How important is GVC participation to export upgrading?*

Abstract

Exporting higher quality and complex products is deemed pathway to economic growth and development. However, producing such products is knowledge-intensive and requires quality intermediate inputs and advanced technologies. Integration into global trade networks is increasingly argued to be among the pathways to obtain such inputs and technologies, although not all countries may benefit equally from such integration. This paper builds on these arguments and investigates how participation in the global value chain (GVC) affects the quality of exported products. Using a highly disaggregated product-level export data from 122 countries, we find that participation in (backward and forward) GVC impacts positively on the quality of exported products and brings the quality level closer to the quality frontier. While this result persists in the sub-sample comprising developed countries, developing countries only benefit from backward GVC participation. Overall, the results indicate that GVC participation matters to export upgrading but points to a potential heterogeneity on the channel of impact across countries at different levels of development.

1 INTRODUCTION

Integration into global value chains (GVCs) has shifted the development paradigm of many countries, presenting countries with new development possibilities, all made possible by widespread removal of policy barriers to trade and foreign investment. Falling trade costs through declines in transportation and communication costs have allowed MNCs in pursuit of efficiency to globalise their production. Nowadays, production is sliced into a different sequence of stages of value-creating activities or tasks that are performed in different cost-saving locations across the world (Foster-McGregor et al., 2015; Grossman & Rossi-Hansberg, 2008). In line with the rise of GVC, there is now a large body of literature suggesting that through GVC participation, firms are provided with essential opportunities to access the international markets, specialise in core tasks, access higher quality and sophisticated inputs, and benefits from new ideas, technology transfer, and spillover to stimulate productivity growth and expand the scale of exports (Collier & Venables, 2007; Criscuolo & Timmis, 2017; Pahl & Timmer, 2019).

However, what is still understudied in the literature is whether, in addition to increasing productivity and expanding the scale of exports, integration into GVC has any significant effect on the quality of exported products. Countries stand to gain a lot from producing and exporting higher quality products. There seems to be a consensus, at least, in the economic literature, that the quality of a country's exported products is essential for economic growth and competitiveness as countries become what they produce (Hausmann et al., 2007; Rodrik, 2006). The signalling effect of the quality of exported products tends to influence global patterns of trade (Hallak, 2006; Papageorgious et al., 2019; Verhoogen, 2008). In this regard, countries with more diversified exports at the higher end of the quality spectrum tend to grow faster, by capitalising on their comparative advantages to boost export revenues while creating jobs (Amiti & Freud, 2010; Broda & Weinstein, 2006; Funkel & Ruhwedel, 2001; Hausmann et al., 2007; Hummels & Klenow, 2005). Hence, it is crucial to understand the effect of participation in GVC and the quality of exported products, especially given the rise in the share of globally traded intermediates.

In this paper, we fill this gap in the literature by focusing on the potential effects of GVC participation on the quality of exported products. In particular, we examine whether improvements in the quality level of exported products are determined by the extent of GVC participation, the type of GVC participation and the development level of the GVC participating country. We answer this question using a highly disaggregated product-level export data from 122 countries, which we merged to industry-level indicators of GVC participation. Our product quality measure is based on the novel method developed by Khandelwal et al. (2013) that infers quality from an empirical demand function. Following the extant literature (see Amendolagine et al., 2019; Banh et al., 2020; Carril-Caccia & Pavlova, 2020; Foster-McGregor et al., 2015; Kummritz et al., 2017; UNCTAD, 2013; Wang et al., 2019), our measure of GVC participation is computed as the sum of the share of foreign value-added used in a country's exports (we call this backward GVC participation) and the share of country's domestic value-added that enters as an intermediate input in the value-added exported by other countries (we call this forward GVC participation).

To preview our findings, results from the full sample analysis indicate that GVC participation is crucial for improving the quality of exported products, and this works through forward and backward GVC participation. Additional analysis based on the full sample revealed that the benefits associated with GVC participation do not only lead to increased export quality but also allow countries to catch-up with the frontier export quality. Also, we find that the positive effect of GVC participation, including backward and forward GVC participation, tends to be higher for more complex and differentiated products. Analysis of two subsamples comprising developed and developing countries reveals different but interesting results. In particular, for the sample comprising developed countries, improvements in the quality of exported products that are associated with GVC participation work through both backward and forward linkages, with the latter exerting a higher influence. In the sample comprising developing countries, however, improvements in the quality of exported products that are associated with GVC participation only work through backward GVC participation. Hence, while our results generally suggest that GVC participation is crucial for quality upgrading of exported products, they underline a potential heterogeneity in the effect and on the channel of impact across countries at different levels of economic development.

To our knowledge, this paper provides the first empirical evidence on the relationship between GVC participation and the quality of exported products. Nevertheless, the paper is related to the rapidly growing literature on the level and growth effects of the recent waves of globalisation. Among others, this literature includes studies examining the impact of input liberalisation on productivity, the quality and scope of exported products (Amiti & Khandelwal, 2013; Bas & Strauss-Khan, 2014; Fan et al., 2015; Xu & Mao, 2018), trade and industrial upgrading (Gereffi, 1999; Kummritz et al., 2017). It also includes studies examining the impact of foreign direct investment (FDI) and imports on productivity (Iršová & Havránek, 2013; Javorcik, 2014) and the quality of exported products (Amighini & Sanfilippo, 2014; Harding & Javorcik, 2012; Zhu & Fu, 2013). Other studies in this literature for which the current study is related to include studies examining the productivity effects of GVC participation (Banh et al., 2020; Constantinescu et al., 2019; Pahl & Timmer, 2019) and the erstwhile case study literature examining GVC and industrial policy (Gereffi & Sturgeon, 2013; Mottaleb & Sonobe, 2011) and governance of GVC, lead firms, standardisation requirements and firm upgrading (Gereffi, 2014).

