## _wp15204

## Source details

**Canonical URL:** [_wp15204](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15204.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15204.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15204.pdf.json)

---

### Introduction and main findings
- Over the past 30 years, production has become increasingly fragmented through the growing prevalence of global value chains (GVC), with components crossing numerous international borders.
- Trade in intermediate inputs has grown faster than trade in final goods: during 1995–2013, Asia’s trade in intermediate goods grew by a factor of six, while trade in final goods grew almost four times. This compares with fourfold and threefold increases, respectively, in the rest of the world.
- The paper uses a unique OECD-WTO trade in value-added database on GVCs covering 57 countries.
- Key high-level findings:
  - Joining GVCs brings positive and significant gains in productivity (empirical evidence cited: Baldwin and Yan, 2014).
  - Capturing a bigger slice of the GVC pie is positively associated with productivity gains and higher per capita growth.

### Benefits and distributional issues of GVC participation
- Benefits beyond traditional trade in final goods:
  - Enables finer task specialization and exploitation of finer comparative-advantage niches.
  - Raises benefits from economies of scale and scope.
  - Empirical micro evidence: Canadian firms entering GVCs gained 5 percent productivity advantage in the first year and cumulative 9 percent after four years; exiting firms lost 1 percent in the first year and cumulative 8 percent over four years (Baldwin and Yan, 2014).
- Distributional patterns within GVCs:
  - Example: In the iPod supply chain, Apple captures between one-third and one-half of an iPod’s retail value; Japanese and Korean firms capture large shares from high-value components; firms and workers in China capture no more than 2 percent from assembling the product.

### Stylized facts for Asia
- GVC expansion particularly pronounced among emerging Asian economies, including ASEAN.
- Asian economies, particularly China, have captured an increasingly larger share of the value added generated in GVCs, even after adjusting for recent rapid growth in relative economic size.
- Some advanced Asian economies, notably Korea, have captured a bigger slice of the GVC pie in high-tech manufacturing.
- Adjusted for relative economic size, shares of value added in GVCs accruing to Japan and advanced economies outside Asia have declined.
- Asian economies, both advanced and emerging, have moved upstream (providing intermediate inputs to other countries) rather than downstream (processing inputs from more upstream countries).
- Within high-tech manufacturing, advanced Asian economies remain significantly more specialized in upstream production than emerging Asian economies.

### Definitions, measurement, and empirical strategy
- GVC participation index (Koopman et al., 2010) = (foreign value added in domestic exports (backward participation) + domestically produced intermediates used in third countries (forward participation)) / gross exports. Excludes exports of final goods with no foreign input content.
- Upstreamness measure: Distance to Final Demand (DFD) (Fally, 2012) — larger DFD = more upstream.
- Data and sample:
  - 57 countries, 1995–2012 (indices interpolated to annual frequency for panel regressions).
  - Tariffs from UNCTAD TRAINS via WITS.
  - Infrastructure index: first principal component of communication, electricity, road density, paved roads, power distribution.
  - Human capital: years of schooling and quality of education system; health expenditure as % of GDP.
  - Governance index: first principal component of six WGI pillars.
  - Trade impediments: distance (weighted), trade restrictiveness, investment restrictiveness (inverses of World Freedom Index measures).
- Estimation approach:
  - Panel regressions with country and time fixed effects; use of lagged explanatory variables (t-1) to address endogeneity.
  - Industry-, country-, and year-fixed-effects specifications used for analyses of captured domestic value added.

