Domestic Value Added, Exports and Employment: An Input–Output Analysis of Indonesian Manufacturing

Abstract

The paper is motivated by the current emphasis on the share of domestic value added in exports (SVEX) as a policy criterion for export development strategy in developing countries. We hypothesise that the policy emphasis on SVEX, which harks back to the import-substitution era, is inconsistent with the objectives of achieving economic growth with employment generation in this era of economic globalisation. We test this hypothesis by examining the relationship of SVEX with both export-induced employment and the total domestic value added (the contribution of exports to gross domestic product) by applying the standard input–output methodology to data from Indonesian manufacturing. Our findings do not support the view, widely held in policy circles, that manufacturing industries characterised by a higher SVEX have the potential to make a greater contribution to employment generation and total domestic value added. The policy inference is that policymakers should focus on the export potential of industries when designing export development policy, rather than on SVEX.

INTRODUCTION

Policymakers in Indonesia emphasise the share of domestic value added in exports (SVEX) when determining sectoral priorities in export development. For instance, ‘increasing the value-added of the manufacturing sector’ was a key objective of the National Long-Term Development Plan, 2005–25, of the Yudhoyono administration (Kuncoro 2018, 407). This policy emphasis resulted in protectionist policies in the form of an array of restrictions on intermediate goods imports. For example, in 2012 the Government of Indonesia imposed tariffs on imports of machinery and materials used in the assembly of automobiles. The regulation stipulates that at least 30% of the total value of machines used must be locally produced.¹ Similar policies are found relating to other industries, such as the garment, footwear, and food and beverage industries (Patunru and Rahardja 2015). In 2017, the Ministry of Industry introduced a regulation requiring a minimum of 30% local content in the manufacturing of fourth-generation mobile phones.² President Joko Widodo recently instructed the Indonesian Ministry of Industry to give priority to development programs with the potential to increase the domestic value added for the benefit of Indonesians (Ministry of Industry 2020). Presidential Regulation 18/2020 states that Indonesia aims to increase the local content in its export products from an average of 43.3% in 2019 to 50% in 2024. To pursue this policy stance, the Ministry of Industry has issued regulations relating to the local content of domestic production of a number of products, including pharmaceuticals, cellular phones, handheld computers and tablets, and some other electronic or telematics devices.

The policy emphasis on SVEX is not unique to Indonesia. In India, a key focus of the Modi government’s grand vision of ‘Make in India’ is to incubate domestic industry rather than expose it to the undue pressure of competition, with a view to broadening and deepening the domestic procurement base of export-oriented industries (Sharma 2015). The past three years have seen the introduction of selective tariff increases and financial incentives to promote domestic intermediate goods production to encourage export producers to turn to domestically produced inputs (Athukorala 2019; Sharma 2015). In Sri Lanka, the new industrialisation strategy developed by the Presidential Task Force on National Revival and Poverty Eradication has emphasised ‘exploring new export opportunities’ to upgrade value added (CBSL 2021, 20). In South Africa, the National Industrial Policy Framework emphasises the need to promote production at the intermediate stages of the production process as a logical progression to strengthen manufacturing value chains in the South African economy (Hausmann, Klinger and Lawrence 2008).

The usual justification given by the proponents of these policies is that increasing domestic input usage as a percentage of gross output of exports (per unit value added)³ will create more domestic employment while boosting the overall growth of the economy (increase in GDP) in terms of total net export earnings (total value added of exports). What this reasoning overlooks is that under export-oriented industrialisation (as opposed to conventional import-substitution industrialisation), direct policy intervention to per unit domestic value added could, in fact, hinder the growth of and employment generated by domestic manufacturing, for three reasons. First, production for the export market requires the use of high-quality inputs procured at world market prices to maintain competitiveness. Second, under the ongoing process of global production sharing (production fragmentation),⁴ per unit value added in exports within global production networks naturally tends to decline everywhere simply because each country specialises in a given slice/ task of the production process of a given product. Therefore, an increase in total net export earnings (that is, the total domestic value added of exports) from participation in global production sharing depends increasingly on the expansion of export volume rather than per unit value added.⁵ Third, the production of intermediate goods is typically more capital intensive than the assembly of final goods, which is more labour intensive. This means that shifting the domestic production structure towards the production of final goods and away from intermediate production would enhance the potential for employment generation through domestic manufacturing in a labour-abundant country (that is, generate ‘pro-poor’ growth) (Little 1999).

