## wpiea2023099-print-pdf

## Source details

**Canonical URL:** [wpiea2023099-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023099-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023099-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023099-print-pdf.pdf.json)

---

### Context and role of Temporary Trade Barriers (TTBs) within the WTO framework
- The GATT/WTO multilateral processes have succeeded in reducing world average tariffs below 3%.
- The WTO allows governments to:
  - act against dumping where there is genuine (“material”) injury to the competing domestic industry;
  - launch investigations and charge extra duty (known as “countervailing duty”) on subsidized imports found to be hurting domestic producers;
  - restrict imports temporarily (safeguard actions) if a domestic industry is seriously injured or threatened with injury caused by a surge in imports.
- TTBs can be applied as temporary measures during investigations; the dataset shows the average time between the start of the investigation and the determination date is 12 months (see Figure A3).
- Empirical literature cited:
  - Bown and Crowley (2013a), Bown and Tovar (2011), Kuenzel (2020), and Bown (2022) on responsiveness of TTBs to macroeconomic conditions, substitution with tariff policy, and disproportionate targeting of China.

### Definition and measurement of retaliation
- Retaliation focus: rules-inconsistent (non-cooperative) retaliation using TTBs.
- Operational definition:
  - Retaliation is TTB actions by country j that occur not more than x days after TTBs taken by trading partner i, where x is the median response time (in days) between j’s actions and those of all its trading partners.
  - Intensity of retaliation by country j = number of country j’s TTB measures following a TTB measure imposed by country i within interval x days.
- Data: daily TTB data from the World Bank’s TTB Database (Bown, 2015) at HS6 product level; high-frequency data key to capture short-interval responses.

### Data, sample, and empirical design
- Sample coverage:
  - 1,220 subsectors (HS4 digits)
  - 25 advanced and emerging economies
  - 1989-2019
- Dataset key features:
  - Daily frequency at HS6 level; measures classified as anti-dumping (AD), countervailing (CVD) and global safeguards (GS).
  - Four-dimensional panel: domestic country j, partner country i, time t, and sectors k.
  - Sectoral aggregation to HS4 to reduce zero/one occurrences and capture intensive margin (number of HS6 products targeted within each HS4 sector).
- Fixed effects and identification:
  - Country–time and sector–time fixed effects employed; robustness check includes country-pair-time fixed effects.
  - Focus on trade authority’s determination date for timing of retaliatory TTBs.

### Country-specific timing and retaliation measure construction
- Distribution of days between opposing measures:
  - Mode around 50 days; median of 165 days for the whole sample.
- Selected country-specific median response times (days):
  - China: 63
  - United States: 81
  - European Union: 100
  - Taiwan: 64
  - India: 103.5
  - Japan: 214
  - Argentina: 186
  - Whole sample median: 165
- Implication: many responses occur within a year; calendar-year windows miss a large share (two thirds) of retaliation.
- Example timeline (China–US 2016–2017) illustrates use of country medians (China median = 63; US median = 81) to determine whether an action counts as retaliation.
- Robustness: baseline results robust to alternative thresholds (whole sample median 165 days or first quartile by country).

### Stylized facts and descriptive patterns
- Prevalence and dispersion:
  - Retaliation is common but widely dispersed across countries.
  - Smaller countries and emerging market economies retaliate less than larger or richer countries.
  - Retaliation represents around 10% of TTB use sample-wide; up to 20% for some countries.
  - Recourse to retaliatory TTBs increased over time, peaking in the early 2000s and after the Global Financial Crisis (GFC).
  - Time evolution: peak share to total measures of 14% in the 2000s; share rose after the GFC, reaching 9% in 2012 and 9% in 2019.
- Sectoral patterning:
  - Same-sector retaliation is uncommon: modal share of same HS4-sector retaliation is close to zero.
  - Retaliation typically occurs across many sectors simultaneously rather than targeting the originally injured sector.
- Intensive margin matters:
  - When foreign country targeted at least 9 products within the last 165 days, probability of foreign measure preceding a domestic one rises.
- Unconditional and conditional probabilities (percent):
  - Probability of at least one domestic measure, whole sample: 3.9
  - Probability of at least one domestic measure, largest users (12 countries): 58
  - Probability of at least one domestic measure preceded by a foreign measure, whole sample: 4.6
  - Probability of at least one domestic measure preceded by a foreign measure, largest users: 26.7
  - Probability of a domestic measure preceded by more than 9 foreign measures, whole sample: 12
  - Probability of a domestic measure preceded by more than 9 foreign measures, largest users: 41
  - Probability of a domestic measure preceded by less than 9 foreign measures (conditional), whole sample: 3.8

### Descriptive statistics (selected, preserved exactly)
- TTB_jik,t mean: .0060998; standard deviation: .1810874; min: 0; max: 29
- TTB_ijk,t mean: .0000156; standard deviation: .007763605; min: 0; max: 3
- TTB_ijk',t mean: .0011662; standard deviation: .167196084; min: 0; max: 84
- ∆Imports_jik,t−1 mean: 40.57498; standard deviation: 13450.24; min: -.9999998; max: 1.27e+07
- ∆RER_ji,t−1 mean: -.0972273; standard deviation: 11.5155; min: -66.255341; max: 139.914
- TradeBalance_ji,t−1 mean: 6.975381; standard deviation: 12.42457; min: -7.013038385; max: 385.05662
- ExportShare_ji,t−1 mean: .7171183; standard deviation: 12.66001; min: -52.595128; max: 283.90308
- Overhang_jik,t−1 mean: 12.45125; standard deviation: 15.17271; min: -358.697; max: 24.17
- ∆Tariff_jik,t−1 mean: .26747833; standard deviation: 3.675996; min: -796.83774; max: 4.34
- Disputes_ji,t−1 mean: .107407; standard deviation: 1.435497; min: 0; max: 906
- Number of observations for baseline TTB estimation: 1,476,615

