## ch4onlineannex - Annex Table 4.1.1). As a benchmark, the same

## Source details

**Canonical URL:** [ch4onlineannex - Annex Table 4.1.1). As a benchmark, the same](https://www.imf.org/-/media/files/publications/weo/2024/april/english/ch4onlineannex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/weo/2024/april/english/ch4onlineannex.pdf.md)
- [Structured JSON version](/-/media/files/publications/weo/2024/april/english/ch4onlineannex.pdf.json)

---

### Idiosyncratic growth surprises and correlations
- Definitions:
  - Growth surprises: estimated residual (γ) of equation (2) (using the April WEO projections).
  - Idiosyncratic growth surprises: residual of a regression (notation in source).
- Findings:
  - Correlation across idiosyncratic growth surprises in advanced economies (AEs) excluding the U.S. is positive and statistically significant throughout the entire period (Online Annex Figure 4.1.2).
  - The role of G20 EMs in the global economy has become more relevant in the last decade (interpreted from the positive correlation).
- Key visuals described:
  - Online Annex Figure 4.1.1: Idiosyncratic Growth Surprises (Percent) — countries: Argentina, Brazil, China, Indonesia, India, Mexico, Russia, Saudi Arabia, Türkiye, South Africa (2000–2022).
  - Online Annex Figure 4.1.2: Correlation across Idiosyncratic Growth Surprises (Percent) — plotted versus idiosyncratic growth surprises in AEs and Average idiosyncratic growth surprises in US for 1990–2013 and 2014–22.

### Regression evidence on AE growth surprises (selected estimates)
- Dependent variable: Growth surprises in AEs.
- Selected coefficient estimates (Online Annex Table 4.1.1):
  - Average growth surprise G-20 EM x I(1990-2013): −0.0459 (0.075); −0.0311 (0.080).
  - Average growth surprise G-20 EM x I(2014-2022): 0.5153** (0.204); 0.4619** (0.217).
  - Average growth surprise U.S. x I(1990-2013): 0.4336*** (0.063) (column 3).
  - Average growth surprise U.S. x I(2014-2022): 0.4444*** (0.133) (column 3).
- Sample and fit:
  - Observations: 297, 264, 264 (across columns).
  - R-squared: 0.023, 0.018, 0.187 (across columns).
  - Sample: AEs; AEs (ex. U.S.); AEs (ex. U.S.) (column-wise).
- Notes:
  - Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
  - Country fixed effects included (Country FE: Y Y Y).
  - Growth surprises defined as the estimated residual (γ) of equation (2).

### G20 EM trade integration (gravity model)
- Methodology:
  - Standard gravity model on IMF DOTS data, 1980–2022; dependent variable: log goods imports from source s to destination d in year t.
  - Key regressor: dummy = 1 if either source or destination is a G20 EM, interacted with year dummies.
  - Fixed effects: source × year, destination × year, country-pair.
- Interpretation and findings:
  - Coefficients β_i measure the difference in trade involving G20 EMs relative to trade among other countries.
  - Trade with G20 EMs accelerated after China’s WTO accession in 2001 and more so in the second half of the 2010s.
  - The increasing integration is not uniquely attributable to China; other G20 EMs also contributed.
- Key visual:
  - Online Annex Figure 4.1.3: Increase in G20 Emerging Markets Global Trade Excluding China (Log points) — coefficients plotted yearly (reference year 2001).

### G20 EM financial integration (US portfolio assets)
- Data and measurement:
  - Security-level US cross-border portfolio holdings from the TIC system; presentations on nationality basis and residence basis as share of total US portfolio assets abroad.
- Findings:
  - US portfolio assets vis-à-vis G20 EMs rose from 3 percent in the early 2000s to around 7 percent in the early 2020s, driven largely by US assets in China and, to a lesser extent, India.
  - On a residence basis, the increase in US portfolio assets in China is not visible, indicating recent US investment in China routed through third countries.
- Key visuals:
  - Online Annex Figure 4.1.4 panel 1: Nationality Basis (Percent of total assets) — series for Argentina, Brazil, China, Indonesia, India, Mexico, Russia, Saudi Arabia, Türkiye, South Africa (2003–2022).
  - Online Annex Figure 4.1.4 panel 2: Residence Basis (Percent of total assets).

### Are G20 EMs small and open? SOE model evidence
- Framework:
  - Updated Fernandez, Schmitt-Grohé and Uribe (2017) SOE model: domestic cyclical GDP in a G20 EM responds to global vector F_i (real prices of agriculture, metals, fuel cyclical components plus one financial variable: 3-month US T-Bill real gross return, US 10-year real gross return, US real effective exchange rate, or Moody’s investment-grade corporate spread).
  - Block exogeneity assumption: global variables affect domestic economy contemporaneously and with lags; domestic cannot affect global prices/financial conditions contemporaneously or with lag.
  - Estimation on quarterly data since 1980 (Russia since 1995); Argentina and Saudi Arabia excluded.
  - Sub-samples: before and after the GFC; GFC and COVID-19 periods excluded.
- Tests and results:
  - Likelihood ratio tests compare unrestricted estimates with those restricted by block-exogeneity.
  - First test (one global variable at a time): after the GFC most G20 EMs affect at least one global price (notably agriculture prices and short-term interest rates).
  - Stricter test (weak exogeneity to entire global vector): SOE assumption rejected for China across all specifications starting in the early 2000s.
  - Some EMs affect financial variables even after accounting for commodity effects: Brazil (post-GFC); South Africa, Mexico, Indonesia when investor risk-appetite included (investment grade corporate spread).
- Related evidence:
  - Online Annex Figure 4.1.5: Times EMs are Mentioned on FED’s FOMC Documents (Percent of document word count) — rising mentions over time.
  - Aguiar and Gopinath (2007)-style panel tests (GMM after Burnside (1999)) for G20 EMs excluding China, India, Mexico: consumption/output volatility ratios and net exports–output correlations consistent with SOE stylized facts (Online Annex Figure 4.1.6).
- Data treatment notes:
  - China: alternative indicators used (San Francisco Fed’s China Cyclical Activity Tracker; Li indicator) transformed to quarterly GDP in levels and seasonally adjusted due to measurement concerns post-GFC.
  - Other countries: quarterly GDP back to 1980 compiled from Aslam and others (forthcoming), WEO, IFS, OECD, national offices; temporal disaggregation used for earlier periods.

### Aggregate spillovers: SVAR analysis (identification and estimation)
- Objective: quantify spillovers from G20 EMs to growth in other EMs and AEs.
- SVAR setup:
  - Country-specific SVARs for triples: a G20 AE, a G20 EM, and another non-G20 EM.
  - Additional global variables: VIX and a commodity price index.
  - Two growth differentials included for identification.
  - Each SVAR contains ten variables (Online Annex Table 4.2.2); for each G20 EM, 37 models estimated.
  - Estimation sample: quarterly data Q1 2001–Q3 2023.
- Identification:
  - Combination of sign and zero restrictions separates domestic demand vs supply and domestic vs foreign shocks.
  - Zero restrictions assume shocks from smaller economies transmit to larger economies only with a one-quarter lag.
  - Example: in U.S.–Mexico–Chile model, shocks from Chile (average weight 0.4 percent of total sample GDP PPP over 2000-2023) affect U.S. real GDP (20.8% share) and Mexico (2.6% share) with a lag of one quarter; shocks from Mexico can affect Chile immediately but U.S. only after one quarter.
- Country coverage and sample weights (average weights in total sample 2000-2023, GDP PPP US$):
  - G20 AEs: United States 20.8%, Japan 5.9%, Germany 4.5%, France 3.2%, United Kingdom 3.2%, Italy 2.8%, Korea 2.1%, Canada 1.9%, Australia 1.3%.
  - G20 EMs: China 18.8%, India 7.8%, Brazil 3.4%, Russia 4.1%, Indonesia 2.9%, Mexico 2.6%, Türkiye 2.1%, Saudi Arabia 1.8%, South Africa 0.8%, Malaysia 0.8%, Philippines 0.8%, Vietnam 0.8%, Argentina 1.0%.
  - Other EM examples: Bulgaria 0.2%, Belarus 0.2%, Chile 0.4%, Colombia 0.7%, Ecuador 0.2%, Georgia 0.05%, Croatia 0.1%, Hungary 0.3%, Kazakhstan 0.4%, Moldova 0.03%, Peru 0.4%, Poland 1.2%, Romania 0.5%, Thailand 1.2%, Tunisia 0.1%, Ukraine 0.5%.
- Identification restrictions (impact-only summary):
  - Real GDP and CPI for each block (G20 AE_i, G20 EM_j, other EM_k) have sign/zero restrictions to distinguish shock types.
  - VIX restricted to have negative sign on impact for some shocks; Commodity Price Index restricted to have positive sign on impact for some shocks.
  - Additional growth-differential variables included with positive restrictions as specified.

### SVAR estimation details and measures of spillovers
- Identification and estimation:
  - Restrictions sharpen shock identification: one-quarter lag transmission from smaller to larger economies' prices; AD shocks from the U.S. and China imply an increase in commodity price index on impact; U.S. expansionary shocks have immediate downward impact on the VIX.
  - SVARs estimated with standard Bayesian methods using Minnesota prior hyper-parameters chosen to maximize marginal data density.
  - Heteroskedasticity correction: first three quarters of 2020 volatility increased by a factor of 10.
  - MATLAB procedures based on Canova-Ferroni toolbox used.
  - Shocks normalized to lead to a one percent increase four quarters ahead in origin country GDP.
- Measures reported:
  - Spillovers to output reported over three years; to consumer prices over one year.
  - Impulse Response Functions (IRFs), Historical Decompositions (HDs), and Forecast Error Variance Decompositions (FEVDs) used.
  - Key FEVD finding: shocks in G20 EMs can explain around 5 percent of real GDP fluctuations in advanced economies over a three-year horizon.