### Empirical results: tariffs and participation drivers
- Correlation:
  - Strong negative correlation between tariff rates on intermediate goods and GVC participation; emerging Asian economies with lower participation impose higher effective tariffs on intermediate imports.
- Impact of intermediate goods tariffs — illustrative magnitudes and context:
  - Moving from the 25th to 75th percentile of the cross-country distribution of tariffs (an increase in tariffs) lowers GVC-linked trade participation by about ¾ of a percentage point to ¼ percentage points, depending on backward/forward linkages.
  - Median backward and forward participation rates in low-tech manufacturing typically around 3–6 percent.
  - Negative impact on backward participation larger than on forward participation; effects strongest in low-tech manufacturing.
- Reported coefficient estimates (all variables in logarithm terms; country & time fixed effects included; significance: * p<0.10, ** p<0.05, *** p<0.01):
  - Panel A (Overall participation index):
    - Log (Tariffs) t-1 = -0.118*** (All sectors), -0.153*** (High-tech), -0.076*** (Low-tech).
  - Panel B (Backward participation):
    - Log (Tariffs) t-1 = -0.221*** (All sectors), -0.173*** (High-tech), -0.188*** (Low-tech).
  - Panel C (Forward participation):
    - Log (Tariffs) t-1 = -0.038** (All sectors), -0.105*** (High-tech), -0.053** (Low-tech).
  - Number of observations reported: 726, 638, 643 (panels reported).
- Drivers of GVC participation (Table 2 highlights; dependent variable: log(PI)):
  - High-tech manufacturing (column 1):
    - Real GDP per capita (lag 1): 0.153***.
    - Infrastructure (lag 1): 0.079***.
    - Quality of education system (lag 1): 0.053**.
    - Health expenditure (lag 1): 0.079**.
    - Governance (lag 1): 0.230***.
    - Distance weighted by economic size (lag 1): -0.325***.
    - Trade restrictiveness (lag 1): -0.115**.
    - Investment restrictiveness (lag 1): -0.364***.
    - Tariff on intermediate goods (lag 1): -0.118***.
    - No. of Obs.: 431; R-squared: 0.993.
  - Low-tech manufacturing (column 2):
    - Real GDP per capita (lag 1): –0.268***.
    - Infrastructure (lag 1): 0.128**.
    - Years of schooling (lag 1): 0.551**.
    - Lax labor regulations (lag 1): 0.264***.
    - Tariff on intermediate goods (lag 1): –0.074*.
    - No. of Obs.: 346; R-squared: 0.824.
  - Interpretation:
    - Better fundamentals (regulatory environment, human capital, basic infrastructure), and lower tariffs/trade barriers raise GVC participation.
    - Industry-specific drivers differ: education quality matters more for high-tech; basic education and labor regulation flexibility matter more for low-tech.

### Capturing a bigger slice of the GVC pie — methodology and core results
- Objective: estimate determinants of a country’s share of world domestic value added (DVA) in industry k at time t; dependent variable = log(DVA_{i,k,t}).
- Key explanatory variables:
  - Log(GDP share) — control for relative economic size.
  - Log(DFD) — upstreamness measure.
  - Log(ECI) — Economic Complexity Index at industry level.
  - Log(Tariff) — intermediate goods tariff at industry level.
- Data: country- and industry-level for 57 countries; years 1995, 2000, 2005, 2008, 2009; DVA from OECD-WTO TiVA 2013 release; GDP from IMF WEO; tariffs from UNCTAD TRAINS via WITS.
- Main regression results (Table 3; industry, country, year fixed effects; cluster S.E. by country and industry):
  - High-tech manufacturing (column 1):
    - Log (GDP): 0.874*** (S.E. 0.12).
    - Log (DFD): 1.065** (S.E. 0.42).
    - Log (ECI): 0.531 (S.E. 0.34).
    - Log (Tariff): –0.359*** (S.E. 0.09).
    - Observations: 723; R-squared: 0.882.
  - Low-tech manufacturing (column 2):
    - Log (GDP): 0.678*** (S.E. 0.10).
    - Log (DFD): 0.860** (S.E. 0.43).
    - Log (ECI): 0.770*** (S.E. 0.20).
    - Log (Tariff): –0.211** (S.E. 0.10).
    - Observations: 939; R-squared: 0.77.
- Interpretations:
  - Upstreamness (larger DFD) increases a country’s captured share of value added; effect larger in high-tech manufacturing.
  - Higher economic complexity (ECI) raises captured value added; magnitude larger for low-tech manufacturing.
  - Higher tariffs on intermediate goods decrease the share of value added captured; negative impact larger in high-tech manufacturing.