The emphasis on domestic value added as a policy criterion has received added impetus from a new wave of literature dealing with the measurement and patterns of manufacturing exports after converting gross data from customs records into ‘value-added’ terms using an input–output methodology.⁶ This literature was originally motivated by a valid concern that gross trade data tend to exaggerate the magnitudes of bilateral trade imbalances under the ongoing process of global production sharing. This concern arose mainly because of the widening trade deficit of the United States with China, underpinned by China’s rise as an ‘assembly centre’ within global production networks (GPNs) (Athukorala and Yamashita 2009; Bergsten et al. 2006; Dedrick, Kraemer and Linden 2010). It was for this reason that the former World Trade Organization (WTO) director general Pascal Lamy initiated a project for the WTO and Organisation for Economic Co-operation and Development on value-added trade (Lamy 2011). However, many researchers and policy advisers have since begun to use the data generated by this project (and that of other research projects that have emerged to generate value-added trade data, such as the University of Groningen’s World Input–Output Database) to make inferences relating to the developmental implications of export-oriented industrialisation and various other facets of global economic integration.⁷

The purpose of this paper is to assess the validity of using SVEX as a performance criterion in designing policies for export-oriented growth. We hypothesise that, in the context of the ongoing process of internationalising production, industries characterised by high-import intensity (that is, low per unit domestic value added) have the potential to make a greater contribution to employment generation and growth of national income than industries that are deeply rooted in the domestic economy. The import intensities of most of the dynamic product areas are largely determined by factors beyond the control of the individual exporting nations. Therefore, the use of SVEX as a policy guide can be both ineffective and counterproductive. We provide evidence in support of this hypothesis by global production networks (GPNs) has contributed to the decoupling of the ongoing process examining the relationship of value added with both the employment intensity of Indonesian exports and the contribution of exports to the country’s national income by applying the standard input–output methodology to the data for the years 1995, 2000, 2005, 2010 and 2016. We believe that patterns depicted by these quinquennial input–output tables⁸ over the period 1995–2016 provide a suitable empirical base for our analysis because global production sharing is not simply a passing phenomenon but an enduring structural change in global production.⁹

What is the rationale behind the specific focus in this paper on manufacturing; in particular, the contribution of that sector to employment and export expansion? Wouldn’t the specific focus of government policy on manufacturing pull resources out of other sectors, lowering their contribution to national income? Our emphasis on manufacturing is guided by the widely acknowledged view in mainstream development economics of the central role of manufacturing in the early stages of growth and structural transformation in labour-abundant economies (Kuznets 1973; Chenery 1986; Newman et al. 2016). Historically, manufacturing is the sector into which resources have first moved in the course of structural transformation. While it is feasible to move significant numbers of workers from agriculture to wellpaid jobs in manufacturing, it is not possible to move them into formal services sectors: employment in manufacturing requires only on-the-job training whereas employment in the formal services sectors requires at least college-level education. About half of the catch-up by developing countries to high-income levels during the post-war era is explained by shifting labour from relatively low-productive agriculture to more productive manufacturing (Duarte and Restuccia 2010). There is evidence that the manufacturing sector, unlike agriculture and services, converges to global best practice productivity levels unconditionally, regardless of poor institutional quality or bad policies (Rodrik 2013). Export-oriented manufacturing, in particular, is a more labour-demanding strategy of development that could deliver faster and more equitable growth. Developing countries with the most rapid and sustained growth during the post-war era have typically experienced a reallocation of labour from agriculture and informal services into export-oriented manufacturing (Panagariya 2019).¹⁰

The rest of the paper is structured in four sections. The next section provides the context for the ensuing analysis by reviewing the emphasis on domestic value added as a policy criterion in industrial development. The methodology of calculating the domestic value added, employment intensity of exports (export-related employment), and net export earnings using the standard input–output framework is then described. In the third section, these indicators derived from the Indonesian input–output tables for 1995, 2000, 2005, 2010 and 2016 are used to assess the validity of the uses of the value-added ratio as a performance criterion in designing policies for industrial development. The key findings are summarised in the final section.

THE ISSUE

The emphasis on domestic value added in manufacturing (alternatively known as ‘domestic content’ and ‘domestic retained value’) was central to the policy debate on industrialisation in developing countries in the first three decades after the Second World War when import-substitution industrialisation (ISI) held sway as the basic tenet of development strategy. This provided the justification for imposing local content requirements on foreign-invested firms in domestic manufacturing, selectivity in tax concessions, and other incentives for firms to use domestic inputs in the production process. Estimating and analysing the determinants of domestic value added or import intensity of exports and identifying industries with a strong domestic supply base in terms of forward and backward input linkages (commonly dubbed ‘the key sectors’) was a main focus of empirical development economics during this period.¹¹

The basic policy thrust of the ISI strategy was to turn inward and seek the key to industrial development in greater interaction among domestic industries while failing to consider ‘factor proportions’ with regard to resource allocations, as advocated by mainstream economists (Hirschman 1958).¹² Therefore, the empirical development literature at the time mainly aimed to help policymakers find ‘an alternative strategy to linking the economy to the rest of the world on the basis of comparative advantage’ (Findlay 1984, 23).

The emphasis on domestic value added as a policy criterion gradually dissipated from the development literature in the late 1970s as a result of an important paradigm shift in development thinking away from import substitution and towards export-oriented industrialisation. This is because, in a labour-abundant economy, attempts to ‘create’ domestic value added through direct policy intervention can stifle the evolution of the export structure of a given country in line with its comparative advantage in the internationalisation of production. This in turn will hinder the achievement of employment and income growth objectives.