The remainder of this paper is organised as follows: Section 2 presents the theoretical background linking GVC participation and export upgrading. Section 3 specifies the model and describes the data sources used in the empirical analysis; Section 4 presents and discusses the results. Section 5 concludes.

2 THEORETICAL FRAMEWORK

The growth and development experience of some developing countries in recent years has demonstrated the importance of the quality of exported products for long-term economic growth. For instance, Rodrik (2006) and Hausmann et al. (2007) present persuasive theoretical and empirical evidence suggesting that countries promoting exports of more sophisticated products grow faster. Higher quality and sophisticated products are less vulnerable to price competition from low-wage producers. This helps to boost export revenues and productivity, which ultimately contributes to the country's economic growth and development (Amiti & Khandelwal, 2013; Henn et al., 2020; Khandelwal, 2010). For this reason, it is widely argued that it is not how much of the products exported that matter, but the quality of exported products. However, producing quality products for export requires a mix of intermediate inputs, some of which are often outside the bounds of a firm.

Several studies have examined the drivers of the quality level of exported products. In these studies, it has been found that the quality of products exported by a country is enhanced by factors such as capital deepening, investment in R&D, FDI, imports, technology transfer, human capital development, institutional quality, factor endowments and access to credit (Amighini & Sanfilippo, 2014; Crino & Ogliari, 2017; Harding & Javorcik, 2012; Xu & Mao, 2018; Zhu & Fu, 2013). While these studies provide useful insights, they do not pay attention to the importance of GVC participation. For instance, Zhu and Fu (2013), in a comprehensive study, acknowledged the importance of imports for upgrading the quality of exported products. However, the authors used imports to GDP ratio as their measure of the importance of sources of foreign embodied knowledge. Harding and Javorcik (2012), on the other hand, examined the importance of foreign knowledge content via FDI on the unit values of exported products. While these studies provide relevant insights on the importance of GVC integration for the upgrading of exported products, none of them employed standard measures of GVC participation. Hence, the mechanisms underlying such a potential relationship are totally ignored. This gap in the literature calls for improvement in the measurement of GVC integration and its subsequent inclusion as one of the main drivers of product quality upgrading.

The nexus between GVC participation and the quality of exported products can be rationalised on the premise that participation in GVC leads to technology transfer, knowledge spillover, and access to cheaper and higher quality intermediate inputs needed for productivity improvement, which are important drivers of export upgrading (Collier & Venables, 2007; Criscuolo & Timmis, 2017; Pahl & Timmer, 2019). The conventional view is that GVC allows firms to use inputs of different quality. In this relation, the ability of a firm to operate at the higher end of the quality spectrum of the export value chain, command a higher price and generate higher sales for its exports in the global competitive market is dependent on the use of a variety of superior quality intermediate inputs in the production of its exports (Bas & Strauss-Khan, 2013; Halpern et al., 2015; Manova & Yu, 2017).

Through the sourcing of intermediate inputs, GVC participation creates opportunities for fast technological learning and skill acquisition for local firms, given that they have built their absorptive capacity (Pietrobelli, 2008; Pietrobelli & Rabelloti, 2011). In this relationship, an upstream GVC partner through interactions is able to transfer technology (through embodied and disembodied capital and intermediate input use) to a firm downstream in the supply chain to aid the latter to produce more efficiently and quality-upgrade their exports (Bas & Strauss-Khan, 2014; Goldberg et al., 2010; Halpern et al., 2015; Harding & Javorcik, 2012; Xu & Mao, 2018). Related to this, Xu and Mao (2018) argue that the use of these imported new materials or services can improve the quality of exported goods either through the ‘variety effect’ or ‘innovation effect’. The variety-effect channel works by providing GVC participating firms with a wider variety of inputs that they can choose from to produce final goods. The innovation-effect channel, on the other hand, works via the advanced technologies embedded in the variety of imported inputs from which producers choose the optimal technology to produce final goods. Thus, policies that inhibit access to foreign intermediates could have adverse effect on the quality of final goods exports (Kasahara & Lapham, 2013).

Other studies also suggest that GVC participation leads to specialisation in core tasks, which contributes positively to firm productivity (for an extensive literature review, see Amador and Cabral (2015)). The underlying argument here is that participation in GVC offers firms the opportunity to outsource activities they have a less comparative advantage and concentrate on core activities where they have a competitive advantage while using limited production resources more efficiently. This may provide an ideal opportunity for firms, particularly those in developing countries to specialise in niche product categories, instead of struggling to build capabilities to master entire production systems. By specialising in their most efficient core activities in the value chain, these firms are able to invest their resources to build technological and other specific capabilities for the specific GVC functions and activities they perform in the value chain in order to meet the global standardisation requirements of lead firms and to remain competitive in the value chain which ultimately affects the quality of their produce and export (Baldwin, 2012; Gereffi, 2014; Lall, 1992; Morrison et al., 2008; Newman et al., 2016; Schwörer, 2013).