### Conclusions and policy implications — key findings and recommended policies
- Summary findings:
  - Asian economies increased GVC participation, captured larger slices of GVC value added, and relocated toward upstream production.
  - Upstreamness (especially in high-tech), higher economic complexity, and lower tariffs on intermediate goods are associated with improved prospects for capturing higher GVC value-added shares.
- Policy lessons and recommendations:
  - Removing trade barriers:
    - Tariffs on intermediate imports reduce GVC participation and hamper the ability to capture a higher share of GVC value added because intermediates cross borders multiple times and trade barriers compound.
    - Removing tariffs and other trade barriers benefits all GVC participants.
    - IMF (2015) guidance: advanced economies should open services markets; emerging economies should avoid import-substitution policies and protectionism via nontariff barriers.
  - Facilitating trade and regional cooperation:
    - Implement trade-facilitating measures (simplify port and customs procedures) to reduce trade costs.
    - Regional trade agreements and cooperation help; ASEAN Economic Community commitments (beginning end-2015) are highlighted as welcome.
  - Enhancing human capital formation and technology development:
    - Upstream repositioning requires knowledge- and technology-enhancing measures: invest in human capital, encourage innovation and R&D.
    - Figure 8 (in source) indicates positive association between R&D expenditure changes and upstreamness in electronics.
  - Improving fundamentals:
    - Efficiency-enhancing structural reforms: better infrastructure, more efficient regulatory framework, stronger economic and legal institutions, and unwinding overly rigid labor market regulations.
  - Mitigating GVC-related risks:
    - GVC participation increases vulnerability to supply shocks propagating across networks (example: 2011 tsunami in Japan); build redundancies, inventories, strengthen resilience to macro shocks, and ensure adequate financial safety nets.
- Overarching message:
  - Policy efforts to reduce tariffs and trade impediments, raise economic complexity, invest in R&D and human capital, and improve fundamentals can both broaden GVC participation and increase the share of value added an economy captures within GVCs.

### Appendix: Economic complexity in Asia — facts, drivers, and BMA results
- ECI concept (Hidalgo and Hausmann, 2009):
  - Captures productive knowledge/capabilities via diversity (number of distinct products a country makes) and ubiquity (how many countries make the same product). Higher ECI → capability to produce diverse, less ubiquitous products.
  - Hidalgo and Hausmann (2009) find a one standard deviation increase in complexity is associated with a subsequent growth acceleration of 1.6 percent per year.
- Regional facts:
  - ECI has generally increased globally, but ECI for Asia is lower compared with economies at similar income levels outside Asia.
  - Key emerging Asian economies (China, India, Indonesia) have relatively low ECIs; advanced Asian economies (Japan, Korea) have lower ECIs than Germany, the United Kingdom, and the United States.
  - Drivers of ECI: better institutional quality, enhanced macroeconomic stability, and greater trade openness.
- Bayesian Model Averaging (BMA) selection and panel specification:
  - Variables selected (BMA): Geographical distance from the rest of the world; Size of government; Trade openness; Composite institutional quality; (Implicitly) GDP per capita (included in the panel regression as lagged GDP per capita).
  - Panel regression for 93 countries during 1980–2010 with country fixed effects:
    - ECI_{c,t} = α_{c,t} + β1 (GDP per capita)_{c,t-1} + β2 (Trade Openness)_{c,t-1} + Β3 (Distance)_{c,t-1} + β4 (Size of Government/GDP)_{c,t-1} + Β5 (Composite Institutional Quality)_{c,t-1} + α_c + ε_{c,t}
  - Two-step least-square approach used to address endogeneity of GDP per capita (GDP-per-capita variable estimated first; predicted values used in ECI regression).
- Main BMA regression results (Table A1 — Drivers of Economic Complexity; lag 1):
  - GDP per capita (lag 1): –0.027 (0.026).
  - Trade openness (lag 1): 0.341*** (0.078).
  - Distance weighted by GDP (lag 1): –0.901*** (0.118).
  - Size of government (lag 1): –0.095*** (0.026).
  - Composite institutional quality (lag 1): 0.170*** (0.025).
  - Additional regression statistics:
    - Observations: 136.
    - R-squared: 0.773.
    - Robust SE: Y.
    - Time dummy: Y.
  - Interpretation:
    - The ECI is positively correlated with greater trade openness and higher institutional quality.
    - The ECI is negatively correlated with geographic distance from the rest of the world and with the size of government.
    - Results robust to country fixed effects and time dummies.

*Source: IMF staff estimates and analysis in the provided content unit.*

### REFERENCES .............................................................................................................

### _wp15204 - REFERENCES .............................................................................................................

