## wpiea2024041-print-pdf

## Source details

**Canonical URL:** [wpiea2024041-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024041-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024041-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024041-print-pdf.pdf.json)

---

### Background and scope
- Focus: empirical investigation of cross-border trade spillovers from domestic subsidies using Global Trade Alert (GTA) data.
- Data coverage: 193 economies between 2009 and 2021; product targeting at HS6 level; records date of announcement, implementation, and removal; no systematic monetary values of subsidies; missing measures introduced before 2009.
- Definition: subsidies are government measures involving an unrequited financial transfer creating an advantage for beneficiaries (IMF, OECD, World Bank and WTO, 2022; UNCTAD, 2019). Paper considers subsidies given to firms (excludes consumption subsidies, export subsidies, and trade finance).

### Data sources, sample construction, and measurement
- Primary policy source: Global Trade Alert (GTA) database (launched end-2008); GTA records credible unilateral policy changes with implementing country, policy instrument (one among 60 types), targeted products (HS 6-digit), announcement/implementation/withdrawal dates, and expected impact on foreign commercial interests.
- Announcement year: earlier of announcement and inception; announcements after July 1st set to following year. Removal year similarly set (after June 31st counted as same year).
- GTA policy classification used:
  - Domestic subsidies: MAST chapters L and P7 (excluding consumption subsidies and export subsidies aggregated under export promotion).
  - Instrument groups: production subsidies; direct transfers (state aid, grants, capital injections); revenue-reducing policies (tax breaks); risk-transfer policies (loans, guarantees).
- Sample restrictions:
  - Policies lacking product detail (mostly services) excluded.
  - GTA horizontal policies (~6 percent of interventions) omitted.
  - Non-governmental policies (~20 percent) excluded; focus on governmental national and supranational policies.
  - Only GTA policies evaluated as “distortive” are the primary focus (across all GTA policies: about 30 percent liberalizing, 7 percent neutral, 63 percent distortive; 82 percent of domestic subsidies in GTA are classified as distortive).
  - Small countries with average population < 1 million over 2009–2021 excluded.
- Trade data:
  - Product-level exports and imports: CEPII BACI through 2021 (HS 2007 → HS 2012).
  - Gravity/bilateral/internal trade: ITPD-E through 2019, aggregated to ISIC Revision 3 2-digit.
  - Bilateral determinants and macro variables from CEPII Gravity database.
- Measurement choices:
  - Subsidy presence measured with a dummy for at least one active intervention in country-product-year.
  - Legacy subsidies introduced before 2009 are unobserved.
  - GTA does not report monetary values of subsidies.

### Empirical approaches
- Difference-in-difference (product-level):
  - Compares changes in exports and imports of a product targeted by subsidies with non-targeted products within the same industry.
  - Baseline specification (log trade value):
    - ln(X_ik,t) = β1 S_ik,t + β2 IP_ik,t + α_ik + α_ik × t + δ_ic,t + μ_k,t + ε_ik,t
    - Fixed effects: country-product, country-product linear trends, country-sector-year, product-year.
  - Identification challenges: endogeneity, selection, pre-trends, lack of monetary subsidy values, legacy subsidies.
- Gravity model (industry-level):
  - Aggregated origin–destination–industry sales including domestic flows; estimates differential effect of subsidies on international (INT=1) versus domestic (INT=0) sales.
  - Specification: X_i j k,t = exp( β S_i k,t × INT_i j + γ GRAV_i j k,t + δ_i k,t + μ_j k,t + ε_i j k,t )
  - Estimation with PPML; controls include exporter-industry-year and importer-industry-year fixed effects, bilateral gravity controls, WTO/PTA/FTA dummies; clustering by importer, exporter and symmetric country pairs.
  - Identification: β measures how subsidies shift international relative to domestic sales.

### Descriptive facts on subsidies (GTA data; Section III stylized facts)
- Fact 1: Domestic subsidies have been on the rise since 2009.
  - Number of subsidy announcements: from 760 in 2009 to about 3000 in 2021.
  - Subsidy share of all GTA policy announcements: increased from 29 to 60 percent between 2009 and 2021; reached 50 percent by 2019.
  - Domestic subsidy share of GTA policies in force: rose from 25 to 45 percent between 2009 and 2021.
  - Cumulative tally of subsidy interventions by end of sample: around 14,000 globally.
  - Pronounced increases in 2020 and 2021 likely linked to COVID crisis responses.
- Fact 2: Subsidies concentrated in G20 economies.
  - In any year 2009–2021, number of subsidy policies in force in non-G20 economies is at most 2 percent of number observed in G20 economies.
  - G20 EMs list: Argentina, Brazil, China, India, Indonesia, Mexico, Russia, Saudi Arabia, Türkiye, South Africa.
  - G20 AEs list: Australia, Canada, Republic of Korea, the EU, Japan, United States.
  - Domestic subsidy share for the average G20 EM reached 67 percent in 2021, consistently above the average G20 AE.
  - 2020–2021: G20 AEs experienced a surge in subsidy announcements (COVID response); G20 EMs did not mirror the same pattern.
- Fact 3: Instrument shift toward direct transfers and loans.
  - Direct transfers (state aid, grants, capital injections, in-kind grants) dominate recent announcements, especially 2020–2021.
  - Loans and risk-transfer policies increased in 2020–2021.
  - Production subsidies remain marginal relative to direct transfers and risk-transfer instruments.
- Fact 4: Manufacturing increasingly targeted.
  - Manufacturing share of domestic subsidy policies in force rose from 25 to 45 percent over the sample.
  - By 2021, approximately 600 subsidy policies affect manufacturing products (about half of all subsidies in force).
  - Industries with highest counts by end of sample include agriculture, food, chemical, machinery, and motor vehicles.
  - Largest increases between 2009 and 2021 (relative to 2009), excluding tobacco: apparel and medical and optical equipment.

### Main empirical findings — product-level difference-in-differences (baseline and dynamic)
- Incidence of GTA policies in sample:
  - Around 30 percent of country-product combinations receive at least one GTA policy during 2009–2021.
  - 9 percent receive domestic subsidies during the sample period.
  - Conditional on receiving a subsidy in a year, median country-product receives 1 subsidy; mean receives 3 subsidy interventions.
- Baseline DID (full sample):
  - Exports of targeted products are 2 percent higher after the subsidy than before, relative to other products.
  - Imports of targeted products are 4 percent higher after the subsidy than before, relative to other products.
  - Other GTA policy indicators: export restrictions significantly inhibit exports; technical barriers, government procurement policies, temporary import restrictions and macro policies have negative effects on imports.
- Event-study (dynamic) evidence:
  - Positive effects on exports and imports persist up to 8 years after subsidy introduction in event-time plots.
  - Strong pre-trends: exports and imports of products slated to receive subsidies increase relative to other products years before announcement—violating strict parallel trends.
  - Average post coefficients from dynamic specification:
    - Average effect on exports = 2.3 percent (standard error = 0.7 percent).
    - Average effect on imports = 3 percent (standard error = 0.7 percent).
  - Interpretation: pre-trends indicate selection into treatment (political economy); full-sample dynamic evidence implies subsidies do not sustain export growth relative to comparators and do not produce import-substitution at product level.

### Heterogeneity by country group (G20 EMs vs G20 AEs)
- G20 EMs:
  - Subsidies durably boost exports on both intensive and extensive margins with limited pre-trends.
  - Estimated increases: exports of targeted products rise by 7.6 percent; probability of exporting a product increases by 2.2 percentage points relative to other products.
  - The 7.6 percent intensive-margin export effect is more than twice the average yearly product-level export change for G20 EMs (3.1 percent).
  - Spillovers strongest for trade with non-G20 economies: exports to non-G20 markets increase and imports from non-G20 economies of targeted products fall in G20 EMs.
- G20 AEs:
  - Results align with full-sample patterns: exports and imports of subsidized products are higher but show significant pre-subsidy increases (selection effects).
  - After bias correction, effects on exports in G20 AEs become small and not significant.

### Intensive vs extensive margins (probability of exporting/importing)
- Full sample and G20 AE:
  - Insignificant or very small coefficients on the extensive margin; negligible effects on probability of exporting/importing.
- G20 EMs:
  - Extensive-margin effects are strong:
    - Probability a product is exported increases by 2.2 percentage points after receiving subsidies (relative to other products).
    - Average effect on probability of importing is 1.5 percentage points.
    - These represent 2.3 percent of the average probability of exporting and 1 percent of the average probability of importing, respectively.
    - Export probability increases to other G20 EMs: 2.8 percentage points (4 percent of average probability); to non-G20s: 2.2 percentage points (7 percent of average probability).
  - Interpretation: for G20 EMs, subsidies raise exports on both intensive and extensive margins, especially toward non-G20 destinations.

