Can Asia sustain an export-led growth strategy in the aftermath of the global crisis? Exploring a neglected aspect

Journal of Asian Economics 29 (2013) 45–61

Contents lists available at ScienceDirect

Journal of Asian Economics

Can Asia sustain an export-led growth strategy in the aftermath of the global crisis? Exploring a neglected aspect§ Gonzalo Hernandez Jimenez a,b,*, Arslan Razmi a a b

Department of Economics, University of Massachusetts, Amherst, MA 01003, United States Universidad Javeriana, Bogota, Colombia

A R T I C L E I N F O

A B S T R A C T

Article history: Received 1 August 2012 Received in revised form 11 August 2013 Accepted 21 August 2013 Available online 4 September 2013

Employing panel data for Asian countries to distinguish between different kinds of exportand tradable-led growth, we find that the proportion of a country’s manufactured exports that is destined for industrialized countries, a variable largely ignored by existing studies, is robustly associated with growth. This finding has crucial implications given the expected deceleration of industrialized country import growth in the coming years. Most importantly, and contrary to some recent studies, prospects for continued growth, now centered on domestic tradable consumption or on developing countries as markets, may be limited. South–South trade may not be a good substitute for South–North trade. ß 2013 Elsevier Inc. All rights reserved.

JEL classification: F43 O11 O53 Keywords: Export-led growth Tradable-led growth Global imbalances Industrialization Capital accumulation

1. Introduction and background This paper empirically investigates future prospects for the so-called ‘‘Asian growth model’’ by attempting to identify the nature of such growth in the years predating the current global economic crisis. In the process, we identify the proportion of a country’s exports that is destined for industrialized countries as a statistically significant determinant of Asian growth. This finding, which has important implications for future growth, can be broadly motivated at the firm level by recent theoretical and empirical work following Melitz (2003). At the macro/economy-wide level, however, this variable has been ignored as a potential determinant of growth. The term ‘‘export-led growth,’’ whether interpreted to mean growth driven by net exports (i.e., trade surpluses) or simply exports, has been closely associated with East Asian countries in recent decades. Indeed the pursuit of export promotion is what, according to some, has distinguished the East Asian performance from that of other less successful countries,1 making it a desirable template for many developing countries across the globe. In particular, relatively rapid growth along with current account surpluses in Asian developing countries following the Asian financial crisis of 1997–98 and the global

§ We would like to acknowledge very useful comments by the two anonymous referees. Earlier versions of this paper were presented in 2012 at the Eastern Economics Association meetings in Boston and the Analytical Political Economy workshop at the University of Massachusetts. We thank the participants for helpful feedback. * Corresponding author. E-mail addresses: [email protected] (G.H. Jimenezy), [email protected] (A. Razmiz). 1 See, for example, Bhagwati (1990).

1049-0078/$ – see front matter ß 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.asieco.2013.08.003

46

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

Fig. 1. Weighted current account as a proportion of GDP for the 23 industrialized countries in our sample (1997–2011). Sources: Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations.

recession in 2001 have generated renewed interest in export-led growth. Unprecedented growth in China along with its rapid accumulation of foreign exchange reserves have only served to confirm the perceived efficacy of such a growth model. Will the export-led growth strategy retain its efficacy, if, as expected, slower industrialized country growth in post-crisis years reduces global imbalances? A bit of background may help analyze this question. Viewpoints on post-crisis Asian prospects range from skeptical to optimistic, partly reflecting varied interpretations of pre-crisis Asian growth. On the optimistic side, some scholars have recently argued that developing countries have ‘‘decoupled’’ from the US and Europe, and that the slowdown in the developed countries should, therefore, not significantly affect the prospects of the South. Canuto, Haddad, and Hanson (2010), for example, suggest that we may be witnessing the emergence of ‘‘Export-Led Growth V2.0,’’ where South–South exports substitute for South–North exports. This interpretation implies that exports, especially manufactured ones, contribute to growth regardless of their destination. A somewhat optimistic message also emerges from Rodrik (2009), who has argued that, rather than exports or trade surpluses, Asian growth successes were based on broader tradable sector growth. Thus, there is something special about this sector which, in Asian countries, is typically associated with industrial production. We refer to this later as the tradable-led growth strategy. If Rodrik’s argument is empirically valid, then shrinkage of global imbalances should not be an obstacle to post-crisis growth since domestic demand for tradables can substitute for exports. Skeptics, however, have pointed out that the existence of a fallacy of composition or adding-up constraint undermines the sustainability and/or universal applicability of the export-led growth strategy. For one country to export more, at least one other country has to import more. Simultaneous pursuit of export-led growth by all developing countries, especially if concentrated in a similar range of products, could only be successful if demand from developed countries grows at a correspondingly rapid pace, and/or if the terms of trade move against the growing countries, thus increasing competitiveness in an imperfect substitutes framework. 2 Such a strategy, in other words, requires that developed countries run trade deficits, which may beyond some point become unsustainable. Thus, universal pursuit of export-led growth is likely to yield diminishing returns. The recent global financial crisis has served to highlight the adding up constraint. For this constraint becomes even more relevant if, as widely expected, developed countries grow at a slower pace or are less willing to run trade deficits following the crisis. Indeed, current account imbalances have already shrunk. This can be seen from Fig. 1, which illustrates weighted current account deficits as a percentage of GDP for the 23 industrialized countries in our sample (see Section 3 and Table 2 for details of the sample).3 Interestingly enough, after a long period of deficits since the Asian crisis of 1997–98, industrialized countries have, as a group, returned to nearly balanced current accounts in the aftermath of the 2007–08 global downturn. Finally, a consideration that has received far less attention is the possibility that exports may have different effects depending on their destination. As discussed in Section 2, a more recent strand of literature, inspired in large part by Melitz (2003), has emphasized the role of exports as harbingers of productivity growth. Insofar as knowledge spillovers, technology transfer, and adoption of new management techniques are more likely to result from manufactured exports to developed

2 Barring the unlikely case where developing country products are perfect substitutes for developed country products, or where there is complete passthrough of exchange rate changes into developing country export prices when measured in domestic currency terms, a devaluation will translate into a deterioration in the terms of trade. In logical terms, the simultaneous pursuit of export-led growth by a number of small developing countries becomes analogous to the large country case. 3 The current account to GDP (CA-GDP) ratio was obtained for the period 1997–2011 from the United Nations Conference on Trade and Development (UNCTAD) STAT database. The annual weights assigned to each developed country for calculating the CA-GDP ratio were based on the share of total manufactured exports to developed countries from the 43 Asian countries in our sample that went to that particular country that year. In other words, we weighed the industrialized countries according to their importance as an export destination for Asian countries.

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

47

countries, growth may also be a function of the proportion of a country’s manufactured exports destined for these countries. The effects of exports, from this perspective, are not destination neutral. The discussion above suggests that, before we can evaluate prospects for the future, it would be helpful to evaluate the past. In particular, we need to clarify terms such as ‘‘export-led growth’’ and ‘‘tradable-led growth.’’ This paper does this by distinguishing between four different strategies: 1. Tradable sector-led growth (TRADABLE): Rodrik (2008), for example, argues that the tradable sector, which in developing countries is associated mainly with industry, is typically afflicted with market failures and institutional weaknesses to a greater extent, leading these countries to devote a sub-optimal share of their resources to this sector. Second-best policies to subsidize tradable production, therefore, are likely to promote growth. 2. Net Export-Led Growth (NET_EXPORT): The term ‘‘export-led growth’’ has traditionally been understood in a Keynesian framework, whereby positive net exports or trade surpluses generate a source of demand for domestic output, and hence cause output growth. It is in this sense that the idea of an adding-up constraint emerges. A logical corollary is that slower growth of demand and greater reluctance to run trade deficits in developed countries will make it harder to pursue this kind of growth. 3. Manufactured export-led growth (MANUF_EXPORT): The hypothesis is that greater international competition, international knowledge spillovers, economies of scale, and other relevant externalities make exports a vehicle for technological change and, hence, economic growth. What makes exports special in this case is not any external accountrelated consideration but rather the presence of externalities associated with the process of exporting. Thus, the emphasis shifts to supply-side factors. Since the kinds of externalities discussed above are generally associated with manufactured exports, we refer to this hypothesis as the manufactured export-led growth strategy. 4. Industrialized country-centered export-led growth (EXPORT_TO_INDUS). Finally, a consideration that has received far less attention is the possibility that all manufactured exports may not be created equal. Insofar as positive effects on productivity are more likely to result from exports to developed countries, growth may also be a function of the proportion of a country’s manufactured exports destined for these markets. This could be seen as a special case of the manufactured export-led growth hypothesis. As discussed in Section 2, existing literature has explored the first three strategies. Treatment of the fourth, at least at an aggregate level, however, is absent to the best of our knowledge. This is surprising in light of the recent literature on learning by exporting. Furthermore, the four growth strategies listed above have different implications for the post-crisis prospects for developing countries. In particular, the degree to which shrinking global imbalances could hamper a continuation of precrisis growth strategies depends on the nature of these strategies. Table 1 lays out a schematic summary of these implications. The NET_EXPORT strategy will face a tighter adding-up constraint in a post-crisis world if global growth is slower, and if developed countries experience smaller trade deficits. The EXPORT_TO_INDUS, MANUF_EXPORT and TRADABLE strategies, by contrast, will not face that constraint since external imbalances are not a factor. Put differently, developing Asia can continue to pursue EXPORT_TO_INDUS, MANUF_EXPORT, and TRADABLE but not NET_EXPORT in a world without Asian trade surpluses. Moreover, the distinctions between various growth models or strategies have other important policy implications. For example, consider a small open economy where workers save a smaller proportion of their income than owners of capital. If there is something special about exports, then redistributing income towards profits will lower domestic consumption, and help obtain the growth benefits of the MANUF_EXPORT strategy by freeing up domestic tradables for export. If, on the other hand, it is the entire tradable/industrial sector that is special – the Rodrik (2009) argument – then lower wages would simply shift the composition of demand for domestic output from domestic to foreign sources without affecting output growth. The key in this case is boosting tradable production regardless of where it is sold. From our perspective, the fact that it is the industrialized countries that are expected to experience shrinking trade deficits in the near future has an interesting implication for the post-crisis world. Since positive net exports provide a boost to demand regardless of destination, at least some countries could continue pursuing NET_EXPORT by substituting trade surpluses with other developing countries for those with developed countries. However, if the destination matters, then Table 1 Different growth strategies.

