## 5.7    Role of economic policies

## Source details

**Canonical URL:** [5.7    Role of economic policies](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022081-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022081-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022081-print-pdf.pdf.json)

---

### Overview of the shock and data
- In April 2020, the value of France’s total exports declined by 78% and imports declined by 70% in year-on-year terms.
- Analysis uses granular monthly French customs data from January 1995 to December 2020, focusing on January 2019 to December 2020 (2019 as benchmark).
- COVID-19 shock measures: lockdown stringency index and number of COVID-19 deaths (OxCGRT); economic support index (OxCGRT); COVID-19 trade policy measures (WTO).
- Products classified using European CN8 nomenclature harmonized over time.

### Margins of trade adjustment — key findings
- Trade adjustment decomposed into three margins: sub-intensive (continuing firms change value of traded varieties), sub-extensive (continuing firms add/drop varieties), and extensive (firm entry/exit).
- Aggregate patterns (peak shock, first-half 2020):
  - Both exporting and importing firms adjusted primarily along the sub-intensive margin.
  - The extensive margin played a small role overall.
  - Exporting firms exhibited a notable sub-extensive adjustment (stopping exports of some products to some destinations).
  - For imports, the sub-intensive margin accounted for almost 90 percent of the drop in aggregate imports in the second quarter of 2020.
- Recovery: trade flows largely returned toward pre-pandemic levels by the second half of 2020, but by December 2020 both imports and exports were still declining relative to 2019.

### Heterogeneity across partner country groups and China
- Country groups analyzed: Advanced Economies (AE), Emerging Markets (EM), Low-Income and Developing Countries (LIDC), and China separately.
- Exports:
  - Sub-intensive margin dominated for AEs and EMs.
  - Sub-extensive margin played a major role for developing partners: 57 percent of export change in EMs and 89 percent in LIDCs.
  - Exports to China dropped earlier due to pandemic origin there.
- Imports:
  - Sub-intensive margin accounted for at least two-thirds of import variation for all groups except China.
  - Imports from China rebounded swiftly in the first half of 2020.
  - Adjustment in imports from China along the extensive margin explains almost all observed change — many firms began importing from China for the first time in 2020Q2.

### Quantified contributions of margins (selected statistics from 2019Q2 → 2020Q2)
- Exports (total change; share sub-intensive; share sub-extensive; share extensive):
  - AE: Total change -34.56; Share sub-intensive 0.74; Share sub-extensive 0.24; Share extensive 0.02
  - EM excl. China: Total change -39.73; Share sub-intensive 0.43; Share sub-extensive 0.57; Share extensive -0.004
  - China: Total change -27.92; Share sub-intensive 0.21; Share sub-extensive 0.02; Share extensive -0.01
  - LIDC: Total change -34.48; Share sub-intensive 0.01; Share sub-extensive 0.89; Share extensive 0.01
- Imports (total change; share sub-intensive; share sub-extensive; share extensive):
  - AE: Total change -28.97; Share sub-intensive 0.87; Share sub-extensive 0.13; Share extensive 0.005
  - EM excl. China: Total change -41.54; Share sub-intensive 0.80; Share sub-extensive 0.14; Share extensive 0.06
  - China: Total change 21.02; Share sub-intensive -0.32; Share sub-extensive 0.36; Share extensive 0.96
  - LIDC: Total change -39.41; Share sub-intensive 0.65; Share sub-extensive 0.05; Share extensive 0.30

### Transmission channels and demand vs supply
- Lockdowns were the main transmission mechanism: implementing the most stringent lockdowns (example cited: Philippines in April of 2020) is associated with a 19 percentage point drop in exports and almost 8 percentage point drop in imports.
- The shock was short-lived: effect of lockdowns dissipated quickly in the third quarter of 2020.
- Trade value decreases were driven primarily by adjustments in volumes, with slower passthrough to prices (unit values).
- Propagation along GVCs: shock propagated quickly; goods closer to final consumers (downstream) were most impacted while intermediate goods were relatively more insulated — evidence consistent with demand-driven collapse and recovery.

### Firm-level heterogeneity and mitigating factors
- Inventory management:
  - Firms in more inventory-intensive industries weathered the shock better.
  - Most inventory-intensive firms tended to stockpile imported inputs.
- Automation (robotization):
  - Proxy: firm-level imports of industrial robots in recent years.
  - Robotization mitigated the impact of the shock on both exports and imports.
- Firm age and size:
  - Older and medium-sized firms fared better during the pandemic.
  - Younger, smaller, and larger firms’ trade flows were more negatively impacted.
- Transportation mode:
  - Firms relying on air transportation were more negatively affected, implying larger initial disruptions to air transport versus road and maritime modes.