### GVAR methodology and estimation choices
- Model structure:
  - GVAR built from 63 country-specific VARX models: 34 AEs and 29 EMs, accounting for over 90 percent of global output.
  - Endogenous variables: real GDP, CPI inflation, real exchange rate, short-term rate, long-term rate.
  - Foreign variables via predetermined bilateral trade weights.
  - Two commodity blocks model oil and metal markets (prices, production, inventory, global demand).
- Data and trade weights:
  - Baseline weight matrix: bilateral trade averages 2000–2019; alternative estimation using 2017–2019.
- Estimation choices:
  - Bayesian VARs with stochastic volatility.
  - Four priors: Minnesota (MN), Stochastic Search Variable Selection (SSVS), Normal-Gamma (NG), Horseshoe (HS).
  - SSVS prior with one lag best on residual autocorrelation and cross-country residual correlation.
- Identification:
  - Local identification by imposing sign restrictions within each country.
  - AD shocks → positive co-movement between output and prices; AS shocks → negative co-movement.
  - Restrictions imposed for quarters zero to three.

### GVAR and SVAR results — output and inflation spillovers
- Relative size and reach (three-year-ahead GDP responses; shocks normalized to a one percent increase in Chinese GDP four quarters ahead):
  - China, positive aggregate demand shock → advanced economies: 0.2 percent (after three years).
  - China, positive aggregate demand shock → emerging market economies: 0.36 percent (after three years).
  - China, aggregate supply shocks → advanced and emerging market economies: both around 0.18 percent (three years ahead).
  - Other G20 EMs (simple average), both AD and AS shocks → about 0.02 percent (three years ahead).
- Cross-country exposure patterns:
  - Examples: Indonesia predominantly exposed to AEs; Argentina and Germany more symmetric; South Africa more impacted by EM shocks.
- Inflation spillovers (one-year-ahead responses to a China shock raising domestic output and inflation by 1 percentage point):
  - Advanced economies: 0.15 percentage point higher inflation after one year (AD shock).
  - Emerging markets: 0.19 percentage point higher inflation after one year (AD shock).
  - China AS shock → reduces inflation by 0.1 percentage point in advanced economies after one year.
  - China AS shock → reduces inflation by 0.19 percentage point in emerging markets after one year.
  - Consumer price variance share: China shocks contribute slightly more than 10 percent of one-year-ahead variation in consumer prices in other emerging markets; G20 EMs explain a small fraction of one-year-ahead variance in AEs.
- Evolution over time:
  - Spillovers from G20 EMs contained until the GFC but increased markedly since then.
  - Using 2017–19 trade weights vs 2001–03: spillovers from China approximately doubled over two decades prior to COVID-19.
  - Some G20 EMs (Türkiye, Saudi Arabia) show slight increases in spillovers to other EMs; spillovers from Russia declined.
  - Cross-country aggregates use PPP GDP weights and significance assessed with 68 percent credible intervals.

### Impact on oil prices (one-year-ahead IRFs)
- Responses to a 1 percentage point per year positive domestic aggregate demand and aggregate supply shocks in each G20 EM:
  - China AD shock → oil prices +2.2 percent.
  - China AS shock → oil prices +1.6 percent.
  - Spillovers from other G20 EMs → oil prices ~0.25 percent.
- Bars represent median IRFs; lines show 68 percent credible intervals.

### Firm-level analysis of spillovers — data, methodology, and main results
- Data and sample:
  - Orbis cross-country firm-level panel covering 42 countries, 2000–2022; analysis focuses on 2000–2019.
  - Final sample: >63 million observations; ~6.5 million unique firms (3.6 million in AEs; 2.9 million in EMs).
  - Baseline regression: ~34 million observations corresponding to 4.8 million firms.
  - Excluding G20 EMs (Brazil, China, India, Indonesia, Mexico, Türkiye) leaves ~55 million observations.
- Identification and exposure measures:
  - Shift-share approach: common shifter = idiosyncratic growth surprise in G20 EMs; exposures = output and input linkages at country-sector level.
  - Output linkages: share of global demand coming from G20 EM consumers/firms.
  - Input linkages: share of total inputs supplied by G20 EM industries.
  - Exposures constructed from EORA Global Supply Chain Database; measures capture direct exposures only.
  - Exposure treated as time invariant via regression of country-sector-year exposures on fixed effects; standardized fixed effects used as exposure.
- Estimation equation (firm revenue growth over horizon j):
  - (y_{isA,t+P} − y_{isA,t−1})/y_{isA,t−1} = β1*(InputExposure)*γ̂_{AiEM} + β2*(OutputExposure)*γ̂_{AiEM} + λ_{isA,t−1} + μ_i + θ_{Ai} + ϖ_{As} + ε_{isA,i}
  - Controls: firm fixed effects μ_i, country-year fixed effects θ_{Ai}, country-sector fixed effects, two lags of dependent variable; SEs clustered at firm level.
  - β1 and β2 measure differential effect of one standard deviation increase in exposure to idiosyncratic growth surprises in G20 EMs on firm revenue growth.
- Main baseline results (Online Annex Table 4.3.4, Panel 1; dependent = firm revenue growth):
  - 1 year: 0.4310*** (0.011)
  - 2 years: 0.4461*** (0.018)
  - 3 years: 0.4063*** (0.022)
  - 4 years: 0.2668*** (0.025)
  - 5 years: 0.2380*** (0.029)
  - Additional interaction coefficients (second block): 1 year 0.0332** (0.015); 2 years −0.0103 (0.025); 3 years 0.0260 (0.032); 4 years 0.0238 (0.036); 5 years 0.0224 (0.041).
  - Observations: 33,787,483 (1 year) down to 16,341,735 (5 years).
  - R-squared: 0.201 (1 year) to 0.532 (5 years).
- Panel 2 (restricted sample, j=5):
  - 1 year: 0.4591*** (0.012); 2 years: 0.4178*** (0.020); 3 years: 0.3559*** (0.024); 4 years: 0.2141*** (0.027); 5 years: 0.2380*** (0.029).
  - Interaction coefficients: 1 year 0.0415** (0.017) ... 5 years 0.0224 (0.041).
  - Observations: 16,105,295 to 16,341,735. R-squared: 0.239 to 0.532.
- Firm-level spillovers from G20 AEs vs G20 EMs:
  - Spillovers to firms reliant on demand from G20 AEs are ~twice as large as to firms dependent on demand from G20 EMs.
  - Downstream spillovers are on average negative, suggesting import-competition effects adversely affect firms more dependent on inputs supplied by G20 AEs.
- Country-specific firm-level coefficients (Online Annex Table 4.3.5) — selected entries (dependent = firm revenue growth):
  - Argentina: 1 year 0.0174*** (0.002); 2 years −0.0088*** (0.002); 3 years −0.0056** (0.003); 4 years −0.0249*** (0.003); 5 years −0.0367*** (0.003).
  - China: 1 year 0.0080 (0.007); 2 years 0.0060 (0.011); 3 years 0.0407*** (0.015); 4 years 0.0235 (0.019); 5 years 0.0088 (0.020).
  - India: 1 year 0.2284*** (0.009); 2 years 0.2584*** (0.014); 3 years 0.1774*** (0.017); 4 years 0.0643*** (0.020); 5 years 0.1694*** (0.022).
  - Mexico: 1 year 0.5650*** (0.014); 2 years 0.8979*** (0.025); 3 years 1.0065*** (0.032); 4 years 1.0386*** (0.039); 5 years 1.0930*** (0.045).
  - Russia: 1 year −0.0758*** (0.029); 2 years −0.2125*** (0.050); 3 years −0.3750*** (0.066); 4 years −0.5846*** (0.081); 5 years −0.7169*** (0.091).
  - Observations for country regressions: 16,105,295 (1 year) to 16,341,735 (5 years).
  - All regressions include two lags of dependent variable; SEs clustered at firm level. ***, **, * denote 1%, 5%, 10%.
- Robustness:
  - Results robust to splitting sample by firm headquarter (AEs vs non-G20 EMs), using firm sales as dependent variable, adding firm size and long-term debt share, excluding countries with lower coverage, computing exposure variables as simple averages, and clustering at country-year level (which yields larger SEs and reduces significance after 4 years for output-dependent firms).
- Spillovers on firm sales (Online Annex Table 4.3.6, Panel 1):
  - 1 year: 0.4463*** (0.011); 2 years: 0.4586*** (0.017); 3 years: 0.4206*** (0.022); 4 years: 0.2866*** (0.025); 5 years: 0.2990*** (0.029).
- Alternative clustering and exposure computations (Online Annex Tables 4.3.9–4.3.10) report similar coefficient patterns with larger SEs when clustering at country-year level.

