## 1. Countries and Effects in the Trade Network

## Source details

**Canonical URL:** [1. Countries and Effects in the Trade Network](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15149.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15149.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15149.pdf.json)

---

### Summary of method and key findings
- Proposes a new method for assessing international spillovers from nominal demand shocks derived from complex network theory.
- Defines rounds of spillovers:
  - First round: direct spillover — drop in demand at the epicenter affects its trading partners.
  - Second round: spill-in — affected trading partners propagate the shock to their partners.
  - Third element: spill-back — countries affected in the first two rounds radiate the shock back to the epicenter.
- Short-run assumption: countries cannot take measures to prevent spillovers.
- Key findings:
  - Network effects can amplify shock spillovers; network effects may exceed the initial shock in magnitude.
  - Impact from a domestic demand shock can be calculated for any other country, region, and the epicenter itself; results can inform policy recommendations.
  - Network effects generally higher for small open economies and lower for large and relatively closed economies.
  - Spillover profile depends on network structure: epicenter size/location, number and characteristics of partners, direction and strength of flows.
  - Country roles:
    - About 40 percent of countries block spillovers.
    - About 40 percent absorb spillovers.
    - About 20 percent amplify shocks.
  - Typical patterns:
    - Most developed countries pass through shocks by amplifying or absorbing them.
    - Low-income countries and commodity exporters typically block shocks.
  - Policy implication: Countries capable of transmitting shocks should bear special responsibility for international economic stability; macroeconomic policies in spillover amplifiers (e.g., United States, Switzerland, Italy, Korea, and India) and some absorbers (e.g., Japan, Germany, France) may help attenuate negative-demand-shock spillovers from large countries.

### Framing of spillovers and channels
- Epicenter shocks: unexpected lower GDP growth versus baseline; origins include:
  - domestic banking crisis,
  - loss of consumer confidence,
  - fiscal contraction,
  - exogenous developments (drop in international prices for main export commodity, natural disasters, geopolitical crises),
  - unilateral policy actions (e.g., import restrictions, sanctions, retaliation).
- Main transmission channels:
  - Trade and financial flows (most important).
  - Supply-side: disruption of global supply chains.
  - Financial spillovers: cross-border bank claims and equity holdings.
  - Other channels: remittances, direct and portfolio investment, tourism, commodity prices.
- Objective: develop a computable network model of international spillovers on bilateral balance-of-payments flows, focusing on nominal demand (import-demand) shocks and sequential transformation of inflow-outflow matrices to capture initial shock, spillin, and spillback. Illustrative applications: import demand shocks in China and Ukraine.

### Network economics of spillovers — mechanism and formalism
- Balance of payments as a multilayer network; trade (current account) treated as single-layer network; extendable to other layers.
- Data constraints: bilateral trade data from UN Comtrade available for 1962–present; capital/financial bilateral data scarcer.
- Economic mechanism (import-demand shock assumptions):
  - Assume exchange rates and prices do not adjust quickly; shock affects nominal imports as aggregate demand falls.
  - Marginal propensity to import (MPM) can be calculated for each country (equation references in source).
  - Import shortfall expressed as deviation when actual growth differs from projected (equations (3) and (4)).
  - Spillover occurs if imports depend on exporters’ export revenue (equation reference (5)); cointegration between exports (X) and imports (M) is a precondition for long-run dependence.
- Pass-through coefficient (β) and classification:
  - Pass-through estimated via elasticity of imports to export revenue (equation (7)); estimated β classifies countries:
    - Amplifier: pass-through parameter > 1 → change in export revenue causes proportionally larger change in imports (expands initial shock).
    - Absorber: 0 < pass-through parameter < 1 → change in export revenue causes proportionally smaller change in imports (reduces shock strength but still spills over).
    - Blocker: pass-through parameter ≤ 0 or not statistically significant → exports revenue is not a constraint for imports; shock does not spill further.
- Sequence of transmission (cascade algorithm):
  1. Initialization: epicenter nominal demand shock ΔYi → assuming MPM = 1, translate to ΔMi; distribute ΔMi proportionally among exporters to epicenter → export revenue shocks for adjacent countries.
  2. First round: adjacent countries’ export revenue losses lead to GDP declines depending on export/GDP shares; their pass-through β determines how much they reduce imports.
  3. Second and subsequent rounds: reduced imports by first neighbors create export revenue shocks for their exporters; process repeats until shocks become insignificant.

### Network representation and topologies
- International trade modeled as a directed, weighted, incomplete, and asymmetric graph:
  - Directed: links represent exports revenue/imports payments.
  - Weighted: links carry payment values.
  - Incomplete: not all countries trade with each other.
  - Asymmetric: out-links vs in-links differ by country.
- Neighbor orders defined:
  - Epicenter, first neighbors, second neighbors, third neighbors, etc.
- Elementary link types (from epicenter A perspective):
  - (a) No link: no direct impact; indirect spillin possible.
  - (b) One-way wrong direction (A → C): import-demand shock in A does not directly affect C; spillins possible.
  - (c) One-way right direction (D → A): import-demand shock in A directly affects D (loss of export revenue); no immediate spillback, indirect possible.
  - (d) Two-way link (A ↔ E): import-demand shock in A affects E and E’s loss in export revenue reduces its imports including from A → immediate spillback.
- Topologies illustrated: star, tree, ring, line, mesh/feedback — real-world topology is much more complex.
- Multiround dynamics example (numeric schematic):
  - Epicenter A: first-round import drop of 100 → exporters B, C, D lose 20, 30, 50 respectively.
  - B = blocker (no secondary spillover); C = amplifier (30 → reduces imports by 40); D = absorber (50 → reduces imports by 30).
  - Second round: C’s 40 → partners reduce imports by 20; D’s 30 → partners reduce imports by 40.
  - Third round: shock dies for C, persists for D at 20; fourth round: shock dies for D.
  - Conclusion: total multistep spillover can exceed immediate first-round effect; secondary spillovers depend critically on network properties.

### Matrix formalism and cascade steps
- Export-import matrix X with element X_ij = exports from country i to country j.
- Initial import demand shock ΔM_A distributed proportionally among exporters to A → export revenue losses ΔR for those exporters.
- Column-normalized transformations used to compute ΔR as matrix multiplications.
- Simplest linear specification for import response: ΔM = β · ΔR (β = pass-through coefficient), possibly with lags; newly generated import shocks become next round’s export revenue shocks.

