## wpiea2019116 - 2014. As pointed out by di Giovanni and Levchenko (2010), examining cross-border BC

## Source details

**Canonical URL:** [wpiea2019116 - 2014. As pointed out by di Giovanni and Levchenko (2010), examining cross-border BC](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019116.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019116.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019116.pdf.json)

---

### Main findings and contributions
- Industry-level analysis of cross-border business cycle (BC) comovement in Korea shows clear international comovement:
  - Industries highly export-linked to a country (either directly or indirectly) exhibit stronger comovement with that country’s GDP growth.
  - Industries highly import-linked to a country exhibit stronger negative comovement with that country’s employment growth.
  - Results are robust across country coverages (42 countries or 10 largest trading partners) but differ across industry sectors (manufacturing or services).
- Quantification of channels for aggregate BC comovement:
  - The increase in trade with China contributed the most to aggregate BC comovement.
  - The impact of trade linkages on BC comovement was propagated domestically via vertical linkages, especially through upstream linkages.
- Policy-relevant implication:
  - The Korean economy can be significantly affected by a few countries that are highly trade-linked and/or a few industries (often dominated by a few large firms) that are highly linked to other industries and account for a large share of international trade.

### Rationale for industry-level approach
- Advantages of industry-level analysis:
  - Inclusion of a rich set of fixed effects (country and industry fixed effects) helps control for unobservable country- and industry-specific characteristics, reducing omitted variable and simultaneity concerns.
  - Industry-level input-output linkages allow investigation of the role of vertical production linkages across industries in international BC comovement.
- Transmission channels from trading partners to domestic industries:
  - Direct channel via direct export/import linkages to a particular country.
  - Indirect channel via upstream/downstream linkages to other industries that are trading with a foreign country.
  - Both channels transmit economic shocks from a trading partner to a domestic industry, producing BC comovement.

### Related literature and positioning
- Network-based theoretical literature: microeconomic shocks at industry- or firm-level propagate to macro fluctuations (examples cited).
- Empirical evidence on input-output transmission highlights import penetration effects on value added and employment and the role of upstream/downstream exposure.
- Trade-comovement literature: trade associated with higher business cycle correlation but critiques exist; micro-level analyses show persistence of trade-comovement after controlling for common shocks.
- Unique contributions of this paper:
  - Applies international BC comovement analysis to the Korean economy.
  - Modifies di Giovanni and Levchenko (2010) and di Giovanni et al. (2018) models to explore impacts of the rise in trade linkages on BC comovement in Korea.
  - Separately identifies (i) export and import channels; and (ii) differentiated roles of trade and vertical linkages in BC comovement for GDP and employment.
  - Analyzes role of changes in trade linkages over time (pre/post GFC), not just levels.

### Data, coverage, and descriptive statistics
- Data source:
  - Industry-level input, output, value added, and international trade data from the WIOD (2016 release), containing WIOT and SEA.
- Coverage and sample:
  - WIOT covers 56 industries for each of the 43 countries over 15 years (2000-2014).
  - Focus on comovement between Korea and its trading partners, with special focus on the largest three trading partners: China, the U.S., and Japan.
  - Empirical work uses 38 industry classifications after combining/dropping some industries for reliability.
- Periods and notes:
  - Data cover 2000-2014 with two subperiods: pre-GFC (2000-07) and post-GFC (2009-14); 2008 excluded.
  - Intermediate inputs account for roughly 60 percent of international trade.
  - Top three trading partners accounted for 42.6 percent of Korea’s total trade (sum of exports and imports) in 2018: China (23.6 percent), the U.S. (11.6 percent), and Japan (7.5 percent).

### Trade and vertical linkage patterns (stylized facts)
- Sectoral patterns:
  - Four broad sectors: agriculture, manufacturing, non-manufacturing, services.
  - Overall export and import linkages to the global economy increased after the GFC in all four sectors.
  - Rise of export linkage particularly prominent in manufacturing.
  - Rise of import linkage relatively pronounced in both manufacturing and non-manufacturing.
  - Export linkage level particularly high in manufacturing; import linkage level substantially higher in manufacturing and non-manufacturing than in agriculture and services.
- Bilateral and industry-level patterns:
  - Export and import linkages to China increased significantly post-GFC; linkages to the U.S. and Japan remained stable through the 2000s.
  - Export linkages to China about three times higher than to the U.S. or Japan; import linkages to China highest among these countries.
  - Top five manufacturing industries by contribution to growth (share of GDP in 2014): electronics, chemicals, motor vehicles, machinery, basic metals.
  - Electronics most highly-linked to China in both exports and imports; motor vehicles highly export-linked to the U.S.; chemicals highly import-linked to Japan.
- Vertical (domestic) linkages definitions and patterns:
  - Upstream linkage: share of intermediate input supply to other industries (connectedness to buyers).
  - Downstream linkage: share of intermediate input purchase from other industries (connectedness to sellers).
  - Industry-level indirect trade-linkage indices defined as UP_iC^EX = sum_j IO_ij × EX_jC_j and DN_iC^IM = sum_j IO_ji × IM_jC_j with IO_ij and IO_ji as input supply/purchase shares.
  - Upstream and downstream linkages remained stable in manufacturing and services; mild increase in agriculture and non-manufacturing.
  - Upstream and downstream linkages relatively high in manufacturing; upstream linkages highest in agriculture.
  - Downstream linkages high in top manufacturing industries; upstream linkages relatively high in chemicals and basic metals; both linkages stable over 2000s in most industries.