In the sections that follow, we bring the forgoing arguments to data within an empirical framework that envisions export upgrading as a function of GVC participation. While the discussions so far focused on firms,11 In fact, in reality, economic actors in the business of GVC participation are firms. for lack of comparable firm-level data across different countries, we test our research idea using industry-level GVC indicators that we merge to product-level quality measures.

3 DATA SOURCE AND MODEL SPECIFICATION

The two most important variables for our analysis are measures of export quality and GVC participation. This section shows how we compute these variables. It also presents our empirical model.

3.1 Measuring GVC participation

Countries, or more specifically firms, participate in GVCs as ‘buyers’ (also known as backward participation) and/or ‘sellers’ (also known as forward participation). Backward participation captures the share of foreign value-added that is embodied in a country's gross exports, while forward participation is the share of domestic value-added in the exports that a country sells to other countries that are used in the production of exports of these countries. In other words, it is the share of domestic value-added that is used in the export of third countries. A measure of a country's overall involvement in GVC is, therefore, best captured by a metric that simultaneously accounts for a country's engagement in GVC through backward and forward participation. Consequently, we follow extant literature (see Amendolagine et al., 2019; Banh et al., 2020; Carril-Caccia & Pavlova, 2020; Foster-McGregor et al., 2015; Kummritz et al., 2017; UNCTAD, 2013; Wang et al., 2019) to define the extent of GVC participation of country c in sector s and period t as:

(1)

where is the share of foreign value-added used in a country's export in a sector, is the share of a country's domestic value-added that enters as inputs in the exports of other countries and is country c gross export in sector s. Larger values of the index indicate more intensive participation in the GVC. Equation 1 is considered a standard measure of GVC participation in the nascent GVC literature as it acknowledges that countries and firms participate in GVC participation either as ‘buyers’ or ‘sellers’ or both (e.g., see Foster-McGregor et al., 2015; UNCTAD, 2013).22 It is important to note that some studies either focus on backward participation (Pahl & Timmer, 2019) or forward participation (Baldwin & Gonzalez-Lopez, 2015) as a proxy measure for GVC participation. However, these studies often cite a dearth of data and the objective of the study as the rationale for adopting these proxy measures. For example, Pahl and Timmer (2019) used only backward linkages as a proxy measure of GVC. In their paper (see page 5), the authors acknowledge the limitation of using this approach as it ignores completely the country's forward participation in GVC. Hence, it gives a less than complete picture of a country's overall GVC participation. However, because we are interested in the mechanisms through which GVC participation impacts the quality of exported products, we also consider the independent effects of backward (i.e., ) and forward GVC participation (i.e., . In line with our definition above, higher values of for industry s in country c are interpreted as higher engagement in GVC through backward participation, while higher values of are interpreted as higher engagement in GVC through forward participation.

To compute the indicators of GVC participation, we rely on the latest release of the EORA MRIO I-O database of the University of Sydney (https://worldmrio.com), which provides the largest coverage of countries (including developed and developing countries) that we are aware of. The database is widely used (see Amendolagine et al., 2019; Aslam et al., 2017; UNCTAD, 2013; Wang et al., 2019; World Bank WDR, 2020) and the reliability and accuracy of the database are discussed in detail in Lenzen et al. (2013) where it compares well with comparator databases such as the GTAP database, OECD–WTO data and the WIOD database (UNCTAD, 2013).

The MRIO table provides information on the amounts of intermediates needed for the production of the gross output. Each column in the MRIO table provides the domestic and foreign share of intermediate in the production of one unit of output. Borrowing the notations in UNCTAD (2013), the MRIO table can be translated into a standard I-O matrix as follows;

(2)

where is gross output, is intermediate demand, is final demand, is the identity matrix and is the technical coefficient matrix and is the Leontief inverse.

(3)

From this framework, we follow Hummels et al. (2001), Koopman et al. (2011), Koopman et al. (2014), and Aslam et al. (2017) to calculate trade in value-added defined as the value-added embodied in gross trade flow. Three components are needed to calculate trade in value-added , namely the Leontief inverse matrix , value-added shares (value-added per unit of output) matrix and total gross exports matrix . The Leontief matrix is recovered from the technical coefficient matrix () by dividing elements of the matrix with corresponding elements of the vector. The value-added shares matrix () is obtained by summing across rows of the () matrix and subtracting all the elements on the diagonal of the square matrix from an identity matrix. is arrived by multiplying the two components and , and the diagonalised row vector of the total gross exports matrix () (Aslam et al., 2017; Foster-McGregor et al., 2015) and is given as:

(4)

with and as () diagonalised row vector of value-added shares and gross exports for each industry and country is the () Leontief inverse of country . Given , we decompose and compute the indicators for GVC participation. First is the DVX, which is the row sum of the matrix, excluding diagonal terms. The second indicator, FVA, is then the column sum of the matrix excluding diagonal terms.