### Global trade model (BHLP) — model exposition, calibration, and scenarios
- Model type and structure:
  - Multi-region, multi-sector general-equilibrium global network model (BHLP following Huo, Levchenko, and Pandalai-Nayar (2019) and Bonadio and others (2021, 2023)).
  - Static global economy with N regions and J sectors; tradeable sectoral output with iceberg trade costs; labor immobile between regions.
  - Final consumption: nested CES across N×J region-sectors; production with representative firm Y_F_P = Z_F_P [K_F_P^α_j H_F_P^(1−α_j)]^η_j λ_F_P^(1−η_j); intermediate goods λ_F_P nested CES.
  - Extensive margin fixed; intensive margin adjusts.
  - Solution via exact hat algebra (Dekle and others (2008)).
- Data and calibration:
  - Data: OECD ICIO Tables (2018); 45 sectors × 66 economies (36 AEs, 26 EMs, 4 developing economies, rest of world).
  - Key calibrated parameters (Online Annex Table 4.4.1):
    - Human capital distribution parameter: 1.5.
    - Intermediate goods substitution elasticity across countries: 4.
    - Final goods substitution elasticity across countries: 4.
    - Intermediate goods production substitution elasticity across sectors: 0.95.
    - Final goods substitution elasticity across sectors: 0.95.
    - Capital Share: 0.
    - Frisch elasticity of labor supply: 1.
- Scenarios simulated (sectoral TFP shocks):
  1. Negative 2.5 percent TFP shock to all sectors in all G20 EMs.
  2. Negative 2.5 percent TFP shock to GVC sectors in all G20 EMs.
  3. Negative 2.5 percent TFP shock to the Construction sector in China.
  4. Positive 2.5 percent TFP shock to the IT sector in India.
- Key qualitative result for Scenario 4 (India IT +2.5%):
  - Sectoral reallocation shows less heterogeneity than Construction shock in China.
  - Most significant impact: reduction in value added in IT sector outside India (advanced economies and EMs).
  - All other sectors expand, particularly energy commodities production — interpreted as expansion in India’s IT sector increasing competition with other countries’ IT sectors.
- Sensitivity to elasticities:
  - Alternative A (γ = ν = 2): impacts roughly twice baseline (examples: from 0.155 (0.093) percent of GDP to 0.29 (0.16) percent of GDP in Baseline (GVC) scenario).
  - Alternative B (ρ = ε = 0.5): impact increases negligibly to 0.16 (0.095) percent of GDP in Baseline (GVC).
  - Interpretation: lower γ, ν (less substitutability across regions) increases impact; lower ρ, ε (less substitutability across sectors) has smaller effect for region-wide shocks.
- Employment sensitivity methodology:
  - For 19 countries (G20 EMs and G20 AEs excluding Australia) compute elasticity of sectoral value added to productivity shocks in all sectors of other countries; multiply by sectoral employment share to get employment elasticities.
  - For each affected country-sector (19×45), keep three largest positive elasticities (complementarity) and three largest negative elasticities (competition); counts by source group reported in Table 4.1.
  - Historical comparison (2000 vs 2018): complementarity source sectors from G20 EMs rise from 3 (2000) to 34 (2018); competition rises from 6 (2000) to 28 (2018), indicating much larger G20 EM trade footprint in 2018.
- Representative employment-spillover counts (Online Annex Table 4.4.3, 2018) — selected entries:
  - Complementarity (Number of Destination Economies Affected): EM Computer, electronic and optical equipment: 7; EM Textiles, wearing apparel, leather and related products: 7; AE Motor vehicles, trailers and semi-trailers: 5.
  - Competition (Number of Destination Economies Affected): EM Computer, electronic and optical equipment: 8; EM Basic metals: 3; EM Machinery and equipment: 3.
  - Entries sum to 57 = 19 economies × 3 source sectors per panel.
- ICIO sector code reference (selected mappings):
  - 01T02 Agriculture, hunting, forestry
  - 05T06 Mining and quarrying, energy producing products
  - 07T08 Mining and quarrying, non-energy producing products
  - 10T12 Food products, beverages and tobacco
  - 13T15 Textiles, textile products, leather and footwear
  - 15 Basic metals
  - 26 Computer, electronic and optical equipment
  - 29 Motor vehicles, trailers and semi-trailers
  - 45T47 Wholesale and retail trade
  - 64T66 Financial and insurance activities
  - 85 Education
  - 94T96 Other service activities

### GIMF G20 EM Upside Scenario — model, calibration, and quantitative impacts
- GIMF model structure and calibration highlights:
  - Annual, multi-region DSGE with 10 regions and an added GVC tradable sector.
  - Households: some non-Ricardian; some liquidity constrained.
  - Firms: monopolistic competition; GVC roundabout production; Bernanke-Gertler-Gilchrist (1999) financial accelerator.
  - Monetary policy: inflation forecast-based targeting; fiscal policy: debt-GDP targeting.
  - Calibration: OECD ICIO 2018 and national accounts.
  - Intertemporal elasticity of substitution = 0.2.
  - Share of liquidity constrained households: 25 percent in United States, EU+, other AEs, China; 50 percent in other EM regions.
  - Trade/production elasticities (All Regions): Consumption 1.5; Investment 1.5; Non-GVC Tradables 1.5; GVC Tradables 0.5 (domestic/imported); Different regions elasticity for GVC Tradables: 0.8.
  - Elasticity between nontradable and tradable goods bundle = 0.5; tradable bundle elasticity = 0.95.
- Domestic sector calibration (percent of region's GDP, Online Annex Table 4.5.2 — selected entries):
  - Share of Global GDP (US$): United States 24.4; European Union + 18.9; Other Advanced Economies 8.7; China 16.7; India 3.2; Southeast Asia 2.3; Middle East and CIS 3.2; Mexico 1.2; Other Latin America 2.3; Rest of the World 16.7.
  - Household Consumption (percent of region's GDP): United States 65.4; European Union + 54.9; China 51.7; India 56.7; Mexico 63.0.
  - Private Investment (percent of region's GDP): United States 17.3; China 18.5; India 26.8; Mexico 22.0.
  - Aggregate Exports (percent of GDP): United States 11.4; European Union + 20.1; China 17.4; India 20.1; Southeast Asia 44.2; Mexico 36.3.
  - GVC Tradable exports (% of GDP): United States 3.8; European Union + 7.1; China 7.5; India 5.8; Mexico 22.7.
  - GVC Tradable imports (% of GDP): United States 5.4; European Union + 7.0; China 9.8; India 12.5; Mexico 17.8.
- Scenario design:
  - G20 EM (excluding China) upside scenario: shocks to short-term aggregate demand (household consumption and private investment) in G20 EMs ex-China over five years (forecast horizon).
  - Shock size represents a likely upside to WEO baseline: 30 percent probability that growth in each EM ex-China could be higher than envisaged in this scenario (or 70 percent probability that growth is lower).
  - For aggregated regions containing G20 EMs and non-G20 EMs, each G20 EM’s shock profile scaled by GDP share within the region to construct aggregated region shock profile.
- Quantitative impacts (70th percentile of upside scenario — deviations of real GDP growth from WEO baseline, Online Annex Table 4.5.4):
  - Year 3 (2026) / Year 5 (2028) deviations (percentage points):
    - All EM Regions: 0.7 / 0.9
    - India: 0.7 / 1.0
    - Southeast Asia: 0.5 / 0.6
    - Middle East and CIS: 1.3 / 1.5
    - Mexico: 2.5 / 3.0
    - Other Latin America: 1.1 / 1.4
    - Other EMDEs: 0.2 / 0.3
  - Regional fifth-year impacts range: 0.3 percentage points (other EMDEs) to 1 percentage point (India), 1.5 percentage points (Middle East and CIS), 3 percentage points (Mexico).
  - Potential aggregate effect: this G20 EM upside could push up global growth, on average, by at most 0.5 percentage points.
  - Spillovers to AEs and China: generally positive but small; both AEs and China growth could be up to 0.1 percent higher on average in the first three years, then subside.
  - Other G20 EM shocks generate spillovers among EMDE regions, accounting for about 15 percent of the upside scenario impact.

### EM spillovers on EMDEs vs advanced economies
- Key finding:
  - EM spillovers on EMDEs are larger than on advanced economies (Online Annex Figure 4.5.2), because the upside scenario is based on higher G20 EM aggregate demand for final goods, of which EMDEs produce more than AEs.
- Largest contributors and channels:
  - For AEs (largest spillovers after 3 and 5 years): Middle East and CIS (hydrocarbon exports); Mexico (two-way trade with United States); Southeast Asia (trade with China and AEs).
  - For EMDEs (largest spillovers after 3 and 5 years): India (supplier of intermediates and final goods via GVCs; major user of raw materials); Middle East and CIS (hydrocarbons); Southeast Asia (supplier of final consumption goods).
- Quantitative presentation:
  - Impacts reported as percentage point deviations from WEO baseline on level of real GDP at horizons of 3 and 5 years.
  - Figure axis values shown in source: −0.05, 0.00, 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35.

*Source: CHAPTER 4, Online Annex (ch4onlineannex - CHAPTER 4), International Monetary Fund | April 2024.*

### Annex Table 4.1.1). As a benchmark, the same

### ch4onlineannex - Annex Table 4.1.1). As a benchmark, the same

### Idiosyncratic growth surprises and correlations
- Growth surprises are defined as the estimated residual (γ) of equation (2) (using the April WEO projections) while idiosyncratic growth surprises are defined as the residual of a regression (notation in source).
- Findings:
  - Correlation across idiosyncratic growth surprises in advanced economies (AEs) excluding the U.S. is positive and statistically significant throughout the entire period (reference: Online Annex Figure 4.1.2).
  - The role of G20 EMs in the global economy has become more relevant in the last decade (interpretation from the positive correlation).
- Key visuals described:
  - Online Annex Figure 4.1.1: Idiosyncratic Growth Surprises (Percent) — countries shown include Argentina, Brazil, China, Indonesia, India, Mexico, Russia, Saudi Arabia, Türkiye, South Africa across years 2000–2022.
  - Online Annex Figure 4.1.2: Correlation across Idiosyncratic Growth Surprises (Percent) — plotted versus Idiosyncratic growth surprises in AEs and Average idiosyncratic growth surprises in US for 1990–2013 and 2014–22.

### Regression evidence on AE growth surprises (table extract)
- Dependent variable: Growth surprises in AEs
- Selected coefficient estimates (from Online Annex Table 4.1.1):
  - Average growth surprise G-20 EM x I(1990-2013): -0.0459 (standard error 0.075) and -0.0311 (standard error 0.080) across columns.
  - Average growth surprise G-20 EM x I(2014-2022): 0.5153** (standard error 0.204) and 0.4619** (standard error 0.217).
  - Average growth surprise U.S. x I(1990-2013): 0.4336*** (standard error 0.063) (column 3).
  - Average growth surprise U.S. x I(2014-2022): 0.4444*** (standard error 0.133) (column 3).
- Sample and fit:
  - Observations: 297, 264, 264 (across columns).
  - R-squared: 0.023, 0.018, 0.187 (across columns).
  - Sample: AEs; AEs (ex. U.S.); AEs (ex. U.S.) (column-wise).
- Notes:
  - Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
  - Country fixed effects included (Country FE: Y Y Y).
  - Growth surprises defined as the estimated residual (γ) of equation (2).

