Asian Trade Flows: Trends, Patterns, and Projections

This paper provides trade flow projections for major Asian developing economies (ADEs) up to 2030 against the backdrop of an in-depth analysis of policy shifts and trade patterns over the past 4 decades. Merchandise trade of ADEs has grown at a much faster rate in the global context, with a distinct intraregional bias. Global production sharing has become a unique feature of the economic landscape of the region, with the People’s Republic of China playing a pivotal complementary role as the premier assembly center within the regional production networks. According to the projections made within the standard gravity modelling framework, total real non-oil trade of ADEs would increase at an average annual rate of 8.2 during the next 2 decades, with a notable convergence of individual countries’ rates to the regional average. The share of intraregional trade in total non-oil trade would increase steadily from 53% in 2010 to 58% in 2030. The trade-to-GDP ratio would increase from 39.4% and 74.4% between these 2 years. These predictions need to be treated with caution as they are based on the assumption that the trade structure pertaining to the estimation period will remain unchanged in the next 2 decades.


INTRODUCTION
The purpose of is paper is two-fold: to examine emerging trends and patterns of merchandise trade in Asia over the past four decades and to generate projection for Asian trade. Relating to the latter factors, a key theme running through this paper is the implications of global production sharing -that is, the geographic separation of activities involved in producing a good (or service) across two or more countries-for the debate on regional versus global integration of these countries. 3 The paper is organized as follows. Section 2 briefly surveys the policy environment shaping countries' participation in the global economy. Section 3 examines Among other issues canvassed, we also aim to contribute to the debate as to whether the emergence of China as the world's fastest growing industrial economy will crowd out other countries' opportunities for integrating into the regional and global economy through fragmentation-based specialization.
1 I am grateful to Archanun Kohpaiboon for an excellent job done in colleting/tabulating trade data and to Majeed and Shahbaz Naseer for invaluable help with econometrics. 2 These countries account for over 96% of total trade of all countries in Northeast and Southeast Asia. 3 An array of alternative terms have been used to describe this phenomenon, including 'international production fragmentation', 'vertical specialisation', 'slicing the value chain and 'outsourcing'. general patterns of trade since 1970, encompassing trade flows over time in aggregate, by major partners, and by major commodity groups. This section also examines geographic patterns of trade, with emphasis on the implications of the growing importance of global production sharing. Section 4 presents the estimates of trade equations for the twelve countries and preliminary trade flow predictions based on these estimates. The final section summarizes the main findings and draws out some general inferences.

THE POLICY CONTEXT
Rapid growth and structural change in DAEs over the past four decades has been underpinned by notable reduction in barriers to international trade. There has been significant trade liberalisation in PRC; Indonesia; the Republic of Korea; Malaysia; the Philippines; Taipei, China; and Thailand since the mid-1980s. India and Vietnam embarked on reforms in the early 1990s. Trade liberalisation in all these countries has been predominantly unilateral and non-discriminatory, and was also aided by multilateral liberalisation under the General Agreement on Trade and Tariff (GATT) and its successor the World Trade Organization (since 1991). Historically countries in Asia have not been enthusiastic towards preferential trading agreements (PTAs), with the exception of some trade preferences within the ASEAN region which presumably had only trivial trade flow effect. Since then, the advent of the European Union and the rise of regionalism in North America led to a proliferation of PTAs in the region. It is too early to assess the trade flow implications of these PTAs, but the available circumstantial evidence suggest that so far the impact would have been rather small. The preference utilization rates of the PTAs remains very low, given the narrow preference margins resulting from the on-going process of multilateral and unilateral tariff reductions and the administrative cost and complications involved in meeting the rules of origins involved for benefiting from tariff preferences (Plummer 2007, Baldwin 2006). This section aims to provide an overview of the process of trade opening in the region and the current state of openness to trade. The discussion is based on three broad sets of indicators of openness to international trade, namely the Sachs-Warner binary classification, revealed trade orientation measured as exports or total trade (exports + imports) as a percentage of GDP, the average tariff rate. All these measures have their shortcomings (as discussed below), but together they enable us to assert with reasonable confidence whether an economy is broadly open. Table 1  Consequently, from about 2005 the 'socialist economic system' characterization remains the only Sachs-Warner closed-economy criterion applicable to these countries. 4 See Table 2, Note 1 for the criteria used in identifying the year of demarcation. The original Sachs-Warner classification covered 100 countries (78 developing and 22 developed countries) over the period 1945 to 1994. Wacziarg and Welch (2003) have updated the classification to 2000, while expanding the coverage to 131 countries. 5 After China's accession to member ship in January 2001 and Vietnam in January 2006, all 12 DACs are now bound by WTO multilateral disciple. in the region have been compensating exporters for duties paid on imported inputs. These typically take the form of duty exemptions or drawbacks, or the establishment of export processing zones. However there is evidence that almost all countries in the region have dismantled most of binding NTBs and scaled down selective incentives for export producers (Panagariya 2007, Krueger 2010).