### Empirical framework (baseline)
- Outcome variable: TTB_jik,t = count of HS6 products aggregated to HS4 sector k targeted by country j against partner i in year t.
- Estimator: fixed-effect Poisson applied to
  TTB_jik,t = β1 TTB_ijk,t + β2 TTB_ijk',t + θ' Z_ji,t−1 + α_jt + δ_it + γ_st + μ_ji + ν_jik,t
  - TTB_ijk,t: measures introduced by j’s partner i in same sector k.
  - TTB_ijk',t: measures introduced by i in other sectors k'.
  - Z_ji,t−1: bilateral-time varying controls.
  - α_jt and δ_it: country-time fixed effects.
  - γ_st: sector-time fixed effects.
  - μ_ji: country-pair fixed effects.
  - Standard errors clustered at country-pair-HS4-sector (jik) level.
- Non-linear interactions allow terms β3 TTB_ijk,t M_ji,t−1 + β4 TTB_ijk',t M_ji,t−1 where M are macro variables (e.g., unemployment, partner size).

### Main empirical findings (summary)
- Retaliation magnitude:
  - A one standard-deviation increase in the number of new TTBs in a given HS4 sector by country i on country j increases the number of newly targeted products by j on i by 1% both in the same sector and in other sectors (column 6 reported).
- Selected coefficients from baseline (Table 4, preserved exactly):
  - TTB_ij,k,t: 0.592*** (column 1); 0.961*** (column 2); 0.610*** (column 3); 0.595*** (column 4); 0.613*** (column 5); 1.212*** (column 6).
  - TTB_ijk′,t: 0.0747*** (column 1); 0.0672*** (column 2); 0.0769*** (column 3); 0.0750*** (column 4); 0.0766*** (column 5); 0.0696*** (column 6).
- Effects of one SD increase (from Table 4):
  - Effect of one SD increase in TTB_ijkt (%): 0.472, 0.767, 0.486, 0.474, 0.489, 0.966 (columns 1–6).
  - Effect of one SD increase in TTB_ijk′t (%): 1.117, 1.005, 1.150, 1.122, 1.145, 1.041 (columns 1–6).
- Observations and fit (Table 4):
  - Observations: 2,044,522; 1,291,826; 2,044,522; 2,044,522; 2,044,514; 1,291,822 (columns 1–6).
  - Pseudo-R2: 0.372, 0.384, 0.373, 0.372, 0.381, 0.394 (columns 1–6).

### Baseline controls and interpretation
- Tariff measures:
  - Overhang_jik,t−1: -0.0142*** (column 5); -0.0136*** (column 6). Interpretation: having room to raise tariffs decreases incentives to resort to TTBs.
  - ∆Tariff_jik,t−1: 0.0114*** (column 5); 0.0112*** (column 6). Interpretation: when overhang is null, an increase in applied tariff increases the use of TTBs.
- WTO disputes:
  - Disputes_ji,t: -0.0988 (column 5); -0.269*** (column 6).
  - Disputes_ji,t+1: 0.233*** (column 5); 0.222*** (column 6).
  - Disputes_ij,t: 0.122** (column 5); 0.119** (column 6).
  - Disputes_ij,t−1: 0.114** (column 5); 0.115** (column 6).
  - Interpretation: current and future disputes are associated with increases in domestic TTBs; expectations of future attacks foster domestic protectionism.
- Comparative advantage:
  - CA_jk: -0.291*** (column 5); -0.261*** (column 6). Interpretation: import protection tends to be lower in sectors with comparative advantage.
  - CA_ik: 0.740*** (column 5); 0.691*** (column 6). Interpretation: countries tend to target comparative advantage sectors of trading partners.
- Inclusion of TTB_ij,t+1: 0.00427** (column 5); 0.00927*** (column 6) — controls for fear of retaliation.

### Fixed-effects comparison and sensitivity
- Dropping country-time FE (Table 5) increases coefficients on TTB_ij,k,t and makes macro variables significant (∆RER, RTA, ExportShare, TradeBalance), indicating country-time FE absorb much macro variation in baseline.
- Estimation alternatives and robustness:
  - OLS confirms significant retaliation effects.
  - Including country-pair-time FE slightly increases impact of a standard deviation.
  - HS6-digit disaggregation confirms cross-product retaliation.
  - Results robust to using only AD, alternative time thresholds (whole sample median 120 days and first quartile), subsample splits (pre/post GFC, AE vs EM).