Our analysis focuses on 10 sectors. Table 1 shows the average indexes of GVC participation in the sample, while Table 2 shows their respective averages over the sample period in each sector. Interesting patterns emerge when we examine Table 2, particularly for the two subsamples of developed and developing countries. In particular, on average, developed countries are more integrated into GVC than developing countries and this tends to persist across all the sectors. When we consider the two subcomponents of GVC, developed countries, on average, outperform developing countries across sectors.

Table 1. Data sources and basic summary statistics

Variable Names	Data Sources	Obs	Mean	Standard Dev.	Min	Max
Product Quality	Author's Calculation using BACI-CEPII Dataset	5,164,855	−0.552	1.121	−8.439	11.132
Backward GVC Participation (FVA)	Author's Calculation using Eora MRIO Database	5,164,855	0.258	0.140	0.003	0.942
Forward GVC Participation (DVX)	Author's Calculation using Eora MRIO Database	5,164,855	0.207	0.133	0.001	0.952
Aggregate GVC Participation [(FVA + DVX)/Gross Export]	Author's Calculation using Eora MRIO Database	5,164,855	0.465	0.162	0.082	0.999
Human Capital	Penn World Table version 9	5,164,855	2.734	0.627	1.053	3.734
Institutional Quality (Rule of Law)	World Governance Indicator	5,164,855	0.450	0.999	−2.123	2.100
Inflation Rate	World Development Bank	5,164,855	6.868	3.666	−8.439	4,146
Financial Development	(Svirydzenka, 2016).	5,164,855	0.456	0.244	0	1

Table 2. Summary statistics of main variables over the sample period

Sector	Full sample				Developed Countries				Developing Countries
	GVC	FVA	DVX	quality	GVC	FVA	DVX	quality	GVC	FVA	DVX	quality
Agriculture	0.430	0.178	0.252	−1.188	0.507	0.237	0.270	−0.927	0.377	0.139	0.239	−1.366
Electrical & Machinery	0.480	0.293	0.187	−0.073	0.511	0.355	0.156	0.270	0.458	0.248	0.210	−0.320
Fishing	0.452	0.211	0.241	−0.545	0.502	0.252	0.250	−0.360	0.409	0.177	0.233	−0.704
Food & Beverages	0.370	0.219	0.151	−1.151	0.423	0.274	0.149	−0.988	0.332	0.180	0.152	−1.271
Metal Products	0.528	0.256	0.272	−0.588	0.598	0.311	0.287	−0.339	0.469	0.209	0.259	−0.799
Minin & Quarrying	0.514	0.184	0.329	−1.800	0.567	0.236	0.331	−1.644	0.470	0.143	0.327	−1.928
Petroleum, Chemical and Nonmetallic	0.497	0.286	0.211	−0.752	0.552	0.358	0.194	−0.498	0.447	0.220	0.226	−0.980
Textile & Wearing Apparel	0.418	0.256	0.163	−0.158	0.506	0.318	0.188	0.020	0.352	0.207	0.145	−0.291
Transport Equipment	0.470	0.278	0.192	−0.018	0.510	0.349	0.161	0.337	0.442	0.227	0.214	−0.269
Wood & Paper	0.476	0.234	0.242	−0.671	0.525	0.282	0.242	−0.417	0.439	0.197	0.241	−0.864

Source: Author's Computation.

3.2 Measuring export quality

The quality of a product is an unobserved attribute and, therefore, needs to be inferred. Hence, to obtain a quality measure, we rely on the novel approach developed by Khandelwal et al. (2013). While the approach infers quality based on estimating an empirical demand function, the underlying intuition behind the approach is that conditional on price, a variety with a higher quantity is assigned higher quality. Following Khandelwal et al. (2013), therefore, we infer the quality of product k shipped to a destination country d by country c in year t, via the following empirical demand equation:

(5)

where is the quantity of a product exported by country c to a destination country, and and are the price and the quality of the exported product. and are the price index and income level of the destination country, while is the elasticity of substitution and . We take logs of the empirical demand equation and then use the residual from the following OLS regression to infer quality:

(6)

where is destination country time-varying fixed effects which absorb both and , and is product fixed effects which capture differences across product categories. is the error term. The inferred quality estimate is, then, given by the estimated residual of Equation (3) as: .33 Note that, as price and quantity are expressed in logarithms in Equation 5, the estimated quality will be in logarithm form as well. We follow Manova and Yu (2017) and set the elasticity of substitution across product variety at the commonly used value of 5.44 The results are qualitatively similar to when we set to 7 or 10. In what follows, we utilise the BACI-CEPII data set (Gaulier & Zignago, 2010) to estimate Equation 3. The data set contains bilateral export values and quantities at the 6-digit Harmonized System Classification (HSC) for a large number of countries. We define a product as a 6-digit HSC, while the price of each product is measured by the product's unit price (i.e., value divided by quantity). After estimating equation 3 using the data set, we construct a country–product-specific measure of quality by averaging across the importing countries. Hence, the quality index is not destination-specific and our empirical analysis does not require destination-specific variables such as bilateral distance. Finally, we use appropriate concordance table and map these products to the corresponding sector as our GVC participation indexes.