There are two key considerations here. First, in an open economy, the factor intensity of production depends not only upon the technology in the final and intermediate stages of domestic production but also upon the structure of foreign trade. This is because participation in international trade provides the economy with the opportunity to specialise in products in which it has a comparative advantage (that is, labour-intensive products in the case of a surplus-labour economy) while relying on world trade for the procurement of intermediate inputs. The production of intermediate goods is typically more capital intensive than the final assembly of products (Riedel 1975, 1976). Therefore, importing intermediate inputs for export production involves an implicit substitution of labour for relatively capitalintensive intermediate products in the production process. For instance, when an economy imports capital-intensive inputs such as machinery, synthetic fibre and industrial chemicals, with foreign exchange earned by exporting labour-intensive products such as garments, footwear and toys, it is implicitly substituting labourintensive goods for capital-intensive goods in the production structure. This would enhance the labour intensity of the overall production process. Thus, resource allocation considerations make a strong case for the development of footloose (that is, loosely linked) export industries in a labour-abundant economy.

Second, the emphasis on achieving greater domestic content in exports can run counter to the objective of increasing income levels through export-oriented industrialisation. In contrast to the closed-economy approach of ISI, the key to success in penetrating world markets for manufactured goods lies in a country’s ability to produce what international buyers demand. For a surplus-labour country, light consumer goods (for example, clothing, footwear and sporting goods) and component production and assembly in vertically integrated global industries are the most promising areas in the early stage of export-led industrialisation. In the production of these light consumer goods, the use of imported inputs is essential to maintain high quality standards. In component production and final assembly within vertically integrated global industries, import content is naturally high and, in many cases, there is virtually no possibility of local substitution of intermediate inputs. Thus, per unit value added of final assembly is generally lower than in import-substitution production and even in traditional export-oriented manufacturing production. Nevertheless, given the vast market potential for the assembled products, total value added, and hence the contribution to GDP could be much higher. The potential for employment generation is also disproportionately high compared to the low per unit value added. This is because assembly activities within GPNs in a given labour-abundant country are generally more labour intensive than in the conventional factory production of a good in that country ‘from the beginning to the end’ (Jones 2000).

There is extensive case study-based literature covering the industrialisation experiences of both the newly industrialised countries in East Asia and the secondtier newly industrialised countries that casts doubt on the use of value added as a policy criterion in the context of export-led industrialisation (Chow and Papanek 1981; Little 1999; Ranis 1973, 1995). One of the strongest inferences worth quoting here is that by Little (1999, 234):

must be declared the enemy of employment and equity. All labour-intensive sectors have their K/L [capital–labour] ratios raised by backward linkages [that is, an increase in domestic content], because all the intermediaries—petrochemical, artificial fibre, steel, non-ferrous metals, etc.—are highly capital intensive. These intermediaries are the curse of developing countries.

The above arguments by no means imply that a labour-surplus country must remain locked into ‘footloose’ manufacturing activities forever. On the contrary, the important message is that attempts to ‘create value added’ through direct intervention could run counter to the objectives of growth and employment generation under an export-oriented development strategy. With the gradual depletion of excess supplies of labour and adjustment in response to competition emanating from greater international specialisation, the industrial structure will gradually shift to more capital- and skill-intensive industries (provided, of course, that the required preconditions, including human capital development, are met). With the further global integration of the manufacturing sector, the quality of intermediate goods produced in the country would also improve through increased international exposure, although global production sharing naturally sets a limit on the substitution of locally produced parts and components compared to those exchanged within cross-border production networks.

METHODOLOGY

Our methodology draws on the standard input–output framework developed by Leontief (1936).¹³ We calculate domestic value added, employment intensity of exports (export-related employment), and net export earnings (that is, the contribution of exports to domestic value added and GDP) based on the Leontief inverse matrix. Export-related employment captures both direct employment in export production and employment generated indirectly by export production through backward linkages with other industries. Likewise, net export earnings (that is, the total domestic value added of exports) are defined as gross exports minus direct and indirect imported inputs embodied in exports.

Let X be an n × 1 vector of gross output and M be an n × 1 vector of imports. Furthermore, Y^D and E are n × 1 vectors of domestic demand (including usage in consumption and investment) and export demand for domestically produced outputs, respectively, and Y^M is an n × 1 vector of the final demand for imported products (for both consumption and investment). We then have the following:

where is an n × n matrix of direct input coefficients of domestic products and is an n × n matrix of direct imported input coefficients. That is,

where and are elements of n × n matrices Z^D and Z^M—the domestic transaction table and the imported intermediate inputs transaction table, respectively—the summation of which is the n × n total transaction matrix Z.

Solving equation 1 for X gives

where the first term on the right-hand side is the Leontief domestic inverse matrix, with I being the identity matrix. The element of this matrix, , is the output required of the ith sector to sustain one unit of the output of the jth sector.