### Introduction and main findings
- Over the past 30 years, production has become increasingly fragmented through the growing prevalence of global value chains (GVC), with components crossing numerous international borders.
- Trade in intermediate inputs has grown faster than trade in final goods: during 1995–2013, Asia’s trade in intermediate goods grew by a factor of six, while trade in final goods grew almost four times. This compares with fourfold and threefold increases, respectively, in the rest of the world.
- The paper uses a unique OECD-WTO trade in value-added database on GVCs covering 57 countries.

### Benefits and distributional issues of GVC participation
- Integration into GVCs brings benefits beyond traditional trade in final goods:
  - Enables finer task specialization and exploitation of finer comparative-advantage niches.
  - Raises benefits from economies of scale and scope.
  - Empirical evidence (Baldwin and Yan, 2014) shows joining GVCs brings positive and significant gains in productivity.
- The GVC “pie” is not distributed equally:
  - Example: In the iPod supply chain, Apple captures between one-third and one-half of an iPod’s retail value; Japanese and Korean firms capture large shares from high-value components; firms and workers in China capture no more than 2 percent from assembling the product.
  - Capturing a bigger slice of the GVC pie is positively associated with productivity gains and higher per capita growth.

### Stylized facts for Asia
- The rise of GVCs has been ubiquitous, but expansion has been particularly pronounced among emerging Asian economies, including ASEAN.
- Asian economies, particularly China, have captured an increasingly larger share of the value added generated in GVCs, even after adjusting for recent rapid growth in relative economic size.
- Some advanced Asian economies, notably Korea, have captured a bigger slice of the GVC pie in high-tech manufacturing.
- Adjusted for relative economic size, shares of value added in GVCs accruing to Japan and advanced economies outside Asia have declined.
- Asian economies, both advanced and emerging, have moved upstream (providing intermediate inputs to other countries) rather than downstream (processing inputs from more upstream countries).
- Within high-tech manufacturing, advanced Asian economies remain significantly more specialized in upstream production than emerging Asian economies.

### Factors associated with higher captured shares and GVC positioning
- Moving toward a more upstream position in production and raising economic complexity are associated with a growing share of GVC value added captured by countries.
- Key factors and policy areas to foster GVC participation and increase captured value include:
  - Reducing trade barriers.
  - Strengthening infrastructure.
  - Enhancing human capital formation.
  - Supporting research and development (R&D).
  - Improving institutions.
  - Strengthening resilience to shocks.

### Research focus and structure
- The paper documents stylized facts about Asia’s GVC participation, positions within GVCs, and how much of the GVC pie Asian economies capture.
- It assesses which factors support GVC participation and help raise the captured share of value added.

*Source: _wp15204 - REFERENCES .............................................................................................................*

### section IV examines how an economy can reap a larger slice of the GVC pie; and section V

### _wp15204 - section IV examines how an economy can reap a larger slice of the GVC pie; and section V

### II. A primer on Asia’s GVC participation: definitions and patterns
- Definition: A global value chain (GVC) is a network of interlinked stages of production for the manufacture of goods and services that straddles international borders; combining imported intermediate goods and domestic goods and services into products exported for use as intermediates in subsequent stages.
- “Smiley-shaped” hypothesis: the relation between production stage and value added suggests most value added accrues at upstream (e.g., R&D) and downstream (e.g., marketing/branding) ends, with lower shares for midstream assembling.
- Measurement: GVC participation index (Koopman et al., 2010) = (foreign value added in domestic exports (backward participation) + domestically produced intermediates used in third countries (forward participation)) / gross exports. Excludes exports of final goods with no foreign input content.
- Empirical patterns (1995–2012):
  - GVC participation relatively high in Asia (including Korea, Malaysia, Philippines); growth faster in Asia, particularly ASEAN, than elsewhere. China’s participation grew significantly during 1995–2012 but is lower than the Asian average.
  - During 1995–2009, both advanced and emerging Asian economies gained domestic value-added shares in GVCs; gains larger in low-tech than high-tech manufacturing. Outside Asia, emerging economies gained while advanced economies lost shares.
  - Japan’s value-added share in high-tech manufacturing eroded; Korea gained. China moved up GVCs mainly in low-tech manufacturing.
- Upstreamness vs downstreamness:
  - Upstreamness = longer distance to final demand (Fally, 2012); downstreamness = shorter distance.
  - In high-tech manufacturing, advanced economies tend to specialize upstream; emerging economies downstream. Advanced Asia is more upstream than counterparts elsewhere; emerging Asia more downstream.
  - In low-tech manufacturing, both advanced and emerging Asian economies moved slightly upward but remained downstream relative to the rest of the world.

