Facilitating Backward Global Value Chain Participation in South Asia: The Role of the South Asian Free Trade Agreement

Abstract

This study examines the impact of the South Asian Free Trade Agreement (SAFTA) on participation in global value chains (GVCs) among South Asian economies, specifically Bangladesh, India, and Pakistan. This research offers new empirical insights into the relatively underexplored relationship between SAFTA and GVCs in the region. The findings indicate that SAFTA has promoted backward GVC participation by increasing the foreign value-added content of exports, particularly from India to Bangladesh and Pakistan, and from Pakistan to Bangladesh. These results suggest untapped potential for expanding regional GVC linkages, as many bilateral GVC connections within South Asia remain underdeveloped.

1. Introduction

Asian economies are at the forefront of global economic development. While East and Southeast Asian economies have experienced substantial growth, South Asian economies continue to lag behind. South Asian economies are often characterized by premature industrialization and weak integration into global value chains (GVCs).

Table 1. Profile of South Asian countries.

	Population in 2022 (in Million)	GDP per Capita in 2022 (in US$)	Income Class in 2023	Manufacturing % of GDP in 2022	GVC Participation in Manufacturing % in 2020
Afghanistan	34.3	422	low	10.2	-
Bangladesh	168.5	2731	lower-middle	21.8	25.2
Bhutan	0.8	3775	lower-middle	8.7	-
India	1417.2	2366	lower-middle	13.1	27.0
Maldives	0.4	15,962	upper-middle	2.0	-
Nepal	30.5	1337	lower-middle	4.8	-
Pakistan	227.0	1651	lower-middle	13.8	12.5
Sri Lanka	22.4	3342	lower-middle	19.7	-
Average	-	-	-	11.8	21.6
Malaysia	32.7	12,466	upper-middle	23.4	42.6
Thailand	70.1	7073	upper-middle	27.1	38.1

Sources: Population and GDP per capita: World Economic Outlook Database. International Monetary Fund. Retrieved from https://www.imf.org/en/publications/weo (accessed on 1 June 2025). Income Classification: World Bank, https://datahelpdesk.worldbank.org/knowledgebase/articles/906519 (accessed on 1 June 2025). Manufacturing, value-added (% of GDP): World Bank Open Data, https://data.worldbank.org/ (accessed on 1 June 2025). GVC Participation: Trade in Value-added (TiVA 2023) database, Organization for Economic Co-operation and Development (OECD), https://data-explorer.oecd.org/ (accessed on 1 June 2025).

compares the economic profiles of South Asian economies (Afghanistan, Bangladesh, Bhutan, India, the Maldives, Nepal, Pakistan, and Sri Lanka1) with those of selected Southeast Asian economies (Malaysia and Thailand). As of 2023, only the Maldives was classified as an upper-middle-income country2, while the rest fall within the lower-middle- or low-income groups. In terms of industrialization, the average manufacturing value-added as a percentage of GDP in South Asian countries was 11.8% in 2022, significantly below the levels in emerging Southeast Asian economies such as Malaysia (23.4%) and Thailand (27.1%). Likewise, the average GVC participation rate3 in manufacturing in 2020 for the selected South Asian economies was only 21.6%, far lower than those in Malaysia (42.6%) and Thailand (38.1%).

Participation in GVCs is critical to driving economic development and industrialization. According to World Bank (2020), GVCs enhance economic growth by enabling production specialization, improving efficiency, and fostering productivity. Durable inter-firm linkages also facilitate technology diffusion throughout the value chain. The World Bank (2020) estimates that a 1% increase in GVC participation can raise per capita income by over 1%. Taguchi and Tsukada (2022) argue that deeper GVC integration may help mitigate premature deindustrialization risk in emerging and developing Asian economies. Therefore, strengthening GVC involvement is essential for South Asia’s industrial and economic transformation.

The concept of the GVC was initially introduced by Hummels et al. (2001) as “vertical specialization.” Koopman et al. (2011, 2014) extended this concept by accounting for the value-added contents in gross exports across multiple countries and sectors, effectively integrating vertical specialization into the value-added trade framework in the literature. This reconceptualization enabled the development of detailed trade databases such as those provided by the Organization for Economic Co-operation and Development (OECD) and World Trade Organization, allowing researchers to trace the value-added contributions of gross exports and perform empirical GVC analyses.

The “fragmentation” model introduced by Jones and Kierzkowski (2005) further explains GVC mechanisms in the context of intra-industry trade. Firms decide to fragment production across borders based on differences in location advantages (e.g., factor prices such as wages) and service link costs. They defined service link costs as bundles of activities connecting fragmented production blocks consisting of coordination, administration, transportation, and financial services. Among these, customs duties are a notable service link cost as they directly impact the cross-border trade of manufacturing parts and components. Free trade agreements (FTAs) can reduce these tariffs, thereby mitigating service link costs and fostering regional GVC activities.

