-specific data on gross output (y), final consumption (f) and intermediate goods and services (A) by industry are required in order to determine the Leontief identity. The corresponding data have been sourced from the World Input-Output Database (WIOD).18 Launched in April 2012, this project is has been funded by the European Commission and provides time-series of world input-output tables. The current edition compiles data for 40 countries, including all EU27 countries as well as 13 other major economies and a model for the rest-of-the-world, covering the time period 1995 to 2011. In total, it covers more than 85 percent

17_{For similar reasons NACE 23 completed Luxembourg has been excluded.}

18_{Please refer to Timmer (2012a) for a thorough discussion of design, construction and sources of WIOD. Access}

(33)

28

of global GDP in 2008. Thereby, WIOD distinguishes key 35 industries according to the NACE classification and 59 product types.

In order to understand the distribution of benefits within value chains, we need to identify the income that is sustained in different parts of the chain, as compared to sheer profits. Opposed to figures of gross value of exports that are subject to double-counting in trade statistics, supply chain analysis stresses the importance of value added, a measure for the difference between output value and input costs, in each link of the value chain (Kaplinsky & Morris, 2001). A recent study on Apple’s iPod by Dedrick, Kraemer and Linden (2008) shows that a distinction between gross export values and value added from activities in the value chain are crucial in order to understand distributional gains at each stage of production of a good or service. They estimate that out of US$150 gross export value, only US$4 of value added are retained by Chinese workers in assembly activities. Besides world input-output tables, WIOD also provides figures of value added generated as the sum of distribution of labour, distinguished by three skill levels, and capital inputs in socio-economic accounts (SEA) that will be used to determine the value-added content by factor (l).

In order to determine how wages have changed over time due to international fragmentation of production, information on the distribution of income between the factors of production is required. In the case of wages this type of data is given by labour compensation in WIOD SEAs’, which is defined as wages and additional non-wage benefits, with a correction for income of own-account workers (Timmer, 2012a). For this purpose, labour compensation for the European countries has been estimated based on EUROSTAT data, while equivalent measures for non-EU countries originate country-specific data sources are used. As an important note, SEA indicators are given in values of national currency. For reasons of coherence, the socio-economic data have therefore been converted to US$ using a dataset of exchange rates accompanying the world input-output tables.

b. Control variables

(34)

29

abroad. In the context of this research, it is used as proxy for the market size of an economy and its potential to attain economies of scale.

Internet users can engage more intensively in a network, which provides therefore a source of competitiveness for businesses. An indicator for the access to worldwide networks and national ICT infrastructure is collected by the World Bank Development Indicators in terms of internet

users per 100 people.

For our purpose, the RoL variables serves as measure of national, regulatory environments. Data for the Rule of Law (RoL) variable come the Worldwide Governance Indicators (WGI)19, a research database summarizing results of private and public sector surveys on the quality of governance. It captures the respect of citizen and the state for institutions that govern economic and social interactions, such as quality of contract enforcement, property rights, and the courts, as well as the likelihood of crime and violence. The composite index consists of 25 individual variables and provides time-series data starting in 1996, which covers all our sample countries. c. Measures of social upgrading – The Quantity and quality of work

Important qualitative aspects of quality of work are hard to quantify. However, the literature has identified several quantitative measures as neat approximations of work conditions (for a comprehensive review, see Winkler and Milberg (2011)). These are presented in the following. i. ICTWSS Database

Trade unions represent a major player in the wage negotiation process of employees. Essentially, a trade union is a group of workers that have organized in order to maintain or improve prevailing conditions of their employment, with respect to wages, additional benefits and working conditions. In order to test Hypothesis 4, data on union density rate, denoted as the union membership as a proportion of wage and salary earners in employment, are gathered from the ICTWSS database20. The data collection contains annual data for all sample countries covering a time interval from 1960 to 2011. Moreover, it provides information on qualitative

19

The aggregate indicators as well as the corresponding source data are publicly available at

www.govindicators.org. Please, refer to Kaufman, Kraay and Mastruzzi (2010) for a thorough discussion of the WGI methodology.

20_{Visser (2013) presents an elaborate discussion of construction and sources of the ICTWSS database. Data publicly}

Economic Upgrading and Social Upgrading in Global Value Chains

__________________________________________________________

Master Thesis