### Data, sample, and pass-through estimation
- Data source: COMTRADE bilateral flows for 1993–2013.
- Original set: 184 countries; 14 excluded due to data deficiencies (ATG, BTN, BWA, ERI, KIR, LSO, MNE, NAM, SRB, SWZ, TLS, TUV, TWN, UVK).
- Final sample: 170 countries.
- Of 28,730 possible bidirectional trade flows, 9,029 (about 31 percent) are absent.
- Estimation period: 1993–2013; some countries have constrained windows (e.g., Belgium and Luxembourg not available for 1993–96; South Africa for 1993–97).
- Pass-through coefficients estimated on a panel with fixed effects; three specifications considered:
  - Model 1: intercept and no lags.
  - Model 2: no intercept and no lags.
  - Model 3: intercept and one lag (exports lagged one year).
- Model selection and robustness:
  - Model 3 performs poorly: "very low overall fit", "relatively large standard errors", most coefficients statistically insignificant; only in eight cases lagged pass-through significant.
  - Models 1 and 2 have substantially better fit and smaller standard errors; most coefficients significant at the 5 percent level.
  - Model 1 (intercept, no lags) chosen in reduced form for pass-through estimation; intercept α captures determinants of imports beyond export revenue.
  - Caveats: annual data and assumption of slow exchange-rate/price adjustment may bias β estimates; an economy’s role may depend on participation in global supply chains.
- Cointegration check:
  - Assuming exports and imports nonstationary in log levels, cointegration between imports and exports assessed.
  - "In 134 countries, imports seem cointegrated with exports"; overlap with regression-based country lists.

### Country classification by pass-through coefficients (Model 1 results)
- Of 170 countries, 95 countries are capable of passing through shocks.
- Counts and shares:
  - 29 countries are potential shock-amplifiers (17 percent of the total).
    - Many with insignificant amplifications of 5–10 percent (examples cited: United States, India, Brazil, Italy, Switzerland).
    - A small subgroup expands original shocks by 30 percent and more (examples cited: Argentina, Thailand, Korea, Hong Kong SAR, Denmark, Indonesia, India).
  - 70 countries (41 percent) are shock-absorbers.
  - 71 countries (42 percent) are shock-blockers (β statistically insignificant).
    - Examples of blockers: Benin, Chad, Central African Republic, Dominica; some oil producers (Qatar, Iran, Nigeria); some financial centers (Cyprus, Luxembourg).
- Group statistics (as presented in source):
  - Amplifiers / Absorbers / Blockers
    - No. of countries: 29 / 70 / 71
    - Percent of total: 17 / 41 / 42
    - Pass-through coefficient (group summary)
      - Amplifiers: Max 1.68, Min 1.01, Mean 1.16, Median 1.08
      - Absorbers: Max 0.99, Min 0.14, Mean 0.62, Median 0.66
    - Group averages
      - GDP per capita (US$, PPP, 2011): Amplifiers 24,248; Absorbers 20,482; Blockers 11,386
      - Standard error (GDP pc): Amplifiers 2,631; Absorbers 2,145; Blockers 2,372
      - Openness (trade % of GDP, 2013): Amplifiers 101; Absorbers 94; Blockers 94
      - Doing business rating (2013): Amplifiers 53; Absorbers 80; Blockers 122
- Typical interpretations:
  - Amplifiers: mainly developed, PPP GDP per capita ≈ US$ 24,000; openness > 100 percent; Doing Business ≈ 53; private sector may overreact to export shortfalls.
  - Absorbers: lower per capita GDP and openness than amplifiers; worse business environment.
  - Blockers: mainly middle/low-income and some oil exporters; imports delinked from export revenue (e.g., financed by sovereign funds or donors).

### Shock calibration and illustrative shocks
- Calibration approach: demand shocks calibrated in percent of epicenter country’s imports based on historical precedents (2014 IMF Spillover Report and October 2014 WEO).
- Historical average: "Average 1993–2013 import demand shocks hover around 10 percent."
- Illustrations use uniform 10 percent drop of nominal imports in 2012.
  - In China case: 10 percent drop of China’s imports amounted to 4.3 percent of its 2012 GDP (also a note earlier that another calibration equated China shock to 1.7 percent of 2012 GDP in a different context).
  - In Ukraine case: shock represented as 11 percent of GDP decline in Ukraine’s imports in one description; elsewhere an example set to 10 percent drop for illustrative uniformity.
- Temporal splitting: annual trade flows split into four equal quarterly flows; shock affects epicenter in Q1 and spills over during remaining three quarters (four-round spillover process). Rationale: intrayear correlation and quarterly adjustment of import values to export proceeds.

### Application: import-demand shock in China
- China network position:
  - China’s in/out degree: 168/166 out of 170.
  - China centrally located; largest trade flows pass through China.
- Shock calibration and magnitudes:
  - Shock: China’s imports drop by 10 percent (noted as 4.3 percent of its GDP in 2012).
  - For 30 countries with largest impact:
    - First-round direct spillover effects average 3.6 percent of their GDP.
    - Largest first-round impacts (percent of GDP): Hong Kong SAR 22.6, Mongolia 8.7, Solomon Islands 8.6.
  - Four-round total effects:
    - Indirect spillovers between countries other than China add an additional 3.5 percent of GDP loss on average.
    - Total impact on average: about 7.1 percent of GDP for all countries.
    - Largest total direct and indirect spillovers: Hong Kong SAR, Singapore, Solomon Islands.
  - Spillin (ricochet) effect:
    - Average absolute spillin ≈ 3.5 percent of GDP.
    - Largest absolute spillin (percent of GDP): Singapore 11.9, Hong Kong SAR 8.2, Sweden 6.8.
    - Highest spillover strength (total/first-round ratio): Trinidad and Tobago 29.5, Brunei Darussalam 14.9, UAE and Qatar 6.6.
  - Spillback to China:
    - Spillback effect on China about 0.2 percent of GDP (ricochet from partners).
  - Import-side analysis notes:
    - Average spillin (difference between total and second-round spillover) about 2.1 percent of GDP.
    - Highest (percent of GDP): Hong Kong SAR 11.3, Singapore 6.3, Kyrgyz Republic 5.1.
    - Strongest spillin ratio (total/second-round): Dominica 30.8, Malta 18.5, Slovak Republic 14.6.
    - Spillback of 0.6 percent of GDP on China itself noted in import-side analysis (reduction of China’s imports on top of original 4.3 percent of GDP shock).

