## wpiea2025027-print-pdf

## Source details

**Canonical URL:** [wpiea2025027-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025027-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025027-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025027-print-pdf.pdf.json)

---

### Introduction — global context and motivation
- Rise of the G20−EMs (Argentina, Brazil, China, India, Indonesia, Mexico, Russia, Saudi Arabia, South Africa, Turkey) reshaped the global economy over the last two decades.
- G20−EMs: average annual growth rate of nearly 6 percent; contributions to approximately one-third of global economic activity and one-quarter of global trade.
- Prior literature concentrated on China; IMF projects medium-term growth rates of 6.5 percent for India.
- Need for cross-country comparative analysis to assess growth spillovers across trade, commodities, and financial flows.

### Research objective and contribution
- Investigates global diffusion of domestic growth shocks originating from the G20−EMs using a Bayesian Global Vector Autoregression (GVAR) model.
- Contributions:
  - Extension of geographical and temporal scope: panel of 63 countries (text reports both "63 countries, of which 30 are emerging economies" and "63 countries, including 34 advanced economies and 29 emerging economies").
  - Evaluation of spillovers across direct trade linkages, regional effects, commodity price fluctuation, and financial integration.
  - Estimation of both demand-side and supply-side shocks within the G20−EMs.
  - Evidence that spillovers from China are relatively larger; broader examination across other emerging economies.

### Model framework and key methodological choices
- Model: Bayesian GVAR building on Pesaran et al. (2004); incorporates Bayesian estimation and stochastic volatility (Feldkircher and Huber, 2016).
- Sample and data:
  - Quarterly data from 2001Q1 to 2023Q2.
  - 63 countries covered.
  - Endogenous variables in country models: real GDP, CPI inflation, the real exchange rate, the short-term rate, and the long-term rate.
  - Country-specific foreign variables: cross-sectional averages using fixed bilateral trade weights matrix {wij} (share of total trade of country i with country j normalized to sum to unity).
  - Data sources: IMF databases, including Direction of Trade Statistics (DOTS), and national authorities.
- Commodity block:
  - Dedicated oil market block: oil prices, production levels, oil inventories, and a measure of global demand.
  - Allows oil markets to influence all country models empirically rather than treating oil as endogenous to a single country.

### Bayesian estimation, priors, and model selection
- Priors tested: Minnesota (MN), Normal-Gamma (NG), and stochastic search variable selection (SSVS).
  - MN prior: nudges variables toward unconditional stationary mean or at least one unit root; minimal hyperparameters; uniform shrinkage.
  - NG prior: global and local shrinkage; tighter shrinkage for higher-lag coefficients.
  - SSVS prior: 'spike and slab' mixture-normal prior with specification Ξi,j | δi,j ∼ δi,j N(0, τi,0^2) + (1−δi,j) N(0, τi,1^2), δi,j ∼ Bernoulli(p); chosen τi,0 = 0.01, τi,1 = 3, prior inclusion probability = 0.5.
- Diagnostics favor the SSVS prior as optimal; modest serial correlation in most equations’ residuals and minimal cross-unit correlation of posterior median residuals.

### Stochastic volatility, estimation details, and posterior inference
- Model accommodates stochastic volatility; structural analysis uses variance-covariance matrix with median volatility on its diagonal.
- Posterior simulation:
  - 30,000 draws following a 10,000 iteration burn-in phase, with a thinning factor of 10.
  - Interpretation focuses on median impulse response values as primary reference.

### Structural identification and shock classification
- Generalized impulse response functions (GIRFs) used for invariance to variable ordering.
- Sign restrictions to distinguish demand and supply shocks:
  - Aggregate demand shock: output (y) ↑, price change (Δp) ↑.
  - Aggregate supply shock: output (y) ↑, price change (Δp) ↓.
  - Restrictions imposed as ≥/≤ over four quarters (binding for 4 quarters).