### Corrections for bias and alternative estimators
- Implements imputation approach of Borusyak, Jaravel, and Spiess (2024) to address bias from staggered treatments.
- Imputation approach results:
  - Full sample and G20 AE: effect of subsidies on exports becomes small and not significant; import effect similar to baseline but contaminated by pre-trends (pre-subsidy coefficients reject zero).
  - G20 EMs: export effect remains positive and larger than in full/G20 AE samples, and not contaminated by pre-trends (though smaller than original baseline).

### Industry- and instrument-level heterogeneity
- Industry effects:
  - Robust pro-trade effects in electrical machinery, apparel, textile, textile, furniture, chemicals, metals, machinery, and motor vehicles in various specifications.
  - Manufacturing subsample mirrors full-sample baseline; primary sector coefficients not significant in product-level DID.
- Instrument effects:
  - Tax breaks (revenue-losing policies) exhibit the most robust positive influence on exports and imports.
  - Direct transfers display a negative coefficient on trade in some specifications.
  - Caveat: GTA lacks monetary sizes—differences across instruments may reflect average magnitude differences rather than pure instrument mechanics.

### Gravity model results — international relative to domestic trade (spillovers)
- Baseline international border effect: international trade is 51 percent lower than domestic trade (Intl. trade flows coefficient: -0.722*** (0.065)).
- Effect of subsidies on international vs domestic trade:
  - Interaction of international dummy and industry subsidy dummy indicates subsidies increase international trade relative to domestic sales.
  - The gap between international and domestic trade diminishes from 51 percent to a 38 percent difference when an industry is targeted by subsidies (subsidies (all) coefficient examples: 0.244*** (0.028); domestic subsidies: 0.229*** (0.028); other specs 0.119*** (0.027); 0.093*** (0.035)).
  - Impact driven entirely by domestic subsidies; export promotion policies show no effect in columns (3)–(5).
  - Controlling for other GTA policies halves the positive effect but it remains statistically significant.
- Heterogeneity:
  - Pro-trade effect largest for trade between different country groups (G20 AEs, G20 EMs, non-G20).
  - Sectoral heterogeneity: interaction effects larger in primary sector; in manufacturing several industries (machinery, furniture, metals, electrical machinery and apparatus) show positive or significant spillovers.
- Robustness:
  - Replacing subsidy dummy with instrument dummies: revenue-losing policies (tax breaks) have the largest international vs domestic impact.
  - Excluding China or the U.S. leaves results robust (e.g., Intl. trade × Domestic subsidies coefficients: No CHN 0.132*** (0.027); No USA 0.141*** (0.030)).
  - Including “neutral” or “liberalizing” policies or NFIs implementation yields consistent positive interaction coefficients.

### Robustness checks and sensitivity
- Inclusion of GTA-coded “neutral” or “liberalizing” policies leaves point estimates virtually unchanged (only 3.5 percent of country-policy-product combos coded as “liberalizing”).
- Adding national financial institutions (NFIs) as implementers: coefficients similar to baseline; NFIs affect 2.5 percent of product-level sample vs 15 percent at intervention level.
- Omitting influential countries:
  - U.S. had 2928 subsidy announcements in force since 2009 (23 percent of subsidies announced globally) as of 2021; China accounts for 39 percent of all subsidies in force in 2021.
  - Results confirmed when dropping U.S. and China: point estimates virtually unchanged.
- Count-based robustness (inverse hyperbolic sine of counts): significance can change when controlling for other GTA policies; interpretation complicated.

### Interpretation and policy implications
- Empirical synthesis:
  - Product-level DID: subsidies associated with higher exports and imports of targeted products in full sample, but strong pre-trends complicate causal interpretation.
  - Gravity analysis: subsidies reallocate sales toward international markets, shrinking the international–domestic sales gap by 16 percent in baseline interpretation (from a 51 percent international penalty to a 38 percent penalty).
  - For G20 EMs, subsidies raise exports on both intensive and extensive margins; for G20 AEs, export effects are small or driven by pre-trends.
  - Spillovers concentrated in certain industries (electrical machinery highlighted) and largest between different country groups (e.g., G20 vs non-G20).
  - Tax breaks and revenue-losing policies show the largest spillover effects.
- Welfare and policy considerations:
  - Domestic subsidies generate significant trade spillovers and can distort international level playing field.
  - Spillovers may provoke retaliatory tit-for-tat strategies (subsidy wars); welfare consequences of subsidy-induced retaliation are complex and unresolved in this study.
  - Enhanced multilateral cooperation and clearer rules of conduct may be warranted to prevent detrimental retaliatory actions.

### Research and data recommendations
- Further work needed to address endogeneity and political economy drivers of subsidy selection.
- Need for data on monetary size of subsidies to assess fiscal ramifications and intensive-margin dynamics; even limited country coverage would aid mechanism identification.
- The study calls for theoretical and empirical advances to better understand firm selection into subsidies, the role of lobbying, and cross-border retaliation dynamics.

*Italic: Source — wpiea2024041-print-pdf (Trade Spillovers of Domestic Subsidies — Working Paper No. WP/2024/041).*

### References .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .

### wpiea2024041-print-pdf - References

### Background and scope
- Focus: empirical investigation of cross-border trade spillovers from domestic subsidies using Global Trade Alert (GTA) data.
- Data coverage: 193 economies between 2009 and 2021; product targeting at HS6 level; records date of announcement, implementation, and removal; no systematic monetary values of subsidies; missing measures introduced before 2009.
- Definition: subsidies are government measures involving an unrequited financial transfer creating an advantage for beneficiaries (IMF, OECD, World Bank and WTO, 2022; UNCTAD, 2019). Paper considers subsidies given to firms (excludes consumption subsidies, export subsidies, and trade finance).

### Empirical approaches
- Difference-in-difference (product-level):
  - Compares changes in exports and imports of a product targeted by subsidies with non-targeted products within the same industry.
  - Identifies export and import side effects but may be confounded by product-specific dynamics and pre-trends.
- Gravity model (industry-level):
  - Estimates subsidies’ effect on international relative to domestic trade, netting out exporter-, importer-, and bilateral-specific shocks.
  - Assesses how subsidies reallocate sales across domestic and international markets (trade spillovers).

### Descriptive facts on subsidies (GTA data)
- Global use of subsidies has substantially increased since 2009.
- Approximately 60 percent of all distortive interventions recorded in the GTA database by the last year of the sample took the form of subsidies.
- Sectoral shift: more subsidies in manufacturing and fewer in primary industries.
- Rising role of G20 emerging markets (G20 EMs): the domestic subsidy share of all GTA policies has risen to 67 percent for the average G20 EM in 2021, consistently surpassing the equivalent share for the average G20 advanced economy (G20 AE).
- Policy instruments: direct transfers (grants and state aid) constitute the primary type of instrument; their importance increased in 2020 and 2021.
- Tax breaks, grants, state aid, and loans are noted instrument types; tax breaks (policies entailing a loss of government revenue) show the most robust positive influence on trade in the analysis.

### Main empirical findings — difference-in-difference estimates
- Full sample:
  - Exports of subsidized products are 2 percent higher after the subsidy than before, relative to other products.
  - Imports of subsidized products are approximately 4 percent higher after the subsidy than before, relative to other products.
  - Strong pre-trends: governments direct subsidies to products already experiencing increases in both exports and imports.
  - Dynamic effects: exports and imports of subsidized products, while higher after the subsidy, stop growing relative to other products; no evidence of import-substitution at the product level.
  - Interpretation: on average, subsidies are not able to shift comparative advantage patterns.
- Heterogeneity by country group:
  - G20 EMs:
    - Subsidies durably boost exports on both intensive and extensive margins without significant pre-trends.
    - Imports react weakly.
    - Estimated effects: increases in value of exports of targeted products by 7.6 percent and increase in probability of exporting a product by 2.2 percentage points, relative to other products.
    - The intensive-margin export effect is equivalent to more than twice the average yearly change in exports from G20 EMs at the product level (3.1 percent growth).
    - Spillovers strongest for trade with non-G20 economies: exports of targeted products increase in non-G20 markets and imports from non-G20 economies of targeted products fall in G20 EMs.
  - G20 AEs:
    - Results align with full-sample patterns: exports and imports of subsidized products are higher and show pre-subsidy increases (selection effects).

