Policy Research Working Paper 9114 World Development Report 2020 Background Paper Conceptual Aspects of Global Value Chains Pol Antràs Development Economics World Development Report 2020 Team January 2020 Policy Research Working Paper 9114 Abstract The paper offers an overview of some key conceptual aspects the key determinants and implications of GVC participa- associated with the rise of global value chains (GVCs). It tion, both at the country level and at the firm level. In the outlines a series of alternative interpretations and defini- process, the paper offers some speculative thoughts about tions of what the rise of GVCs entails, and it traces the the future of GVCs in light of the advent of an array of implications of these alternative conceptualizations for the new technologies. measurement of the phenomenon, as well as for elucidating This paper is a product of the World Bank’s World Development Report 2020 Team, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/ prwp. The author may be contacted at pantras@fas.harvard.edu. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Conceptual Aspects of Global Value Chains Pol Antràs Harvard University 1 Introduction In the last few decades, a series of technological, institutional, and political developments have fueled a signi…cant globalization of production processes across countries.1 More and more …rms now organize production on a global scale and choose to o¤shore parts, components or services to producers in foreign and often distant countries. The typical “Made in” labels in manufactured goods have become archaic symbols of an old era. These days, most goods are “Made in the World.” Some aspects of this new wave of globalization are not particularly novel. Signi…cant and sus- tained increases in the trade-to-GDP ratio had been experienced in the past. The period 1870-1914, for instance, witnessed a major increase in international trade ‡ows, largely fueled by the invention of the steamship, and that period is often referred to as the “First Globalization”. Similarly, inter- national trade in raw materials and intermediate inputs has been a prominent feature of world trade ‡ows since time immemorial. For example, Assyrian merchants who settled Kanesh (in modern-day This paper was written as a background paper for the 2020 World Development Report entitled “Trading for Development in the Age of Global Value Chains.” I am grateful to my fellow co-directors, Caroline Freund and Aaditya Mattoo, and to Penny Goldberg, Daria Taglioni and the rest of the core WDR team for helpful comments. Any errors are my own. 1 Three types of developments were particularly key: (i) the information and communication technology (ICT) revolution, (ii) an acceleration in the rate of reduction in man-made trade barriers, and (iii) political developments that brought about a remarkable increase in the share of world population participating in the capitalist system (see Antràs, 2015, for more details). Turkey) in the 19th Century BCE imported luxury fabrics and tin from Aššur, and they also traded copper and wool within Anatolia (Barjamovic et al., 2019). Despite these precedents, there is a common-held view that the transformation of the world economy since the 1980s has some distinctive features, and that interpreting the so-called rise of global value chains (GVCs) as simply an intensi…cation of trade integration across countries misses several key dimensions of this phenomenon. The aim of this paper is to o¤er a succinct overview of some key conceptual aspects associated with the rise of GVCs. The paper will o¤er alternative interpretations and de…nitions of what the rise of GVCs entails, and it will later trace the implications of these alternative conceptualizations for the measurement of the phenomenon, as well as for elucidating the key determinants and implications of GVC participation, both at the country level but also at the …rm level. In the process, this paper will also o¤er some speculative thoughts about the future of GVCs in light of the advent of an array of new technologies. Outline of the Paper The paper is structured as follows. It will …rst present a broad conceptualization of the rise of GVCs, one that interprets this phenomenon as an increase in the extent to which the goods and services transacted across borders are intermediate inputs rather than …nal goods (as emphasized in traditional conceptual frameworks of international trade). According to this broad de…nition, GVCs are tightly related to the use of foreign value added (embodied in materials, intermediate inputs or “tasks”) in production, particularly for exports. The unit of analysis in this broad approach is typically the country-industry, thus allowing this body of work to connect with a recent empirical literature focused on computing and documenting the observed growth in the extent to which foreign value added is used in production in speci…c countries and industries. s overwhelming focus on country- and industry-level studies, it will be Despite this literature’ argued that one can similarly apply this broad de…nition of GVCs to …rm-level analyses of inter- national trade. At the theoretical level, the paper will highlight that it is fruitful to conceptualize GVC participation at the …rm-level, particularly in environments in which …rms have some market power and production processes feature increasing returns to scale. In other words, GVC par- ticipation (even when interpreted in a broad sense) is ultimately a …rm-level phenomenon, and hence much can be learned from conceptualizing it in this manner. The measurement of GVC participation at the …rm-level is at an infant stage relative to the much more mature literature measuring GVC participation at the country-industry level, but this paper will argue that many of the measures that have been developed in world Input-Output analyses can be fruitfully adopted at the …rm-level, thereby opening the door for empirical analyses of the causes and consequences of GVC participation at the …rm-level (the original plan was to carry these analyses in Chapters 2 and 3 of the 2020 World Development Report). This incursion into the measurement of GVC participation will also lead to a critical reevaluation of the merits and limitations of the most widely used measures of GVC participation. 2 The broad conceptualization of the rise of GVCs might suggest that there is nothing funda- mentally new about this latest wave of globalization. It just entails more (or deeper) integration across countries, but it is shaped by the same factors as traditional trade ‡ows and it carries largely the same implications. Although it will be highlighted that even this broad view of GVCs delivers many novel insights, this paper will also develop a narrower de…nition of GVCs that emphasizes several distinctive characteristics of the rise of GVCs. This narrower conceptualization of GVCs highlights that GVC participation is often (and increasingly) associated with transactions that are very di¤erent in nature from the type of anonymous, one-shot transactions that permeate tradi- tional trade theory. The various …rms and plants participating in a GVC often exchange highly customized inputs on a repeated basis, with the contracts governing these relationships being highly incomplete and hard to enforce. Furthermore, …rms spend signi…cant time and resources designing the organizational structure of these production networks (e.g., whether transactions occur within or across …rm boundaries). From this alternative relational conceptualization of GVCs emerges a richer analysis of GVCs, one that puts at the center stage the major actors (multinational …rms, lead …rms in GVCs, etc.) that play a leading role in shaping GVC activity and foreign direct investment (FDI) ‡ows, and one that underscores the importance of institutional factors in shaping the location of global produc- tion. By explicitly modeling the mechanisms by which the division of the gains from specialization are divided across …rms, this relational approach also delivers novel lessons regarding the implica- tions of GVC participation for inequality and for development. Finally, this novel approach also provides a rich set of predictions regarding how an increase in automation or the adoption of digital technologies might a¤ect the landscape of the international economy and a¤ect di¤erent agents in society. On the empirical front, this relational approach to GVC activity has largely focused on case studies and, in a few cases, on more representative datasets that provide some information on the ownership decisions of …rms (see Antràs, 2015). A burgeoning literature is attempting to build more systematic measures of relational GVC activity, and later, this paper will build on this literature to suggest a measure of relational GVC participation based on transaction-level customs data from s Export Dynamics Database (EDD). It will also be discussed how one can build the World Bank’ on transaction-level trade data to construct country- and industry-level narrow GVC participation measures analogous to those developed in the literature embracing the broad approach to GVC activity. Beyond expositing these alternative conceptualizations of GVCs, this paper will put them to work to provide some guidance for the empirical work in various chapters of the 2020 World De- velopment Report. First, it will delineate various key determinants of GVC participation and will present hypothesis regarding the relative importance of these determinants depending on the type of GVC activity (broad or relational) one focuses on. Second, it will also build on the conceptual framework to highlight several notable implications of GVC participation for economic performance at the …rm level, but also for growth, poverty, inequality, and market structure at the country level. 3 Table 1. An Overview of Some Key Concepts    Broad / Traditional  Narrow / Relational  Definition  Basic:  Increased used of foreign  Basic:  Increased used of foreign  value added in production,  value added in production,  especially in production  especially in production  destined for exports  destined for exports, but    focusing on inter‐firm and intra‐ firm transactions involving  customized inputs and  relational contracting (distinct  from anonymous, spot trades in  homogeneous goods)    Conceptual framework  GVCs essentially entail a finer  GVCs entail a finer international  international division of labor  division of labor, but also  (trade in parts and components,  involve:    tasks)  ‐ nontrivial matching    between importers and    exporters  ‐ relationship‐specific  investments by all parties  ‐ inter‐firm and intra‐firm  flows of goods, technology  and credit in environments  with limited contractual  security  ‐ governed by incomplete  contracts enforcement     Empirical measures  ‐ Country‐ and industry‐level  ‐ Case studies;  measures of (i) foreign  ‐ Analyses of intrafirm trade  value added in production  flows or global ownership  and in exports (related to  patterns;  backward and forward GVC  ‐ Empirical work on  participation indices);  persistence in firm‐level  ‐ Analogous measures at the  trading relationships;  firm‐level   ‐ Measures based on HS    products (anonymous vs  relational)      4 Third, it will use the conceptual framework to speculate on the future of GVCs. Table 1 o¤ers a succinct summary of some of the key ideas in this paper (see also Appendix for a more extensive set of concepts and hypotheses put forth by Daria Taglioni). 2 A Broad View of Global Value Chains In presenting the broad or traditional conceptualization of the rise of GVCs in recent decades, it is useful to begin with a broad de…nition of a GVC and of GVC participation: De…nition: “A global value chain or GVC consists of a series of stages involved in producing a product or service that is sold to consumers, with each stage adding value, and with at least two stages being produced in di¤erent countries. A …rm participates in a GVC if it produces at least one stage in a GVC.” As is clear from these de…nitions, a GVC is de…ned as a production process that embodies value added (e.g., labor services) from at least two countries. As such, this notion of GVCs naturally relates this phenomenon to the increasing use of foreign value added in production, especially when that production is destined for exports. The de…nition above is agnostic about the speci…c form in which foreign value added is embodied in production, though it is often associated with either international trade in raw materials (e.g., tin or aluminum), in intermediate inputs (e.g., car parts), or in tasks (e.g., back-o¢ ce services). Similarly, the above de…nition is consistent with various con…gurations of global value chains, including simple “spider-like” structures – in which multiple parts and components converge to an assembly plant – and “snake-like” structures – in which value is created sequentially in a series of stages (see Baldwin and Venables, 2013, or Antràs and de Gortari, 2019). Regardless of the speci…c shape GVCs take, the possibility of fragmenting production across borders gives rise to a …ner international division of labor and greater gains from specialization. GVCs allow resources to ‡ow to their most productive use, not only across countries and sectors, but also within sectors across stages of production. As a result, GVCs magnify the growth, employment, and distributional impacts of standard trade. Under this “hyper-specialization” interpretation of GVCs, traditional determinants of international trade (such as factor endowments, geography, institutions, market size) are naturally also relevant for the extent to which countries and industries get embedded in GVCs, though the way in which these traditional determinants a¤ect GVC ‡ows is sometimes distinct from how they a¤ect traditional trade ‡ows, as will be discussed below. In sum, unlike in traditional conceptualizations of international trade, which focus on interna- tional transactions that involve only two countries (an exporting country and an importing country), GVCs entail production processes that often cross borders multiple times and that often involve more than two countries. This leads to a rich set of determinants and consequences of GVC partic- ipation, but it also creates important challenges for the measurement of GVC activity in the world. Before overviewing the main determinants and consequences of GVC participation, it is useful to explain these measurement challenges and how they have been surmounted in economic research. 