### International BC comovement — descriptive evidence
- GDP growth correlations:
  - On average, GDP growth correlation with 42 countries significantly increased after the GFC, including China, the U.S., and Japan.
  - Stronger BC comovement with most sample countries after the GFC at the individual country level.
- Employment growth correlations:
  - On average, employment growth correlation increased somewhat after the GFC, with substantial variation across trading partners.
  - Employment correlation with some countries (including the U.S. and Japan) increased; with others (including China) declined.
  - Employment correlation with China turned from positive to negative.
  - Divergence between GDP and employment comovement suggests different driving forces for GDP growth comovement and employment growth comovement.

### Empirical specification and decomposition framework
- Aggregate value-added growth relation:
  - y_At = sum_i w_it × y_it, where y_it is growth rate of industry i value added and w_it is share of i’s value added.
- Aggregate correlation decomposition:
  - ρ(y_At, y_Ct) = sum_i w_it (σ_i / σ_A) ρ(y_it, y_Ct)
- First-differenced change between pre-GFC and post-GFC (assuming constant σ̄_A, σ̄_i, and w̄_i):
  - Δρ(y_At, y_Ct) = sum_i w̄_i (σ̄_i / σ̄_A) Δρ(y_it, y_Ct)_i
- Direct and indirect channels:
  - Direct: EX_iC (export intensity of i with country C) and IM_iC (import intensity).
  - Indirect: UP_iC^EX = sum_j IO_ij × EX_jC_j and DN_iC^IM = sum_j IO_ji × IM_jC_j.
- Industry-country level empirical specifications (pre and post and first-differenced) and extended specification including ΔUP_iC^EX and ΔDN_iC^IM.
- Decomposition to aggregate using estimated coefficients:
  - Direct-only decomposition and direct + indirect decomposition formulas using estimated β̂ coefficients.

### Estimation results — main findings
- Estimation setup:
  - Dependent variables: industry-level GDP growth correlation or employment growth correlation with foreign countries.
  - Specifications include industry fixed effects, country fixed effects, and country×sector fixed effects; some include indirect trade linkage variables.
- GDP growth comovement (full sample and top 10 trading partners):
  - Increase of export linkages (direct) has positive and significant coefficients for GDP comovement in direct-only specifications.
  - Import linkages coefficients for GDP are positive but not significant in direct-only specifications.
  - When indirect channels are included, indirect network effects of export linkages via upstream linkages are positive and significant.
  - For Korea’s top 10 trading partners, increase of direct export linkages contributed to rise of GDP growth comovement (significant at the 10 percent level); indirect channels not significant for this subsample.
- Employment growth comovement:
  - Direct-only specifications find negative and significant coefficients on import linkages for employment comovement (increased import linkages contributed to decline in employment comovement).
  - When indirect channels are introduced, the direct negative effect of import linkage on employment becomes insignificant—indirect vertical linkages can offset direct negative employment effects.
  - Indirect upstream effects from exports are positive for employment in the estimates.
- Sectoral heterogeneity:
  - Manufacturing:
    - Increase of export linkages (direct and indirect) contributed to rise of GDP growth comovement.
    - Increase of import linkages played an important role in employment growth desynchronization.
  - Services:
    - Roles of direct and indirect trade linkages in BC comovement not clearly identified; trade linkages relatively low in services.

