VoxEU Column International trade

Has production become more fragmented? International vs domestic perspectives

As the oft-cited iPhone example illustrates, production has become increasingly fragmented across countries. This column presents recent research, however, suggesting that this trend may be reversing for manufacturing plants in the US. It shows that intermediate goods account for a decreasing fraction of output value, while industries that are closer to the final consumer contribute to an increasing share of GDP.

Production seems more complex and fragmented today than ever before. For instance, airplanes are made of zillions of parts involving many suppliers from various countries (see eg www.newairplane.com). It has also become difficult to keep track of the production of relatively more simple goods such as the Barbie doll (Feenstra 1998) or the iPhone (see Xing 2011 on this site).

Recent news has also pointed to the cross-border nature of production and the interdependence of countries. Last year’s tsunami in Japan has stopped major automobile factory lines all over the world. A few months ago, floods in Thailand disrupted supply chains in the hard drive and car industry – the hardest hit companies included Western Digital and Toyota. Apparently simple components can also stir some trouble. A screw supply shortage from an Italian supplier severely hit the French car manufacturer Peugeot last summer. International trade is no longer just about final goods but involves an increasing share of intermediate goods. Indeed, as Grossman and Rossi-Hansberg say “it’s not wine for cloth anymore” as production chains are now organised across many countries (Grossman and Rossi-Hansberg 2006).

Production fragmentation across plants

While international trade economists have emphasised the fragmentation of production across borders, little is known about the fragmentation of production across plants, whether these plants are within the same or different countries. This is not a second-order issue: input-output linkages (‘vertical’ linkages) have been shown to play a key role in models of economic development, and also matter for agglomeration (Ellison et al 2010 cite input-output linkages as the most important force of agglomeration), business cycles transmission, etc.

A full picture on fragmentation across plants is impossible to get without transaction-level data between suppliers and buyers, which are unfortunately not available. Nonetheless, interesting information on the length of supply chains can be extracted from input-output tables (Fally 2011). From these tables, we could learn more about the degree of vertical fragmentation, ie the number of plants involved sequentially in production (number of ‘stages’). In particular, we can obtain a measure of the average number of plants involved sequentially weighted by the contribution of each plant to value-added.

The evolution of this measure, computed using benchmark input-output tables for the US between 1947 and 2002 (available about every 5 years) is shown in Figure 1.

Figure 1. Value-added weighted average of the number of production stages (excluding services)

Before looking at the evolution, a first take from this measure is that production is not very fragmented. On average, value-added has gone through less than two stages. This hides substantial heterogeneity across industries: food and transportation-equipment industries are more vertically fragmented, while services are less fragmented. Interestingly, high-tech industries are not more fragmented than apparently more simple industries such as the food industry. (An explanation is that production of more complex goods is more difficult to coordinate and to break down across plants.)

Moreover, this measure shows an overall decrease over the past 60 years for the US, which means that production today is less vertically fragmented than in 1947. This decline can be partially explained by the shift of production and consumption towards service industries, since service industries are less vertically fragmented and more integrated. But this is not the whole story – a similar decline in vertical fragmentation can be observed by focusing on commodities and manufacturing goods. Most industries exhibit a decline across years on average: food industries, chemical, aircraft manufactures, etc. Exceptions include electronic industries, which became more vertically fragmented, and auto industries.

Part of the measured decline in vertical fragmentation corresponds to a shift of value-added towards industries that are closer to the final consumer (ie more downstream). Equivalently, industries that are further from final demand (basic commodities, parts and components) now account for a smaller share of total manufacturing value-added. The growth of value-added has been relatively stronger in industries that are more intensive in advertising, industries with more differentiated goods, and less capital-intensive industries – those are industries with fewer production stages before reaching final demand.

The hugely successful iPods and iPhones provide examples of value chains where the downstream end of the chain (Apple itself) contributes to a large fraction of the final value of these products. While the production of iPods and iPhones involves many suppliers in different countries (it is difficult to even count them!) the value-weighted number of sequential production stages is not large (between 2.5 and 3). For both the iPod and the iPhone, almost half of the value is being ‘added’ by Apple1. This corresponds to marketing, design, distribution, etc. Putting Apple at the end of the value chain, this means that production is not in fact very fragmented if we look at where the value comes from. A few key electronic components capture a large value (eg display, memory) with a large part being directly added by these first-tier suppliers. A multitude of other components are involved but they correspond to a much smaller fraction of value-added and are assembled in China for a very small cost.

At least two remarks are in order. First, ‘vertical’ fragmentation is only one aspect of fragmentation, which might be increasing with respect to other dimensions. Second, the measure of vertical fragmentation used in my study (Fally 2011) may be sensitive to changes in the relative price of intermediate vs final goods.