### G20 EM trade integration (gravity model)
- Methodology:
  - A standard gravity model for bilateral goods trade is estimated on annual IMF Direction of Trade Statistics (DOTS) data from 1980 to 2022.
  - Dependent variable: logarithm of goods imports from source country s to destination country d in year t.
  - Key regressor: dummy equal to one if either source or destination country is a G20 EM, interacted with year dummies (t).
  - Fixed effects: source x year, destination x year, and country-pair fixed effects to absorb time-varying push/pull factors and time-invariant bilateral heterogeneity.
- Interpretation:
  - Estimated coefficients β_i measure the difference in trade involving G20 EMs compared with trade flows among other countries.
  - Trade with G20 EMs accelerated following China’s accession to the WTO in 2001 and more so in the second half of the 2010s.
  - The increasing trade integration of G20 EMs is not uniquely attributable to China; other G20 EMs also contributed.
- Key visual described:
  - Online Annex Figure 4.1.3: Increase in G20 Emerging Markets Global Trade Excluding China (Log points) — model coefficients plotted yearly for a dummy for bilateral pairs involving G20 EMs (reference year 2001).

### G20 EM financial integration (US portfolio assets)
- Data and measurement:
  - Security-level US cross-border portfolio holdings from the TIC system (Bertaut, Bressler and Curcuru (2019)) are used.
  - Two presentations: nationality basis and residence basis, both as share of total US portfolio assets abroad.
- Findings:
  - US portfolio assets vis-à-vis G20 EMs rose from 3 percent in the early 2000s to around 7 percent in the early 2020s, driven largely by US assets in China and, to a lesser extent, India.
  - On a residence basis, the increase in US portfolio assets in China is not visible, signaling that most recent US investment in China has been channeled through third countries.
- Key visuals described:
  - Online Annex Figure 4.1.4 panel 1: US Portfolio Assets vis-à-vis G20 EMs on a Nationality Basis (Percent of total assets) — shares for Argentina, Brazil, China, Indonesia, India, Mexico, Russia, Saudi Arabia, Türkiye, South Africa over 2003–2022.
  - Online Annex Figure 4.1.4 panel 2: US Portfolio Assets vis-à-vis G20 EMs on a Residence Basis (Percent of total assets) — analogous series showing the residence-based pattern.

### Are G20 EMs small and open? SOE model evidence
- Frameworks used:
  - An updated implementation of Fernandez, Schmitt-Grohé and Uribe (2017) SOE model: domestic cyclical GDP in a G20 EM is influenced by a vector of global variables (F_i) that includes cyclical components of real prices of agriculture, metals, and fuel, plus one of several financial variables (real gross return of 3-month US T-Bills, real gross return on US 10-year government bonds, US real effective exchange rate, or Moody’s investment-grade corporate spread).
  - Block exogeneity assumption: global variables can influence domestic economy contemporaneously and with lags, but domestic economy cannot affect commodity prices or international financial conditions contemporaneously or with lag (appropriate for SOEs).
  - Estimation on quarterly data since 1980 (except Russia since 1995); Argentina and Saudi Arabia are excluded due to data limitations.
  - Two sub-sample periods reported: before and after the Global Financial Crisis (GFC); GFC and COVID-19 periods excluded.
- Tests and results:
  - Likelihood ratio tests compare unrestricted estimates with those restricted by the block-exogeneity assumption to identify which emerging markets satisfy the SOE assumption.
  - First test (impact on one global variable at a time): after the GFC most G20 EMs affect at least one global price, notably agriculture commodity prices and short-term interest rates.
  - Stricter test (weak exogeneity with respect to entire global vector): the SOE assumption is rejected for China across all model specifications starting in the early 2000s.
  - Some evidence other EMs affect financial variables even after controlling for commodity effects: Brazil in post-GFC; South Africa, Mexico, and Indonesia when global vector includes investor risk-appetite variables (investment grade corporate spread).
- Related evidence:
  - Online Annex Figure 4.1.5: Times Emerging Markets are Mentioned on FED’s FOMC Documents (Percent of document word count) — rising mentions over time (series from 1982 through 2012/2017 indicated).
  - Aguiar and Gopinath (2007)-style tests: for G20 EMs excluding China, India, and Mexico, panel estimates (GMM after Burnside (1999)) indicate these economies continue to exhibit emerging market SOE properties — specifically, ratio of consumption to output volatilities and correlation between net exports and output consistent with permanent income hypothesis stylized facts (Online Annex Figure 4.1.6).
- Notes on data treatment:
  - China: due to measurement concerns post-GFC, alternative indicators are used (San Francisco Fed’s China Cyclical Activity Tracker; Li indicator) transformed to quarterly GDP in levels and seasonally adjusted.
  - For other countries, quarterly GDP back to 1980 compiled using datasets from Aslam and others (forthcoming), WEO, IFS, OECD, national statistics offices, with temporal disaggregation techniques for earlier periods.

### Aggregate spillovers: VAR analysis (SVAR methodology and identification)
- Objective: quantify spillovers from the G20 EMs to growth in other emerging market and advanced economies.
- SVAR setup:
  - Country-specific medium-scale SVARs estimated for triples: a G20 advanced economy (AE), a G20 emerging market (EM), and another non-G20 EM.
  - Additional global variables: VIX (proxy for international uncertainty and financial cycles) and a commodity price index (global commodity price cycles).
  - Two growth differentials between country pairs included to sharpen identification.
  - Each SVAR includes ten variables (specified in Online Annex Table 4.2.2); for each G20 EM, 37 models are estimated.
  - Estimation sample: quarterly data from Q1 2001 to Q3 2023.
- Identification strategy:
  - Combination of sign and zero restrictions (see Online Annex Table 4.2.2).
  - Sign restrictions separate domestic demand from domestic supply shocks; zero restrictions differentiate domestic from foreign shocks by assuming shocks from smaller economies are transmitted to larger economies only with a lag of one quarter.
  - Example implication: in a model with the U.S., Mexico, and Chile, shocks from Chile (average weight 0.4 percent of total sample GDP PPP over 2000-2023) can affect U.S. real GDP (share 20.8 percent) and Mexico (share 2.6 percent) with a lag of one quarter; shocks from Mexico can exert an immediate effect on Chile but impact U.S. real GDP only after one quarter.
- Country coverage and sample weights (Online Annex Table 4.2.1 — average weights in total sample 2000-2023, GDP PPP US$):
  - G20 Advanced Economies: United States 20.8%, Japan 5.9%, Germany 4.5%, France 3.2%, United Kingdom 3.2%, Italy 2.8%, Korea 2.1%, Canada 1.9%, Australia 1.3%.
  - G20 Emerging Economies: China 18.8%, India 7.8%, Brazil 3.4%, Russia 4.1%, Indonesia 2.9%, Mexico 2.6%, Türkiye 2.1%, Saudi Arabia 1.8%, South Africa 0.8%, Malaysia 0.8%, Philippines 0.8%, Vietnam 0.8%, Argentina 1.0%.
  - Other Emerging Economies (sample examples and weights): Bulgaria 0.2%, Belarus 0.2%, Chile 0.4%, Colombia 0.7%, Ecuador 0.2%, Georgia 0.05%, Croatia 0.1%, Hungary 0.3%, Kazakhstan 0.4%, Moldova 0.03%, Peru 0.4%, Poland 1.2%, Romania 0.5%, Thailand 1.2%, Tunisia 0.1%, Ukraine 0.5%.
- Identification restrictions (impact only; summary from Online Annex Table 4.2.2):
  - Variables listed for each block (G20 AE_i, G20 EM_j, other EM_k): real GDP and CPI with sign/zero restrictions to distinguish supply vs demand shocks and domestic vs foreign shocks.
  - VIX restricted to have negative sign on impact for some shocks; Commodity Price Index restricted to have positive sign on impact for some shocks.
  - Additional growth-differential variables included with positive restrictions as specified.

*Source: IMF staff calculations.*

### CHAPTER 4

### CHAPTER 4

### Identification and Estimation
- Structural VAR (SVAR) identification:
  - Imposed restrictions sharpen shock identification (e.g., one-quarter lag transmission from smaller to larger economies' prices; AD shocks from the U.S. and China imply an increase in the commodity price index on impact; U.S. shocks have an immediate impact on the VIX—downwards if expansionary).
  - Shocks from economies other than the U.S. and China may affect VIX and commodity prices immediately but with no imposed sign.
- Estimation details:
  - SVARs estimated with standard Bayesian methods using the Minnesota prior hyper-parameters chosen to maximize the marginal data density.
  - Heteroskedasticity correction for the first three quarters of 2020 assuming volatility increases by a factor of 10.
  - MATLAB procedures based on the Canova-Ferroni toolbox used.
- Shock normalization:
  - Shocks are normalized to lead to a one percent increase four quarters ahead in GDP of the country of origin of the shock.

### Measures of Spillovers (IRFs, HDs, FEVDs)
- Output horizons:
  - Spillovers to output reported over three years.
  - Spillovers to consumer prices reported over one year.
- Instruments and interpretations:
  - Impulse Response Functions (IRFs) capture the impact of aggregate demand (AD) and aggregate supply (AS) shocks on GDP and CPI and control variables.
  - Historical Decompositions (HDs) illustrate each estimated shock’s contribution to variables each quarter; contributions of domestic shocks derived as weighted averages of domestic AD and AS shocks from a selected SVAR for each G20 EM country; contributions of foreign shocks derived as residual component not explained by domestic shocks.
  - Forecast Error Variance Decompositions (FEVDs) quantify the fraction of GDP variance explained by each shock, accounting for estimated size and frequency of shocks.
- Key FEVD finding:
  - Shocks in G20 EMs can explain around 5 percent of real GDP fluctuations in advanced economies, over a three-year horizon.