### Decomposition of aggregate BC correlation — quantified contributions
- Decomposition methods:
  - Table 4: decomposition of changes in aggregate correlations from direct export and import linkages using Equation (14) and specification [1] in Table 1.
  - Table 5: decomposition including indirect linkages via upstream and downstream using Equation (15) and specification [3] in Table 1.
- Main decomposition findings for GDP growth correlations (Table 4):
  - On average, the increase of export linkages made significant contributions to the rise of GDP growth correlation after the GFC.
  - Rise of GDP growth correlation mainly driven by the increase of GDP growth correlation with China, mostly explained by increased export linkages with China.
  - GDP growth correlations with the U.S. and Japan declined due to decline of export and import linkages with these countries.
- Main decomposition findings for employment growth correlations (Table 4):
  - On average, changes of import linkages made important contributions to the increase of employment growth desynchronization.
  - Employment growth desynchronization mainly driven by rise of desynchronization with China, mostly explained by increased import linkages with China.
  - Employment synchronization with the U.S. and Japan slightly increased due to decline of import linkages with those countries.
- Direct and indirect linkages (Table 5):
  - Indirect export linkages with China via upstream linkages were the most important contributor to increase of GDP growth correlations.
  - Direct import linkages with China contributed the most to employment growth desynchronization.
  - Overall estimated contribution larger for GDP growth correlations but smaller (in absolute value) for employment growth correlations relative to direct-only decomposition.
  - Indirect linkages can act as offsetting factors for negative employment effects because increased import in one industry can create employment in vertically linked industries.
- Selected numeric decomposition entries (reported):
  - Table 4, Country rows (GDP / Employment decomposition components):
    - China: 0.628, 0.244, 0.219, 0.176, 0.043, -0.289, -0.063, -0.225
    - U.S.: -0.041, -0.083, -0.026, -0.012, -0.014, 0.074, 0.008, 0.066
    - Japan: -0.119, -0.110, -0.043, -0.022, -0.021, 0.123, 0.026, 0.096
    - Average: 0.156, 0.017, 0.050, 0.048, 0.003, -0.031, -0.010, -0.021
  - Table 5, selected aggregated reported values (printed format preserved):
    - China0.6280.24437.3410.470.4400.080-0.0130.3270.046-0.109-0.123-0.2140.303-0.074
    - U.S.-0.041-0.083-1.684-4.189-0.030-0.0050.004-0.010-0.0190.0920.0150.063-0.0140.029
    - Japan-0.119-0.110-4.869-5.813-0.065-0.0100.007-0.032-0.0290.1350.0510.092-0.0480.040
    - Average0.1560.01710.260.1540.1150.022-0.0010.095-0.0010.039-0.019-0.0200.080-0.002

### Robustness checks (Appendix B and top-five manufacturing)
- Robustness motivations:
  - Checks include (i) industries including manufacturing and non-manufacturing (Appendix B) and (ii) industries including only five major manufacturing industries (Table B.2).
  - Top-five industry robustness motivated by concentration: top five industries accounted for about 75 percent of total manufacturing value added and 24 percent of GDP in 2017.
- Empirical robustness patterns:
  - Table B.2 highlights important role of indirect trade channels (upstream and downstream) for top five manufacturing industries on GDP growth correlations, similar to manufacturing overall.
  - Confirms that a few large industries exhibit higher trade and vertical linkages and contribute significantly to international BC comovement.
- Regression summary statistics (selected robustness numbers):
  - Appendix B, Table B.1 / Table B.2 report Observations: 1,596 (in many specifications); Number of countries: 42.
  - R2 examples for GDP regressions in Table B.2: 0.596, 0.675, 0.597, 0.676.
  - Smaller-sample top 5 manufacturing regressions: Observations: 210; Number of countries: 42; R2 examples: 0.483, 0.511, 0.340, 0.354.
  - Significance notation: ***, **, * indicate significance at 1%, 5%, 10%, respectively.
- Selected coefficient magnitudes (reported examples from Table B.2):
  - Coefficients: -0.003, -0.527***, -0.023, -0.496** (with standard errors in tables).
  - Other reported coefficients: 0.190, -1.358**, -0.548**, -2.313***; 2.120***, 1.845**; 5.833***, 6.668** (standard errors printed in tables).

### Main conclusions and analytical implications
- Trade and vertical linkages evolution:
  - Korea’s participation in global trade increased dramatically during the 1990s and 2000s; trade openness almost doubled between 1990 and 2011.
  - Production structure became more vertical as intermediate input accounts for an increasing share in total output, mainly via rise of intermediate input trade.
  - Korea’s export and import linkages showed a large increase in all industry sectors, particularly in manufacturing, through the 2000s.
  - Korea’s GDP growth correlation with many foreign countries increased over the same period.
- Quantified impacts and channel-specific conclusions:
  - Industry-level international BC comovement in Korea:
    - (i) An industry highly export-linked to a country (directly or indirectly) exhibits stronger comovement with that country’s GDP growth.
    - (ii) An industry highly import-linked to a country exhibits stronger negative comovement with that country’s employment growth.
    - (iii) Effects differ across industry sectors (manufacturing vs services).
  - Relative importance of channels for aggregate BC comovement:
    - (i) Increase of trade with China contributed the most to aggregate BC comovement.
    - (ii) Impact of trade linkages on BC comovement propagated domestically via vertical linkages.
  - Implication: Korean economy vulnerability concentrated in a few highly trade-linked countries and a few large, highly interconnected industries.
- Contributions and suggested extensions:
  - Extends di Giovanni and Levchenko (2010) and di Giovanni et al. (2018) by analyzing role of changes in trade linkages (not only levels) and separately identifying export/import channels and vertical linkage roles for GDP and employment.
  - Suggested extensions: use value-added exports to assess GVC participation; consider higher-order interconnections across industries to capture cascade effects of shock propagation.