### Role of trade and economic support policies
- COVID-19 trade policies (WTO-coded: import promotion; import ban; export promotion; export ban) — dummy variables indicate a measure implemented on the targeted product by the foreign economy.
- Empirical findings (Table 12):
  - Exports: lockdown stringency = -19.394 ∗∗∗ (All); COVID-19 deaths = -0.171 ∗∗ (All).
  - Imports: lockdown stringency = -7.785 ∗∗ (All); COVID-19 deaths = 0.015 (All).
  - Import promotion: exports coefficient 2.769 (All); import ban: exports coefficient -3.865 (All).
  - Export ban: imports coefficient -5.904 (All).
  - Interpretation: import-promoting and export-banning measures had little impact on aggregate trade with AEs and EMs; measures were particularly effective in LIDCs where both exports to and imports from LIDCs significantly dropped in response to their trade policy changes.
- Economic support policies (OxCGRT economic support index) — included as control (Table 13):
  - Exports: economic support index = 0.003 (All).
  - Imports: economic support index = 0.010 (All).
  - Interpretation: economic support measures do not seem to have had an impact on the economic activity of French firms during 2020; likely explanation is most policy effect already captured by lockdown stringency and deaths.

### Econometric specification and baseline regression results
- Dependent variable: mid-point year-on-year growth rate of trade flows between firm i and country c of product p in month t:
  - y_ipct = (flow_ipct − flow_ipc(t−12)) / (1/2 (flow_ipct + flow_ipc(t−12))).
- Main regressors: Stringency_ct (monthly average lockdown stringency index), Deaths_ct (monthly change in COVID-19 deaths per 1,000 people).
- Controls and fixed effects: α_i (firm fixed effects), γ_pc (product-country fixed effects), τ_t (month-year fixed effects). Robust standard errors clustered by country and product.
- Baseline coefficients (Table 3):
  - Aggregate exports: lockdown stringency = -19.391 (4.161)***; COVID-19 deaths = -0.170 (0.082)**.
    - One standard deviation increase in monthly lockdown stringency (about .25 on a scale from 0 to 1) → about 4.8 percentage point drop in aggregate exports in 2020.
    - Implementing the most stringent lockdowns implies a drop in exports of 19 percentage points (when controlling for deaths).
    - One standard deviation increase in monthly new deaths (about 7 in the sample) → associated with 1.2 percentage point decline in exports.
  - Aggregate imports: lockdown stringency = -7.843 (3.559)**.
    - One standard deviation increase in partner lockdown stringency → about 1.9 percentage point decline in French imports; most stringent lockdown implies almost 8 percentage point drop in imports.
- Dynamics:
  - Three-month rolling-window estimates show the effect of lockdowns dissipates quickly for both exports and imports; initial impact larger for exports.

### Unit values (prices vs. quantities)
- Dependent variable: mid-point yoy growth in unit values (values divided by units for continuing products).
- Table 4 coefficients:
  - Export unit values: lockdown stringency = -0.800 (0.635) (ns).
  - Import unit values: lockdown stringency = 0.581 (0.339), * p<0.1.
  - COVID-19 deaths: exports -0.035 (0.011)***; imports -0.025 (0.008)***.
- Finding: Aggregate impact of COVID-19 shock on average unit values in 2020 was limited; trade declines driven mainly by volumes.

### COVID-19 shock propagation via supply chains (GVCs and upstreamness)
- Sample restricted to firms that both export and import.
- Exports regressed on lagged imports (Table 5 top): lagged import growth coefficient = 6.989 (0.321)***.
  - A 10 percentage point drop in imports in the previous period → almost 0.7 percentage point drop in current exports.
- Imports regressed on lead export growth (Table 5 bottom): lead export growth coefficient = 2.647 (0.134)***.
  - A 10 percentage point drop in future exports → about 0.26 percentage point decrease in current imports.
- Upstreamness interaction (Table 6):
  - Range about 1.4 to 3.1, average about 2.
  - Exports: lockdown stringency = -49.594 (7.451)***; lockdown stringency × upstreamness = 15.018 (2.729)***.
  - Imports: lockdown stringency = -35.106 (7.098)***; lockdown stringency × upstreamness = 13.636 (2.633)***.
  - Interpretation: more upstream products were more insulated from lockdown-driven declines; intermediate and capital goods relatively more insulated than consumer goods.