### III. How can economies increase their GVC participation? — literature and empirical strategy
- Benefits of GVC participation:
  - Knowledge spillovers, technology transfer, cost-savings, productivity gains in tradable sectors.
  - Micro evidence: Baldwin and Yan (2014) — Canadian firms gained 5 percent productivity advantage in the first year after entering a GVC vs non-GVC firms; cumulative 9 percent after four years. Firms exiting GVCs suffered a 1 percent loss in first year, cumulative 8 percent over four years.
- Drivers and impediments (WTO/OECD reviews; Hummels et al.):
  - Barriers: tariffs (especially on intermediates), inadequate infrastructure, limited access to trade finance, standards compliance, regulatory environment, business environment, transportation infrastructure, labor skills.
  - Fragmentation raises the “effective rate of protection”; tariffs particularly harmful when intermediates cross borders multiple times (Blanchard, 2013).
- Empirical strategy:
  - Objectives: (i) assess impact of tariffs on GVC participation (total, backward, forward) for high-tech and low-tech manufacturing; (ii) identify drivers of participation by industry.
  - Panel regressions with country and time fixed effects, annual indices from 1995–2012, use of lagged explanatory variables to address endogeneity.
  - Data: 57 countries, 1995–2012; GVC participation indices from OECD-TiVA (interpolated to annual frequency following Duval et al. (2014)); tariffs from UNCTAD TRAINS via WITS; infrastructure index (first principal component of communication, electricity, road density, paved roads, power distribution); human capital (years of schooling, quality of education system), health expenditure (% of GDP); governance index (first principal component of six WGI pillars); trade impediments: distance (weighted), trade restrictiveness, investment restrictiveness (inverses of World Freedom Index measures).

### III.C–D. Empirical results: tariffs and participation drivers
- Correlation: strong negative correlation between tariff rates on intermediate goods and GVC participation; emerging Asian economies with lower participation impose higher effective tariffs on intermediate imports.
- Impact of intermediate goods tariffs (Table 1 summary):
  - Tariffs on intermediate goods associated with significant negative effects on GVC participation (overall, backward, forward) in both high-tech and low-tech manufacturing.
  - Illustrative magnitude: if a country moves from the 25th to 75th percentile of the cross-country distribution of tariffs (an increase in tariffs), GVC-linked trade participation is lowered by about ¾ of a percentage point to ¼ percentage points, depending on backward/forward linkages.
  - Median backward and forward participation rates in low-tech manufacturing typically around 3–6 percent.
  - Negative impact on backward participation larger than on forward participation; effects strongest in low-tech manufacturing.
- Reported coefficient estimates (Table 1; specified panels):
  - Panel A (Overall participation index): Log (Tariffs) t-1 = -0.118*** (All sectors), -0.153*** (High-tech), -0.076*** (Low-tech).
  - Panel B (Backward participation): Log (Tariffs) t-1 = -0.221*** (All sectors), -0.173*** (High-tech), -0.188*** (Low-tech).
  - Panel C (Forward participation): Log (Tariffs) t-1 = -0.038** (All sectors), -0.105*** (High-tech), -0.053** (Low-tech).
  - Note: No. of Obs. reported as 726, 638, 643; significance: * p<0.10, ** p<0.05, *** p<0.01. All variables in logarithm terms. Country & time fixed effects included.