Tradable-led growth (TRADABLE) Net export-led growth (NET_EXPORT) Export-led growth

Driven by manufactured exports (MANUF_EXPORT) Driven by exports to industrialized countries (EXPORT_TO_INDUS)

Shrinking trade deficits necessarily bad

Shrinking industrialized country demand necessarily bad

No Yes

No No

No

No

No

Yes

48

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

Table 2 Data and sample definitions. Code

Definition

Source

Coverage

GRGDPCH RGDPCH TRADABLE MANUF _ EXPORT

Geometric growth rate of (chained) real GDP per capita (Chained) real GDP chain per capita Industry value added (% of GDP) Manufactured exports (% of GDP). Calculation based on manufactured exports (% of merchandise exports), merchandise exports (current US$) and GDP (current US$) External balance on goods and services (% of GDP) Manufactured exports (SITC 5–8) to developed countries as a proportion of manufactured exports to World FDI inflows as a share of GDP Government spending as a share of GDP Terms of trade Saving as a proportion of GDP Openness [(exportsþimports)/GDP] Gross Fixed Capital Formation as a proportion of GDP Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Luxembourg, Netherlands, Norway, New Zealand, Portugal, Spain, Sweden, Switzerland, UK, and USA Afghanistan, Azerbaijan, Bahrain, Bangladesh, Bhutan, Brunei Darussalam, Cambodia, China, Hong Kong SAR, Macao SAR, India, Indonesia, Iran, Iraq, Israel, Jordan, Kazakhstan, Kuwait, Kyrgyzstan, Lao, Lebanon, Malaysia, Maldives, Mongolia, Nepal, Oman, Pakistan, Philippines, Qatar, Rep. of Korea, Saudi Arabia, Singapore, Sri Lanka, Syria, Tajikistan, Thailand, Timor-Leste, Turkey, Turkmenistan, United Arab Emirates, Viet Nam, Yemen, Uzbekistan Brunei Darussalam, Cambodia, China, Hong Kong SAR, Indonesia, Rep. of Korea, Laos, Macao SAR, Malaysia, Philippines, Singapore, Thailand,Timor-Leste, Viet Nam

PWT 7.0 PWT 7.0 WDI Authors’ calculations based on WDI

1950–2009 1950–2009 1960–2009 1960–2009

WDI UN COMTRADE

1960–2009 1962–2010

UNCTADstat WDI WDI WDI PWT 7.0 WDI

1962–2010 1960–2006 1980–2006 1960–2006 1960–2009 1960–2006

NET _ EXPORT EXPORT _ TO _ INDUS FDI GG TOT SAV _ GDP OPENC INV _ GDP Developed countries

Asian developing countries/regions

East and Southeast Asian countries/regions

lower export growth to industrialized countries will undermine the EXPORT_TO_INDUS strategy, hampering continuation of pre-crisis growth. Put slightly differently, a region will be affected negatively by slower developed country growth in a post crisis world if that region pursued EXPORT_TO_INDUS, but not necessarily if it pursued MANUF_EXPORT (and could, therefore, replace exports to industrialized countries with those to other developing countries – the Canuto et al.’s (2010) argument). This study seeks to econometrically distinguish between the NET_EXPORT, MANUF_EXPORT, EXPORT_TO_INDUS, and TRADABLE strategies using panel data for pre-crisis years. Although we explore post-crisis growth prospects for Asian countries, our study should not be seen as a comprehensive study of growth determinants. We focus instead on the external channels that are directly relevant to growth in Asian countries, most of which experienced the recent global downturn through their linkages to the US and Europe, rather than through internal banking or financial collapses. Our contribution to the literature on export-led growth and global rebalancing is twofold: First, we distinguish between four growth strategies, identifying historically the most relevant ones for Asia, and drawing conclusions for the future. Second, we find that the proportion of a country’s exports that is destined for industrialized countries (a proxy for EXPORT_TO_INDUS) has a statistically robust positive effect on output growth. This variable, which gains added importance in the present global context, rarely appears in the literature. Section 2 provides an overview of the main issues and related literature. The next two sections develop the empirical strategy and present the econometric estimates. Section 5 concludes. 2. Literature review The recent international financial crisis has served as a big shock to the global trade and financial architecture. As illustrated by Fig. 2, Asian countries in particular had enjoyed rapid growth and trade surpluses in the years leading to the crisis.4 Also interesting is the upward evolution, since the late seventies, of manufactured exports as a proportion of GDP and that of the proportion of manufactured exports destined for industrialized countries. The size of the industrial sector as a proportion of GDP has, on the other hand, stayed more or less the same since the mid-eighties.

4 Due to the unbalanced nature of our panel, we display the means of our variables of interest. More details about the composition of our sample follow in Section 3.

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

49

Fig. 2. Mean of real per capita GDP and other variables for Asian countries (1953–2009). Sources: Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations.

The rapid growth in the years leading up to 2007 was widely perceived as having been based on surging exports. Especially impressive in this regard has been growth in China over the last three decades which has existed alongside huge current account surpluses in recent years. A logical corollary is that, given that developed countries are likely to grow at a slower pace following the crisis, and that deficit countries may increasingly resort to protectionist measures, the model of growth based on exporting manufactures to developed countries may have outlived its utility.5 Discussion of the sustainability of the growth model cannot be separated from that of the nature of the growth model. Traditionally export-led growth has been interpreted to mean trade surplus-led growth. This type of growth is subject to the fallacy of composition critique that becomes particularly relevant in the post-crisis world where a shortage of international demand originating from developed countries is likely. Evidence on the existence of a fallacy of composition has thus far been suggestive although not conclusive. For example, based on panel data estimates for 22 major developing country exporters of manufacturers, Razmi (2007) finds the presence of significant demand-side constraints on export growth. Furthermore, the estimates suggest that rapid Chinese export growth has had a significant impact in this regard. Eichengreen, Rhee, and Tong (2007) confirm the tendency for China’s exports to crowd out those of other Asian countries but find a difference in the impact of China on low income versus middle and high income Asian countries. This is because the effect is felt mainly in markets for consumer goods which are exported by lower income Asian countries. China’s simultaneous tendency to absorb large volumes of capital good imports from its Asian neighbors, on the other hand, has benefited the more advanced Asian economies. A different basis for export-led growth was offered by a strand of literature following Feder (1983). In Feder’s two sector model, the output of the non-export sector depends not only on the factors of production (labor and capital) but also on exports. This captures the externality associated with factors unique to exports such as higher quality labor, internationally competitive management, etc. Moreover, the marginal product of factors in the export sector is greater than that in the nonexport sector. Thus, exports, from this perspective, can potentially influence productivity and growth independently of their impact on the external balance. More recently, several studies following Melitz (2003) have analyzed the relationship between firm heterogeneity, trade, and exports at a more micro level. A relevant empirical finding is that exporting firms tend, on average, to be larger and more productive. This suggests either that more productive firms self-select into export markets (due to extra costs imposed by the process of exporting), and/or that firms that export become more productive. The latter may happen due to several reasons such as economies of scale, dynamic learning, technological spillovers, and competitive pressures. Pack (2001), for example, notes that international competition allowed purchasers abroad to exert heavy pressure on East Asian exporters, producing under contract, to cut costs and increase efficiency. Exporting firms may have easier access to new technologies thanks to their international links. Moreover, exporting firms may receive technical guidance on how to meet higher quality standards from their clients in importing countries. Easier transfer of managerial skills may also be a factor. While empirical evidence for self-selection tends to be quite robust, that for learning-by-exporting appears to be significant only for developing countries. This is not surprising since these countries tend to be farther away from the technological frontier, and hence have greater scope for learning.