There has been a universal trend towards lower tariffs across the ADCs over the past two decades. In some countries, the declines have been very large, more than Singapore have of course always had negligible protection. The others mostly range up to 10%. Importantly, though not adequately recorded here, the East Asian economies were much quicker than the two major South Asian countries to adopt partial reforms that enabled exporters to operate on an effective free-trade footing.
The export/GDP 6 6 Exports are generally regarded as preferable to total trade (or imports) as the numerator in calculating this ratio because restrictiveness of a given country's policy regime is presumably better captured by export performance. ratio, the standard revealed openness measure, is reported in Table 3. The trade/GDP ratio is a comparison between a net and a gross concept; trade is measured in gross terms (intermediate material inputs + value added) whereas GDP is essentially measured on a value-added basis (that is, net of intermediate material inputs).
Thus, the measured change in trade orientation is sensitive to changes in import intensity of export production. As we will see later in this chapter, over the past decade there has been a palpable shift in the export composition away from primary products and towards labour intensive light manufacturing and, more recently, the ongoing process of international production fragmentation within high-tech industries. The increase in measured trade orientation could partly reflect the fact that these new product lines are relatively more import-intensive compared to the former. Another limitation for crosscountry comparisons is that the ratios need to be adjusted for size, in recognition of the fact that small countries by definition will trade more than larger ones. However, as we have already noted, the usefulness of this measure in its own right as an indicator of trade openness is limited because, by construct, it is driven by structural shifts in production and trade patterns. Of particular relevance in this connection is the ongoing process of international production fragmentation, which involves small value added additions at various stages of the production process of a given final good in various countries, thus resulting in inflated trade values relative to GDP. Even in small countries, at least 60% of GDP is generated by non-tradable sectors. Thus an export share of much more than 30-40% can arise only when export production involves adding fairly small amounts of value to imported inputs (Krugman 1995, p. 335). These limitations notwithstanding, the inter-country differences and the time profile revealed by this measure are broadly consistent with those we have already observed.
Openness to trade is a necessary but not sufficient condition for successful global economic integration. Equally important is the conduciveness of the business environment. International competitiveness requires high quality infrastructure, both hard and soft, especially for successful participation in time-sensitive global production and purchasing networks. Labour markets need to reflect underlying supply and demand conditions, with wage growth and differentials driven by productivity. Prudent macroeconomic management is required to provide a stable and predictable commercial policy environment, and to ensure that exchange rate outcomes do not impair competitiveness. Above all, political stability and policy certainty figure prominently among prerequisites for profitable long-term investment, particularly for MNEs. The ease of doing business data summarized in Table 4 confirm the superiority of East Asia over South Asia, with the four Asian NIEs together with Malaysia and Thailand ranking the highest; the two city-states (Hong Kong, China and Singapore) have the highest ranking. China ranks ahead of India by a wide margin. The differences between Northeast and Southeast Asia are not significant. Nevertheless this ranking exercise is at best indicative and it also presumably reflects the development paradigms of the institution preparing the data. Communist states such as China and Vietnam fare poorly, in spite of the sweeping reforms of the past two decades, and even when they are obviously commercially attractive to foreign investors. Naturally the country ranking in terms of the LPI (Table 5) is remarkably similar to that based on the Doingbusiness database. Singapore tops the overall logistics quality raking. The other major transshipment hub in the region, Hong Kong is forth on the global ranking and second among DACs. The data also reveal significant differences among countries at similar levels of development. Countries in Northeast and Southeast Asia have higher rankings.
Other East Asian contraries, including China, also compare very favorably compared to developing countries in other parts of the world.

TRADE PATTERNS
Rapid trade expansion has been the hallmark of Asia's rise in the global economy. The combined share of ADEs in total world merchandise exports increased continuously from less than 5% in 1970 to 22% by 2008 (Figure 1a). 7 The region accounted for over a third of total increment in world exports over this period. Focusing on world non-oil trade (trade net of oil and gas) to gain a better comparative picture, 8 7 Trade data throughout the paper are measured in current US dollars unless otherwise indicated. the increase in ADEs' world export share turns out to be sharper, from about 4% in the early 1970s to over nearly 25% in 2008 (Figure 1b).