The South Asian Association for Regional Cooperation (SAARC) has been instrumental in regional integration efforts, leading to the establishment of the South Asian Free Trade Agreement (SAFTA) in 20064. Several bilateral FTAs has also been established among SAARC members, such as between India and Sri Lanka (2000), Pakistan and Sri Lanka (2005), India and Bhutan (2006), and India and Nepal (2009). Beyond South Asia, FTAs include those between India and the Association of Southeast Asian Nations (ASEAN) (2010), India and Japan (2011), India and Korea (2010), India and Singapore (2005), Pakistan and China (2007), and Pakistan and Malaysia (2008).5

This study investigates the impact of FTAs, particularly SAFTA, on the development of GVCs in South Asia. Using the OECD’s6 2023 Trade in Value-added (TiVA) database, we construct GVC participation variables and apply a structural gravity trade model to estimate the agreement’s effects. Due to data availability constraints, the analysis focuses on Bangladesh, India, and Pakistan to examine the effects of FTAs on SAFTA.

This research offers new empirical evidence on the link between FTAs and GVCs in South Asia, a subject that remains underexplored. GVCs, which are characterized by vertical trade, can be expressed by ordinary gross trade values and value-added trade values. Kimura et al. (2007) employed the gross trade values of parts and components in a gravity trade model to examine the vertical trade of fragmented manufacturing products in an intra-industry setting. However, gross trade values do not necessarily represent vertical trade accurately because traded parts and components could be used to fulfill domestic final demands and not exclusively for processing for export. In contrast, value-added trade data precisely denote vertical trade in GVC linkages and are suitable for examining the relationship between FTAs and GVCs because the rule of origin in FTAs often applies the value-added rule. Therefore, this study uses the OECD TiVA 2023 database to provide supporting evidence regarding the relationship between FTAs and GVCs in South Asia.

Additionally, this study uses the structural gravity trade model, which is a refinement of the traditional model. The traditional gravity trade model explains bilateral trade flows in terms of economic size and geographical distance. However, Piermartini and Yotov (2016) contend that this approach yields biased and inconsistent results. They recommend a theoretically grounded econometric approach, namely a structural model.

The remainder of this paper is organized as follows. Section 2 reviews prior literature on the trade effects of South Asian FTAs and highlights this study’s contributions. Section 3 presents empirical analyses, including trends in GVC participation and gravity model estimations. Finally, Section 4 concludes the paper.

2. Literature Review and Contributions

Few studies have rigorously examined the effects of South Asian FTAs, partly due to the relatively recent evolution of both FTAs and GVC integration in the region, especially when compared with East and Southeast Asian. The existing literature presents mixed findings, often depending on the nature of FTA (regional or bilateral) and traded products involved. This study addresses a gap in the literature by offering updated and empirical evidence on the link between SAFTA and GVC participation in South Asia.

We start with theoretical foundation on the role of FTAs in trade flows. The trade effects of FTAs, as Viner (1950) initially argued, are decomposed into the trade creation effect and trade diversion effect. The trade creation takes place when joining an FTA leads to the replacement of high-cost domestic production by imports from within the FTA members. In this situation, trade is increased and/or created within member countries. The trade diversion, conversely, occurs when joining an FTA leads to the replacement of cheap imports from outside the FTA members by more expensive imports from inside. Under this scenario, trade is reduced and/or even eliminated with non-members. As a seminal empirical studies, Urata and Okabe (2014) comprehensively examined the impacts of FTAs on global trade flows focusing on their trade creation and diversion effects and identified trade creation effect in FTA implementation for many products. As a recent study, Singh (2021) demonstrated that the India-ASEAN FTA leads to a trade creation in total bilateral trade in terms of exports and imports.

This study focuses on the FTA effect on GVC backward participation (as detailed in Section 3.1). The specific research question is how the foreign value-added inputs in exports are affected by FTA implementation. According to the fragmentation model proposed by Jones and Kierzkowski (2005), the incorporation of foreign value-added inputs in exports rises with the implementation of FTAs, as these agreements diminish service link costs via tariff reductions and standard harmonization. These effects are categorized within the aforementioned trade creation effect. A limited number of empirical studies have examined the FTA impact on GVC involvement. Zhang et al. (2021) demonstrated that the depth of FTAs has significant positive effects on exports of foreign value-added globally. This study aligns with ours by concentrating on foreign value-added trade, although it focuses on the export side rather than the input side of value-added trade.