5

See, for example, the discussions in UNCTAD (2010) and Adams and Park (2009).

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

50

Other recent studies too have pointed to the potentially special nature of exports. In an econometric study of nine African countries, Van Biesebroeck (2005) finds evidence that manufactured exports facilitate productivity growth. The study shows that scale economies plays an important role in this regard. In a study of British manufacturing firms, Greenaway and Kneller (2007) find that exporting firms experience productivity growth relative to non-exporters. Moreover, the magnitude of divergence across industries appears to be driven by differences in the scope for learning. The export effect is greater if the distance to the technological frontier is large. These issues are likely to be more relevant for developing countries, as are the potential gains from imitation.6 The special nature of the tradable sector, which in developing countries consists mainly of the industrial and agriculture sectors, need not be limited to exports, however. Rodrik (2008) presents an AK-type model of endogenous growth in which the tradable sector is special in the sense that it is characterized to a greater degree by institutional weaknesses and market failures (information and coordination externalities), leading to a bias against this sector in the allocation of resources. Second-best policies to subsidize tradable production, therefore, could promote growth.7 In a recent contribution that perhaps comes closest to the spirit of our paper, Rodrik (2009) tests the tradable-led and the export-led growth hypotheses by running a horse race between: (1) the industrial share of GDP (used as a proxy for the size of the tradable sector) and the exports to GDP ratio on the one hand and, (2) the former and trade surpluses as a proportion of GDP on the other. The panel consists of both developed and developing countries. The paper finds evidence that the industrial share of GDP matters more, especially for developing countries. We end this section with a brief look at an issue that is directly relevant to our empirical findings as they relate to postcrisis prospects for developing countries. As mentioned earlier, some literature has suggested that the emerging economies in Asia and elsewhere have ‘‘decoupled’’ from the developed world, and are, therefore, immune to slower growth in the latter. Noting the growth in South–South trade, Canuto et al. (2010), discuss the possible evolution of a new version of export-led growth, in which South–South trade picks up the slack through middle income countries importing more from low-income ones. The authors term this scenario ‘‘export-led growth v2.0.’’ This, however, raises a new set of questions. Since it is the developed countries that are expected to limit their trade deficits in the post-crisis years, is there anything special about exporting to these countries? In other words, is learning-by-exporting more significant in the case of manufactured exports to developed countries, perhaps due to the presence of more stringent product quality expectations, a greater proportion of sophisticated manufactured products in the export basket, more technical guidance and greater diffusion of best management practices from buyers, greater learning in the presence of competitors who operate close to technological frontier, or other factors? Indeed, existing literature does provide some supportive evidence in this regard. For example, Pack (2001) notes that export-oriented production encouraged East Asian countries to move toward more sophisticated technology to meet the complex contractual requirements from Western industrial countries. Insofar as broadly similar goods are differentiated in quality and consumers in higher income countries have a greater willingness to pay for quality, developing country exporters are likely to export higher quality products to Northern markets. Theoretical and empirical support comes from Verhoogen (2008), who develops a model in which differential quality valuation on the part of consumers leads Southern exporters to produce higher quality goods for export and upgrade their technologies. Based on an econometric exercise, the paper finds support for this prediction in the Mexican case.8 De Loecker (2007) concludes, based on an econometric study, that productivity gains from exporting are greater for firms exporting to high income countries.9 If this is indeed the case, that is, if exporting manufactures to high income industrialized countries has growthinducing effects over and above those originating from exporting manufactures, then a (post-crisis) Export-Led Growth v2.0, which involves other developing countries replacing developed countries as export destinations, may not be a good substitute for (pre-crisis) Export-Led Growth v1.0. We find, as we probe these issues empirically in the next section, that there may indeed be something special about exports to industrialized destinations. 3. Data and econometric strategy Based on the various forms of tradable- and export-led growth strategies identified in Section 1, we begin with a baseline regression of the form: 2 2 2 X X X GRGDPCHjt ¼ a þ b0 lnRGDPCHjt1 þ di TRADABLEjti þ g i MANUF EXPORT jti þ li NET EXPORT jti i¼0

þ

2 X

pi EXPORT TO INDUSjti þ f t þ f j þ ejt

i¼0

i¼0

(1)

i¼0

6 See also, among other recent studies, Hiep and Ohta (2009) for the case of Vietnamese manufacturing firms and Park, Yang, Shi, and Jian (2009) for China. Wagner (2007), Pedro and Yang (2009), and Silva, Africano, and Alfonso (2010) comprehensively review studies of the learning-by-exporting channel. 7 See also Razmi, Rapetti, and Skott (2012) for a model of an economy that features tradable-led growth in an environment of underemployment of labor resources. 8 See also Brambilla, Lederman, and Porto (2010) and Bastos and Silva (2010) for related evidence. 9 See also Pedro and Yang (2009) and Silva, Africano, and Alfonso (2010).

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

51

The dependent variable is the average annual rate of real (chained) GDP per capita growth, RGDPCHit1 (real GDP per capita in the previous period) captures the convergence term, ft time specific effects, fj country specific effects, while eit is the error term. Real GDP growth was obtained from the Penn World Tables version 7.0. The GDP share of industry is denoted by TRADABLE. Following Rodrik (2009), among other studies, we use this as a proxy for the size of the tradable sector. The variable NET _ EXPORT represents the trade balance as a proportion of GDP, and captures the effects of net exports on growth. Manufactured exports, i.e., exports of SITC categories 5, 6, 7, and 8, as a proportion of GDP is represented by the variable MANUF _ EXPORT. Data for these four variables were obtained from the World Bank’s World Development Indicators (WDI) online database. Finally, EXPORT _ TO _ INDUS is the proportion of manufactured exports destined for developed countries. Data for the construction of this variable were obtained from the United Nation’s COMTRADE database. Table 2 provides a data dictionary along with a list of the countries included in the sample. It may be warranted here to re-visit the choice of our main variables of interest, as included in Eq. (1). Our focus is on exploring the nature of Asia’s growth strategy as it relates to external factors. More specifically, whether Asian growth can be identified either as tradable-led or export-led (or both), and, if so, what implications does the past pattern of growth have for a future in which slower developed country growth translates into less global demand? The motivation behind the inclusion of a proxy for the tradable sector is obvious in light of the discussion in Section 1. If, as Rodrik (2009) argues, pre-crisis Asian growth was tradable-led, then exports are not the main driver of growth. Indeed one could subsidize tradable production for domestic consumption to substitute for exports. If, however, pre-crisis growth was export-led, then this may not be true and shrinking global imbalances and/or reduced global demand become more serious concerns. The trade balance as a proportion of GDP captures Keynesian demand-side net export-led growth stimulus. As discussed in Section 1, this is only one channel – and perhaps not the most important one at that – through which exports could facilitate growth, and exports, especially manufactured ones, could be special for other reasons. This provides the grounds for including MANUF _ EXPORT as an explanatory variable. However, as discussed earlier, the size of the manufacturing export sector relative to the economy may be only part of the story. If exports to industrialized countries feature to a greater degree the positive externalities associated with knowledge spillovers, competition, learning-by-exporting, and quality control, then more limited demand from these countries in the post-crisis environment could become a significant constraint on developing country growth. An additional variable, i.e., EXPORT _ TO _ INDUS, is required to accommodate the possibility of manufactured exports to industrialized countries generating growth effects over and above those produced by manufactured exports in general. Our sample consists of a maximum of 43 Asian developing countries and 23 industrialized countries over the time period 1953–2009, although data are available only for shorter intervals for some series. In order to remove short-run cyclical effects, we use data averaged over three year intervals. 10 We pursue a general-to-specific (GTS) estimation strategy, which is particularly useful given our limited sample size. In each case, we first estimate the general form based on Eq. (1). The variables that are not significant at the 10 percent level are then eliminated in a step-wise manner. A more parsimonious specification allows us to increase the available degrees of freedom. Given the asymmetrical availability of data for each key variable, moreover, the specific model relaxes the limits on the maximum number of observations used in a common sample. 11 Some of the variables in our sample could potentially be endogenous in the sense that these are jointly determined with the dependent variable. For example, the share of industry in the economy may not be exogenous to the GDP growth rate. Moreover, some of the variables are likely to exhibit hysteresis or persistence over time. To address the robustness of our baseline OLS estimates to potential endogeneity/simultaneity issues, we therefore, carry out dynamic panel estimations using the Arellano–Bover General Method of Moments (GMM) approach. We specify the second and third lags of the dependent variable as instruments in addition to the third lags of TRADABLE, NET _ EXPORT, MANUF _ EXPORT, and EXPORT _ TO _ INDUS. Consistent with our OLS strategy, we specify time and cross-section effects, and pursue a more parsimonious specification based on eliminating variables that are not statistically significant. The Sargan test of overidentifying restrictions is employed to test the validity of our instruments. We report both the OLS and GMM results for our baseline regressions (Tables 3 and 4) but only the GMM results for the later exploration of robustness and related issues.12 Fig. 3 captures characteristics of the variables of primary interest with the help of histograms. These characteristics later inform some of our robustness tests. Asia had an impressive mean growth rate of 3 percent per year in real GDP per capita. Almost two-thirds of the observations for GRGDPCH lie between 4.5 percent (i.e., one standard deviation). Industry as a proportion of GDP ranges from a minimum of 7.7 percent to a maximum of 90.2 percent, with a mean of 34.7 percent. Most of the observations lie within the 20–45 percent range. The distribution of manufactured exports as a percentage of GDP is much more skewed with most values clustered in the 0–10 percent range and very few beyond 50 percent. Moreover there is a significant difference between the mean (16.6 percent) and the median (6.7 percent), indicating that a relatively small number of countries pulls the average up. Only a few values lie above the 60 percent level. The trade balance as a proportion of GDP is centered around zero percent, as one would expect. The highest number of values lie between negative 5 percent to zero. The mean is a trade deficit

10

The 3 year average for GRGDPCH, a variable in growth rate form, was calculated using the following formula: GRGDPCH = [(RGDPCHt/RGDPCHt1)(1/ ]  1. 11 See Cuthbertson, Hall, Stephen, and Taylor (1992) for a more detailed discussion of the philosophy underlying the GTS strategy. 12 The OLS results are available in a separate ‘‘Available On Request’’ appendix.