### China: export vs import shock profiles (averages)
- Export shock profile:
  - First round: on average China’s partners immediately lose about 1 percent of their GDP.
  - Second–fourth rounds add on average about 0.5 percent of GDP to first-round shock.
  - Total after four rounds: about 2.3 percent of GDP on average for all countries.
  - Export shock remains broadly unchanged after four rounds (does not decay).
- Import shock profile:
  - First round: spillover zero for all countries (only China experiences shock).
  - Second round: average magnitude ≈ 0.6 percent of GDP.
  - Third and fourth rounds: ≈ 0.3 percent of GDP each.
  - After four iterations total import shock reached 1.2 percent of GDP (compared to export shock of 2.3 percent).
  - Contributing reasons: ~70 countries block spillovers before next import shock; not all import shocks lead to next export shock; amplifiers play key role in second round.

### Application: import-demand shock in Ukraine
- Ukraine network and trade structure:
  - Ukraine’s in/out degree: 151/156 out of 170.
  - Ukraine does not trade with at least 10 percent of countries.
  - Ukraine’s exports to main 11 partners = 60 percent of exports; imports from 11 main partners = 77 percent of imports.
  - Russia dominant export destination and source of imports; China and Germany important exports but less as import sources; Turkey, Czech Republic, United States, Hungary important export destinations but not top import partners; Egypt, India, Spain important import sources but not top export destinations.
- Pass-through characteristics and immediate effects:
  - Most of Ukraine’s trading partners can pass through shocks.
  - On import side, all partners except Belarus and Egypt can pass through potential spillovers.
  - Russia and Kazakhstan absorb part of the shock; Bulgaria blocks.
  - Turkey, Italy, Poland, Spain can amplify their share of the shock.
- Shock calibration and magnitudes:
  - Example shock: an 11 percent of GDP decline in Ukraine’s imports (presented in text as a shock representation).
  - For countries with impact above average:
    - First-round direct spillovers average 0.2 percent of their GDP.
    - Largest first-round impacts: Belarus 2.2 percent of GDP (export revenue loss), Lithuania 0.6 percent of GDP.
  - Four rounds:
    - Indirect spillovers add additional 0.2 percent on average; total average impact ≈ 0.4 percent of GDP.
    - Examples after four rounds: Belarus 2.7 percent of GDP; Lithuania ≈ 1.0 percent of GDP.
    - Average absolute spillin ≈ 0.2 percent of GDP.
    - Strength of spillin (ratio total/first-round) highest for medium-size open economies and some small oil-producers.
    - Visible spillback on Ukraine itself ≈ 0.2 percent of GDP.
- Import shock into partners and downstream effects:
  - Import shock becomes nontrivial only from second round.
  - Shock blockers (e.g., Belarus, Egypt) halt spillovers when affected by first round export revenue drop.
  - Second round hardest impact: Lithuania reduces imports by 0.7 percent of GDP; Poland, Estonia, Guyana reduce further imports by 0.3 percent of GDP.
  - Largest overall impact after four rounds: Kyrgyz Republic 1.1 percent of GDP; Lithuania 1.0 percent; Estonia and Latvia ≈ 0.8–1.0 percent of GDP.
  - Average spillin (difference between total and second-round) ≈ 0.2 percent of GDP; largest for Kyrgyz Republic ≈ 0.9 percent of GDP.
  - Strongest spillin ratios for medium-size open economies: Hong Kong SAR, Malta, Sweden, Singapore.
- Export vs import shock averages for Ukraine:
  - Export shock:
    - First round: partners lose 0.05 percent of GDP on average.
    - Second round adds ≈ 0.03 percent; third and fourth ≈ 0.02 percent each.
    - Total after four rounds ≈ 0.12 percent of GDP on average.
    - Export shock decays relatively fast; may not be statistically different from zero after four rounds.
  - Import shock:
    - First round: zero for all partners.
    - Second and third rounds: ≈ 0.02–0.03 percent of GDP.
    - Fourth round: < 0.02 percent of GDP.
    - Total import shock ≈ 0.07 percent of GDP.
    - Import shock totals a little more than half of export shock for Ukraine’s partners on average.

### Network findings and policy implications (conclusions)
- Network model captures second-round network effects often disregarded by traditional methods.
- Network effects originate from feedback starting at second round; strength depends on:
  - magnitude of initial shock at epicenter,
  - epicenter centrality and network properties,
  - positions of main trading partners,
  - domestic economic structures of partners,
  - compounding strength of signals propagating in same direction,
  - offsetting signals in opposite directions.
- When compounded, network effects can be comparable to and often exceed initial shock at epicenter.
- Classification recap:
  - ≈ 20 percent of countries amplify spillovers.
  - ≈ 40 percent absorb spillovers.
  - ≈ 40 percent block spillovers.
- Typical patterns:
  - Most developed and major trading nations tend to pass through shocks.
  - Most commodity exporters, especially oil producers, and low-income countries tend to block spillovers.
- Capacity to amplify/absorb/block depends on development, openness, and business environment: higher development/openness → higher probability to pass through shocks.
- Policy implications:
  - Macroeconomic policies of spillover-conducive countries are important for global stability.
  - Core amplifiers identified include United States, Switzerland, Italy, Korea, India.
  - Spillover absorbers with relatively high pass-through include Japan, Germany, France.
  - Coordinated macroeconomic policies among core countries can mitigate negative external shock spillovers.
  - Policy levers include altering export-import matrix properties: in-/out-degrees, weighted centrality, flow weights, pass-through coefficients.
- Research agenda:
  - Quantify how changes in degrees, centrality, flow weights, and pass-through affect spillover magnitudes and profiles.
  - Model dynamic evolution of export-import matrix across successive shock waves.
  - Identify policy designs to optimally reshape network properties to limit adverse spillovers without significant distortions.

*Source: _wp15149 - 1. Countries and Effects in the Trade Network (extracted from the provided IMF PDF content unit).*

### 1. Countries and Effects in the Trade Network ........................................................................11

### 1. Countries and Effects in the Trade Network ........................................................................11

### Summary
- Proposes a new method for assessing international spillovers from nominal demand shocks derived from complex network theory.
- Defines rounds of spillovers:
  - First round: direct spillover effect — drop in demand at the epicenter affects its trading partners.
  - Second round: spill-in effect — affected trading partners propagate the shock to their partners.
  - Third element: spill-back effect — countries affected in the first two rounds radiate the shock back to the epicenter.
- Assumes that in the short run countries cannot take measures to prevent spillovers.
- Key findings:
  - The strength of shock spillovers can be amplified by network effects; network effects may exceed the initial shock in magnitude.
  - The impact from a domestic demand shock can be calculated for any other country, region, and the epicenter country itself; results can inform policy recommendations.
  - The size of the network effects is generally higher for small open economies and lower for large and relatively closed economies.
  - The profile of spillovers depends on the network structure: size and location of the epicenter country in the network, the number and economic characteristics of its partners, and the direction and strength of economic flows among them.
  - Individual countries may amplify, absorb or block spillovers: About 40 percent of countries block, 40 percent absorb and 20 percent amplify shocks.
  - Most developed countries pass through shocks by either amplifying them or absorbing part of their strength. Low-income countries and commodity exporters typically block shocks.
  - Policy implication: Countries capable to transmit shocks should bear special responsibility for international economic stability. Economic policies in spillover amplifiers (e.g., US, Switzerland, Italy, Korea, and India) and some spillover absorbers (e.g., Japan, Germany, France) may help attenuate the impact of spillovers from negative demand shocks in large countries.