### Nonlinearities and heterogeneity (section 5.3)
- Trade-policy environment:
  - Membership in same RTA associated with stronger retaliation: interaction RTA_ji,t−1 coefficient: 0.668*** (standard error 0.0906).
  - Past reductions in tariffs not mitigating retaliation: interaction ∆Tariff_jik,t−1: 0.00259 (standard error 0.00851) — not significant.
  - Presence of unresolved disputes does not reduce retaliatory TTBs; selected interactions:
    - Disputes_ji,t interaction: 0.0596 (standard error 0.0367) — not significant.
    - Disputes_ji,t+1 interaction: 0.371** (standard error 0.145).
    - Disputes_ij,t−1 interaction: 0.313*** (standard error 0.0654).
- Macroeconomic amplifiers:
  - Higher unemployment strengthens retaliation: TTB_ij,t ∗ Unemployment_j,t−1 coefficient: 0.0144*** (standard error 0.00523).
- Large-user heterogeneity:
  - Other TTB_j,t ∗ Large user_j: -0.0539 (standard error 0.0537) — not significant.
  - Other TTB_j,t ∗ USA_j: -0.0530** (standard error 0.0248).
  - Other TTB_j,t ∗ EU_N_j: 0.666*** (standard error 0.0882).
  - Other TTB_j,t ∗ CHN_j: 1.255*** (standard error 0.137).
- Targeting by comparative advantage:
  - TTB CA_j ij,t coefficient: 0.763*** (standard error 0.0448).
  - TTB CA_i ij,t coefficient: 1.059*** (standard error 0.178).
  - Interpretation: retaliatory TTBs more likely when foreign economy first targeted its own comparative advantage; retaliation tends to target comparative advantage sectors of trading partners.
- Representative coefficients across robustness tables (preserved):
  - Table 6: Effect of one SD increase in TTB_ijkt (%) across columns: 0.357, 2.528, 0.491, 0.628, 0.453, 0.390.
  - Table 6: Effect of one SD increase in TTB_ijk′t (%) across columns: 0.7482, 1.241, 1.1538, 0.5052, 2.692, 1.196.
  - Table 8: TTB_ij,t coefficients across columns (1)–(6): 0.0789***, 0.0722***, 0.0811***, 0.0793***, 0.0808***, 0.0744***.
  - Table 9: Baseline TTB_ij,t coefficients across columns: 0.0692***, 0.0732***, 0.0572***, 0.0643***, 0.112***, 0.0547***.

### Interpretation and implications emphasized in the paper
- Strategic behavior:
  - Use of TTBs for retaliation suggests actions driven more by perceptions of unfair foreign policies than by sector-specific injury.
  - The temporary nature of TTBs enables use as a safety valve or “vigilante justice” for countries unwilling or unable to pursue WTO litigation.
- Substitutability and complementarity:
  - Overhang negative and ∆Tariff positive implies having tariff room reduces TTB use, while applied tariff increases can complement TTBs when overhang is null.
- Effectiveness of WTO dispute mechanism:
  - Filing disputes does not appear to reduce rules-inconsistent retaliation; disputes can correlate with increased retaliatory responses.
- Policy-relevant heterogeneity:
  - Retaliation larger when tariffs cannot be used (e.g., within RTAs), larger when both countries are in a trade dispute, and larger during periods of higher unemployment.
  - Major economies (United States, European Union, China) display markedly different retaliation intensities.

*Source: wpiea2023099-print-pdf*

### 1.  Introduction .......................................................................................................

### wpiea2023099-print-pdf - 1.  Introduction

### Context and role of Temporary Trade Barriers (TTBs) within the WTO framework
- The GATT/WTO multilateral processes have succeeded in reducing world average tariffs below 3%.
- The WTO allows governments to:
  - act against dumping where there is genuine (“material”) injury to the competing domestic industry;
  - launch investigations and charge extra duty (known as “countervailing duty”) on subsidized imports found to be hurting domestic producers;
  - restrict imports temporarily (safeguard actions) if a domestic industry is seriously injured or threatened with injury caused by a surge in imports.
- TTBs can be applied as temporary measures during investigations; the dataset shows the average time between the start of the investigation and the determination date is 12 months (see Figure A3).
- Empirical literature documents that TTBs respond to macroeconomic conditions and may substitute for tariff policy: Bown and Crowley (2013a) show responsiveness to macroeconomic conditions; Bown and Tovar (2011) show TTBs can compensate for an official tariff decrease; Kuenzel (2020) provides evidence of substitution between bound tariff decreases and increased TTB use.
- Concerns exist that TTBs may be used for retaliation or protection beyond their WTO-allowed purposes; Bown (2022) documents disproportionate targeting of China.

### Definition and measurement of retaliation used in this paper
- The paper focuses on rules-inconsistent (non-cooperative) retaliation using TTBs.
- Retaliation is defined as TTB actions by country j that occur not more than x days after TTBs taken by trading partner i, where x is the median response time (in days) between j’s actions and those of all its trading partners.
- The intensity of retaliation by country j is measured as the number of country j’s TTB measures following a TTB measure imposed by country i within an interval of x days.
- Daily data on TTBs from the World Bank’s TTB Database (Bown, 2015) at the HS6 product level are used; high-frequency data are key to capture actions occurring within short intervals, addressing concerns that retaliatory measures within a year would otherwise be missed.

### Data, sample, and empirical design
- Sample coverage:
  - 1,220 subsectors (HS4 digits)
  - 25 advanced and emerging economies
  - 1989-2019
- Data sources and structure:
  - Daily TTB data at HS6 level from the World Bank’s TTB Database (Bown, 2015).
  - Use of sectoral (HS4) data permits distinguishing same-sector versus cross-sector retaliation.
  - Four-dimensional panel structure: domestic country j, partner country i, time t, and sectors k.
- Identification and fixed effects:
  - Country–time and sector–time fixed effects are employed to control for country- and sector-level shocks.
  - As a robustness check, the authors also control for country-pair-time fixed effects.
- Rationale for focusing on determination dates:
  - Public records provide dates for each step; the paper focuses on the trade authority’s determination date to examine the timing of retaliatory TTBs.

