## _wp11139

## Source details

**Canonical URL:** [_wp11139](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11139.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11139.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11139.pdf.json)

---

### Introduction and objectives
- New acts of protectionism since the start of the global financial crisis have been documented by two main monitoring exercises (WTO and Global Trade Alert (GTA)) which reach different prevalence conclusions.
- Aim: quantify the impact of crisis protectionist measures on merchandise trade flows using a dummy-variable identification strategy because reliable measures of the magnitude of many non-tariff measures are unavailable.
- Methodological constraint: product-level identification relies on dummy variables indicating whether a product has been affected by protectionism at the 4-digit HS level.

### Data sources, scope, and sample construction
- Trade data:
  - Monthly bilateral 4-digit HS merchandise trade data from Global Trade Information Services (GTIS).
  - Reporters: EU and fourteen other major G-20 reporting countries (all G-20 except Saudi Arabia when accounting for EU membership).
  - Coverage: some 80 percent of global merchandise trade.
  - Time coverage: for many reporters July 2007-April 2010; data for all reporters available through December 2009.
  - Dependent-variable series (year-on-year log differences) starts in July 2008.
  - Total observations for regressions: 9.9 million monthly observations of import values in country-pair/product combinations.
  - Import volumes available and used in robustness analysis.
- GTA measures:
  - GTA provided implementing country, affected partners, 4-digit HS categories, month of implementation (and removal if any), and description.
  - Inclusion: only measures reported by GTA as implemented during the study period and classified as almost certainly discriminating against foreign interests (“red” measures).
  - As of beginning of June 2010, 508 GTA measures met these criteria; 314 featured all information necessary for inclusion in the analysis.
  - Excluded 38 percent of measures were mostly behind-the-border measures for which affected partners or merchandise categories could not be identified.
- Measure classification (sample construction):
  - Import restrictions: (a) tariffs and import bans, (b) trade defense measures—anti-dumping, countervailing duty, safeguard measures, (c) non-tariff barriers, (d) discriminatory purchasing policies.
  - Export restrictions and export support measures (each less than 20 measures) largely excluded from main analysis.
  - Behind-the-border measures: bailouts, domestic subsidies, investment subsidies.
  - Most analysable measures are border measures; behind-the-border measures less numerous in the analysable sample.

### Identification strategy and econometric specification
- Identification intuition:
  - Track trade for a 4-digit product in months following a new import restriction by measuring trade of affected country-pairs as a share of global trade in the same product, normalizing market share to 100 in T-1.
- Estimation equation (first-differenced gravity equation):
  - ln(Imports)_{ijpt}Δ = β · ProtDummy_{ijpt}Δ + TVFE_{ijpt}Δ + ε_{ijpt}
  - ln(Imports)Δ is the 12-month change in the U.S. dollar value of log imports.
  - ProtDummy counts the number of crisis protectionist measures applied to a given trade flow (0, 1, 2, ...).
  - TVFE denotes one or more sets of time-varying fixed effects.
  - First-differencing: controls for time-invariant determinants and addresses product-level seasonality by focusing on 12-month changes.
- Time-varying fixed effects considered:
  - TVP (product), TVIM (importer), TVEX (exporter), TVCP (country-pair), TVIMP (importer-product), TVEXP (exporter-product).
  - Practical approach: include two high-dimensional TVFEs at a time using Guimaraes and Portugal’s iterative technique; first-differencing eliminates one set.
- Protectionist dummy construction and aggregation:
  - ProtDummy counts measures per observation; β interpretable as average trade impact of one measure.
  - Aggregation: multiply estimated percentage reduction in product-level trade by the amount of trade affected and sum across protectionist dummies.
  - Empirical simplification: majority of affected observations have one measure only (observations affected by one import restriction numbered 56,050 = 0.57 percent of sample; two, three, four restrictions numbered 4,780; 218; and 2 respectively; remainder 9,817,431 observations unaffected).

### Baseline specification and estimation choices
- Preferred specification: time-varying product & country-pair fixed effects (Specification 3) after comparing six FE setups.
  - Specification progression: TV product FEs only (1) → add importer FEs (2) → add country-pair FEs (3, preferred). Specifications 4–6 generalize by allowing product-varying demand shocks within countries but Specification 3 is preferred.
  - F-statistics: reject spec 2 in favor of spec 3 with value 1.80; reject spec 2 in favor of spec 4 with F-value 1.12.
- Note on R-squared: low R-squared values expected given volatile 12-month differences; protectionist dummies are highly statistically significant but rare and explain a small fraction of variance.

### Product-level findings (key quantitative regression results)
- Import restrictions (border measures):
  - Affected trade flows fell by 5 percent in response to border measures (Regression 1: roughly 5 percent = e^-0.048; preferred estimate Specification 3: decrease of 5 percent = e^-0.051).
  - More parsimonious specifications corroborate the 5 percent result; detailed importer-product FEs imply impacts on the order of 8 percent in some specifications.
- Behind-the-border measures:
  - Affected trade flows fell by 7 percent in response to behind-the-border measures (preferred Specification 3 estimate).
  - Graphical analogs show an 8 percent average immediate decline; Regression 1 reports roughly 15 percent = e^-0.165 in the least controlled specification.
  - Estimates may be sensitive to inclusion of importer fixed effects because many behind-the-border measures discriminate between domestic and foreign firms, not among partners.
- Disaggregation by income groups (Table 7 reported estimates):
  - Affected lower middle income country exports reduced by 8 percent.
  - Affected upper middle income country exports reduced by 5 percent.
  - Affected exports of low income countries (LICs) estimated to be reduced by over 10 percent, but coefficient statistically insignificant (fewer measures affected LICs).
- Measure granularity caveat:
  - GTA coding at 4-digit HS leads to downward bias of product-level estimates because many measures are targeted at 6- or 8-digit lines or specific firms.

