Fragmentation and East Asia's information technology trade

Abstract

This article studies the growth and determinants of information technology (IT) trade in the Asia-Pacific region. We argue that the rise of IT trade must be understood within the context of increasing vertical fragmentation of production processes that has occurred over the past two decades. To evaluate this empirically, we estimate a set of pooled bilateral IT export equations for eight Asian countries, the USA and the EU, where foreign direct investment (FDI) inflows are introduced as a proxy for fragmentation. We apply a panel cointegration approach that allows for heterogeneity in short-run dynamics and in fixed effects. Consistent with production fragmentation, we find that the evolution of IT trade can be explained in part by traditional income and relative price effects but also by FDI inflows.

Acknowledgements

Special thanks to the University of Hawaii New Economy Research Grant program for financial support. Thanks also to seminar and conference participants at the University of Hawaii at Manoa, Singapore Management University, Yokohama National University, UN Project LINK, and the Western Economic Association International.

Notes

¹ For our purposes, IT includes data processing machines (computers, photocopy machines, calculators, etc.), radio and telecommunications equipment and electronic components (semiconductors, electronic tubes and valves, etc.). These correspond to standard international trade classification (Rev. 2) categories 75, 76 and 776. This admittedly rough classification is necessitated by limitations in international trade data.

² The concentration of IT production in East Asia has made many economies in the region uniquely vulnerable to IT industry cycles. Abeysinghe (2000) looks at the case of Singapore, and Daraisami (2004) looks at the case of Malaysia.

³ East Asia's share of exports remains higher than that of the United States even if we exclude intraregional East Asian trade, since nearly two-thirds of East Asia's IT-related exports are interregional (Table 3).

⁴ We identify three of the eight three-digit SITC IT sectors as predominantly ‘Parts and Components’ and the remaining five categories as primarily ‘Final Products’. ‘Parts and Components’ comprise SITC 759 (parts of and accessories suitable for office machines and automatic data processing machines and units), 764 (telecommunication equipment and parts) and 776 (thermionic, cold and photo-cathode valves and tubes). ‘Final Products’ include 751 (office machines), 752 (automatic data processing machines and units), 761 (television receivers), 762 (radio-broadcast receivers) and 763 (gramophones, dictating, sound recorders etc.). This rough measure is the best one can do with three-digit data.

⁵ The export RCA index is calculated as the ratio of two ratios, the ratio of exports for each subsection of IT in an economy to that economy's total IT exports, relative to the ratio of world exports for each corresponding section to world total IT exports. The index reveals the pattern of export specialization for an economy relative to worldwide patterns. The greater a sector's RCA, the more an economy specializes in that sector's exports relative to world specialization patterns. The import RCA index is defined analogously. While such indices by definition measure the degree of specialization of exports in product categories, the term ‘revealed comparative advantage’ is an unfortunate misnomer. Because trade may be highly distorted by trade barriers and implicit or explicit domestic taxes and subsidies, such measures may ‘reveal’ little about the actual comparative advantage of countries.

⁶ FDI is an imperfect proxy for fragmentation, since multinational enterprises can also move production stages across borders by outsourcing production to foreign firms. Nevertheless, considering the limited availability of data, FDI may be the best available proxy.

⁷ Over the years, a number of empirical studies have focused on the question of whether aggregate trade and FDI are complements or substitutes in East Asia. Urata (2001) uses country-level aggregate trade and investment data and finds that inward FDI stock promotes trade. Liu et al . (2002) find that FDI inflows Granger cause Chinese exports. Head and Ries (2001) use firm-level data for Japanese manufacturing firms and find a net complementary effect between trade and FDI, with substitution effects occurring for firms that do not export intermediate inputs. Ramstetter (1999, 2002), finally, uses plant-level data and finds that foreign plants have higher trade propensities than local plants in Indonesian, Thai and Singapore manufacturing.

⁸ This well-known model can be derived from a simple imperfect substitutes trade model, where the quantity demanded is a function of income in the destination market, own product price and prices of competing products. See Armington (1969). The notation here loosely follows Goldstein and Khan (1985). Because a log-linear form imposes constant elasticities, the functional form is not without its critics, for example Marquez (1994). This form is nevertheless dominant in empirical practice.

⁹ Representation in this form requires that demand be homogenous in prices and that foreign consumers choose between country i goods and a composite of all other countries’ goods. Because we are modelling demand for a specific subgroup of commodities, ideally we would like to include prices of all goods entering the foreign consumption basket. As a practical matter, we do not have reliable time series of prices for IT products, and so we restrict our attention to aggregate price deflators, specifically consumer (or overall export) prices.

¹⁰ Branson (1968) as cited in Goldstein and Khan (1985). But note that the interpretation of price elasticity coefficients for exports are different with this specification, since a rise in domestic export prices will raise the value of a given export volume one-for-one. So the value elasticity = volume elasticity + 1.

¹¹ A number of studies have estimated import demand equations that augment the usual relative price and income elasticities with an exporter supply term. Sato (1977) includes manufacturing capacity in the exporting country as a supply term. Helkie and Hooper (1988) include the relative capital stock between the exporting and importing country. Bayoumi (1999) and Gagnon (2003) include export-country real GDP as a proxy for supply. The justification of the inclusion of a supply term in these studies is different from ours. These studies based the inclusion of a supply term in the context of new trade theories, which focus on the implications for trade of increasing returns to scale in production and the desire of consumers for variety in consumption. The studies generally find that the inclusion of the exporter supply term significantly reduces the income elasticity.

¹² We have omitted five bilateral flows because of very short samples: Indonesian exports to Korea, the Philippines and Taiwan; Philippines exports to Indonesia and Singapore exports to Indonesia. We have also interpolated the missing 1989 values for trade between the Philippines and Taiwan.

¹³ The CPI is a potentially poor proxy for relative prices of tradable goods; we will also consider a specification using relative export prices in future research.

¹⁴ Based on Monte Carlo simulations reported in (Pedroni, 2004), we should expect that the group ADF statistic has the best power in samples such as ours with small T and moderate N, followed by the panel ADF. Pedroni also considers rho and variance-ratio tests. He reports Monte Carlo simulations indicating that the panel rho and group rho have power ranging from 0% to 20% for samples with T = 20, N = 20. He finds that the variance-ratio tests consistently produced low-power and large-size distortions.

¹⁵ That is, estimates converge to their true values at rate T as opposed to the conventional (Stock, 1987).

¹⁶ In addition to the standard problem of omitted supply side, there are good theoretical reasons to expect trade levels to influence FDI. Investments may be made in response to anticipated trade opportunities, or trade may initially open a market for a firm, which may later have an incentive to move production into the market once scale economies are realized. In addition, common actors, such as per capita income levels, appear to explain both trade and FDI in cross-sectional analyses (Eaton and Tamura, 1995).

¹⁷ The panel OLS estimator converges to the true value at rate (Pedroni, 1996).

¹⁸ Unlike the Phillips and Hansen (1990) single equation FMOLS, panel FMOLS estimators are normally distributed due to the averaging of the unit root distributions over the cross-section under the maintained assumption of independent idiosyncratic error processes.

¹⁹ The common time dummy is intended to capture the impact of cross-sectional dependencies. If the IT surge of the past decade was caused in part by an exogenous shock to productivity growth, this would lead to increases in the export of IT products by all countries. If this common causes is not modelled, the strong export growth may instead be linked to growing income, which itself may be rising due to the productivity growth.