### Generalized Impulse Response Analysis (Section 4.1) — aggregate and country-level spillovers
- Aggregate spillover effects from G20−EMs primarily attributed to China.
- China’s aggregate spillovers (reported magnitudes):
  - To AEs: 0.65 percent
  - To EMs: 0.88 percent
  - To MCD region: 1.6 percent (based on a sample of only four countries)
  - To Asia (excluding MCD): 0.9 percent
  - To Europe: approximately 0.7 percent
  - To the Western Hemisphere: approximately 0.7 percent
- Spillovers from other G20−EMs: remain below 0.1 percent on average.
- Notable bilateral and country-level spillovers (selected reported values):
  - Mexico to AEs: 0.11 percent
  - India and Brazil to other EMs: nearing 0.1 percent
  - Russia, Indonesia, Saudi Arabia: around 0.05 percent
  - Turkey, South Africa, Argentina: below 0.025 percent
  - China to Hong Kong and Singapore: both exceeding a magnitude of one
  - China to Japan: exceptionally large among AEs
  - Indonesia to Singapore: approximately 0.36
  - India to Singapore: approximately 0.17
  - India to Saudi Arabia: 0.36
  - Brazil to Argentina: 1.1
  - Brazil to Chile: 0.25
  - Mexico to Argentina: 0.18; to Colombia: 0.12; to Chile: 0.09
  - Mexico to US and Canada: around 0.19
  - Russia to Kazakhstan: 0.51; to Belarus: 0.47; to Moldova: 0.46
  - Turkey to Greece and several Eastern European countries: 0.1 to 0.23
  - Saudi Arabia maximum bilateral spillovers: around 0.11 percent
  - South Africa to UK: 0.03; to Singapore and Germany: 0.025 each; to Saudi Arabia: 0.07; to Argentina: 0.05

### Structural Shocks: Demand versus Supply (Section 4.2)
- Shocks normalized to trigger a 1 percent increase in source country output over 1 year.
- One-year horizon GDP responses:
  - China’s demand-side spillovers:
    - To EMs: 0.39 percent
    - To AEs: 0.24 percent
  - China’s supply-side spillovers:
    - To EMs: 0.27 percent
    - To AEs: 0.16 percent
  - Other G20−EMs:
    - Demand spillovers to EMs: 0.01 percent
    - Supply spillovers to EMs: 0.04 percent
    - Demand spillovers to AEs: 0.02 percent
    - Supply spillovers to AEs: 0.015 percent
  - US spillovers (for comparison):
    - To AEs: demand 0.23, supply 0.18
    - To EMs: demand 0.27, supply 0.35
- Medium-term (1–3 year) GDP responses:
  - China’s demand shocks to EMs: 0.8
  - China’s supply shocks to EMs: 0.39
  - Other G20−EMs to EMs: demand 0.04, supply 0.05
  - China’s demand shocks to AEs: 0.46
  - China’s supply shocks to AEs: 0.4
  - Other G20−EMs to AEs: demand 0.04, supply 0.05
- Country-level notes: most spillovers from other G20−EMs are below 0.05, except Brazil, Mexico, Turkey, Russia, India, Indonesia (primarily due to supply-side shocks); Russia shows significant demand spillovers to AEs (pre-war trade ties with Europe).
- Inflation responses (one-year horizon):
  - China’s positive demand shocks raise inflation by:
    - Advanced economies: approximately 0.15 percentage points
    - Emerging markets: approximately 0.2 percentage points
  - China’s supply shocks reduce inflation by:
    - Advanced economies: -0.1 percentage points
    - Emerging markets: -0.2 percentage points
  - Spillovers from other G20−EMs to inflation: minimal for both shock types

### Regional Spillovers (Section 4.3)
- Regional aggregates and median impulse responses at 1-year horizon (68% threshold).
- China’s regional spillovers (1-year horizon, medians reported):
  - Asia: demand 0.32 percent; supply 0.36 percent
  - Middle East and Central Asia: approximately 0.15 percent (each for demand and supply); China’s impact on Saudi Arabia through oil markets: approximately 0.55
  - Europe: demand 0.26 percent; supply 0.07 percent
  - Western Hemisphere: 0.35 and 0.30 (demand and supply)
- Other large EMs by region:
  - Asia: Indonesia medians up to 0.05 percent; India medians up to 0.02 percent
  - Europe: Russia demand-side impacts 0.10 percent; supply-side 0.02 percent
  - Western Hemisphere: Brazil 0.18 and Mexico 0.10 (largest regional spillovers after China)
  - Turkey: larger spillovers to European countries predominantly through demand channels (0.03 percent)
- Interpretation: China’s spillovers have global reach; other EMs tend to have more pronounced regional impacts.

