## wpiea2026070-source-pdf

## Source details

**Canonical URL:** [wpiea2026070-source-pdf](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026070-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026070-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026070-source-pdf.pdf.json)

---

### Key empirical findings on firm-level stock returns
- A firm with high exposure on both trade-sensitivity and uncertainty-sensitivity experienced returns that were 1.05 percentage points higher (in magnitude) than those of a firm with low exposure on both margins around the initial escalation.
- After trade agreements and declining expected tariffs and trade policy uncertainty, comparing firms at the 90th and 10th percentiles:
  - Higher trade sensitivity → 0.33 percentage point additional increase in cumulative returns.
  - Higher uncertainty sensitivity → 0.34 percentage point additional increase in cumulative returns.
  - A firm with high exposure along both dimensions enjoys an additional cumulative return of about 0.67 percentage points from trade deals.
  - The 0.67 percentage point gain is close to 10 percent of the three-day gain of 8.5 percent around the deal dates for tradable firms over the same range.
- Trade and uncertainty measures are proxies constructed from historical data and are subject to measurement error; estimated effects are likely conservative and should be interpreted as lower-bound estimates.

### April 2, April 9, and Trade-Deal event results
- April 2 announcement (universal tariff shock) — estimated coefficients (Table 3, April 2):
  - TradeSensitivity_i: -0.00815 (col (1)), -0.00764 (col (2)), -0.00789 (col (3)), -0.00789 (col (4)); significance retained in industry-clustered SE (col (5)).
  - UncertaintySensitivity_j: -0.0552 (col (1)), -0.0504 (col (2)), -0.0542 (col (3)), -0.0542 (col (4)); remains significant under clustering.
  - Interdecile magnitudes:
    - Difference 90th–10th percentile TradeSensitivity_i = 1.383 → additional 1.09 percentage points decline in three-day cumulative returns.
    - Difference 90th–10th percentile UncertaintySensitivity_j = 0.186 → additional 1.01 percentage points decline.
- April 9 announcement (90-day pause excluding China) — estimated coefficients (Table 3, April 9):
  - TradeSensitivity_i: 0.00410 (col (1)), 0.00380 (col (2)), 0.00374 (col (3)), 0.00374 (col (4)); significance retained.
  - UncertaintySensitivity_j: 0.0319 (col (1)), 0.0295 (col (2)), 0.0283 (col (3)), 0.0283 (col (4)); loses significance under industry clustering in col (5).
  - Interdecile magnitudes:
    - 90th–10th percentile trade sensitivity → additional 0.52 percentage point return.
    - 90th–10th percentile uncertainty sensitivity → additional 0.53 percentage point return.
- Trade deals — estimated coefficients (Table 3, Trade Deals):
  - TradeSensitivity_i: 0.00256 (col (1)), 0.00244 (col (2)), 0.00237 (col (3)), 0.00237 (col (4)); significance attenuates under clustering (col (5)).
  - UncertaintySensitivity_j: 0.0180 (col (1)), 0.0169 (col (2)), 0.0180 (col (3)), 0.0180 (col (4)); statistically significant.
  - Interdecile magnitudes:
    - 90th–10th percentile trade sensitivity → 0.33 percentage point excess return.
    - 90th–10th percentile uncertainty sensitivity → 0.34 percentage point relative gain.
  - Asymmetry: trade-deal recovery is roughly one-third the size of the initial April 2 drop, indicating incomplete reversal and continuing residual risk premium for global fragmentation.