To measure net export earnings, the import intensity of domestic production must first be subtracted from gross exports. Import intensity is calculated as

where M is the import inverse matrix, and R is the diagonal matrix of imported input coefficients (that is, the share of imported inputs in the total output of the given sectors). An element of M, m_ij, represents both direct and indirect imports required to produce one unit of product j domestically. Thus, the increase of the imported inputs in sector j when the final demand of sector j increases by one unit is given by

The total imports embodied in sector j’s total exports (denoted by e_j) (‘foreign content of exports’ [National Research Council 2006]) is

Accordingly, we can derive the ‘domestic content of export’ or ‘net export earnings’ of sector j as

Finally, SVEX (or, more precisely, per unit domestic content of exports) is given by the ratio of net exports and gross exports, as follows:

Export-related employment is measured by a similar approach. That is,

where L is the employment inverse matrix and G is the diagonal matrix of labour input coefficients. An element of L, l_ij, represents both the direct and indirect employment required to produce one unit of product j domestically. Thus, the increase of employment in sector j when the final demand of that sector increases by one unit is given by

Finally, is export-related employment (MPEX) in sector j as follows:

The dataset for the empirical analysis is constructed by bringing together the input–output tables of Indonesia for 1995, 2000, 2005, 2010 and 2016, and employment data from the National Labour Force Survey (Sakernas) for the same years from Statistics Indonesia (BPS). Indonesia is one of the few developing countries that has produced ‘complementary import type’ input–output tables¹⁴ every five years for more than three decades. The number of sectors in each table varies from 172 to 185. We standardised the tables to 163 sectors (including 82 manufacturing sectors)¹⁵ to allow for sector-by-sector intertemporal comparison. Employment data from the labour force survey are classified by the same sectors in order to calculate export-related employment.

It is worth noting that input–output tables for most countries (including the United States, China, India and Vietnam) are of the competitive import type. For these countries, the calculation of import intensity and net export earning requires separating the intra-industry metric into domestic and imported input matrices by employing the stringent ‘import similarity assumption’—within the product categories of the input–output table, the mixes of imports and domestically made goods are the same.¹⁶ This assumption can lead to significant biases in the estimated domestic content of exports if the exports are concentrated in some manufacturing sectors that depend heavily on imported inputs (such as electronics, electrical goods and cars) (Patunru and Athukorala 2021). The presence of duty drawback schemes and other government initiatives that facilitate duty-free access for the intermediate inputs used in export production could compound such biases. Fortunately, our analysis does not suffer from this limitation, because the Indonesian input–output tables, as noted, are of the complementary import type, with separate domestic and imported input matrices. Both tables are constructed using input-structure data collected from the annual industry survey.

It is pertinent to mention that our estimation procedure may lead to an underestimation of the import intensity of exports, for two reasons. First, the import content of exports produced in each industry is identical to the average import intensity of the industry’s total production (the assumption on which equation 3 is based). This assumption is not entirely consistent with reality. The usual pattern is that, even when industries are finely classified, the import content in an industry’s production for export is higher than in its production for the home market. Second, since the estimates are based on the inter-industry transaction table, they incorporate only the direct import requirements of export production. These estimates do not capture the import intensity of domestic investment (that is, capital formation) in export-producing industries.

RESULTS

We computed the domestic content of exports (net export earnings) and exportrelated employment using the Indonesian input–output tables for 1995, 2000, 2005, 2010 and 2016. The estimates for the 82 manufacturing industries and the supporting statistical tables are given in appendix 1. Table 1 provides the summary indicators derived from these tables for total manufacturing and manufacturing sectors that are closely associated with global production networks (which we call ‘GPN industries’ for brevity). In the data tabulation, we distinguish between industries operating within buyer-driven and producer-driven GPNs. ¹⁷ As discussed below, there are notable differences between these two forms of GPNs in terms ofthe way domestic manufacturing is linked to global manufacturing and the diffusion of technology and marketing and managerial practices. Understanding these differences is important for formulating policies for enhancing a country’s gains from export-oriented industrialisation.

Table 1. Indonesian Manufacturing: Summary Data on Domestic Value Added, Net Exports and Export-Related Employment

Source: Based on the tables in appendix A.

Notes: ^a Based on the Indonesian Standard Industrial Classification (KBLI), which is broadly comparable to the International Standard Industrial Classification (ISIC). Under this classification, products belonging to divisions 15–37 in the ISIC system are treated as manufactured (industrial) products. ^b Export-weighted average.

Buyer-driven networks are common in traditional labour-intensive manufacturing industries such as clothing, footwear, travel goods, toys and sporting goods where production technology is not a significant barrier to entry. In these production networks, the ‘lead firms’ are international buyers (large retailers such as Walmart, Marks & Spencer and H&M) or brand manufacturers (such as Victoria’s Secret, Gap, Zara and Nike). Global production sharing in these networks takes place predominantly through arm’s length relationships, with global sourcing companies (value chain intermediaries, such as Hong Kong-based Li & Fung and Mast Industries) playing a key role in linking producers and lead firms. Thus, there is room for local firms to engage directly in exporting through links established with foreign buyers and to substitute local inputs for imported inputs, depending of course on the ability of local suppliers to meet the required quality standards.