Table 1 shows the average export quality in our sample, while Table 2 shows the average quality in each sector in the sample. Table 2 shows that the sector with the highest average product quality is Transport equipment and Electric machinery, which is somewhat expected since those sectors contain highly differentiated products that tend to have a higher scope for quality adjustment. For the two subsamples of developed and developing countries, the average product quality in each sector is consistently lower in developing countries than in developed countries. This is consistent with the conventional view that developed countries tend to export higher quality and sophisticated products than developing countries (Hausmann et al., 2007; Hummels & Klenow, 2005).

3.3 Model specification

To examine the importance of GVC participation in export upgrading, we consider the following empirical model:

(7)

where denotes inferred quality level of product k country c exports in period t, is the logarithm of aggregate (or the backward and forward) GVC participation indicator of country i in sector s and in period t. We emphasise that while the export quality measure is at the product level, GVC participation is measured at the sector level. is a vector of country-specific time-varying characteristics. It includes human capital, inflation rate, institutional quality and financial development. The measurement and data sources of these variables are described in Table 1. The inclusion of these variables in our equation is informed by extant studies on the drivers of export quality (Amighini & Sanfilippo, 2014; Harding & Javorcik, 2012; Henn et al., 2020). Considering that our model includes several sources of heterogeneity that need consideration, we follow Harding and Javorcik (2012) and Amighini and Sanfilippo (2014) and include country–product fixed effects () and country–sector () fixed effects to account for any time invariant characteristic that may affect the quality level of exported products. We also include time dummies () to absorb varying effects that are common across countries. Finally, is the error term.

Although we argue that GVC participation increases export quality, it may well be that sectors with higher export quality level self-select into GVC, leading to reverse causality. While we acknowledge this as a potential problem, we believe that measuring export quality at the product level ameliorates this concern since it is less likely that export quality at a highly disaggregated 6-digit HSC would significantly drive the intensity of GVC participation of an entire sector. An exception to this is if only a few of these products account for the sector's entire export quality. To reduce the scale of this concern, therefore, we lagged the GVC participation variables by a period. Moreover, besides mitigating the reverse causality concern, this empirical setting enables us to account for the time needed for knowledge and technology acquired through GVC participation to be used in industrial activities that would lead to quality upgrading. Besides reverse causality, Equation 7 may also suffer from omitted viable bias. Hence, in addition to estimating Equation 7 with a panel fixed effect model, following Bahn et al. (2020), we also estimate it using an instrumental variable approach wherein we instrument GVC indicators with average GVC measure at the world sector level. Because our variable of interest varies at the country–sector–year level, while our outcome variable is at the more disaggregated country–product–year level, we cluster standard errors at the country–sector–year level following Harding and Javorcik (2012).

Our final sample comprises 122 countries over the period 1996–2015, and with a total observation of 5,164,855 (i.e., Country*Product*Year). The number of developed countries in the sample is 33 (with a total observation of 2,266,134), while the number of developing countries in the sample is 89 (with a total observation of 2,898,721). The choice of the countries in the sample and time period analysed is solely determined by data availability. Our definition of developed and developing countries follows the IMF classification wherein ‘Advanced Markets’ are classified as developed countries, while ‘Emerging Markets and Low Income Countries’ are all classified as developing countries.

4 EMPIRICAL RESULTS

This section proceeds in three steps. The first two subsections present the full sample results. The third section explores the differential effect of GVC participation on the export upgrading of differentiated and homogenous products, while the fourth section examines the export quality effects of GVC participation in developed and developing countries.

4.1 Aggregate GVC participation and quality upgrading: Full sample

We begin by discussing the results for the global sample, which we report in Table 3. Columns 1 and 2 show the panel fixed effects results. In particular, column 1 shows the result when we regress the quality level of exported products on aggregate GVC participation without controlling for other regressors. The result shows a significantly positive relationship between GVC participation and the quality level of exported products. In column 2, we include other regressors as contained in Equation 7. We find that our initial result remains unchanged in qualitative terms. Columns 3 and 4 show the second-stage results of the IV regression where we treat GVC participation to be endogenous. The results reported in the two columns are consistent with those reported in the previous columns. Regarding the appropriateness of the employed instrument, the Kleibergen-Paap statistic as reported in the lower part of the panel is between 243 and 245, suggesting that our instrument performs well and strongly predicts industry-level GVC participation. Hence, the results presented in Table 3 indicate that GVC participation increases the quality of exported products. Our result is consistent with the broader GVC-related literature arguing that GVC participation provides access to higher quality and sophisticated intermediate inputs, new ideas and technologies (Collier & Venables, 2007; Criscuolo & Timmis, 2017), which have been underscored in another strand of literature to be important drivers of export upgrading (Amighini & Sanfilippo, 2014; Harding & Javorcik, 2012; Xu & Mao, 2018).

Table 3. Baseline regression: GVC participation and export quality

	Panel Fixed Effects		IV
	[1]	[2]	[3]	[4]
GVC participation	0.0637*** [0.004]	0.0628*** [0.004]	0.1373*** [0.018]	0.1360*** [0.018]
Rule of Law		0.0670 [0.042]		0.0410 [0.042]
Human capital		0.0766*** [0.017]		0.0667*** [0.017]
Inflation		−0.0002*** [0.000]		−0.0002*** [0.000]
Financial Development		0.2445*** [0.049]		0.2307*** [0.048]
Kleibergen-Paap Wald F sta.	-	-	245.258	242.99
Observations	5,164,855	5,164,855	5,164,855	5,164,855
R-squared	0.22	0.22	-	-

Note

• Standard errors clustered at the country–sector–year level in squared brackets. Significance at the ***1%, **5% and *10% level. The dependent variable is the inferred quality (in logs) of the 6-digit HSC product from country c in year t, and the GVC participation indicator is measured at the sector level. All variables are lagged by a period. All regressions include country–product, country–sector and year fixed effects.