We next review the literature on the effects of FTAs focusing on South Asia. Islam et al. (2014) found no statistically significant evidence supporting SAFTA’s overall trade creation effect on member economies. Suhail and Sreejesh (2011), and Taguchi and Rubasinghe (2019) identified a trade creation effect in the FTA between India and Sri Lanka but not in SAFTA. Regarding South Asian FTA impacts on specific industries, Shaikh et al. (2015) observed a positive effect of SAFTA on Pakistan’s textile and rice exports to India. Mahmood and Jongwanich (2018) concluded that Pakistan-related FTAs had a greater impact on agricultural products than on manufacturing products. The FTA effects on GVC involvement in South Asia have rarely been studied, and Wijesinghe and Yogarajah (2022) explicitly examined the impact of SAFTA on GVC participation but found it to be negligible in most sectors.

The literature reviewed indicates that the impacts of FTAs on trade creation have been a primary focus of scholarly studies and extensively examined for decades (e.g., Urata & Okabe, 2014; Singh, 2021); however, their effects on GVC activities have been infrequently examined from both theoretical and empirical viewpoints (Zhang et al., 2021). Particularly, South Asia has not garnered much scholarly attention, owing to the relatively recent development of both FTAs and GVC integration in the region, and the direct linkage between FTAs and GVCs has been rarely discussed from the viewpoint of the reduction in service link costs (Wijesinghe & Yogarajah, 2022). This study fills a gap in the literature by offering updated and empirical evidence on the direct link between SAFTA and GVC participation in South Asia, positing explicitly that SAFTA diminishes service link costs via tariff reductions and standard harmonization, hence improving GVC integration. This study also utilizes a novel approach and database, leveraging the structural gravity trade model and data from the OECD’s TiVA 2023 database.

It should be noted that the contemporary GVCs have encountered multiple exogenous shocks. Alvarez et al. (2021) contended the resilience and adaptability of GVCs to geopolitical, environmental, and pandemic risks. The GVC activities in South Asia should also be monitored in light of the recent escalating risks.

3. Empirical Analyses

This section presents our empirical investigation, which includes an analysis of GVC trends and an econometric estimation of SAFTA’s effect on GVC participation in Bangladesh, India, and Pakistan. Concentrating on these three economies, owing to the data availability limitations of the OECD’s TiVA 2023 database, would restrict the representativeness in assessing SAFTA’s impact. However, from the perspective of trade volume, three economies constituted almost 90% of South Asian trade (exports plus imports) values in 2020. India accounts for 76.1%, Bangladesh for 10.2%, and Pakistan for 8.1%, while other economies represent less than 10%: Sri Lanka at 3.1%, Nepal at 1.3%, Afghanistan at 0.9%, Maldives at 0.3%, and Bhutan at 0.2%.7 Thus, three sample economies as predominant traders can exemplify region-wide effects of SAFTA.

3.1. GVC Trend in South Asian Economies

This subsection examines the trends in GVCs and the origin of foreign value-added in gross exports, particularly in relation to SAFTA’s implementation in 2006. The analysis focuses on Bangladesh, India, and Pakistan, using data from the OECD’s TiVA 2023 database. According to Koopman et al. (2011), there are two types of GVC participation in a vertical specialization chain, which are defined as follows:

GVC Participation = FVA/E + IVA/E,

where FVA, IVA, and E represent “foreign value-added embodied in gross exports,” “domestic value-added embodied as intermediate inputs in other countries’ gross exports,” and “gross exports,” respectively. The first item (FVA/E) reflects backward GVC participation (downstream engagement). The second item (IVA/E) reflects forward participation (upstream engagement), consistent with definitions from the World Bank (2020).

This study concentrates on backward GVC participation, given that industrialization and GVC linkages in South Asian economies are still developing, as shown in Table 1. Their manufacturing exports depend on foreign inputs and have limited capacity to supply industrial inputs (manufacturing materials, parts, and components) for third countries’ exports. This trend indicates that manufacturing in South Asian economies contributes to downstream engagement, rather than upstream engagement in GVCs. Backward GVC participation is significant for the industrialization of developing economies, including South Asian countries, because it involves intermediate inputs containing foreign technology. This type of participation boosts the competitiveness of their exports by combining foreign technology with local labor, capital, and technology (World Bank, 2016).

The OECD’s TiVA 2023 database enables the disaggregation of the value of gross exports into home and foreign countries’ value-added origins. The trends in backward GVC participation (foreign value-added as a percentage of gross exports) in Bangladesh, India, and Pakistan from 1995 to 2020 are analyzed in terms of total industry and manufacturing participation. Figure 1 reveals that the GVC participation ratios of the three countries range from 5% to 25%, which are lower than those in Malaysia and Thailand (see Table 1). No clear trends are apparent, except for India, which exhibits a moderate increase from 10.7% in 1995 to 17.3% in 2020 in total industry and from 7.6% in 1995 to 13.6% in 2020 in manufacturing.