### Effects by measure types (preferred TV product & country-pair FE specification)
- Border measures disaggregated (Regression 7):
  - (i) Tariffs and import bans
  - (ii) Trade defense measures (anti-dumping, countervailing duty, safeguards)
  - (iii) Non-tariff barriers (mainly licensing requirements and sanitary and phytosanitary restrictions)
  - (iv) Discriminatory purchasing measures (local content provisions, public procurement, consumption subsidies)
- Key findings by type:
  - Tariffs: impact statistically insignificant despite 67 tariff measures recorded.
  - Quotas and import bans: decreased trade in affected products by 24 percent and 13 percent respectively (coefficients borderline significant; aggregate contribution low due to narrow targeting).
  - Trade defense measures: 102 duties imposed; trade in affected 4-digit products decreased by 16 percent, implying greater collapse at sub-4-digit levels.
  - NTBs (new licensing requirements): 9 percent trade decrease at product level.
  - Discriminatory purchasing measures: most harmful among border measures; consumption subsidies caused product-level decreases of -9 percent and reduced global trade by -0.03 percent; local content requirements covering entire domestic market were similarly harmful; public procurement measures did not show a trade impact in this analysis.
- Behind-the-border disaggregation:
  - Bailouts and domestic subsidies each decreased trade in affected products by around 7 percent.
  - Bailouts aggregate impact: -0.09 percent.
  - Domestic subsidies aggregate impact: -0.03 percent.
  - Investment subsidies did not cause a contemporaneous statistically significant trade reduction.

### Aggregate results, scenarios, and alternative calculations
- Baseline aggregate impact:
  - Crisis protectionism measures are decreasing global trade by at least $30-35 billion, or 0.2 percent, annually.
  - Import restrictions estimated to have reduced world trade by 0.21 percent in the last quarter 2009.
    - In the sample (covering 80 percent of world trade) this 0.21 percent reduction translates into a quarterly trade loss of $4.6 billion when applied to quarterly trade flows in the last quarter of 2009 using Specification 3 product-level coefficients.
    - “Affected quarterly trade” stood at $77.7 billion in the last quarter of 2009.
  - Border vs behind-the-border contributions (last quarter of 2009): border measures reduced trade by $2.1 billion; behind-the-border measures reduced trade by $2.6 billion.
  - A typical single behind-the-border measure distorted trade about seven times more than a typical border measure (40 behind-the-border import restrictions vs 239 border measures in sample).
- Comparative significance:
  - Removing the crisis protectionist measures included in the study could increase aggregate global trade by about 1/7 of the amount that could be expected from a Doha Round conclusion (0.21/1.50 = 1/7, where Doha draft estimated to boost world trade by 1.5 percent).
- Data-limitation adjustments and alternative scenarios:
  - Exclusion of 38 percent of GTA “red” measures likely underestimates affected trade and the 0.21 percent reduction.
  - If excluded measures were exactly as restrictive as included ones, impact would rise to 0.34 percent or $50-60 billion annually (calculation reference: 0.21*(508/314)).
  - Extreme case: if all excluded measures were behind-the-border measures and same product-level coefficients apply, total trade impact could reach 0.75 percent of world trade or $110-125 billion a year, with behind-the-border measures accounting for most distortion (calculation reference: 0.11*((508-314+40)/40)+0.10).
- Robustness checks:
  1. Adding export measure dummies: inclusion does not change import restriction coefficients; export measures insignificant (fewer than 20 export restrictions and fewer than 20 export support measures).
  2. Dependent variable as trade volumes: weaker magnitudes and significance but main results broadly hold; in TV product & country-pair FE specification, border restrictions reduce trade by 2.5 percent (vs 5 percent in baseline); behind-the-border measures lose statistical significance.
  3. Protectionist dummies coded 0-1 only: baseline results almost unchanged.