What I define as vertical fragmentation only captures the number of sequential stages. If production chains take the form of ‘snakes’ (in keeping with the Baldwin and Venables 2011 analogy), then the index above measures the length of snakes. It could be possible that Internet and communication technologies have allowed production chains to become more like ‘spiders’ in the sense that more suppliers are involved simultaneously (in ‘parallel’ if we draw an analogy to electric circuits).

My results seem to indicate however that inter-industry linkages – measured as the number of different products purchased as intermediate goods by each downstream industry – evolves in the same direction as the number of sequential stages, showing no evidence of a substitution between snake- and spider-organisational forms of production. It would be interesting also to confirm these results with plant-level data, but the time coverage allowed by such data is usually very short (see Fort 2011 for a recent analysis of fragmentation across plant using US Manufacturing Census data).

We could wonder whether the shift of value-added towards final stages corresponds to an erosion of bargaining power of component suppliers, inducing components to become relatively cheaper than final goods. However, price data provided by the Bureau of Labour and Statistics (BLS) suggest that parts and components have not become relatively cheaper than final goods in the US over the past decades.2 In short, the shift of value-added towards final stages seems to reflect an evolution of technology and preferences rather than a change in relative prices or bargaining power along production chains.

Can globalisation explain the decline in vertical specialisation and the shift of value-added towards final stages? While the number of sequential stages may have decreased, an increasing fraction of these stages is now performed abroad (Hummels et al 2001). In fact, industries with a larger share of imported intermediate goods are those where the number of stages has increased relatively more. Foreign suppliers are not simply substitutes for domestic sources but now also supply inputs that used to be produced within the same plant. Hence, globalisation could explain an increase in vertical fragmentation. This effect is however not very large quantitatively, otherwise the number of production stages would have increased over the past decades.

Another related question is whether the shift of value-added towards final stages in the US could be explained by a specialisation of the US economy in more downstream activities. Evidence on this matter is scarce and does not support this hypothesis. Hummels et al (2001) have documented an increasing extent of ‘vertical specialisation’ (ie specialisation at specific stages along the production chains), but there is only limited evidence on whether different groups of countries specialise in more downstream vs upstream activities. To answer this question, in work with colleagues (Antras et al forthcoming) and independently (Fally 2011), I measure the average ‘upstreamness’ (number of stages to final demand) of a country’s exports. It appears however that the average upstreamness of US exports – and those of richer countries in general – is not significantly different from other countries’ exports. Moreover, the shift towards final stages documented for the US seems to be a common trend: global production has become more downstream over the past 50 years.


Antras, Pol, Davin Chor, Thibault Fally, and Russell Hillberry (forthcoming), “Measuring Upstreamness of Production and Trade”, American Economic Review Papers & Proceedings.

Baldwin, Richard, and Anthony Venables (2010), “Relocating the Value Chain: Offshoring and Agglomeration in the Global Economy”, NBER Working Papers 16611.

Ellison, Glenn, Edward L Glaeser, and William Kerr (2010), “What Causes Industry Agglomeration? Evidence from Coagglomeration Patterns”, American Economic Review, 100(3):1195-1213.

Fally, Thibault (2011), “On the Fragmentation of Production in the US”, University of Colorado-Boulder mimeo, October.

Feenstra, Robert (1998) “Integration of Trade and Disintegration of Production in the Global Economy", Journal of Economic Perspectives, Fall: 31–50.

Fort, Teresa (2011) “Breaking up is hard to do: why firms fragment production across locations”, University of Maryland, mimeo.

Grossman, Gene and Esteban Rossi-Hansberg (2006), "The Rise of Offshoring: It's Not Wine for Cloth Anymore", The New Economic Geography: Effects and Policy Implications, Jackson Hole Conference Volume, Federal Reserve Bank of Kansas City.

Hummels, Davis, Jun Ishii, and Kei-Mu Yi (2001), “The Nature and Growth of Vertical Specialization in World Trade", Journal of International Economics, 54(1):75-96.

Kraemer, Kenneth L, Greg Linden, and Jason Dedrik (2011a), “Capturing Value in Global Networks: Apple’s iPad and iPhone”, The Centre for Research on Information Technology and Organizations, July.

Kraemer, Kenneth L, Greg Linden, and Jason Dedrik (2011b), “Innovation and Job Creation in a Global Economy: The Case of Apple’s iPod”, The Centre for Research on Information Technology and Organizations.

Xing, Yuquing (2011), “How the iPhone widens the US trade deficit with China”, VoxEU.org, 10 April.

1 See the work of Kraemer et al (eg 2011a and 2011b) for detailed studies of these two products.

2 We should however remain cautious given that data on the evolution of prices are not available at a sufficiently disaggregated level and do not perfectly account for the introduction of new varieties (eg new types of components provided by overseas suppliers).

5,144 Reads