### Spillovers to Commodity Prices (Section 4.4)
- Oil price responses to a 1pp per year positive domestic aggregate demand (AD) and aggregate supply (AS) shocks:
  - China AD shocks: approximately 2.2 percent increase in oil prices after one year
  - China AS shocks: approximately 1.6 percent increase in oil prices after one year
  - Three-year horizon for China: AD 3.1 percent; AS 2.5 percent (effects significant only for China during this period)
  - Other G20−EMs: minimal oil price spillovers around 0.25 percent
  - India AD shocks: around 1 percent impact on oil prices at one year
  - US shocks (for context): 2.5 percent increase in oil prices after one year
- Oil production responses:
  - China demand and supply shocks increase oil production by approximately 0.5 to 0.6 percent after one year
  - Reactions to shocks from other G20−EMs are significantly smaller
- Metal prices:
  - China’s AD and AS shocks increment metal prices by about 0.25-0.3 percent after one year
  - Other G20−EMs show little to no similar effect
- Conclusion: commodity price channel is particularly potent for China.

### Changes in the Size of Spillovers over Time (Section 4.5)
- Baseline trade weights: 2017 to 2019; robustness includes earlier trade-weight comparisons (2001–2003).
- China’s trade integration evolution:
  - Contribution to global merchandise exports: from approximately 4% to over 14%
  - Import share: from around 4% to 11%
  - Share of global manufacturing gross production: 35% (and 29% of value-added)
  - China is the largest export destination for 33 countries and the primary source of imports for 65 countries
- Temporal findings:
  - Spillovers from China have roughly doubled over the last twenty years for both demand and supply fluctuations.
  - Example magnitudes (advanced economies):
    - Aggregate supply shocks: from 0.19 to 0.4
    - Aggregate demand shocks: from 0.25 to 0.46
  - Example magnitudes (emerging economies):
    - Supply shocks: from 0.15 to 0.39
    - Demand shocks: from 0.23 to [text truncated in source; described as rising markedly]
- Reported summary statistics and headline magnitudes:
  - A one percentage point uptick in China’s GDP growth elicits:
    - an approximate increase in the output of advanced economies by 0.2−0.4 percentage points;
    - an approximate increase in the output of other emerging economies by 0.3−0.8 percentage points.
  - Demand shocks from China generate larger spillovers than supply shocks:
    - 0.3 percent versus 0.2 percent at the one-year horizon on global output;
    - three-year cumulative effects: 0.6 percent from demand-side shocks versus 0.4 percent from supply-side shocks on world output.
  - Paper’s GIRFs indicate a global impact of about 0.75 percent at the one-year horizon.
  - Additional reported numeric tag: "0.79 for demand shocks."

### Major findings on spillovers (summary)
- China’s spillovers are large, widespread, and have increased substantially over the last two decades.
- Commodity exporters in emerging Asia and Latin America, especially oil-exporting nations, experience significant spillovers from China.
- Spillovers from other G20−EMs are generally modest (typically below 0.1 percentage points), though certain commodity exporters (e.g., Russia, Saudi Arabia) can trigger larger regional effects.
- Demand-side transmission channels from China dominate supply-side channels in magnitude.

### Robustness, identification, and methodological insights
- Robustness checks:
  - Bayesian GVAR covering 63 countries; priors MN, SSVS, NG; lag lengths from one to four lags; models with and without time trends and stochastic volatility; alternative sign-restriction durations; traditional non-Bayesian GVAR on smaller sets.
- Robustness conclusions: core conclusions—particularly China dominance and stronger demand shocks—are robust across specifications, though quantitative magnitudes vary.
- Geweke convergence diagnostic: out of 218460 variables, 25169 (11.52%) have Z-values exceeding the 1.96 threshold.
- Posterior median residuals indicate acceptably low cross-country residual correlation for structural analysis.

### Data, sample, and diagnostics (detailed)
- Quarterly observations from 2001Q1 to 2023Q2.
- 63 countries included.
- Baseline weight matrix: annual bilateral trade flows averaged over 2017 to 2019.
- Variables: real GDP (log), consumer price inflation (difference of log CPI), short-term interest rates (0.25×log(1+Rit/100)), long-term interest rates (similar transform for 10-year yields), real exchange rate (log, deflated using CPI), equity prices (log nominal equity price index deflated by CPI), oil price (log nominal price of oil in US dollars), oil production (log mil. Barrels/Day), oil inventories (log forward consumption in OECD in days), global activity proxy (global real economic activity index in industrial countries).
- Country coverage highlights:
  - Europe: majority of sample (26 advanced, 10 emerging).
  - Asia-Pacific: 13 countries (6 advanced, 7 emerging).
  - Western Hemisphere: 9 countries (2 advanced, 7 emerging).
  - Middle East and Central Asia: 5 emerging economies.
  - Africa: limited coverage.