### Country Coverage and Model Blocks
- SVAR small-open-economy block:
  - Each selected SVAR includes: real GDP and CPI data for the U.S., the G20 EM of interest, and a smaller emerging market economy (South Africa for models with China/India/Russia/Brazil/Indonesia/Mexico/Türkiye/Saudi Arabia/Argentina as the G20 EM of interest; Tunisia for the model with South Africa as the G20 EM of interest), plus a commodity price index and the US VIX.
- Aggregation:
  - Aggregates of the G20 EMs excluding China are derived as weighted averages of median estimates of all other nine G20 EMs, regardless of statistical significance.

### GVAR Methodology
- Model structure:
  - Global Vector Autoregressive (GVAR) model built from 63 country-specific VARX models: 34 advanced economies and 29 emerging market economies, collectively accounting for over 90 percent of global output.
  - Endogenous variables: real GDP, CPI inflation, the real exchange rate, the short-term rate and the long-term rate.
  - Foreign variables incorporated through predetermined bilateral trade weights.
  - Two distinct commodity blocks separately model oil and metal markets (prices, production, inventory, global demand).
- Data and trade weights:
  - Baseline weight matrix constructed using bilateral trade averages from 2000 to 2019.
  - Alternative estimation using trade data spanning 2017 to 2019.
- Estimation choices:
  - Bayesian VARs with stochastic volatility employed.
  - Four alternative priors explored: Minnesota (MN), Stochastic Search Variable Selection (SSVS), Normal-Gamma (NG), Horseshoe (HS).
  - SSVS prior with one lag structure exhibits best performance on residual autocorrelation and cross-country residual correlation.
- Identification:
  - Local identification of structural shocks by imposing sign restrictions within the country of interest.
  - Aggregate demand shocks assumed to generate positive co-movement between output and prices; aggregate supply shocks assumed to generate negative co-movement.
  - Restrictions imposed for four quarters (quarters zero to three).

### GVAR and SVAR Results — Output Spillovers
- Relative size and reach:
  - China’s shocks have more sizable spillovers than those from other G20 EMs; spillovers from China are larger for emerging market economies than for advanced economies.
- Quantified three-year-ahead GDP responses (shocks normalized to a one percent increase in Chinese GDP four quarters ahead):
  - China, positive aggregate demand shock → advanced economies: 0.2 percent (after three years).
  - China, positive aggregate demand shock → emerging market economies: 0.36 percent (after three years).
  - China, aggregate supply shocks → advanced and emerging market economies: both around 0.18 percent (three years ahead).
  - Other G20 EMs (simple average), both AD and AS shocks → about 0.02 percent (three years ahead).
- Cross-country comparisons:
  - The relative importance of G20 advanced and emerging markets (excluding the U.S. and China) in explaining GDP fluctuations is broadly comparable for several countries; exposure patterns differ by country (examples cited: Indonesia predominantly exposed to AEs; Argentina and Germany more symmetric; South Africa more impacted by EM shocks).

### GVAR and SVAR Results — Inflation Spillovers
- One-year-ahead inflation responses to a China shock:
  - China, positive aggregate demand shock that raises domestic output and inflation by 1 percentage point → advanced economies: 0.15 percentage point higher inflation after one year.
  - Same China shock → emerging markets: 0.19 percentage point higher inflation after one year.
  - China, positive aggregate supply shock → reduces inflation by 0.1 percentage point in advanced economies after one year.
  - China, positive aggregate supply shock → reduces inflation by 0.19 percentage point in emerging markets after one year.
- Consumer price variance share:
  - Shocks from China contribute to slightly more than 10 percent of the one-year-ahead variation in consumer prices in other emerging markets.
  - Shocks from G20 EMs explain a small fraction of the one-year-ahead variance of consumer prices in advanced economies.

### Evolution of Spillovers Over Time
- Time variation:
  - Spillovers from G20 EMs were contained until the Global Financial Crisis (GFC) but increased markedly since then.
  - Spillovers from advanced economies remained significant throughout the past two decades.
- Trade-weight sensitivity:
  - Using more recent trade weights (2017–19) versus earlier trade weights (2001–03) shows that spillovers from China approximately doubled over the two decades prior to the COVID-19 pandemic.
  - Some other G20 EMs (notably Türkiye and Saudi Arabia) show slight increases in spillovers to other EMs; spillovers from Russia have declined in the comparison.
- Note on reported aggregates:
  - Reported cross-country aggregate results use PPP GDP weights and consider impulse responses significant on the basis of 68 percent credible intervals.

*Source: CHAPTER 4, Online Annex (IMF staff calculations). International Monetary Fund | April 2024*

### 1. Growth Spilloversto Advanced Economies

### 1. Growth Spilloversto Advanced Economies

### Impact on Oil Prices
- One-year-ahead impulse responses of oil price to a 1 percentage point per year positive domestic aggregate demand and aggregate supply shocks in each G20 EM:
  - Aggregate demand shock originating in China raises oil prices by about 2.2 percent.
  - Aggregate supply shock originating in China increases oil prices by 1.6 percent.
  - Spillovers from shocks in other G20 EMs are small (around 0.25 percent).
- Note: Bars represent median impulse responses and lines identify the 68 percent credible intervals.

*Source: IMF staff calculations (Online Annex Figure 4.2.7).*

### Firm-level Analysis of Spillovers — Data and Sample
- Original Orbis cross-country firm-level panel covers 42 countries from 2000 to 2022; analysis focuses on 2000 to 2019.
- Dataset includes corporate balance sheet and income statement information; cleaning follows Kalemli-Özcan and others (2024) and Díez, Fan and Villegas-Sanchez (2021).
- Sample exclusions:
  - Firms with less than five yearly observations.
  - Financial firms and public sector firms.
  - Firms in sectors not covered by the EORA input-output tables.
- Final sample counts and composition:
  - More than 63 million observations.
  - Almost 6.5 million unique firms: 3.6 million in advanced economies and 2.9 million in emerging market economies.
  - Excluding G20 EMs from the sample (Brazil, China, India, Indonesia, Mexico and Türkiye) leaves almost 55 million observations.
  - Baseline regression run on almost 34 million observations corresponding to 4.8 million firms.
- Sample distribution highlights (selected):
  - Yearly observations (Online Annex Table 4.3.3): 2002: 736,673 (1.88%); 2019: 2,952,886 (7.52%).
  - Sectoral shares (Online Annex Table 4.3.2): Construction 6,546,135 (19.37%); Retail Trade 5,315,806 (15.73%); Wholesale Trade 5,368,671 (15.89%).

### Methodology
- Identification: shift-share approach where the common shifter is the idiosyncratic growth surprise in G20 EMs and exposures are measures of output and input linkages at the country-sector level.
- Exposure measures:
  - Output linkages measure the share of global demand coming from G20 EM consumers and firms (sales of intermediate inputs and final goods to emerging market economies).
  - Input linkages measure the share of total inputs supplied by G20 EM industries.
  - Indicators constructed using input-output tables from the EORA Global Supply Chain Database.
- Caveat: measures capture only direct exposures to G20 EMs, missing indirect ones.
- Exposure variable treated as time invariant by regressing country-sector-year exposures against country-sector and year fixed effects; standardized estimated country-sector fixed effects used as exposure measures.
- Estimation equation (growth in firm revenue y over horizon t+j):
  - (푦푦_{isA,t+P} − 푦푦_{isA,t−1})/푦푦_{isA,t−1} = 훽1 * (InputExposure) * 훾̂_{AiEM} + 훽2 * (OutputExposure) * 훾̂_{AiEM} + λ_{isA,t−1} + μ_i + θ_{Ai} + ϖ_{As} + ε_{isA,i}
  - Firm fixed effects (μ_i), country-year fixed effects (θ_{Ai}), and country-sector fixed effects included; two lags of the dependent variable included; standard errors clustered at the firm level.
- Interpretation:
  - Coefficients 훽1 and 훽2 trace the differential effect of a one standard deviation increase in exposure to idiosyncratic growth surprises in G20 EMs on firm revenue growth.

### Main Results (selected)
- Baseline results for G-20 EMs (Online Annex Table 4.3.4, Panel 1: whole sample; dependent variable: firm revenue growth):
  - 1 year horizon: coefficient 0.4310*** (0.011)
  - 2 years horizon: coefficient 0.4461*** (0.018)
  - 3 years horizon: coefficient 0.4063*** (0.022)
  - 4 years horizon: coefficient 0.2668*** (0.025)
  - 5 years horizon: coefficient 0.2380*** (0.029)
  - Additional interaction coefficient row (second block in table) for Panel 1: 1 year 0.0332** (0.015); 2 years −0.0103 (0.025); 3 years 0.0260 (0.032); 4 years 0.0238 (0.036); 5 years 0.0224 (0.041).
  - Observations: 33,787,483 (1 year) down to 16,341,735 (5 years).
  - R-squared: 0.201 (1 year) up to 0.532 (5 years).
- Panel 2 (restricted sample, j=5) highlights:
  - 1 year: 0.4591*** (0.012); 2 years: 0.4178*** (0.020); 3 years: 0.3559*** (0.024); 4 years: 0.2141*** (0.027); 5 years: 0.2380*** (0.029).
  - Interaction coefficients second block: 1 year 0.0415** (0.017); 2 years −0.0176 (0.027); 3 years 0.0292 (0.034); 4 years 0.0689* (0.039); 5 years 0.0224 (0.041).
  - Observations: 16,105,295 (1 year) to 16,341,735 (5 years).
  - R-squared: 0.239 (1 year) to 0.532 (5 years).
- Firm-level spillovers from G20 advanced economies to firm revenue growth (Online Annex Figure 4.3.1):
  - Spillovers to firms particularly reliant on demand from G20 advanced economies are about twice as large as those to firms dependent on demand from G20 EMs.
  - Downstream spillovers are on average negative, suggesting import-competition effects adversely affect firms more dependent on inputs supplied by G20 advanced economies.