3)

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

52 Table 3 Baseline growth regressions, 1953–2009.

Dependent variable: GRGDPCH (Growth rate of real GDP per capita)a

Constant

(1)

(2)

(3)

(4)

(5)

OLS general

OLS specific

OLS specific standardized

GMM general

GMM specific

0.1601 (1.35)

0.2631*** (3.58)

GRGDPCHt1 0.0287*** (3.07) 0.0025*** (2.68) 0.0024** (2.44)

0.9571*** (3.07) 0.8553*** (2.68) 0.8109** (2.44)

0.0006** (2.26) 0.0006** (2.02)

0.3411** (2.26) 0.3363** (2.02)

Time dummies Country dummies

0.0194 (1.38) 0.0031** (2.46) 0.0012 (0.87) 0.0010 (1.11) 0.0007* (1.94) 0.0013*** (2.74) 0.0008** (2.52) 0.0003 (0.54) 0.0012 (1.60) 0.0005 (1.47) 0.0023 (0.11) 0.0022 (0.10) 0.0174 (1.59) Yes Yes

0.0264** (2.07) Yes Yes

0.1323** (2.07) Yes Yes

Long-run coefficients TRADABLE Wald statistic p-value MANUF _ EXPORT Wald statistic p-value NET _ EXPORT Wald statistic p-value EXPORT _ TO _ INDUS Wald statistic p-value

0.0009 1.72 [0.19] 0.0002 1.07 [0.30] 0.0010 1.05 [0.31] 0.0220 0.61 [0.43]

0.0001 0.04 [0.85] 8.35E06 0.00 [0.97]

0.0445 0.04 [0.85] 0.0049 0.00 [0.97]

0.0264 4.30 [0.04]

0.1323 4.30 [0.04]

0.55

0.53

0.53

LnRGDPCHt1 TRADABLEt TRADABLEt1 TRADABLEt2 MANUF _ EXPORTt MANUF _ EXPORTt1 MANUF _ EXPORTt2 NET _ EXPORTt NET _ EXPORTt1 NET _ EXPORTt2 EXPORT _ TO _ INDUSt EXPORT _ TO _ INDUSt1 EXPORT _ TO _ INDUSt2

Adjusted R-squared J-statistic Instrument rank Sargan test (p-value) Cross-sections included Observations

32 214

32 239

32 239

0.1357 (0.80) 0.0057 (0.32) 0.0027 (1.40) 0.0089*** (2.91) 0.0040 (1.51) 0.0012 (0.99) 0.0024 (1.42) 0.0008 (0.71) 0.0013 (1.12) 0.0012 (0.83) 0.0004 (0.65) 0.0425 (0.51) 0.1599 (1.35) 0.1984** (2.57) Yes Yes

0.0040*** (3.60) 0.0055*** (6.14)

0.0018* (1.93) 0.0018** (2.24) 0.0020** (2.28)

0.1192* (1.97) 0.1657*** (3.35) Yes Yes

0.0025 1.07 [0.30] 0.0005 0.50 [0.48] 0.0006 0.13 [0.72] 0.0938 1.75 [0.19]

0.0015 2.28 [0.13] 6.57E05 0.03 [0.86] 0.0020 5.22 [0.02] 0.0465 2.05 [0.15]

13.44 41 0.57 28 149

20.27 41 0.57 29 152

Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations. Long-run GMM estimates correspond to the sum of short-run coefficients divided by one minus the estimate for GRGDPCHt-1. Variables defined in Table 2. a t-statistic. * p < 0.10. ** p < 0.05. *** p < 0.01.

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

53

Table 4 Growth regressions that include an Asian crisis dummy, 1953–2009. Dependent variable: GRGDPCH (Growth rate of real GDP chain per capita)a

Constant

(1)

(2)

(3)

(4)

OLS general

OLS specific

GMM general

GMM specific

0.1322* (1.76)

0.1895*** (3.31)

GRGDPCHt1 LnRGDPCHt1 TRADABLEt TRADABLEt1 TRADABLEt2 MANUF _ EXPORTt MANUF _ EXPORTt1 MANUF _ EXPORTt2 NET _ EXPORTt NET _ EXPORTt1 NET _ EXPORTt2 EXPORT _ TO _ INDUSt EXPORT _ TO _ INDUSt1 EXPORT _ TO _ INDUSt2 ASIAN _ CRISIS (1998–2000 =1) Country dummies

0.0139 (1.50) 0.0035*** (3.22) 0.0014 (0.98) 0.0019** (2.15) 0.0006* (1.68) 0.0010** (2.24) 0.0005* (1.74) 0.0003 (0.80) 0.001 (1.47) 0.0011*** (4.20) 0.0159 (0.85) 0.0085 (0.38) 0.0438*** (3.75) 0.0265*** (5.42) Yes

Long-run coefficients (sum of the individual coefficients) TRADABLE 0.0002 Wald statistic 0.11 p-value [0.74] MANUF _ EXPORT 8.30E05 Wald statistic 0.10 p-value [0.76] NET _ EXPORT 0.0002 Wald statistic 0.06 p-value [0.80] EXPORT _ TO _ INDUS 0.0364 Wald statistic 2.50 p-value [0.11] Adjusted R-squared J-statistic Instrument rank Sargan test (p-value) Cross-sections included Observations

0.53

32 214

0.0184** (2.51) 0.0023*** (2.95)

0.0029*** (4.41)

0.0008*** (4.78)

0.0544*** (4.34) 0.0290*** (10.63) Yes

0.0006 1.78 [0.18]

0.0008 22.80 [0.00] 0.0544 18.82 [0.00]

0.0244 (0.21) 0.0246* (1.73) 0.0017 (0.97) 0.0051** (2.59) 0.0019 (1.02) 0.0011 (1.18) 0.0021* (1.82) 0.0006 (0.70) 0.0006 (0.66) 0.0009 (1.06) 0.0009* (1.68) 0.0611 (1.12) 0.0158 (0.20) 0.0774 (1.56) 0.0205* (1.70) Yes

0.0035*** (6.00)

0.0013** (2.30) 0.0011** (2.09)

0.0722*** (3.61) 0.0353*** (3.97) Yes

0.0016 1.14 [0.29] 0.0004 0.67 [0.41] 0.0007 0.52 [0.47] 0.0329 0.49 [0.49]

0.0035 36.02 [0.00]

14.31 29 0.10 28 149

15.88 28 0.86 30 162

0.0002 0.12 [0.73] 0.0722 13.04 [0.00]

0.52

35 233

Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations. Long-run GMM estimates correspond to the sum of short-run coefficients divided by one minus the estimate for GRGDPCHt-1. Variables defined in Table 2. a t-statistic. * p < 0.10. ** p < 0.05. *** p < 0.01.

54

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

Fig. 3. Histograms of main variables of interest. Sources: Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations.

of 0.23 percent although the median (1.93 percent) suggests that a relatively small number of countries with large surpluses characterizes the series.13 Finally, the proportion of manufactured exports that is destined for developed countries ranges from almost zero to nearly 100 percent. The mean is 34 percent. Since Japan itself is an Asian country, albeit a high income industrialized one, we exclude it from the list of industrialized countries while calculating the series EXPORT _ TO _ INDUS. As a robustness test, we also estimate regressions with Japan included among the industrialized countries, and show that such a change does not qualitatively affect our results. Including Japan, however, raises the mean to almost 40 percent. 4. Estimates 4.1. Baseline regressions Columns (1)–(3) of Table 3 present the results of our baseline OLS regressions, proceeding from the most general form based on Eq. (1) to more specific/parsimonious specifications based on the strategy discussed earlier. The upper half of the table reports the individual coefficient estimates while the lower half details the long-run coefficients along with their statistical significance (where applicable, i.e., only in the cases where one or more of the contemporary and lagged instances

13 There was one value that was so implausibly high that we excluded it from the outset. The trade deficit to GDP ratio for Kazakhstan was reported as 10,133 percent for 1989–91!