### I. Introduction — framing of spillovers
- International spillovers reflect the impact of macroeconomic changes in one country on other countries through integrated international flows captured in balance of payments.
- Epicenter shocks are typically unexpected lower GDP growth versus baseline projections and can stem from:
  - domestic banking crisis,
  - loss of consumer confidence,
  - fiscal contraction,
  - exogenous developments (drop in international prices for main export commodity, natural disasters, geopolitical crises),
  - unilateral policy actions (e.g., import restrictions, sanctions, retaliation).
- Channels for spillovers:
  - Trade and financial flows are the most important for most countries.
  - Supply-side effects: disruption of global supply chains can negatively affect production in partner countries.
  - Financial spillovers via cross-border bank claims and equity holdings.
  - Other channels: remittances, direct and portfolio investment, tourism, commodity prices.
- Objective of the paper: develop a method to assess network effects in cross-border shock spillovers, focusing on nominal demand shocks and sequential transformation of inflow-outflow matrices of bilateral flows; method captures initial shock and subsequent network effects (spillin and spillback). Illustrative applications: import demand shocks in China and Ukraine.

### II. The Network Economics of Spillovers — literature and economic mechanism
- Literature highlights:
  - Network cascade models (Jackson, 2010; Bass model ideas).
  - Epidemic models on networks (Newman, 2010; SI, SIR, SIS models).
  - Diffusion as branching processes (Easley and Kleinberg, 2010).
  - Sparse prior studies on international spillovers in a network context:
    - Cerdeiro and Wirkierman (2008): linear general interdependence model for propagation through trade.
    - Kali and Reyes (2010): global trading system as an interdependent complex network; crisis amplification when epicenter is well integrated; target countries better able to dissipate impact if well integrated.
    - Vidon (2011): impacts of changes in US imports including knock-on effects via interconnectedness.
    - Fronczak and Fronczak (2012): spillover model based on a fluctuation response theorem linking relative changes in bilateral trade to changes in GDP of trade partners.
    - Fagiolo and others (2014): Leontief input-output matrices; impact of shocks depends on nature of shock and country size; shocks to final demand are large but homogeneous; shocks changing input-output structures are large but more heterogeneous.
- Contributions of this paper:
  - Develops a computable network model of international spillovers usable on any bilateral balance of payments flows.
  - Identifies and estimates network effects of international shock spillovers that can significantly amplify the initial shock and are largely untraceable by existing methodology.
  - Proposes the concept and estimation of a pass-through coefficient to quantify shock percolation through individual countries and classify their ability to amplify, absorb, or block shocks.
  - Proposes macroeconomic interpretations of network concepts and metrics.

- Balance of payments as a multilayer network:
  - Trade (current account) treated as a single-layer network; methodology extendable to other layers of the balance of payments.
  - Data constraints: bilateral trade data available from UN Comtrade (detailed bilateral exports and imports of goods from 1962 to present for over 200 countries and areas). Capital and financial account bilateral data are scarcer; Coordinated Portfolio Investment Survey covers 74 countries for 2001–2012.

- Economic mechanism of import-demand shocks:
  - Assume exchange rate and prices do not adjust quickly; shock affects nominal imports as aggregate demand falls.
  - Marginal propensity to import can be calculated for each country (equation reference (1)).
  - Import demand projections depend on GDP growth and marginal propensity to import (equation reference (2)).
  - Import shortfall expressed as deviation when actual growth differs from projected (equations (3) and (4)).
  - Spillover to other countries occurs if imports depend on export revenue of exporters (equation reference (5)): a country’s capacity to pay for imports depends on export revenue, possibly with lags.
  - Cointegration between exports (X) and imports (M) is a precondition for imports depending on exports in the long run; cointegration evidence:
    - X and M cointegrated in 35 out of 50 countries in some studies (Arize, 2002), including the United States, Australia, and some developed countries.
    - Narayan (2005) found cointegration in 6 out of the 22 countries in that sample.
  - Pass-through can be estimated via an elasticity of imports to export revenue (equation reference (7)); the estimated pass-through coefficient (denoted as parameter in text) classifies countries:
    - Spillover amplifying if pass-through parameter > 1: a change in export revenue of neighbors leads to a proportionally larger change in imports, expanding the initial shock.
    - Spillover absorbing if 0 < pass-through parameter < 1: change in export revenue leads to a proportionally smaller change in imports; the shock is reduced but still spills over.
    - Spillover blocking if pass-through parameter ≤ 0 or not statistically significant: exports revenue is not a constraint for imports; the shock to exports revenue does not affect imports and thus does not spill further.

- Sequence of shock transmission (summary of the paper’s framework):
  - Initialization: initial shock to epicenter country i is decline in its nominal demand ΔYi; assuming marginal propensity to import is unity, this translates into a decline in its imports of ΔMi; this becomes a loss of export revenue for adjacent countries, redistributed proportionally to shares in the epicenter’s imports.
  - First round: loss of export revenue for adjacent countries leads to a decline in their GDP; the impact depends on the share of exports in their GDP.
  - Pass-through behavior: countries are classified by their pass-through parameter into amplifiers, absorbers, or blockers as above.
  - Second and sequential rounds: variably lower GDP growth of immediate trading partners translates into demand shocks for their trading partners proportional to the decline in export revenue of each immediate partner; assuming marginal propensity to import equals unity, imports of first neighbors from their immediate neighbors decline in proportion to change in their export revenue.

*Source: _wp15149 - 1. Countries and Effects in the Trade Network ........................................................................11*

### 19. International trade can be presented in a network form. Each country would be

### _wp15149 - 19. International trade can be presented in a network form. Each country would be

### Network representation of international trade
- International trade is represented as a directed, weighted, incomplete, and asymmetric graph:
  - Directed: links represent flows (exports revenue, imports payments) from one country to another.
  - Weighted: links reflect payment values that vary by flow and country.
  - Incomplete: not all countries are connected with each other through trade.
  - Asymmetric: number of export partners (out-links) differs from number of import partners (in-links) for most countries.