The rise of the PRC has been a dominant factor behind the share increase in ADEs world market shares from about the early 1990s. But the upward trend in world market shares of the other countries that began in the early 1970s has continued unabated throughout the period. Thus, on first inspection, there is no indication of China 'crowding out' its neighbours─ China's market share gains has been at the expense of that of the rest of the world, not from the rest of Asia. This observation is consistent with the inferences coming from a number of recent studies which have systematically examined the impact of China's rise on exporter performance of the other countries in the region (Athukorala 2009, Greenaway at al. 2008Eichengreen et al. 2007 ) At the individual country level, market share gains have varied notably over time (Table 1). 9 Rapid export growth in Developing Asia (DAC) has been underpinned by a pronounced shift in export structure away from primary commodities and toward manufactures (Table 6)). From about the early 1990s manufactures accounted for over a four-fifths of total merchandise exports from these countries, up from 84.3% in 1969/70 four decades ago. Given the nature of their resource endowments, the four Asian newlyindustrialized economies (NIEs) (Hong Kong, Taiwan, Korea and Singapore) relied very heavily on manufacturing for export expansion from the outset. However, beginning in the 1970s, a notable shift towards manufacturing is observable across all countries, at varying speeds and intensity. The combined shares of the ASEAN countries other than Singapore increased from a mere 11% to 71.0% between these two time points. Among individual countries Indonesia and Vietnam has a significantly lower share of Among the nine largest ADEs economies only Hong Kong, Indonesia and the Philippines have smaller world trade shares than India, which is by far the dominant South Asian economy. Notwithstanding the notable export expansion in recent years, India still accounts for a mere 1.1% of total world trade, equivalent to less than 5% of DACs total trade. Pakistan is the only country among the DAC12 to record a decline in market share (other than Hong Kong, whose export production base has 'migrated' to the Mainland China since about the early 1990s). 9 In this and other trade data tables, data are presented as two-year averages to smooth out the impact of yearly fluctuations in trade. manufactures in their exports, reflecting both their comparative advantage and their later adoption of export-oriented industrialization strategies.
Within manufacturing, machinery and transport equipment (SITC 7) have played a pivotal role in the structural shift in the export composition of DACs (Tables 7 and 8).
The share of machinery and transport equipment in the export structures of some of the more industrialized economies of East Asia is particularly high. By contrast, that for Indonesia, Vietnam and all of South Asia is much smaller. Within the machinery and transport equipment category, ICT products have been the most dynamic component of Asian export expansion. By 2007/8, over 58% of total world ICT exports originated from Asia, up from 30.8% in 1994/5 (Table 8); China accounted for 25.4% of total world ICT exports, up from 4.2% in 1994/5. In electrical goods, China's world market share increased from 3.1% to 20.6% between these two years.
Asia's share in the other main product categories has also increased over time, though at a slower rate. Of particular interest here is the notable increase in the region's share in miscellaneous manufacturing. This mostly consists of standardized labourintensive manufactured goods, in particular clothing and footwear. China has accounted for much of this increase but, in contrast to ICT exports, the geographic participation has been broader, by Indonesia, Vietnam and India (and also a number of low-wage countries in Southeast and South Asia, including Indonesia, Vietnam, India, Sri Lanka, Bangladesh, and Cambodia not covered in the table) have all recorded impressive gains in market share.

Global production sharing
The fast growth of machinery trade in Asia has been driven by rapid growth of international fragmentation of production in world trade and the increasingly deep Semiconductors and Texas Instruments, to set up plants in Singapore to assemble semiconductor devices (Athukorala 2008, Goh 2003. From about the late 1970s, the MNEs with production facilities in Singapore began to relocate some low-end assembly activities in neighbouring countries (particularly in Malaysia, Thailand and the Philippines) in response to the rapid growth of wages and land prices. Many newcomer MNEs to the region also set up production bases in these countries, bypassing Singapore.
From about the early 1990s the emergence of China as the 'global factory' of electrical and electrical goods assembly based on parts and component imported from other countries has contributed to rapid expansion of production networks in the region. More recently regional production networks have begun to expand to Vietnam. Over the past three decades, the process of global production sharing has created a new division of labour among countries in the region, based on skill differences involved in different stages of the production process and relative wages, and improved communication and transport infrastructure (Ando and Kimura 2010). As we will see below, the formation of production networks has dramatically transformed the spatial patterns of international trade in the region, with a notable 'magnification' effect on recorded trade flows operating through multiple border-crossing of parts and components on the expansion of intra-regional trade.