### Effects by country group, sector, and time period
- Effects by country group:
  - Border measures:
    - Advanced countries’ border measures reduced affected product-level imports by 8 percent; developing countries’ measures reduced them by about half that amount.
    - Advanced countries account for about ¾ of aggregate trade distortion from border measures.
    - Regional drivers: North America drives results; Latin American border measures reduced affected imports by -9 percent (small trade coverage).
    - East Asian border measures show highest impact on aggregate world trade though product-level coefficient small and statistically insignificant.
  - Behind-the-border measures:
    - Developing countries’ behind-the-border measures more trade restrictive at product level (-9 percent), driven by upper middle income countries and Central and East Asia.
    - Advanced countries’ behind-the-border measures more harmful on aggregate because they targeted larger import flows; advanced countries bore more than 85 percent of aggregate impact of behind-the-border measures.
    - Much of behind-the-border estimated impact originates in North America (possible bias from crisis-unrelated factors).
  - Exports:
    - Developing country exports hurt somewhat more by border measures; aggregate reduction about 0.05 percent of world trade for both developed and developing country exports.
    - Lower middle income countries experienced considerably higher impact than upper middle income countries; LICs estimated to have fallen by even more but with statistically insignificant coefficients.
    - Regional export impacts: Africa grouping faces most restrictive border measures; North America faces more restrictive measures among advanced countries; Western Europe suffers larger aggregate impact.
    - Behind-the-border export impacts mainly within advanced countries: product-level reduction of 8 percent for advanced country exports facing behind-the-border measures; Latin America faced highest product-level impact for new behind-the-border measures with 14 percent reduction; Africa and Asia seemingly unscathed by behind-the-border measures.
- Effects by sector:
  - Textiles and machinery: border measures reduced affected imports by -7 and -5 percent respectively; machinery accounts for more than half of reduction in world imports among border measures due to larger affected trade flows.
  - Behind-the-border measures mainly obstructed machinery and transport equipment, reducing affected flows by 12 percent on average.
  - Transport sector accounts for the lion’s share of aggregate behind-the-border impact, with aggregate impact tallied at -0.14 percent reduction in world trade.
  - HS 2-digit to sector mapping used (e.g., Agriculture 01-15; Machinery 84, 85, 90, 91, 93, 95, 96; Transportation 86-89).
- Effects by time periods:
  - Time period definitions:
    - Trade collapse: up to January 2009.
    - Trade stabilization: February-May 2009.
    - Trade recovery: from June 2009 onwards.
  - Measures implemented during first nine months after Lehman collapse (September 2008) were particularly harmful.
  - Border measures implemented before trade recovery (before June 2009) decreased affected trade flows by 5-6 percent on average.
  - Trade stabilization phase (February-May 2009) saw most and broadest measures; accounted for almost entire aggregate impact for both border and behind-the-border measures, reducing world trade by roughly 0.1 percent each.
  - Behind-the-border measures implemented during stabilization reduced trade flows by 8 percent on average.
  - Measures implemented after trade recovery began did not show a statistically significant effect (sample limitations noted).
  - Persistence: measures remaining in effect during recovery continued to harm trade—estimated product-level reductions during recovery: 4 percent for border and 6 percent for behind-the-border measures.
  - Removal of crisis measures still in effect could boost global trade by at least 0.2 percent annually.

### Interpretation, caveats, and policy implications
- Interpretation:
  - Affected trade flows fell by 5 percent in response to border measures and 7 percent in response to behind-the-border measures; these impacts may be underestimated due to 4-digit aggregation mismatch with measure granularity.
  - Traditional trade measures (notably tariffs) had little impact; antidumping duties and unconventional measures (NTBs, discriminatory purchasing policies, bailouts, domestic subsidies) substantially reduced affected trade flows.
  - Aggregate world trade impacts are moderated by the small share of global trade subject to new measures.
- Caveats:
  - Product-level estimates are lower bounds because GTA codes at 4-digit HS while many measures apply at 6- or 8-digit levels or target specific firms.
  - Exclusion of 38 percent of GTA “red” measures likely understates affected trade and aggregate impact.
  - Some regional behind-the-border estimates may be biased by omitted non-protectionist crisis-related factors.
- Policy recommendation:
  - Policymakers should remain attentive to protectionist pressures and consider removing crisis protectionist measures to aid the global economic recovery and avoid perpetuating associated trade losses.
  - The limited aggregate interference with the global recovery reflects restraint by countries in adopting measures, not that the individual measures are innocuous.

*Source: Excerpt from IMF working paper content unit _wp11139.*

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

### _wp11139 - References

### Introduction and objectives
- New acts of protectionism since the start of the global financial crisis have been well documented.
- Two main monitoring exercises—WTO and Global Trade Alert (GTA)—reach different conclusions on prevalence; WTO reports acknowledge “instances of trade restrictive measures” while GTA data suggest protectionism rose during the slump and as trade recovered.
- Aim of the paper: go beyond stocktaking to quantify the impact of crisis protectionist measures on merchandise trade flows.
- Methodological constraint: estimation strategy relies on dummy variables to identify whether a product has been affected by protectionism because reliable measures of the magnitude of many non-tariff measures are unavailable.

### Product-level findings
- Affected trade flows fall in response to discriminatory measures:
  - Affected trade flows fell by 5 percent in response to border measures.
  - Affected trade flows fell by 7 percent in response to behind-the-border measures.
  - These impacts may be somewhat underestimated (reason: mismatch between HS 4-digit aggregation and measure granularity).
- Relative impacts by measure type:
  - Tariffs and other traditional trade measures have a relatively small impact.
  - Antidumping duties and unconventional measures (non-tariff barriers, discriminatory purchasing policies, bailouts, domestic subsidies) have substantially reduced affected trade flows.
- Distributional effects across country groups:
  - Protectionism was very harmful for exports of developing countries, particularly poorer nations within this group.
  - Border restrictions most hurt the developing world.
  - Developing country exports were little affected by trade partners’ bailouts; advanced countries’ behind-the-border measures appear to fall mainly on their peers.
- Income-group differential estimates (reported in Table 7):
  - Affected lower middle income country exports are estimated to be reduced by 8 percent.
  - Affected upper middle income country exports are estimated to be reduced by 5 percent.
  - Affected exports of low income countries (LICs) are estimated to be reduced by over 10 percent, but the respective coefficient is statistically insignificant (potentially because LICs are affected by a much smaller number of measures).