### Policy implications and conclusions
- Policymakers should monitor China’s demand and supply conditions closely given substantial estimated spillovers to both advanced and emerging economies.
- Key recommendations:
  - Remain vigilant to developments in China’s demand and supply conditions.
  - Monitor commodity markets and GVC linkages closely, given their role in transmission.
  - Use detailed country-level GVAR insights to tailor regional and country-specific responses rather than relying on aggregate EM measures alone.
- The potential for broad adverse global ramifications from a Chinese economic deceleration is high and unlikely to be fully offset by other emerging markets in the foreseeable future.

*Source: IMF Working Papers — Spillovers from Large Emerging Economies: How Dominant Is China? Working Paper No. WP/2025/027*

### 1. Introduction

### 1. Introduction

### Global context and motivation
- Over the last two decades, the global economic landscape has been reshaped by the rise of the G20 large emerging economies (G20−EMs): Argentina, Brazil, China, India, Indonesia, Mexico, Russia, Saudi Arabia, South Africa, Turkey.
- The G20−EMs have experienced an average annual growth rate of nearly 6 percent, increasing their contributions to approximately one-third of global economic activity and one-quarter of global trade.
- Much prior research has concentrated on China given its integration into the global economy; studies document China's large impacts on global trade patterns, commodity markets, and global value chains.
- The emphasis on China has left less attention on other G20−EMs, which could play a larger role if China’s growth moderates and other G20−EMs (e.g., India) sustain strong growth—IMF projects medium-term growth rates of 6.5 percent for India.
- A cross-country comparative analysis is therefore important to assess growth spillovers from all G20−EMs across multiple channels (trade, commodities, financial flows).

### Research objective and contribution
- This paper investigates global diffusion of domestic growth shocks originating from the G20−EMs using a Bayesian Global Vector Autoregression (GVAR) model.
- Contributions emphasized:
  - Extension of geographical and temporal scope by encompassing 63 countries (text reports both "63 countries, of which 30 are emerging economies" and "63 countries, including 34 advanced economies and 29 emerging economies").
  - Evaluation of spillovers across all pertinent channels: direct trade linkages, regional effects, commodity price fluctuation, and financial integration.
  - Estimation of both demand-side and supply-side shocks within the G20−EMs to allow nuanced structural transmission analysis.
  - Evidence that spillovers from China are relatively larger, while broadening examination of spillover avenues to a wider set of emerging economies.

### Model framework and key methodological choices
- Model used: Bayesian GVAR building on Pesaran et al. (2004) and incorporating Bayesian estimation techniques to improve estimation over shorter horizons and handle stochastic volatility (Feldkircher and Huber, 2016).
- Sample and data:
  - Quarterly data from 2001Q1 to 2023Q2.
  - Panel of 63 countries covering a mix of advanced and emerging economies (see the reported counts above).
  - Endogenous variables in country models: real GDP, CPI inflation, the real exchange rate, the short-term rate, and the long-term rate.
  - Country-specific foreign variables constructed as cross-sectional averages of other countries’ domestic variables using a fixed bilateral trade weights matrix {wij} based on bilateral trade weights (computed as the share of total trade of country i with country j as a share of total trade of country i, normalized to sum to unity).
  - Data sources: IMF databases, including Direction of Trade Statistics (DOTS), and national authorities.
- Commodity block:
  - Dedicated commodity block captures oil market dynamics: oil prices, production levels, oil inventories, and a measure of global demand.
  - Commodity modeling contrasts with approaches that treat oil prices as endogenous to a single country (often the US); the framework allows oil markets to influence all country models empirically.