- Drivers of GVC participation (Table 2 highlights; dependent variable: log(PI)):
  - High-tech manufacturing (column 1):
    - Real GDP per capita (lag 1): 0.153***.
    - Infrastructure (lag 1): 0.079***.
    - Quality of education system (lag 1): 0.053**.
    - Health expenditure (lag 1): 0.079**.
    - Governance (lag 1): 0.230***.
    - Distance weighted by economic size (lag 1): -0.325***.
    - Trade restrictiveness (lag 1): -0.115**.
    - Investment restrictiveness (lag 1): -0.364***.
    - Tariff on intermediate goods (lag 1): -0.118***.
    - No. of Obs.: 431; R-squared: 0.993.
  - Low-tech manufacturing (column 2):
    - Real GDP per capita (lag 1): –0.268***.
    - Infrastructure (lag 1): 0.128**.
    - Years of schooling (lag 1): 0.551**.
    - Lax labor regulations (lag 1): 0.264***.
    - Tariff on intermediate goods (lag 1): –0.074*.
    - No. of Obs.: 346; R-squared: 0.824.
  - Interpretation: Better fundamentals (regulatory environment, human capital, basic infrastructure), and lower tariffs/trade barriers raise GVC participation; industry-specific drivers differ (e.g., education quality matters more for high-tech; basic education and labor regulation flexibility matter for low-tech).

### IV. How can an economy capture a bigger slice of the GVC pie? — methodology and results
- Objective: assess factors underlying an economy’s ability to acquire a greater share of value added along GVCs; differentiate high-tech vs low-tech industries (Eurostat definitions).
- Estimation: panel regression with industry, country, and year fixed effects. Dependent variable: log share of domestic value added (DVA) of country i over world in industry k at time t.
- Key explanatory variables:
  - Log(GDP share) — control for relative economic size.
  - Log(Distance to Final Demand, DFD) — upstreamness measure.
  - Log(Economic Complexity Index, ECI) — productive capabilities measure (constructed at industry level).
  - Log(Tariff) — intermediate goods tariff at industry level.
- Data: country- and industry-level for 57 countries, years 1995, 2000, 2005, 2008, 2009; DVA from OECD-WTO TiVA 2013 release; GDP from IMF WEO; DFD from TiVA; tariffs from UNCTAD TRAINS via WITS.
- Main results (Table 3; dependent variable: log(DVA)):
  - High-tech manufacturing (column 1):
    - Log (GDP): 0.874*** (S.E. 0.12).
    - Log (DFD): 1.065** (S.E. 0.42).
    - Log (ECI): 0.531 (S.E. 0.34).
    - Log (Tariff): –0.359*** (S.E. 0.09).
    - Observations: 723; R-squared: 0.882; industry, country, year FE; cluster S.E. by country and industry.
  - Low-tech manufacturing (column 2):
    - Log (GDP): 0.678*** (S.E. 0.10).
    - Log (DFD): 0.860** (S.E. 0.43).
    - Log (ECI): 0.770*** (S.E. 0.20).
    - Log (Tariff): –0.211** (S.E. 0.10).
    - Observations: 939; R-squared: 0.77; industry, country, year FE; cluster S.E. by country and industry.
- Interpretations:
  - Upstreamness (larger DFD) and higher ECI increase a country’s captured share of value added in GVCs.
  - Impact of upstreamness larger in high-tech manufacturing than in low-tech manufacturing (consistent with higher value added of upstream tasks like R&D in high-tech).
  - Economic complexity raises captured value added; magnitude larger for low-tech manufacturing than high-tech.
  - Higher tariffs on intermediate goods decrease the share of value added captured; negative impact larger in high-tech manufacturing than low-tech.

### V. Conclusions and policy implications — key findings and recommended policies
- Summary findings:
  - Asian economies increased GVC participation, captured larger slices of GVC value added, and relocated toward upstream production.
  - Upstreamness (especially in high-tech), higher economic complexity, and lower tariffs on intermediate goods are associated with improved prospects for capturing higher GVC value-added shares.
- Policy lessons and recommendations:
  - Removing trade barriers:
    - Tariffs on intermediate imports reduce GVC participation and hamper the ability to capture a higher share of GVC value added; because intermediates cross borders multiple times, trade barriers compound and act effectively as taxes on a country’s own exports.
    - Removing tariffs and other trade barriers benefits all GVC participants.
    - IMF (2015) guidance: advanced economies should open services markets; emerging economies should avoid import-substitution policies and protectionism via nontariff barriers.
  - Facilitating trade and regional cooperation:
    - Implement trade-facilitating measures (simplify port and customs procedures) to reduce trade costs.
    - Regional trade agreements and cooperation help; ASEAN Economic Community commitments (beginning end-2015) are highlighted as welcome.
  - Enhancing human capital formation and technology development:
    - Upstream repositioning requires knowledge- and technology-enhancing measures: invest in human capital, encourage innovation and R&D.
    - Figure 8 indicates positive association between R&D expenditure changes and upstreamness in electronics.
  - Improving fundamentals:
    - Efficiency-enhancing structural reforms: better infrastructure, more efficient regulatory framework, stronger economic and legal institutions, and unwinding overly rigid labor market regulations.
  - Mitigating GVC-related risks:
    - GVC participation increases vulnerability to supply shocks propagating across networks (example: 2011 tsunami in Japan); build redundancies, inventories, strengthen resilience to macro shocks, and ensure adequate financial safety nets.
- Overarching message: policy efforts to reduce tariffs and trade impediments, raise economic complexity, invest in R&D and human capital, and improve fundamentals can both broaden GVC participation and increase the share of value added an economy captures within GVCs.