### Types of countries and neighbor structure (shock-spillover perspective)
- Definitions and neighbor orders:
  - Epicenter: country affected by a domestic demand shock.
  - First neighbors: immediate trading partners of the epicenter.
  - Second neighbors: countries directly connected to first neighbors.
  - Third neighbors and beyond: iterative extensions via neighbors' links.
- Examples of first-neighbor counts:
  - Small developing countries (e.g., Burundi, Tonga, Guinea-Bissau, Solomon Islands, São Tomé and Príncipe, Comoros, Vanuatu): export to and import from not more than 40–50 other countries.
  - Largest trading nations (China, United States, Germany, Netherlands, Turkey): trade with 165–169 countries.
- Spillover terminology:
  - Direct spillover: shock transmission from epicenter to first neighbors.
  - Indirect spillover: transmission from first neighbors to second through nth neighbors.
  - Spillback: ricochet impact from any neighbor back to the epicenter.
  - Spillin: ricochet impact on first neighbors from second to nth neighbors.

### Elementary link types and immediate consequences for spillovers
- Four elementary link options (from epicenter A perspective):
  - (a) No link in any direction: no direct impact from A on B; indirect impact via spillin still possible.
  - (b) One-way link in "wrong" direction: A exports to C (A → C) but does not import from C; an import demand shock in A would not directly affect C; spillins possible.
  - (c) One-way link in "right" direction: A imports from D (D → A); an import demand shock in A directly affects D (loss of export revenue); no direct spillback from D to A but indirect spillbacks possible.
  - (d) Two-way link: A imports from and exports to E; an import demand shock in A affects E and there will be immediate spillback from E to A because E's loss in export revenue reduces E's imports from all its partners including A.
- The strength of initial spillover, spillback, and spillin depends on relative weights of each link.

### Network topologies and implications for spillovers
- Possible topologies described (epicenter A with neighbors):
  - Star (A with first neighbors B, C, D): only some links can spill over.
  - Tree: directed spillovers without loops (e.g., D → F, G).
  - Ring: fully interlinked countries (e.g., E, K, L) where spillover and spillin depend on link weights.
  - Line: sequential links with no spillins (e.g., D → F → H → G).
  - Mesh/feedback: sequential links with feedback between nodes (e.g., L ↔ M ↔ N).
- Consequences:
  - Some countries may not pass through shocks (shock blockers), others may amplify or absorb shocks.
  - A shock can die out for some branches while persisting for others for several rounds.
  - Real-world trade topology is much more complex than simple illustrations.

### Types of shocks in the spillover process
- Import shock: drop in import demand in the epicenter country for any reason (an exit shock).
- Export revenue shock: drop in export revenue of epicenter’s trading partners resulting from the import shock (an entrance shock).
- Distinction and dynamics:
  - Import shock sends a signal out from the epicenter to first, second, ... nth neighbors.
  - Export shock affects neighbors following an import shock at the epicenter.
  - Export shock for each country will always be nonzero; import shock will be nonzero only for countries where import depends on export revenue, and zero otherwise.
  - First neighbors face both the direct shock from the epicenter and spillins from other first neighbors.

### Multiround spillover dynamics (illustrative numeric example)
- Example setup:
  - Epicenter A has three first neighbors B, C, D.
  - A’s imports drop by 100 at first round; immediate losses in export revenue for B, C, D are 20, 30, and 50, respectively.
  - Pass-through behaviors:
    - B does not pass through shocks (shock blocker): no secondary spillover.
    - C amplifies shock: loss of exports revenue 30 → C reduces imports by 40 (amplification).
    - D absorbs part of shock: loss of exports revenue 50 → D reduces imports by 30 (partial absorption).
- Second round:
  - Only C and D spillover further:
    - C’s 40 loss in export revenue → their partners reduce imports by 20 (attenuation).
    - D’s 30 loss in export revenue → their partners reduce imports by 40 (amplification).
- Third round:
  - Shock dies for C but persists for D at 20.
- Fourth round:
  - Shock dies for D.
- Conclusion: total multistep spillover effect may be substantially larger than the immediate (first-round) effect; secondary spillovers depend critically on network properties.

### Summary of spillover mechanism
- Immediate spillover:
  - List of immediate trading partners and their marginal propensity to import (MPM) is well known; distribution based on share of epicenter in partners’ exports is straightforward.
- Secondary spillovers:
  - Complex and less understood due to network effects and heterogeneous pass-through across countries.
- Algorithmic cascade description:
  - Each round consists of:
    1. Distributing the epicenter’s import demand shock proportionally among exporters to the epicenter to create export revenue shocks for those exporters.
    2. Export revenue shocks lead to reduced import demand by affected countries, creating new import shocks for their exporters in the next round.
  - The process repeats until shocks become insignificant (often after four rounds in examples).

### Network model formalism (matrix representation and cascade steps)
- Export-import matrix notation:
  - Export-matrix: rows = exports of a country to all other countries; columns = imports of each country from all other countries.
  - Matrix element X_ij stands for exports from country i to country j.
  - For fixed i, vector X_i· is exports of country i; for fixed j, vector X_·j is imports into country j.
- Cascade step (schematic; equations referenced in source):
  - Initial import demand shock in epicenter (ΔM_A) is distributed proportionally among exporters to the epicenter, creating export revenue shocks ΔX for those exporters.
  - Export revenue shocks create reduced demand for imports and a cascade of import shocks in first neighbors; newly generated import shocks become export revenue shocks for the next round.
- Matrix transformations:
  - An initial export-import matrix X transforms after the first round according to the proportional reduction in export receipts to exporters of the epicenter.
  - Export revenue loss vector ΔR can be written as matrix multiplication involving a column-normalized matrix (each column normalized by its sum).
- Key assumption on spillover dynamics:
  - For some, but not all, countries a decline in export revenue leads to a drop in imports (contemporaneously or with lag).
  - Simplest (linear) specification: import shock ΔM generated by export revenue shock ΔR is determined on average by ΔM = β · ΔR (β = pass-through coefficient), possibly with lags.
  - Newly generated import demand shock becomes the next round’s export revenue shock.