5 Figure 1: The Structure of a World Input-Output Table Input use & value added Final use Total use Country 1 Country J Country 1 Country J Industry 1 Industry S Industry 1 Industry S 11 1S 11 1S 1 1 Industry 1 Z11 Z11 Z1 J Z1 J F11 F1J Y11 rs rs Intermediate Country 1 Z11 Z1 J S1 SS S1 SS S S Industry S Z11 Z11 Z1 J Z1 J F11 F1J Y1S rs s inputs Zij Fij Yjs 11 1S 11 1S 1 1 1 Industry 1 ZJ 1 ZJ 1 ZJJ ZJJ FJ 1 FJJ YJ rs rs supplied Country J ZJ 1 ZJJ S1 SS S1 SS S S S Industry S ZJ 1 ZJ 1 ZJJ ZJJ FJ 1 FJJ YJ Value added V A11 V A S 1 V A s j V A 1 J V A S J Gross output Y11 Y1S Yjs YJ1 YJS Measurement The main challenge facing the measurement of GVC arises from the fact that customs data, the standard source for international trade ‡ows, provides information on where the transacted good or service was produced, but not on how it was produced, i.e., which countries contributed value to it. Similarly, customs data record where the transacted good is ‡owing to, but not how it will be used, i.e., whether it will be fully consumed (absorbed) in the importing country, or whether it will be re-exported after the importing country adds value to it. With the goal of tracing value-added trade ‡ows across countries, a body of work has combined information from customs o¢ ces together with national Input-Output tables to construct global Input-Output tables. The most widely used world Input-Output tables are: (a) the World Input-Output Database (WIOD), a collaborative project led by researchers at the University of Groningen; (b) the OECD TiVA database, and (c) the Eora Global Supply Chain Database, constructed by a team of researchers at the University of Sydney. At a very broad level, this collaborative projects can be thought of as a “scaled up” version of product-level studies (such as Dedrick, Kraemer and Linden’s 2010 well-known case study of the iPod), attempting to break out the distribution of the …nancial value embedded in a product across the many participants in its supply chain. Figure 1 provides a schematic version of one such WIOT.2 With these global Input-Output 1 tables at hand, it is then straightforward to devise alternative measures to document the extent to which production process have become globalized in recent years, and the extent to which various countries and sectors participate in global value chains. At the world level, Borin and Mancini (2019) develop a natural measure of the importance of GVC trade in total international trade. Building on global Input-Output tables, they identify the share of a country’s exports that ‡ow through at least two borders.3 It is important to emphasize that 2 The WIOT in Figure 1 considers a world economy with J countries (indexed by i or j ) and S sectors (indexed by r or s). In its top left J S by J S block, the WIOT contains information on intermediate purchases by industry rs s in country j from sector r in country i. In the table, these intermediate input ‡ows are denoted by Zij . To the right of this block, the WIOT contains an additional J S by J block with information on the …nal-use expenditure r in each country j on goods originating from sector r in country i. These …nal consumption ‡ ows are denoted by Fij in the table. The sum of the (J S ) + J terms in each row of a WIOT represents the total use of output of sector r from country i, and naturally coincides with gross output in that sector and country (denoted by Yir ). 3 See also Wang et al. (2013). Other important papers on the measurement of GVC participation include the pioneering work of Hummels et al. (2001), Johnson and Noguera (2012), and Koopman et al. (2014). 6 Figure O.1 GVC trade grew rapidly in lower trade barriers induced manufac the 1990s but stagnated after the 2008 production processes beyond nationa global financial crisis Figure 2: The Importance of GVC Trade in World Trade O.1). GVC growth was concentrated electronics, and transportation, and 55 specializing in those sectors: East Asia GVC share of global trade (%) and Western Europe. Most countries 50 participate in complex GVCs, produ manufactures and services, and enga 45 activities (map O.1). By contrast, ma Africa, Latin America, and Central A 40 commodities for further processing in In recent years, however, trade an 35 have slowed (figure O.1). One reason overall economic growth, and especi 30 Another reason is the slowing pace an 70 75 80 85 90 95 00 05 10 15 of trade reforms. Furthermore, the fr 19 19 19 19 19 19 20 20 20 20 Sources: WDR 2020 team, using data from Eora26 database; Borin and production in the most dynamic regi Mancini (2019); and Johnson and Noguera (2017). See appendix A for a has matured. China is producing mo description of the databases used in this Report. the United States, a booming shale Note: See figure 1.2 in chapter 1 for details. Unless otherwise specified, GVC participation measures used in this and subsequent figures throughout the oil imports by one-fourth between 20 these exports encompass Report two follow the methodology broad types offrom GVC and Mancini Borin trade. (2015, On the one hand, slightly 2019). GVC tradereduced the incentives to outs includes transactions in which a country’ s exports embody value added that it has previously production. turingimported 2 from abroad. This type of GVC The expansion participation of referred is often GVCsto could Recent increases as backward GVC participation. On in protection cou the other hand, GVC tradestall alsounless comprises policy transactionspredictability in which a country’ evolution of GVCs. s exports are not fully Protectionism coul absorbed in the importing restored is country, and instead are embodied in the importing ing of existing country’ GVCs s exports to or their shifts t Unless policy predictability is restored third countries. The latter form of GVC participation is often dubbed forward GVC participation. GVCs have existed for centuries. But they grew swiftly of GVCs is likely to remain on hol As Figure 2 indicates, the overall share of GVC trade in total world trade grew very signi…cantly from 1990 to 2007 as technological advances—in trans- access to markets is uncertain, firms h in the 1990s and early 2000s, but it appears to have stagnated or even declined in the last 10 years. portation, information, and communications—and to delay investment plans until uncert Still, about one-half of world trade appears to be related to global value chains. The two components of GVC participation, backward and forward, can also easily be computed at the country and evenMap O.1 All countries country-sectoral level. When participate in GVCs—but doing so, it becomes clear thatnot the in the same way expansion of GVC activity has occurred in an uneven way in the world. On the one hand, there are regions in the world (such as Europe and East Asia) that are deeply involved in GVCs, while GVC participa- tion is much smaller in other parts of the world, most notably Latin America and Africa. On the other hand, even when countries actively participate in GVCs, the sectoral composition of these GVC ‡ ows are also quite diverse. Some countries have largely specialized in agricultural GVCs (such as Ethiopia) or in the natural resource segments of GVCs (such as Chile or Norway). Other countries are largely involved in the manufacturing segments of GVCs, with relatively less devel- oped economies (such as Tanzania) specializing in low-tech (or limited) manufacturing, and more developed economies (such as Mexico, Slovakia or China) participating in high-tech (or advanced) manufacturing processes. GVCInlinkages, addition,2015 it is useful to distinguish a subset of countries (e.g., India) that have largely specialized inLow participation the services embodied in GVCs, and a small set of very advanced Limited commodities economies (e.g., United States, Germany or Japan) playing a signi…cant role in the provision of High commodities Limited manufacturing Advanced manufacturing 7 and services Innovative activities Data gaps IBRD 4 Figure 3: Uneven sectoral specialization in GVCs Map 1.1 All countries participate in GVCs—but not in the same way GVC linkages, 2015 Low participation Limited commodities High commodities Limited manufacturing Advanced manufacturing and services Innovative activities Data gaps IBRD 44640 | AUGUST 2019 Source: WDR 2020 team, based on the GVC taxonomy for 2015 (see box 1.3). Note: The type of a country’s GVC linkages is based on the country’s extent of backward GVC participation, measured as the portion of imports embodied in manufacturing exports as a percentage of a country’s total exports, combined with the country’s sector specialization of domestic value added in exports and engagement in innovation. Countries in the commodities group have a small share of manufacturing exports and limited backward GVC integration. Their share of commodity exports can be low, medium, or high. Countries specialized in limited manufacturing GVCs engage in some manufacturing exports, often alongside commodities exports, and exhibit medium backward GVC integration. Countries specialized in advanced innovative goods and services. Figures 3 illustrates the extent to which GVC participation has manufacturing and services GVCs have a high share of manufacturing and business services exports and high backward GVC integration. Countries specialized in innovative GVC activities spend a large share of GDP on research and development, receive a large share of GDP from intellectual property, and exhibit high backward GVC integration. been uneven across countries. export widespread Despite their further raw materials foruse in processing; economic others 1.5  Figure it research, isCountry transitions important between different to emphasize two key limi- import inputs for assembly and exports; and still others types of GVC participation, 1990–2015 tations of global Input-Output produce tables. complex goods and services. First, In addition, because they rely on fairly aggregated Input-Output some are heavily reliant data, the resulting sectoralon GVCs for trade, whereas others disaggregation of GVC ‡ows is pretty coarse, and thus these data Innovative export largely domestic goods for consumption. To activities a signi…cant sources miss capture these distinct features of of amount GVC activity participation, coun- occurring within these broadly-de…ned sectors. tries are classified into four main types—commodities, Austria, Canada, Finland, Ireland, one can For instance,limited compute the origin of “fabricated manufacturing, advanced manufacturing and Singapore, Spain metal products”in Israel, Italy, Republic of Korea, Advanced the production of “mo- manufacturing services, tor vehicles”in the and innovative United activities—based States, on the prod- but where and services tires, car engines or windshield wipers originate from Czech Republic ucts they export and their participation in GVCs. The cannot be inferred from these rules for classification datasets. are described 1.3. second key in box A shortcoming of world Input-Output tables is China, Estonia, India, Lithuania, This taxonomy reveals clear distinctions among Philippines, Poland, Portugal, that, in constructing them, researchers are forced to regions. East Asia, Europe, and North America are impose strong assumptions Romania, Thailand, Turkey Limited to back out certain manufacturing engaged in advanced bilateral intermediate input trade ‡ manufacturing and services ows that cannot be readily read from either customs data or GVCs and innovative GVC activities, whereas Africa, national Input-Output Central Asia, and tables. Latin America are mostly in com- modities and A di¤erent, more granular limited manufacturing approachGVCs. to measuring the degree to which production Jordan, Argentina, Armenia, Lesotho processes Bosnia and Herzegovina, GVC participation intensified between 1990 and Cambodia, Costa Rica, Cyprus, El Salvador, are fragmented2015,across countries as illustrated by the many was …rst that countries suggested tran- byIndonesia, Yeats (2001) and consists of computing the Ethiopia, Kenya, Nepal, Serbia, sitioned up into more sophisticated forms of GVC South Africa, Tanzania Commodities share of trade ‡ ows accounted participation for by (figure 1.5). Transitions especially categories that can be safely assumed to contain industry were in East Asia common inputs only intermediate (as and Europe, re‡ ected by countries where the use of the words “Parts of” at the beginning of the were heavily engaged in the sectors most amenable to GVCs, such Yeats product description). (2001) as electronics found that and machinery. Among intermediate input categories accounted for about advanced countries, small open economies tended to 1990 2015 30% of OECD merchandise exports of machinery and transport equipment in 1995, and that this share had steadily increased from its 26.1% value in 1978. Yeats classi…cation has continued to be The new face of trade | 21 re…ned in recent years based on the Broad Economic Categories (BEC) product classi…cation made available by UNCTAD.4 4 See https://unstats.un.org/unsd/tradekb/Knowledgebase/50090/Intermediate-Goods-in-Trade-Statistics. 