### Appendix: Economic complexity in Asia — facts and drivers
- ECI concept (Hidalgo and Hausmann, 2009): captures productive knowledge/capabilities via diversity (number of distinct products a country makes) and ubiquity (how many countries make the same product). Higher ECI → capability to produce diverse, less ubiquitous products.
- Growth association: Hidalgo and Hausmann (2009) find a one standard deviation increase in complexity is associated with a subsequent growth acceleration of 1.6 percent per year.
- Regional facts:
  - ECI has generally increased globally, but ECI for Asia is lower compared with economies at similar income levels outside Asia.
  - Key emerging Asian economies (China, India, Indonesia) have relatively low ECIs; advanced Asian economies (Japan, Korea) have lower ECIs than Germany, the United Kingdom, and the United States.
  - Drivers of ECI: better institutional quality, enhanced macroeconomic stability, and greater trade openness.
- Empirical approach to drivers: Bayesian Model Averaging (BMA) used to select from a wide range of socioeconomic variables to identify key drivers of ECI (details of selected variables and probabilities of inclusion are reported in the Appendix of the source).

*Source: IMF staff estimates and analysis in the provided content unit.*

### 0.5 eliminated from the selection. Based on Bayesian Model Averaging, five variables are

### _wp15204 - 0.5 eliminated from the selection. Based on Bayesian Model Averaging, five variables are

### Variables selected (BMA)
- Geographical distance from the rest of the world
- Size of government
- Trade openness
- Composite institutional quality
- (Implicitly) GDP per capita (included in the panel regression as lagged GDP per capita)

### Econometric specification
- Panel regression for 93 countries during 1980–2010 with country fixed effects:
  - ECI_{c,t} = α_{c,t} + β1 (GDP per capita)_{c,t-1} + β2 (Trade Openness)_{c,t-1} + Β3 (Distance)_{c,t-1} + β4 (Size of Government/GDP)_{c,t-1} + Β5 (Composite Institutional Quality)_{c,t-1} + α_c + ε_{c,t}
- Note on endogeneity:
  - A two-step, least-square approach is used where the GDP-per-capita variable is estimated in the first step and the corresponding predicted values are used for the ECI regression.

### Main results (signs and interpretation)
- The ECI is:
  - Positively correlated with greater trade openness.
  - Positively correlated with higher institutional quality.
  - Negatively correlated with geographic distance from the rest of the world.
  - Negatively correlated with the size of government.
- The main results are reported as robust to country fixed effects and time dummies.

### Table A1 — Drivers of Economic Complexity (Panel BMA Best Specification)
- Dependent Variable: ECI
- Coefficients and standard errors (lag 1)
  - GDP per capita (lag 1): –0.027 (0.026)
  - Trade openness (lag 1): 0.341*** (0.078)
  - Distance weighted by GDP (lag 1): –0.901*** (0.118)
  - Size of government (lag 1): –0.095*** (0.026)
  - Composite institutional quality (lag 1): 0.170*** (0.025)
- Additional regression statistics
  - Observations: 136
  - R-squared: 0.773
  - Robust SE: Y
  - Time dummy: Y
- Note: Economic Complexity Index; BMA = Bayesian Model Averaging.
- Significance notation: *** p<0.01, ** p<0.05, * p<0.1

*Source: IMF staff estimates.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15204.pdf_