### Bayesian estimation, priors, and model selection
- Bayesian GVAR advantages noted: handles stochastic volatility and mitigates overfitting despite large dataset; yields enhanced forecasting precision relative to traditional cointegrated VARs.
- Priors tested: Minnesota (MN), Normal-Gamma (NG), and stochastic search variable selection (SSVS).
  - MN prior: nudges variables toward unconditional stationary mean or at least one unit root; minimal hyperparameters; uniform shrinkage across country VARs.
  - NG prior: introduces global and local shrinkage; tightens coefficients toward zero unless evidence suggests otherwise; favors stronger shrinkage for higher-lag coefficients.
  - SSVS prior: flexible, accounts for model uncertainty, allows more country-specificity via a stacked coefficient matrix Ξi and 'spike and slab' mixture-normal prior on each coefficient.
    - Specification provided: Ξi,j | δi,j ∼ δi,j N(0, τi,0^2) + (1−δi,j) N(0, τi,1^2), δi,j ∼ Bernoulli(p).
    - Prior variances semi-automatically scaled with OLS standard errors from the full model.
    - Chosen parameters: τi,0 = 0.01, τi,1 = 3, prior inclusion probability = 0.5.
- Diagnostics and selection:
  - Models yield closely comparable results; diagnostics favor the SSVS prior as optimal.
  - Diagnostics reveal modest serial correlation in most equations’ residuals and minimal cross-unit correlation of posterior median residuals—supporting the GVAR assumption of weak cross-sectional correlation of idiosyncratic shocks.
  - SSVS aligns with prior findings of superior out-of-sample predictive performance and better accommodation of country-specific nuances.

### Stochastic volatility, estimation details, and posterior inference
- Model accommodates stochastic volatility to allow residual variances to vary over time—important due to half the sample being emerging economies and the sample period containing significant shocks.
- Structural analysis uses the variance-covariance matrix with median volatility (over the sample period) on its diagonal.
- Posterior simulation details:
  - Posterior results based on 30,000 draws following a 10,000 iteration burn-in phase, with a thinning factor of 10.
  - Interpretation focuses on the model yielding median impulse response values as the primary reference (following Fry and Pagan, 2011).

### Structural identification and shock classification
- Impulse-response identification strategy:
  - Primary use of generalized impulse response functions (GIRFs) for consistency with the GVAR literature; GIRFs are invariant to variable ordering and neutralize concurrent shocks through the observed shock distribution.
  - GIRFs interpreted as encompassing generic domestic growth shocks that encapsulate both demand and supply-side factors.
- Structural identification via sign restrictions:
  - Sign restrictions imposed locally within each country model to distinguish demand and supply shocks, following Feldkircher and Huber (2016); Eickmeier et al. (2015); Dees et al. (2007).
  - Restrictions grounded in aggregate demand-and-supply frameworks and aligned with dynamic stochastic general equilibrium models.
  - Identification rule:
    - Aggregate demand shock: output (y) ↑, price change (Δp) ↑.
    - Aggregate supply shock: output (y) ↑, price change (Δp) ↓.
  - Constraints on output (y) and price dynamics (Δp) are binding for 4 quarters (restrictions imposed as ≥/≤ over four quarters) to enhance differentiation and exclude transient comovements.
  - Note: some literature applies different durations (e.g., Feldkircher and Huber (2016) used restrictions binding for the first 10 quarters for U.S. aggregate supply shocks).

### Main empirical emphasis and findings summarized in the introduction
- The analysis assesses spillovers from the G20−EMs across trade, regional spillovers, commodity-price channels, and financial integration.
- Findings highlighted:
  - Relatively larger spillovers from China are corroborated and expanded by examining all spillover avenues across a broader set of emerging economies.
  - The commodity price channel (via the dedicated oil market block) is emphasized as an important transmission mechanism.
  - Estimation of both demand-side and supply-side shocks within G20−EMs provides a more nuanced understanding of structural transmission of external shocks.