### Numerical example (Box 1) — schematic highlights
- World of three countries A, B, C with bilateral trade represented in an export-import matrix.
- A experiences a demand shock: A’s imports drop by 10 percent (values: drop by 1 from B and by 2 from C; total import shock in A is 3).
- Transformation of export-import matrix yields first-round export revenue losses for B and C (sums across corresponding rows).
- Pass-through behaviors specified:
  - B: shock blocker (does not pass through).
  - C: shock absorber (passes through 0.5 of initial shock).
  - D (in example nomenclature): shock amplifier (pass-through coefficient 1.5).
- Newly generated import shocks are computed for subsequent rounds and applied to the transformed export-import matrix; rounds continue until shock is insignificant, usually after four rounds.

### Data, sample, and pass-through coefficient estimation
- Data:
  - Derived from COMTRADE bilateral flows for 1993–2013.
  - Original set: 184 countries (nodes); due to data deficiencies 14 countries excluded: ATG, BTN, BWA, ERI, KIR, LSO, MNE, NAM, SRB, SWZ, TLS, TUV, TWN, UVK (Annex 1).
  - After exclusions, sample contains 170 countries.
  - Of 28,730 possible bidirectional trade flows, 9,029 (about 31 percent) are absent (no trade in either direction or one direction).
  - Estimation period: 1993–2013; some countries have constrained data windows (e.g., Belgium and Luxembourg not available for 1993–96; South Africa for 1993–97).
  - Exports and imports for each country were aggregated into adjacency matrices representing weighted directed graphs.
- Pass-through coefficients:
  - Estimated on a panel with fixed effects covering 170 countries with nominal exports and imports for 1993–2013.
  - Eq. (7) estimated on contemporaneous changes in nominal exports and imports.
  - Log-linear specification considered preferable but problematic because nominal intercept α in many cases becomes dominant and overshadows β if logged.
  - Three specifications considered:
    - Model 1: intercept and no lags.
    - Model 2: no intercept and no lags.
    - Model 3: intercept and one lag (exports lagged one year relative to imports to test if imports react to export revenue with a lag).

*Source: IMF working paper content unit*

### 34. The model with an intercept and no lags seems superior to others. Model 3 with

### _wp15149 - 34. The model with an intercept and no lags seems superior to others. Model 3 with

### Model selection and pass-through estimation
- Model comparison results:
  - Model 3 (one lag of the independent variable) performs poorly: "very low overall fit", "relatively large standard errors" and "most coefficients are statistically insignificant".
  - Only in eight cases the lagged pass-through coefficient was statistically significant, suggesting on average the values of import react contemporaneously to changes in export revenues.
  - Models 1 and 2 (with and without the intercept α and lagged dependent variable correspondingly) have substantially better overall fit, smaller standard errors, and most coefficients are statistically significant at the 5 percent level.
  - Both Models 1 and 2 produce an almost identical list of countries with β>1.
  - Fundamental differences between Models 1 and 2 are in:
    - the number of statistically insignificant β,
    - the number of countries with statistically significant β<1,
    - the significance of α in model 1 in 44 cases.
  - In model 1, coefficients α and β are simultaneously statistically significant in 13 cases.
  - Statistical significance of the intercept α "cannot be ignored as it captures all determinants that may affect imports, in addition to the country’s export revenue."
- Model chosen for pass-through estimation:
  - A reduced model 1 (imports depend contemporarily mainly on exports revenue and all other factors captured by α) was selected for the estimation of pass-through coefficients.
- Methodological note:
  - "In estimating the pass-through coefficients from exports to imports, only trade data is included and an assumption is made that the exchange rates and prices do not adjust quickly." Using annual data under that assumption "may lead to biases in the pass-through coefficients."
  - "Also, an economy’s status as shock amplifier, absorber, and blocker may depend on its participation in the global supply chain."
- Sample and estimation scope:
  - "Average statistics for model selection; 170 country-specific coefficients were estimated."

### Robustness: cointegration check
- Cointegration results:
  - Assuming exports and imports are not stationary in log levels, a cointegration relationship between imports and exports (Eq. 6) was assessed.
  - "Based on this approach, in 134 countries, imports seem cointegrated with exports; and the country list overlaps with those in the regressions with and without the intercept."

### Country classification by pass-through coefficients
- Classification and counts (based on Model 1 and calculated pass-through coefficients):
  - Of 170 countries, 95 countries are capable of passing through shocks.
  - 29 countries are potential shock-amplifiers (17 percent of the total).
    - Many pass through shocks with insignificant amplifications of 5–10 percent (examples: United States, India, Brazil, Italy, Switzerland).
    - A small subgroup of strong shock amplifiers can expand original shocks by 30 percent and more (examples: Argentina, Thailand, Korea, Hong Kong SAR, Denmark, Indonesia, India).
  - 70 countries (41 percent) are shock-absorbers.
  - 71 countries (42 percent) are shock-blockers (pass-through coefficients are statistically insignificant).
    - Examples of shock-blockers: Benin, Chad, Central African Republic, Dominica; some oil producers (Qatar, Iran, Nigeria); some financial centers (Cyprus, Luxembourg).
  - "The distinction between shock amplifiers, absorbers and blockers depends only on each country’s economic structure and is unrelated to the structure of the network and to the location of each country in the network."
- Summary statistics and group averages (as presented):
  - Amplifiers / Absorbers / Blockers
    - No. of countries 29 70 71
    - Percent of total 17 41 42
    - Pass-through coefficient
      - Max 1.68 0.99 ...
      - Min 1.01 0.14 ...
      - Mean 1.16 0.62 ...
      - Median 1.08 0.66 ...
    - Group averages
      - GDP per capita (US$, PPP, 2011)24,24820,48211,386
      - Standard error2,6312,1452,372
      - Openness (trade in % of GDP, 2013)1019494
      - Standard error1566
      - Doing business rating (2013)5380122
      - Standard error865
- Typical characteristics by group:
  - Amplifiers:
    - "Mainly developed countries with the PPP-based GDP per capita of about US$ 24,000."
    - High openness: ratio of exports plus imports to GDP exceeding 100 percent.
    - Average Doing Business rating of 53.
    - Interpretation: private sector may overreact to export revenue shortfall, amplifying shocks.
  - Absorbers:
    - Statistically close to amplifiers; pass through shocks but absorb part of strength.
    - Lower average per capita GDP and openness; worse business environment than amplifiers.
    - Interpretation: weaker macroeconomic policies, less openness and deficiencies in doing business put natural brakes on transmission.
  - Blockers:
    - "Mainly middle and low-income countries and some oil exporters."
    - Openness comparable to absorbers, but substantially lower per capita income and worse business environment.
    - Interpretation: imports delinked from export revenue; imports financed by sovereign funds or donor resources.