8 2.1 Determinants of GVC Participation In the workhorse conceptual frameworks developed to understand traditional trade ‡ows, such as the Heckscher-Ohlin model of trade, factor endowments are a key determinant of the structure of international trade. Skilled-labor abundant countries gain comparative advantage in producing skilled-intensive goods, while they bene…t from importing low-skill-labor intensive goods from low- skill labor abundant economies. Similarly, countries with a large availability of arable land or of natural resources (relative to their endowment of other factors of production) are expected to specialize and export primary products. In a world of global value chains, factor endowments play an analogous role in shape specializa- tion, but they also a¤ect the positioning of countries in GVCs. For instance, natural resource rich countries are expected to feature high levels of forward GVC participation because their exports of natural resources are used in a variety of downstream production processes that typically cross several borders. Similarly, and although agricultural products are much closer to …nal consumption than natural resources are, tari¤ escalation practices often leave less-developed countries special- izing in the most upstream stages of agricultural GVCs, that is, exporting raw products before processing.5 By the same logic, one would expect physical capital abundant countries or skilled-labor abun- dant countries to specialize upstream or downstream depending on the relative physical capital intensity or relative skilled-labor intensity of upstream versus downstream stages. According to a widely used U.S.-based sectoral measure of upstreamness (see Antràs et al., 2012), upstream sectors in U.S. manufacturing appear to be more capital intensive and more skilled-labor intensive than downstream sectors. One might then expect physical capital-abundant and skilled-abundant countries to feature relatively high levels of forward GVC participation and relatively low levels of backward GVC participation. The evidence in Chapter 2 of the 2020 World Development Report …nds support for the predicted link between low-skilled-labor abundance and downstream GVC positioning, but this same chapter shows that physical-capital scarce countries tend to specialize upstream rather than downstream. Nevertheless, the link between physical capital abundance and GVC positioning is blurred by the fact that GVCs often come hand-in-hand with FDI capital in- ‡ows. To provide a speci…c example, Nigeria is a relatively physical-capital scarce country, and oil extraction is a relatively physical-capital intensive process, but large FDI in‡ows have conferred Nigeria comparative advantage in that relatively upstream production stage. As a result, Nigeria features a large level of forward GVC participation relative to its backward GVC participation level (over the 1990-2015, the former’ s average was 0.09). s average was 0.26, while the latter’ The international exchange of goods and services is subject to a large degree of contractual insecurity (Antràs, 2015). Weak contract enforcement is a signi…cant deterrent of traditional trade ‡ows, but its incidence on GVC trade is likely to be disproportionately large. Part of the reason for this magni…ed e¤ect will be elucidated later, when discussing our relational conceptualization 5 Tari¤ escalation is the practice of setting higher import duties on semi-processed products than on raw materials, and higher still on …nished products (see WTO Glossary). 9 of GVCs. Yet, even when sticking to the view of GVCs as mere engines of hyper-specialization, multi-stage production processes with signi…cant complementarities across production stages will tend to be particularly sensitive to contractual institutions (see Acemoglu et al., 2007). In plain words, the performance of a global value chain is often severely impacted by the strength of its weakest link, and thus production delays or mishaps driven by weak contract enforcement might be particularly harmful in GVCs. In sum, the quality of a country’s institutions and its political stability are expected to be important determinants of GVC participation and might a¤ect GVC trade disproportionately more than traditional trade. Chapter 2 of the 2020 World Development Report …nds support for this hypothesis. Empirical work in international trade emphasizes that trade costs have a signi…cant negative e¤ect on trade ‡ows. There are various possible sources of trade costs, ranging from geographical features (such as remoteness), ine¢ cient infrastructure, regulatory barriers (e.g., tari¤s and quotas), or delays in clearing customs. There are at least two reasons to believe that trade costs might have a disproportionately negative e¤ect on the GVC component of trade ‡ows. First, the “weakest-link” mechanism invoked in the previous paragraph applies also to production delays associated with impediments to trade (such as customs delays). Second, and relative to traditional trade, higher trade costs not only increase the prices at which imported goods are consumed, as in traditional trade, but they also increase the cost of imported intermediate inputs, which get passed down the value chain and translate into higher costs associated with a country’s exports, thereby further depressing GVC participation. Regional trade agreements, or trade agreements more broadly, are a particularly e¤ective mech- anism to reduce trade barriers between the signing countries. Consequently, one would expect that GVCs are particularly active among countries that have signed regional trade agreements. Indeed, GVC activity is particularly intense among EU and ASEAN members, but it should be noted that other regional trade agreements, such as Mercosur, appear to have been much less successful in generating GVC participation among its members. Trade costs not only a¤ect the overall GVC participation of countries, but they might have a signi…cant e¤ect in the positioning of countries in GVCs. In sequential (or snake-like) GVCs, trade costs compound along the value chain and have a higher incidence on downstream stages than on upstream stages. As Antràs and de Gortari (2019) demonstrate, this leads remote countries to specialize in upstream stages, and more central countries to specialize in more downstream stages. An implication of this fact is that we might expect the e¤ect of trade costs to be more signi…cant for backward GVC participation indices than for forward GVC participation indices. Empirical work in international trade also emphasizes the role of market size in shaping bilateral trade ‡ows. Larger economies are expected to export more (since they produce more), and they are also expected to import more (since their income is larger). An often-underappreciated aspect of the gravity model of trade, the most successful empirical model of bilateral trade ‡ows, is that it provides a straightforward explanation for why the ratio of trade to output tends to be smaller 10 for larger economies.6 Whether a larger market size is associated with a higher or lower level of GVC participation is less clear, however. On the one hand, with sequential multi-stage production processes, larger countries naturally tend to attract a larger set of stages than small countries do. Furthermore, to minimize cross-hauling of semi-processed goods, the set of stages countries specialize in are often contiguous. As a result, larger countries are less likely to use imported inputs in their exports and should, other things equal, record lower levels of backward GVC integration. On the other hand, by their sheer size, large countries are likely to be geographically close to world demand for …nal goods, and thus their more “central” location should make them more prone to specialize downstream (see Antràs and de Gortari, 2019), and thus record higher backward GVC integration on account of their centrality. 2.2 Consequences of GVC Participation How has the rise of GVCs a¤ected the economies in which GVC participation has grown dispro- portionately? How might a rise in protectionism, by undoing some of the forces in recent decades, a¤ect those same economies? Is protectionism costlier in the age of GVCs? These are the questions that will next concern us. A …rst important insight already anticipated above is that by a allowing a …ner international division of labor, the growth of GVCs is associated with greater income gains from trade than a commensurate expansion of traditional trade. Intuitively, GVCs allow countries to bene…t from the comparative advantage of other countries not only at the sectoral level, but also at the stage level within sectors. As a result, models of trade featuring global Input-Output links (such as Caliendo and Parro, 2013, or Antràs and de Gortari, 2019) typically deliver larger gains from trade than models without those links.7 Furthermore, the resulting magni…ed costs of protectionism in a world of GVCs are exacerbated by environments in which the multiple stages involve in production feature especially high levels of complementarity (see Fally and Sayre, 2018, or Baqaee and Farhi, 2019). As we will discuss more extensively in section 3, when conceptualizing GVCs at the …rm level, rather than at the country-sector level, it will also be clear that part of the reason for these magni…ed real income gains stems from the fact that, by lowering input costs and allowing an expansion of …rm scale, GVCs tend to increase the productivity of …rms. It is well understood that the aggregate income implications of trade liberalization are also s growth rate ampli…ed in environments in which, via enhanced technology di¤usion, an economy’ might also be positively a¤ected by trade opening (see, for instance, Sampson, 2015). If increased GVC participation enhances technology di¤usion across countries, it would thus seem plausible that the rise of GVCs has a larger e¤ect on growth rates than traditional trade does. It is fair to say, 6 P between countries i and j are given by Xij = Yi Yj =k, then Note that, in a world without input trade, if exports k must be equal to world GDP, and the trade share is j 6=i Xij =Yi = 1 (Yi =Yworld ). 7 Arkolakis, Costinot and Rodríguez-Clare (2012) show that the costs of moving to autarky are shaped by the share of imports in a country’ s total spending. In their model of sequential GVCs, Antràs and de Gortari (2018) instead obtain a formula in which the costs of moving to autarky depend on the share of spending in goods that embody any foreign value (regardless of whether those goods themselves are imported or not). 