Regarding the control variables, the estimated coefficients of the inflation rate are negative and statistically significant at conventional levels in all the columns in the table, suggesting a distortionary impact macroeconomic instability may exert on productive economic activities. The estimated coefficients of human capital and financial development are both significantly positive, suggesting that skill-abundant countries and countries with a developed financial market export higher quality products. Finally, we obtain a statistically insignificant effect of institutional quality, as measured by rule of law, on export quality.

Next, in Table A1 in the appendix, we perform additional analysis to examine whether, in addition to improving the quality of exported products, GVC participation also contributes to bringing the quality level of exported products closer to the quality of the frontier. To do this, we compute a new outcome variable that is expressed relative to the quality frontier. Following Harding and Javorcik (2012), the frontier quality is defined as the 95th percentile of the distribution of inferred quality of product k exported in period t by all countries in our data set. The new dependent variable is then defined as the ratio of the quality level of product k exported by country c in period t to product k's frontier quality in the same period. The higher the value of the outcome variable, the closer the exporter is to the quality frontier. Hence, a positive estimated coefficient of the GVC participation index would then be a validation of the claim that GVC participation enables exporters to catch-up to the quality frontier. As reported in Table A1 in the appendix, we find strong evidence that GVC participation brings the quality level of exported products closer to the frontier quality. This additional result leads to the further conclusion that besides raising the level of exported products, participation in GVC brings the quality of exported products closer to the frontier quality.

4.2 Backward and forward GVC participation, and quality upgrading: Full sample

While the results reported in the previous section suggest an export quality gain due to a country's insertion into GVC, Morrison et al. (2008) argue that not all insertions into GVCs are beneficial. A similar argument has been raised by Banga (2013), who further suggests a disentangling of the GVC participation impact into forward and backward participation to provide more useful insights into the gains that go to a country from participating in GVC. To this end, we focus on the two subcomponents of GVC participation (i.e., backward and forward GVC participation) and examine their individual effects on the quality level of exported products. Table 4 reports the results of this exercise. Columns 1 and 3 present the panel fixed effects results. The estimated coefficients of both backward and forward GVC participation all turn out significantly positive in both columns. Columns 2, 4 and 5 show the IV regression results. The results are consistent with the panel fixed effects results. Hence, Table 4 results suggest that the export quality effects of aggregate GVC participation we observed in Table 3 accrue from both backward and forward GVC participation. Regarding the control variables, the results are consistent with those reported in Table 3.

Table 4. Backward and forward GVC participation, and export quality

	Panel Fixed Effects	IV	Panel Fixed Effects	IV	IV
	[1]	[2]	[3]	[4]	[5]
Backward GVC [ln]	0.0585*** [0.004]	0.1030*** [0.019]			0.0669** [0.031]
Forward GVC [ln]			0.0463*** [0.003]	0.1449*** [0.018]	0.0788** [0.035]
Human Capital	0.0679 [0.042]	0.0516 [0.042]	0.0793* [0.042]	0.0578 [0.042]	0.0476 [0.041]
Rule of Law	0.0754*** [0.017]	0.0680*** [0.018]	0.0809*** [0.017]	0.0719*** [0.017]	0.0668*** [0.017]
Inflation	−0.0002*** [0.000]	−0.0002*** [0.000]	−0.0002*** [0.000]	−0.0001** [0.000]	−0.0002*** [0.000]
Financial Development	0.2463*** [0.049]	0.2386*** [0.049]	0.2458*** [0.049]	0.2232*** [0.049]	0.2267*** [0.048]
Kleibergen-Paap Wald F sta.	-	133.98		245.757	54.244
Observations	5,164,855	5,164,855	5,164,855	5,164,855	5,164,855
R-squared	0.22	-	0.22	-	-

Note

• Standard errors clustered at the country–sector–year level in squared brackets. Significance at the ***1%, **5% and *10% level. The dependent variable is the inferred quality (in logs) of the 6-digit HSC product from country c in year t, and the GVC participation indicator is measured at the sector level. All variables are lagged by a period. All regressions include country–product, country–sector and year fixed effects.

4.3 GVC Participation, complex products and quality upgrading

Thus far, our analysis found a positive association between GVC participation and the quality level of exported products. While the analysis that led to this result relied on the implicit assumption that products within an industry respond evenly to variations in the extent of industry-level GVC participation, there are at least two reasons to believe otherwise. First, compared to homogenous products, differentiated products have greater scope for quality adjustment (Crino & Ogliari, 2017; Fan et al., 2015), meaning that they would be more responsive to variations in the extent of GVC participation. Second, most differentiated products are complex in nature, requiring producers to source specialised and sophisticated inputs and know-how, which being inserted into GVC easily avails the opportunity. This occurs either directly through outright knowledge and technology transfer or indirectly through knowledge spillover. Against this backdrop, we further examine whether the association between GVC participation and the quality level of exported products tends to be stronger in differentiated products. To address this question, we modify equation 1 by introducing an interaction variable comprising the GVC participation index and a dummy variable that equals 1 if a product is differentiated and zero if otherwise.