Regarding the impact of SAFTA, GVC activities in South Asia are more important than their general trends. Figure 2 analyzes the interaction between FVA origins and gross exports among Bangladesh, India, and Pakistan (FVA origins are shown as a percentage of total FVA in exporting countries). FVA shares from Indian origin in the exports of Bangladesh and Pakistan are relatively large and reveal an acceleration following SAFTA enforcement in 2006. In contrast, those from Bangladesh and Pakistan as FVA origins in India’s exports and Bangladesh’s FVA origin in Pakistan’s exports remain negligible (less than one percent), exhibiting no significant post-SAFTA increase. This pattern suggests strong GVC linkages between the exports of Bangladesh and Pakistan, and FVA inputs from India, as well as the incremental impacts of SAFTA on these GVCs. However, the impact of SAFTA on GVCs should be verified using a sophisticated econometric model, as GVC dynamics are influenced by multiple factors beyond FTAs.

3.2. Estimation of the Structural Gravity Trade Model

This subsection outlines the estimation model used to assess SAFTA’s impact on GVCs in South Asia, followed by a discussion of the sample data.

3.2.1. Specification of the Estimation Model

This study employs the structural gravity model estimation approach proposed by Piermartini and Yotov (2016), using panel data with bilateral time-invariant fixed effects and multilateral time-varying price resistance terms. The estimation uses both ordinary least squares (OLS) in Equation (1) and Poisson pseudo maximum likelihood (PPML) in Equation (2).

ln FVA_ij,t = α₁ SAFTA + α₂ FTAs + π_i,t + χ_j,t + μ_ij + ε_ij,t

(1)

FVA_ij,t = exp [β₁ SAFTA + β₂ FTAs + π_i,t + χ_j,t + μ_ij] + ε_ij,t

(2)

Here, the subscripts i, j, and t denote host exporting economies (receiving foreign value-added in exports), origin economies (offering foreign value-added in exports), and trading years, respectively. FVA represents the foreign value-added in exports for the total industry (FVA_T) and manufacturing (FVA_M). SAFTA and FTAs are dummy variables denoting the presence of SAFTA and other FTAs between the sample countries in South Asia and their partners8. They take on a value of one if an agreement was in effect at time t and a value of zero otherwise. π_i and χ_j are the directional time-varying fixed effects of host economies i and origin economies j, respectively, that control for unobservable multilateral resistances. μ_ij represents the paired time-invariant fixed effects between economies i and j for addressing FTA endogeneity and time-invariant bilateral trade costs. ε_ij is an error term. α₁, α₂, β₁, and β₂ are the estimated coefficients of Equations (1) and (2). Additionally, “ln” demotes a logarithmic form and “exp” denotes an exponential form.

This study focuses on the entire industry and manufacturing as export sectors in exporting economies, without conducting comprehensive sectoral disaggregation, such as in food, textiles, and machinery. The rationale for using aggregate level data is the lack of diversification in the export sectors of sample economies; for example, textile exports constitute roughly 80% and 60% of total manufactured exports in 2020 in Bangladesh and Pakistan, respectively, as per the OECD’s TiVA 2023 database. Consequently, sectoral disaggregation may not yield sufficiently robust results to illustrate the sector-specific idiosyncrasies in GVC mechanisms. Revealing the sector-specific eccentricities necessitates individual case studies, as will be discussed in Section 5 on the limits of this study.

Equations (1) and (2) in the structural gravity model align with Piermartini and Yotov’s (2016) six recommendations: (i) use panel data that allow for the treatment and estimation of the effects of time-invariant bilateral trade costs; (ii) use interval data to allow for adjustment in trade flows; (iii) include intra-national trade flows that ensure consistency with gravity theory and capture the effects of globalization on international trade; (iv) use directional time-varying fixed effects to control for unobservable multilateral resistances; (v) employ paired fixed effects to address FTA endogeneity and time-invariant bilateral trade costs9; and (vi) estimate gravity using the PPML to address any trade data issues such as heteroscedasticity and zero trade flows. The core variables in traditional gravity trade model are addressed in the structural model as follows: GDP and population constitute components of directional time-varying fixed effects in (iv); distance serves as a component of paired fixed effects in (v); and the other common control variables are classified as either directional time-varying fixed effects in (iv) or paired fixed effects in (v).

A significant challenge is addressing the causal identification of the relationship between FTA implementation and GVC participation, specifically the endogeneity bias associated with FTA enforcement. Multiple methodologies have been employed, including the application of instrumental variables and a difference-in-difference approach. Baier and Bergstrand (2007) contended that conventional cross-section methodologies employing instrumental variables and control functions failed to yield stable estimates of FTA effects amid endogeneity, highlighting the importance of utilizing “bilateral fixed effects” in panel data settings for analyzing endogenous FTA treatment effects. The methodological illustration of Baier and Bergstrand (2007) is reflected in the recommendation (v) presented by Piermartini and Yotov (2016).