Producer-centred production networks are common in vertically integrated global industries such as those for electronics, electrical goods, automobiles, and scientific and medical devices. In these industries, production technology is normally specific to the lead firm and is closely protected to prevent imitations. Moreover, the production of final goods in these industries requires highly customised and specialised parts and components whose quality cannot be assured by a third party. Given these specific features of the production process, global production sharing takes place predominantly through the lead firms’ global branch networks of subsidiary companies set up in individual countries based on direct foreign investment decisions. Thus, within producer-centred networks, opportunities for increasing the domestic value-added ratio are limited compared to the opportunities offered through specialisation within buyer-driven production networks. However, trade within producer-driven GPNs accounts for the lion’s share of total world GPN trade,¹⁸ and hence specialisation within these production networks opens up much greater opportunity for increased total export-related value added (addition to GDP) and total manufacturing employment than specialisation within buyer-driven GPNs (Athukorala 2019). Moreover, there is evidence that exports from these firms, compared to those operating within buyer-driven GPNs, are less susceptible to the vicissitudes of the domestic investment climate and supply-side interruptions as these GPNs operate under long-term supply arrangements governed by the lead firm within the global value chain (Arndt and Huemer 2007; Athukorala and Khan 2016).

The export-weighted average of the value-added share of manufacturing has remained within the narrow margin of 0.77%–0.80% without showing any clear trend (table 1, panel A). As we hypothesised, both the total net export earnings (the net addition to GDP) and export-related employment exhibit quite distinct patterns. Net export earnings in 2016 stood at Rp 561 trillion, compared to Rp 472 trillion in 1995. Total export-related employment increased from almost 4.9 million jobs to more than 6.2 million over the same period. However, export-related employment was slightly lower in 2005, 2010 and 2016 than in 2000. The disaggregated data reported in appendix 1 suggest that the garments (IO code 76) and plywood (IO code 82) predominantly explained the slowing of employment. Employment in the garment industry contracted by almost 30% between 2000 and 2005 (from 1.03 million jobs to 742, 000) and remained around the 2000 level in 2010 and 2016. The decline in employment in the plywood industry was even sharper: the average level of employment in 2005, 2010 and 2016 in that industry was only a third of the number in 2000 (almost 1.2 million).

The decline in employment in the garment industry is generally ascribed to competition from China following the abolition of the Multi-Fiber Arrangement (MFA) as of January 2005 (Patunru and Rahardja 2015). However, it is important to note that some other countries in the region, such as Bangladesh, India, Cambodia and Sri Lanka, recorded notable increases in garment export after MFA quota restrictions were lifted. A comparative analysis of the Indonesian experience suggests that Indonesia’s failure to reap gains from the removal of quota restrictions was mainly due to supply constraints that hindered the diversification of the product mix and direction of exports in line with changing global demand patterns in the post-MFA era (Pane 2019, chap. 3) In the plywood industry, the initial growth spurt in the 1990s was propelled by the ban imposed by the Indonesian government on log exports (export substitution of plywood for logs). This export boom subsequently dissipated because of the shortage of domestic log supply caused by rapid deforestation and the smuggling of logs to circumvent the ban, coupled with a demand constraint due to ‘tropical timber certification’ (eco-labelling) introduced by the major importing countries (Thee 2009, 145).

The share of export earnings from GPN products in total net export earnings increased from 13.2% in 1995 to nearly 34.5% in 2016 (table 1, panel B).¹⁹ The share of export-related employment generated from these products in total exportrelated employment increased at an even faster rate between those two years, from 21.2% to 37.1%. This supports our assumption of greater employment intensity in vertical specialisation within GPNs compared to the conventional horizontal export orientation. As expected, the average export-weighted value-added ratio of GPN industries (about 70%) (panel B) is smaller than the overall industry average of about 80% (panel A).

The estimates of domestic value added, net exports and export-related employment for the two-digit GPN industries are summarised in table 2. As expected, the average SVEX of GPN products from buyer-driven networks (76%–80%) is larger than that of products from producer-driven networks (67%–74%). Notwithstanding the lower domestic value-added share of producer-driven GPN products, both the net export earnings and export-related employment from these products have grown faster than those from buyer-driven GPN products. Between 1995 and 2016, the net exports of GPN products from producer-driven networks increased at a compound annual rate of 8.9% compared to 5.7% for buyer-driven GPN products. The growth rates of export-related employment due to GPN products from producer-driven and buyer-driven networks were 4.5% and 3.5%, respectively.

Table 2. Indonesian Manufacturing: Domestic Value Added, Net Exports and Export-Related Employment from GPN Industries

Source: See the appendix tables.