*Source: wpiea2025027-print-pdf - 1. Introduction*

### 4. Empirical Results

### 4. Empirical Results

### 4.1. Generalized Impulse Response Analysis
- Aggregate spillover effects from G20−EMs to both advanced economies (AEs) and emerging markets (EMs) are primarily attributed to China.
- China’s spillovers:
  - To AEs: 0.65 percent
  - To EMs: 0.88 percent
  - To MCD region: 1.6 percent (based on a sample of only four countries)
  - To Asia (excluding MCD): 0.9 percent
  - To Europe: approximately 0.7 percent
  - To the Western Hemisphere: approximately 0.7 percent
- Spillovers from other G20−EMs remain below 0.1 percent on average.
- Notable country-level spillovers:
  - Mexico to AEs: 0.11 percent
  - India and Brazil to other EMs: nearing 0.1 percent
  - Russia, Indonesia, Saudi Arabia: around 0.05 percent
  - Turkey, South Africa, Argentina: below 0.025 percent
- Large bilateral and regional patterns:
  - China to Hong Kong and Singapore: both exceeding a magnitude of one
  - China to Japan: exceptionally large among AEs
  - China to commodity-producing EMs (Argentina, Saudi Arabia, Russia, Kazakhstan, Brazil, Chile): pronounced
  - China to GVC-integrated Asian countries (Thailand, Philippines): significant
  - Indonesia to Singapore: approximately 0.36
  - India to Singapore: approximately 0.17
  - India to Saudi Arabia: 0.36
  - India to Argentina and South Africa: around 0.18 each
  - Indonesia to regional EMs (Thailand, Philippines, Malaysia): 0.17 to 0.14
  - Brazil to Argentina: 1.1 (Brazil’s spillovers to Argentina)
  - Brazil to Chile: 0.25
  - Mexico to Argentina: 0.18; to Colombia: 0.12; to Chile: 0.09
  - Mexico to US and Canada: around 0.19
  - Brazil to Portugal: 0.06
  - Russia to Lithuania: 0.25; to Finland: 0.15; to Greece: 0.15
  - Russia to Kazakhstan: 0.51; to Belarus: 0.47; to Moldova: 0.46
  - Turkey to Greece and several Eastern European countries: 0.1 to 0.23
  - Saudi Arabia maximum bilateral spillovers: around 0.11 percent
  - South Africa to UK: 0.03; to Singapore and Germany: 0.025 each; to Saudi Arabia: 0.07; to Argentina: 0.05

### 4.2. Structural Shocks: Demand versus Supply
- Shocks normalized to trigger a 1 percent increase in source country output over 1 year.
- At the 1-year horizon (GDP responses):
  - China’s demand-side spillovers:
    - To EMs: 0.39 percent
    - To AEs: 0.24 percent
  - China’s supply-side spillovers:
    - To EMs: 0.27 percent
    - To AEs: 0.16 percent
  - Other G20−EMs:
    - Demand spillovers to EMs: 0.01 percent
    - Supply spillovers to EMs: 0.04 percent
    - Demand spillovers to AEs: 0.02 percent
    - Supply spillovers to AEs: 0.015 percent
  - US spillovers (for comparison):
    - To AEs: demand 0.23, supply 0.18
    - To EMs: demand 0.27, supply 0.35
- Medium-term (1–3 year) GDP responses:
  - China’s demand shocks to EMs: 0.8
  - China’s supply shocks to EMs: 0.39
  - Other G20−EMs to EMs: demand 0.04, supply 0.05
  - China’s demand shocks to AEs: 0.46
  - China’s supply shocks to AEs: 0.4
  - Other G20−EMs to AEs: demand 0.04, supply 0.05
- Country-level observations:
  - Most spillovers from other G20−EMs are below 0.05, except Brazil, Mexico, Turkey, Russia, India, Indonesia (primarily due to supply-side shocks).
  - Russia shows significant demand spillovers to AEs (pre-war trade ties with Europe).
- Inflation responses (one-year horizon):
  - China’s positive demand shocks raise inflation by:
    - Advanced economies: approximately 0.15 percentage points
    - Emerging markets: approximately 0.2 percentage points
  - China’s supply shocks reduce inflation by:
    - Advanced economies: -0.1 percentage points
    - Emerging markets: -0.2 percentage points
  - Spillovers from other G20−EMs to inflation: minimal for both shock types

### 4.3. Regional Spillovers
- Method: regional aggregates for Europe, Asia, Western Hemisphere, Middle East and Central Asia (MCD); median impulse responses at 1-year horizon, with interquartile ranges, based on significant responses (68% threshold).
- China’s regional spillovers (demand and supply, 1-year horizon):
  - Asia: demand 0.32 percent; supply 0.36 percent
  - Middle East and Central Asia: approximately 0.15 percent (each for demand and supply); China’s impact on Saudi Arabia through oil markets: approximately 0.55
  - Europe: demand 0.26 percent; supply 0.07 percent
  - Western Hemisphere: 0.35 and 0.30 (demand and supply, driven by US and Latin American commodity exporters)
- Other large EMs by region:
  - Asia: Indonesia and India medians up to 0.05 and 0.02 percent, respectively
  - Europe: Russia demand-side impacts 0.10 percent; supply-side 0.02 percent
  - Western Hemisphere: Brazil 0.18 and Mexico 0.10 (largest regional spillovers after China), with Brazil linked to commodity exports and Mexico to GVC integration with the United States
  - Turkey: larger spillovers to European countries predominantly through demand channels (0.03 percent)
- Interpretation:
  - China’s spillovers have global reach; other EMs tend to have more pronounced impacts within their geographical areas.