### Country-specific Firm-level Results (selected coefficients from Online Annex Table 4.3.5)
- Panel 1: Argentina (dependent variable: firm revenue growth)
  - 1 year: 0.0174*** (0.002)
  - 2 years: −0.0088*** (0.002)
  - 3 years: −0.0056** (0.003)
  - 4 years: −0.0249*** (0.003)
  - 5 years: −0.0367*** (0.003)
- Panel 3: China
  - 1 year: 0.0080 (0.007)
  - 2 years: 0.0060 (0.011)
  - 3 years: 0.0407*** (0.015)
  - 4 years: 0.0235 (0.019)
  - 5 years: 0.0088 (0.020)
- Panel 5: India
  - 1 year: 0.2284*** (0.009)
  - 2 years: 0.2584*** (0.014)
  - 3 years: 0.1774*** (0.017)
  - 4 years: 0.0643*** (0.020)
  - 5 years: 0.1694*** (0.022)
- Panel 6: Mexico
  - 1 year: 0.5650*** (0.014)
  - 2 years: 0.8979*** (0.025)
  - 3 years: 1.0065*** (0.032)
  - 4 years: 1.0386*** (0.039)
  - 5 years: 1.0930*** (0.045)
- Panel 7: Russia
  - 1 year: −0.0758*** (0.029)
  - 2 years: −0.2125*** (0.050)
  - 3 years: −0.3750*** (0.066)
  - 4 years: −0.5846*** (0.081)
  - 5 years: −0.7169*** (0.091)
- Observations for country-specific regressions: 16,105,295 (1 year) to 16,341,735 (5 years).
- Note: All regressions include two lags of the dependent variable; standard errors clustered at firm level. ***, **, * denote significance at 1%, 5%, 10%.

### Robustness Tests
- Results robust to:
  - Splitting sample between firms headquartered in advanced economies and in non-G20 EMs (Online Annex Figure 4.3.2).
  - Taking firm sales as the dependent variable (Online Annex Table 4.3.6).
  - Including firm size (log of assets) and share of long-term debt in total liabilities as additional controls (Online Annex Table 4.3.7).
  - Excluding countries with lower or less stable coverage in sales (Online Annex Table 4.3.8).
  - Computing exposure variables as simple averages over sample periods of input and output linkages (Online Annex Table 4.3.9).
  - Clustering standard errors at the country-year level: in this specification, standard errors are larger and the effect on output-dependent firms becomes no longer significant after 4 years.
- Spillovers on firm sales (Online Annex Table 4.3.6, Panel 1: whole sample):
  - 1 year: 0.4463*** (0.011)
  - 2 years: 0.4586*** (0.017)
  - 3 years: 0.4206*** (0.022)
  - 4 years: 0.2866*** (0.025)
  - 5 years: 0.2990*** (0.029)
  - Interaction coefficients: 1 year 0.0590*** (0.014); 2 years 0.0145 (0.024); 3 years 0.0464 (0.030); 4 years 0.0293 (0.034); 5 years −0.0031 (0.039).
- Additional-controls specification (Online Annex Table 4.3.7, Panel 1):
  - 1 year: 0.4327*** (0.011)
  - 2 years: 0.4405*** (0.017)
  - 3 years: 0.3810*** (0.021)
  - 4 years: 0.1955*** (0.024)
  - 5 years: 0.1488*** (0.027)
  - Interaction coefficients: 1 year 0.0275* (0.015); 2 years −0.0343 (0.024); 3 years −0.0030 (0.030); 4 years 0.0038 (0.034); 5 years 0.0054 (0.037).
- Restricted-sample results (Online Annex Table 4.3.8, Panel 1: whole sample):
  - 1 year: 0.5128*** (0.012)
  - 2 years: 0.5160*** (0.019)
  - 3 years: 0.4681*** (0.024)
  - 4 years: 0.3255*** (0.028)
  - 5 years: 0.3298*** (0.032)
  - Interaction coefficients: 1 year 0.1453*** (0.017); 2 years 0.1231*** (0.028); 3 years 0.1511*** (0.036); 4 years 0.1360*** (0.041); 5 years 0.0948** (0.046).

*Source: Online Annex 4.3, “Firm-level Analysis of Spillovers,” World Economic Outlook, International Monetary Fund | April 2024.*

### CHAPTER 4

### CHAPTER 4

### Firm-level spillovers — alternative exposure variables and clustering
- Online Annex Table 4.3.9 and Online Annex Table 4.3.10 present firm-level regressions of "Firm revenue growth over: 1 year, 2 years, 3 years, 4 years, 5 years" with panel specifications and clustering alternatives.
- Panel 1: whole sample (exposure variables computed as simple averages over the sample period of exposures defined at the country-sector-year level; regressions include two lags of the dependent variable: firm sales in t-1 and t-2; standard errors clustered at the firm level in Table 4.3.9; clustered at the country-year level in Table 4.3.10).
- Panel 2: restricted sample.
- Selected coefficient estimates and precision (exact values as reported):
  - Table 4.3.9, Panel 1 (whole sample, firm clustering):
    - 1-year: 0.4346*** (0.011)
    - 2-year: 0.4518*** (0.018)
    - 3-year: 0.4136*** (0.022)
    - 4-year: 0.2745*** (0.026)
    - 5-year: 0.2459*** (0.029)
    - Additional coefficient row: 0.0365** (0.015); -0.0017 (0.025); 0.0396 (0.032); 0.0403 (0.037); 0.0396 (0.041)
    - Observations: 33,787,483; 28,179,190; 23,418,680; 19,593,920; 16,341,735
    - R-squared: 0.2010; 0.3380; 0.4200; 0.4830; 0.532
  - Table 4.3.9, Panel 2 (restricted sample, firm clustering):
    - 1-year: 0.4635*** (0.012)
    - 2-year: 0.4246*** (0.020)
    - 3-year: 0.3645*** (0.024)
    - 4-year: 0.2223*** (0.027)
    - 5-year: 0.2459*** (0.029)
    - Additional coefficient row: 0.0466*** (0.017); -0.0053 (0.028); 0.0470 (0.035); 0.0882** (0.039); 0.0396 (0.041)
    - Observations: 16,105,295; 15,503,783; 15,380,060; 15,454,487; 16,341,735
    - R-squared: 0.2390; 0.3500; 0.4300; 0.4940; 0.532
  - Table 4.3.10 (alternative clustering at country-year level) reports very similar coefficient patterns with larger standard errors in many cells; selected values:
    - Panel 1 (whole sample): 1-year 0.4310***; 2-year 0.4461***; 3-year 0.4063***; other entries include 0.2668 (0.113), 0.2380 (0.151)
    - Panel 2 (restricted sample): 1-year 0.4591***; 2-year 0.4178**; 3-year 0.3559**; other entries include 0.2141 (0.185), 0.2380 (0.211)
    - Observations and R-squared values match Table 4.3.9 as listed above.
- Fixed effects included in all regressions: Country-Year FE = Y, Country-Sector FE = Y, Firm FE = Y.
- Significance notation: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

### Global Trade Model — BHLP model exposition
- Model type:
  - Multi-region, multi-sector general-equilibrium global network model (BHLP model following Huo, Levchenko, and Pandalai-Nayar (2019) and Bonadio and others (2021, 2023)).
  - Static global economy with N regions and J sectors; sectoral output is tradeable between regions with iceberg trade costs and balanced trade (model allows exogenous trade imbalances but that feature is not used here).
  - Labor is immobile between regions.
- Households and final consumption:
  - Household utility increasing and linear in final consumption ℱ(ω) and decreasing and concave in labor supply.
  - Final consumption: nested CES functional form across N×J region-sectors.
    - Upper nest: ℱ(ω) = [∑ζ_F_P^(1/ρ) ℱ_P(ω)^(ρ−1)/ρ]^(ρ/(ρ−1)), where ζ_F_P is a taste parameter and ρ is the elasticity of substitution.
    - Lower nest across source regions: ℱ_P(ω) = [∑μ_D,F_P^(1/γ) ℱ_D,F_P^(γ−1)/γ]^(γ/(γ−1)), where μ_D,F_P is a taste parameter and γ is the elasticity of substitution across regions.
- Firms and production:
  - Representative firm in sector j in region I with production function:
    - Y_F_P = Z_F_P [K_F_P^α_j H_F_P^(1−α_j)]^η_j λ_F_P^(1−η_j)
    - Z_F_P = region-sector TFP; K_F_P = physical capital; H_F_P = human capital; λ_F_P = aggregate of intermediate inputs.
    - Parameters α_P and η_P determine the capital share of value added and the value-added share of revenue.
  - Intermediate goods aggregate λ_F_P is a nested CES across N×J region-sectors with elasticities ε and ν in upper and lower nests and technological/taste parameters ϖ and μ_D_i_F_P.
- Trade costs:
  - Iceberg trade costs τ̃_D,F_P^f and τ̃_D_i,F_P^x for final goods and intermediates depend on source region and destination region-sector (and on source region-sector for intermediates).
- Extensive margin:
  - The extensive margin (which economy-sectors source from which others) is fixed; adjustments occur along the intensive margin only.
- Solution method:
  - Exact hat algebra approach of Dekle and others (2008) used to compute changes in region real output, region-sector value added, and sector prices in response to shocks.
  - Most model parameters enter via measurable expenditure shares; only a subset of parameters must be set explicitly: η_P, α_P, ρ, γ, ε, ν, ψ (labor supply elasticity), and μ (variance of labor productivity endowments across sectors).
  - α_P is set to 0 for all j (equipped labor parametrization, no explicit role for physical capital).
  - η_P set to match sector-level value-added share of output averaged across countries.