### Heterogeneous effects: inventory intensity and automation
- Inventory intensity (Table 7):
  - Exports: lockdown stringency × inventory intensity = 22.428 (All); 17.896 (AE); 35.891 (EM); 28.505 (LIDC).
  - Imports: lockdown stringency × inventory intensity = 33.486 (All); 36.110 (AE); 33.852 (EM); 33.814 (LIDC).
  - Interpretation: firms in more inventory-intensive industries were more insulated from lockdowns; most inventory-intensive firms tended to stockpile imported inputs.
  - Robustness: median-based inventory measure results were not robust to other inventory measures.
- Automation (Table 8):
  - Robot indicator: binary = 1 if firm imported industrial robots between 2017 and 2020.
  - Share using robots: about 7 percent of firms exporting to AEs; around 11 percent exporting to EMs and LIDCs. On imports: 8 percent for firms importing from AEs; close to 18 percent importing from LIDCs.
  - Exports: lockdown stringency × robots = 9.541 (All); 11.961 (AE); 8.713 (EM); -14.444 (LIDC).
  - Imports: lockdown stringency × robots = 11.757 (All); 10.034 (AE); 16.864 (EM); 17.056 (LIDC).
  - Interpretation: robotization mitigated the impact of lockdowns on trade for many country groups; effect mixed for LIDCs on exports.

### Key numerical highlights
- One standard deviation in monthly lockdown stringency ≈ .25 (scale 0–1).
- Export effect (aggregate): one std dev lockdown → about 4.8 percentage point drop; most stringent lockdown → 19 percentage points drop.
- Deaths effect: one std dev monthly new deaths ≈ 7 → associated with 1.2 percentage point decline in exports.
- Selected coefficients:
  - exports lockdown stringency = -19.391 (4.161)***; exports COVID-19 deaths = -0.170 (0.082)**;
  - imports lockdown stringency = -7.843 (3.559)**;
  - lagged import growth → exports = 6.989 (0.321)***;
  - lead export growth → imports = 2.647 (0.134)***;
  - lockdown stringency × upstreamness (exports) = 15.018 (2.729)***; (imports) = 13.636 (2.633)***.

### Concluding remarks and policy-relevant implications
- Aggregate findings:
  - Lockdown stringency of trading partners had a larger impact on French firm-level performance than number of deaths, except for LIDCs.
  - Firms adjusted primarily along the intensive margin; firms trading with LIDCs were more prone to stop exchanging with these countries.
  - The shock propagated along supply chains.
- Heterogeneity in firm responses:
  - Amplifying factors of the shock: air transportation for exports and road transportation for imports.
  - Firms better able to cope: medium-size and older firms; firms trading more upstream goods; firms using robots in production; firms holding higher inventories.
- Role of policies:
  - No evidence that economic support policies and trade policies played a significant role in mitigating the shock for French firms in 2020, with the notable exception of trade measures affecting LIDCs.
- Implication on globalization:
  - Results support the view that deglobalization of trade following the COVID-19 shock is unlikely.
- Future work:
  - The analysis focuses on the first year of the pandemic; authors intend to extend the analysis using 2021 data to capture subsequent shocks (rise in shipping costs, repeated waves of COVID-19 variants and lockdowns).

*Source: wpiea2022081-print-pdf (section 5.7) — IMF working paper content excerpt.*

### 5.1    Baseline  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .

### 5.1    Baseline

### 5.2    Unit values
- Page: 14

### 5.3    COVID-19 shock propagation via supply chains
- Page: 14

### 5.4    Heterogeneous effects:  inventory intensity and automation
- Page: 17

### 5.5    Heterogeneous effects:  firm characteristics
- Page: 19

### 5.6    Heterogeneous effects:  transportation mode
- Page: 22

*Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022081-print-pdf.pdf*

### 5.7    Role of economic policies .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  

### 5.7    Role of economic policies

### Overview of the shock and data
- In April 2020, the value of France’s total exports declined by 78% and imports declined by 70% in year-on-year terms.
- Analysis uses granular monthly French customs data from January 1995 to December 2020, focusing on January 2019 to December 2020 (2019 as benchmark).
- COVID-19 shock measures: lockdown stringency index and number of COVID-19 deaths (OxCGRT); economic support index (OxCGRT); COVID-19 trade policy measures (WTO).
- Products classified using European CN8 nomenclature harmonized over time.

### Margins of trade adjustment — key findings
- Trade adjustment decomposed into three margins: sub-intensive (continuing firms change value of traded varieties), sub-extensive (continuing firms add/drop varieties), and extensive (firm entry/exit).
- Aggregate patterns (peak shock, first-half 2020):
  - Both exporting and importing firms adjusted primarily along the sub-intensive margin.
  - The extensive margin played a small role overall.
  - Exporting firms exhibited a notable sub-extensive adjustment (stopping exports of some products to some destinations).
  - For imports, the sub-intensive margin accounted for almost 90 percent of the drop in aggregate imports in the second quarter of 2020.
- Recovery: trade flows largely returned toward pre-pandemic levels by the second half of 2020, but by December 2020 both imports and exports were still declining relative to 2019.