### Main empirical findings — gravity estimates (international vs. domestic trade)
- When an exporter implements a subsidy, the disparity between international trade and domestic sales in the industry diminishes by 16 percent.
- Subsidies increase international trade relative to domestic sales, indicating trade spillovers and outward-oriented policy effects.
- Heterogeneity: subsidies increase international trade especially between countries of different groups (between G20 AEs, G20 EMs, and non-G20 countries) versus domestic trade, highlighting potential for contentious cross-country responses.

### Industry- and instrument-level heterogeneity
- Industry effects:
  - Robust pro-trade effects in the electrical machinery industry, and in apparel and textile industries.
- Instrument effects:
  - Tax breaks exert the most robust and positive influence on trade compared to state aid, grants, and loans.
  - Caveat: GTA data lack monetary size of subsidies, so differences across instruments may partly reflect differences in average magnitude.

### Interpretation and policy implications
- Domestic subsidies generate significant trade spillovers by reallocating sales toward international markets and altering trade between country groups.
- Such spillovers can distort the international level playing field and may provoke retaliatory tit-for-tat strategies (subsidy wars).
- Enhanced multilateral cooperation may be warranted to prevent detrimental retaliatory actions; the welfare consequences of subsidy-induced retaliation are complex and not resolved in this study.

### Contribution and relation to literature
- Provides cross-country, multi-industry empirical evidence on subsidies’ effects on exports, imports, and international versus domestic sales.
- Complements country-specific case studies and theoretical models on optimal subsidies in open economies.
- Aligns with prior work finding subsidies can boost exports and reallocate production; extends analysis to compare international relative to domestic trade flows using broader GTA coverage.

*Source: wpiea2024041-print-pdf - References (GTA-based analysis of subsidies and trade; 193 economies, 2009–2021).*

### Section III presents some stylized facts about the variation in the use of subsidies over time,

### wpiea2024041-print-pdf - Section III presents some stylized facts about the variation in the use of subsidies over time,

### Data: sources, sample construction, and measurement
- Primary policy source: Global Trade Alert (GTA) database, launched at the end of 2008; records credible unilateral policy changes with details including implementing country, policy instrument (one among 60 types, including subsidies), targeted products (HS 6-digit), announcement dates, implementation dates, withdrawal dates, and expected impact on foreign commercial interests (distortive, neutral or liberalizing).
- Empirical sample: exploits GTA policy announcements from 2009 onward; announcement year defined as the earlier of announcement and inception, with announcements after July 1st set to the following year; removal year set similarly (after June 31st counted as same year, otherwise prior year).
- Definition of subsidies: policies classified under MAST chapters L (domestic policies) and P7 (export promotion); excludes consumption subsidies; domestic subsidies include corporate subsidies except export subsidies; export promotion policies (export subsidies, trade finance, other export incentives) are lumped together.
- Further instrument categorization (within domestic subsidies): production subsidies; subsidies transferring resources to firms (direct transfers / state aid / grants); subsidies resulting in losses in government revenues (tax breaks); policies where government assumes risk (loans, loan guarantees, state loans, interest payment subsidies).
- Sample restrictions and exclusions:
  - Policies lacking product detail (mostly services) excluded.
  - Policies classified as horizontal by GTA (6 percent of interventions) omitted.
  - Non-governmental policies (e.g., by financial institutions, ~20 percent of sample) excluded; focus on governmental national and supranational policies.
  - Only GTA policies evaluated as “distortive” are the primary focus (across all GTA policies, about 30 percent are liberalizing, 7 percent neutral, and 63 percent distortive; 82 percent of domestic subsidies in GTA are classified as distortive).
  - Small countries excluded: those with average population < 1 million over 2009 - 2021.
- Trade data:
  - Product-level exports and imports: CEPII BACI dataset through 2021 (HS 2007 converted to HS 2012 via UN correspondence).
  - Gravity/bilateral trade (and internal trade): ITPD-E database through 2019, covering 170 industries; aggregated to ISIC Revision 3 2-digit sectors to align with GTA product aggregation.
  - Bilateral determinants (distance, colonial relationship, contiguity, common language), GDP, population from CEPII Gravity database.
- Measurement choices:
  - Presence of subsidies measured with a dummy for at least one active intervention in country-product-year (favored over counts to reduce intensive-margin measurement error).
  - Legacy subsidies introduced before 2009 are not observed; analysis exploits variation in presence of subsidies announced from 2009 onward.
  - GTA does not report monetary values of subsidies.

### Stylized facts on the variation in the use of subsidies (Section III)
- Fact 1: Domestic subsidies have been constantly on the rise since 2009.
  - Number of subsidy announcements: from 760 in 2009 to about 3000 in 2021.
  - Subsidy share of all GTA policy announcements: increased from 29 to 60 percent between 2009 and 2021; reached 50 percent by 2019.
  - Domestic subsidy share of GTA policies in force: rose from 25 to 45 percent between 2009 and 2021.
  - Cumulative tally of subsidy interventions by end of sample: around 14,000 globally.
  - Note: pronounced increases in 2020 and 2021 likely linked to COVID crisis policy responses.
- Fact 2: Subsidies are largely used by both advanced and emerging G20 economies.
  - In any given year 2009–2021, number of subsidy policies in force in non-G20 economies is at most 2 percent of the number observed in G20 economies.
  - G20 groupings used:
    - G20 EMs: Argentina, Brazil, China, India, Indonesia, Mexico, Russia, Saudi Arabia, Türkiye, and South Africa.
    - G20 AEs: Australia, Canada, Republic of Korea, the EU, Japan, and the United States.
  - Both G20 EMs and G20 AEs rapidly increased subsidy interventions between 2009 and 2021.
  - Domestic subsidy share of all GTA policies for the average G20 EM escalated to 67 percent in 2021, consistently surpassing the corresponding share for the average G20 AE.
  - In 2020–2021, G20 AEs experienced a surge in subsidy announcements (COVID response), whereas G20 EMs did not follow the same pattern.
- Fact 3: State aid and grants have become more common in recent years.
  - Domestic subsidy instruments broken into four groups: production subsidies; direct transfers (state aid and grants); revenue-reducing policies (tax breaks); and risk-transfer policies (loans, loan guarantees).
  - Direct transfers (state aid, grants, capital injections, in-kind grants) dominate recent announcements, especially in 2020–2021.
  - Loans and other risk-transfer policies also increased in popularity in 2020–2021.
  - Production subsidies play a marginal role relative to direct transfers and risk-transfer instruments.
  - Export promotion policies (trade finance and export subsidies) are a smaller component; most export promotion policies are trade finance interventions, but announcements in this area are overall declining over time (see appendix referenced in source).
- Fact 4: Subsidies increasingly target manufacturing industries.
  - Manufacturing share of domestic subsidy policies in force rose from 25 to 45 percent over the sample period.
  - By 2021, approximately 600 subsidy policies affect manufacturing products (about half of all subsidies).
  - Primary sector: 278 subsidy policies in 2009 vs. 158 measures announced in manufacturing in 2009; cumulative primary-sector subsidies since 2009 total 2391 (constituting 40 percent of subsidy policies in force in manufacturing by 2021).
  - Industries with highest number of domestic subsidy policies by end of sample: agriculture, food, chemical, machinery, and motor vehicles.
  - Largest increases between 2009 and 2021 (relative to 2009 levels), excluding tobacco: apparel and medical and optical equipment.
  - Missing product information pertains to services or interventions for which targeted product lists could not be collected.

### Empirical strategy overview
- Two complementary approaches to identify trade responses and spillovers of domestic subsidies:
  1. Product-level difference-in-differences (DID) using HS 6-digit product-country-year observations.
     - Baseline empirical specification (log trade value as dependent variable):
       ln(X_ik,t) = β1 S_ik,t + β2 IP_ik,t + α_ik + α_ik × t + δ_ic,t + μ_k,t + ε_ik,t
       - X: exports from or imports by country i in HS 6-digit product k at time t.
       - S: dummy for presence of at least one subsidy policy in country i, product k, year t.
       - IP: indicators for presence of other GTA policies (aggregated categories).
       - α_ik: country-product fixed effects.
       - α_ik × t: country-product linear time trends.
       - δ_ic,t: country-sector-year fixed effects (sector = ISIC 2-digit).
       - μ_k,t: product-year fixed effects.
       - Standard errors clustered by country and product.
     - Identification challenges discussed:
       - Endogeneity due to productivity shocks, lobbying, and selection of products into treatment.
       - GTA records policy changes but not monetary values; presence/absence dummy used.
       - Legacy subsidies prior to 2009 not observed; may attenuate estimated effects.
  2. Gravity estimations (bilateral) using ITPD-E aggregated at ISIC Revision 3 2-digit level (described in source; complementary but not elaborated in detail in the provided excerpt).
- Data coverage specifics reiterated:
  - Product-level DID uses CEPII BACI exports/imports through 2021.
  - Gravity analysis uses ITPD-E through 2019, 170 industries.
  - Country-sector-year fixed effects control for shocks specific to a sector within a country and over time.