Notes: IO = input–output, GPN = global production network.

To supplement this broad-brush discussion of the association between the domestic value-added ratio, and export-related employment and the total domestic value added of exports, we estimated the following regression using a panel data set constructed by putting together the data for five years of Indonesia’s input– output tables (1995, 2000, 2005, 2010 and 2016).

where TVEX is the total domestic value added of exports, MPEX is the exportrelated employment, SVEX is the share of value added from exports, DGPN is a dummy variable that takes the value of 1 for GPN products and 0 otherwise, PROD is productivity, δ_i is the unobservable fixed characteristics of industries’ productspecific effects, γ_t is the unobservable time-specific effect, ϵ_it is the disturbance term, i =1,2, … ,N is the product category and t = 1,2, … t is the is time unit in years.

The main variable of interest is SVEX, which, according to research advocating the use of value-added share as a policy criterion, is postulated to have a positive effect on both TVEX and MPEX. PROD is included as a control variable to capture the efficiency of production. The ideal measure of efficiency of production is total factor productivity—change in production over and above all factors of production used in the production process. Unfortunately, the lack of data at this level of industry disaggregation prevents us from using this measure. We instead use labour productivity, which is measured as real output (value added) per worker, as our measure of PROD. Labour productivity by construct captures the technical efficiency of machinery, the other capital equipment available for the worker to use and the worker’s efficiency in using this equipment in production.²⁰ The intercept and slope dummies for GPN products are included to test whether the hypothesised relationships vary between these products and total manufacturing.²¹ The expected sign of the coefficient of PROD is positive in the TVEX equation and negative in the MPEX. The four variables—MPEX, TVEX, SVEX and PROD—are measured at constant 2010 prices; MPEX, TVEX and PROD are used in natural logarithms and SVEX is a ratio measured in decimal form.

We estimated equation 14 using estimators for fixed effects and random effects, and compared the results using the Durbin-Wu-Hausman test. This test decisively rejected the null hypothesis that unobserved explanatory variables (the unobserved effects) are not distributed independently of the explanatory variables, favouring the use of the fixed-effect estimator. The results are reported in table 3, and summary statistics are given in table 4 to facilitate the interpretation of the results.

Table 3. Value-Added Share of Exports, Total Value Added and Export-Related Employment in Indonesian Manufacturing

Source: Authors’ calculation.

Notes: Heteroscedasticity-corrected (robust) standard errors are in parentheses. a Time (year ) dummy with the year 1995 as the base dummy.

*** p < 0.01, ** p < 0.05, and * p < 0.1.

Table 4. Summary Statistics

Source: Authors’ calculations.

Notes: TVEX = total value added in exports. MPEX = export-related employment. SVEX = share of domestic value added in exports. PROD = productivity. SD = standard deviation.

In table 3, model 1 is the base model (without the GPN dummies). In model 2, the GPN dummy interaction variables cover all 17 GPN products identified in table 2. The fixed-effect estimator automatically dropped the GPN intercept dummy because of its perfect correlation with the industry fixed effects.²² We prefer the fixed-effect estimator because it captures not only the GPN dummy but also other industry-specific unobservable characteristics.

In both sets of equations, the coefficients of SVEX are not significant (table 3). Thus, the results clearly do not reject the null hypothesis that SVEX has no statistically significant association with the total contribution of exports to GDP (total value added) and employment generation. The results are remarkably insensitive to the inclusion of the dummy interaction variables for GPN products.

The signs of the coefficients of the interaction term of SVEX with DGPN are positive in both the TVEX and MPEX equations. This is consistent with what we observed in the simple comparison between total manufacturing and GPN products in table 2. However, the coefficients have failed to achieve statistical significance, presumably because of the limitations involved in the identification of GPN products at this level of commodity disaggregation.

The coefficients of PROD in both models of the TVEX equation indicate a statistically significant positive association between productivity and total value added in exports (contribution of exports to GDP), as expected. However, there is no statistically significant difference between total manufacturing and GPN-related industries as regards this relationship. The coefficients of PROD in the two models of the MPEX equation are negative and statistically significant, suggesting a plausible trade-off between improvement in labour productivity and total employment. Interestingly, the magnitude of this trade-off seems greater for GPN products: the negative coefficient of PROD for GPN products is larger in magnitude, but this result must be taken with caution because, as noted, PROD is also a widely used proxy for capital intensity. From that point of view, the results permit the alternative interpretation that specialisation in global production sharing has greater employment potential than engagement in traditional horizontal specialisation. Unfortunately, it is not possible to distinguish between these two interpretations owing to a paucity of data.

There is obviously an issue of possible reverse causality (endogeneity) relating to PROD (and arguably to a lesser extent, SVEX) in both equations. This is particularly the case across the time dimension, but it possibly poses an issue across sectors as well, to the extent that some workers (in this case, relatively more skilled workers) are ‘fixed’ to given industries. Unfortunately, external instrumental variables for these two variables are hard to come by. We therefore use as a robustness check the generalised method of moments (GMM) approach based on internal instruments, which is intended particularly for studies with small time (T) and large section (N) dimensions (6 and 404 in our case) (Arellano and Bond 1991). We implemented this methodology using the simplified procedure developed by Roodman (2009).