### Heterogeneity across partner country groups and China
- Country groups analyzed: Advanced Economies (AE), Emerging Markets (EM), Low-Income and Developing Countries (LIDC), and China separately.
- Exports:
  - Sub-intensive margin dominated for AEs and EMs.
  - Sub-extensive margin played a major role for developing partners: 57 percent of export change in EMs and 89 percent in LIDCs (Table 2).
  - Exports to China dropped earlier due to pandemic origin there.
- Imports:
  - Sub-intensive margin accounted for at least two-thirds of import variation for all groups except China.
  - Imports from China displayed a distinct pattern: Chinese imports rebounded swiftly in the first half of 2020.
  - Adjustment in imports from China along the extensive margin explains almost all observed change — many firms began importing from China for the first time in 2020Q2 (Table 2).

### Quantified contributions of margins (selected statistics from 2019Q2 → 2020Q2)
- Exports total percent changes (percent total) and margin shares (Table 2):
  - AE: Total change -34.56; Share sub-intensive 0.74; Share sub-extensive 0.24; Share extensive 0.02
  - EM excl. China: Total change -39.73; Share sub-intensive 0.43; Share sub-extensive 0.57; Share extensive -0.004
  - China: Total change -27.92; Share sub-intensive 0.21; Share sub-extensive 0.02; Share extensive -0.01
  - LIDC: Total change -34.48; Share sub-intensive 0.01; Share sub-extensive 0.89; Share extensive 0.01
- Imports total percent changes and margin shares (Table 2):
  - AE: Total change -28.97; Share sub-intensive 0.87; Share sub-extensive 0.13; Share extensive 0.005
  - EM excl. China: Total change -41.54; Share sub-intensive 0.80; Share sub-extensive 0.14; Share extensive 0.06
  - China: Total change 21.02; Share sub-intensive -0.32; Share sub-extensive 0.36; Share extensive 0.96
  - LIDC: Total change -39.41; Share sub-intensive 0.65; Share sub-extensive 0.05; Share extensive 0.30

### Transmission channels and demand vs supply
- Lockdowns were the main transmission mechanism: implementing the most stringent lockdowns (example cited: Philippines in April of 2020) is associated with a 19 percentage point drop in exports and almost 8 percentage point drop in imports.
- The shock was short-lived: effect of lockdowns dissipated quickly in the third quarter of 2020.
- Trade value decreases were driven primarily by adjustments in volumes, with slower passthrough to prices (unit values).
- Propagation along GVCs: shock propagated quickly; goods closer to final consumers (downstream) were most impacted while intermediate goods were relatively more insulated — evidence consistent with demand-driven collapse and recovery.

### Firm-level heterogeneity and mitigating factors
- Inventory management:
  - Firms in more inventory-intensive industries weathered the shock better.
  - Most inventory-intensive firms tended to stockpile imported inputs.
- Automation (robotization):
  - Proxy: firm-level imports of industrial robots in recent years.
  - Robotization mitigated the impact of the shock on both exports and imports.
- Firm age and size:
  - Older and medium-sized firms fared better during the pandemic.
  - Younger, smaller, and larger firms’ trade flows were more negatively impacted.
- Transportation mode:
  - Firms relying on air transportation were more negatively affected, implying larger initial disruptions to air transport versus road and maritime modes.

### Role of policies
- COVID-19 trade policies (export restrictions and import facilitation):
  - Overall impact limited.
  - Exception: trade with low-income countries, where import and export bans dampened trade.
- Economic support policies (cash transfers to households, stimulus spending):
  - No significant impact detected on demand for French exports nor on supply of imports to France.

### Interpretation and implications
- The bulk of trade adjustment occurred along the intensive margin, indicating a transitory shock rather than permanent deglobalization.
- Findings align with theoretical and empirical literature suggesting GVCs transmitted shocks but also provided resilience — absence of comprehensive and permanent realignment of supply chains is likely given intensive-margin adjustment patterns.
- Pandemic likely to accelerate trends in automation; inventory management practices played a notable role in firms’ resilience.
- Heterogeneity across partner countries and firm characteristics implies targeted policy considerations where low-income partner trade and air-transport–dependent firms may need more attention.