### Product-level difference-in-differences results (main findings)
- Incidence of GTA policies in sample:
  - Around 30 percent of country-product combinations receive at least one GTA policy during 2009–2021.
  - 9 percent receive domestic subsidies during the sample period.
  - Conditional on receiving a subsidy in a year, median country-product receives 1 subsidy; mean receives 3 subsidy interventions.
- Average (baseline) DID estimates (summary interpretation):
  - Introducing subsidies is associated with higher exports and imports at the product level.
  - Estimates with full fixed effects and country-product linear time trends indicate:
    - Exports in targeted products are 2 percent higher than in comparable non-targeted products (post-introduction).
    - Imports in targeted products are 4 percent higher than in comparable non-targeted products (post-introduction).
  - Domestic subsidies drive the positive effects on trade; export promotion policies show no discernible impact in these product-level estimates.
  - Other GTA policy indicators: export restrictions significantly inhibit exports; technical barriers to trade, government procurement policies, temporary import restrictions and macro policies have significant and negative effects on imports (detailed coefficients in appendix referenced in source).
- Event-study (dynamic) evidence and diagnostics:
  - Event-study specification estimates period-specific coefficients from p = -12 to p = +12 years around subsidy announcement.
  - Two salient patterns emerge:
    - Positive effects on exports and imports persist up to 8 years after subsidy introduction.
    - Compelling evidence of pre-trends: exports and imports in products slated to receive subsidies increase relative to other products years before the subsidy announcement—violating strict parallel trends.
  - Average post coefficients (weighted averages of post-period coefficients) from dynamic specification:
    - Average effect on exports = 2.3 percent (standard error = 0.7 percent).
    - Average effect on imports = 3 percent (standard error = 0.7 percent).
  - Interpretation caution: pre-trends suggest product selection into treatment via political economy mechanisms (e.g., “winners picking government policy” or “losers picking government policy”), complicating causal interpretation. Product-level pre-trends indicate targeted products were already experiencing differential trade dynamics relative to others within the same sector prior to subsidy introduction.
  - Overall implication from full-sample dynamic coefficients: subsidies are unable to sustain export growth and fail to curb imports in targeted products when accounting for pre-trends.

### Heterogeneity by country group: G20 AEs vs G20 EMs
- Separate event-study estimates by country group reveal divergent patterns:
  - G20 AEs: estimates resemble full-sample results, including significant pre-trends and limited sustained positive export effects post-subsidy.
  - G20 EMs: subsidies have a positive and substantial effect on exports with little evidence of significant pre-trends; effect increases over time.
    - Estimated average export effect for G20 EMs after receiving the subsidy: 7.6 percent higher exports in targeted products relative to other products.
    - Contextual magnitude: the average yearly percent change in exports from G20 EMs at the product level is 3.1 percent (the 7.6 percent effect exceeds twice this average yearly change).
- Interpretation: heterogeneity suggests that subsidies may be more effective at generating export gains in G20 emerging economies than in G20 advanced economies, though selection and political economy dynamics remain relevant.

### Additional empirical and inferential notes
- Fixed effects strategy:
  - Country-product fixed effects and country-product linear trends absorb time-invariant and linear time-varying product-country factors.
  - Country-sector-year fixed effects control for sector-specific shocks and reporting biases that vary over time within country-sectors.
  - Product-year fixed effects capture global shocks to HS 6-digit products.
- Limitations emphasized in source:
  - GTA lacks monetary subsidy values, limiting analysis of intensive-margin effects.
  - Reporting and coding biases in GTA possible; rich fixed effects mitigate biases specific to country, sector, and product.
  - Legacy subsidies pre-2009 unobserved; may attenuate estimated effects by attributing weaker change when subsidies were already present historically.
  - Selection into treatment and political economy factors generate pre-trends complicating causal claims.

*Italic source: Extracted from wpiea2024041-print-pdf (Section III and associated Data, Empirical Analysis, and Results) as provided.*

### 37.2 percent to the non-subsidized products, receiving subsidies boosted exports by a signif-

### Effects of domestic subsidies on trade flows (product-level findings)

### Main empirical findings from difference-in-difference and event-study analyses
- Receiving subsidies boosted exports by a significant 3 percentage points relative to the non-subsidized products (baseline comparison reported in the text).
- For G20 emerging markets (G20 EMs):
  - Exports increased strongly and the positive effect is not driven by pre-trends.
  - Exports are especially driven by shipments to non-G20 countries; subsidies curb imports from non-G20 economies.
- For G20 advanced economies (G20 AEs):
  - Subsidies increase exports relative to non-targeted products only to other G20 AEs.
  - The positive but declining effect on imports is concentrated on imports originating from G20 EMs and shows evidence of significant pre-trends.
- Pre-trends are present in the full sample: products with rising exports and imports are more likely to receive subsidies, even after controlling for country-product linear time trends.
- The average effect on imports in the full sample is on average negative but poorly estimated and contaminated by significant pre-trends – imports in subsidized products relative to those in other products peak the year before the subsidy and then go down to their previous level.

### Corrections for potential bias and alternative estimators
- Recognizing bias in standard two-way fixed-effects difference-in-difference estimators with heterogeneous and staggered treatments, the study implements the imputation approach of Borusyak, Jaravel, and Spiess (2024).
  - Method: regress outcome on fixed effects and controls in the non-treated sample, predict treated outcomes, and compute observation-specific treatment effects.
- Results of the imputation approach:
  - In the full sample and in the G20 AE subsample, the effect of subsidies on exports becomes small and not significant; the effect on imports is similar to baseline but still affected by pre-trends (pre-subsidy coefficients reject zero).
  - For G20 EMs, the export effect remains positive, larger than in full and G20 AE samples, and not contaminated by pre-trends, although smaller than the original baseline.

### Heterogeneity across sectors, industries, and policy instruments

### Sector and industry patterns
- Sector split:
  - Manufacturing sector: baseline estimates are essentially identical to the full sample; manufacturing received most subsidies in recent years.
  - Primary sector: effects on exports and imports are not significant.
- Industry-level (ISIC 2-digit) findings within manufacturing:
  - None of the coefficients on exports are negative and significant; positive association between product-level exports and subsidies holds.
  - Most robust positive export relationships observed in: textile, furniture, chemicals, and apparel industries.
  - On the import side, positive and significant effects are also found in equipment and machinery industries.

### Comparative advantage and targeted products
- Constructed a revealed comparative advantage (RCA) index (based on Leromain and Orefice (2014) using a gravity-derived productivity estimation with trade elasticity set to 6.53).
- Difference-in-difference results by RCA subsamples:
  - Positive baseline export effect is driven by products with RCA>1 (comparative advantage).
  - Export effect is insignificant for products with RCA<1.
  - Positive import effects of subsidies are confirmed in both RCA subsamples.
- Interpretation: subsidies reinforce existing comparative advantage patterns rather than reversing them.

### Heterogeneity by subsidy instrument
- Domestic subsidy category split into four groups: production subsidies; direct transfers to firms (excluding production subsidies); policies transferring risk to the government; policies leading to losses in government revenues.
- Findings by instrument:
  - Tax breaks and other revenue-reducing policies (losses in government revenues) exhibit substantial positive impacts on both exports and imports.
  - Direct transfers display a negative coefficient on trade.
- Caveat: observed instrument differences may reflect differences in intervention intensity (e.g., monetary value of tax breaks typically larger than direct transfers) rather than pure instrument-specific mechanisms.

### Robustness checks and additional analyses

- Including GTA-coded “neutral” or “liberalizing” policies in the policy set leaves point estimates virtually unchanged.
  - Note: only 3.5 percent of country-policy-product combinations are coded as “liberalizing” in the GTA database.
- Adding national financial institutions (NFIs) as implementing jurisdictions produces coefficients on all subsidies and domestic subsidies very similar to baseline.
  - Only 2.5 percent of the product-level sample is affected by NFIs versus 15 percent at the intervention level, implying NFIs’ policies are more targeted.
- Omitting influential countries:
  - As of 2021, the U.S. has 2928 subsidy announcements in force since 2009 (representing 23 percent of subsidies announced globally), while China accounts for 39 percent of all subsidies in force in 2021.
  - Results are confirmed when dropping the U.S. and China: point estimates remain virtually unchanged.