Equations (1) and (2) fulfill recommendations (i), (iv), (v), and (vi). For recommendation (ii), this study uses data pooled over consecutive years, rather than interval data, due to the short sample period (1995 to 2020). Instead, tests lagged by one year are conducted to allow for adjustments in value-added trade flows in response to FTA enforcement, thereby determining the phase-in effects of FTAs. Regarding recommendation (iii), this study includes domestic value-added (DVA = FVA_ij, i = j) in exports, as DVA corresponds to recommended intra-national trade flows. Concerning recommendation (vi), OLS is used in Equation (1) and PPML is used in Equation (2) as a robustness check, following Head and Mayer (2014).

The principal focus in the estimation pertains to the SAFTA effects (α₁ SAFTA and β₁ SAFTA). Our research hypothesis posits that the execution of SAFTA results in increased GVC backward participation within South Asia. This notion is corroborated by the argument regarding the decrease in service link costs proposed by fragmentation theory (Jones & Kierzkowski, 2005) and the empirical research conducted by Zhang et al. (2021); however, its research methodology differs for our more sophisticated structural gravity model. Therefore, we assume the significant and positive coefficients α₁ and β₁.

3.2.2. Sample Data

FVA data are sourced from the OECD’s TiVA 2023 database, covering Bangladesh, India, and Pakistan as host countries. The original FVA economies are selected from the 20 major host country partners (see

Table 2. FVA origins of Bangladesh, India, and Pakistan in 2020.

	Babgladesh		India		Pakistan
Foreign Origins	Value	Share	Value	Share	Value	Share
	mil. USD	% of Total FVA	mil. USD	% of Total FVA	mil. USD	% of Total FVA
Australia	58	0.7	1689	2.0	35	1.1
Canada	58	0.7	728	0.8	23	0.7
Japan	179	2.1	2360	2.8	62	2.0
Korea	150	1.8	1902	2.2	56	1.8
Mexico	56	0.7	789	0.9	9	0.3
United States	414	4.9	8517	9.9	206	6.5
Bangladesh	-	-	151	0.2	6	0.2
Brazil	60	0.7	804	0.9	38	1.2
China	2576	30.3	9978	11.6	832	26.3
India	673	7.9	-	-	51	1.6
Indonesia	134	1.6	2197	2.6	85	2.7
Malaysia	28	0.3	1079	1.3	32	1.0
Pakistan	142	1.7	75	0.1	-	-
Russia	71	0.8	1561	1.8	39	1.2
Saudi Arabia	53	0.6	7181	8.4	169	5.4
Singapore	65	0.8	1631	1.9	50	1.6
South Africa	28	0.3	841	1.0	62	2.0
Chinese Taipei	161	1.9	719	0.8	39	1.2
Thailand	279	3.3	952	1.1	43	1.3
European Union	1333	15.7	10,766	12.6	363	11.5
Total	6516	76.7	53,920	62.9	2201	69.7
World	8495		85,754		3158

Source: OECD’s TiVA 2023 database.

), comprising over 60% of the total FVA in each host country. The sample period is from 1995 to 2020, covering the full sample from the OECD’s TiVA 2023 database. Panel data are constructed for the 26-year period, combining host and origin economies (26 years × 3 host countries × 20 origin economies = 1560).10

For the subsequent panel estimation, this study uses panel unit root tests to investigate the stationary properties of the constructed panel data of ln FVA_T and ln FVA_M. The Levin, Lin, and Chu test (Levin et al., 2002) serves as the common root test, while the Fisher-ADF, Fisher-PP (Choi, 2001; Maddala & Wu, 1999), and Im, Pesaran, and Shin tests (Im et al., 2003) are used as individual unit root tests. The common unit root test assumes that there is a common unit root process across cross-sections. In contrast, the individual unit root tests allow for individual unit root processes that vary across cross-sections. These tests operate under the null hypothesis that the panel data level has a unit root by including an “intercept” in the test equations.

Table 3. Panel unit root tests.

	Common Unit Root	Individual Unit Root
	Levin, Lin, and Chu Test	Fisher-ADF Chi-Square	Fisher-PP Chi-Square	Im, Pesaran, and Shin W-Stat
ln (FVA_T)	−13.150 ***	193.591 ***	113.386	−5.149 ***
ln (FVA_M)	−13.407 ***	200.572 ***	127.399	−5.392 ***

Note: *** denotes statistical significance at the 99% level. Source: Authors’ estimations.

reveals that all tests, except the Fisher PP test, reject the null hypothesis of a unit root at the 99% significance level for ln FVA_T and ln FVA_M. These results validate the use of level panel data for estimation.