### Data and empirical specifications (selected definitions)
- Data sources and variable definitions (Table A.1):
  - GDP (YC): Real gross domestic product of country C — IMF World Economic Outlook
  - Real value added (Yi): Industry i's gross value added / i's price level in gross value added — World Input-Output Database (2016)
  - Employment (EC, Ei): Number of persons engaged (total industry / industry i) — World Input-Output Database (2016)
  - Export linkage (EXiC): Industry i's input export to country C / i's gross output — World Input-Output Database (2016)
  - Import linkage (IMiC): Industry i's input import from country C / i's gross output — World Input-Output Database (2016)
  - Upstream linkage (IOij) and Downstream linkage (IOji): Industry input supply/purchase ratios — World Input-Output Database (2016)
- Industry classification: ISIC Rev.4 categories across Agriculture (3), Manufacturing (14), Services (17), Non-manufacturing (4).

*Source: wpiea2019116 (IMF Working Paper content provided).*

### 2014. As pointed out by di Giovanni and Levchenko (2010), examining cross-border BC

### wpiea2019116 - 2014. As pointed out by di Giovanni and Levchenko (2010), examining cross-border BC

### Main findings and contributions
- Industry-level analysis of cross-border business cycle (BC) comovement in Korea shows clear international comovement:
  - Industries highly export-linked to a country (either directly or indirectly) exhibit stronger comovement with that country’s GDP growth.
  - Industries highly import-linked to a country exhibit stronger negative comovement with that country’s employment growth.
  - Results are robust across country coverages (42 countries or 10 largest trading partners) but differ across industry sectors (manufacturing or services).
- Quantification of channels for aggregate BC comovement:
  - The increase in trade with China contributed the most to aggregate BC comovement.
  - The impact of trade linkages on BC comovement was propagated domestically via vertical linkages, especially through upstream linkages.
- Policy-relevant implication:
  - The Korean economy can be significantly affected by a few countries that are highly trade-linked and/or a few industries (often dominated by a few large firms) that are highly linked to other industries and account for a large share of international trade.

### Rationale for industry-level approach
- Advantages of examining cross-border BC comovement at the industry level:
  - Inclusion of a rich set of fixed effects (country and industry fixed effects) helps control for unobservable country- and industry-specific characteristics, reducing omitted variable and simultaneity concerns.
  - Industry-level input-output linkages allow investigation of the role of vertical production linkages across industries in international BC comovement.
- Transmission channels from trading partners to domestic industries:
  - Direct channel via direct export/import linkages to a particular country.
  - Indirect channel via upstream/downstream linkages to other industries that are trading with a foreign country.
  - Both channels are crucial for transmitting economic shocks from a trading partner to a domestic industry, producing BC comovement.

### Related literature and positioning
- Network-based theoretical literature: macroeconomic fluctuations arise from microeconomic shocks at industry- or firm-level (examples: Carvalho 2010; Gabaix 2011; Acemoglu et al. 2012).
- Empirical evidence on input-output transmission:
  - Acemoglu et al. (2016a, 2016b): increased Chinese import penetration had negative impacts on value added and employment in US industry; larger negative impact for industries with high upstream exposure to Chinese imports.
  - Wang et al. (2018): employment effects from trading with China can be positive due to downstream channels in non-manufacturing sectors.
  - Ahn and Duval (2016): Chinese trade yields positive impacts on productivity and adverse impacts of Chinese imports on employment.
  - Lee (2019): domestic industry shocks transmitted mainly through downstream linkages; external export shocks propagated through direct export linkages in Korea.
- Trade-comovement literature:
  - Frankel and Rose (1998): countries that trade more exhibit higher business cycle correlation.
  - Imbs (2004): critiques that common shocks might drive the Frankel and Rose result.
  - “Trade-comovement puzzle” highlighted by Kose and Yi (2006); Johnson (2014).
  - Micro-level analyses (di Giovanni and Levchenko 2010; di Giovanni et al. 2018) show positive trade-comovement relationships persist after controlling for common shocks, driven by transmission of shocks at industry- or firm-level through trade and vertical linkages.
- This paper’s unique contributions:
  - Applies international BC comovement analysis to the Korean economy.
  - Modifies empirical models of di Giovanni and Levchenko (2010) and di Giovanni et al. (2018) to explore impacts of the rise in trade linkages on BC comovement in Korea.
  - Separately identifies (i) export and import channels; and (ii) differentiated roles of trade and vertical linkages in BC comovement for GDP and employment.
  - Analyzes the role of changes in trade linkages over time, not just levels across industries or firms.