### Main empirical findings (summarized)
- Prevalence and dispersion:
  - Retaliation is common but widely dispersed across countries.
  - Smaller countries and emerging market economies retaliate less than larger or richer countries.
  - Recourse to retaliatory TTBs increased over time, peaking in the early 2000s and after the Global Financial Crisis (GFC).
- Sectoral patterning:
  - Retaliation appears more focused on protecting non-injured sectors than the injured sector.
  - Retaliation tends to protect many sectors simultaneously, indicating responses are not limited to the directly injured sector.
- Magnitude and robustness:
  - A one standard-deviation increase in the number of new TTBs in a given HS4 sector by country i on country j increases the number of newly targeted products by j on i by 1% both in the same sector and in other sectors.
  - This result is robust to controls for other trade policy instruments, such as tariff variations and trade disputes.
- Conditional amplifiers:
  - Retaliatory actions are larger when tariffs cannot be used (e.g., in the presence of a trade agreement).
  - Retaliation is larger when both countries are in a trade dispute, or when the domestic economy is anticipating further retaliation.
  - Retaliation is also larger in periods of higher unemployment.

### Interpretation and implications emphasized in the introduction
- The use of TTBs for retaliation suggests strategic behavior driven more by perceptions of unfair foreign policies than by sector-specific injury.
- The temporary nature of TTBs enables their use as a safety valve or as “vigilante justice” (direct unilateral retaliation) for countries unwilling or unable to pursue WTO litigation.
- The paper aims to provide evidence of rules-inconsistent retaliation by improving measurement (high-frequency sectoral TTB data) and controlling for confounding shocks via rich fixed effects.

*Source: wpiea2023099-print-pdf - 1. Introduction*

### 3.2    Temporary trade barriers data

### 3.2 Temporary trade barriers data

### Dataset and key features
- Panel of bilateral measures of import protection for 25 advanced and emerging market economies for the period 1989-2019.
- Daily data on temporary trade barriers (TTBs) drawn from the World Bank’s TTB Database (based on Bown, 2015), at the 6-digit Harmonized System (HS6) product level for 1989-2019, classified as anti-dumping (AD), countervailing (CVD) and global safeguards (GS).
- Three principal advantages of the dataset:
  - Daily frequency permits construction of within-year measures of retaliation.
  - Four-dimensional structure (k sectors, j domestic country, i partner country, t time periods) allows inclusion of country–time and sector–time fixed effects to control for aggregate and sectoral shocks, unobserved cross-country heterogeneity, trade deflection, and indirect impacts via other trading partners.
  - Sectoral disaggregation permits distinguishing retaliation in the same sector versus other sectors and captures intensive margin (number of HS6 products targeted) as well as extensive margin.

### Dependent variable and aggregation
- Dependent variable: count of HS6 imported products on which government of j implements a TTB against trading partner i in year t (TTB_jik,t).
- Aggregation: counts aggregated to HS4 product level to reduce zero/one occurrences and to capture intensive margin (number of HS6 products targeted within each HS4 sector).
- Retaliation measure: count of TTBs implemented by country i against j within x days of j’s action, where x is country-specific median response time.

### Country-specific reaction thresholds and timing patterns
- Distribution of days between opposing measures: mode around 50 days; median of 165 days for the whole sample.
- Country-specific median response times (selected examples from Table 1):
  - China: 63
  - United States: 81
  - European Union: 100
  - Taiwan: 64
  - India: 103.5
  - Japan: 214
  - Argentina: 186
  - Whole sample median: 165
- Implication: many responses occur within a year; approaches using previous calendar year windows miss a large share of retaliation (two thirds of TTBs would not be classified as retaliation under a previous-calendar-year measure).

### Stylized facts and patterns of trade retaliation
- Unconditional and conditional probabilities (Table 2, probabilities in percent):
  - Probability of at least one domestic measure, whole sample: 3.9
  - Probability of at least one domestic measure, largest users (12 countries): 58
  - Probability of at least one domestic measure preceded by a foreign measure, whole sample: 4.6
  - Probability of at least one domestic measure preceded by a foreign measure, largest users: 26.7
  - Probability of a domestic measure preceded by more than 9 foreign measures, whole sample: 12
  - Probability of a domestic measure preceded by more than 9 foreign measures, largest users: 41
  - Probability of a domestic measure preceded by less than 9 foreign measures (conditional), whole sample: 3.8
- Intensive margin matters:
  - When foreign country targeted at least 9 products within the last 165 days, probability of foreign measure preceding a domestic one rises (12% whole sample, 41% largest users).
  - Probability of targeting more than 9 foreign products is larger when foreign economy also targeted several products (29% vs 25%).
- Share of retaliation:
  - Retaliation represents around 10% of TTB use sample-wide.
  - Heterogeneity across countries: retaliation represents up to 20% of TTB use for some countries.
- Time evolution:
  - Retaliation increased during the 1990s with a peak share to total measures of 14% in the 2000s.
  - After a decline, share rose again after the GFC, reaching 9% in 2012 and 9% in 2019.
- Sectoral targeting:
  - Same-sector retaliation is uncommon: modal share of same HS4-sector retaliation is close to zero; retaliation typically occurs across many sectors simultaneously rather than targeting the same sector originally targeted by the trading partner.