4. Results and Discussion

Table 4. Estimation results: phase-in effects of SAFTA: The phase-in effects of SAFTA were identified in the PPML estimation.

Total Inductry
Estimation	a_i	a_ii	a_iii	a_iv
Methodology	OLS		PPML
SAFTA	0.526 ***	0.496	0.134 ***	−0.106
	(4.366)	(1.556)	(5.124)	(−1.629)
SAFTA (−1)		0.130		0.147 *
		(0.302)		(1.854)
SAFTA (−2)		0.058		0.200 ***
		(0.135)		(2.938)
SAFTA (−3)		−0.177		−0.101 **
		(−0.557)		(−2214)
Other FTAs
India & ASEAN	0.189	0.189	0.228 ***	0.228 ***
India & Japan	0.028	0.028	0.084 **	0.083 **
India & Korea	0.931 ***	0.931 ***	1.176 ***	1.176 ***
India & Singapore	−0.125	−0.125	0.083	0.083
Pakistan & China	0.349	0.349	0.195 ***	0.194 ***
Pakistan & Malaysia	0.836 ***	0.835 ***	1.088 ***	1.087 ***
it fixed effects	Yes	Yes	Yes	Yes
jt fixed effects	Yes	Yes	Yes	Yes
i.j fixed effects	Yes	Yes	Yes	Yes
Adjusted R−squared	0.945	0.945	−	−
Manufacturing
Estimation	b_i	b_ii	b_iii	b_iv
methodology	OLS		PPML
SAFTA	0.617 ***	0.499	0.108 ***	−0.150 ***
	(4.096)	(1.254)	(80.417)	(−43.254)
SAFTA (−1)		0.170		0.149 ***
		(0.314)		(34.800)
SAFTA (−2)		0.174		0.193 ***
		(0.323)		(52.908)
SAFTA (−3)		−0.240		−0.076 ***
		(−0.604)		(−31.787)
Other FTAs
India & ASEAN	0.471 **	0.471 **	0.393 ***	0.393 ***
India & Japan	0.206	0.206	0.230 ***	0.229 ***
India & Korea	1.077 ***	1.077 ***	1.377 ***	1.376 ***
India & Singapore	0.012	0.012	0.442 ***	0.442 ***
Pakistan & China	0.654 **	0.654 **	0.417 ***	0.417 ***
Pakistan & Malaysia	0.854 ***	0.853 ***	0.564 ***	0.564 ***
it fixed effects	Yes	Yes	Yes	Yes
jt fixed effects	Yes	Yes	Yes	Yes
i.j fixed effects	Yes	Yes	Yes	Yes
Adjusted R−squared	0.920	0.920	−	−

Note: ***, **, and * denote statistical significance at the 99%, 95%, and 90% levels, respectively. T-statistics are shown in parentheses. Source: Authors’ estimations.

presents the estimation results of the structural gravity trade model, focusing on the phase-in effects of SAFTA on GVC participation in total industry and manufacturing.

Table 5. Estimation results: decomposition effects under SAFTA: SAFTA has facilitated the backward GVC participation of Bangladesh and Pakistan with India and that of Bangladesh with Pakistan (in manufacturing).

Estimation	c_i	c_ii
Industry	Total	Manufacturing
Methodology	PPML	PPML
exporter/FV origin in SAFTA
Bangladesh/India	0.283 ***	0.081 ***
Bangladesh/India (−1)	0.248 **	0.204 ***
Bangladesh/India (−2)	0.262 ***	0.327 ***
Aggregate of Significant effects	0.793	0.612
Bangladesh/Pakistan	0.074	0.076 ***
Bangladesh/Pakistan (−1)	0.122	0.102 ***
Bangladesh/Pakistan (−2)	0.074	−0.001
Aggregate of Significant effects	−	0.178
India/Bangladesh	−1.628 ***	−1.518 ***
India/Bangladesh (−1)	0.138	0.107 ***
India/Bangladesh (−2)	−0.088	−0.196 ***
Aggregate of Significant effects	−1.628	−1.607
India/Pakistan	−0.245	0.169 ***
India/Pakistan (1)	−0.238	−0.096 ***
India/Pakistan (2)	0.021	−0.387 ***
Aggregate of Significant effects	−	−0.314
Pakistan/Bangladesh	−2.028 ***	−3.908 ***
Pakistan/Bangladesh (−1)	0.378	0.551 ***
Pakistan/Bangladesh (−2)	−0.486	0.044***
Aggregate of Significant effects	−2.028	−3.401
Pakistan/India	1.457 ***	1.403 ***
Pakistan/India (−1)	0.263	0.327 ***
Pakistan/India (−2)	−0.197 *	−0.071 ***
Aggregate of Significant effects	1.260	1.659
Other FTAs
India & ASEAN	0.213 ***	0.380 ***
India & Japan	0.063	0.216 ***
India & Korea	1.142 ***	1.349 ***
India & Singapore	0.060	0.430 ***
Pakistan & China	0.195 ***	0.417 ***
Pakistan & Malaysia	1.081 ***	0.557 ***
it fixed effects	Yes	Yes
jt fixed effects	Yes	Yes
i,j fixed effects	Yes	Yes

Note: ***, **, and * denote statistical significance at the 99%, 95%, and 90% levels, respectively. T-statistics are shown in parentheses. Source: Authors’ estimations.

disaggregates SAFTA effects into bilateral impacts among Bangladesh, India, and Pakistan. Both tables report results using a log-link function.