### Extensive-margin (probability of exporting/importing) responses

- Intensive-margin versus extensive-margin:
  - Main estimates capture intensive-margin responses (conditional on positive trade flows).
  - Extensive-margin (entry into new exported/imported products) is analyzed via dummies for strictly positive trade flows (linear probability models).
- Full-sample and G20 AE results:
  - Insignificant or very small coefficients on subsidies for the extensive margin; negligible effects on the probability of exporting/importing.
- G20 EM results:
  - Strong effects on the extensive margin:
    - Products are, on average, 2.2 percentage points more likely to be exported from G20 EMs after receiving subsidies than before, relative to other products.
    - Average effect on the probability of importing is 1.5 percentage points.
    - These represent 2.3 percent of the average probability of exporting and 1 percent of the average probability of importing, respectively.
  - Distribution by destinations for G20 EMs:
    - Increase in export probability to other G20 EMs: 2.8 percentage points (4 percent of the average probability of a G20 EM exporting to other G20 EMs).
    - Increase in export probability to non-G20s: 2.2 percentage points (7 percent of the average probability of a G20 EM exporting to non-G20 countries).
    - Import probability effects are most pronounced for countries outside the G20.
- Interpretation: for G20 EMs, subsidies raise exports notably along both the intensive and extensive margins, particularly toward non-G20 destinations.

### Gravity-model approach: international relative to domestic trade (spillovers)

- Motivation:
  - Product-level DD estimates exhibit pre-trends and cannot directly measure displacement of domestic sales by international sales.
  - A gravity framework can control for time-varying shocks specific to country-industry and identify how subsidies affect international relative to domestic trade flows.
- Gravity specification (aggregate industry-level sales data, international and domestic flows):
  - X_i j k,t = exp( β S_i k,t × INT_i j + γ GRAV_i j k,t + δ_i k,t + μ_j k,t + ε_i j k,t )
    - X: value of sales from origin i to destination j in industry k in year t.
    - S_i k,t: dummy equal to one if origin country has at least one subsidy policy announced since 2009 that is in force in industry k at year t.
    - INT_i j: indicator equal to one for international flows (zero for domestic flows).
    - β measures differential effect of subsidies on international vs domestic trade (i.e., displacement/spillover).
  - GRAV contains bilateral trade determinants (e.g., distance, contiguity, common language, colonial ties, WTO/PTA membership); baseline replaces time-invariant bilateral shifters with asymmetric country-pair-industry fixed effects.
- Identification:
  - Because subsidies are country-industry-year specific, origin-industry-year fixed effects preclude estimating a direct effect on bilateral trade but allow estimation of β (interaction with INT) to identify international vs domestic differential effects.

*Source: wpiea2024041-print-pdf (excerpt provided).*

### introduction of subsidies (Head and Mayer, 2014; Piermartini and Yotov, 2016). To further

### Effects of subsidies on international relative to domestic trade flows (gravity model results)

### Methodology and identification
- Gravity equation estimated with the PPML estimator of Silva and Tenreyro (2006) to include zeros and account for heteroskedasticity.
- Regression controls:
  - Interactions between the international trade flows dummy and indicators for other GTA policies, GDP, GDP per capita, and country-specific dummies for membership in the EU and WTO.
  - Importer-industry-year and exporter-industry-year fixed effects (origin-industry-year and destination-industry-year fixed effects) capturing multilateral resistance, output and expenditure, and average effect of other country-specific variables.
  - Asymmetric country-pair-industry fixed effects in columns (2)–(5).
  - Dummies for FTA and WTO memberships; bilateral distance (in logs) and dummies for contiguity, common official language, colonial relationship post 1945 (in some columns).
- Standard errors clustered by importer, exporter and symmetric country pairs.
- The gravity model identifies effects on international trade (exports plus imports) relative to domestic trade; directional trade effects (exports vs imports) are not separately estimated due to directional fixed effects.

### Main quantitative findings (summary)
- Baseline international border effect:
  - International trade is 51 percent lower than domestic trade (column (1) Intl. trade flows coefficient implies a 51 percent difference).
- Effect of subsidies on international vs domestic trade:
  - Interaction between the international trade dummy and an indicator for the industry being targeted by at least one subsidy policy introduced since 2009 indicates international trade increases with subsidies relative to domestic sales.
  - The gap between international and domestic trade shrinks to a 38 percent difference (from the 51 percent average difference).
- Role of subsidy types and other policies:
  - Impact driven entirely by domestic subsidies; export promotion policies show no effect in columns (3)–(5).
  - Controlling for other GTA policies halves the positive effect of subsidies, but it remains statistically significant.
- Table 2 reported coefficients (selected):
  - Intl. trade flows coefficient: -0.722*** (0.065)
  - Subsidies (all) coefficient: 0.244*** (0.028)
  - Domestic subsidies coefficient(s): 0.229*** (0.028); 0.119*** (0.027); 0.093*** (0.035) across specifications
  - Export promotion coefficients: 0.006 (0.043); 0.009 (0.036); -0.021 (0.039) (not significant)
  - Observations vary across columns: 502018, 4500200, 6750200, 6750200, 674873429 (as reported)
  - Note: significance markers ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significant at 1%.
- Count-based robustness (inverse hyperbolic sine of counts):
  - Coefficient on subsidy interactions loses significance when controlling for other GTA policies (results available upon request); interpretation complicated because higher counts imply stronger interventions.

### Heterogeneity: country groups and sectoral patterns
- Country-group heterogeneity (Table 3):
  - Pro-trade effect of subsidies driven by trade flows between different country groups: G20 AEs and G20 EMs, and between each of these groups and non-G20 countries.
  - For G20 AEs:
    - Trade between those countries and others is 68 percent lower than domestic trade; when one country has an industry subsidy the difference diminishes to 64 percent.
  - Subsidies generate the largest trade spillovers between countries of different groups (e.g., G20 EMs and non-G20 economies).
  - Selected coefficients from Table 3 (examples):
    - “Both in group” coefficients: 0.116*** (0.147); 0.376*** (0.120); -0.793*** (0.155)
    - “One in group” coefficients: -1.167*** (0.216); -1.152*** (0.203); -1.494*** (0.232)
    - Domestic subsidies under “One in group”: 0.114*** (0.031); 0.136*** (0.025); 0.119*** (0.031)
    - Observations reported across columns: 502018, 4500200, 6750200, 6750200, 2018450200, 675020067 (as reported)
- Sectoral heterogeneity:
  - Results confirmed in primary and manufacturing subsamples; border effect and interaction with domestic subsidy dummy larger in the primary sector.
  - Manufacturing-industry regressions (ISIC 2-digit) show coefficients that are either positive or statistically insignificant for the interaction between the international trade dummy and domestic subsidy indicator.
  - Industries with more substantial effects include: machinery, furniture, and metals; electrical machinery and apparatus highlighted as particularly affected.
  - Figure 9 reports coefficients and 90 percent confidence intervals by manufacturing industry (examples listed in the source text).

### Robustness and additional checks
- Replacing subsidy dummy with dummies for four main types of domestic subsidies (production subsidies, direct transfers, risk transfers, losses in government revenue):
  - Policies that entail a loss in government revenues such as tax breaks have the largest impact on international relative to domestic trade.
- Excluding country pairs involving China (column (2) of Table A.11) and the U.S. (column (3)):
  - Results robust to exclusion, confirming baseline findings despite influential roles of China and the U.S. in subsidy adoption.
- Modifying GTA policy definitions:
  - Including “neutral” or “liberalizing” policies (column (4)) or including policies implemented by national financial institutions (column (5)) yields a positive and significant coefficient on the interaction between the international trade flows dummy and the domestic subsidy dummy, confirming baseline findings.

### Interpretation and implications
- Combined difference-in-difference and gravity results:
  - Introduction of subsidies is associated with heightened export and import levels of targeted products relative to non-targeted ones.
  - Subsidies do not generally change the direction of comparative advantage at the aggregate sample level (both exports and imports increase for targeted products), except evidence that for G20 EMs subsidies lead to higher exports on the intensive and extensive margins, suggesting potential impacts on comparative advantage for these countries.
  - Significant trade spillovers from domestic subsidies, concentrated in particular industries (electrical machinery in particular), and largest for trade between different country groups (e.g., G20 and non-G20 members).
  - Trade effects most important for tax breaks and other subsidy policies that involve losses in government revenues, surpassing impacts of state aid, grants and loans.