*Italic: Source: wpiea2022081-print-pdf (section 5.7) — IMF working paper content excerpt.*

### 2020.  This is consistent with the earlier onset of the virus spread in China and subsequent recovery, while

### wpiea2022081-print-pdf - 2020.  This is consistent with the earlier onset of the virus spread in China and subsequent recovery, while

### Econometric specification
- Dependent variable: mid-point year-on-year growth rate of trade flows between firm i and country c of product p in month t, defined as
  - y_ipct = (flow_ipct − flow_ipc(t−12)) / (1/2 (flow_ipct + flow_ipc(t−12))).
- Main regressors:
  - Stringency_ct: monthly average of the country-specific daily lockdown stringency index (Oxford COVID-19 Government Response Tracker).
  - Deaths_ct: monthly change in the number of COVID-19 deaths per 1,000 people.
- Fixed effects and controls:
  - α_i: firm fixed effects.
  - γ_pc: product-country fixed effects.
  - τ_t: month-year fixed effects.
  - Robust standard errors clustered by country and product.
- Heterogeneity:
  - Interactions of Stringency_ct and Deaths_ct with time-invariant firm, product, and country characteristics (e.g., upstreamness).

### Baseline results (exports and imports)
- Aggregate exports:
  - Table 3 coefficient: lockdown stringency = -19.391 (standard error 4.161), *** p<0.01.
  - Interpretation: one standard deviation increase in monthly lockdown stringency (about .25 on a scale from 0 to 1) is associated with about 4.8 percentage point drop in aggregate exports in 2020.
  - Implementing the most stringent lockdowns implies a drop in exports of 19 percentage points (when controlling for deaths).
  - COVID-19 deaths coefficient (aggregate exports): -0.170 (standard error 0.082), ** p<0.05.
    - One standard deviation increase in monthly new deaths (about 7 in the sample) is associated with 1.2 percentage point decline in exports.
  - Heterogeneity by country group:
    - AEs: lockdown stringency = -13.423 (5.441), **.
    - EMs: lockdown stringency = -24.047 (8.565), ***.
    - LIDCs: lockdown stringency = -18.624 (12.483) (not statistically significant).
    - COVID-19 deaths: AEs -0.184 (0.075) **; EMs 0.104 (0.213) (ns); LIDCs -17.659 (9.630) * (note large standard error).
- Aggregate imports:
  - Table 3 coefficient: lockdown stringency = -7.843 (3.559), ** p<0.05.
  - One standard deviation increase in partner lockdown stringency is associated with about 1.9 percentage point decline in French imports; most stringent lockdown implies almost 8 percentage point drop in imports.
  - Heterogeneity by country group:
    - AEs: lockdown stringency = -8.916 (6.193) (ns).
    - EMs: lockdown stringency = -14.203 (3.933), ***.
    - LIDCs: lockdown stringency = 7.455 (7.435) (ns).
  - COVID-19 deaths on imports: aggregate -0.015 (0.037) (ns); AEs -0.078 (0.032) **; EMs 0.120 (0.104) (ns); LIDCs -6.727 (3.550) *.
- Dynamics:
  - Three-month rolling-window estimates (Figure 5) show the effect of lockdowns dissipates quickly for both exports and imports.
  - Initial impact larger for exports, leading to a sharper recovery path; toward end-2020, import stringency coefficients turn positive with wide confidence intervals.
- Interpretation:
  - Lockdown policies were the primary driver of the demand shock; COVID-19 deaths had a smaller and less consistent effect.
  - Differences between exports and imports suggest demand shocks in partner countries dominated supply shocks, subject to the identifying assumptions described in the text.

### Unit values (prices vs. quantities)
- Dependent variable: mid-point yoy growth in unit values (values divided by units for continuing products).
- Table 4 coefficients:
  - Export unit values: lockdown stringency = -0.800 (0.635) (ns).
  - Import unit values: lockdown stringency = 0.581 (0.339), * p<0.1.
  - COVID-19 deaths: exports -0.035 (0.011) ***, imports -0.025 (0.008) ***.
- Findings:
  - Aggregate impact of COVID-19 shock on average unit values in 2020 was limited.
  - Small negative effect of rising COVID-19 deaths on unit values; lockdown stringency effect on unit values is economically small and not statistically significant for exports.