### Data and descriptive statistics
- Data source:
  - Industry-level input, output, value added, and international trade data are taken from the WIOD (2016 release), which contains the World Input-Output Tables (WIOT) and the Socio-Economic Accounts (SEA).
- Coverage and sample:
  - WIOT covers 56 industries for each of the 43 countries over 15 years (2000-2014).
  - The paper focuses on comovement between Korea and its trading partners, with special focus on the largest three trading partners: China, the U.S., and Japan.
  - For empirical work, some industries are combined or dropped based on data reliability, resulting in 38 industry classifications.

### Structure of the paper (overview)
- Section II: Reviews related literature and discusses contributions.
- Section III: Presents data and stylized facts on trade and vertical linkages of the Korean economy and BC comovement with the global economy.
- Section IV: Estimates the role of trade and vertical linkages in determining BC comovement and quantifies the relative importance of various channels for aggregate BC comovement.
- Section V: Concludes and provides discussion of policy implications.

*Source: wpiea2019116 - 2014. As pointed out by di Giovanni and Levchenko (2010), examining cross-border BC (IMF Working Paper content provided).*

### Appendix A for detailed industry classifications used in this paper).

### Appendix A for detailed industry classifications used in this paper

### Data and periods
- Data cover the time period of 2000-2014 with classification of two subperiods by the GFC: pre-GFC (2000-07) and post-GFC (2009-14).
- The year 2008 is excluded in classification of two subperiods to mitigate large disturbances in data owing to the GFC.
- Intermediate inputs account for roughly 60 percent of international trade.
- Top three trading partners accounted for 42.6 percent of Korea’s total trade (sum of exports and imports) in 2018: China (23.6 percent), the U.S. (11.6 percent), and Japan (7.5 percent).

### Trade linkages — sectoral patterns
- Focus on four broad sector classifications: agriculture (agriculture, forestry, and fishing), manufacturing, non-manufacturing (mining/quarrying, electricity/gas, water supply, and construction), and services.
- Changes between pre-GFC (2000-07) and post-GFC (2009-14):
  - Overall export and import linkages to the global economy increased after the GFC in all four sectors.
  - The rise of export linkage is particularly prominent in manufacturing.
  - The rise of import linkage is relatively pronounced in both manufacturing and non-manufacturing.
  - Level differences across sectors:
    - Export linkage level is particularly high in manufacturing.
    - Import linkage level is substantially higher in manufacturing and non-manufacturing than in agriculture and services.

### Trade linkages — bilateral and industry-level patterns
- Bilateral linkages to Korea’s three major trading partners (China, the U.S., Japan):
  - Export and import linkages to China showed a significant increase in the post-GFC period, while those to the U.S. and Japan remained stable through the 2000s.
  - The substantial increase of trade linkages to China over the 2000s is associated with China’s growing presence in the global economy (noted event: China joined the WTO in 2001).
  - As a result of the large increase, the level of export linkages to China is about three times higher than that to the U.S. or Japan; import linkages to China are also the highest among these countries.
- Top five manufacturing industries by contribution to growth (share of GDP in 2014): electronics, chemicals, motor vehicles, machinery, basic metals.
  - Export and import linkages to China showed the greatest increase through the 2000s in most major industries.
  - Industry-level linkage levels:
    - Electronics are the most highly-linked to China in both exports and imports.
    - Motor vehicles are highly export-linked to the U.S.
    - Chemicals are highly import-linked to Japan.

### Vertical (domestic) linkages
- Definitions:
  - Upstream (downstream) linkage: share of intermediate input supply (purchase) to (from) other industries; upstream linkage denotes connectedness to buyers of an industry; downstream denotes connectedness to sellers.
  - Industry-level indirect trade-linkage indices:
    - UP_iC^EX = sum over j of IO_ij × EX_jC_j
    - DN_iC^IM = sum over j of IO_ji × IM_jC_j
    - IO_ij (upstream linkage) = share of i’s input supply to j in i’s gross output.
    - IO_ji (downstream linkage) = share of j’s input supply to i in i’s gross output.
- Empirical patterns over the 2000s:
  - Both upstream and downstream linkages remained stable in manufacturing and service sectors.
  - Upstream and downstream linkages showed a mild increase in agriculture and non-manufacturing.
  - Interpretation: significant increase of vertical linkages mostly occurred during the 1990s; over the 2000s, economic interconnectedness rose mostly through the rise of trade linkages.
  - At the level of vertical linkages:
    - Both upstream and downstream linkages are relatively high in manufacturing.
    - Upstream linkages are highest in agriculture.
- Industry-level vertical linkage patterns (top 5 manufacturing industries):
  - Downstream linkages are high in these industries relative to overall manufacturing.
  - Upstream linkages are relatively high in chemicals and basic metals.
  - Both upstream and downstream linkages were stable over the 2000s in most individual industries.