Regarding the SAFTA phase-in effects on GVC participation in the OLS estimations for total industry (a_i-ii) and manufacturing (b_i-ii), only the no-lag effects (a_i and b_i) are statistically significant and positive. In contrast, PPML estimations for both total industry (a_iii-iv) and manufacturing (b_iii-iv) indicate positive and significant SAFTA effects with no lag (in a_iii and b_iii), and with one-year and two-year lags (in a_iv and b_iv). This discrepancy reflects some inconsistency between the OLS and PPML approaches. We appreciate the PPML methodology for the subsequent reasons. The PPML is a superior estimator compared to the OLS as it effectively handles the issues of heteroscedasticity and zero trade flows in trade data as Piermartini and Yotov’s (2016) argued. Secondly, the PPML provides more realistic estimation results with the FTA phase-in effects, as these effects have been corroborated by prior empirical research, including works by Baier and Bergstrand (2007) and Taguchi (2015). Consequently, we concentrate on the PPML estimation outcomes with the FTA phase-in effects in the following discussion. Regarding the other FTA effects in the estimations of a_iv and b_iv, all save the India–Singapore FTA in total industry demonstrate significantly positive effects on GVC participations.

Table 5 breaks down SAFTA’s effects from Table 4 (in a_iv and b_iv) into bilateral impacts, focusing on combinations of FVA inputs from Bangladesh, India, and Pakistan in each other’s exports in total industry (c_i) and manufacturing (c_ii). The estimations presented encompass the phase-in effects up to two-year lags, which were identified as having significantly positive effects in Table 4, and also exhibit the aggregate of the significant lagged effects.

Focusing on the aggregate of lagged effects in each combination, positive effects can be observed for the FVA inputs from India in the exports of Bangladesh and Pakistan, and in the FVA inputs from Pakistan in Bangladesh’s manufacturing exports in the PPML estimations. These findings indicate that SAFTA has facilitated the backward GVC participation of Bangladesh and Pakistan with India and that of Bangladesh with Pakistan (in manufacturing). The results partly align with our research hypothesis in that they substantiate the role of the FTA in GVC integration, supported by the fragmentation model, in selected bilateral combinations within SAFTA. However, the results contrast with previous research by Wijesinghe and Yogarajah (2022), as well as Islam et al. (2014), which indicated no significant effects of SAFTA on GVC and trade flows in South Asia.

The effects of the other bilateral combinations such as FVA inputs from Bangladesh and Pakistan in India’s exports, and from Bangladesh in Pakistan’s exports, are insignificant or even significantly negative, suggesting sluggish GVC activities in these combinations under SAFTA. These insignificant and negative effects of SAFTA on GVC in the other bilateral combinations contradict our research hypothesis, necessitating thorough investigations of tariff and trade structures.

The trade flows behind GVC activities are affected by the tariff rates set by the host countries under FTAs.

Table 6. Comparison of average tariff rates in SAFTA.

Total	Average Tariff Rate for Common Items
	Number of Items	Tariff Rate
Bangladesh, total, 2015
MFN	4250	13.3
SAFTA (India & Pakistan)		2.8
Bangladesh, manufacturing, 2015
MFN	3268	11.5
SAFTA (India & Pakistan)		2.4
India, total, 2008
MFN	3269	16.6
SAFTA (Pakistan)		11.6
SAFTA for LDCs (Bangladesh)		3.4
India, manufacturing, 2008
MFN	2663	13.3
SAFTA (Pakistan)		10.1
SAFTA for LDCs (Bangladesh)		3.2
Pakistan, total, 2008
MFN	1651	18.2
SAFTA (India)		15.6
SAFTA for LDCs (Bangladesh)		10.5
Pakistan, manufacturing, 2008
MFN	1395	18.9
SAFTA (India)		16.1
SAFTA for LDCs (Bangladesh)		10.8

Note: SAFTA differentiates the preferential tariff rate for the least-developed countries, as defined by the United Nations. Bangladesh belongs to this category. Source: Authors’ estimations are based on World Integrated Trade Solutions. See https://wits.worldbank.org/ (accessed on 1 June 2025).

compares the average preferential tariff rates under SAFTA with the Most Favored Nation (MFN) rates as a benchmark for the common items at the eight-digit Harmonized Commodity Description and Coding System code level in total industry and manufacturing for Bangladesh, India, and Pakistan. Data on tariff rates are sourced from the World Integrated Trade Solutions of the World Bank. Table 6 reveals that the preferential tariff rates under SAFTA, including those for the least-developed countries, are lower than the MFN rates for any combination of host and partner countries among the three target countries.