### COVID-19 shock propagation via supply chains (GVCs and upstreamness)
- Sample restricted to firms that both export and import (integrated in GVCs).
- Exports regressed on lagged imports:
  - Table 5 top panel: lagged import growth coefficient = 6.989 (0.321), ***.
    - A 10 percentage point drop in imports in the previous period is associated with almost 0.7 percentage point drop in current exports.
    - Stronger for firms trading with AEs and EMs; more than 50 percent lower for firms exporting to LIDCs.
- Imports regressed on lead export growth:
  - Table 5 bottom panel: lead export growth coefficient = 2.647 (0.134), ***.
    - A 10 percentage point drop in future exports is associated with about 0.26 percentage point decrease in current imports.
  - Interpretation: firms adjust imports of intermediates in anticipation of downstream demand; shocks propagate upstream and downstream along supply chains.
- Upstreamness interaction (industry-level upstreamness, range about 1.4 to 3.1, average about 2):
  - Exports (Table 6):
    - lockdown stringency = -49.594 (7.451), ***.
    - lockdown stringency × upstreamness = 15.018 (2.729), ***.
    - COVID-19 deaths = -0.817 (0.490), *.
    - COVID-19 deaths × upstreamness = 0.327 (0.215) (ns).
    - Interpretation: more upstream products were more insulated from lockdown-driven declines in exports; positive interaction indicates attenuation with higher upstreamness.
  - Imports (Table 6):
    - lockdown stringency = -35.106 (7.098), ***.
    - lockdown stringency × upstreamness = 13.636 (2.633), ***.
    - COVID-19 deaths = -0.029 (0.292) (ns).
    - COVID-19 deaths × upstreamness = 0.004 (0.136) (ns).
    - Interpretation: similar insulation for imports—intermediate and capital goods were relatively more insulated than consumer goods.
  - Heterogeneity:
    - Interaction effects are generally consistent across AEs and EMs; less clear for LIDCs (lockdown stringency × upstreamness for exports to LIDCs not statistically significant).
    - COVID-19 deaths show stronger differential impacts by upstreamness for developing countries in some specifications (e.g., imports and EM/LIDC subsamples).

### Key numerical highlights
- One standard deviation in monthly lockdown stringency ≈ .25 (scale 0–1).
- Export effect (aggregate): one std dev lockdown → about 4.8 percentage point drop; most stringent lockdown → 19 percentage points drop.
- Deaths effect: one std dev monthly new deaths ≈ 7 → associated with 1.2 percentage point decline in exports.
- Table 3 coefficients (selected): exports lockdown stringency = -19.391 (4.161)***; exports COVID-19 deaths = -0.170 (0.082)**; imports lockdown stringency = -7.843 (3.559)**.
- Table 5: lagged import growth → exports = 6.989 (0.321)***; lead export growth → imports = 2.647 (0.134)***.
- Table 6: lockdown stringency × upstreamness (exports) = 15.018 (2.729)***; (imports) = 13.636 (2.633)***.

*Source: wpiea2022081-print-pdf - 2020.  This is consistent with the earlier onset of the virus spread in China and subsequent recovery, while*

### 5.4    Heterogeneous effects:  inventory intensity and automation

### 5.4    Heterogeneous effects:  inventory intensity and automation

### Inventory intensity — empirical findings
- Dependent variable: mid-point yoy growth in trade flows.  Robust standard errors clustered by country and product.  All regressions include firm, country-year, and month fixed effects.
- Key coefficients from Table 7 (Exports, top panel):
  - lockdown stringency: -24.898 (All); -17.711 (AE); -33.141 (EM); -25.446 (LIDC)
  - lockdown stringency × inventory intensity: 22.428 (All); 17.896 (AE); 35.891 (EM); 28.505 (LIDC)
  - COVID-19 deaths: -0.247 (All); -0.276 (AE); 0.021 (EM); -20.299 (LIDC)
  - COVID-19 deaths × inventory intensity: 0.218 (All); 0.272 (AE); 0.220 (EM); 7.087 (LIDC)
  - Observations (exports): 132896 (All); 129860 (AE); 46129 (EM); 78734? (LIDC) [as printed: 1328961298604612907873461479 — observations row formatting in source]
- Key coefficients from Table 7 (Imports, lower panel):
  - lockdown stringency: -15.980 (All); -0.145 (AE); 2.629 (EM); 44.139 (LIDC)
  - lockdown stringency × inventory intensity: 33.486 (All); 36.110 (AE); 33.852 (EM); 33.814 (LIDC)
  - COVID-19 deaths: -0.063 (All); -0.115 (AE); 0.154 (EM); -14.813 (LIDC)
  - COVID-19 deaths × inventory intensity: 0.213 (All); 0.168 (AE); -0.131 (EM); 35.860 (LIDC)
  - Observations (imports): 168863 (All); 95116 (AE); 46500 (EM); 471952? (LIDC) [as printed: 16886395116465004719524491037 — observations row formatting in source]

- Interpretation and substantive findings:
  - Using industry-level medians of inventories-to-sales ratio (median industry ratio around 0.83 for the most inventory-intensive industries), exports and imports of firms in more inventory-intensive industries were more insulated from the pandemic shock.
  - The interaction coefficients (lockdown stringency × inventory intensity) are sizable and positive for both exports and imports, suggesting inventories mitigated the impact of lockdowns.
  - Firms in the most inventory-intensive industries increased their imports during the pandemic (consistent with stockpiling behavior), but the evidence should be interpreted with caution: other literature shows inventory intensity can amplify trade collapses in crises.
- Robustness note from source:
  - Other measures of inventory intensity were tested (industry average ratio, aggregate inventories-to-sales), and results based on the median measure were not robust to other measures.