### Aggregate results and interpretation
- Despite substantial product-level impacts, aggregate world trade impacts are moderated by the small share of global trade subject to new measures.
- Estimated aggregate annual impact:
  - Crisis protectionism measures are decreasing global trade by at least $30-35 billion, or 0.2 percent, annually.
- Comparative significance:
  - Removing the crisis protectionist measures included in the study could increase aggregate global trade by about 1/7 of the amount that could be expected from a Doha Round conclusion.
- Policy implication emphasized: the limited aggregate interference with the global recovery reflects restraint by countries in adopting measures, not that the individual measures are innocuous.

### Data sources and scope
- Trade data:
  - Monthly bilateral 4-digit HS merchandise trade data from Global Trade Information Services (GTIS).
  - Data include external imports and exports reported by the European Union (EU) and fourteen other major G-20 reporting countries (all G-20 except Saudi Arabia when accounting for EU membership).
  - The data cover some 80 percent of global merchandise trade.
  - For many reporters, data cover July 2007-April 2010; data for all reporters are available through December 2009.
  - When year-on-year log differences are constructed, the dependent-variable series starts in July 2008.
  - Total observations for regressions: 9.9 million monthly observations of import values in country-pair/product combinations.
  - Import volumes are available and used in robustness analysis; volume data are somewhat less reliable but confirm main results.
- GTA measures:
  - GTA provided implementing country, trade partners affected, 4-digit HS product categories affected, month of implementation (and removal, if any), and a description.
  - Analysis includes only measures reported by GTA as implemented during the study period and classified as almost certainly discriminating against foreign interests (“red” measures).
  - As of beginning of June 2010, 508 GTA measures met these criteria; of these, 314 featured all information necessary to be included in the analysis.
  - The excluded 38 percent of measures were mostly behind-the-border measures (e.g., financial sector bailouts) for which GTA could not identify affected partners or merchandise categories.

### Measure classification and sample construction
- Measures classified (Table 1) as:
  - Import restrictions (subcategories: (a) tariffs and import bans, (b) trade defense measures—anti-dumping, countervailing duty, safeguard measures, (c) non-tariff barriers, (d) discriminatory purchasing policies such as government procurement provisions and consumption subsidies tied to domestic products).
  - Export restrictions.
  - Export support measures.
  - Behind-the-border measures (subcategories: bailouts, domestic subsidies, investment subsidies).
- Empirical sample:
  - Most protectionist measures in the sample were border measures aimed at restricting imports.
  - Behind-the-border measures are fewer in the analysable sample partly because many could not be included.
  - Export restrictions and export support measures number less than 20 each and are excluded from much of the analysis due to limited power in volatile disaggregate trade data (their exclusion does not change main results on import restrictions and behind-the-border measures).

### Identification strategy and graphical evidence
- Identification intuition:
  - Track trade for a 4-digit product in months following a new import restriction by measuring trade of affected country-pairs as a share of global trade in the same product, normalizing that market share to 100 in T-1 (the month prior to implementation).
  - This controls for worldwide product-specific influences (e.g., differential cyclical patterns across durable vs. nondurable goods).
- Graphical findings (Figures 1–3):
  - Raw data series for import restrictions implemented through March 2009 show the market share of trade covered by restrictions declining after implementation; the decline in restricted trade outpaced that in global trade for the same product and appears persistent.
  - Averaged series suggest import restrictions may have decreased trade by 10 to 20 percent depending on averaging technique (by number of observations affected or by value of trade affected).
  - Regression estimates are qualitatively similar but somewhat lower than the graphical 10–20 percent implied reductions.

### Caveats and methodological notes
- Granularity limitation:
  - GTA codes affected products at the 4-digit HS level; many measures are more narrowly targeted (6-digit or higher), so estimated product-level effects are likely underestimated when only parts of 4-digit categories are affected.
  - Use of more detailed trade data (6-digit) would likely improve precision but was not possible given GTA coding.
- Export effects of behind-the-border measures:
  - Behind-the-border measures (e.g., bailouts) can both decrease imports and increase exports relative to the counterfactual; the authors’ estimates indicate only the import effects are economically important, so export effects are dropped in most regressions (analogs including export effects are reported in Table 3).
- Period coverage and implementation timing:
  - The implementation month of the first protectionist measures included in the GTA database is October (year not restated in the provided excerpt).
  - Early implementation months provide the longest post-implementation series and affected larger amounts of trade than measures implemented later.