### Constructing retaliation variable: country-specific threshold approach
- Define country-specific reaction threshold x equal to the median number of days taken by country j to implement a TTB following measures introduced by all its trading partners.
- Example timeline China–US (2016–2017) illustration:
  - US initiates measures on Jul. 20th; China responds on Nov. 18th (121 days); US implements another set on Jan. 11th (54 days later); China implements measures on Jan. 12th (1 day later).
  - Using medians: China median = 63 days; US median = 81 days. When considering determinants of Chinese measures, the Jul. 20th US set is not counted as retaliation (121 > 63) but the Jan. 11th US set is counted (1 day gap relative to China). When considering determinants of US measures, Chinese Nov. 18th measures are counted as retaliation (54 < 81).
- Robustness: baseline results are robust to alternative thresholds (whole sample median 165 days or first quartile by country).

### Empirical framework
- Outcome: TTB_jik,t = count of HS6 products aggregated to HS4 sector k targeted by country j against partner i in year t; over-dispersed non-negative count.
- Baseline estimation: fixed-effect Poisson estimator applied to
  TTB_jik,t = β1 TTB_ijk,t + β2 TTB_ijk',t + θ' Z_ji,t−1 + α_jt + δ_it + γ_st + μ_ji + ν_jik,t
  - TTB_ijk,t: measures introduced by j’s partner i in same sector k.
  - TTB_ijk',t: measures introduced by i in other sectors k'.
  - Z_ji,t−1: bilateral-time varying controls.
  - μ_ji: country-pair fixed effects.
  - α_jt and δ_it: country-time fixed effects.
  - γ_st: sector-time fixed effects.
  - Standard errors clustered at country-pair-HS4-sector (jik) level.
- Non-linear interactions specification:
  - Adds β3 TTB_ijk,t M_ji,t−1 + β4 TTB_ijk',t M_ji,t−1 terms to allow retaliation effects to vary with macro variables M (e.g., unemployment, partner size).

### Controls and additional considerations
- Controls included:
  - WTO disputes: present and future domestic WTO disputes and past and present foreign WTO disputes.
  - Share of country j’s exports towards country i (ExportShare_ji,t−1) to capture reluctance to retaliate when export exposure is high.
  - Macroeconomic variables and import levels/growth from CEPII’s BACI dataset (∆Imports_jik,t−1).
  - Variation of real exchange rate from USDA (∆RER_ji,t−1).
  - Bilateral trade balance (TradeBalance_ji,t−1).
  - Tariff levels, tariff changes, tariff overhang (difference between bound and applied tariffs) from UNCTAD TRAINS (Overhang_jik,t−1, ∆Tariff_jik,t−1).
  - Dummy for same Regional Trade Agreement (from CEPII Gravity).
  - Comparative advantage index (CA_ik, CA_rk) and dummies indicating comparative advantage.
- Accounting for alternative responses:
  - Inclusion of WTO dispute measures accounts for rule-consistent avenues as substitutes for retaliatory TTBs.
  - Sector-time fixed effects help absorb third-party effects and trade diversion that could otherwise be misattributed to retaliation.

### Descriptive statistics (Table 3)
- TTB_jik,t mean: .0060998; standard deviation: .1810874; min: 0; max: 29
- TTB_ijk,t mean: .0000156; standard deviation: .007763605; min: 0; max: 3
- TTB_ijk',t mean: .0011662; standard deviation: .167196084; min: 0; max: 84
- ∆Imports_jik,t−1 mean: 40.57498; standard deviation: 13450.24; min: -.9999998; max: 1.27e+07
- ∆RER_ji,t−1 mean: -.0972273; standard deviation: 11.5155; min: -66.255341; max: 139.914
- TradeBalance_ji,t−1 mean: 6.975381; standard deviation: 12.42457; min: -7.013038385; max: 385.05662
- ExportShare_ji,t−1 mean: .7171183; standard deviation: 12.66001; min: -52.595128; max: 283.90308
- Overhang_jik,t−1 mean: 12.45125; standard deviation: 15.17271; min: -358.697; max: 24.17
- ∆Tariff_jik,t−1 mean: .26747833; standard deviation: 3.675996; min: -796.83774; max: 4.34
- Disputes_ji,t−1 mean: .107407; standard deviation: 1.435497; min: 0; max: 906
- Number of observations for baseline TTB estimation: 1,476,615

*Source: wpiea2023099-print-pdf - 3.2 Temporary trade barriers data*

### 5.1    Baseline results

### 5.1    Baseline results

### Main retaliation estimates (baseline specification, equation 1)
- On average, a one standard deviation increase in the number of TTBs in a HS4-sector by country i on country j increases the number of newly targeted products by 0.97% in the same sector and 1% in another sector (column 6).  
- Coefficients for the variables of interest in Table 4 (selected):
  - TTB_ij,k,t: 0.592*** (column 1); 0.961*** (column 2); 0.610*** (column 3); 0.595*** (column 4); 0.613*** (column 5); 1.212*** (column 6).
  - TTB_ijk′,t: 0.0747*** (column 1); 0.0672*** (column 2); 0.0769*** (column 3); 0.0750*** (column 4); 0.0766*** (column 5); 0.0696*** (column 6).
- Effects of a one SD increase in TTB_ijkt and TTB_ijk′t (from Table 4):
  - Effect of one SD increase in TTB_ijkt (%): 0.472, 0.767, 0.486, 0.474, 0.489, 0.966 (columns 1–6).
  - Effect of one SD increase in TTB_ijk′t (%): 1.117, 1.005, 1.150, 1.122, 1.145, 1.041 (columns 1–6).
- Observations and fit (Table 4):
  - Observations: 2,044,522; 1,291,826; 2,044,522; 2,044,522; 2,044,514; 1,291,822 (columns 1–6).
  - Pseudo-R2: 0.372, 0.384, 0.373, 0.372, 0.381, 0.394 (columns 1–6).
- Estimator and fixed effects: fixed-effect Poisson with country-time (jt and it), country-pair (ji), and HS2 sector-time (st) fixed effects; standard errors clustered at country-pair-hs4 sector (jik).