### Automation (robotization) — empirical findings
- Construction of robot indicator:
  - Binary indicator: firm imported industrial robots between 2017 and 2020 → value one; otherwise zero.
  - Share using industrial robots: about 7 percent of firms exporting to AEs; around 11 percent exporting to EMs and LIDCs. On the import side: 8 percent for firms importing from AEs; close to 18 percent importing from LIDCs.
- Key coefficients from Table 8 (Exports, top panel):
  - lockdown stringency: -23.608 (All); -18.449 (AE); -27.988 (EM); -12.765 (LIDC)
  - lockdown stringency × robots: 9.541 (All); 11.961 (AE); 8.713 (EM); -14.444 (LIDC)
  - COVID-19 deaths: -0.192 (All); -0.206 (AE); 0.127 (EM); -22.208 (LIDC)
  - COVID-19 deaths × robots: 0.061 (All); 0.071 (AE); -0.070 (EM); 10.355 (LIDC)
  - Observations (exports): 134125 (All); 58996 (AE); 88792 (EM); 187184? (LIDC) [as printed: 1341255899688792918718463977]
- Key coefficients from Table 8 (Imports, lower panel):
  - lockdown stringency: -12.584 (All); -13.231 (AE); -20.321 (EM); 3.401 (LIDC)
  - lockdown stringency × robots: 11.757 (All); 10.034 (AE); 16.864 (EM); 17.056 (LIDC)
  - COVID-19 deaths: -0.042 (All); -0.116 (AE); 0.189 (EM); -4.773 (LIDC)
  - COVID-19 deaths × robots: 0.044 (All); 0.073 (AE); -0.208 (EM); -8.305 (LIDC)
  - Observations (imports): 169756 (All); 34117 (AE); 130204 (EM); 738066? (LIDC) [as printed: 16975634117130204738066494784]

- Interpretation and substantive findings:
  - Interaction coefficients between lockdown stringency and the robots indicator are positive for most country groups (exports to AEs and EMs) — suggesting firms using industrial robots were less impacted by lockdowns for exports to AEs and EMs. Exception: negative coefficient for exports to LIDCs.
  - On imports, automation is associated with higher resilience: lockdown stringency × robots positive across country groups, indicating firms using industrial robots experienced smaller declines in imports.
  - Evidence on COVID-19 deaths × robots is mixed, with negative coefficients for imports from EMs and LIDCs, but on net imports of firms using robots declined significantly less.

*Source: IMF Working Paper chapter section 5.4 (figures and tables cited in the source).*

### 5.7    Role of economic policies

### 5.7    Role of economic policies

### Trade policy measures and data
- WTO recorded an increasing number of trade policy measures as early as January 2020, with a number of targeted products peaking in March 2020.
- Many countries lowered trade barriers on medical supplies (e.g., surgical masks), and on agricultural and food products.
- Trade policy measures collected and classified for the analysis: (i) import promotion; (ii) import ban; (iii) export promotion; (iv) export ban.
- Dummy variables indicate a measure implemented on the targeted product by the foreign economy.
- Notes on econometrics (from tables): Dependent variable is a mid-point yoy growth in trade flows. Robust standard errors clustered by country and product are in parentheses. All regressions include firm, country-year, and month fixed effects. *** p<0.01, ** p<0.05, * p<0.1.