*Italic source attribution line.*

### 2008. However, measures implemented in October 2008 affected very small trade flows. As a consequence, the

### _wp11139 - 2008. However, measures implemented in October 2008 affected very small trade flows. As a consequence, the

### Measures, coverage, and graphical findings
- Measures implemented up to March 2009 cover more than 80 percent of total trade affected by protectionist measures covered in this study.
- Figure-based analogs and regression analogs:
  - Our closest regression analog to the figures is Regression 1 (Table 2). It reports the trade-reducing impact of import restrictions at roughly 5 percent (=e^-0.048).
  - The resulting market share series for measures implemented in October 2008 are very volatile and were omitted from Figure 1.

### Behind-the-border measures (definition, coverage, and graphical findings)
- Definition: Behind-the-border measures are defined as direct assistance to domestic firms that is discriminatory, i.e. not available to foreign firms exporting to the domestic market.
- Coverage: To cover the most relevant behind-the-border measures while retaining time series length, measures implemented up to June 2009 are included; by including measures through June 2009, coverage of more than 80 percent of imports affected by behind-the-border measures throughout the entire sample is achieved.
- Graphical result (Figure 3 analog):
  - Figure 3 indicates an 8 percent average decline in imports as a result of behind-the-border measures in the months immediately after implementation.
  - From five months after implementation onwards, trade seems to return to its normal level.
- Regression analog:
  - Our closest regression analog to the figures is Regression 1 (Table 2). It reports an average import-reducing impact of behind-the-border measures of roughly 15 percent (=e^-0.165).
- Additional note:
  - While behind-the-border measures could also be expected to increase exports, this does not seem to have been the case in practice, likely because firms receiving bailouts were too fragile to make further inroads into export market.

### Export restrictions and export support measures
- Analog figures for export restrictions and export support measures are presented in the appendix (Figures A1 and A2).
- Reliability: Results on export measures are less reliable given the small number of implemented measures, reflected in their insignificant regression coefficients.
- Graphical patterns:
  - Trade affected by export support measures gradually increased, as expected.
  - Contrary to initial intuition, trade in products subject to export restrictions actually increased; possible explanation: high incentives for exports, such as a positive world-to-domestic price differential, when export restrictions are imposed.
- Regression coefficients for export measures are generally not significant.

### Estimation strategy (overview)
- Two-step approach:
  - Step 1: Estimate protectionism’s trade impact at the product-level.
  - Step 2: Multiply product-level estimates with amounts of trade affected by measures to calculate impacts on aggregate global trade.
- Rationale for product-level identification:
  - Global trade experienced an unprecedented collapse as the global crisis broke out, which for the most part was unrelated to protectionism.
  - The scope of new protectionist measures was not widespread: WTO estimates that between late 2008 and late 2010, new trade restrictions accumulated to cover 1.9 percent of global trade in goods (WTO, 2010b).
  - In our sample, as of the last quarter of 2009, 3½ percent of trade was affected.
- Aggregate detection difficulty:
  - Given that large scale protectionism was prevented, it would be near impossible to detect any protectionist impact in aggregate trade data, particularly given the contemporaneous trade collapse due mostly to other causes.

### Econometric specification
- Estimation equation (first-differenced gravity equation):
  - ln(Imports)_{ijpt}Δ = β · ProtDummy_{ijpt}Δ + TVFE_{ijpt}Δ + ε_{ijpt}
  - Where ln(Imports)Δ is the 12-month change in the U.S. dollar value of log imports.
  - TVFE stands for one or more sets of time-varying fixed effects.
  - ProtDummy counts the number of crisis protectionist measures applied to any given trade flow (takes positive integer values of 0, 1, 2, ...).
  - i, j, p, and t index importers, exporters, 4-digit HS product categories, and time (months), respectively.
- Justification for first-differencing:
  - Controls for long-run determinants of trade that are constant over time (e.g., distance, common language, colonial relationships).
  - Addresses product-level seasonality by focusing on 12-month changes (compares each of the 12 months after imposition to the corresponding month one year earlier).
  - Computational advantage: differencing eliminates time-invariant country-pair-product specific determinants, allowing inclusion of two sets of time-varying fixed effects given high-dimensional fixed effects constraints.

### Time-varying fixed effects considered
- Time-varying product fixed effects (TVP): control for global shocks to specific products (closest analog to graphical analysis).
- Time-varying importer fixed effects (TVIM): control for changes in an importer’s trade determinants that affect all products equally, and importer-specific multilateral resistance.
- Time-varying exporter fixed effects (TVEX): control for exporter-side supply shocks and exporter-specific multilateral resistance.
- Time-varying country-pair fixed effects (TVCP): combine TVIM and TVEX and control for changes in bilateral trading costs that affect all products.
- Time-varying importer-product fixed effects (TVIMP): allow product fixed effects to vary by importer.
- Time-varying exporter-product fixed effects (TVEXP): allow product fixed effects to vary by exporter.
- Practical note: Guimaraes and Portugal’s (2009) iterative technique is used to accommodate two high-dimensional fixed effects at a time; first-differencing is crucial to eliminate one set and enable inclusion of two TVFEs.