The GMM estimates are reported in table 5. A comparison of these results with those reported in table 4 suggests the use of the standard fixed-effect estimator results in a mild underestimation of endogeneity bias on the coefficient of SVEX in both equations. However, the coefficients of this variable in both equations are not statistically significant even at the 10% level. Thus the results confirm our previous inference that SVEX is not a significant determinant of total value added (addition to GDP) and the new employment effects of export-oriented manufacturing. Interestingly, in the equation for TVEX, productivity is significant only for sectors that are associated with the global production network. In the MPEX equation, the coefficient of the dependent variable is statistically significant with the positive sign indicating a strong lingering effect of labour absorption in Indonesian manufacturing.

Table 5. GMM Estimates for Value-Added Share of Exports, Total Value Added and Export-Related Employment in Indonesian Manufacturing

Source: Authors’ calculations

Notes: ^a Time (year ) dummy with the years 1995 and 2000 absorbed in differencing process. t-statistics are in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

CONCLUDING REMARKS

We have examined the implications of using the share of domestic value added (per unit value added) as a criterion in designing national policy for export-oriented industrialisation in this era of economic globalisation. The key hypothesis is that, given the increased cross-border spread of production processes within vertically integrated industries, policy emphasis on increasing the domestic value-added ratio in exports, which harks back to the era of import-substitution industrialisation, runs counter to the national objectives of achieving economic growth and generating employment under economic globalisation. Production for competitive export markets requires the use of high-quality inputs procured at world market prices. Moreover, given the growing importance of global production sharing as the prime mover of manufacturing export expansion over the past few decades, per unit value added in exports naturally tends to decline everywhere, and national gains from export expansion are fundamentally dependent on volume expansion, not on the increase in domestic content in a given country. Finally, since intermediate production of goods is typically more capital intensive than the assembly of final goods, the domestic value-added ratio is likely to correlate positively with the employment creation (and hence poverty reduction) potential of export-oriented industrialisation at the early stage of industrialisation in developing countries.

We have provided evidence in support of this hypothesis by applying the standard input–output methodology to data for the Indonesian economy. The findings clearly show that export expansion and the growth of export-related employment in the Indonesian economy during 1995–2016 occurred in a context where domestic value added, as usually measured by the domestic content of exports as a percentage of gross exports, remained virtually unchanged. The econometric analysis failed to detect a statistically significant association between the standard valueadded ratio of manufacturing production and net export earnings (contribution to GDP) and export-related employment. The findings become even more striking when we recall that they are based on an estimation procedure that could perhaps have led to an underestimation of the import intensity of export production.

The policy inference of our findings is that, in this era of economic globalisation, policymakers should focus on the export potential of industries rather than on the share of domestic value added of exports in designing export development policy. Using the value-added ratio as a criterion in industrial approval and attempting to engineer value added through other direct policy interventions could run counter to the objectives of growth and employment generation under an export-oriented development strategy. The gradual depletion of the domestic production base through global integration would improve the quality of intermediate goods produced in the country, resulting in an increase in domestic value added in exports. However, the rapid expansion of global production sharing naturally sets a limit on the substitution of locally produced parts and components for those exchanged within cross-border production networks. In this context, an increase in the domestic value added of exports (net export earnings) and employment expansion depends crucially on export volume expansion and the ability of manufacturing firms to move towards high-value tasks and segments in the global manufacturing value chain.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the excellent comments from the two anonymous referees. We also thank Sulistiyo Ardiyono for his excellent research assistance.

Notes

1 Minister of Finance Regulation 76/2012.

2 Minister of Industry Regulation 29/2017.

3 Henceforth, we use ‘value added’ to imply ‘per unit value added’ for brevity, except when explicit distinction is needed.

4 In recent years, in particular following the onset of the Covid-19 pandemic that temporarily disrupted global supply chains, there has been concern about a possible reversal of the internationalisation of industrial production. However, there is convincing evidence that global production sharing is an enduring structural phenomenon, notwithstanding some isolated cases of ‘reshoring’ that figure prominently in the media (Ando, Kimura and Yamanouchi 2022; Antràs 2020; World Bank 2020). Contrary to the popular perception, there is no evidence to suggest that China’s emergence as the premier assembly centre within global production networks (GPNs) has contributed to the decoupling of the ongoing process of global production sharing between China and the United Sates (Athukorala 2017).

5 Therefore, it is important to distinguish between total value added (addition to GDP) and per unit value added (value added ratio) when analysing gains from exports: the latter may decline while the former goes up.