To capture products that offer greater scope for differentiation, we follow Harding and Javorcik (2012), Fan et al. (2015), and Crino and Ogliari (2017) by using Rauch (1999) product classification. Rauch defined differentiated products as those not having a reference price or whose price is not quoted on organised exchanges. Rauch suggested two definitions, a conservative (which minimises the number of products that are classified as homogeneous) and a liberal one (which maximises the number of products that are classified as homogeneous). We employ both classifications and report the results in Table 5. In all the columns reported in the Table, the estimated coefficients of the interaction variable between the respective GVC participation indexes and differentiated product dummy are consistently positive and statistically significant at all conventional significance levels. Hence, the results lead to the further conclusion, that while GVC participation leads to export upgrading, this effect tends to be higher for more complex and differentiated products.

Table 5. GVC participation, differentiated products and export quality

	Panel Fixed Effects
	Liberal Classification			Conservative Classification
	[1]	[2]	[3]	[4]	[5]	[6]
GVC*Differentiated	0.0207*** [0.001]			0.0206*** [0.001]
Backward*Differentiated			0.0218*** [0.001]			0.0217*** [0.001]
Forward*Differentiated		0.0221*** [0.001]			0.0219*** [0.001]
GVC	0.0486*** [0.004]			0.0478*** [0.004]
Backward GVC		0.0536*** [0.005]	0.0402*** [0.005]		0.0535*** [0.005]	0.0393*** [0.005]
Forward GVC		−0.0074* [0.004]	0.0057 [0.004]		−0.0081** [0.004]	0.0056 [0.004]
Human Capital	0.0710* [0.043]	0.0719* [0.043]	0.0718* [0.043]	0.0715* [0.043]	0.0723* [0.043]	0.0723* [0.043]
Rule of Law	0.0763*** [0.017]	0.0752*** [0.017]	0.0752*** [0.017]	0.0762*** [0.017]	0.0751*** [0.017]	0.0751*** [0.017]
Inflation	−0.0002*** [0.000]	−0.0002*** [0.000]	−0.0002*** [0.000]	−0.0002*** [0.000]	−0.0002*** [0.000]	−0.0002*** [0.000]
Financial Development	0.2469*** [0.049]	0.2480*** [0.049]	0.2482*** [0.049]	0.2468*** [0.049]	0.2479*** [0.049]	0.2482*** [0.049]
Observations	5,164,303	5,164,303	5,164,303	5,164,303	5,164,303	5,164,303
R-squared	0.23	0.23	0.23	0.23	0.23	0.23

Note

• Standard errors clustered at the country–sector–year level in squared brackets. Significance at the ***1%, **5% and *10% level. The dependent variable is the inferred quality (in logs) of the 6-digit HSC product from country c in year t, and the GVC participation indicators are measured at the sector level. All variables are lagged by a period. All regressions include country–product, country–sector and year fixed effects.

4.4 GVC participation and quality upgrading: Development level

Table 6 reports the panel fixed effects results for two subsamples: developed and developing countries. In both samples, the estimated coefficients of the aggregate GVC participation index are significantly positive (see columns 1, 2, 4 and 5), although the positive effect tends to be higher in developed countries as suggested by the magnitude of the estimated coefficients. Columns 3 and 6 report the results for the different components of the aggregate GVC participation index. As the result indicates, backward and forward GVC participation is associated with improvements in the quality level of exported products from developed countries. The results are, therefore, consistent with those of the global sample where we find that GVC participation increases the quality level of exported products through both backward and forward GVC participation.

Table 6. GVC participation and export quality: Developing and developed countries

	Panel Fixed Effects
	Developed Countries			Developing Countries
	[1]	[2]	[3]	[4]	[5]	[6]
GVC Participation	0.1153*** [0.002]	0.0797*** [0.002]		0.0824*** [0.001]	0.0722*** [0.002]
Backward GVC			0.0104*** [0.004]			0.0473*** [0.003]
Forward GVC			0.0738*** [0.004]			0.0031 [0.003]
Human Capital		0.0994*** [0.009]	0.0642*** [0.009]		0.0604*** [0.005]	0.0000 [0.029]
Rule of Law		0.0164** [0.007]	0.0202*** [0.007]		0.0457*** [0.005]	0.0411*** [0.010]
Inflation		−0.0183*** [0.002]	−0.0173*** [0.001]		−0.0003*** [0.000]	−0.0002*** [0.000]
Financial Development		0.3321*** [0.021]	0.2698*** [0.021]		0.0349 [0.030]	0.5563*** [0.046]
Observations	2,266,134	2,266,134	2,266,134	2,898,721	2,898,721	2,898,721
R-squared	0.69	0.69	0.69	0.48	0.48	0.50

Note

• Standard errors clustered at the country–sector–year level in squared brackets. Significance at the ***1%, **5% and *10% level. The dependent variable is the inferred quality (in logs) of the 6-digit HSC product from country c in year t, and the GVC participation indicators are measured at the sector level. All variables are lagged by a period. All regressions include country–product, country–sector and year fixed effects.