11 however, that the broad conceptualization of GVCs developed so far does not elucidate whether GVCs are a particularly e¤ective vehicle of technology di¤usion. We will return to this point later in the paper. Although GVCs are expected, on average, to generate increases in aggregate income, they also generate nontrivial e¤ects along the income distribution. Much of the literature on the implications of production fragmentation has in fact focused on studying how it shapes the distribution of income, and more speci…cally, the distribution of wage income within countries. The Stolper-Samuelson theorem, one of the key tenets of the canonical conceptualization of traditional international trade, indicates that deeper trade integration is likely to increase wage inequality (the relative wage of skilled versus unskilled workers) in relatively advanced, skilled-labor abundant countries, while it is expected to decrease wage inequality in less developed, skilled-labor scarce countries. In a world of fragmentation, however, the validity of the Stolper-Samuelson theorem is under- mined. More speci…cally, it is widely accepted both theoretically and empirically, that increased production fragmentation leads to increased wage inequality in both advanced and less-developed countries (see Goldberg and Pavcnik, 2007). There are at least three reasons for this. First, when production is o¤shored to less developed economies, the labor force in those economies …nds themselves being employed in new production processes and tasks that might have been perceived to be low-skilled labor intensive in advanced countries, but that are instead skilled-labor intensive relative to the outside opportunities of workers in less developed countries (see Feenstra and Hanson, 1996, 1997). Hence, o¤shoring increases the relative demand for skilled workers in less developed economies and puts upward pressure on wage inequality. A second force towards increased wage inequality in less developed economies stems from the fact that GVCs are often more skill sensitive than traditional trade ‡ows, partly because they often produce goods that are destined for quality-sensitive consumers in rich countries (Verhoogen, 2008), and partly due to high complementarities among the various stages of production carried in di¤erent countries (Kremer and Maskin, 2003, Antràs, Garicano and Rossi-Hansberg, 2006).8 A third force toward increase wage inequality in skilled-labor scarce countries is related to the fact that …rms in GVCs tend to adopt more capital-intensive techniques than comparable domestic …rms (Bernard et al., 2018). Physical capital deepening and upgrading contributes to the increase in the relative demand for skilled workers due to capital-skill complementarity, namely the fact that physical capital (and capital equipment, in particular) is less substitutable with skilled labor than it is with unskilled labor (Griliches, 1969, Krusell et al., 2000). Although GVCs may increase income inequality in less developed economies, their impact on poverty is much less clear cut. On the one hand, the high-quality sensitivity of GVCs often lead them to marginalize the least skilled agents in society, and the increased opportunities they provide for more skilled agents might further deteriorate the livelihood of the less well o¤ (e.g., by bidding up the price of housing or of other services). On the other hand, the positive e¤ects of GVC 8 Production fragmentation also quali…es the link between trade integration and wage inequality in advanced economies. Grossman and Rossi-Hansberg (2008) argue that o¤shoring may increase wage inequality in advanced economies by less than traditional trade would. 12 participation on overall income and income growth are likely to trickle down to everyone in society. Those directly favoring from GVC integration will use their larger incomes to demand more local goods and services, which will generate novel opportunities for other agents in society, even those with relatively few skills. Furthermore, higher aggregate income can support a deeper welfare state that can ensure that the gains from integration are more evenly spread. Finally, GVC integration in certain regions of a country, can incentivize internal migration within countries, which can be a powerful mechanism to reduce poverty. In practice, it appears that GVCs have contributed to lifting millions of individuals out of poverty in some countries (such as China), but they may have failed to do so in other less developed countries. Chapter 3 of the 2020 World Development Report presents evidence that in Vietnam, poverty reduction was greater in locations with a higher presence of GVC activity. For particularly underdeveloped countries, it is also often argued that GVCs may facilitate industrialization by reducing the range of required “capabilities” that these countries must be endowed with to be able to produce and export industrial goods. For example, in the auto industry, countries can participate in GVCs even when they might not have any domestic car makers or any domestic provider of car engines. On the other hand, more sophisticated tasks in value chains require skills and capabilities that many underdeveloped countries lack. As a result, the rise of GVCs might lead less developed economies to specialize in relatively low-value added segments of production with little scope for upgrading. In addition, tari¤ escalation practices by developed economies tend to reduce the value added that can be captured by less developed producers when attempting to remain competitive vis à vis other countries.9 In sum, it may be simpler to “industrialize”in the age of GVCs, but the returns to doing so might not be as high as they were in the past. Furthermore, these e¤ects interact with the gradual increase in automation, as will be discussed later in the paper. The impact of trade liberalization on the environment also needs to be reevaluated in a world of GVCs. The transportation of goods across long distances generates CO2 emissions that directly harm the environment, but in GVCs, the same value added is often shipped multiple times before reaching …nal consumers. Thus, the volume of CO2 emissions generated in transporting goods in GVCs appear to be associated with a larger environmental harm than traditional trade. Similarly, the fact that GVCs foster “hyper-specialization”would appear to indicate that the pollution haven hypothesis might apply particularly intensely to GVCs, as lead …rms from large industrialized nations may locate “dirty” production stages in countries where environmental norms are laxer, thus avoiding the cost of stringent environmental regulations. It is important to emphasize, however, that there are at least two reasons that suggest that environmental concerns associated with globalization might instead be alleviated in the age of 9 Consider the following example. A car assembler in a small, less developed economy faces …xed foreign input costs of $7,000, while a …xed sale price of $9,000 in rich countries. It can thus capture value added of $2,000 per car. Nevertheless, if rich countries set a 20% import tari¤ on assembled cars, to remain competitive, the car assembler will need to reduce its export price to $7,500, and thus local value added is reduced to $500. Thus, a 20% import tari¤ on …nal goods can reduce value added in less developed economies by 75%. 13 GVCs. First, as argued above, GVCs have the potential to accelerate growth in the economies that participate in them actively, and it is well understood that the demand for environmental regulation tends to increase when income rises (or, in economic jargon, environmental quality is a normal good). Second, GVCs are often an e¤ective vehicle of technology transfer, so it seems plausible that they will also constitute an e¤ective vehicle of clean technology transfer. In order to better elucidate this mechanism, however, it will be necessary to develop richer conceptual frameworks of GVCs, a task to which this paper will shortly turn to. This section concludes with a very brief overview of some other macroeconomic consequences of GVCs (these are explained in greater detail in Chapter 4 of the 2020 World Development Re- port). Relative to traditional trade, in which producers compete head-to-head to service foreign markets, GVCs are associated with a higher degree of complementarity in production across coun- tries, as productivity and demand shocks travel upstream and downstream along value chains. This translates into a faster and more intense transmission of shocks across countries (as exempli…ed by natural disasters such as the 2011 Tohoku earthquake), and at a more aggregate level, it also leads to higher co-movement of output and prices across countries (i.e., larger business-cycle synchro- nization and in‡ation spillovers).10 GVCs also weaken the e¤ects of movements in exchange rates on the trade balance. For instance, the positive e¤ect of depreciations on the competitiveness of exports, is stymied by the increased cost of the foreign value added used in production. 3 A Firm-Level Approach to Global Value Chains A common feature of both the conceptual framework and empirical measures described so far is that they advocate an analysis at the country level or, at best, at the country-industry level. Indeed, in neoclassical frameworks in which production technologies feature constant returns to scale, there is little hope to generate predictions at a more granular level (i.e., at the …rm or plant level). Furthermore, Input-Output tables and publicly available international trade statistics are reported at the industry- (or product-level) rather than at the …rm level. In the real world, however, it is not countries or industries that participate in international trade, but rather …rms. In line with this simple observation, economic research in international trade has undergone a dramatic transformation in the last twenty years, one that has placed …rm- level international strategies at the center stage. This intellectual revolution was fueled by the increased availability of longitudinal plant and …rm-level datasets that permitted researchers to unveil a series of new facts that challenged the validity of existing models (see Bernard and Jensen, 1995). At the theoretical level, the seminal paper in the literature is that of Melitz (2003), which focuses on the exporting decisions of heterogeneous …rms within an industry. In Melitz’framework, …rms are assumed to produce di¤erentiated products under technologies featuring increasing returns to scale. Product di¤erentiation confers market power to …rms, while scale economies are associated with …rms facing …xed cost of production and of distribution. The decision of a …rm on whether 10 In the presence of inventories, GVCs are also expected to generate di¤erential volatility at di¤erent stages of the chain (see Ferrari, 2019). 14 to export to a given foreign market is shaped by a comparison of the potential operating pro…t obtained in that foreign market with the …xed costs associated with distributing products in that market. Although the canonical model in this …rm-level approach to international trade was written in a setting in which international trade involves the exchange of only …nal goods, as in traditional conceptualizations of trade, an active literature has adopted similar ideas to understand the rise of GVCs. In the presence of …xed costs of engaging in global sourcing (i.e., of importing intermediate inputs), one would expect that the use of imported inputs in production will demand that importers attain a minimum e¢ cient scale of production, with smaller and less productive …rms in an industry being excluded from GVC participation (see Antràs and Helpman, 2004, Halpern, Koren, and Szeidl, 2015, Gopinath and Neiman, 2014, Antràs, Fort and Tintelnot, 2017). The combination of scale economies and …xed costs of importing and exporting also provides a natural explanation for the fact that …rms that are large enough to be able to amortize the …xed costs associated with importing will also tend to be large enough to …nd exporting successful. Selection into importing thus naturally is associated with …rms engaged in backward GVC participation according to the de…nition developed in our broad conceptualization of GVCs (i.e., the use of foreign value added in exporting). Similarly, …rms exporting intermediate inputs, are likely to engage in forward GVC participation, as the …rms importing their products are likely to be exporters themselves. When adopting a …rm-level approach one can also distinguish global value chains that are orga- nized by a lead …rm, which incurs the bulk of the …xed costs associated with setting up the network of producers in a given production process, and global value chains that are more decentralized in nature, with individual producers incurring costs to set up links upstream and downstream from them (see Bernard, Moxnes, and Ulltveit-Moe, 2018, Antràs and de Gortari, 2019). 3.1 Measurement Firm-level datasets containing information on the import and export transactions of …rms can be fruitfully used to construct measures of GVC participation similar in nature to those discussed above based on the country- industry information in global Input-Output tables. More speci…cally, transaction-level customs datasets of the type available from the World Bank’s Export Dynamics Database can be used to identify the set of …rms in a country that participate in trade, further distinguishing …rms that export, …rms that import and …rms that both export and import. When a given …rm in a given country both imports and exports, it is natural to conclude that this …rm participates in GVCs. To map this de…nition more precisely to the de…nition of backward GVC participation developed in country-industry studies, one would ideally also resort to product-level information to verify that the goods imported by an exporting …rm are indeed intermediate inputs (rather than …nal goods), so that one can more comfortably conclude that this …rm is indeed using foreign value-added in their production destined for exports. Without linking customs-level data across countries, it is much harder to come up with analogous …rm-level measures of forward GVC participation. The 15 reason for this is that even when a …rm is identi…ed as an exporter of intermediate inputs (rather than of …nal goods), it is virtually impossible to establish whether those inputs are fully absorbed in the importing country, or whether they are re-exported to third markets by the importing …rms after having added value to them. We envision a future in which researchers will be able to conduct more satisfactory analyses with linked customs-data datasets from a variety of countries, but this future is not yet here. It should be stressed that the above …rm-level measures only identify the extensive margin of GVC participation, while industry-level measure based on global Input-Output tables also capture the intensity of GVC participation. Computing intensive