### Policy and research recommendations
- Further theoretical and empirical work needed to address endogeneity of subsidies:
  - Firms expanding export markets or facing heightened import competition are more prone to receiving subsidies; further research should unravel political economy drivers behind subsidy selection and how selection affects subsidy efficacy.
- Need for data on monetary size of subsidies:
  - Data on amounts, even for a limited set of countries, would enable assessment of fiscal ramifications and exploration of additional spillover transmission channels.
- Multilateral cooperation and rules of conduct:
  - Strong evidence of spillovers through trade from domestic subsidies calls for better understanding and design of rules of conduct to support multilateral trade cooperation amid rising state intervention.

*Italic: Source — wpiea2024041-print-pdf (introduction of subsidies section), IMF.*

### REFERENCES

### REFERENCES

### Bibliographic entries
- Aghion, Philippe, Jing Cai, Mathias Dewatripont, Luosha Du, Ann Harrison, and Patrick Legros, 2015, “Industrial policy and competition,” American economic journal: macro-economics, Vol. 7, No. 4, pp. 1–32.
- Anderson, James E, and Yoto V Yotov, 2010, “The changing incidence of geography,” American Economic Review, Vol. 100, No. 5, pp. 2157–2186.
- Arkolakis, Costas, Sharat Ganapati, and Marc-Andreas Muendler, 2021, “The extensive margin of exporting products: A firm-level analysis,” American Economic Journal: Macroeconomics, Vol. 13, No. 4, pp. 182–245.
- Attinasi, MG, L Boeckelmann, and B Meunier, 2023, “Unfriendly friends: Trade and relocation effects of the US Inflation Reduction Act,” VoxEU. org, Vol. 3.
- Bacchetta, Marc, and Michele Ruta, 2011, The WTO, Subsidies and Countervailing Measures (Edward Elgar Publishing).
- Bagwell, Kyle, and Robert W Staiger, 2001, “Domestic policies, national sovereignty, and international economic institutions,” The Quarterly Journal of Economics, Vol. 116, No. 2, pp. 519–562.
- Bagwell, Kyle, and Robert W Staiger, 2006, “Will international rules on subsidies disrupt the world trading system?” American Economic Review, Vol. 96, No. 3, pp. 877–895.
- Baldwin, Richard, and Paul Krugman, 1988, “Industrial policy and international competition in wide-bodied jet aircraft,” in Trade policy issues and empirical analysis, pp. 45–78.
- Baldwin, Richard E, and Frédéric Robert-Nicoud, 2007, “Entry and asymmetric lobbying: why governments pick losers,” Journal of the European Economic Association, Vol. 5, No. 5, pp. 1064–1093.
- Bartelme, Dominick G, Arnaud Costinot, Dave Donaldson, and Andres Rodriguez-Clare, 2019, “The textbook case for industrial policy: Theory meets data,” Working Paper 26193, National Bureau of Economic Research.
- Barwick, Panle Jia, Myrto Kalouptsidi, and Nahim Bin Zahur, 2019, “China’s Industrial Policy: an Empirical Evaluation,” Working Paper 26075, National Bureau of Economic Research.
- Besedeš, Tibor, and Thomas J Prusa, 2011, “The role of extensive and intensive margins and export growth,” Journal of development economics, Vol. 96, No. 2, pp. 371–379.
- Beverelli, Cosimo, Alexander Keck, Mario Larch, and Yoto V Yotov, 2023, “Institutions, trade, and development: identifying the impact of country-specific characteristics on international trade,” Oxford Economic Papers.
- Blanga-Gubbay, Michael, Paola Conconi, and Mathieu Parenti, 2020, “Lobbying for globalization,” Techn. rep.
- Bombardini, Matilde, 2008, “Firm heterogeneity and lobby participation,” Journal of International Economics, Vol. 75, No. 2, pp. 329–348.
- Borchert, Ingo, Mario Larch, Serge Shikher, and Yoto V Yotov, 2021, “The international trade and production database for estimation (ITPD-E),” International Economics, Vol. 166, pp. 140–166.
- Borusyak, Kirill, Xavier Jaravel, and Jann Spiess, 2024, “Revisiting event study designs: Robust and efficient estimation,” arXiv preprint arXiv:2108.12419.
- Bown, Chad P, 2023a, “Industrial policy for electric vehicle supply chains and the US-EU fight over the Inflation Reduction Act,” Peterson Institute for International Economics Working Paper, , No. 23-1.
- Bown, Chad P, 2023b, “Modern industrial policy and the WTO,” Peterson Institute for International Economics Working Paper, , No. 23-15.
- Cherif, Reda, Marc Engher, and Fuad Hasanov, 2024, “Crouching beliefs, hidden biases: The rise and fall of growth narratives,” World Development, Vol. 173, p. 106246.
- Cherif, Reda, and Fuad Hasanov, 2019, “The return of the policy that shall not be named: principles of industrial policy,” Working paper, International Monetary Fund.
- Choi, Jaedo, and Andrei A Levchenko, 2021, “The long-term effects of industrial policy,” Working Paper 29263, National Bureau of Economic Research.
- Conte, Maddalena, Pierre Cotterlaz, Thierry Mayer, and others, 2022, The CEPII gravity database (CEPII).
- Costinot, Arnaud, Dave Donaldson, and Ivana Komunjer, 2012, “What goods do countries trade? A quantitative exploration of Ricardo’s ideas,” The Review of economic studies, Vol. 79, No. 2, pp. 581–608.
- Criscuolo, Chiara, Ralf Martin, Henry G Overman, and John Van Reenen, 2019, “Some causal effects of an industrial policy,” American Economic Review, Vol. 109, No. 1, pp. 48–85.
- Datt, Mohini, Bernard Hoekman, and Mariem Malouche, 2011, “Taking stock of trade protectionism since 2008.”
- De Chaisemartin, Clément, and Xavier d’Haultfoeuille, 2023, “Two-way fixed effects and differences-in-differences with heterogeneous treatment effects: A survey,” The Econometrics Journal, Vol. 26, No. 3, pp. C1–C30.
- Disdier, Anne-célia, Lionel Fontagné, and Enxhi Tresa, 2021, “Economic drivers of public procurement-related protection,” The World Economy, Vol. 44, No. 11, pp. 3072–3090.
- Evenett, Simon, Martin Fernando, Adam Jakubik, and Michele Ruta, 2024, “The return of industrial policy in data,” Working paper wp/24/1, International Monetary Fund.
- Evenett, Simon J, and Johannes Fritz, 2020, “The global trade alert database handbook,” Version: October 2022, Vol. 14.
- Gaulier, Guillaume, and Soledad Zignago, 2010, “Baci: international trade database at the product-level (the 1994-2007 version).”
- Girma, Sourafel, Yundan Gong, Holger Görg, and Zhihong Yu, 2009, “Can Production Subsidies Explain China’s Export Performance? Evidence from Firm-level Data,” Scandinavian Journal of Economics, Vol. 111, No. 4, pp. 863–891.
- Girma, Sourafel, Holger Görg, and Ignat Stepanok, 2020, “Subsidies, spillovers and exports,” Economics Letters, Vol. 186, p. 108840.
- Görg, Holger, Michael Henry, and Eric Strobl, 2008, “Grant support and exporting activity,” The review of economics and statistics, Vol. 90, No. 1, pp. 168–174.
- Harris, Richard, Ian Keay, and Frank Lewis, 2015, “Protecting infant industries: Canadian manufacturing and the national policy, 1870–1913,” Explorations in Economic History, Vol. 56, pp. 15–31.
- Head, Keith, and Thierry Mayer, 2014, “Gravity equations: Workhorse, toolkit, and cookbook,” in Handbook of international economics, Vol. 4, pp. 131–195.
- Heid, Benedikt, Mario Larch, and Yoto V. Yotov, 2021, “Estimating the effects of non-discriminatory trade policies within structural gravity models,” Canadian Journal of Economics/Revue canadienne d’économique, Vol. 54, No. 1, pp. 376–409.
- Hoekman, Bernard M, and Douglas Nelson, 2021, Subsidies, Spillovers, and Multilateral Cooperation (Brookings Institution Press).
- IMF, 2024, Fiscal Monitor.
- IMF, OECD, World Bank and WTO, 2022, Subsidies, Trade, and International Cooperation.
- Irwin, Douglas A, and Nina Pavcnik, 2004, “Airbus versus Boeing revisited: international competition in the aircraft market,” Journal of international economics, Vol. 64, No. 2, pp. 223–245.
- Juhász, Réka, 2018, “Temporary protection and technology adoption: Evidence from the napoleonic blockade,” American Economic Review, Vol. 108, No. 11, pp. 3339–3376.
- Juhász, Réka, Nathan Lane, Emily Oehlsen, and Verónica C Pérez, 2022, “The Who, What, When, and How of Industrial Policy: A Text-Based Approach,” What, When, and How of Industrial Policy: A Text-Based Approach (August 15, 2022).
- Juhász, Réka, Nathan J Lane, and Dani Rodrik, 2023, “The new economics of industrial policy,” Techn. Rep. 31538, National Bureau of Economic Research.
- Kalouptsidi, Myrto, 2018, “Detection and impact of industrial subsidies: The case of Chinese shipbuilding,” The Review of Economic Studies, Vol. 85, No. 2, pp. 1111–1158.
- Kinzius, Luisa, Alexander Sandkamp, and Erdal Yalcin, 2019, “Trade protection and the role of non-tariff barriers,” Review of World Economics, Vol. 155, pp. 603–643.
- Krueger, Anne O, and Baran Tuncer, 1982, “An empirical test of the infant industry argument,” The American Economic Review, Vol. 72, No. 5, pp. 1142–1152.
- Kucheryavyy, Konstantin, Gary Lyn, and Andrés Rodríguez-Clare, 2023, “Grounded by gravity: A well-behaved trade model with industry-level economies of scale,” American Economic Journal: Macroeconomics, Vol. 15, No. 2, pp. 372–412.
- Lampe, Markus, Kevin Hjortshøj O’Rourke, Lorenz Reiter, and Yoto V. Yotov, 2023, “The Empire Project: Trade Policy in Interwar Canada,” NBER Working Papers 31238, National Bureau of Economic Research, Inc.
- Lane, Nathan, 2022, “Manufacturing revolutions: Industrial policy and industrialization in South Korea,” Available at SSRN 3890311.
- Lashkaripour, Ahmad, and Volodymyr Lugovskyy, 2023, “Profits, scale economies, and the gains from trade and industrial policy,” American Economic Review, Vol. 113, No. 10, pp. 2759–2808.
- Lee, G.M., 2016, “Chapter 4 - Subsidies and Countervailing Duties,” Handbook of Commercial Policy, Vol. 1, pp. 161–210 (North-Holland).
- Lee, Soo Jeong, and Jeffrey M Wooldridge, 2023, “A Simple Transformation Approach to Difference-in-Differences Estimation for Panel Data,” Available at SSRN 4516518.
- Leromain, Elsa, and Gianluca Orefice, 2014, “New revealed comparative advantage index: Dataset and empirical distribution,” International Economics, Vol. 139, pp. 48–70.
- Liu, Licheng, Ye Wang, and Yiqing Xu, 2024, “A practical guide to counterfactual estimators for causal inference with time-series cross-sectional data,” American Journal of Political Science, Vol. 68, No. 1, pp. 160–176.
- Manelici, Isabela, and Smaranda Pantea, 2021, “Industrial policy at work: Evidence from Romania’s income tax break for workers in IT,” European Economic Review, Vol. 133, p. 103674.
- McCallum, John, 1995, “National borders matter: Canada-US regional trade patterns,” The American Economic Review, Vol. 85, No. 3, pp. 615–623.
- Munch, Jakob, and Georg Schaur, 2018, “The effect of export promotion on firm-level performance,” American Economic Journal: Economic Policy, Vol. 10, No. 1, pp. 357–387.
- Navarra, Elisa, 2023, “The Effects of Corporate Subsidies Along Supply Chains,” Working Paper. Mimeo.
- Pack, Howard, and Kamal Saggi, 2006, “Is there a case for industrial policy? A critical survey,” The World Bank Research Observer, Vol. 21, No. 2, pp. 267–297.
- Piermartini, Roberta, and Yoto V Yotov, 2016, “Estimating trade policy effects with structural gravity.”
- Roth, Jonathan, Pedro HC Sant’Anna, Alyssa Bilinski, and John Poe, 2023, “What’s trending in difference-in-differences? A synthesis of the recent econometrics literature,” Journal of Econometrics.
- Silva, JMC Santos, and Silvana Tenreyro, 2006, “The log of gravity,” The Review of Economics and statistics, Vol. 88, No. 4, pp. 641–658.
- UNCTAD, 2019, International classification of non-tariff measures (UN).
- Volpe Martincus, Christian, and Jerónimo Carballo, 2010, “Entering new country and product markets: does export promotion help?” Review of World Economics, Vol. 146, pp. 437–467.
- Volpe Martincus, Christian, and Jerónimo Carballo, 2008, “Is export promotion effective in developing countries? Firm-level evidence on the intensive and the extensive margins of exports,” Journal of International Economics, Vol. 76, No. 1, pp. 89–106.
- Wooldridge, Jeffrey M, 2021, “Two-way fixed effects, the two-way mundlak regression, and difference-in-differences estimators,” Available at SSRN 3906345.
- World Bank, 2023, “Unfair Advantage : Distortive Subsidies and Their Effects on Global Trade,” Equitable Growth, Finance and Institutions Insight, World Bank Group.