### Shock calibration and illustrative shocks
- Shock calibration approach:
  - "The demand shocks are calibrated in percent of the epicenter country’s imports based on historical precedents."
  - Examples of demand shocks taken from the 2014 IMF Spillover Report and the October 2014 World Economic Outlook.
- Historical context and chosen shock size:
  - "Average 1993–2013 import demand shocks hover around 10 percent."
  - For illustration, individual shocks were set uniformly to a 10 percent drop of nominal imports in 2012.
  - Applied to China (where it amounted to 1.7 percent of its 2012 GDP) and Ukraine (4.7 percent of its 2013 GDP).
  - Sample shocks are applied to the 2012 trade data.
- Temporal splitting for modeling:
  - Annual trade flows were split into four equal quarterly flows.
  - The shock affects the epicenter country in the first quarter and spills over to trading partners during the remaining three quarters (four-round spillover process).
  - Rationale: captures intrayear correlations between export revenue and import flows; "adjustment of import values to intrayear changes in export proceeds also takes place within the year, probably on a quarterly basis."

### Application: shock in a large country (China)
- China’s centrality and network position:
  - China’s in/out degree is 168/166 out of the maximum of 170.
  - Visualization using Fruchterman-Reingold layout shows China centrally located; "the largest trade flows of the world pass through China."
- Asymmetries and implications:
  - China’s export and import partner lists differ; many partners are asymmetrically important as exporters or importers.
  - Consequence: an import demand shock in China leads to immediate drops in export revenue of its partners proportional to China’s share in their exports.
  - Amplification likely because many of China’s main partners (United States, Hong Kong SAR, Korea, Italy, India) are spillover amplifiers.
- China shock calibration and effects:
  - Shock represented by a drop of China’s imports by 10 percent (4.3 percent of its GDP in 2012).
  - Considering 30 countries with largest impact, the first-round direct spillover effects average 3.6 percent of their GDP.
    - Largest first-round impacts (percent of GDP): Hong Kong SAR 22.6, Mongolia 8.7, Solomon Islands 8.6.
  - Four-round total effects:
    - Indirect spillovers between countries other than China add an additional 3.5 percent of GDP loss on average.
    - Total impact on average: about 7.1 percent of GDP for all countries.
    - Largest total direct and indirect spillovers: Hong Kong SAR, Singapore, Solomon Islands.
  - Spillin (ricochet) effect:
    - Average absolute spillin effect amounts to 3.5 percent of GDP.
    - Largest absolute spillin (percent of GDP): Singapore 11.9, Hong Kong SAR 8.2, Sweden 6.8.
    - Highest spillover strength (total/first-round ratio): Trinidad and Tobago 29.5, Brunei Darussalam 14.9, United Arab Emirates and Qatar 6.6.
  - Spillback to China:
    - Spillback effect on China of about 0.2 percent of GDP (ricochet impact from partners back to China).
  - Additional details on rounds:
    - Import-side spillovers (starting from second round) and spillin measured relative to second round:
      - Average spillin (difference between total and second-round spillover) about 2.1 percent of GDP of affected countries.
      - Highest (percent of GDP): Hong Kong SAR 11.3, Singapore 6.3, Kyrgyz Republic 5.1.
      - Strongest spillin ratio (total/second-round): Dominica 30.8, Malta 18.5, Slovak Republic 14.6.
    - Spillback of 0.6 percent of GDP on China itself in the import-side analysis (reduction of China’s imports on top of original 4.3 percent of GDP shock).

*Italic attribution: Content summarized from the provided IMF PDF content unit.*

### 52. Originated in China export and import shock spillovers have different profiles.

### 52. Originated in China export and import shock spillovers have different profiles.

### China: export shock spillover profile
- The export shock is the strongest at the first round: on average all China’s partners immediately lose about 1 percent of their GDP (Figure 11a).
- Once at the second round the drop in revenue from exports transforms into an import demand shock for countries that can pass it through and then again into an export revenue shock for their partners.
- The second through the fourth rounds add on average about 0.5 percent of GDP to the first round shock to arrive to the total shock magnitude after all four rounds of spillovers of about 2.3 percent of GDP on average for all countries.
- Countries capable of amplifying the shock would experience above average spillovers, compared to countries that absorb or block shocks.
- The export shock does not decay and remains broadly unchanged after four rounds.

### China: import shock spillover profile
- At the first round the spillover is zero for all countries, as only China itself experiences the shock (Figure 11b).
- On average, the shock magnitude at the second round is approximately twice as high (0.6 percent of GDP) as the shock at the third and fourth rounds (0.3 percent of GDP).
- Average spillovers through shock amplifiers are the highest at the second round, decaying fast but remaining persistent through the ensuing rounds.
- The spillovers through shock absorbers are at about the average level.
- Shock blockers do not pass through spillovers at all.
- After four iterations the import shock reached 1.2 percent of GDP, compared to the export shock of 2.3 percent of GDP.
- Contributing factors: the shock to China itself is not taken into account; about 70 countries block spillovers before they transform into the next import shock; not all import shocks lead to the next round of export revenue shocks for all countries; countries that amplify shocks play an important role at the second round.

### Shock in a medium-size country (Ukraine): network and trade structure
- Ukraine’s in/out degree is 151/156 out of maximum of 170, that is, Ukraine does not trade with at least 10 percent of countries in the world.
- Ukraine’s exports to main 11 partners amount to 60 percent of its exports.
- Ukraine’s imports from 11 main partners amount to 77 percent of its imports.
- Russia is dominant as Ukraine’s main export destination and the source of imports; China and Germany are important export destinations but much less so as import sources; Turkey, the Czech Republic, United States, and Hungary are among important export destinations but not top import partners; Egypt, India, and Spain are important import sources but not among the top export destinations.
- The Fruchterman-Reingold (1991) force-directed layout algorithm is used to visualize network structure; node areas are proportional to the share of a partner in Ukraine’s exports and imports and link widths are proportional to the value of trade in each direction.

### Ukraine: pass-through characteristics and immediate effects
- Most of Ukraine’s trading partners can pass through shocks.
- On the import side, all partners, with the exception of Belarus and Egypt, can pass through potential spillovers; Belarus and Egypt are peripheral in the network.
- The largest countries from where Ukraine obtains its imports, Russia and Kazakhstan, absorb part of the shock and Bulgaria blocks spillovers altogether.
- Turkey, Italy, Poland, and Spain can amplify their share of the shock.
- When at the second round the import shock transforms into an export shock, there will be even fewer blockers (only Belarus); all other countries would pass through the spillover by either amplifying or absorbing part of it.