### Protectionist dummy construction and aggregation to aggregate impact
- ProtDummy counts the number of protectionist measures affecting each observation (0, 1, 2, ...); β coefficients are interpretable as the average trade impact of one measure.
- Data constraints limit incorporation of magnitude of measures (e.g., tariff increases); robustness partly addressed by splitting dummies by types of measures and implementing countries.
- Aggregation procedure:
  - Multiply estimated percentage reduction in product-level trade by the amount of trade affected.
  - Sum across all protectionist dummies (e.g., tariff dummy, antidumping dummy etc).
- Empirical simplification justified: large majority of observations affected by protectionism are affected by one measure only.
  - Observations affected by one import restriction numbered 56,050 (0.57 percent of the total sample).
  - Observations affected contemporaneously by two, three, and four import restrictions numbered 4,780; 218; and 2, respectively.
  - The remainder of the sample (9,817,431 observations) was unaffected by import restrictions.
- Robustness: results do not vary if the protectionist dummy is instead coded as taking the values of 0 and 1 only.

### Baseline results and preferred specification
- Model selection:
  - Six different fixed-effect setups were scrutinized to arrive at the preferred TV product & country-pair FE specification.
  - Specification progression and fit:
    - Specification 1: TV product FEs only (closest to graphical analysis).
    - Specification 2: adds importer FEs; statistically outperforms Specification 1.
    - Specification 3: adds country-pair FEs; outperforms Specification 2 and becomes preferred.
    - Specification 4: generalizes Specification 2 by allowing demand shocks to vary by product within country; preferred to Specification 2 but less so than Specification 3.
    - Specifications 5 and 6: generalizations of Specification 4; Specification 3 remains preferred.
  - F-statistics:
    - The F-statistic rejects specification 2 in favor of specification 3 with a value of 1.80.
    - Specification 2 is rejected in favor of specification 4 with an F-value of 1.12.
- R-squared and interpretation:
  - Low R-squared values are not surprising because the dependent variable is volatile 12-month differences in detailed monthly product-level trade flows.
  - Protectionist dummies are highly statistically significant but explain only a small fraction of total variance given their rarity in the sample.

### Key quantitative regression findings (graphical-regression analogs and interpretation)
- Import restrictions:
  - Regression 1 reports the trade-reducing impact of import restrictions at roughly 5 percent (=e^-0.048).
- Behind-the-border measures:
  - Figure-based average immediate decline in imports: 8 percent.
  - Regression 1 reports an average import-reducing impact of roughly 15 percent (=e^-0.165).
- Sample coverage:
  - Measures implemented up to March 2009 cover more than 80 percent of total trade affected by protectionist measures in the study.
  - By including measures through June 2009, coverage of more than 80 percent of imports affected by behind-the-border measures is achieved.
  - In the sample, as of the last quarter of 2009, 3½ percent of trade was affected.

*Source: _wp11139 (content excerpt).*

### 30.      Our baseline focuses on average effects of all import restrictions and all behind-

### _wp11139 - 30.      Our baseline focuses on average effects of all import restrictions and all behind-

### Baseline estimates of protectionist measures
- The baseline focuses on average effects of all import restrictions and all behind-the-border measures; the protectionist dummy is not split by subcategory (tariffs, antidumping duties, nontariff barriers, etc.).
- Border (import-restrictive) measures:
  - Estimated trade decline in response to import-restrictive border measures is 5 percent, but may be as high as 8 percent (Table 2, upper panel).
  - Preferred estimate (Specification 3) indicates a decrease in affected imports by 5 percent (=e^-0.051).
  - More parsimonious Specifications 1 and 2 corroborate the 5 percent result; all estimates are highly statistically significant.
  - Detailed importer-product fixed effects in Specifications 4-6 imply an impact on the order of 8 percent, suggesting countries imposed barriers in products where trade fell less (or rose) during the crisis.
- Behind-the-border measures:
  - Estimated to decrease trade by 7 percent according to preferred estimates (Specification 3).
  - Specification 1 (no country-specific shocks) indicates a much higher impact of -15.2 percent; much of this effect disappears when importer fixed effects are added in Specification 2.
  - When TV Importer-Product FEs are included (Specifications 4-6), impacts of behind-the-border measures tend to be absorbed by these fixed effects because such measures do not discriminate among trading partners but between domestic and foreign firms.

### Interpretation and measurement caveats
- Protectionist coefficients are semi-elasticities because changes in log imports are the dependent variable while protectionist variables are dummies.
- Product-level estimates are lower bounds:
  - Many measures (e.g., anti-dumping duties) are applied at 6- or 8-digit tariff lines or to specific firms; estimates are coded at the 4-digit level for data availability.
  - If estimation were at 6- or 8-digit levels, product-level estimates would likely be higher because most of the 4-digit category may be unaffected by the protectionist measure.
  - This downward bias at the product level is addressed through multiplication used to derive aggregate results.