measures of GVC participation at the …rm- level data is however challenging, particularly in situations in which complementary census-level information is not available. The reason for this is that customs-level data have no information on …rms’ domestic purchases of inputs and on …rms’ domestic sales of goods. As a result, it is di¢ cult to infer the ratio of foreign inputs used in production, and it is even more challenging to disentangle the foreign-input content of exports from the foreign-content of overall production (see Kee and Tang, 2016, for an attempt using processing trade in China). Firm-level information on importing and exporting can also be used to shed light on the extent to which global Input-Output tables provide an accurate description of value-added trade ‡ows across countries. Even when the entries of these tables provide an accurate account of the origin of inputs s industrial production, the standard methods used to compute bilateral value-added in a country’ trade ‡ows from these tables assume that the same combination of inputs is used in production regardless of the destination of sales of a country and industry’s output. As demonstrated by de Gortari (2018) with …rm-level evidence from Mexico, in practice, …rms selling output to di¤erent markets use very di¤erent combinations of sources of inputs, and this has signi…cant implications for the type of bilateral value-added trade ‡ows one infers from global Input-Output tables. For instance, because Mexican exports to the U.S. embody a disproportionate amount of U.S. value added relative to Mexican exports to other countries, de Gortari estimates that the share of U.S. value in U.S. imported Mexican manufactures is 30%, rather than the 17% one would infer from standard techniques applied to global Input-Output tables.11 We now turn to elucidate the determinants and consequences of GVC participation through the lens of this …rm-level approach to GVCs. To be clear, many of the forces outlined in the initial traditional conceptualization of GVCs continue to hold under this novel approach, so the focus will be on outlining distinct implications that arise in the present context. 3.2 Determinants of GVC Participation Firm-level GVC participation is shaped by many of the same forces that shape GVC participation at the country-industry level. Firms in countries with large endowments of factors that are used 11 Apart from qualifying the type of implications that one can draw from aggregated Input-Output tables, …rm-level data can also be fruitfully used to test more systematically the validity of the “proportionality” assumptions that go into the construction of those data. 16 intensively in certain segments of GVCs will be more prone to participate in GVCs, while trade costs and institutional quality continue to play a key role in shaping the likelihood that GVCs ‡ow across a given country. The …rm-level perspective, however, carries at least two important new lessons. First, for some of the determinants of GVC activity, the relevant endowments might be better measured at the …rm level than at the country level. For instance, even in countries with relatively low levels of educational attainment, some …rms might be able to participate in relatively skilled- labor intensive value chains if certain …rms in that country manage to attract a critical mass of skilled workers to work for them. On the other hand, it is much harder for individual …rms to surmount the obstacles posed by an unfavorable geography, by de…cient infrastructure, or by weak contract enforcement (see, however, our discussion of relational GVC below). Similarly, market size certainly matters for whether …rms participate or not in GVCs, but often what is relevant is that …rm size is su¢ ciently high, both to be able to amortize the …xed costs associated with GVC participation, and also to be able to ful…ll large-volume orders from comparably large importers in other countries. The second key lesson arising from a …rm-level conceptualization of GVCs is that to better understand the extensive margin of participation, it is fruitful to separate determinants that are likely to a¤ect the …xed cost of participation from determinants that are likely to impact GVC pro…tability conditional on participating.12 This is because the relative magnitude of these e¤ects is likely to shape the characteristics (size, productivity, etc.,) of …rms participating in GVCs in a given country. 3.3 Consequences of GVC Participation The consequences of GVC participation also become signi…cantly richer when adopting a …rm-level approach to GVCs. Such a focus makes it clear that participation in GVCs generates aggregate income gains partly by increasing the productivity of …rms. There are in turn two forces that shape these productivity improvements. On the one hand, there is a direct e¤ect coming from the fact that the use of foreign value added in production is associated with …rms being able to secure inputs from abroad at lower cost than they would from domestic suppliers. On the other hand, this reduction in costs coupled with the exporting associated with GVCs tend to increase the scale of operation of …rms engaged in GVCs, and this tends to reduce their average costs in the presence of scale economies. In sum, GVC participation enhances …rm-level productivity, a fact that has been con…rmed empirically by Amiti and Konings (2007), Goldberg et al. (2010) and De Loecker et al. (2016), among many others. These productivity e¤ects imply that import competition shocks can carry signi…cantly richer implications when they entail surges in imports of …nal goods (as in traditional trade) and when 12 Antràs, Fort and Tintelnot (2017) identify a signi…cant divergence between the extent to which foreign countries generate cost savings for U.S. …rms o¤shoring in those countries and the …xed costs associated with sourcing from these countries. For instance, China is estimated to o¤er signi…cant cost savings, but the initial set up cost to be able to source from that country appears to be especially high. 17 they partly entail an increase in imports of intermediate inputs. In the latter case, the positive productivity e¤ects at the …rm-level can lead to unexpected implications for scale and employment at the …rm level. For instance, Antràs, Fort and Tintelnot (2017) show that U.S. …rms that started s accession to the WTO also increased their sourcing from importing from China after this country’ domestic suppliers in the United States. Envisioning GVC participation as a …rm-level phenomenon also helps conceptualize why GVCs might be a particularly powerful vehicle of technology transfer. The reason for this is that unlike in formulations of traditional trade, in which …rms from di¤erent countries compete against each other, GVCs constitute networks of …rms with common goals, such as minimizing production costs or maximizing pro…ts associated with the production chain. Surely, the incentives of agents in GVCs are not always aligned, and the division of the gains generated by GVCs naturally generates distributional con‡ict, but downstream …rms typically gain from productivity enhancements of upstream producers in their chains, and vice versa. A direct implication of this simple observation is that …rms from advanced countries importing or exporting goods to less developed economies might …nd it bene…cial to share process and product innovations with their GVC co-participants in those less advanced economies. In sum, technology transfer is particularly ‡uid within GVCs. s participation in GVCs can As emphasized in our broad conceptualization of GVCs, a country’ also constitute a signi…cant source of increased income inequality. More speci…cally, the growth in aggregate income stemming from countries joining GVCs is often coupled with nontrivial increases in wage inequality. The …rm-level approach to GVCs provides a more ‡eshed-out version of this mechanism, as the type of large …rms often participating in GVCs (especially in manufacturing) tend to be larger, more skilled intensive, and more capital intensive than other types of …rms (Bernard et al., 2008). A study of GVC activity at the …rm level further introduces novel dimensions associated with the distributional consequences of trade integration. For instance, in the presence of scale economies and …xed costs of participation, large …rms will tend to bene…t disproportionately from GVC participation due the cost-savings associated with importing and the expansion in scale a¤orded by exporting. As a result, the size distribution of …rms is likely to be signi…cantly more skewed in a world of GVCs than in a world without them. This suggests the intriguing hypothesis that the rise of “superstar” …rms in the U.S. and other advanced economies (see Autor et al., 2017) might be partly associated with the rise of GVCs. Furthermore, it is a well-established fact that large …rms tend to face lower price elasticities of demand than smaller …rms, and that cost reductions are often only partially passed on to prices, and particularly so by large companies. As a result, the growth of GVC activity appears to be a potential contributor to the recently documented widespread rise in average markups and in the dispersion of these markups (see De Loecker et al., 2019). In fact, De Loecker et al. (2016) provide direct evidence from India showing that input trade liberalization was associated with an increase in the markups charged by Indian …rms importing inputs from abroad. Chapter 3 of the 2020 World Development Report also provides preliminary evidence that increasing GVC participation 18 appears to be associated with rising markups in developed countries but with falling markups in developing countries. The latter fact suggests that the relative bargaining power of agents in GVCs might be key for the distribution of the gains from GVC activity, an issue to which we will return in section 4. A rise in markups associated with GVCs is likely to re‡ect more than the necessary increase in price-cost margins required to cover the increased …xed costs associated with a more complex sourcing or exporting strategy. As a result, GVC participation is also likely to increase the pro…t rate of these companies, thereby generating a force towards a lower share of an economy’s income being paid to labor. Similarly, lead …rms in GVCs often transfer their relatively capital-intensive production techniques and automation practices to their foreign production facilities in less devel- oped economies, which might result in further shifts in the distribution of income away from labor and on to capital. There are of course many possible explanations for the observed global decline in the labor share (see Karabarbounis and Neiman, 2013), but the rise in GVC activity appears to be a likely contributor to this trend. In terms of other macroeconomic implications of the rise of GVCs, the fact that …rms partic- ipating in GVCs tend to be large and tend to engage in both importing and exporting provides a natural explanation for the fact that following large depreciations, import growth often tracks export growth closely, thus reducing the e¤ectiveness of depreciations in reducing trade imbalances (see Blaum, 2018). 4 A Relational View of Global Value Chains Although research adopting the broad approach to GVCs has provided valuable novel insights, modeling global production sharing as simply an increase in the extent to which foreign inputs (or foreign value added) is used in production misses important distinctive characteristics of the recent rise of GVCs. In other words, the de…nition of GVCs adopted so far may be too broad to do full justice to the novel landscape that has emerged in the world economy in the last thirty years. The rise of GVCs entails much more than the intensi…cation of the type of trade in raw materials and homogeneous intermediate inputs that has existed since the Bronze Age. It is also much more than importing and exporting …rms transacting with each other in world markets. The expansion of GVCs entails a …ner international division of labor but it also involves a number of additional features. Four of these distinctive features are particularly important. First, the process by which agents co-participating in GVCs match with each other is not frictionless. Fixed costs of exporting and importing partly re‡ect the costs of …nding suitable s products. For this reason, these …xed suppliers of parts and components or suitable buyers of one’ costs are better understood as sunk costs, which naturally create a “stickiness”among participants in a GVC. A second source of lock-in in GVC relationships is related to the fact that GVC participants often undertake numerous relationship-speci…c investments (such as purchasing specialized equipment or 19 customizing products) which would obtain a much-depressed return were GVC links to be broken. The need to customize inputs, coupled with quality sensitivity considerations, renders the above matching between buyers and sellers particularly important. If a …rm suddenly faces an increase in the demand for their goods, it cannot easily scale up by buying more foreign inputs from some centralized market. There are typically only a handful of suppliers worldwide that can provide the additional necessary customized inputs necessary to scale