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

55

of a variable form part of the reported specification).14 Consistent with standard expectations, the convergence term (LRGDPCHt1) has a negative sign and is generally significant at the 1 percent level.15 The most general form in column (1) has few significant coefficients, the contemporary coefficient of TRADABLE, and the contemporary and lagged coefficients of MANUF _ EXPORT. This is likely due to the number of lags specified for a relatively small panel. Column (2) reports estimates for the more specific form. Only the contemporary and lagged instances of TRADABLE and MANUF _ EXPORT, and the twicelagged EXPORT _ TO _ INDUS survive. Moreover, the Wald test indicates that the summed coefficients of TRADABLE and MANUF _ EXPORT are not significant at the 10 percent level. Thus, EXPORT _ TO _ INDUS emerges as the only significant longrun variable. In order to facilitate comparison, column (3) presents the standardized coefficients based on the specific regression in Column (2). A one standard deviation variation in EXPORT _ TO _ INDUS boosts growth by 0.13 standard deviations. This suggests that a 22 percent rise in the proportion of a country’s exports destined for industrialized countries – which corresponds approximately to one standard deviation change in EXPORT _ TO _ INDUS – is associated with an increase in annual growth of real GDP per capita of 0.6 (= 0.13 standard deviation of GRGDPCH) percentage points. This is a substantial effect, although one has to bear in mind that such a large change in a country’s export composition takes time to occur. As mentioned earlier, endogeneity is a concern with most macroeconomic studies of this kind. In order to further explore causality, we turn next to the Arellano–Bover GMM approach, which also allows us to address persistence effects by including the lagged dependent variable. Columns (4) and (5) present the results of the robustness tests using this approach. Column (4) reports the most general regression, which again yields very few significant variables.16 Moving to the more parsimonious regression reported in column (5), again, the second lag of EXPORT _ TO _ INDUS turns out to be positive and significant, and the effect is larger than in the OLS case. The long-run coefficient, although positive, is significant only at the 15 percent level. The coefficients on the contemporary instance and first lag of TRADABLE are significant as are those on the contemporary instance of NET _ EXPORT and the first and second lags of MANUF _ EXPORT, although again, the summed coefficients are not significant at the 10 percent level with the exception of NET _ EXPORT, which is small and negative. The Sargan tests of overidentifying restrictions, reported for all three regressions, do not raise any concerns at the standard levels of significance. In sum, both the OLS and GMM approaches suggest that, of the variables in our baseline regression, only the second lag of EXPORT _ TO _ INDUS and the contemporary instance of TRADABLE have a consistently positive and significant effect. Moreover, only EXPORT _ TO _ INDUS has a significant and positive long-run effect on GDP growth, although the GMM estimate is significant only at 15 percent. The coefficient on this variable is larger in the GMM regressions. The trade balance or magnitude of manufactured exports as a proportion of GDP do not appear to significantly affect growth in the long-run. Finally, the finding that it is the second lag of EXPORT _ TO _ INDUS that is statistically significant suggests that the effect takes time to develop. Growth of exports to industrialized countries leads to output growth farther in the future. 4.2. Taking the Asian crisis into account As is well known, the Asian crisis of 1997–99, which began with a speculative run on the Thai Baht and quickly spread to other parts of Asia had a drastic negative impact on income and employment. Does this effect show up in our data? To explore this dimension more directly,17 we re-ran the baseline regressions with a dummy for the period 1998–2000. Time fixed effects were now excluded from the model in Eq. (1) for obvious reasons. Again, we estimated using both OLS and GMM techniques. Table 4 summarizes the results. As expected, the Asian crisis had a negative and significant impact on Asian growth regardless of the estimation technique. The coefficient on this dummy variable ranges from 0.02 to 0.04. The proportion of exports destined for industrialized countries continues to be positive and significant as a long-run determinant. In terms of individual coefficients, it is the second lag that is consistently significant. Industry as a proportion of GDP has a negative long-run effect, although it is statistically significant only in the GMM case. In qualitative terms, the only major difference from the baseline regression is that the long-run effect of NET _ EXPORT too now becomes significant in the OLS case, the significant coefficient on the second lag suggesting that trade surpluses may have a positive impact on future growth. 4.3. The influence of Japan and China As mentioned earlier, we excluded Japan from the list of industrialized countries while calculating the variable EXPORT _ TO _ INDUS. Exports to Japan have been a major area of growth for other Asian countries. Are our estimates robust to the inclusion of Japan in the list of developed countries? Table 5 addresses this question. Starting with the estimates derived without controlling for the Asian crisis (columns (1) and (2)), notice first that the second lag of EXPORT _ TO _ INDUS

14 The long-run coefficients are the sum of the short-run ones in the OLS case and the sum divided by one minus the coefficient on the lagged dependent variable in the GMM case. 15 This remains true for most of the regressions reported below, although the magnitude of the estimate varies. 16 Notice that we are down to 149 observations in this case. 17 Notice that the time fixed effects in earlier specifications should capture this Asia-wide shock. In our baseline regression, the time fixed effect is the largest for the period 1998–2000, and is 0.025 in magnitude.

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

56

Table 5 (GMM) Growth regressions run after including Japan as a destination country, 1953–2009. Dependent variable: GRGDPCH (Growth rate of real GDP chain per capita)a (1) General GRGDPCHt1 LnRGDPCHt1 TRADABLEt TRADABLEt1 TRADABLEt2 MANUF _ EXPORTt MANUF _ EXPORTt1 MANUF _ EXPORTt2 NET _ EXPORTt NET _ EXPORTt1 NET _ EXPORTt2 EXPORT _ TO _ INDUSt EXPORT _ TO _ INDUSt1 EXPORT _ TO _ INDUSt2

0.2138 (1.18) 0.0261 (1.09) 0.0011 (0.55) 0.0079*** (2.73) 0.0047* (1.74) 0.0022* (1.76) 0.0029 (1.57) 0.0009 (0.71) 0.0004 (0.29) 0.0011 (0.77) 0.0007 (1.14) 0.0066 (0.08) 0.1669 (1.31) 0.1667*** (2.84)

(2) Specific

0.0042*** (5.98)

0.0779*** (3.37)

ASIAN _ CRISIS (1998–2000 =1) Time dummies Country dummies

Yes Yes

Yes Yes

(3)

(4)

(5)

(6)

Dummy Asian Crisis

Dummy Asian Crisis

Excluding China

Excluding China

General

Specific

General

Specific

0.0062 (0.05) 0.0150 (0.98) 0.0026 (1.54) 0.0064*** (3.68) 0.0019 (1.00) 0.0016* (1.74) 0.0021* (1.78) 0.0004 (0.52) 0.0006 (0.54) 0.0007 (0.81) 0.0009 (1.63) 0.0488 (0.97) 0.0566 (0.57) 0.0436 (0.64) 0.0176 (1.39) No Yes

0.0022** (2.37) 0.0051*** (6.66)

0.0645*** (3.68) 0.0363*** (4.98) No Yes

Long-run coefficients (sum of the individual coefficients) TRADABLE 0.0027 0.0042 Wald statistic 0.93 35.81 p-value [0.33] [0.00] MANUF _ EXPORT 0.0002 Wald statistic 0.12 p-value [0.73] NET _ EXPORT 9.33E06 Wald statistic 0.00 p-value [0.996] EXPORT _ TO _ INDUS 0.0082 0.0779 Wald statistic 0.01 11.35 p-value [0.92] [0.00]

0.0019 2.46 [0.12] 0.0001 0.05 [0.83] 0.0004 0.11 [0.74] 0.0518 1.43 [0.23]

0.0029 22.35 [0.00]

J-statistic Instrument rank Sargan test (p-value) Cross-sections included Observations

15.69 29 0.33 28 150

13.01 41 0.60 28 150

29.41 39 0.25 31 175

0.0027 (0.02) 0.0140 (0.86) 0.0025 (1.42) 0.0065*** (3.87) 0.0020 (1.06) 0.0014 (1.63) 0.0017 (1.60) 0.0001 (0.16) 0.0008 (0.71) 0.0005 (0.59) 0.0008 (1.49) 0.0404 (0.79) 0.0628 (0.60) 0.0368 (0.52) 0.0133 (0.91) No Yes

0.0033*** (5.53)

0.0012* (1.96) 0.0014*** (2.78)

0.0798*** (3.04)

0.0385*** (4.47) No Yes

0.0033 30.62 [0.00]

0.0645 13.52 [0.00]

0.0020 2.62 [0.11] 0.0001 0.08 [0.77] 0.0005 0.21 [0.65] 0.06 1.98 [0.16]

27.28 27 0.24 31 172

14.39 29 0.42 27 145

26.19 26 0.29 27 172

0.0002 0.16 [0.68] 0.0798 9.25 [0.00]

Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations. Long-run GMM estimates correspond to the sum of short-run coefficients divided by one minus the estimate for GRGDPCHt-1. Variables defined in Table 2. a t-statistic. * p < 0.10. ** p < 0.05. *** p < 0.01.

continues to be positively and significantly associated with growth (see column (2)). Second, TRADABLE has a small but negative effect that is statistically significant. The other two variables representing the trade balance and the manufactured exports’ share of GDP are insignificant. The inclusion of a dummy variable for the Asian crisis makes the contemporary coefficient on TRADABLE significant while the other results remain qualitatively the same. The long-run coefficient on TRADABLE is significant but small and negative. Reassuringly, the inclusion of Japan does not significantly affect our results and the long-run coefficient on EXPORT _ TO _ INDUS continues to be positive and significant.