### Empirical findings from Table 12 (COVID-19 trade policies)
- Exports regressions (coefficients reported as in table):
  - lockdown stringency: -19.394 ∗∗∗ (All); -13.408 ∗∗ (AE); -24.042 ∗∗∗ (EM); -18.413 (LIDC)  
    (standard errors: (4.160)(5.448)(8.573)(12.469))
  - COVID-19 deaths: -0.171 ∗∗ (All); -0.184 ∗∗ (AE); 0.103 (EM); -17.423 ∗ (LIDC)  
    (standard errors: (0.082)(0.075)(0.216)(9.609))
  - import promotion: 2.769 (All); 3.767 ∗ (AE); 0.665 (EM); -11.549 ∗∗ (LIDC)  
    (standard errors: (2.302)(2.133)(2.907)(5.522))
  - import ban: -3.865 (All); 0.000 (AE); 1.604 (EM); -38.240 ∗∗∗ (LIDC)  
    (standard errors: (3.111)(.)(1.193)(9.392))
  - Observations: 134125 (All); 58996 (AE); 88792 (EM); 91871 (LIDC); 8463977 (note: table lists a long sequence under Observations)
- Imports regressions:
  - lockdown stringency: -7.785 ∗∗ (All); -8.811 (AE); -14.221 ∗∗∗ (EM); 7.491 (LIDC)  
    (standard errors: (3.534)(6.155)(3.906)(7.434))
  - COVID-19 deaths: 0.015 (All); -0.078 ∗∗ (AE); 0.121 (EM); -6.713 ∗ (LIDC)  
    (standard errors: (0.037)(0.032)(0.104)(3.561))
  - export promotion: -2.961 (All); -4.151 (AE); 4.777 (EM)  
    (standard errors: (2.418)(2.479)(3.461))
  - export ban: -5.904 (All); -0.712 (AE); -7.277 (EM); -17.951 ∗ (LIDC)  
    (standard errors: (4.504)(5.633)(4.986)(9.739))
  - Observations: 169756 (All); 34117 (AE); 302047 (EM); 380664 (LIDC); 94784 (note: table lists a long sequence under Observations)
- Interpretation summarized in text:
  - Import-promoting and export-banning measures had little impact on aggregate trade when considering advanced and emerging trade partners.
  - These measures were particularly effective in LIDCs: both exports to and imports from LIDCs significantly dropped in response to their trade policy changes.
  - Import-banning and export-promoting measures were rarely used.

### Economic support policies and data
- Economic support policies sourced from the Oxford COVID-19 Government Response Tracker (OxCGRT); index includes policies such as income support to households and debt relief measures.
- Economic support index included as a control variable in regressions.

### Empirical findings from Table 13 (Economic Support Policies)
- Exports regressions (coefficients reported as in table):
  - lockdown stringency: -19.424 ∗∗∗ (All); -13.430 ∗∗ (AE); -23.278 ∗∗∗ (EM); -18.422 (LIDC)  
    (standard errors: (4.203)(5.643)(8.107)(13.372))
  - COVID-19 deaths: -0.170 ∗∗ (All); -0.184 ∗∗ (AE); 0.088 (EM); -17.832 ∗ (LIDC)  
    (standard errors: (0.082)(0.074)(0.208)(9.618))
  - economic support index: 0.003 (All); 0.001 (AE); -0.031 (EM); 0.010 (LIDC)  
    (standard errors: (0.028)(0.045)(0.032)(0.062))
  - Observations: 134072 (All); 83996 (AE); 88792 (EM); 91871 (LIDC); 8459685 (note: table lists a long sequence under Observations)
- Imports regressions:
  - lockdown stringency: -8.119 ∗∗ (All); -8.937 (AE); -13.279 ∗∗∗ (EM); 6.447 (LIDC)  
    (standard errors: (3.901)(6.357)(4.356)(7.622))
  - COVID-19 deaths: -0.013 (All); -0.077 ∗∗ (AE); 0.094 (EM); -6.873 ∗ (LIDC)  
    (standard errors: (0.037)(0.031)(0.117)(3.972))
  - economic support index: 0.010 (All); 0.003 (AE); -0.020 (EM); 0.079 (LIDC)  
    (standard errors: (0.023)(0.047)(0.021)(0.073))
  - Observations: 169717 (All); 97117 (AE); 302047 (EM); 380664 (LIDC); 94608 (note: table lists a long sequence under Observations)
- Interpretation summarized in text:
  - Economic support measures do not seem to have had an impact on the economic activity of French firms during 2020.
  - Likely explanation: most of the effect of governmental policies is already captured by the lockdown stringency index and the number of deaths.

### Concluding remarks and policy-relevant implications
- Aggregate findings:
  - Lockdown stringency of trading partners had a larger impact on French firm-level performance than number of deaths, except for LIDCs.
  - Firms adjusted primarily along the intensive margin; firms trading with LIDCs were more prone to stop exchanging with these countries.
  - The shock propagated along supply chains.
- Heterogeneity in firm responses:
  - Amplifying factors of the shock: air transportation for exports and road transportation for imports.
  - Firm characteristics associated with better coping: medium-size and older firms; firms trading more upstream goods; firms using robots in production; firms holding higher inventories.
- Role of policies:
  - No evidence that economic support policies and trade policies played a significant role in mitigating the shock for French firms in 2020.
- Implication on globalization:
  - Results support the view that deglobalization of trade following the COVID-19 shock is unlikely.
- Future work:
  - The paper focuses on trade adjustment in the first year of the pandemic. Several pandemic-related shocks in 2021 (rise in shipping costs, repeated waves of COVID-19 variants and lockdowns) posed roadblocks to trade recovery. The authors intend to extend the analysis using 2021 data to capture these developments.

*Trade and the COVID-19 Pandemic: Lessons from French Firms Working Paper No. WP/2022/081*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022081-print-pdf.pdf_