The most ubiquitous indicator of the intensity of fragmentation-based specialization in world trade is the share of parts and components in total manufacturing trade ( Table 9, Panel A). 10 In an inter-country comparison, there is a remarkable similarity of component shares in manufacturing exports and imports across all East Asian countries other than China, reflecting overlapping specialisation patterns in component assembly and testing Over the past two decades there has been a sharp increase in the share of parts and components (henceforth referred to as 'components' for brevity) in world manufacturing trade from 19.3% in 1992/3 to 27.1% in 2006/7 (Table 9). And this share has increased at much faster rate in DACs, from 17.3% to 34.0%. Components share is particularly high among the countries in ASEAN, with all countries in East Asia recording shares well above the world average. The combined component share in manufacturing exports from the ASEAN countries in 2006/7 amounted to 44.2%, up from 22.7% in 1992/93. In spite of its intrinsic comparative advantage, India still remains a minor player in this new form of international exchange (Krueger 2010). 10 Henceforth, for the sake of brevity, the term 'components' in place of 'parts and components' and 'machinery' in place of 'machinery and transport equipment' are used. among countries in the region. China's manufacturing trade patterns differ from its East Asian neighbours. In particular, the components share in its total manufacturing imports of China (44% in 2006/07) is much larger compared to the corresponding share in its manufacturing exports (25.6%). This difference between China and the other countries in the region is consistent with our earlier observation that China's rise in world trade has brought about a notable shift in the division of labour within regional production networks, with ASEAN countries playing an increasing role in producing parts and components for the rapidly growing final assembly activities in China. The bulk of components used in final assembly in PRC come from other countries in the region. At the same time, final goods (total exports minus components) account for an overwhelming share of PRC's exports to the rest of the world, mostly to the US and EU.
A notable outcome of the rapid expansion of production networks encompassing an increasing number of countries in the region has been the growing importance of components in intra-regional trade compared to extra-regional trade of the countries in the regions (Table 9, panel B). For instance, in 2006/7 components accounted for 53.9% of intra-DAC exports (59.5 of imports) compared to 34.0% in the regions total exports (44.2% of total imports). Component share in intra-DAC trade is much higher compared to that of intra-regional trade in NATA and EU15. As we will see below this has a significant implication for analyzing inter-regional versus extra-regional trade integration of countries in the region.

Direction of Trade
What have been the implications of the structural change in trade patterns of DACs for the geographic composition of trade? Has the relative importance of intra-regional and extra regional markets changed over time?
There is a vast literature dealing with these issues, which unequivocally points to a persistent increase in intra-regional trade in East Asia, whether or not Japan is included, from about the early 1980s. 11 11 See Yoshitomi (2007) and Park and Shin (2009) and the works cited therein. This evidence figures prominently in the current regional debate concerning the establishment of regional trading arrangements covering some or all countries in East Asia. In particular, the proponents of expanding AFTA to encompass Japan, China and South Korea (the ASEAN+3 proposal), and more broadly towards an 'Asian Economic Community', and of various proposals for monetary integration in the region, often refer to deepening economic interdependence, as reflected in intra-regional trade among these countries, as evidence of likely success of these initiatives. Another implication of the highly publicized apparent trade integration in the region was the so called 'decoupling' thesis, which was a popular theme in the Asian policy circles in the first decade of the new millennium until the onset of the recent financial crisis. This thesis held that East Asian region had become a self-contained economic entity with potential for maintaining its own growth dynamism independent of the economic outlook for the traditional developed market economies.
As can be seen in figure 2, intra-regional trade patterns of DACs are remarkably consistent with this 'conventional' view. Intra regional trade (export + imports as a percentage of GDP) increase continuously from about 20% in 1985 to over 36% by the mid 1990s (When the data coverage is expanded to Japan (that was commonly done in previous studies) the increase was even sharper from 23% to nearly 50%). Both imports/GDP and exports/ GDP shares have closely moved together. However, the time pattern has changed notably from then on. While, import to GDP ratio has continued to increase at even a faster rate, the export-to-GDP ratio has virtually stagnated or recorded a mild decline in some years. Consequently, the rate of increased in the trade-to-GDP ratio has remarkably slowed.