### Appendix A — Figures and Tables (captions and selected reported values)
- Figure A.1. Number of export promotion policies over time
  - Panels: (a) Policy announcements; (b) Cumulative sum of announcements.
  - Note: In panel (a), we count the number of new policies by year of announcements. If the announcement happens on or after July 1st, the following year is the year of announcement. In panel (b) we count the number of policies introduced since 2009 and in force at a given year. A policy is a country-policy combination (e.g., policies adopted at the supranational level (EU for instance) are counted as many times as the number of countries affected), counted once regardless of the number of products affected. Direct transfers are subsidies that transfer resources to firms. Trade finance includes policies classified under the “trade finance” policy instrument category in the GTA. The export subsidy group includes policies classified as “Export subsidies”, "Tax-based export incentive" and "Other export incentive" in the GTA database.

- Figure A.2. Number of subsidies in the primary sector and the primary share of subsidies over time
  - Panels: (a) Policy announcements; (b) Cumulative sum of announcements.
  - Note: In panel (a), we count the number of new policies by year of announcements. If the announcement happens on or after July 1st, the following year is the year of announcement. In panel (b) we count the number of policies introduced since 2009 and in force at a given year. A policy is a country-policy combination (e.g., policies adopted at the supranational level (EU for instance) are counted as many times as the number of countries affected), counted once regardless of the number of products affected. The primary sector is identified by the ISIC 2-digit code to which the targeted products belong (chapters before 15 are considered primary sector).

- Table A.1. Domestic subsidies by industry
  - Note: Each policy is counted once for each industry it targets. Announcements in force include those made since 2009. Column (7) reports the difference in the number of announcements in force between 2021 and 2009, relative to the count in 2009.
  - Selected entries (Industry — Announcements 2009 — Announcements 2015 — Announcements 2021 — 2021-2009 change):
    - Agriculture, hunting and related service activities — 260 — 565 — 186 — 260135218626.16
    - Forestry, logging and related service activities — 314 — 366 — 316 — 823376.67
    - Fishing, aquaculture and service activities incidental to fishing — 312 — 179 — 315 — 022373.33
    - Extraction of crude petroleum and natural gas — 633 — 266 — 861 — 8830.33
    - basic metals — 1779 — 1117 — 2319 — 3453.94
    - motor vehicles, trailers and semi-trailers — 361 — 100 — 1433 — 6369124633.61
  - (Table contains many industry rows; each policy counted once per targeted industry.)

- Table A.2. Summary statistics for the variables used in the difference-in-difference specifications
  - Observations: 9931350 for product-level variables.
  - Exports: Mean 21.115, Std. dev. 20.880, Min 0.000, 25th p. 0.000, 75th p. 0.552, Max 62674.56.
  - Imports: Mean 21.084, Std. dev. 95.95, Min 0.000, 25th p. 0.013, 75th p. 3.193, Max 14813.66.
  - Subsidies (all): Mean 0.06, Std. dev. 0.25, Min 0.000, 25th p. 0.000, 75th p. 0.001, Max 1.00.
  - Domestic subsidies: Mean 0.06, Std. dev. 0.23.
  - Production subsidies: Mean 0.00, Std. dev. 0.04.
  - Direct transfers: Mean 0.02, Std. dev. 0.12.
  - Revenue-losing policies: Mean 0.04, Std. dev. 0.20.
  - Export promotion: Mean 0.01, Std. dev. 0.11.
  - Note: Exports and imports are in millions of current US$.