### International business cycle (BC) comovement — descriptive evidence
- GDP growth correlations:
  - On average, GDP growth correlation with 42 countries significantly increased after the GFC, including China, the U.S., and Japan.
  - At the individual country-level, stronger BC comovement with most sample countries after the GFC.
- Employment growth correlations:
  - On average, employment growth correlation increased somewhat after the GFC, but showed substantial variation across trading partners.
  - Employment correlation with some countries (including the U.S. and Japan) increased; with others (including China) declined.
  - Notably, employment correlation with China turned from positive to negative.
  - The divergence between GDP and employment comovement suggests different driving forces for GDP growth comovement and employment growth comovement.

### Empirical specification — structure and channels
- Aggregate value-added growth relation:
  - y_At = sum_i w_it × y_it
    - y_it: growth rate of value added of industry i
    - w_it: share of i’s value added in total value added
- Aggregate correlation decomposition:
  - ρ(y_At, y_Ct) = sum_i w_it (σ_i / σ_A) ρ(y_it, y_Ct)
- First-differenced change between pre-GFC and post-GFC (assuming constant σ̄_A, σ̄_i, and w̄_i):
  - Δρ(y_At, y_Ct) = sum_i w̄_i (σ̄_i / σ̄_A) Δρ(y_it, y_Ct)_i
- Direct and indirect channels from a foreign country C to domestic industry i:
  - Direct channel via EX_iC (export intensity of i with country C) and IM_iC (import intensity).
  - Indirect channel via upstream/downstream linkages to other industries that trade with country C:
    - UP_iC^EX = sum_j IO_ij × EX_jC_j
    - DN_iC^IM = sum_j IO_ji × IM_jC_j
- Empirical specifications (industry-country level), pre and post:
  - ρ_pre(y_it, y_Ct) = α_pre + β1 EX_iC_pre + β2 IM_iC_pre + μ_i_pre + δ_C_pre + ε_iC_pre
  - ρ_post(y_it, y_Ct) = α_post + β1 EX_iC_post + β2 IM_iC_post + μ_i_post + δ_C_post + ε_iC_post
  - First-differenced specification:
    - Δρ(y_it, y_Ct) = α + β1 ΔEX_iC + β2 ΔIM_iC + μ_i + δ_C + ε_iC
  - Extended to include indirect channels:
    - Δρ(y_it, y_Ct) = α + β1 ΔEX_iC + β2 ΔIM_iC + β3 ΔUP_iC^EX + β4 ΔDN_iC^IM + μ_i + δ_C + ε_iC
- Decomposition to aggregate level using estimated coefficients (hats denote estimates):
  - Direct-only decomposition:
    - Δρ̂(y_At, y_Ct) = sum_i w̄_i (σ̄_i / σ̄_A) (β̂1 ΔEX_iC + β̂2 ΔIM_iC)_i
  - Direct + indirect decomposition:
    - Δρ̂(y_At, y_Ct) = sum_i w̄_i (σ̄_i / σ̄_A) (β̂1 ΔEX_iC + β̂2 ΔIM_iC + β̂3 ΔUP_iC^EX + β̂4 ΔDN_iC^IM)_i

### Estimation results — main findings (summary of Tables 1–3)
- Estimation setup:
  - Dependent variables: industry-level GDP growth correlation or employment growth correlation with foreign countries.
  - Specifications vary by inclusion of industry fixed effects (all), country fixed effects, and country×sector fixed effects (sector = agriculture, manufacturing, non-manufacturing, services).
  - Some specifications include indirect trade linkage variables (ΔUP_iC^EX, ΔDN_iC^IM).
- GDP growth comovement (Tables 1 and 2; full sample and top 10 trading partners):
  - Increase of export linkages (direct) has positive and significant coefficients for GDP comovement in specifications considering only direct linkages.
  - Import linkages coefficients for GDP are positive but not significant in direct-only specifications.
  - When indirect channels are included, indirect network effects of export linkages via upstream linkages are positive and significant.
  - Interpretation: rise of export linkages (direct or indirect via upstream suppliers) contributed importantly to the increase in Korea’s GDP growth comovement with trading partners after the GFC.
  - For the sample of Korea’s top 10 trading partners, increase of direct export linkages contributed to rise of GDP growth comovement (significant at the 10 percent level); indirect channels were not significant for this subsample.
- Employment growth comovement:
  - Direct-only specifications find negative and significant coefficients on import linkages for employment comovement, implying increased import linkages contributed to decline in employment comovement.
  - This aligns with evidence (in other literature) that increased import penetration can reduce domestic employment opportunities.
  - However, when indirect channels (upstream/downstream) are introduced, the direct negative effect of import linkage on employment becomes insignificant.
    - Interpretation: direct negative employment effects from import penetration can be offset by indirect positive effects (for example, increased intermediate input supply to other industries).
  - Indirect upstream effects from exports are identified as positive for employment in the estimates.
- Sectoral heterogeneity (Table 3 — manufacturing vs services):
  - Manufacturing:
    - Increase of export linkages (direct and indirect) contributed to rise of GDP growth comovement.
    - Increase of import linkages played an important role in employment growth desynchronization (movement in opposite direction).
  - Services:
    - Roles of direct and indirect trade linkages in BC comovement are not clearly identified; trade linkages are relatively low in services compared to manufacturing.