Regarding the sample comprising developing countries, the estimated coefficients of backward and forward GVC participation are positive as reported in column 6. However, only the estimated coefficient of backward GVC participation is statistically significant at the conventional significance level. This suggests that if developing countries are ever gaining from the insertion into GVC in terms of an improvement in the quality level of exported products, it is only through backward GVC participation as they are allowed the opportunity to benefit from outright knowledge and technology transfer, or indirectly through knowledge spillover. One of the possible explanations for the insignificant effect of forward GVC participation could be that compared to developed countries, developing countries, as shown by Foster-McGregor et al. (2015), participate in GVC more strongly through forward integration. However, despite their high participation through this channel, they end up supplying primary inputs that have less scope for quality adjustment. Therefore, they may be locked-up into low value-added activities in the value chain, which limits the opportunity for upgrading and hence might not benefit from GVC integration through this channel to improve the quality of their exports.

5 CONCLUSION

Economic growth and development require the transformation of a country's economic structure. This relies upon, among other things, the improvement in the quality of products produced and exported. Producing and exporting higher quality products often requires access to advanced technologies and higher quality intermediate inputs. Participation in GVC could offer a pathway to access these technologies and intermediate inputs providing economies with opportunities for fast-track development through export upgrading. Consequently, the current study examined the impact of GVC participation on export quality across countries at varying levels of development. We specifically examine if improvements in the quality level of exported products are conditional on the type of GVC participation and the development level of the country.

Results from the full sample analysis indicate that improvements in the quality levels of exported products are positively associated with the extent of GVC participation, and this works through forward and backward GVC participation. Additional analysis on the full sample revealed that the benefits associated with GVC participation do not only lead to increased export quality but also allow countries to catch-up with the frontier export quality. Also, we find that the positive effect of GVC participation, including backward and forward GVC participation, tends to be higher for more complex and differentiated products. Analysis of two subsamples comprising developed and developing countries reveals different but interesting results. In particular, for the sample comprising developed countries, improvements in the quality level of exported products that are associated with GVC participation work through both backward and forward linkages, with the latter exerting a higher influence. In the sample comprising developing countries, however, improvements in the quality level of exported products that are associated with GVC participation only work through backward GVC participation. Our findings, therefore, indicate that GVC participation matters for improving the quality levels of exported products, but the channels underlying the impact differ across countries of different development levels.

Our findings have implications for policy, especially on the supply of intermediate inputs through GVC participation. Inadequate supply of quality foreign intermediate inputs could be a constraining factor to export upgrading. Hence, policies that inhibit a country's participation in GVC could be strategically reformed to allow firms, particularly those in developing economies, to integrate into the global supply chain to access these crucial foreign intermediate inputs. Achieving this will require exerting effort towards gaining greater market access to the global trade network through favourable trade agreements, such as preferential tariffs on imported intermediates and implementing policies that would attract lead firms and global suppliers in the value chain. Such policies would also include adopting better trade and investment promotion policies, competitive exchange rate regimes, favourable and attractive FDI policies, improved business environment together with analog complements such as building infrastructure, and strengthening connectivity.

We conclude with a few caveats that inform potential areas for further research. First, our study only considered heterogeneity that may arise due to differences in development level. Further studies could consider the role of other country characteristics such as human capital and the broader business climate and their interaction with GVC participation. Second, our export quality indicator only captures within product quality upgrading, that is, the production of similar products with higher quality. GVC participation could also lead to other types of upgrading, such as movements to higher quality products within industries or movement into other sectors using the knowledge acquired in another value chain. Theoretical discussions on these types of upgrading and how they are related to GVC participation are contained in Humphrey–Schmitz (2002).55 Also see Timmer et al. (2019) for discussion on functional specialisation. However, the empirical analysis that tests these relationships is lacking. Our study is just one in that direction and more needs to be done in this area of research.

Appendix

Table A1. GVC participation and quality frontier

	Panel Fixed Effects		IV
	[1]	[2]	[3]	[4]
GVC Participation	0.0570*** [0.003]	0.0562*** [0.003]	0.0753*** [0.017]	0.0745*** [0.017]
Human Capital		0.0695** [0.034]		0.0630* [0.035]
Rule of Law		0.0775*** [0.016]		0.0750*** [0.016]
Inflation		−0.0002*** [0.000]		−0.0002*** [0.000]
Financial Development		0.1678*** [0.042]		0.1643*** [0.042]
Kleibergen-Paap Wald F sta.	-	-	245.258	242.99
Observations	5,164,855	5,164,855	5,164,855	5,164,855
R-squared	0.14	0.14	-	-

Note

• Standard errors clustered at the country–sector–year level in squared brackets. Significance at the ***1%, **5% and *10% level. The dependent variable is the quality frontier defined as the ratio of the quality level of product k exported by country c in period t to product k's frontier quality in the same period, where frontier quality is defined as the 95th percentile of the distribution of inferred quality of product k exported in period t by all countries in our data set. The GVC participation indicator is measured at the sector level. All variables are lagged by a period. All regressions include country–product, country–sector and year fixed effects.