- Table A.3. Effects of subsidies and other GTA policies on product-level trade (dependent variable ln(exports) / ln(imports))
  - Subsidies (all): column (1) ln(exports) coefficient 0.074** (0.031); column (2) ln(imports) coefficient 0.029*** (0.010).
  - Domestic subsidies: column (3) ln(exports) 0.019*** (0.006); column (4) ln(imports) 0.039*** (0.006).
  - Export restrictions: ln(exports) -0.139*** (0.032).
  - Technical barriers to trade: ln(imports) -0.060** (0.027).
  - Temporary import restrictions: ln(imports) -0.034** (0.013).
  - Macro restrictions: ln(imports) -0.081*** (0.024).
  - Observations vary by specification; example Obs = 61386116 in some specifications.
  - R2 values reported: e.g., 0.86, 0.88, 0.91, 0.91, 0.88, 0.89, 0.92, 0.92.
  - Note: Subsidies and other GTA policies are dummies equal to one if there is at least one active intervention targeting a product in a country and year. Standard errors clustered by country and product. ∗ significant at 10%; ∗∗ significant at 5%; ∗∗∗ significant at 1%.

- Figure A.3. Effects of domestic subsidies on trade flows by country group and destination
  - Panels show event-time effects (horizontal axis: time before and after a country-product is targeted) with 90 percent confidence intervals.
  - Definitions: “G20 AEs” include Australia, Canada, Republic of Korea, France, Germany, Italy, the United Kingdom, Japan and the United States. “G20 EMs” include Argentina, Brazil, China, India, Indonesia, Mexico, Russia, Saudi Arabia, Türkiye and South Africa. Non-G20 economies are countries that are not in the G20.

- Table A.4. Effects of subsidies on product-level trade – imputation method (dependent variable Ln(exports) / Ln(imports))
  - Full sample: Domestic subsidy coefficient on Ln(exports) 0.004 (0.005); Ln(imports) 0.026*** (0.006).
  - G20AE: Ln(exports) 0.014 (0.012); Ln(imports) 0.021** (0.010).
  - G20EM: Ln(exports) 0.033* (0.020); Ln(imports) -0.008 (0.020).
  - F-stat for pre coefficients = 0 test: Full sample 04.05 (p-value (0.00)); G20AE 1.31 (0.20); G20EM 1.21 (0.27); imports Full sample 27.32 (0.00).
  - Note: Imputation method of Borusyak, Jaravel, and Spiess (2024). Standard errors from clustered bootstrap with 500 replications. ∗ significant at 10%; ∗∗ at 5%; ∗∗∗ at 1%.

- Table A.5. Effects of subsidies on product-level trade by sector (Manufacturing vs Primary)
  - Manufacturing: Subsidies (all) Ln(exports) 0.023*** (0.006), Ln(imports) 0.041*** (0.006).
  - Primary: Subsidies (all) Ln(exports) -0.012 (0.022), Ln(imports) -0.002 (0.018).
  - Domestic subsidies (Manufacturing) Ln(exports) 0.020*** (0.006), Ln(imports) 0.043*** (0.006).
  - Export promotion (Primary) Ln(exports) -0.120** (0.050).
  - Observations and R2 reported across columns (examples: Obs 5571871; R2 0.91).

- Table A.6. Effects of domestic subsidies on product-level trade – robustness checks
  - Domestic subsidies coefficients (selected specifications):
    - RCA<1 Ln(exports) 0.009 (0.012)
    - RCA>1 Ln(exports) 0.022*** (0.006)
    - No US Ln(exports) 0.020*** (0.006)
    - No CHN Ln(exports) 0.020*** (0.006)
    - RCA<1 Ln(imports) 0.040*** (0.008)
    - RCA>1 Ln(imports) 0.041*** (0.007)
    - No US Ln(imports) 0.041*** (0.006)
    - No CHN Ln(imports) 0.041*** (0.006)
  - Direct transfers: Ln(exports) -0.024** (0.010); Ln(imports) -0.020*** (0.007) in some specs.
  - Revenue-losing policies: Ln(exports) 0.027*** (0.008); Ln(imports) 0.059*** (0.007) in some specs.
  - Notes describe samples restricted by revealed comparative advantage (RCA), and exclusions of U.S. or China.

- Table A.7. Effects of subsidies on product-level trade – additional robustness
  - Subsidies (all) Ln(exports) 0.018*** (0.006); Ln(imports) 0.035*** (0.007).
  - Domestic subsidies Ln(exports) 0.018*** (0.006); Ln(imports) 0.037*** (0.007).
  - Columns consider all policy changes and those including National Financial Institutions (NFIs).

- Table A.8. Effects of subsidies on export and import probabilities (dependent variable Exports>0 / Imports>0)
  - Subsidies (all): Exports>0 coefficient -0.001* (0.001); Imports>0 0.002*** (0.001).
  - Domestic subsidies: Exports>0 -0.001 (0.001); Imports>0 0.003*** (0.001).
  - Observations: 9931350. R2 around 0.74–0.77.
  - Note: Linear probability models with country-product fixed effects and controls.

- Figure A.4. Effects of domestic subsidies on export and import probabilities of G20 EMs by destination
  - Event-time plots showing effects on Prob(exports>0) and Prob(imports>0) for G20 EMs vis-à-vis G20 AEs, G20 EMs, and non-G20 partners.
  - Note: “EMEs” includes Argentina, Brazil, China, India, Indonesia, Mexico, Russia, Saudi Arabia, Türkiye and South Africa.

- Table A.9. Summary statistics for the variables used in the gravity specifications
  - Observations: 5032232 for bilateral exports and GTA policy dummies.
  - Bilateral exports: Mean 283.204, Std. dev. 436.25, Min 0.000, 25th p. 0.000, 75th p. 0.721, Max 1815773.19 (millions of current US$).
  - Subsidies (all): Mean 0.30, Std. dev. 0.46.
  - Domestic subsidies: Mean 0.27, Std. dev. 0.45.
  - Import restrictions: Mean 0.56, Std. dev. 0.50.
  - PTA: Mean 0.19, Std. dev. 0.40.
  - WTO: Mean 0.80, Std. dev. 0.40.
  - ln(distance): Mean 8.56, Std. dev. 0.87, Min 1.61, Max 9.90.
  - Note: An observation corresponds to a country-pair-industry-year combination. Bilateral exports are in millions of current US$.

- Table A.10. Effects of subsidies on international relative to domestic trade flows (manufacturing and primary sectors)
  - Manufacturing: Intl. trade flows × coefficient -0.682*** (0.064) in one specification; Subsidy (all) 0.237*** (0.030).
  - Domestic subsidy (Manufacturing): 0.105*** (0.032).
  - Primary: Intl. trade flows × coefficient -1.036*** (0.098); Subsidy (all) 0.326*** (0.051).
  - Domestic subsidy (Primary): 0.240*** (0.042).
  - Export subsidy coefficients reported (some not significant); example Export subsidy (Primary) -0.080* (0.047).
  - Note: Gravity estimates at ISIC 2-digit; columns include importer-industry-year and exporter-industry-year fixed effects and controls. Standard errors clustered by importer, exporter and symmetric country pairs.

- Table A.11. Effects of subsidies on international relative to domestic trade flows – additional results
  - Intl. trade flows × Domestic subsidies coefficients in robustness specs:
    - No CHN: 0.132*** (0.027)
    - No USA: 0.141*** (0.030)
    - All GTA: 0.110*** (0.029)
    - NFIs: 0.119** (0.027)
  - Production subsidies: 0.071*** (0.021) in one specification.
  - Direct transfers: 0.048** (0.020) in one specification.
  - Revenue-losing policies: 0.109*** (0.024) in one specification.
  - Observations vary (examples: 5020067; 4927083).
  - Note: Columns exclude China or the U.S. in some specifications; others include all GTA policy changes or those implemented by national financial institutions. Standard errors clustered by importer, exporter and symmetric country pairs. ∗ significant at 10%; ∗∗ at 5%; ∗∗∗ at 1%.

*Trade Spillovers of Domestic Subsidies — Working Paper No. WP/2024/041*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024041-print-pdf.pdf_