### Key analytical implications
- Trade linkages growth after the GFC is an important driver of Korea’s increased GDP comovement with the global economy, with exports (direct and indirect via upstream linkages) playing a central role.
- Import penetration is associated with employment comovement decline in direct specifications; indirect vertical linkages can mitigate or offset these direct employment effects.
- Vertical connectedness (upstream linkages) amplifies the transmission of foreign demand shocks via export channels, increasing GDP comovement.
- Sectoral differences (manufacturing vs services) are important: effects are clearer and stronger in manufacturing where trade and vertical linkages are larger.

*Italic: Source — Appendix A for detailed industry classifications used in this paper (wpiea2019116), World Input-Output Database; Author's calculations.*

### Appendix B), and (ii) industries including only five major manufacturing industries (Table B.2 in

### wpiea2019116 - Appendix B), and (ii) industries including only five major manufacturing industries (Table B.2 in

### Robustness of industry-level results
- Motivation:
  - Robustness checks include (i) industries including both manufacturing and non-manufacturing industries (Appendix B), and (ii) industries including only five major manufacturing industries (Table B.2 in Appendix B).
  - The former robustness is meaningful because Korea’s exports in final products—which account for a significant portion of international trade—can also be affected by the global economy.14
  - The latter robustness is motivated by the highly concentrated and interconnected industrial structure of the Korean economy.15
- Key numeric facts:
  - Top five industries accounted for about 75 percent of total manufacturing value added and 24 percent of GDP in 2017.
- Empirical patterns:
  - Table B.2 highlights the important role of indirect trade channels (both upstream and downstream) for the top five manufacturing industries on the GDP growth correlations, similar to the results for the manufacturing sector overall.
  - The results confirm that a few large industries exhibit higher trade and vertical linkages and contribute significantly to international BC comovement, consistent with di Giovanni et al. (2017, 2018) and the literature on large firms in aggregate fluctuations (e.g., Gabaix 2011; di Giovanni et al. 2014; Carvalho and Grassi 2019).

### Decomposition of aggregate business cycle (BC) correlation
- Method:
  - Table 4: decomposition of contributions to changes in aggregate correlations into those from direct export and import linkages by countries, computed from Equation (14) using estimated coefficients from specification [1] in Table 1.
  - Table 5: decomposition using Equation (15) with estimated coefficients from specification [3] in Table 1; decomposes contributions from indirect export and import linkages via upstream and downstream linkages as well as direct linkages.
- Main decomposition findings for GDP growth correlations (Table 4):
  - (i) On average, the increase of export linkages made significant contributions to the rise of GDP growth correlation in the aftermath of the GFC.
  - (ii) The rise of GDP growth correlation was mainly driven by the increase of GDP growth correlation with China, which can be mostly explained by the increase of export linkages with China.
  - (iii) GDP growth correlations with the U.S. and Japan have declined due to the decline of export and import linkages with these countries.
- Main decomposition findings for employment growth correlations (Table 4):
  - (i) On average, the changes of import linkages made important contributions to the recent increase of employment growth desynchronization.
  - (ii) Employment growth desynchronization was mainly driven by the rise of employment growth desynchronization with China, which can be mostly explained by the increase of import linkages with China.16
  - (iii) Employment synchronization with the U.S. and Japan slightly increased due to the decline of import linkages with those countries.
- Direct and indirect linkages (Table 5) — key points:
  - (i) Indirect export linkages with China via upstream linkages were the most important contributor to the increase of GDP growth correlations.
  - (ii) Direct import linkages with China contributed the most to employment growth desynchronization.
  - Overall estimated contribution would be larger for GDP growth correlations but smaller (in absolute value) for employment growth correlations relative to the decomposition that considers only direct linkages.
  - Note: indirect export and import linkages can be additional channels of GDP correlations but can act as offsetting factors for negative employment effects because increased import in a certain industry can also create additional employment in its vertically linked industries.
- Selected numeric decomposition entries (as reported):
  - Table 4, Country rows (GDP / Employment decomposition components):
    - China: 0.628, 0.244, 0.219, 0.176, 0.043, -0.289, -0.063, -0.225
    - U.S.: -0.041, -0.083, -0.026, -0.012, -0.014, 0.074, 0.008, 0.066
    - Japan: -0.119, -0.110, -0.043, -0.022, -0.021, 0.123, 0.026, 0.096
    - Average: 0.156, 0.017, 0.050, 0.048, 0.003, -0.031, -0.010, -0.021
  - Table 5, selected aggregated (concatenated) reported values (as printed):
    - China0.6280.24437.3410.470.4400.080-0.0130.3270.046-0.109-0.123-0.2140.303-0.074
    - U.S.-0.041-0.083-1.684-4.189-0.030-0.0050.004-0.010-0.0190.0920.0150.063-0.0140.029
    - Japan-0.119-0.110-4.869-5.813-0.065-0.0100.007-0.032-0.0290.1350.0510.092-0.0480.040
    - Average0.1560.01710.260.1540.1150.022-0.0010.095-0.0010.039-0.019-0.0200.080-0.002