### Data and calibration
- Data source and coverage:
  - Uses data for the year 2018 from the OECD Inter-Country Input-Output (ICIO) Tables (OECD 2021).
  - Decomposes activity into 45 sectors in each of 66 economies: 36 advanced economies, 26 emerging market economies, 4 developing economies, and an aggregate rest of the world.
  - Aggregation across economies uses GDP at purchasing power parity.
- Calibration:
  - Calibrated parameters are drawn from medium-/long-term counterfactual literature; specific values and sources documented in Online Annex Table 4.4.1.
  - When bringing the model to the data, zeroes in the global input-output matrix are replaced by the minimum of observed positive values to avoid convergence issues (negligible impact).
- Selected calibrated parameter values (as reported in Online Annex Table 4.4.1):
  - Human capital distribution parameter: 1.5 (source: Bonadio and others (2023), taken from estimates in Galle and others (2023))
  - Intermediate goods substitution elasticity across countries: 4 (Caliendo and others (2022), which uses a value of 5)
  - Final goods substitution elasticity across countries: 4 (Caliendo and others (2022) which uses a value of 5)
  - Intermediate goods production substitution elasticity across sectors: 0.95 (Caliendo and Parro (2015), which uses Cobb-Douglas)
  - Final goods substitution elasticity across sectors: 0.95 (Caliendo and Parro (2015) which uses Cobb-Douglas)
  - Capital Share: 0 (Equipped labor model)
  - Frisch elasticity of labor supply: 1 (Bonadio and others (2023), taken from estimates in Chetty and others (2011))

### Scenarios simulated
- Four sectoral TFP shock scenarios (each simulated as described):
  1. Negative 2.5 percent TFP shock to all sectors in all G20 EMs.
  2. Negative 2.5 percent TFP shock to GVC sectors in all G20 EMs (GVC sectors defined based on GIMF analysis and Online Annex Table 4.5.1).
  3. Negative 2.5 percent TFP shock to the Construction sector in China.
  4. Positive 2.5 percent TFP shock to the IT sector in India.
- Key qualitative result for Scenario 4 (positive IT shock in India):
  - Sectoral reallocation shows significantly less heterogeneity compared to the Construction shock in China.
  - Most significant impact: reduction in value added in the IT sector outside of India, driven by advanced economies and EMs.
  - All other sectors expand, particularly energy commodities production.
  - Interpretation: expansion in India’s IT sector mainly increases competition with other countries’ IT sectors.

### Sensitivity analysis (elasticities)
- Two alternative lower-elasticity parameterizations compared with baseline:
  - Alternative A: γ = ν = 2 (lower substitutability across regions in final consumption and production).
  - Alternative B: ρ = ε = 0.5 (lower substitutability across sectors in final consumption and production).
- Impact on global GDP excluding G20 EMs (baseline vs GVC scenarios):
  - Baseline parametrization impacts reported in figure (percent scale shown).
  - With lower γ, ν (Alternative A): impact about twice as large; reported numeric examples:
    - From 0.155 (0.093) percent of GDP to 0.29 (0.16) percent of GDP in the Baseline (GVC) scenario.
  - With lower ρ, ε (Alternative B): impact increases only negligibly to 0.16 (0.095) percent of GDP in the Baseline (GVC) scenario.
- Interpretation:
  - Lower γ, ν increases the impact because output of G20 EMs is less substitutable with other countries’ output.
  - Lower ρ, ε has smaller effect because shocks are to all sectors in a particular region rather than to all regions in a particular sector.
  - While impact on global GDP outside of G20 EMs increases with lower elasticity, overall impact on global GDP decreases very slightly.
- Sectoral reallocation sensitivity:
  - With lower trade elasticity (γ, ν = 2): fewer sectors expand, and those that do expand by less.
  - With lower cross-sector elasticity (ρ, ε = 0.5): more sectors expand and those that do expand by more (higher prices and larger supply response elsewhere).

### Employment sensitivity — methodology and additional results
- Methodology to construct Table 4.1 (main text):
  - For each of 19 countries (G20 EMs and G20 AEs excluding Australia), compute elasticity of sectoral value added to productivity shocks in all sectors of other countries by solving a first-order approximation to the trade model around no-shock equilibrium.
  - Multiply that elasticity by the sectoral employment share within the affected country to get elasticity of the country’s total employment via the sector in question to all possible sectoral productivity shocks in the other 18 countries.
  - For each affected country-sector (19×45), list employment elasticities to the 18×45 other country-sector productivity shocks.
  - For complementarity table: keep three largest positive elasticities per affected country-sector; for competition table: keep three negative elasticities largest in magnitude per affected country-sector.
  - For each affected country, this yields three source country-sectors with the most important employment impact via a single sector in the affected country; counts by source group (advanced economy, emerging market) are reported in Table 4.1.
- Robustness and historical comparison:
  - Online Annex Table 4.4.2 repeats the exercise using data for 2000 instead of 2018:
    - Complementarity: in 2000 only three of the 57 affected country-sectors are most affected by shocks to sectors in G20 EMs (one in advanced economies and two in emerging markets) versus 34 when using 2018 data.
    - Competition: in 2000 six of the sectors are most affected by a sector in the G20 EMs, compared to 28 when using 2018 data.
    - Interpretation: G20 EMs had a much smaller global trade footprint in 2000 than in 2018.
  - Online Annex Table 4.4.3 reports sectors in G20 economies with the three largest positive and negative spillovers on aggregate employment (2018):
    - Results broadly similar for complementary linkages, with commodities playing a more important role for spillovers to advanced economies.
    - Competition results show a smaller role for wholesale and retail trade as a concentrated employment spillover source, despite the sector accounting for approximately 15 percent of employment.
  - Online Annex Table 4.4.4 shows results using 2000 data; reduced prominence of G20 EMs confirms their increased potential for generating spillovers between 2000 and 2018.

### Representative tabulated findings (employment-spillover counts excerpt)
- Online Annex Table 4.4.3 (Sectors in G20 Economies with the Largest Employment Spillovers, 2018) — selected entries (Number of Destination Economies Affected):
  - Complementarity panel examples:
    - EM Computer, electronic and optical equipment: 7
    - EM Textiles, wearing apparel, leather and related products: 7
    - AE Motor vehicles, trailers and semi-trailers: 5
    - AE Financial and insurance activities: 3
    - EM Mining and quarrying, energy producing products: 3
    - AE Coke and refined petroleum products: 1
  - Competition panel examples:
    - EM Computer, electronic and optical equipment: 8
    - EM Basic metals: 3
    - EM Machinery and equipment: 3
    - EM Mining and quarrying, energy producing products: 3
    - AE Financial and insurance activities: 2
    - EM Coke and refined petroleum products: 2
  - Note: Entries sum to 57 = 19 economies × 3 source sectors for each panel; AE = advanced economy; EM = emerging market.

*Source: CHAPTER 4, Online Annex (ch4onlineannex - CHAPTER 4), International Monetary Fund | April 2024.*

### 1. Complementarity

### 1. Complementarity

### Key findings on sectoral employment spillovers
- The analysis identifies source economy-sectors that generate the largest employment spillovers (positive co-movement = "complementarity"; negative co-movement = "competition"). Entries in the two columns "Number of Destination Sectors Affected" (panel 1 and panel 2) sum to 57 = 19 economies x 3 sectors.
- AE = advanced economy; EM = emerging market.
- Sources used: Bonadio and others (2021, 2023); Huo, Levchenko, and Pandalai-Nayar (2019); OECD Inter-Country Input-Output Tables; OECD Trade in Employment Database; and IMF staff calculations.
- Sample covers G20 economies, excluding Australia; regional aggregates for Asia and Pacific, Middle East and Central Asia, Europe, and Western Hemisphere; and a rest of the world aggregate.

- Representative entries (complementarity, number of destination sectors affected) reported in the source include:
  - AE Motor vehicles, trailers and semi-trailers: 9
  - AE Computer, electronic and optical equipment: 6
  - AE Financial and insurance activities: 7
  - AE Wholesale and retail trade: 18
  - AE Agriculture, hunting, forestry: 1
  - AE Machinery and equipment: 1
  - AE Basic metals: 1
  - AE Professional, scientific and technical activities: 1
  - EM Computer, electronic and optical equipment: 1
  - EM Wholesale and retail trade: 1
  - EM Mining and quarrying, energy producing products: 1
  - EM Mining and quarrying, non-energy producing products: 1

- Representative entries (competition, number of destination sectors affected) reported in the source include:
  - AE Computer, electronic and optical equipment: 10
  - AE Motor vehicles, trailers and semi-trailers: 4
  - AE Financial and insurance activities: 3
  - AE Wholesale and retail trade: 7
  - AE Machinery and equipment: 3
  - AE Textiles, textile products, leather and footwear: 3
  - AE Manufacturing; repair and installation of machinery: 2
  - AE Education: 2
  - AE Food products, beverages and tobacco: 1
  - AE Agriculture, hunting, forestry: 8
  - EM Agriculture, hunting, forestry: 3
  - EM Motor vehicles, trailers and semi-trailers: 1
  - EM Mining and quarrying, energy producing products: 1
  - EM Mining and quarrying, non-energy producing products: 1

- For the metric "Number of Destination Economies Affected" (panel 1 and panel 2), the entries also sum to 57 = 19 economies x 3 source sectors. Representative entries include:
  - Complementarity (by economy affected count): AE Computer, electronic and optical equipment: 10; AE Motor vehicles, trailers and semi-trailers: 8; AE Financial and insurance activities: 4; EM Mining and quarrying, energy producing products: 2.
  - Competition (by economy affected count): AE Computer, electronic and optical equipment: 9; AE Motor vehicles, trailers and semi-trailers: 6; AE Financial and insurance activities: 4; AEChemical and chemical products: 3; AEMachinery and equipment: 3; AEWholesale and retail trade: 3.