up. Third, …rms participating in GVCs do not only engage in trade in tangible goods with other members of their value chains. GVCs often involve large ‡ows of intangibles, such as technology, intellectual property and credit. The exchange of these intangibles is signi…cantly more complex than that of simple goods or services. Fourth, the prevalence of lock-in e¤ects and ‡ows of intangibles within GVCs is made particu- larly relevant by the limited contractual security governing transactions within these chains. GVC often involve transactions for which a strong legal environment is particularly important to bind producers together and to preclude technological leakage. And yet, GVC are often conducted in situations in which this strong legal environment is missing because cross-border exchange of goods cannot generally be governed by the same contractual safeguards that typically accompany similar exchanges occurring within borders. As a result, GVC participants are left to employ repeated interactions among them to build a governance that provides implicit contract enforcement. As in the case of matching frictions and relationship-speci…city, this force contributes to the “stickiness” of GVC relationships. In sum, these considerations lead to a novel, relational conceptualization of GVCs in which the focus is shifted away from the mere allocation of value added across countries resulting from anonymous, spot exchanges of goods and services. Instead, a new paradigm emerges in which the identity of the speci…c agents participating in a GVC is crucial. Within these GVCs, contracting is often relational in nature, and thus more likely to exhibit persistence than in transactions involving raw materials and homogeneous inputs. An extreme version of this type of relational contracting arises when parties involved in a GVC altogether by-pass the market mechanism and decide to transact within …rm boundaries, by having the buyer vertically integrate the seller or vice versa. The prevalence of intra…rm trade ‡ows in world trade ‡ows exempli…es the importance of relational aspects in the growth of GVCs. For instance, U.S. customs data suggest that, in recent years, close to one-half of U.S. imports involve related-party transactions (see Antràs, 2003). At the global level, intra…rm trade has been estimated to account for about one-third of world trade ‡ows. Nevertheless, the internalization of transactions in a GVC is just one of the many organizational responses to the contractual vagaries associated with cross-border transactions. In an in‡uential study, Gere¢ , Humphrey, and Sturgeon (2005) elaborate a much more extensive taxonomy of potential governance forms within GVCs, and various researchers have built on their work to shed light on the relative prevalence of these governance forms through a number of interesting case studies. 20 Although the relational approach to the study of GVCs is concerned with the same phenomena as the more traditional approach discussed earlier in this paper, these literatures have largely evolved in isolation. These separate paths are partly explained by the signi…cantly di¤erent sources of data used in these di¤erent literatures, as discussed next. 4.1 Measurement Measuring participation of …rms in relational GVCs is notoriously di¢ cult. First of all, because relational GVCs activity entails …rm-to-…rm links, it presents the same di¢ culties outlined in section 3 regarding measuring participation at the …rm level. But more importantly, the distinctive characteristics of relational GVCs call for a narrower empirical de…nition of this type of GVCs, one that excludes …rm-level GVC participation associated with the exchange of fairly homogeneous goods in spot transaction. Given these di¢ culties, it is no surprise that progress on this area has been rather scant. The most in‡uential work on the relational nature of GVCs tends to focus on particular case studies of speci…c sectors and countries, which allow the researcher to paint a more colorful and realistic picture of the web of relationship sustaining global production in that sector. The extensive and in‡uential of Gary Gere¢ provides a good example of this (see for instance, Gere¢ , 1994, on the apparel industry, Sturgeon et al., 2008, on the automobile industry or Bamber and Gere¢ , 2013, on the medical device industry in Costa Rica). More recently, Rocco Macchiavello and Ameet Morjaria have studied in detail the relational nature of contracting in various agricultural markets in less developed economies, such as their work on the ‡ower industry in Kenya (see Macchiavello and Morjaria, 2013) or their work on the co¤ee value chain in Rwanda (Machiavello and Morjaria, 2019). One of the initial goals of the 2020 World Development Report was to attempt to provide a bridge between this case-study-based literature and the broader empirical literature on GVC participation. The idea was to develop …rm-level and product-level measures of the extent to which the GVC participation of …rms is relational in nature, and the extent to which international transactions associated with a given product tend to be relational in nature. More speci…cally, the aim was to exploit the panel nature of some of the datasets available from the World Bank’s Exporter Dynamics Database (EDD) to construct measures of the observed persistence of …rm- to-…rm transactions, perhaps partialling out various natural determinants of persistence (such as exchange rate volatility). This is very much in line with the measure of “relationship stickiness” proposed by Martin et al. (2018), which they compute using detailed …rm-to-…rm export data from France. The hope was to “scale up”this approach and apply it to various countries, with the initial goal of verifying whether there was a signi…cant positive rank-correlation in product-level stickiness across countries. Such a …nding would indicate that there is a signi…cant product-level dimension to stickiness that is likely to render GVC participation in certain goods particularly prone to be relational in nature. With such …nding at hand, one could use standard industry concordances to construct a global Input-Output table that separates trade ‡ows associated with traditional 21 versus relational transactions, thereby allowing one to construct broad industry-level measures of relational GVC participation for the comprehensive set of countries in the Eora Global Supply Chain Database. Due to time constraints, such analysis was not conducted in time for the 2020 World Development Report, but hopefully it will be carried out in future work. 4.2 Determinants of GVC Participation In a world of relational GVCs, the e¤ect of factor endowments on GVC participation is largely analogous to the one developed above for the broad notion of GVCs (both at the country-industry and at the …rm level). A key novel aspect, however, is that as mentioned above, the relational approach to GVCs provides a straightforward explanation for why GVC activity and FDI ‡ows go hand in hand. When tight control over foreign production processes is necessary (perhaps because the legal environment does not su¢ ce to discipline the behavior of suppliers or to avoid the leakage of intellectual property), lead …rms might decide to rely on integrated suppliers and assemblers in foreign countries. This results in intra…rm trade and FDI ‡ows. For this same reason, countries that put in place policies that are FDI friendly (such as establishing credible mechanisms to reduce the risk of expropriation or o¤ering tax breaks for new investments by foreign companies) are more likely to be able to participate in GVCs. The stickiness of relational GVCs also makes them particularly vulnerable to supply chain disruptions. For this reason, …rms participating in relational GVCs are likely to be particularly sensitive to the availability of skilled labor in the economies in which they operate. Furthermore, in relational GVCs, informational ‡ows and communication are particularly important, which suggests that language skills (such as a good command of English as a second language) might be particularly relevant for GVC participation. The relational nature of GVCs also considerably reinforces the role of institutional quality as a signi…cant determinant of GVC participation. Having said this, institutional quality and relational GVCs interact in subtle ways. On the one hand, the emphasis this approach places on contractual insecurity naturally implies that production processes involving high degrees of customization (i.e., more relationship-speci…c investments), and for which a thick “secondary” market is missing, are likely to be located in countries with strong institutional quality (Levchenko, 2007, Nunn, 2007). The reason for this is that in those countries, lead …rms are in a better position to discipline the behavior of producers and avoid costly production delays or quality debasements. Nevertheless, and as argued above, the same forces that make relational GVC rely intensively on institutional quality also make GVC links particularly “sticky”, which fosters the emergence of reputational mechanisms of cooperation which might partly substitute for the absence of formal contracting. In addition, under some circumstances, vertical integration may serve as a direct (albeit imperfect) substitute for strong contract enforcement in countries hosting GVCs. Because relational GVCs feature signi…cant ‡ s …nancial institutions and ows of credit and intellectual property, a country’ IPR protection policies also constitute important determinants of GVC participation. The e¤ects of trade costs on GVC participation need also be quali…ed when GVCs are rela- 22 tional in nature. As reiterated above, supply chain disruptions are particularly costly when …rms participating in these chains cannot easily resort to alternative producers when some of the links in the chain fail to provide components or services in time or under pre-speci…ed terms. With that in mind, trade delays associated with ine¢ cient or corrupt customs o¢ ces might act as a particularly large deterrent for relational GVCs requiring coordination and in-time delivery. Furthermore, in the presence of weak contract enforcement, the emergence of cooperation among producers in GVCs is fostered by repeated interactions among the several agents participating in the chain, interactions that may be severely curtailed by remoteness or insu¢ cient air connectivity. The lock-in e¤ects associated with costly search and relationship-speci…c investments also have implications for the role of market size in attracting GVC activity. Beyond the factors identi…ed by the above broad conceptualization of GVCs, in settings with relational GVCs, a large market might serve the role of reducing search frictions (Grossman and Helpman, 2005), and may also facilitate resorting to alternative suppliers in the presence of production disruptions. 4.3 Consequences of GVC Participation How will economies be impacted by participating in relational GVCs? Many of the e¤ects discussed under the broad conceptualization of GVCs continue to apply, but one can identify three particularly distinctive aspects of relational GVCs that have signi…cant rami…cations for understanding the consequences of GVCs. First, the relational approach clari…es the role of GVCs as vehicles of technology transfer. GVC participation is not just about using foreign value added in production or about engaging in importing and exporting. A GVC is made up of inter-…rm and intra-…rm relationships which govern the transfer of tangible goods, but also the transfer of information and technology involved in making a product or providing a service. The scope for absorbing foreign technology is thus particularly large in relational GVCs. Furthermore, the sticky nature of relational GVCs make them particularly prone to bene…ts from learning-by-importing and learning-by-exporting via repeated interactions with highly productive …rms in advanced economies. Similarly, the transfer of clean environmental standards to less developed economies might be most e¢ ciently carried out via relational GVCs. A second key consideration is that the combination of incomplete contract enforcement and the lock-in e¤ects stemming from search frictions and relationship-speci…c in relational GVCs give rise to transaction prices between buyers and sellers that tend to be bilaterally negotiated, and that are thus not fully disciplined by market-clearing conditions. This observation has a number of implications for the consequences of relational GVCs. Countries should concern themselves not only with fostering the participation of their local …rms in GVCs, but should also be mindful of the bargaining power that their …rms will have vis à vis large foreign …rms. The implications of GVCs for the emergence of superstar …rms with huge scale, high market power and large pro…t rates are exacerbated by the disproportionate bargaining power that these large …rms might have vis à vis their suppliers. More speci…cally, large lead …rms might be able to buy parts and components at 23 relatively lower prices (via volume discounts, playing suppliers against each other, etc.), while also being able to sell at higher prices (if they are themselves key suppliers to downstream producers). This strong bargaining power enhances the pro…tability of large …rms engaged in relational GVCs, but might come at the cost of a lower “share of the pie”accruing to smaller and less powerful …rms in less developed economies. These e¤ects might be compounded in countries with unfavorable geographical features or weak institutions, as large …rms might react to the risk of facing production disruptions in those countries by setting up “back-up”suppliers elsewhere, thereby further eroding the bargaining power of suppliers in distant or weak contracting countries. Indeed, as mentioned above, Chapter 3 of the 2020 World Development Report reports evidence suggesting that increasing GVC participation appears to be associated with falling markups in developing countries. Beyond distributional concerns, Antràs and Staiger (2012) also show that the fact that international prices are often bilaterally negotiated carries important implications for the role of design of international trade agreements. A third set of novel mechanisms by which relational GVCs might a¤ect inequality relates to the disproportionate importance of the matching between buyers and sellers. Because the identity of these producers matters, especially in situations in which sensitivity to quality is high, relational GVCs may set o¤ “a war for talent” in which the price of particularly attractive producers, or the wage of particularly skilled individuals might be disproportionately bid up relative to a world without relational GVCs. This constitutes another mechanism by which GVCs might exacerbate inequality in the distribution of income. 