### 4.4. Spillovers to Commodity Prices
- Oil price responses to a 1pp per year positive domestic aggregate demand (AD) and aggregate supply (AS) shocks:
  - China aggregate demand shocks: approximately 2.2 percent increase in oil prices after one year
  - China aggregate supply shocks: approximately 1.6 percent increase in oil prices after one year
  - Three-year horizon for China: AD 3.1 percent; AS 2.5 percent (effects significant only for China during this period)
  - Other G20−EMs: minimal oil price spillovers around 0.25 percent
  - India aggregate demand shocks: around 1 percent impact on oil prices at one year
  - US shocks (for context): 2.5 percent increase in oil prices after one year
- Oil production responses:
  - China demand and supply shocks increase oil production by approximately 0.5 to 0.6 percent after one year
  - Reactions to shocks from other G20−EMs are significantly smaller
- Metal prices:
  - China’s AD and AS shocks increment metal prices by about 0.25-0.3 percent after one year
  - Other G20−EMs show little to no similar effect
- Conclusion: the commodity price channel is particularly potent for China, affecting commodity exporters, global prices, and production.

### 4.5. Changes in the Size of Spillovers over Time
- Trade-weight baseline: results reported used trade weights for 2017 to 2019; robustness includes earlier trade-weight comparisons.
- China’s trade integration evolution (early 2000s to present):
  - Contribution to global merchandise exports: from approximately 4% to over 14%
  - Import share: from around 4% to 11%
  - Share of global manufacturing gross production: 35% (and 29% of value-added)
  - China is the largest export destination for 33 countries and the primary source of imports for 65 countries
- Other G20−EMs: more modest expansion in trade shares; heterogeneous bilateral developments (e.g., India, Indonesia, Mexico slight increases; Argentina, Russia declines).
- Comparative re-estimation using trade weights: averages for 2001–2003 versus 2017–2019, GDP-PPP weighted aggregates, demand and supply shocks.
- Key temporal findings:
  - Spillovers from China have roughly doubled over the last twenty years (post-WTO accession) for both demand and supply fluctuations.
  - Example magnitudes (advanced economies):
    - Aggregate supply shocks: from 0.19 to 0.4
    - Aggregate demand shocks: from 0.25 to 0.46
  - Example magnitudes (emerging economies):
    - Supply shocks: from 0.15 to 0.39
    - Demand shocks: from 0.23 to [text truncated in source; evolution described as rising markedly]

*International Monetary Fund*

### 0.79 for demand shocks.

### wpiea2025027-print-pdf - 0.79 for demand shocks.

### Major findings on spillovers
- China’s spillovers are large, widespread, and have increased substantially over the last two decades.
- A one percentage point uptick in China’s GDP growth elicits:
  - an approximate increase in the output of advanced economies by 0.2−0.4 percentage points;
  - an approximate increase in the output of other emerging economies by 0.3−0.8 percentage points.
- Demand shocks from China generate larger spillovers than supply shocks:
  - 0.3 percent versus 0.2 percent at the one-year horizon on global output;
  - at the three-year horizon, cumulative effects: 0.6 percent from demand-side shocks versus 0.4 percent from supply-side shocks on world output.
- Global impact estimates and comparisons with prior studies:
  - Prior non-structural studies report a range of 0.15 to 0.5 percent decrease in global output for a 1 percent reduction in Chinese GDP growth over one to three years.
  - The paper’s GIRFs indicate a global impact of about 0.75 percent at the one-year horizon.
  - Copestake et al. (2023) report supply shocks induce a cumulative 0.6 percent decrease in GDP in other countries after two years, and demand shocks a smaller 0.4 percent decline after four years; the present study finds relatively larger demand-side effects.
- Regional and sectoral heterogeneity:
  - Some countries (particularly in Asia) have prior estimates of impacts as large as 1.6 percent.
  - Commodity exporters in emerging Asia and Latin America, especially oil-exporting nations, experience significant spillovers.
  - China’s demand and supply shocks materially affect oil and metal prices; Cashin et al. (2016) find a one standard deviation decrease in China’s GDP results in a 2.8 percent reduction in commodity prices after three years.
- Spillovers from other G20−EMs are generally more modest (typically below 0.1 percentage points), though certain commodity exporters (e.g., Russia, Saudi Arabia) can trigger larger and more persistent regional effects.
- Variation over time and weighting: aggregate responses of output to G20−EM demand and supply shocks were evaluated using different weighting schemes (2001-2003 and 2017-2019 average trade weights), with PPP GDP weighted impulse responses three year ahead based on responses significant at 68 percent credible intervals.