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

57

From the other side of the income divide, one may justifiably suspect that, as a rapidly growing large exporter, China may be driving our results. Columns (5) and (6), derived after excluding China from the sample of developing countries, confirm that this is not the case. The results are qualitatively the same with minor differences in the coefficient values. 4.4. Regional and temporal asymmetries Much of the debate surrounding global imbalances and export-led growth has involved the East Asian tigers and the South East Asian export dynamos that followed their lead in what is sometimes called a ‘‘flying geese’’ formation. Do these Table 6 (GMM) Growth regressions for cross-sectional and temporal sub-samples. Dependent variable: GRGDPCH (Growth rate of real GDP per capita)a (1)

(2)

(3)

(4)

Rest of Asia

Excluding Middle East

1953–1997

Specific

Specific

General

General

0.0329*** (5.25)

0.1499*** (4.17) 0.0921*** (7.78) 0.0015*** (3.98)

0.5155** (2.16) 0.0225 (1.54) 0.0016 (0.80) 0.0058** (2.22) 0.0030 (1.34) 0.0002 (0.20) 0.0011 (0.54) 0.0006 (0.44) 0.0035* (1.78) 0.0038 (1.39) 0.0002 (0.15) 0.0189 (0.31) 0.0801 (1.10) 0.1141* (1.85) Yes Yes

East and South East Asia General

Time dummies Country dummies

0.5281** (2.21) 0.0194 (1.20) 0.0027 (1.42) 0.0062** (2.09) 0.0027 (0.88) 0.0004 (0.44) 0.0020 (1.35) 0.0009 (0.85) 0.0035* (1.93) 0.0034 (1.56) 5.31E05 (0.04) 0.0426 (0.38) 0.0349 (0.37) 0.0900 (1.22) Yes Yes

Long-run coefficients TRADABLE Wald statistic p-value MANUF _ EXPORT Wald statistic p-value NET _ EXPORT Wald statistic p-value EXPORT _ TO _ INDUS Wald statistic p-value

0.0018 0.37 [0.56] 0.0016 1.40 [0.24] 6.54E05 0.00 [0.98] 0.0266 0.02 [0.88]

J-statistic Instrument rank Sargan test (p-value) Cross-sections included Observations

16.76 40 0.27 10 71

GRGDPCHt1 LnRGDPCHt1 TRADABLEt TRADABLEt1 TRADABLEt2 MANUF _ EXPORTt MANUF _ EXPORTt1 **

MANUF _ EXPORTt2

NET _ EXPORTt NET _ EXPORTt1 NET _ EXPORTt2 EXPORT _ TO _ INDUSt EXPORT _ TO _ INDUSt1 EXPORT _ TO _ INDUSt2

0.0004*** (3.64)

0.0003 (0.99)

0.0826*** (3.25) Yes Yes

Yes Yes

0.0017 16.02 [0.00] 0.0004 13.28 [0.00] 0.0003 1.00 [0.32] 0.0826 10.58 [0.00] 21.64 39 0.60 11 99

118.73 123 0.19 25 149

(5) Specific

(6)

(7)

(8)

(9)

1989–2009 Specific

General

Specific

0.0989*** (4.32) Yes Yes

0.0352 (0.15) 0.0080 (0.18) 0.0046 (1.50) 0.0103** (2.35) 0.0057** (1.86) 0.0021 (1.40) 0.0024 (0.98) 0.0011 (0.56) 0.0027 (1.35) 0.0018 (1.44) 0.0001 (0.12) 0.0640 (0.48) 0.0360 (0.26) 0.0981 (1.07) Yes Yes

0.0026 0.53 [0.47] 0.0015 1.79 [0.18] 0.0009 0.06 [0.80] 0.1092 1.31 [0.25]

0.0009 1.05 [0.31] 0.0007 5.23 [0.02] 0.0030 21.47 [0.00] 0.0989 18.67 [0.00]

3.17E05 0.00 [0.995] 0.0008 0.17 [0.68] 0.0011 0.32 [0.57] 0.0019 0.00 [0.98]

0.0030 5.44 [0.02] 0.0008 3.49 [0.06] 0.0021 3.19 [0.07] 0.0065 0.02 [0.88]

0.0043 1.30 [0.26] 0.0005 0.11 [0.74] 0.0004 0.03 [0.87] 0.1932 1.33 [0.25]

0.0746 4.35 [0.04]

12.98 41 0.60 20 120

22.33 41 0.56 21 126

8.49 29 0.29 19 76

16.77 29 0.21 19 76

7.65 26 0.18 27 115

18.82 23 0.17 33 161

0.0067*** (4.45) 0.0058*** (4.49)

0.0007** (2.29)

0.0030*** (4.63)

0.1508** (2.41) 0.1573*** (3.16) Yes Yes

0.0172 (0.08) 0.0269 (0.59) 0.0008 (0.20) 0.0059 (1.02) 0.0007 (0.17) 0.0016 (0.71) 0.0023 (1.30) 0.0001 (0.11) 1.26E05 (0.01) 9.29E05 (0.04) 0.0005 (0.48) 0.0676 (0.34) 0.0692 (0.21) 0.1981 (0.78) Yes Yes

0.0063*** (5.59) 0.0073*** (4.93) 0.0040*** (3.05) 0.0008* (1.87)

0.0038*** (2.94) 0.0018* (1.83)

0.0845*** (4.93)

0.0746** (2.08) Yes Yes

Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations. Long-run GMM estimates correspond to the sum of short-run coefficients divided by one minus the estimate for GRGDPCHt-1. Variables defined in Table 2. a t-statistic. * p < 0.10. ** p < 0.05. *** p < 0.01.

58

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

countries behave differently than the rest of Asia in terms of our main variables of interest? In order to explore this possibility, we divided the sample into East and South East (ESE) countries on the one hand and the rest of Asia (ROA) on the other. Columns (1)–(3) of Table 6 summarize the estimates derived for these groups. Focusing again on the parsimonious form estimates (columns (2) and (3)), there is some evidence of differing behavior. While the proportion of exports destined for industrialized countries plays a positive and statistically significant role in boosting real per capita GDP growth in the ESE countries, that appears not to be the case for the ROA countries, where only the long-run coefficient on the industry share of GDP is positive and significant. As is generally the case with our previous regressions, the trade balance and share of manufactured exports are either insignificant and/or have a negative impact on output growth. Thus, the main finding reported by Rodrik (2009), that is, the existence of a positive association between the share of the industrial/tradable sector, holds for the rest of Asia but not for East and South East Asian countries. These results provide suggestive evidence for the widely perceived export-led basis of East Asian growth. To explore the issue further, columns (4) and (5) present the regressions after excluding Middle Eastern countries. These countries are, on average, mainly exporters of primary commodities and may have systematically different structural characteristics from many of the other countries in the sample.18 The results are similar to those obtained for the ESE countries, suggesting that the rest of Asia behaves like this group of countries once the Middle East is excluded. Columns (6)–(9) present results for regressions run with the sample period split into two overlapping periods, 1953– 1997 and 1989–2009. The periods were allowed to overlap in order to have evenly split and reasonably large sub-samples.19 Our general finding that the proportion of exports sold in industrialized country markets is robustly and positively associated with long-run output growth holds only for the second sub-period. For the earlier period, the individual coefficient on the second-lagged instance is positive and significant but is offset by the negative coefficient on the first lagged instance so that the long-run effect is not significant. For the first sub-period, however, the long-run coefficient on TRADABLE is significantly and positively associated with growth, a result that is in line with the findings of Rodrik (2009). Thus, the proportion of exports destined for industrialized countries appears to have mattered only in recent decades. Given the small sizes of the sub-samples, however, this evidence should only be seen as suggestive. 4.5. Excluding outliers Table 7 addresses potential concerns raised by the presence of outliers. One such concern is that our results could be driven by a handful of high income oil exporting countries. Suppose, for example, that commodity exporters have, on average, a lower proportion of exports destined for developed countries. Since some of these exporters are relatively high income, and since high income countries may, on average, grow slower, this introduces a bias in favor of finding a positive impact of EXPORT _ TO _ INDUS. A scatter plot of EXPORT _ TO _ INDUS versus RGDPCH (not shown but available on request) conveys the concern. The (negative) slope of the fitted line is driven by Qatar and Brunei Darussalam, two small oil and gas exporting countries that have a relatively low proportion of exports destined for developed countries. Notice first that this concern should be addressed in principle by our inclusion of a convergence term. Second, as mentioned earlier, our results are robust to the exclusion of Middle Eastern countries. Finally, the negative correlation between RGDPCH and EXPORT _ TO _ INDUS weakens noticeably once we restrict the sample to countries below the 20,000 threshold of real per capita GDP. Re-running our GMM regression with this more limited sample delivers results similar to our baseline regression that includes all data points (compare column (5) of Table 3 and column (2) of Table 7), although the coefficient on EXPORT _ TO _ INDUS is larger.20 As discussed earlier, and as highlighted by Fig. 3, a few small open economies in our sample (mainly Singapore and Hong Kong but also Macao and Malaysia) have exceptionally high proportions of manufactured exports as a share of GDP. Could these historically fast growing economies be driving our results? Columns (3) and (4) of Table 7 present the estimates derived once we limit the sample to values of MANUF _ EXPORT less than or equal to 60 percent. Again the results are similar to our baseline regressions (compare Table 3 and column (4) of Table 7). The long-run coefficient on TRADABLE is significant and negative. The long-run coefficient on EXPORT _ TO _ INDUS is, if anything, larger in magnitude than the baseline GMM case. Finally, we noticed while discussing Fig. 3 that a few countries export almost entirely to developed countries. Could these countries be driving our results? To investigate this aspect, we re-estimate our baseline growth equation after excluding data points with EXPORT _ TO _ INDUS greater than 60 percent. Columns (5) and (6) of Table 7 present the results. Once again the estimates are qualitatively very similar to those derived for the full sample (see column (5) of Table 3). As previously, EXPORT _ TO _ INDUS emerges as the only significant and positive long-run influence on real per capita GDP growth.