This notably asymmetry in the growth of intra-regional exports and imports (and the consequent slow expansion in total intra-regional trade) is reflection of East Asia's unique role within global production networks, and in particular China's role as the premier assembly centre within these networks based on parts and components procured from the rest of East Asia. As can be seen in Figure 3, the intra-regional share of China's exports has continuously declined from about the early 1990s (reflecting its increased reliance on extra regional markets for final (assembled) goods, in a context where the intra-regional share in imports has continuously increased. Given that China is catering for a much larger extra-regional market for final (assemble goods), its total intra-regional trade share continuously declined. So far we have looked at only the implications of the emerging asymmetry in intra-regional exports and imports arising from the ongoing process of global production sharing for intra-regional trade integration. In addition, and more importantly, increased 'component intensity' of trade flows in the region has an important direct implication for a meaningful assessment of the relative importance of intra-regional versus extra-regional trade for the growth dynamism of countries in the region.
We have noted two important peculiarities of trade in East Asia compared to global trade patterns. First, component trade has played a much more important role in trade expansion in East Asia compared to the rest of the world. Second, trade in components accounts for a much larger share in intra-regional trade than is the case for the rest of the world. Given these two peculiarities, conventional trade flow analysis which does not make a distinction between components and final goods is bound overstate the relative importance of intra-regional trade, as compared to global trade, for growth in East Asia. This is because growth based on assembly activities depends on the demand for final goods, which in turn depends on extra-regional growth.
To illustrate these arguments intra-regional trade shares estimated separately for total manufacturing trade component trade and final manufacturing trade (that is, total manufacturing trade less component trade) are reported in Table 11. The table covers trade in East Asia 12 12 There is no notable difference between intra-regional trade patterns of Asia (East Asia + South Asia) and East Asia given that South Asia accounts for a tiny share in total Asian trade. and three sub-regions therein which relate to contemporary Asian policy debate on regional integration. Data for NAFTA and EU are reported for comparative purposes. Estimates are given for total trade (imports + exports) as well as for exports and imports separately in order to illustrate possible asymmetry in trade patterns resulting from East Asia's increased engagement in fragmentation-based international exchange. According to our estimates, on the export side, the intra-regional share of final goods declined sharply from 45.0% in 1996/7 to 28.7% in 2006/7, whereas intra-regional import share increased at a slower rate, from 41.4% to 46.5%.
Consequently the trade-to=GDP ratio declined from 43.2% to 37.6% between these two time points.
In sum, these data support the hypothesis that, where fragmentation-based trade is expanding rapidly, the standard trade flow analysis can generate misleading inferences regarding the process of economic integration through trade. When data on assembly trade are excluded from trade flows, these estimates suggest that extra-regional trade is much more important than intra-regional trade for continued growth in DACs. Thus, the rising importance of product fragmentation seems to have strengthened the case for a global approach to trade and investment policymaking rather than a regional one. This

MODELLING TRADE FLOWS
This section reports results of an econometric exercise undertaken to predict trade flows of the 12 countries to 2030 based on trade equations estimated at the country level using data for the period 1985 to 2008. The analytical tool used in this section for modeling trade flows of DACs is the gravity model, which has become the 'workhorse' for modeling bilateral trade flows. The standard gravity model postulates that trade between two countries, like the gravitational force between two masses, is a function of their economic size and the geographic distance between them. 13 We augment this basic model by adding a number of explanatory variables which have found in previous studies 14 to improve the explanatory power of the estimated trade equations.
Our specification of the gravity model is:

CLN
A dummy variable which is unity if i and j a common language (a measure of cultural affinity) (+) 13 For an introduction to the gravity model and recent methodological and theoretical advances in its applications to trade flow modeling see Bergeijk and Brakman (2010). 14  The coefficients of the two standard gravity variables (GDP in pairs, and the distance) are statistically significant with the hypotheses signs in all cases. The coefficient of the distance variable is well within the range of 0.7 to 1.20 commonly found in various gravity model applications. The coefficient of GDP in pairs is consistency closer to unity. The LPI performs remarkably well in explaining both imports and export, with statistically significant and positive coefficients in all cases. The other control variables are not uniformly significant across all countries, both on import and export sides. We have retained these variables in a given equation only it the coefficients carried the expected sign with a t-ratio of more than unity. In terms of the overall fit the export equations generally performs better (with R 2 s of closer to 0.80) than the import equations (R 2 s of around 0.65). As one would expect, estimated equations (both exports and imports) for non-oil trade generally exhibit a better overall fit compared to those for total trade.