### Controls and interpretation of trade-policy interactions
- Tariff overhang and applied tariff:
  - Overhang_jik,t−1: -0.0142*** (column 5); -0.0136*** (column 6). Interpretation: having room to raise tariffs decreases incentives to resort to TTBs.
  - ∆Tariff_jik,t−1: 0.0114*** (column 5); 0.0112*** (column 6). Interpretation: when overhang is null, an increase in applied tariff increases the use of TTBs, suggesting complementarity between instruments.
  - Overhang * ∆Tariff interactions included (coefficients very small and not significant in baseline columns reported).
- Disputes (WTO dispute settlement filings):
  - Disputes_ji,t: -0.0988 (column 5); -0.269*** (column 6).
  - Disputes_ji,t+1: 0.233*** (column 5); 0.222*** (column 6).
  - Disputes_ij,t: 0.122** (column 5); 0.119** (column 6).
  - Disputes_ij,t−1: 0.114** (column 5); 0.115** (column 6).
  - Interpretation: current domestic and foreign disputes are associated with an increase in the number of domestic TTBs; future disputes (Disputes_ji,t+1) are associated with higher domestic TTBs, suggesting expectations of future attacks tend to foster domestic protectionism.
- Other controls (selected):
  - ∆Imports_jik,t−1: coefficients around -0.000126 to -0.000211 (not statistically significant in baseline).
  - ∆RER_ji,t−1: 0.0389 to 0.0500 (not robustly significant in baseline with full FE).
  - TradeBalance_ji,t−1: -0.000175 (column 5); -0.0182** (column 6).
  - CA_jk: -0.291*** (column 5); -0.261*** (column 6). Interpretation: import protection tends to be lower in sectors with comparative advantage (governments protect less competitive sectors).
  - CA_ik: 0.740*** (column 5); 0.691*** (column 6). Interpretation: countries tend to target comparative advantage sectors of trading partners.
  - TTB_ij,t+1: 0.00427** (column 5); 0.00927*** (column 6) — included to control for potential fear of retaliation.

### Fixed-effects comparison and alternative specification (Table 5: sector-time and country-pair FE only)
- When re-estimating with only sector-time and country-pair fixed effects (dropping country-time FE), main points:
  - TTB_ij,k,t coefficients increase: 0.774***, 1.157***, 0.772***, 0.789***, 0.805***, 1.216*** (columns 1–6 in Table 5).
  - TTB_ijk′,t: 0.0466***, 0.0415***, 0.0491***, 0.0477***, 0.0500***, 0.0502*** (columns 1–6).
  - ∆RER_ji,t−1: 0.00658***, 0.0105***, 0.00649***, 0.00718***, 0.00604***, 0.0106*** (columns 1–6). Interpretation: a real appreciation of the domestic currency increases the number of TTBs against the exporting country when country-time FE are not included.
  - RTA_ji,t−1: 0.566*** (column 5); 0.559*** (column 6). Interpretation: being in the same RTA is associated with more use of alternative trade barriers, perhaps because tariffs cannot be deployed.
  - ∆Tariff_jik,t−1: 0.0153*** (column 5); 0.0137*** (column 6).
  - Overhang_jik,t−1: -0.0131*** (column 5); -0.0121*** (column 6).
  - ExportShare_ji,t−1: 0.0147* (column 5); 0.0232* (column 6). Interpretation: retaliation tends to occur more frequently against countries where the retaliating country exports more.
  - TradeBalance_ji,t−1: -0.00503* (column 5); -0.0248*** (column 6). Interpretation: a reduction in the trade balance is associated with an increase in TTBs, consistent with use of TTBs to restore competitiveness.
- Observations and fit (Table 5):
  - Observations: 3,095,363; 1,829,299; 3,095,363; 3,095,360; 3,090,684; 1,829,292 (columns 1–6).
  - Pseudo-R2: 0.290, 0.318, 0.291, 0.291, 0.300, 0.330 (columns 1–6).
- Comparison: dropping country-time FE shows macro factors (∆RER, TradeBalance, RTA, ExportShare) become statistically significant, implying much of their variation is captured by country-time fixed effects in the baseline.

### Robustness checks (overview leading into 5.2)
- Estimation alternatives:
  - Re-estimation using OLS confirms the statistically significant retaliation effects from the baseline.
  - Including country-pair-time fixed effects (ijt) to account for bilateral political tensions increases the estimated impact of a standard deviation slightly.
  - Using HS6-digit level disaggregation confirms retaliation occurs across different products.
- Dependent-variable construction and timing thresholds:
  - Re-estimating using only AD yields results not statistically different from the baseline.
  - Sensitivity to retaliation time threshold: using the whole sample median (120 days) and each country’s first quartile produces results that confirm the baseline findings.
- Subsample analyses:
  - Splitting the sample before and after the GFC, and between advanced and emerging market economies, reconfirms findings.
  - Evidence suggests retaliation seems stronger in the recent decade, mainly coming from advanced economies.