### Main conclusions
- Trade and vertical linkages:
  - Korea’s participation in global trade increased dramatically during the 1990s and 2000s; trade openness almost doubled between 1990 and 2011.
  - The production structure became more vertical as intermediate input accounts for an increasing share in total output, mainly by the rise of intermediate input trade.
  - Korea’s export and import linkages showed a large increase in all industry sectors, particularly in manufacturing, through the 2000s.
  - Korea’s GDP growth correlation with many foreign countries increased over the same period.
- Quantified impacts and channels:
  - International BC comovement is found at the industry-level in Korea:
    - (i) An industry that is highly export-linked to a country (either directly or indirectly) exhibits stronger comovement with that country’s GDP growth.
    - (ii) An industry that is highly import-linked to a country exhibits stronger negative comovement with that country’s employment growth.
    - (iii) There is differentiation across industry sectors (manufacturing or services).
  - Relative importance of channels for aggregate BC comovement:
    - (i) The increase of trade with China contributed the most to the aggregate BC comovement.
    - (ii) The impact of trade linkages on BC comovement was propagated domestically via vertical linkages.
  - Implication: the Korean economy can be significantly affected by a few countries that are highly trade-linked and/or a few industries (dominated by a few large firms) that are highly linked to other industries and that account for a large share of international trade.
- Contributions to literature and extensions:
  - The paper extends di Giovanni and Levchenko (2010) and di Giovanni et al. (2018) by analyzing the role of changes in trade linkages in determining international BC comovement (not only levels).
  - It separately identifies export and import channels and differentiated roles of trade and vertical linkages for GDP growth and employment growth.
  - Suggested extensions:
    - Use value-added exports as an alternative measure of trade linkages to assess participation in GVCs.
    - Consider higher-order interconnections across industries to capture “cascade effects” whereby productivity shocks propagate beyond immediate downstream customers.

### Data and empirical specifications (selected)
- Data sources and definitions (Table A.1):
  - GDP (YC): Real gross domestic product of country C — IMF World Economic Outlook
  - Real value added (Yi): Industry i's gross value added / i's price level in gross value added — World Input-Output Database (2016)
  - Employment (EC, Ei): Number of persons engaged (total industry / industry i) — World Input-Output Database (2016)
  - Export linkage (EXiC): Industry i's input export to country C / i's gross output — World Input-Output Database (2016)
  - Import linkage (IMiC): Industry i's input import from country C / i's gross output — World Input-Output Database (2016)
  - Upstream linkage (IOij) and Downstream linkage (IOji): Industry input supply/purchase ratios — World Input-Output Database (2016)
- Industry classification (Table A.2): lists ISIC Rev.4 categories and descriptions across Agriculture (3), Manufacturing (14), Services (17), Non-manufacturing (4).
- Regression summary statistics (selected from robustness tables):
  - Appendix B, Table B.1 / Table B.2 specifications report:
    - Observations: 1,596 (in many specifications reported)
    - Number of countries: 42
    - R2 examples: 0.596, 0.675, 0.597, 0.676 (for GDP regressions reported in Table B.2)
  - Smaller-sample top 5 manufacturing regressions:
    - Observations: 210
    - Number of countries: 42
    - R2 examples: 0.483, 0.511, 0.340, 0.354
- Significance notation:
  - ***, **, * indicate levels of significance at 1%, 5%, 10%, respectively.
- Selected coefficient magnitudes from robustness tables (as reported):
  - Example entries (Table B.2 GDP / Employment, top 5 manufacturing industries):
    - Coefficients: -0.003, -0.527***, -0.023, -0.496** (with reported standard errors in parentheses)
    - Other coefficients reported: 0.190, -1.358**, -0.548**, -2.313***; 2.120***, 1.845**; 5.833***, 6.668** (with standard errors printed in tables).

*Source: Appendix B and related sections of the IMF working paper content unit wpiea2019116 - Appendix B), and (ii) industries including only five major manufacturing industries (Table B.2 in (source PDF).

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019116.pdf_