### Sector numbering reference (ICIO codes)
- The source provides the ICIO code-to-sector mapping used in the analysis (Online Annex Table 4.4.5). Selected mappings include:
  - 01T02 Agriculture, hunting, forestry
  - 05T06 Mining and quarrying, energy producing products
  - 07T08 Mining and quarrying, non-energy producing products
  - 10T12 Food products, beverages and tobacco
  - 13T15 Textiles, textile products, leather and footwear
  - 15 Basic metals
  - 17T18 Paper products and printing
  - 26 Computer, electronic and optical equipment
  - 29 Motor vehicles, trailers and semi-trailers
  - 31T33 Manufacturing nec; repair and installation of equipment
  - 36T39 Water supply; sewage, waste management and remediation activities
  - 45T47 Wholesale and retail trade
  - 55T56 Accommodation and food service activities
  - 64T66 Financial and insurance activities
  - 85 Education
  - 86T88 Human health and social work activities
  - 94T96 Other service activities

### Methodological note
- Computation uses the contribution to total employment from each economy-sector’s response to all possible positive productivity shocks from the source economy-sector (panel 1) and negative co-movements (panel 2). The three source economy-sectors driving the top three sector responses by economy are summarized.

---

### Modeling the G20 EM Upside Scenario (GIMF summary and calibration)

### Model structure (GIMF, with GVC sector)
- GIMF is an annual, multi-region, micro-founded dynamic stochastic general equilibrium model (DSGE).
- For this chapter, GIMF comprises 10 regions: United States, EU+, other advanced economies, China, India, Southeast Asia, Middle East and CIS (also referred to as “oil exporters”), Mexico, other Latin America, and the other EMDEs.
- A tradable sector related to global value chains (GVC) was added ("the GVC sector").
- Households: some are non-Ricardian, finitely lived overlapping generations; others are liquidity constrained (consume all income every period).
- Firms: profit-maximizing, monopolistic competition, produce non-tradable, tradable, and GVC goods.
- Investment financing follows a Bernanke, Gertler and Gilchrist (1999) financial accelerator; firms borrow from financial intermediaries and face endogenous corporate risk premia.
- GVC sector features roundabout production: produced GVC goods are inputs into final goods and into other GVC goods, with both imported (1) and domestically sourced (2) inputs.
- Regions trade final goods (consumption and investment), and GVC and non-GVC tradable intermediate goods. Flows are tracked bilaterally; trade reacts to demand, supply, terms of trade, and bilateral real exchange rates.
- Monetary policy: inflation forecast-based targeting. Fiscal policy: debt-GDP ratio targeting.

### Calibration highlights
- Calibration based on OECD Inter-Country Input-Output Tables for 2018 and national accounts and fiscal ratios.
- Intertemporal elasticity of substitution (consumption) = 0.2 (common across regions).
- Share of liquidity constrained households:
  - 25 percent for United States, EU+, other advanced economies, and China.
  - 50 percent for the other EM regions.
- Trade and production elasticities (Online Annex Table 4.5.3, All Regions):
  - Capital-Labor / GVC: (dash in source for many entries; preserved layout indicates capital-labor / GVC elasticities not listed as numeric in that column)
  - Domestic / Imported elasticities:
    - Consumption: 1.5
    - Investment: 1.5
    - Non-GVC Tradables: 1.5
    - GVC Tradables: 0.5
  - Different Regions elasticities:
    - Consumption: 1.5
    - Investment: 1.5
    - Non-GVC Tradables: 1.5
    - GVC Tradables: 0.8
  - Note: * Elasticity between domestic and imported when using GVC tradables in the production of final goods or of other GVC tradables is 0.8 (as annotated in the source).
- For final consumption and investment goods:
  - Elasticity of substitution between nontradable goods and tradable goods bundle = 0.5.
  - Elasticity for tradable goods bundle (assembled from non-GVC and GVC tradable intermediate goods) = 0.95.

### Domestic sector calibration (percent of region's GDP, Online Annex Table 4.5.2)
- Share of Global GDP (%, US$):
  - United States: 24.4
  - European Union +: 18.9
  - Other Advanced Economies: 8.7
  - China: 16.7
  - India: 3.2
  - Southeast Asia: 2.3
  - Middle East and CIS: 3.2
  - Mexico: 1.2
  - Other Latin America: 2.3
  - Rest of the World: 16.7

- Domestic Demand (percent of region's GDP):
  - Household Consumption:
    - United States: 65.4
    - European Union +: 54.9
    - Other Advanced Economies: 56.3
    - China: 51.7
    - India: 56.7
    - Southeast Asia: 57.9
    - Middle East and CIS: 52.2
    - Mexico: 63.0
    - Other Latin America: 63.0
    - Rest of the World: 61.2
  - Private Investment:
    - United States: 17.3
    - European Union +: 12.3
    - Other Advanced Economies: 11.9
    - China: 18.5
    - India: 26.8
    - Southeast Asia: 26.4
    - Middle East and CIS: 23.0
    - Mexico: 22.0
    - Other Latin America: 17.1
    - Rest of the World: 19.0

- Trade (Aggregate Exports and Imports, and component shares) — selected entries:
  - Aggregate Exports:
    - United States: 11.4
    - European Union +: 20.1
    - Other Advanced Economies: 23.4
    - China: 17.4
    - India: 20.1
    - Southeast Asia: 44.2
    - Middle East and CIS: 34.5
    - Mexico: 36.3
    - Other Latin America: 16.2
    - Rest of the World: 25.5
  - GVC Tradable (exports as % of GDP):
    - United States: 3.8
    - European Union +: 7.1
    - Other Advanced Economies: 10.5
    - China: 7.5
    - India: 5.8
    - Southeast Asia: 8.0
    - Middle East and CIS: 17.0
    - Mexico: 22.7
    - Other Latin America: 16.3
    - Rest of the World: 7.2
  - Aggregate Imports (shares mirror Aggregate Exports in the table):
    - Aggregate Imports:
      - United States: 11.4
      - European Union +: 20.1
      - Other Advanced Economies: 23.4
      - China: 17.4
      - India: 20.1
      - Southeast Asia: 44.2
      - Middle East and CIS: 34.5
      - Mexico: 36.3
      - Other Latin America: 16.2
      - Rest of the World: 25.5
  - GVC Tradable (imports as % of GDP):
    - United States: 5.4
    - European Union +: 7.0
    - Other Advanced Economies: 9.8
    - China: 9.8
    - India: 12.5
    - Southeast Asia: 21.2
    - Middle East and CIS: 7.4
    - Mexico: 17.8
    - Other Latin America: 6.2
    - Rest of the World: 8.9

---

### G20 EM Upside Scenario: design and quantitative impacts

### Scenario design and calibration
- The G20 EM (excluding China) upside scenario is based on shocks to short-term aggregate demand (household consumption and private investment) in the G20 EMs excluding China, over a period of five years (the forecast horizon).
- The size of the shock represents a likely upside to the current WEO baseline: there is a 30 percent probability that growth in each EM ex-China could be higher than envisaged in this scenario, or alternatively 70 percent probability that growth is lower than envisaged in this scenario.
- For regions that aggregate G20 EMs with non-G20 EMs (Middle East and CIS; other Latin America; Southeast Asia; other EMDEs), each G20 EM’s shock profile is scaled according to their GDP share within the region to construct the aggregated region shock profile.

### Quantitative impacts (70th percentile of upside scenario — deviation of real GDP growth from WEO baseline, Online Annex Table 4.5.4)
- Year 3 (2026) / Year 5 (2028) deviations:
  - All EM Regions: 0.7 / 0.9
  - India: 0.7 / 1.0
  - Southeast Asia: 0.5 / 0.6
  - Middle East and CIS: 1.3 / 1.5
  - Mexico: 2.5 / 3.0
  - Other Latin America: 1.1 / 1.4
  - Other EMDEs: 0.2 / 0.3

- Regional fifth-year impacts highlighted in the source:
  - Maximum fifth-year impacts on real GDP growth range from 0.3 percentage points for the other EMDEs to 1 percentage points for India, 1.5 percentage point for the Middle East and CIS, and 3 percentage points for Mexico.
- Potential aggregate effect:
  - This G20 EM upside could push up global growth, on average, by at most 0.5 percentage points.
- Spillovers to advanced economies and China:
  - EM spillovers on advanced economies and China would be generally positive but small from the upside scenario, primarily through GVCs.
  - Both advanced economies and China growth could be up to 0.1 percent higher on average in the first three years, but then the effect would subside.
  - The other G20 EMs shocks would also generate spillovers among the EMDE regions, accounting for about 15 percent of the impact of the upside scenario.

*Source: ch4onlineannex - 1. Complementarity (Online annex material as provided).*

### 2026. The EM spillovers on EMDEs is larger

### 2026. The EM spillovers on EMDEs is larger

### Key finding
- The EM spillovers on EMDEs is larger than those on advanced economies (Online Annex Figure 4.5.2), since the upside scenario is based on higher G20 EM aggregate demand for final goods, of which EMDEs produce more than advanced economies.

### Contributors and transmission mechanisms
- Largest contributors to spillovers vary between advanced economies and EMDEs and reflect both the likelihood of larger shocks in specific G20 EMs and domestic characteristics of those EMs.
- For advanced economies, the largest spillovers after 3 and 5 years would be from:
  - Middle East and CIS (hydrocarbon exports)
  - Mexico (extensive two-way trade relationship with the United States)
  - Southeast Asia (significant trade with China and advanced economies)
- For EMDEs as a group, the largest spillovers after 3 and 5 years would be from:
  - India (supplier to other EMDEs of intermediate goods through GVCs, and final consumption and investment goods, and major user of their raw materials)
  - Middle East and CIS (hydrocarbon exports)
  - Southeast Asia (supplier of final consumption goods)

### Quantitative signals and horizons
- Impacts are reported as percentage point deviations from the WEO baseline on the level of real GDP.
- Time horizons presented: Impact after 3 years; Impact after 5 years.
- Figure axis values shown: –0.05, 0.00, 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35.

### Notes and definitions
- Abbreviations as used in the figure and text:
  - AEs = advanced economies
  - EMs = emerging markets
  - EMDEs = emerging markets and developing economies
  - CIS = Commonwealth of Independent States

*Source: IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/weo/2024/april/english/ch4onlineannex.pdf_