5 The Future of GVCs This paper has so far focused on overviewing some key conceptual aspects related to GVCs, and has attempted to develop a set of predictions regarding the key determinants and implications of the rise of GVCs observed in the last thirty years. In this concluding section, this chapter will leave the comfort zone of predicting the past, and will venture into the much more treacherous terrain of attempting to predict the future.13 Although the current political environment would be rife for a discussion of how a continuing escalation of trade tensions between the U.S. and other countries might a¤ect the future geography of global production, this concluding section will instead focus on speculating on the future of GVCs in light of the advent of an array of new technologies, such as digital platforms, blockchain, automation, and 3D printing. Some readers will be disappointed by this focus, but there are at least two reasons for it. First, new technologies are here to stay, while the current trade “strifes” between the U.S. and its major trading partners are very much in ‡ux and, as of the time of writing (December of 2019), it is not clear that they will develop into an all out “trade war”. Second, if an increase in trade barriers indeed materializes for a signi…cant amount of time, the e¤ects it would generate would largely be the mirror image of those following the decline in trade barriers in the 13 As Niels Bohr famously put it,“prediction is very di¢ cult, especially if it’s about the future”. 24 last thirty years, as already hinted in section 2.2. Instead, the e¤ects of new technologies are quite distinct in nature and deserve a separate discussion. How will the future of GVCs be shaped by new technologies? Will the recent (post 2008) slowdown and retrenchment in GVC participation continue in the next few decades? Or are we perhaps in the cusp of a new wave of globalization? And how will new technologies reshape the role of GVCs as a possible engines of development? This …nal section will apply the conceptual framework developed in previous sections to o¤er some tentative answers to these questions. 5.1 Digital Technologies Consider …rst the case of digital technologies. It is clear that global value chains are rapidly changing under the pressure of digital innovation. First and foremost, digital technologies encourage GVC participation by reducing many of the barriers that …rms face when attempting to join GVCs. For instance, digital platforms (such as Amazon, Alibaba or Mercado Libre) facilitate the matching of buyers and sellers, thus reducing the initial …xed costs associated with GVC participation.14 Extending access to high-speed internet and expanding e-commerce thus has the potential to greatly facilitate increased GVC participation by relatively small …rms, and also for …rms in countries with bad infrastructure (which now gain the ability to specialize in segments of global value chains that specialize in the provision of services via digital technologies rather than the provision of physical goods via transport infrastructure). These same technologies also enhance the management of inventories, and of logistics more broadly, thereby improving participation even in manufacturing segments of GVCs (see Fort, 2017). Furthermore, digital platforms (via rating systems) and open distributed ledgers (such as blockchain) enhance veri…cation and monitoring, thus reducing informational frictions and opening the door for countries with weak institutions to bypass a key factor limiting their participation in GVCs. Similarly, in situations in which language barriers remain signi…cant (e.g., in the provision of certain services), the application of big data and machine learning techniques has the potential to provide much more e¢ cient translation services (see Brynjolfsson et al., 2019). In sum, one would expect the unstoppable advance of digital technologies to provide a new tailwind to ensure the continuing growth in GVC activity worldwide. Apart from these e¤ects on the e¢ ciency of GVCs, it is also important to acknowledge some potential distributional e¤ects originating speci…cally from these novel technologies. For instance, the same reputation mechanisms GVCs rely on to verify seller and buyer quality may foster con- centration, thus making it harder for entrants to compete. Within existing relational …rm-to-…rm GVC activity, novel technologies might also have implications for the relative bargaining power of the di¤erent participants in GVCs. For instance, digital platforms might allow large buyers in rich countries to gain information on a larger number of potential suppliers, thus enhancing their ability to have these suppliers compete with each other. This in turn may lead to better terms of trade 14 Note that many digital platform companies o¤er parallel business-to-consumers and business-to-business plat- forms. 25 for lead …rms in rich countries, at the expense of a lower share of the gains from GVCs accruing to producers in less developed economies. Furthermore, digital platforms themselves have been accu- mulating vast amounts of information on the users of their platforms, and this certainly enhances their ability to use this information to their advantage, either by locking in buyers with particularly well tailored recommendations, or by price discriminating in particularly e¤ective ways. As a result, digital platform …rms also pose new challenges for regulators seeking to ensure fair competition and prevent abuse of market power. The speci…c features of relational GVCs emphasized in section 4 suggest that the e¢ ciency and distributional consequences following from the increasing adoption of digital technologies might not be orthogonal to each other. More speci…cally, because veri…cation and monitoring serve as a substitute for the need for implicit contract enforcement, they may lead to an erosion of the “stickiness” of GVC relationships. This in turn may well generate negative e¤ects on the productivity of …rm-to-…rm transactions. Indeed, recent work by Machiavello and Morjaria (2019) suggests that in co¤ee value chains in Rwanda, increased competition was associated with a lower use of relational contracts between mills and farmers, and this breakdown in relational contracts lowered mills’e¢ ciency and output quality. 5.2 Automation and 3D Printing Although new technologies have the potential to raise productivity, they sometimes can prove to be quite disruptive, especially when they lead to a reduction in the demand for workers. The example of industrial automation and robotics is a case in point. At …rst glance, automation constitutes an alternative to o¤shoring for …rms in rich countries seeking to lower their labor costs. Because automation and o¤shoring appear to be substitutes, one would then expect improvements in au- tomation to lead to an increasing amount of reshoring over time. This fact might be particularly concerning to less developed economies, which might view automation as a threat to their ability to leverage their cheap labor to get a foot on the door of GVCs. These fears might be compounded by the fact that large multinational companies typically design their production processes with their home market factor prices in mind, and might then impose these same production processes possibly involving large amounts of automation, in their host countries (see Rodrik, 2018). In other words, participation in GVCs might lead to the di¤usion of automation to less developed economies, thereby aggravating the reduction in demand for labor caused by automation in less developed economies. The substitutability between automation and o¤shoring is, however, much less clear-cut in practice. Automation by …rms in developed countries tend to decrease their costs, enhance their productivity and thereby increase their demand for intermediate inputs, many of which continue to be sourced from less developed economies. Whether automation increases or reduces the extent to which …rms in less developed economies participate in GVCs is thus an empirical matter. Building on recent work by Artuc et al. (2018), Chapter 6 of the 2020 World Development Report presents preliminary evidence suggesting that automation in industrial countries appears to have, in fact, 26 boosted imports from developing countries, although the e¤ect is heterogeneous across sectors and countries. Even when automation might not have a negative impact on GVC participation, it is important to emphasize that it is likely to aggravate the e¤ects of increased GVC participation on inequality. The mechanism here is closely related to the capital-skill complementarity e¤ect already mentioned in section 2.2. Similarly to physical capital investments, automation typically complements skilled labor while substituting unskilled labor, thereby exerting upward pressure on the relative demand for skilled workers, and thus on wage inequality. As a result, even when the e¤ects on employment might be attenuated by productivity e¤ects, automation is likely to be associated with a decreasing share of workers in less developed economies directly gaining from GVCs ‡owing through their economies. Automation is often associated with industrial robots, but there has recently been some debates about the role of 3D printing on the future of GVCs. Will 3D printing lead to large volumes of reshoring therefore depressing GVC participation and trade ‡ows? In answering this questions, the mechanisms at play are very much similar to those applying to automation more broadly. The trade-reducing e¤ects of 3D printing are obvious, but one should also take into account their positive e¤ect on productivity and input demand, and the fact that 3D printers do not print goods out of thin air. In addition, by making consumer goods produced by 3D printers cheaper, the demand for these same goods is increased, and as long as 3D printers are not available everywhere, they may generate a short-run spur to trade ‡ows. Indeed, this is consistent with the …ndings of Freund et al. (2018), who show that the dramatic shift of production of hearing aids via 3D printing actually increased international trade in hearing aids by roughly 60 percent. 27 References Acemoglu, Daron, Pol Antràs, and Elhanan Helpman (2007), “Contracts and Technology Adoption,”Amer- ican Economic Review, 97.3: pp. 916-943. Amiti, Mary and Jozef Konings (2007), “Trade Liberalization, Intermediate Inputs, and Productivity: Evidence from Indonesia,” American Economic Review, 97 (5), pp:1611-1638. Antràs, Pol (2003), “Firms, Contracts, and Trade Structure,” Quarterly Journal of Economics, 118:4, pp. 1375– 1418. Antràs, Pol (2015). Global Production: Firms, Contracts, and Trade Structure. Princeton University Press. Antràs, Pol, Luis Garicano and Esteban Rossi-Hansberg (2006), “O¤shoring in a Knowledge Economy,” Quarterly Journal of Economics, Vol. 121, No. 1, pp. 31-77. Antràs, Pol, and Davin Chor (2013), “Organizing the Global Value Chain,” Econometrica 81(6): 2127-2204. Antràs, Pol, Davin Chor, Thibault Fally, and Russell Hillberry, (2012), “Measuring the Upstreamness of Production and Trade Flows,” American Economic Review Papers & Proceedings 102(3): 412-416. Antràs, Pol, Teresa Fort, and Felix Tintelnot (2017), “The Margin of Global Sourcing: Theory and Evidence from U.S. Firms,” American Economic Review, 107.9, 2514-64. Antràs, Pol and Alonso de Gortari (2019), “On the Geography of Global Value Chains,”manuscript Harvard University. Antràs, Pol and Elhanan Helpman (2004), “Global Sourcing,” Journal of Political Economy, 112, pp.552– 580. Antràs, Pol and Robert W. Staiger (2012), “O¤shoring and the Role of Trade Agreements,” American Economic Review 102.7: pp. 3140-83. Arkolakis, Costas, Arnaud Costinot, and Andrés Rodríguez-Clare. “New trade models, same old gains?” American Economic Review 102.1 (2012): 94-130. Artuc, Erhan, Paulo S. R. Bastos, and Bob Rijkers. 