18 We thank one of the referees for this suggestion. We identified the following countries for exclusion based on the World Bank’s World Development Indicators: Bahrain, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Oman, Qatar, Saudi Arabia, Syria, United Arab Emirates, and Yemen. 19 Much less data are available for the earlier period so that even though it spans more years, the number of observations is less than that for the second sub-period. 20 We also ran regressions with interaction terms to explore whether the impact of EXPORT _ TO _ INDUS varies with real per capita GDP. The interaction terms were found to be insignificant.

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

59

Table 7 (GMM) Growth regressions excluding outliers (1953–2009). Dependent variable: GRGDPCH (Growth rate of real GDP per capita)a (1)

(2)

RGDPCH  20,000 General GRGDPCHt1 LnRGDPCHt1 TRADABLEt TRADABLEt1 TRADABLEt2 MANUF _ EXPORTt MANUF _ EXPORTt1 MANUF _ EXPORTt2 NET _ EXPORTt NET _ EXPORTt1 NET _ EXPORTt2 EXPORT _ TO _ INDUSt EXPORT _ TO _ INDUSt1 EXPORT _ TO _ INDUSt2 Time dummies Country dummies

0.1316 (0.75) 0.0202 (0.77) 0.0024 (1.50) 0.0045 (1.64) 0.0017 (0.92) 0.0024** (2.11) 0.0044** (2.26) 0.0024* (1.82) 0.0004 (0.38) 7.19E05 (0.05) 0.0004 (0.44) 0.0034 (0.04) 0.1084 (1.03) 0.1304* (1.79) Yes Yes

Specific

0.0017* (1.97) 0.0043*** (5.16)

0.0013* (1.78) 0.0023** (2.59) 0.0010* (1.67)

0.0491* (1.94) Yes Yes

Long-run coefficients (sum of the individual coefficients) TRADABLE 0.0004 0.0026 Wald statistic 0.02 9.98 p-value [0.89] [0.001] MANUF _ EXPORT 0.0005 1.74E05 Wald statistic 1.10 0.00 p-value [0.30] [0.96] NET _ EXPORT 0.0011 Wald statistic 0.60 p-value [0.44] EXPORT _ TO _ INDUS 0.0292 0.0491 Wald statistic 0.15 3.77 p-value [0.70] [0.05] J-statistic Instrument rank Sargan test (p-value) Cross-sections included Observations

21.59 41 0.12 24 132

27.35 40 0.16 23 139

(3)

(4)

(5)

(6)

MANUF_EXPORT 60%

EXPORT_TO_INDUS60%

General

General

0.1454 (0.82) 0.0080 (0.36) 0.0017 (0.84) 0.0076** (2.23) 0.0036 (1.43) 0.0006 (0.43) 0.0029 (1.62) 0.0013 (0.95) 0.0005 (0.28) 0.0010 (0.56) 0.0006 (0.81) 0.0699 (0.91) 0.1823 (1.45) 0.2014** (2.58) Yes Yes

Specific

0.0037*** (4.08)

0.0009** (2.26)

0.0856*** (3.02) Yes Yes

0.0028 0.80 [0.37] 0.0012 2.67 [0.10] 1.65E05 0.00 [0.99] 0.1041 2.24 [0.13]

0.0037 16.65 [0.00] 0.0009 5.12 [0.02]

12.35 41 0.65 27 139

0.0659 (0.37) 0.0067 (0.37) 0.0029 (1.52) 0.0087*** (2.73) 0.0036 (1.25) 0.0015 (1.35) 0.0024 (1.42) 0.0006 (0.53) 0.0015 (1.28) 0.0014 (0.99) 0.0008 (1.13) 0.0571 (0.82) 0.1123 (0.96) 0.1712** (2.24) Yes Yes

Specific

0.0030** (2.56) 0.0091*** (5.14) 0.0034** (2.29) 0.0017* (1.85) 0.0018* (1.93)

0.0020* (1.96) 0.0020** (2.52) 0.0010* (1.69)

0.1212*** (2.64) Yes Yes

0.0856 9.10 [0.00]

0.0023 0.98 [0.32] 0.0003 0.24 [0.62] 0.0009 0.44 [0.51] 0.1242 2.77 [0.10]

0.0026 7.12 [0.01] 5.71E05 0.02 [0.88] 0.0010 0.79 [0.37] 0.1212 6.95 [0.01]

20.67 40 0.71 28 147

13.44 41 0.57 26 128

18.02 41 0.59 26 128

Penn World Tables 7.0, World Development Indicators, UN COMTRADE, and authors’ calculations. Long-run GMM estimates correspond to the sum of short-run coefficients divided by one minus the estimate for GRGDPCHt-1. Variables defined in Table 2. a t-statistic. * p < 0.10. ** p < 0.05. *** p < 0.01.

4.6. Controlling for possible omitted influences So far we have investigated factors that could help us gauge prospects for the future of tradable- and export-led growth in Asia. We found EXPORT _ TO _ INDUS to be the most robust correlate of growth. Now we briefly explore the effects of other control variables that are typically included in growth regressions as determinants of growth. Given the focus of this paper, we continue to focus mainly on variables that work through interactions with the external accounts. We add twice lagged values of six variables to a baseline regression: (1) openness (OPENC), as measured by the ratio of the sum of exports and

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

60

Table 8 Robustness to additional variable(s) ‘‘Z’’ (GMM panel regressions). Dependent variable: GRGDPCH (Growth rate of real GDP per capita)a (1)

(2)

(3)

(4)

(5)

(6)

(7)

OPENCt2

SAV _ GDPt2

INV _ GDPt2

GGt2

TOTt2

FDIt2

0.0830 (1.47) 0.0575*** (8.01) 0.0352* (1.89) 0.0006 (1.56) Yes Yes

0.4146 (0.82) 0.0521*** (6.40) 0.0336 (1.54) 7.51E05 (0.15) Yes Yes

0.0836 (1.61) 0.0453*** (7.54) 0.0346** (2.27) 0.0022*** (3.81) Yes Yes

0.0883 (1.14) 0.0722*** (3.66) 0.0869** (2.15) 0.0006** (2.43) Yes Yes

0.0881 (1.25) 0.0676*** (6.48) 0.0687** (2.18) 0.0021 (1.26) Yes Yes

90.91 86 0.05 34 208

32.41 86 0.03 33 200

165.62 151 0.04 33 205

16.98 21 0.11 15 87

32.41 33 0.03 32 187

Baseline

Time dummies Country dummies

Yes Yes

0.1454** (2.53) 0.0451*** (5.59) 0.0452* (1.97) 0.0002* (1.79) Yes Yes

J-statistic Instrument rank Sargan test (p-value) Cross-sections included Observations

27.26 38 0.25 34 225

28.36 39 0.20 34 222

GRGDPCHt1 LnRGDPCHt1 EXPORT _ TO _ INDUSt2

0.0927 (1.54) 0.0547*** (7.19) 0.0686*** (2.76)

Z

Penn World Tables 7.0, World Development Indicators, UNCTAD STAT, UN COMTRADE, and authors’ calculations. OPENC, TOT, SAV_GDP, INV_GDP, GG, FDI denote openness, terms of trade, saving, investment, government spending, and foreign direct investment, respectively. The last four variables are expressed as a share of GDP. Variables defined in Table 2. a t-statistic. * p < 0.10. ** p < 0.05. *** p < 0.01.