*Source: IMF Working Paper (section 5.1 Baseline results and associated tables).*

### 5.3    Nonlinearities

### 5.3    Nonlinearities

### Role of trade-policy environment and alternative instruments
- Countries retaliate using multiple available trade-policy instruments; retaliation is not limited to the same instrument used by the foreign partner.
- Membership in the same RTA is associated with stronger retaliation:
  - Interaction *RTA ji,t−1* coefficient in Table 9: 0.668*** (standard error 0.0906).
- Past reductions in tariffs do not mitigate retaliation:
  - Interaction *∆Tariff jik,t−1* in Table 9: 0.00259 (standard error 0.00851) — not statistically significant.
- The presence of unresolved trade disputes does not reduce retaliatory TTBs; instead, disputes are associated with increased retaliation:
  - Interaction *Disputes ji,t* in Table 9: 0.0596 (standard error 0.0367) — not significant.
  - Interaction *Disputes ji,t+1* in Table 9: 0.371** (standard error 0.145).
  - Interaction *Disputes ij,t−1* in Table 9: 0.313*** (standard error 0.0654).
- Current disputes initiated by the target country are positively correlated with increased retaliation from the responding country (Table 6 and Table 9 patterns).
- The prospect of future foreign measures increases retaliation:
  - Column (6) narrative: future number of foreign measures increases retaliation; Table 6 shows positive and significant coefficients on *Disputes ji,t+1* and *TTB ij,t+1* in several specifications.

### Macroeconomic conditions and heterogeneity across countries
- Retaliation intensity varies with the macroeconomic environment:
  - Retaliation is stronger during periods of higher unemployment:
    - From Table 10, interaction *TTB ij,t ∗ Unemployment j,t−1* coefficient: 0.0144*** (standard error 0.00523).
- Large TTB users do not necessarily retaliate the most:
  - Interaction *Other TTB j,t ∗ Large user j* in Table 10: -0.0539 (standard error 0.0537) — not significant.
- However, specific large users show significantly different retaliation intensities:
  - Interaction *Other TTB j,t ∗ USA j* in Table 10: -0.0530** (standard error 0.0248).
  - Interaction *Other TTB j,t ∗ EU N j* in Table 10: 0.666*** (standard error 0.0882).
  - Interaction *Other TTB j,t ∗ CHN j* in Table 10: 1.255*** (standard error 0.137).
- The intensity of retaliation varies across countries both by overall TTB activity and by country identity (United States, European Union, China highlighted).

### Targeting by comparative advantage and sectoral patterns
- Retaliatory TTBs are more likely when the foreign economy first targeted its own comparative advantage sector:
  - From Table 10, *TTB CA j ij,t* coefficient: 0.763*** (standard error 0.0448).
  - From Table 10, *TTB CA i ij,t* coefficient: 1.059*** (standard error 0.178).
- Retaliatory TTBs are not generally confined to a single injured sector but tend to occur in many sectors simultaneously, suggesting use of TTBs to combat perceived general unfair governmental policies (narrative finding).
- Evidence indicates TTBs tend to be used more against the comparative sector of the trading partner (column (3) narrative and Table 10 coefficients above).

### Robustness and additional empirical patterns (selected key coefficients and patterns)
- Baseline and robustness estimations consistently confirm the existence of retaliation across specifications (see Tables 6–10).
- Selected coefficients from Tables (preserving values exactly as reported):
  - Table 6: Effect of one SD increase in TTB ijkt (%) across columns: 0.357, 2.528, 0.491, 0.628, 0.453, 0.390.
  - Table 6: Effect of one SD increase in TTB ijk′ t (%) across columns: 0.7482, 1.241, 1.1538, 0.5052, 2.692, 1.196.
  - Table 8: TTB ij,t coefficient across columns (1)–(6): 0.0789***, 0.0722***, 0.0811***, 0.0793***, 0.0808***, 0.0744*** (standard errors listed in table).
  - Table 9: Baseline TTB ij,t coefficients across columns: 0.0692***, 0.0732***, 0.0572***, 0.0643***, 0.112***, 0.0547*** (standard errors listed in table).
  - Table 10: Baseline TTB ij,t entries by column: (1) -0.00845 (standard error 0.0315), (2) 0.105*** (standard error 0.0216), (3) 0.121** (standard error 0.0526).
- Other consistent findings across tables:
  - *∆Tariff jik,t−1* typically positive and significant (e.g., Table 6 and Table 9 show 0.0111***, 0.0112*** etc.).
  - *Overhang jik,t−1* typically negative and significant (e.g., Table 6 and Table 9 show coefficients around -0.0130*** to -0.0137***).

### Implications and interpretation
- Countries retaliate using all measures available despite WTO dispute mechanisms; filing disputes does not appear to reduce rules-inconsistent retaliation and may correlate with more retaliatory responses.
- Retaliation is counter-cyclical with respect to unemployment and is sensitive to the sectoral comparative-advantage structure of the foreign partner.
- Certain large economies (United States, European Union, China) exhibit significantly different retaliation intensities, implying heterogeneity in strategic responses across major players.