### Robustness, identification, and methodological insights
- Model and estimation choices explored:
  - Bayesian GVAR model covering 63 countries.
  - Multiple prior specifications: Minnesota (MN), Stochastic Search Variable Selection (SSVS), and Normal-Gamma (NG).
  - Lag lengths explored from one to four lags.
  - Models estimated with and without time trends and stochastic volatility.
  - Structural identification of demand and supply shocks tested under alternative restrictions and durations.
  - A traditional (non-Bayesian) GVAR with a smaller set of countries and variables was also analyzed.
- Robustness conclusions:
  - Core conclusions—particularly the large magnitude of China-originating spillovers and their escalation over the last two decades—are robust across specifications.
  - Quantitative magnitudes vary across specifications but the qualitative pattern (China dominance, stronger demand shocks) is consistent.
- Differences with other studies:
  - The study’s higher-end spillover estimates likely reflect use of the most recent data sample and structural identification.
  - Copestake et al. (2023) used a simpler 4-variable VAR and a proxy for Chinese GDP (China Cyclical Activity Tracker), while this paper uses official GDP figures and a more comprehensive dataset.

### Data, sample, and diagnostics
- Dataset and sample:
  - Quarterly observations from 2001Q1 to 2023Q2.
  - 63 countries included.
  - Baseline weight matrix based on annual bilateral trade flows averaged over the period 2017 to 2019.
- Variables in country models:
  - real GDP (logarithm of real GDP);
  - consumer price inflation (difference of log CPI);
  - short-term interest rates (0.25×log(1+Rit/100), with Rit denoting 3-months money market rates);
  - long-term interest rates (similar transform for 10-year government bond yields);
  - real exchange rate against the US dollar (logarithm, deflated using consumer prices);
  - equity prices (logarithm of nominal equity price index deflated by CPI);
  - additional global variables: oil price (log nominal price of oil in US dollars), oil production (log mil. Barrels/Day), oil inventories (log forward consumption in OECD in days), global activity proxy (global real economic activity index in industrial countries).
- Country coverage highlights:
  - Europe: majority of sample (26 advanced, 10 emerging).
  - Asia-Pacific: 13 countries (6 advanced, 7 emerging).
  - Western Hemisphere: 9 countries (2 advanced, 7 emerging).
  - Middle East and Central Asia: 5 emerging economies.
  - Africa: limited coverage due to lack of quarterly data.
- Model diagnostics and selection:
  - The SSVS prior with a two-lag structure exhibited enhanced performance (residual autocorrelation and cross-country residual correlation).
  - Geweke convergence diagnostic: out of 218460 variables, 25169 (11.52%) have Z-values exceeding the 1.96 threshold.
  - Posterior median residuals indicate acceptably low cross-country residual correlation for structural analysis.

### Policy implications and conclusions
- Policymakers should view China’s economic developments as having substantial and global implications, given:
  - the sizable estimated spillovers to both advanced and emerging economies (explicit magnitudes given above);
  - the predominance of demand-side transmission channels relative to supply-side channels;
  - China’s growing role in commodity markets (oil and metals) and global value chains.
- The potential for broad adverse global ramifications from a Chinese economic deceleration is high and unlikely to be fully offset by other emerging markets in the foreseeable future.
- Recommended stance for policymakers:
  - Remain vigilant to developments in China’s demand and supply conditions;
  - Monitor commodity markets and GVC linkages closely, given their roles in transmission;
  - Use the detailed country-level GVAR insights to tailor regional and country-specific responses rather than relying on aggregate EM measures alone.

*Source: IMF Working Papers — Spillovers from Large Emerging Economies: How Dominant Is China? Working Paper No. WP/2025/027*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025027-print-pdf.pdf_