2018. “Robots, Tasks, and Trade,” Policy Research Working Paper 8674, World Bank, Washington, DC. Autor, David, David Dorn, Lawrence F. Katz, Christina Patterson, and John Van Reenen (2017), “The Fall of the Labor Share and the Rise of Superstar Firms,” forthcoming Quarterly Journal of Economics. Baldwin, Richard, and Anthony J. Venables (2013), “Spiders and Snakes: O¤shoring and Agglomeration in the Global Economy,” Journal of International Economics 90.2 : 245-254. Bamber, Penny, and Gary Gere¢ (2013), “Costa Rica in the Medical Devices Global Value Chain: Oppor- tunities for Upgrading,”Research Report (August 20), Global Value Chains Center, Duke University, Durham, NC. Baqaee, David and Emmanuel Farhi (2019), “Networks, Barriers, and Trade” mimeo Harvard University. Barjamovic, Gojko, Thomas Chaney, Kerem Cosar, and Ali Hortacsu (2019), “Trade, Merchants and the Lost Cities of the Bronze Age,” Quarterly Journal of Economics 134.3: 1455-1503. Bernard, A.B. and Jensen, J.B. (1995), “Exporters, Jobs, and Wages in US Manufacturing: 1976–1987,” Brookings Papers on Economic Activity, Microeconomics, pp. 67– 112. 28 Bernard, A.B., Jensen, J.B., Redding, S.J. and Schott, P.K., (2018), “Global Firms,” Journal of Economic Literature, 56(2), pp.565-619. Bernard, Andrew B., Andreas Moxnes, and Karen Helene Ulltveit-Moe (2018), “Two-Sided Heterogeneity and Trade,” Review of Economics and Statistics 100.3 (2018): 424-439. Bernard, Andrew B., Stephen J. Redding and Peter K. Schott, (2007), “Firms in International Trade,” Journal of Economic Perspectives, 21:3, 105-130. Blaum, Joaquin (2018), “Global Firms in Large Devaluations”, mimeo Brown University. Brynjolfsson, Erik, Xiang Hui, and Meng Liu (2019), “Does Machine Translation A¤ect International Trade? Evidence from a Large Digital Platform,” Management Science. Caliendo, Lorenzo, and Fernando Parro (2015), “Estimates of the Trade and Welfare E¤ects of NAFTA,” Review of Economic Studies, 82 (1): 1-44. Dedrick, J., Kraemer, K. L., & Linden, G. (2010). “Who Pro…ts from Innovation in Global Value Chains? A Study of the iPod and Notebook PCs,” Industrial and Corporate Change, 19(1), 81-116. Disentangling Global Value Chains,” mimeo Harvard University. de Gortari, Alonso (2018), ‘ De Loecker, Jan, Pinelopi K. Goldberg, Amit K. Khandelwal, and Nina Pavcnik (2016), “Prices, Markups, and Trade Reform,” Econometrica, 84 (2), 445-510. De Loecker, Jan, Jan Eeckhout, and Gabriel Unger (2019). “The Rise of Market Power and the Macroeco- nomic Implications,” forthcoming Quarterly Journal of Economics. Fally, Thibault, and James Sayre (2018). “Commodity Trade Matters,” National Bureau of Economic Research Working Paper No. 24695. Feenstra, Robert C. and Gordon H. Hanson, (1996), “Foreign Investment, Outsourcing and Relative Wages,” in R.C. Feenstra, G.M. Grossman and D.A. Irwin, eds., The Political Economy of Trade Policy: Papers in Honor of Jagdish Bhagwati, Cambridge, MA: The MIT Press. Feenstra, Robert C. and Gordon H. Hanson (1997), “Foreign Direct Investment and Relative Wages: Evi- dence from Mexico’ s Maquiladoras,” Journal of International Economics, vol. 42(3-4), pp. 371-393. Ferrari, Alessandro (2019), “Global Value Chains and the Business Cycle,” mimeo European University Institute. Fort, Teresa C. 2017. “Technology and Production Fragmentation: Domestic versus Foreign Sourcing,” Review of Economic Studies 84 (2): 650–87. Freund, Caroline L., Alen Mulabdic, and Michele Ruta. 2018. “Is 3D Printing a Threat to Global Trade? The Trade E¤ects You Didn’ t Hear About,” Working paper, World Bank, Washington, DC. Gere¢ , Gary (1999), “International Trade and Industrial Upgrading in the Apparel Commodity Chain,” Journal of International Economics, 48:1, pp. 37-70. Gere¢ , Gary, John Humphrey, and Timothy Sturgeon (2005), “The Governance of Global Value Chains,” Review of International Political Economy 12.1: 78-104. Goldberg, Pinelopi K. and Nina Pavcnik (2007), “Distributional E¤ects of Globalization in Developing Countries,” Journal of Economic Literature, 45:1, 39-82. 29 Goldberg, Pinelopi Koujianou, Amit Kumar Khandelwal, Nina Pavcnik, and Petia Topalova (2010), “Im- ported Intermediate Inputs and Domestic Product Growth: Evidence from India,” Quarterly Journal of Economics, 125 (4), 1727-1767. Gopinath, Gita and Brent Neiman (2014), “Trade Adjustment and Productivity in Large Crises,” American Economic Review, 104 (3), 793-831. Griliches, Zvi. (1969). “Capital-Skill Complementarity,” Review of Economics and Statistics : 465-468. Grossman, Gene M., and Elhanan Helpman (2005), “Outsourcing in a Global Economy,” Review of Eco- nomic Studies, 72(1): 135–59. Grossman, Gene M. and Esteban Rossi-Hansberg, “Trading Tasks: A Simple Theory of O¤shoring,” Amer- ican Economic Review, 2008, 98 (5), 1978-97. Halpern, Laszlo, Miklos Koren, and Adam Szeidl (2015), “Imported Inputs and Productivity,” American Economic Review, 105 (12), 3660-3703. Hummels, D., Ishii, J. and Yi, K.-M. (2001), “The Nature and Growth of Vertical Specialization in World Trade,” Journal of International Economics 54, p. 75-96. Johnson, R. C. and Noguera, G. (2012), “Accounting for Intermediates: Production Sharing and Trade in Value Added” , Journal of International Economics, 86(2), pp. 224-236. Karabarbounis, Loukas, and Brent Neiman (2013) “The Global Decline of the Labor Share,” Quarterly Journal of Economics 129.1: 61-103. Kee, Hiau Looi, and Heiwai Tang. “Domestic Value Added in Exports: Theory and Firm Evidence from China,” American Economic Review 106.6 (2016): 1402. Koopman, R., Wang, Z. and Wei, S.-J. (2014), “Tracing Value-Added and Double Counting in Gross Exports,” American Economic Review 104(2), pp. 459-494. Kremer, Michael, and Eric Maskin (2006), “Globalization and Inequality,” unpublished manuscript. Krugman, P. (1980). “Scale Economies, Product Di¤erentiation, and the Pattern of Trade,” American Economic Review, 70(5), 950-959. Krusell, P., Ohanian, L. E., Ríos-Rull, J. V., & Violante, G. L. (2000). “Capital-Skill Complementarity and Inequality: A Macroeconomic Analysis,” Econometrica, 68(5), 1029-1053. Levchenko, Andrei (2007), “Institutional Quality and International Trade,” Review of Economic Studies, 74:3, 791-819. Macchiavello, Rocco, and Ameet Morjaria (2015), “The Value of Relationships: Evidence from a Supply Shock to Kenyan Rose Exports,” American Economic Review, 105:9, pp. 2911-45. Macchiavello, Rocco, and Ameet Morjaria (2017), “Competition and Relational Contracts: Evidence from Rwanda’ s Co¤ee Mills,” Bu¤ett Institute Global Poverty Research Lab Working Pape No. 17-103. Martin, Julien, Isabelle Méjean, and Mathieu Parenti (2018), “Relationship Stickiness: Measurement and Applications to International Economics,” unpublished working paper. Melitz, Marc J. (2003), “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity,” Econometrica, 71:6, pp. 1695-1725. 30 Nunn, Nathan (2007) “Relationship Speci…city, Incomplete Contracts and the Pattern of Trade,” Quarterly Journal of Economics, Vol. 122, No. 2, pp. 569-600. Rodrik, Dani (2018), “New Technologies, Global Value Chains, and Developing Economies,” National Bureau of Economic Research Working Paper No. 25164. Sampson, Thomas (2015). “Dynamic Selection: An Idea Flows Theory of Entry, Trade, and Growth,” Quarterly Journal of Economics 131.1: 315-380. Sturgeon, Timothy, Johannes Van Biesebroeck, and Gary Gere¢ (2008), “Value Chains, Networks and Clusters: Reframing the Global Automotive Industry,” Journal of Economic Geography, 8:3, pp. 297- 321. Verhoogen, Eric A. (2008) “Trade, Quality Upgrading, and Wage Inequality in the Mexican Manufacturing Sector,” Quarterly Journal of Economics 123.2: 489-530. Yang C., Dietzenbacher E., Pei J., Chen X., Zhu K., Tang Z. (2015), “Processing Trade Biases the Mea- surement of Vertical Specialization in China,” Economic Systems Research, 27(1):60-76. Wang, Z., S. Wei and K. Zhu (2013), “Quantifying International Production Sharing at the Bilateral and Sector Levels,” NBER Working Paper, No. 19677. ’ 31 Appendix A Country-wide (aggregate) impacts on economic growth, productivity, and industrialization Discussion of impact Outcomes from greater GVC participation Broad measure: neoclassical (trade in Narrow measure: relational (scale, market tasks) – power, and governance of the value chain) Aggregate domestic value Efficiency gains from optimal allocation added: gains from trade larger of input factors is magnified compared than in a world without to trade fragmentation of production Effects on exports: greater Trade costs, trade frictions, and sensitivity to trade costs, trade uncertainty make foreign inputs to frictions, and uncertainty. production more expensive Lower export sensitivity to Suppliers to GVCs located in export Lead firms in the GVC fill gaps in technology, negative domestic environment processing zones benefit of a better training and credit while ensuring continued (exchange rate, access to credit, trade and business environment demand for exports at scale. skills, and quality of institutions) Firm productivity: positive Better access to a greater variety, Relationship-specific investment (inter-firm effect of backward GVC better quality, or to cheaper inputs and intra-firm) participation Firm capabilities: greater because of flow of technology, goods and Learning by exporting/importing as in knowledge spillovers than in credit to environments with limited trade, by at the task level trade.. contractual security ..but mixed effects on incentives Greater spillovers to domestic technology to innovate and technology frontier firms but lower incentives to diffusion innovate in some governance structures Bargaining may make value added low for Industrialization through GVCs: developing countries whose only comparative simpler to industrialize but advantage is low cost production returns to doing so may not be Modularization of production and platform- as high as in the past, or as fast based business models lower entry costs to to materialize technology intensive industry for low skilled firms but make upgrading and product differentiation harder. 32 1 Employment and wages Discussion of impact Outcomes from greater GVC participation Narrow: relational (scale, market power, Broad: neoclassical (trade in tasks) and governance of the value chain) Employment: Positive effects The employment effects may differ if the come primarily through indirect multinational enterprise outsources to links and service inputs, and Reallocation of workers to less routine tasks long-term suppliers versus short term from overall expansion of suppliers economic activity… …and improvement in working conditions for workers in firms Global buyers require higher compliance most closely connected to the with health, safety, and treatment of GVCs of international brands workers (but not for those at the periphery of GVCs) Wage polarization: Skill Sensitivity to quality associated to trade in premium increases in both tasks and greater servicification of Matching between buyers and sellers advanced and developing production internationalize labor markets: leads to war for talent countries, with crunch for both demand for scarce talent and wage middle-income workers… competition at the low-skill end increase Search for quality and scale leads to ..but shift in labor demand automation of production, which further towards skilled jobs.. tilts labor demand towards skilled workers ..and more skewed firm distribution (discussed above) Imperfect labor markets and rent-sharing leads to more skewed wage distribution. 33 Other distributional impacts and inequality Discussion of impact Outcomes from greater GVC participation Relational (scale, market power, and Neoclassical (trade in tasks) governance of the value chain) Disproportionate bargaining power of large firms (volume discounts, monopoly Market concentration: large Complementarities between imports and and monopsy power over prices). firms benefit disproportionately exports increase the benefits and risk for Bargaining powers is heightened for but are also more exposed to large firms platform companies due to hyperscale, shocks network effects and data analytics abilities) Allocation of gains to consumers: cost reductions are Disproportionate bargaining power of often only partially passed on to large firms prices and particularly so by large companies Place-specific effects: benefits Value added and employment may disproportionately go to decrease for locations trying to remain locations with deep and broad competitive while facing fixed input and pool of skilled workers, and output prices good infrastructure Labor intensive GVCs employ disproportionately female workers, but Gender considerations women are less represented in higher segments of the value chain Global buyers require higher Future generations: environmental compliance , but oversight environmental sustainability is is effective for firms most closely needed to ensure good connected to the international brands outcomes also for future (but not for those at the periphery of generations GVCs) Countries with weak institutions are Welfare and inequality: affected negatively by greater integration institutions matter in GVCs or trade openness 34