6 See the literature surveys of Johnson (2014) and Timmer et al. (2014).

7 For a critique of this approach, see the work of Patunru and Athukorala (2021).

8 The latest input–output table (2016) is six years apart from the preceding table (2010). However, according to Statistics Indonesia (BPS), the input–output stature of Indonesian manufacturing remained virtually unchanged between 2015 and 2016.

9 See note 6.

10 Based on the official employment data, one could argue that Indonesia does not have a serious ‘employment problem’: the unemployment rate hovered around 4% in the pre-Covid-19 decade. However, in a developing country such as Indonesia, unemployment figures are a deceptive indicator of the extent to which the labour force is productively deployed: poverty and the lack of a social security system dictate that most people have to scratch a living somehow in order to survive. A significant portion of the labour force is crammed into the unorganised sector doing low-productive work (Pratomo and Manning 2022; Rothenberg et al. 2016), and ‘less than half of those in employment collect a recognizable wage or salary [and] the rest mostly work for themselves or their families’ (Economist 2018, 56).

11 See the work of Hazari (1970), Acharya and Hazari (1971), Bulmer-Thomas (1978) and the literature cited therein. Surprisingly, these papers are missing from the reference lists of recent works on value-added trade, even though there is no real novelty in the methodology used compared to this early literature.

12 The mainstream policy advocacy, which is based on the theory of comparative advantage, is that the industrialisation strategy for improving economic welfare for a country is to produce goods (and services) whose production intensively use the factors of productions in which the country is abundantly endowed compared to its trading-partner countries.

13 For an excellent treatment of the input–output analysis with the latest developments in the subject area, see the textbook by Miller and Blair (2009).

14 Input–output tables take two forms: the ‘complementary import’ type and the ‘competitive import’ type. The former comprises two intra-industry matrices, one for domestic inputs and another for imported inputs. That is, the import content of each inter-industry transaction is identified separately and allocated to a separate import matrix. In the latter, imported inputs and domestically procured inputs are lumped together in a single intraindustry transaction table.

15 According to the Indonesian Standard Industrial Classification used by BPS, animal and vegetable oil, which is dominated by palm oil (input–output sector 55), petroleum processing (99), and smoked and crumb rubber (100) are treated as ‘manufacturing’. We excluded these three sectors from our manufacturing classification because standard (unprocessed or semi-processed) primary products account for over 90% of production in these sectors.

16 For instance, if 30% of the gross output of agriculture is used in the food-processing industry, then 30% of agricultural imports are also used in food processing. Similarly, if 40% of the gross output of the mineral sector goes to the iron and steel industry, so does 40% of the mineral imports.

17 The classification system used for delineating GPN products and further distinguishing between ‘producer-driven’ and ‘buyer-driven’ GPNs (see below) is discussed in the work of Athukorala (2019). It is important to note as a caveat that application of this classification to industries identified based on the two-digit input–output classification (based on the International Standard Industrial Classification) does not permit the precise delineation of the characteristics of GPN products. This is because the output of some ‘GPN industries’ identified at the two-digit level is a combination of production based on global production sharing and production for the domestic market, and normally, the import content of the former tends to be higher than that of the latter (Brumm et al. 2019; Koopman, Wang and Wei 2014).

18 In the late 2010s, buyer-driven GPN products made up less than 8% of total GPN exports from South Korea, Malaysia, Singapore and Thailand (Athukorala 2019, table 2).

19 There was an increase in industrial protection (import tariffs) during 2011–15 (Patunru and Rahardja 2015). However, it is unlikely that this protectionist tendency would have contributed to the increase in the domestic value added of GPN exports, for two reasons. First, the implied structure of GPN industries is determined by the structure of global production sharing rather than by changes in the trade policy regime of a given country. Second, the trade policy regime in Indonesia provides domestic firms involved in GPNs with duty-free access to imported inputs under the duty rebate provisions.

20 For this reason, value added per worker is also used as an alternative measure of capital intensity: capital deepening tends to increase the relative output of the sector with a greater capital share (Acemoglu and Guerrieri 2008).

21 As noted above, firms in GPN-related industries have the opportunity to specialise in a given task within the global manufacturing value chain that fits with its relative cost advantage.

22 Alternative estimates of the model with the GPN intercept dummy and without industry fixed effects are reported in appendix 2 for comparison.

APPENDIX

Table A1. Indonesia’s Manufacturing Sectors, 1995–2016: Domestic Value Added, Employment and Export Earnings

Source: Computed from the Indonesian input–output tables, provided by BPS.

Note: Cells are blank where data were not available. The IO code is the product code used in the IO table, based on the Indonesian Standard Industrial Classification.

Table A2. Value-Added Share of Exports, and Total Value Added in Indonesian Manufacturing: Alternative Estimates with a GPN Gummy Instead of Industry Fixed Effects

Source: Authors’ calculation.

Notes: Heteroscedasticity-corrected (robust) standard errors are in parentheses. ^a Time (year ) dummy with the year 1995 as the base dummy.

*** p < 0.01, ** p < 0.05, * p < 0.1.