imports to GDP, (2) saving as a proportion of GDP (SAV _ GDP), (3) gross fixed capital formation as a proportion of GDP (INV _ GDP), (4) government expenditure as a proportion of GDP (GG), (5) the terms of trade (TOT), and (6) FDI inflows as a proportion of GDP (FDI). National saving and investment correlate with the trade balance as a matter of identity. The inclusion of government spending can be justified both on the basis that most growth studies do so, but also on the grounds that it is the component of aggregate spending that is often the most biased toward non-tradables. The use of twice lagged instances is designed to avoid endogeneity problems.21 Data for TOT, GG, and SAV _ GDP came from the WDI while that for OPENC was obtained from the Penn World Tables (version 7). FDI data came from the UNCTAD STAT database. We only introduce one control variable at a time in order to conserve degrees of freedom. Table 8 summarizes the results. Positive terms of trade shocks and an increase in the saving to GDP ratio have a small positive effect on output growth, although the latter is statistically indistinguishable from zero. Greater openness has a small negative effect, although again it is statistically insignificant, as is FDI. The coefficient on government spending is small, negative, and significant. The coefficient on our main variable of interest, i.e., EXPORT _ TO _ INDUS, is consistently positive and statistically significant. The only exception is the regression that includes INV _ GDP, where it is significant only at the 12 percent level (INV _ GDP itself has a negative and insignificant coefficient). 5. Conclusions and implications Our effort involves a rather ambitious question: is it likely that Asian countries will be able to pursue the pre-crisis patterns of rapid growth? Unlike previous studies, we tackle this question by attempting to establish the characteristics of pre-crisis growth, focusing on the trade- and export-related characteristics. More specifically, to what extent was Asian growth tradable sector-led, net export-led, or export-led in some other sense? As we have stressed, the answers have implications for a future in which industrialized countries are likely to grow at a slower pace and global external account imbalances are likely to shrink. We ran a series of growth regressions to derive OLS and GMM estimates, to take into account the Asian crisis of 1997–99, to test robustness for sub-samples, to the exclusion of outliers, and to control for possible influences missing from our benchmark regressions. Our main finding is that, among our variables of interest, the proportion of total Asian manufactured exports that is destined for industrialized countries is the most robust determinant of real per capita GDP growth. Moreover, the influence is not contemporaneous but rather takes time to develop, increasing our confidence in the direction of

21 For example, EXPORT _ TO _ INDUSt2 could affect the degree of openness in the next period, which could then have a positive impact on growth. The variable underlying growth would still be the proportion of exports destined for industrialized countries but some of this effect will now indirectly show up as a positive coefficient on the measure of openness.

G.H. Jimenezy, A. Razmiz / Journal of Asian Economics 29 (2013) 45–61

61

causality. The industrial share of GDP appears occasionally as a positive influence too, but the overall effect is statistically insignificant in most cases. The other two variables of interest, that is, the share of manufactured exports in GDP and the trade balance as a proportion of GDP generally appear to play no significant role in influencing output growth in Asian countries. We find some suggestive evidence that the proportion of exports destined for industrialized countries may have mattered more for the East and South East Asian countries, a group that is distinguished by its high growth rate over past decades. It is perhaps not surprising that, for developing countries that are well inside the technological frontier, manufactured exports to industrialized countries can facilitate growth through knowledge and technology spillovers and the effects of international competition. Moreover, the lagged nature of the mechanism is not surprising given that one would expect these effects to take hold gradually. Indeed our main finding is consistent with the body of recent literature that has found some evidence for exports leading to productivity growth. Most of this literature, however, is based on firm-level data. We, on the other hand, find evidence at the macroeconomic/national level and explore the effect on aggregate growth. These findings have important implications for the post-crisis global economic architecture. To the extent that our findings suggest some role for the size of the tradable sector in promoting growth, the post-crisis world could still witness rapid growth in Asian countries, albeit one based on tradable production for domestic consumption rather than exports. Industrial policy such as subsidies for tradable production may then have to substitute for export subsidies. Furthermore, policies that penalize domestic consumption in order to generate exports will have to be reversed in the face of shrinking external demand. In line with the argument made by Rodrik (2009), shrinking global imbalances need then not be a pressing concern. Our central finding, i.e., that the proportion of exports sold to industrialized countries is, among our variables of interest, the most robustly (and positively) associated with growth, however, has less sanguine implications. Since pre-crisis global imbalances largely involved industrial country trade deficits, contraction of such imbalances will almost certainly require a decline in these deficits. In principle such a decline could occur through greater industrialized country export growth without an accompanying fall in import growth. However, add to this the near certainty that slow industrialized country income growth will cause demand from these countries to grow at a slower clip, and we get the important implication that Asian exports to industrialized countries are likely to decelerate, which in light of our main finding, is a cause for concern. Put differently, the fact that tradable production for domestic consumption or for export to other developing countries may not be good substitutes for exports to industrialized countries magnifies the challenges facing sustained Asian growth in the coming years. Export-led growth v2.0 in this sense may not be a good substitute for export-led growth v1.0. Our study has focused on the growth determinants that relate to tradable and exportable sector issues. A more exhaustive analysis, beyond the scope of our study, will incorporate other variables that are typically seen as influencing growth. A future effort may also want to investigate the specific channels through which exports to industrialized countries impact growth. It might be interesting too to extend the analysis to investigate whether other developing countries behave differently than our Asian sample. We hope to pursue these questions in future work. References Adams, C., & Park, D. (2009]). Causes and consequences of global imbalances: Perspectives from developing Asia. Asian Development Review, 26(1), 19–47. Bastos, P., & Silva, J. (2010, November). The quality of a firm’s exports: Where you export to matters. Journal of International Economics, 82(2), 99–111. Bhagwati, J. N. (1990]). Export promotion trade strategy: Issues and evidence. In C. R. Milner (Ed.), Export promotion strategies: Theory and evidence from developing countries. New York: New York University Press. Brambilla, I., Lederman, D., & Porto, G. (2010]). Exports, export destinations, and skills. Cambridge: National Bureau of Economic Research (Working Paper 15995). Canuto, O., Haddad, M., & Hanson, G. (2010]). Export-led growth v2. 0. Economic Premise, 3: 1–5. Cuthbertson, K., Hall, G., Stephen, P., & Taylor, M. (1992]). Applied econometric techniques. Ann Arbor: The University of Michigan Press. De Loecker, J. (2007, September). Do exports generate higher productivity? Evidence from Slovenia. Journal of International Economics, 73(1), 69–98. Eichengreen, B., Rhee, Y., & Tong, H. (2007]). China and the exports of other Asian countries. Review of World Economics, 143(2), 201–226. Feder, G. (1983]). On exports and economic growth. Journal of Development Economics, 12(1–2), 59–73. Greenaway, D., & Kneller, R. (2007, October). Industry differences in the effect of export market entry: Learning by exporting? Review of World Economics, 143(3), 416–432. Hiep, N., & Ohta, H. (2009, May). Superiority of exporters and the causality between exporting and firm characteristics in Vietnam. Kobe: Research Institute for Economics and Business Administration, Kobe University (Discussion Paper Series 239). Melitz, M. (2003, November). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6), 1695–1725. Pack, H. (2001]). Technological change and growth in East Asia: Macro versus micro perspectives. In J. Stiglitz & S. Yusuf (Eds.), Rethinking the East Asian miracle. Oxford: Oxford University Press. Park, A., Yang, D., Shi, X., & Jiang, Y. (2009]). Exporting and firm performance: Chinese exporters and the Asian financial crisis. Cambridge: National Bureau of Economic Research (Working Paper 14632). Pedro, M., & Yang, Y. (2009]). The impact of exporting on firm productivity: A meta-analysis of the learning-by-exporting hypothesis. Review of World Economics, 145, 431–445. Razmi, A. (2007, December). Pursuing manufacturing-based export-led growth: Are developing countries increasingly crowding each other out? Structural Change and Economic Dynamics, 18(4), 460–482. Razmi, A., Rapetti, M., & Skott, P. (2012]). The real exchange rate and economic development. Structural Change and Economic Dynamics, 23(1), 151–169. Rodrik, D. (2008]). The real exchange rate and economic growth. Brookings Papers on Economic Activity, 2, 365–412. Rodrik, D. (2009]). Growth after the crisis. Washington, DC: Commission on Growth and Development (Working Paper 65). Silva, A., Africano, A. P., & Alfonso, O. (2010]). Learning-by-exporting: What we know and what we would like to know. Porto: Faculdade de Economia da Universidade do Porto (Working Paper 364). UNCTAD. (2010]). Trade and Development Report. Geneva: United Nations Commission on Trade and Development (Technical report). Van Biesebroeck, J. (2005, December). Exporting raises productivity in Sub-Saharan African manufacturing firms. Journal of International Economics, 67(2), 373–391. Verhoogen, E. (2008, May). Trade, quality upgrading, and wage inequality in the Mexican manufacturing sector. Quarterly Journal of Economics, 123(2), 489–530. Wagner, J. (2007]). Exports and productivity: A survey of the evidence from firm level data. World Economy, 30, 60–82.