### Aggregate results
- Import restrictions estimated to have reduced world trade by 0.21 percent in the last quarter 2009.
  - In the sample (covering 80 percent of world trade) this 0.21 percent reduction translates into a quarterly trade loss of $4.6 billion when applied to quarterly trade flows in the last quarter of 2009 using product-level coefficients of Specification 3.
  - “Affected quarterly trade” stood at $77.7 billion in the last quarter of 2009.
- Border vs behind-the-border contributions (last quarter of 2009):
  - Border measures reduced trade by $2.1 billion.
  - Behind-the-border measures reduced trade by $2.6 billion.
  - There were 40 behind-the-border import restrictions compared to 239 border measures; a typical single behind-the-border measure distorted trade about seven times more than a typical border measure.
- Potential gains from removing crisis measures:
  - Complete removal of crisis protectionist measures implemented up to early 2010 could boost annual world trade by some $30-35 billion.
    - A 0.21 percent reduction in annual global trade amounts to just this amount in non-crisis years.
  - Using base case estimates, removing crisis measures is equivalent to about 1/7 (=0.21/1.50) of the aggregate trade impact under the July 2008 Doha draft modalities for agriculture and industrial goods (Doha draft estimated to boost world trade by 1.5 percent).
- Data-limitation adjustments and alternative scenarios:
  - Inadequate information on 38 percent of GTA “red” measures implemented during the study period leads to underestimation of affected trade and likely of the 0.21 percent reduction.
  - If excluded measures were exactly as restrictive as those in the estimation sample, impact would rise to 0.34 percent or $50-60 billion annually.
    - Calculation reference: 0.21*(508/314), where 508 is total GTA measures and 314 is measures included in estimation sample (including export measures).
  - If all excluded measures were behind-the-border measures (extreme case) and the same product-level coefficients apply, total trade impact could reach 0.75 percent of world trade or $110-125 billion a year, with behind-the-border measures accounting for most distortion.
    - Calculation reference: 0.11*((508-314+40)/40)+0.10, where 40 is the number of behind-the-border measures in estimation sample and 0.10 percent is estimated trade distortion from border measures.

### Robustness checks
- Three focused robustness checks on baseline Specifications 3 and 6:
  1. Adding dummies for protectionist export measures:
     - Inclusion of export measures does not change coefficients for import restrictions.
     - Neither export restrictions nor export support measures significantly affected trade flows during the crisis.
     - Less than 20 export restrictions and less than 20 export support measures were registered in the sample.
  2. Dependent variable as changes in trade volumes (instead of trade value):
     - Results are weakened in magnitude and statistical significance but main results broadly hold.
     - In TV product and country-pair FE specification, border import restrictions still significantly discourage trade with an estimated reduction of 2.5 percent (compared to 5 percent in baseline).
     - Behind-the-border measures retain negative signs but lose statistical significance.
  3. Protectionist dummies coded as 0-1 rather than 0-1-2-...:
     - Restricting dummies to 0 or 1 leaves baseline results almost unchanged (very few observations subject to more than one new measure).

### Effects by measure types (preferred TV product & country-pair FE specification)
- Disaggregation categories for border measures (Regression 7):
  - (i) tariffs and import bans
  - (ii) trade defense measures
  - (iii) non-tariff barriers (NTBs): mainly licensing requirements and sanitary and phytosanitary restrictions
  - (iv) discriminatory purchasing measures: local content provisions, public procurement, consumption subsidies
- Disaggregation for behind-the-border measures:
  - (v) bailouts (directed to specific firm(s))
  - (vi) domestic subsidies (apply to an entire sector)
  - (vii) investment subsidies (discriminatory investment incentives)
- Key findings by measure type:
  - Tariffs:
    - Tariff measures’ impact is statistically insignificant; contribute little to overall impact despite 67 tariff measures recorded.
  - Quotas and import bans:
    - Decreased trade in affected products by 24 percent and 13 percent respectively; coefficients borderline statistically significant but aggregate contribution low due to narrow targeting.
  - Trade defense measures:
    - Narrowly focused by design; 102 different duties imposed.
    - Trade in affected 4-digit products decreased by 16 percent, implying greater collapse at sub-4-digit levels.
  - Non-tariff barriers (NTBs):
    - New licensing requirements drove NTB impact, with a 9 percent trade decrease at the product level; narrow application limited aggregate impact.
  - Discriminatory purchasing measures:
    - Most harmful to aggregate imports of all border measures.
    - Consumption subsidies caused the largest import decreases at the product level (-9 percent) and reduced global trade by -0.03 percent.
    - Local content requirements covering entire domestic market were similarly harmful.
    - Public procurement measures did not show a trade impact in this analysis.

*Italic source attribution: Excerpt from IMF working paper content unit _wp11139 (pages 30–25 of the supplied chapter).*

### 43.      Among behind-the-border measures, bailouts and domestic subsidies were

### _wp11139 - 43.      Among behind-the-border measures, bailouts and domestic subsidies were

### Summary of core findings
- Bailouts and domestic subsidies each decreased trade in affected products by around 7 percent.
- Whether discriminatory aid was directed to specific firms or entire sectors did not change the deterrent effect on imports.
- Aggregate impacts reflect closely the amount of trade covered by these measures:
  - Bailouts aggregate impact: -0.09 percent.
  - Domestic subsidies aggregate impact: -0.03 percent.
- Discriminatory subsidies encouraging investment by domestic firms only did not cause a contemporaneous statistically significant trade reduction (though future effects remain possible as new capacity comes online).

### Overall protectionist impact (disaggregation)
- Disaggregate regressions produced estimates of reductions in aggregate world trade of 0.20 and 0.19 percent—very much in line with the baseline estimate of 0.21 percent.
- Estimates cluster in a relatively tight interval around the 0.21 percent baseline estimate.