*Source: IMF working paper chapter section 5.3 Nonlinearities (extracted tables and narrative).*

### References

### wpiea2023099-print-pdf - References

### References cited
- Bibliographic entries on trade policy, antidumping, temporary trade barriers, retaliation, and related empirical methods, including:
  - Aiyar, S., J. Chen, C. H. Ebeke, R. Garcia-Saltos, T. Gudmundsson, A. Ilyina, A. Kangur, T. Kunaratskul, S. L. Rodriguez, M. Ruta, T. Schulze, G. Soderberg, and J. P. Trevino (2023). Geo-economic fragmentation and the future of multilateralism. Staff Discussion Notes 2023(001).
  - Bagwell, K. and R. W. Staiger (1990). A theory of managed trade. The American Economic Review 80(4), 779–795.
  - Bown, C. P. (2015). Temporary trade barriers database. World Bank.
  - Bown, C. P. and M. A. Crowley (2013a). Import protection, business cycles, and exchange rates: Evidence from the great recession. Journal of International Economics 90, 50–64.
  - Fajgelbaum, P. D., P. K. Goldberg, P. J. Kennedy, and A. K. Khandelwal (2020). The Return to Protectionism*. The Quarterly Journal of Economics 135(1), 1–55.
  - Grossman, G. M. and E. Helpman (1994). Protection for sale. American Economic Review 84(4), 833–850.
  - Irwin, D. (1998). The smoot-hawley tariff: A quantitative assessment. The Review of Economics and Statistics 80(2), 326–334.
  - Silva, J. M. C. S. and S. Tenreyro (2006). The log of gravity. The Review of Economics and Statistics 88(4), 641–658.
  - Staiger, R. W., F. A. Wolak, R. E. Litan, M. L. Katz, and L. Waverman (1994). Measuring industry-specific protection: Antidumping in the united states. Brookings Papers on Economic Activity. Microeconomics 1994, 51–118.
- Additional references cover topics such as protectionism during crises, the Smoot-Hawley trade war, WTO dispute settlement, antidumping filings, retaliation and safeguards, gravity model estimation, and datasets used (e.g., Temporary trade barriers database).

### Appendix: Additional figures — key notes and captions
- Figure A1: Total number of new investigations per country (panels by country over 1989–2020).
- Figure A2: Total number of new investigations per country (cont.) (panels by country over 1989–2020).
- Figure A3: Average duration of a TTB investigation
  - Kernel density of average duration computed over all countries in the sample.
  - Note: "Average duration of a TTB investigation computed over all countries in our sample. All investigations are considered, including those where no injury was found by the trade authority."
- Figure A4: Number of days before investigation (country-level density panels).
- Figure A5: Average number of TTBs per month
  - Plots the average number of HS6 products targeted by a TTBs in the sample.
  - Note: "Two thirds of them are implemented from June to December."
- Figure A6: Share of retaliatory measures across countries
  - Plots distribution of the average share of retaliation; computation: share of TTB identified as potential retaliation out of the total number of new TTB.
  - Large (small) importers defined as those with average imports above sample median. AE (EM) grouping is based on IMF classification.
- Figure A7: Share of retaliatory measures across partners
  - Similar distributional plot by partner groups (AE partners, EM partners); same computation and group definitions as Figure A6.

### Appendix: Additional tables — key contents
- Table B1: List of countries in the sample
  - Countries listed in the sample include: Argentina, Australia, Brazil, Canada, Chile, China, Colombia, Costa Rica, European Union, India, Indonesia, Israel, Japan, Korea, Mexico, Malaysia, New Zealand, Pakistan, Peru, Philippines, Russia, Thailand, Turkey, United States, Taiwan, South Africa, Venezuela.
- Table B2: Total number of investigations per HS2 sector (columns: HS2, Sector, Nber of investigations, Largest investigator)
  - Selected entries (preserving original numeric values and text exactly as in source):
    - HS2 1 Animals; live — USA3; MEX3
    - HS2 2 Meat and edible meat offal — USA54
    - HS2 4 Edible products of animal origin — CHL350
    - HS2 7 Vegetables and certain roots and tubers; edible — KOR100
    - HS2 11 Products of the milling industry; malt, starches, inulin, wheat gluten — CHL39
    - HS2 17 Sugars and sugar confectionery — CHN50
    - HS2 19 Preparations of cereals, flour, starch or milk; pastrycooks’ products — KOR25
    - HS2 20 Preparations of vegetables, fruit, nuts or other parts of plants — ARG26
    - HS2 24 Tobacco and manufactured tobacco substitutes — AUS1
    - HS2 25 Salt; sulphur; earths, stone; plastering materials, lime and cement — PHL30
    - HS2 28 Inorganic chemicals; compounds of precious or rare metals — IND211
    - HS2 29 Organic chemicals — IND610
    - HS2 30 Pharmaceutical products — CAN5; AUS5
    - HS2 39 Plastics and articles thereof — IND156
    - HS2 40 Rubber and articles thereof — USA76
    - HS2 44 Wood and articles of wood; wood charcoal — USA57
    - HS2 48 Paper and paperboard — USA142
    - HS2 52 Cotton — IDN160
    - HS2 54 Man-made filaments, textile materials — TUR206
    - HS2 64 Footwear; gaiters and the like; parts of such articles — ARG871
    - HS2 72 Iron and steel — USA3565
    - HS2 73 Iron or steel articles — USA484
    - HS2 84 Nuclear reactors, boilers, machinery and mechanical appliances — USA237
    - HS2 85 Electrical machinery and equipment — MEX161
    - HS2 95 Toys, games and sports requisites — BRA241
    - HS2 96 Miscellaneous manufactured articles — TUR27
  - Table B2 continues across HS2 sectors listing the "Nber of investigations" and the "Largest investigator" for each sector as in the source.

*Retaliation through Temporary Trade Barriers — Working Paper No. WP/2023/099*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023099-print-pdf.pdf_