### Ukraine: magnitudes and round profiles
- The shock originates in a drop of Ukraine’s imports: an 11 percent of GDP decline in Ukraine’s imports would spillover and be distributed among countries that export to Ukraine proportionally to Ukraine’s share in their exports.
- Considering only countries where the impact from the nominal shock in Ukraine is above the average, the first round direct spillover effects would amount on average to 0.2 percent of their GDP.
- The largest first-round impacts: Belarus of 2.2 percent of GDP (export revenue loss) and Lithuania of 0.6 percent of GDP.
- Taking into account four rounds, indirect spillovers between countries other than Ukraine would add additional 0.2 percent to export revenue losses to all its trading partners; the total average impact would then be about 0.4 percent of GDP.
- Example totals after four rounds: Belarus 2.7 percent of GDP; Lithuania about 1 percent of GDP.
- The absolute size of the spillin effect (difference between total spillover after four rounds and direct spillover after the first round) amounts on average to 0.2 percent of GDP.
- Strength of spillin effect (ratio of total spillover to spillover at the first round) is highest for medium-sized open economies such as Singapore, Hong Kong SAR, Sweden, and small oil-producers such as Equatorial Guiney, Republic of Congo, and Libya.
- There is a visible spillback effect on Ukraine itself of about 0.2 percent of its GDP.

### Ukraine: import shock into partners and downstream effects
- The import shock becomes nontrivial only starting from the second round, as the first round is the drop of Ukraine’s own imports.
- Shock blockers (e.g., Belarus, Egypt) will halt spillovers when affected by the drop of export revenue from the first round.
- At the second round the hardest impact will be on Lithuania, which would have to reduce its imports from all other countries by 0.7 percent of GDP; Poland, Estonia, Guyana would reduce their demand for further imports by 0.3 percent of GDP.
- Largest overall impact after four rounds: Kyrgyz Republic (1.1 percent of GDP), Lithuania (1 percent of GDP), Estonia and Latvia (about 0.8–1.0 percent of GDP).
- The average level of the spillin effect (difference between total spillover and spillover at the second round) is about 0.2 percent of GDP of affected countries; the largest for the Kyrgyz Republic of about 0.9 percent of GDP.
- The strongest spillin effect (ratio of total effect to the second round effect) is highest for medium-sized open economies such as Hong Kong SAR, Malta, Sweden, and Singapore.

### Ukraine: export vs import shock round profiles (averages)
- Export shock:
  - On average all partners immediately lose 0.05 percent of their GDP at the first round.
  - The second round adds on average a bit less than 0.03 percent of GDP.
  - The third and the fourth rounds contribute about 0.02 of GDP each.
  - Total shock magnitude after all four rounds: about 0.12 percent of GDP on average for all countries.
  - The export shock decays relatively fast and may not be statistically different from zero after four rounds.
- Import shock:
  - At the first round the spillover is zero for all countries.
  - Shock magnitudes at the second and third rounds are approximately between 0.02–0.03 percent of GDP.
  - At the fourth round the spillover decays to below 0.02 percent of GDP.
  - Total import shock: 0.07 percent of GDP.
  - Average spillovers through shock amplifiers are the highest and relatively persistent; absorbers are about the average; blockers do not pass through spillovers at all.
- The total size of the import shock to Ukraine’s trading partners is a little more than half of the export shock.

### Network findings and policy implications (Conclusions)
- A network model captures second round network effects of spillovers that can be substantial and often have been disregarded.
- Network effects originate from the feedback process starting from the second round; their strength depends on:
  - the relative magnitude of the initial shock at the epicenter,
  - the epicenter country’s centrality and other network properties,
  - the position of its main trading partners in the network,
  - their domestic economic structure,
  - the relative compounding strength of spillover signals spreading in the same direction,
  - the offsetting strength of signals spreading in opposite directions.
- When compounded through different rounds, network effects can become comparable to and often exceed the initial shock at the epicenter country.
- Classification of countries by spillover behavior:
  - About 20 percent of countries amplify spillovers.
  - About 40 percent are shock absorbers (pass through but reduce strength).
  - About 40 percent are spillover blockers (spillovers die out on reaching them).
- Most developed countries and major international trading nations tend to pass through shocks; most commodity exporters, in particular oil producing countries and low-income countries, tend to block spillovers.
- The capacity to amplify, absorb, or block shocks depends on development, openness to trade, and business environment: higher development/openness implies higher probability to pass through shocks.
- Macroeconomic policies of spillover-conducive countries are important for global stability:
  - Core spillover amplifying countries include the United States, Switzerland, Italy, Korea, and India.
  - Spillover absorbers with relatively high pass-through coefficients include Japan, Germany, France.
  - Coordinated macroeconomic policies of core countries can mitigate negative shock spillovers originated outside the core.
- The profile of shock spillovers largely depends on the epicenter country and pass-through characteristics of its key trading partners:
  - Roughly half of countries can transmit shocks, but only a few play an important role in diffusion.
  - A large country with non-transmitting first neighbors may see shocks die at the first round.
  - A modest shock in a medium-sized country with transmitting/amplifying neighbors can lead to major spillovers.
  - Direct trade connections are needed for the epicenter country to feel the spillback effect.
  - For small economies, the spillin effect can be substantial and rebound from just a few trading partners.

*Source: _wp15149 - 52. Originated in China export and import shock spillovers have different profiles.*

### 73. The strength and the profile of spillovers depend on the properties of the export-

### _wp15149 - 73. The strength and the profile of spillovers depend on the properties of the export-

### Key finding
- The strength and profile of shock spillovers depend on the properties of the export-import matrix.
- Shocks alter the network structure at each wave of the shock in one iteration.

### Mechanisms and network properties identified
- Macroeconomic policies can change shock spillovers by influencing export-import matrix properties, including:
  - in- and out-degrees
  - weighted centrality
  - in- and outflow weights
  - pass-through coefficients
  - other network properties

### Policy implications
- Policies that alter countries’ positions or linkages in trade and financial networks can amplify or dampen international shock transmission.
- Targeting elements of the export-import matrix (for example, reducing excessive dependence on a few partners or altering pass-through mechanisms) is a lever for changing spillover patterns.

### Research agenda
- Substantial further research is required to:
  - quantify how specific changes in in- and out-degrees, centrality measures, flow weights, and pass-through coefficients affect spillover magnitude and profile
  - model the dynamic evolution of the export-import matrix across successive shock waves and iterations
  - identify policy designs that optimally reshape network properties to limit adverse spillovers without introducing significant distortions

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15149.pdf*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15149.pdf_