Trade projections
The estimated equations are used for predicting trade flows from the period 2010-2030.
The methodology involved estimating total trade for each country as the sum of bilateral trade flows estimated using the estimated import and export equations. In making these projections, GDP projections generated in this research project are used for the DAC countries. GDP projections for the other trading partner countries of DAC countries come from the USDA database.
Trade flow projections (US$ billion) for total merchandise trade and non-oil merchandise trade for the 1 DAC countries are reported in Table A2 in the statistical appendix. It is important note that these perditions are based on the assumption that the past trade patterns will continue unchanged for the ensuing two decades. In reality, the Annual growth rates of projected trade flows are reported in Table 12. Total nonoil exports from DACs are projected to growth at an annual rate of 7.9% during 2010-20 and 9.9% during 2020-30 (Table 12, panel a). Predicted growth rates of imports for the two sub-periods are 8.4% and 6.8% respectively. A comparison of the estimates on export and import sides points to a mild tendency for narrowing of the overall trade deficit of the region over the years. Export growth of China is projected to decline from 10.4% during 2010-2020 to 8.1% between the two decades. Predicted import and export growth rates varies notably among the 12 countries, which significant slowing down of trade expansion in the three East Asian NIEs (Korea, Taipei, China and Singapore) compared to the other countries. Overall the estimates point to a notable convergence in the rate of trade expansion among the countries over the years.
Intra regional non-oil trade in ADCs are projected to growth at a slightly faster rate compared to their overall (global) trade (compare figures reported in Panel A and B in Table 12). Exports from these countries to regional markets during 2010-30 will grow at 9.2% compared to 8.5% growth in the region's total exports. The comparable figures on the import side are 8.4% and 7.1% respectively. During these two decades China's intra-regional exports are protected to grow at 10.5 compared to an overall export growth rate of 9.2%. In all other countries too intra-regional trade would grow at a faster rate compared to overall trade, driven by trade expansion associated with faster economic growth.
Reflecting these growth rate differentials between intra-regional and total trade, intra-regional share in total trade of DACs will increase continuously during 200-2030 (Table 13, Figure 3). Intra-regional non-oil export share in DACS would reach 64.9% in 2030 from 56.9% in 2010. The increase on the import-side would be from 53.5% to 59.5%. Intra-regional trade shares are predicted to increase in all countries in the region, with those of the second-tier DACs increasing at a faster rate compared to Korea, Taipei, China and Singapore. China's intra-regional export and import shares are predicted to increase from 51.6% to 59.6% and 41.7% to 81.9% respectively between 2010 and 2030.
Trade openness of the region, measured by the trade to GDP ratio, is projected to According to our projections based on the standard gravity modelling framework, total non-oil exports and imports from DAC countries would increase at an annual rate of 8.5% and 7.8% during 2010-2030, exhibiting a mild slow down in the rate of growth over time. The growth of Intra-regional trade would be about 1.2 percentage points faster, resulting in an increase in intraregional share in total exports and imports of countries in the regional from 53.5% to 9.5%, and 56.9% to 63.3% respectively. These predictions need to be treated with caution as they are based on the assumption that patterns of trade pertaining to our estimation period 1986-2008 will continue unchanged during the ensuing two decades.

Trade Data Compilation
The data used in the analysis of trends and patterns trade flows for all countries other than To analyze the growing importance of regional production networks in determining trade patterns, we rely on detailed (5-digit) data for the period 1992 to 2008.
Although the SITC Rev. 3 was introduced in the mid-1980s, a close examination of country-level data shows that data recording systems in many countries had considerable gaps in the coverage parts and components trade until the early 1990s. Therefore we use 1992 as the starting year of our data disaggregation.
In previous studies of international production fragmentation and trade patterns There is no hard and fast rule applicable to distinguishing between parts and components and assembled products in international trade data. The only practical way of doing this is to focus on the specific industries in which network trade is heavily concentrated. Once these industries are identified assembly trade can be tentatively estimated as the difference between parts and components, directly identified based on our list, and recorded trade in these product categories (Krugman 2008). This is the procedure we follow here. Guided by the available literature on production sharing, we identify seven product categories: office machines and automatic data processing but these are lumped together in the UN data system with 'special transactions' under SITC 9. However, the magnitude of the bias resulting from the failure to cover these items is unlikely to be substantial because network trade in final assembly is heavily concentrated in the product categories covered in our decomposition.