Regarding GVCs, the preferential tariff rates under SAFTA appear to be effective at fostering GVC backward participation for Bangladesh and Pakistan with India and for Bangladesh with Pakistan, but not for India with Bangladesh, India with Pakistan, or Pakistan with Bangladesh. The possible reasons for this sluggish GVC participation even under SAFTA include the issue of comparative advantages in the FVA offered by the origin economy. As shown in Figure 2, the shares of Bangladesh and Pakistan as FVA origins in India’s exports and Bangladesh’s as an FVA origin in Pakistan’s exports are negligible (less than one percent), indicating that Bangladesh and Pakistan have fewer comparative advantages in providing FVA to India’s and Pakistan’s exports. The issue of comparative advantage cannot be addressed by sector-aggregation analysis as conducted in this study and may be revealed through sector-specific case studies.

Under SAFTA enforcement, India and Pakistan may experience a substitution effect in the origins of FVA in their exports from other SAFTA members that were not sampled in this study (e.g., Sri Lanka). Therefore, SAFTA providing preferential tariff rates may not, or may negatively, contribute to GVC participation.

Additionally, the trade flows behind GVC activities are impacted by both tariff rates and non-tariff measures (NTMs). The FTA effects are not restricted to tariff reduction and elimination but may also include other aspects such as non-tariff elimination, coordination of rules of origin, and FTA usability improvements (e.g., Okabe, 2015; Rahman & Strutt, 2023). Therefore, a comprehensive evaluation of SAFTA, including both tariff rates and NTMs, is necessary to understand the impact of SAFTA on GVCs. This study’s focus on tariff effects alone is a limitation in terms of capturing the full impact of SAFTA on GVCs.

Therefore, the main policy implication is that there is further potential to explore GVC activities in South Asia. For example, the current sluggish backward GVC participation of India with Bangladesh and Pakistan, and of Pakistan with Bangladesh could be stimulated by the industrial diversification of Bangladesh and Pakistan to build their comparative advantages and by the improvement in the NTMs among them. As discussed in Section 3.2.1, textile exports accounted for around 80% and 60% of total manufactured exports in Bangladesh and Pakistan, respectively, in 2020. Thus, industrial diversification is an urgent issue to be addressed for them. To achieve this objective, a rapid strategy is to welcome foreign direct investment in the emerging industrial sectors designated for development. The practical recommendation is to create special economic zones to entice foreign investors by offering them incentives, as has been extensively practiced by China and East Asian countries.

5. Conclusions

This study assessed the effects of SAFTA on backward GVC participation in Bangladesh, India, and Pakistan by applying a structural gravity trade model and utilizing the OECD’s TiVA 2023 dataset. It contributes to the existing literature by presenting new evidence on the connection between FTAs and GVCs in South Asia, a topic rarely covered in previous studies. Our findings indicate that SAFTA facilitates backward GVC participation by increasing FVA inputs from India in Bangladesh and Pakistan’s exports and the FVA input from Pakistan in Bangladesh’s exports.

The main policy implication is that there is further potential to explore GVC activities in South Asia. For example, the current sluggish backward GVC participation of India with Bangladesh and Pakistan, and of Pakistan with Bangladesh could be stimulated by the industrial diversification of Bangladesh and Pakistan to build their comparative advantages and by the improvement in the NTMs among them.

Limitations of this study are noted as follows. The first one is the lack of detailed, sector-specific analyses of GVCs in South Asian economies. Future research should conduct case studies across individual sectors and countries (including the other countries than this study’s sample countries) to examine the relationship between GVCs and FTA enforcement to validate the evidence found in this study and develop more concrete recommendations for facilitating South Asian GVCs. The second limitation is this study’s limited focus on NTMs within the SAFTA framework and their effects on GVC activities. A thorough investigation of NTMs is crucial for comprehensively evaluating the role of SAFTA in facilitating GVCs. Third, there is potential for methodological enhancements. For instance, to address the incompatibility in the FTA phase-in effects between OLS and PPML estimation, one may examine alternate dynamic models, such as a distributed-lag model and regressions incorporating nonlinear time effects, to achieve more robust results. For addressing the FTA endogeneity bias, alternative approaches such as the application of instrumental variables and a difference-in-difference estimation can be employed to enhance the robustness of the result of bilateral fixed effect approach.