### Effects by country group
- Border measures:
  - Advanced countries’ border measures reduced affected product-level imports by 8 percent; developing countries’ measures reduced them by about half that amount.
  - Advanced countries account for about ¾ of the aggregate trade distortion implied by border measures.
  - Regional drivers: results largely driven by North America; Latin American border measures reduced affected imports by -9 percent (but covered a relatively small amount of trade).
  - East Asian border measures show the highest impact on aggregate world trade though the product-level coefficient is small and statistically insignificant.
- Behind-the-border measures:
  - Developing countries’ behind-the-border measures were more trade restrictive at the product level (-9 percent), mainly driven by upper middle income countries and Central and East Asia.
  - Advanced countries’ behind-the-border measures were more harmful on aggregate because they targeted larger import flows; advanced countries bore more than 85 percent of the aggregate impact of behind-the-border measures.
  - A substantial part of the estimated behind-the-border impact originates in North America, though this estimate may be biased by omitted protectionism-unrelated factors (e.g., a large crisis-induced reduction in US vehicle purchases).
- Exports:
  - Developing country exports were hurt somewhat more by border measures: each of developed and developing country exports were reduced by about 0.05 percent (of world trade) on aggregate.
  - The shock was relatively more accentuated for developing countries because they account for less than half of world exports.
  - Lower middle income countries experienced a considerably higher impact than upper middle income countries; low income countries (LICs) estimated to have fallen by even more but with statistically insignificant coefficients (likely due to fewer measures affecting them).
  - Regional estimates show exports from the Africa grouping (defined to include both MENA and SSA countries) face the most restrictive border measures.
  - On the advanced country side, North America faces the more restrictive measures, while Western Europe suffers the larger aggregate impact.
- Behind-the-border export impacts remained largely within advanced countries:
  - Product-level trade reduction for advanced country exports facing behind-the-border measures: 8 percent.
  - Much of the impact concentrated in advanced and emerging Europe; Latin America faced the highest product-level impact from new behind-the-border measures with a 14 percent reduction in affected exports.
  - Africa and Asia were seemingly unscathed by behind-the-border measures.

### Effects by sector
- Sectors with largest impacts (Table 9 classification):
  - Textiles and machinery exhibited statistically significant reductions in affected imports among border measures: -7 and -5 percent, respectively.
  - Machinery accounts for more than half of the reduction in world imports among border measures because affected trade flows are higher than those in textiles.
  - Behind-the-border measures mainly obstructed imports of machinery and transport equipment, reducing affected flows by 12 percent on average.
  - Transport sector accounts for the lion’s share of the aggregate behind-the-border impact, with an aggregate impact tallied at -0.14 percent reduction in world trade.
- HS 2-digit mapping to broad sectors used in analysis:
  - Agriculture (01-15), Processed Food (16-24), Minerals (25-27), Metals (68-83), Wood (44-49, 92, 94, 97), Chemicals (28-40), Textiles (41-42, 50-67), Machinery (84, 85, 90, 91, 93, 95, 96), Transportation (86-89).

### Effects by time periods
- Time periods defined by trend in global trade flows:
  - Trade collapse: up to January 2009.
  - Trade stabilization: February-May 2009.
  - Trade recovery: from June 2009 onwards.
- Measures implemented during first nine months after the Lehman collapse (September 2008) were particularly harmful:
  - Border measures implemented before the trade recovery started in June 2009 decreased affected trade flows by 5-6 percent on average.
  - The trade stabilization phase (February-May 2009) saw the most and broadest measures; these accounted for almost the entire aggregate impact for both border and behind-the-border measures, reducing world trade by roughly 0.1 percent each.
  - Behind-the-border measures implemented during stabilization reduced trade flows by 8 percent on average.
  - Measures implemented after trade was already recovering did not show a statistically significant effect (note: sample limitations for post-June 2009 measures may affect identification).
- Persistence:
  - Measures remaining in effect during the trade recovery continued to harm trade:
    - Estimated product-level reductions for measures in effect during the recovery: 4 percent for border and 6 percent for behind-the-border measures.
  - Removal of crisis measures still in effect could boost global trade by at least 0.2 percent annually.

### Conclusion and policy implications
- Methodological approach:
  - Quantified trade impact of a broad set of crisis protectionist measures using product-level trade data and extensive fixed effect specifications; aggregate impacts derived by multiplying product-level coefficients by the amount of trade affected using a dummy variable approach.
- Main quantitative conclusions:
  - Affected trade flows fell by 5 percent in response to border measures and 7 percent in response to behind-the-border measures (with these impacts possibly being somewhat underestimated).
  - Traditional trade measures, most notably tariffs, had little impact; antidumping duties and unconventional protectionism (non-tariff barriers, discriminatory purchasing policies, bailouts, domestic subsidies) substantially reduced affected trade flows.
  - Crisis protectionism measures included in the study are estimated to be reducing global trade by at least $30-35 billion or 0.2 percent annually.
- Policy recommendation:
  - Policymakers should remain attentive to protectionist pressures and consider removing crisis protectionist measures to aid the global economic recovery and avoid perpetuating associated trade losses.

*Source: _wp11139 - 43.      Among behind-the-border measures, bailouts and domestic subsidies were*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11139.pdf_