### Robustness and alternative specifications
- Four-day window ([−1, +2]):
  - April 2 four-day estimates strengthen:
    - TradeSensitivity_i: -0.0166 (col (4)/(5) consistent) → top 10th percentile incurred additional 2.30 percentage point decline over four days relative to bottom 10th percentile.
    - UncertaintySensitivity_j: -0.0715 (col (4)/(5)) → additional 1.33 percentage point decline.
  - April 9 four-day: 90th percentile trade sensitivity → additional 0.47 percentage point increase.
  - Trade deals four-day: trade = 0.12 percentage point; uncertainty = 0.36 percentage point for 90th percentile.
- One percent winsorization (robustness to tails; Table 5, col (5)):
  - April 2:
    - TradeSensitivity_i: -0.00253 (attenuated from baseline -0.00789) but statistically significant at 1 percent.
    - UncertaintySensitivity_j: -0.0570 (nearly identical to baseline -0.0542).
  - April 9:
    - TradeSensitivity_i: 0.00119.
    - UncertaintySensitivity_j: 0.0235.
  - Trade Deals:
    - TradeSensitivity_i: 0.000837 (loses significance).
    - UncertaintySensitivity_j: 0.0192 (remains significant).
  - Interpretation: trade sensitivity is susceptible to tail estimation noise; uncertainty sensitivity is stable across winsorization.
- All tradable firms (imputed sample; 16,267 firms, 142 SIC3 industries):
  - April 2:
    - TradeSensitivity_i: -0.00506 (col (1)), -0.00470 (col (2)), -0.00504 (col (3)/(4)) — significant.
    - UncertaintySensitivity_j: -0.0516 (col (1)), -0.0489 (col (2)), -0.0557 (col (3)/(4)) — significant.
    - Interdecile: trade = 1.807 → additional 0.91 percentage point decline; uncertainty = 0.186 → additional 1.04 percent decline.
  - April 9:
    - TradeSensitivity_i: 0.00834 (col (1)), 0.00799 (col (2)), 0.00762 (col (3)/(4)) — robust.
    - UncertaintySensitivity_j: 0.0512 (col (1)), 0.0467 (col (2)), 0.0469 (col (3)/(4)) — weakens in some specs.
  - Trade Deals: results lose statistical significance (attenuation consistent with imputation measurement error).
- Abnormal returns (CAR, Market Model; Table 7, three-day window [−1, +1]):
  - April 2:
    - TradeSensitivity_i: -0.00316 (col (1)), -0.00289 (col (2)), -0.00317 (col (3)/(4)/(5)) — significant.
    - UncertaintySensitivity_j: -0.0289 (col (1)), -0.0271 (col (2)), -0.0289 (col (3)/(4)/(5)) — negative and significant, attenuated vs raw returns.
    - Economic magnitude: 90th percentile trade sensitivity → additional 0.44 percentage point decline in CAR relative to 10th percentile.
  - April 9 and Trade Deals: positive but often lose significance under industry clustering for trade; uncertainty effects remain robust for deals (90th percentile → 0.31 percentage point excess CAR).

### Key channels and measurement
- TradeSensitivity_i:
  - Constructed by firm-by-firm regressions over 2017–2023: ∆Sales_it = η_0 + η_1i ∆USExports_jkt + ε_ijkt; η_1i saved as TradeSensitivity_i for tradable sectors.
  - Interpreted as proxy for net trade exposure capturing market access shock and correlated intermediate input cost shock.
- UncertaintySensitivity_j:
  - Based on Kim and Kung (2017) sectoral asset redeployability index computed from BEA capital flow table across 123 industries; index inverted and one added so higher values indicate greater sensitivity to uncertainty.
  - Firms in sectors with low redeployability face higher adjustment costs under uncertainty.
- Distinctness:
  - Correlation between average trade sensitivity (SIC3) and uncertainty sensitivity = 0.0158; firm-level correlation = 0.1082 — channels are largely orthogonal.

### Data, sample, and summary statistics
- Data sources and coverage:
  - Firm-level data from Datastream (Worldscope): primary operating sector (SIC 4-digit), daily closing stock prices from February 28, 2025 to August 1, 2025, financial characteristics for year 2023, firm-level annual revenue data 2017–2023.
  - Sectoral U.S. export data 2017–2023 from World Bank WITS.
  - Sectoral asset redeployability index by Kim and Kung (2017) at SIC 3-digit.
- Baseline sample: 11,741 firms across 138 SIC 3-digit industries in 48 economies.
- Winsorization and imputation:
  - TradeSensitivity_i and UncertaintySensitivity_j winsorized at 5 percent; all other variables winsorized at 1 percent.
  - Missing redeployability values imputed with SIC 2-digit average.
- Selected summary statistics (Table 2):
  - CRRLib_i[−1, +1]: Observations 11,741; Mean -0.0303; S.D. 0.0572; Min -0.2281; Max 0.1503
  - CRRPause_i[−1, +1]: Observations 11,741; Mean 0.0257; S.D. 0.0534; Min -0.1502; Max 0.2298
  - CRRDeal_i[−1, +1]: Observations 8,086; Mean 0.0126; S.D. 0.0399; Min -0.1025; Max 0.1797
  - TradeSensitivity_i: Observations 11,741; Mean 0.2281; S.D. 0.9388; Min -3.0913; Max 5.6148
  - UncertaintySensitivity_j: Observations 11,741; Mean 0.6767; S.D. 0.0598; Min 0.5755; Max 0.8320
  - MTB_i: Observations 11,737; Mean 0.0068; S.D. 0.0263; Min -0.0155; Max 0.2012
  - Size_i: Observations 11,740; Mean 12.9959; S.D. 2.1622; Min 5.2012; Max 18.5000
  - ROA_i: Observations 11,718; Mean 2.291515; S.D. 15.3350; Min -186.4500; Max 34.1600
  - DA_i: Observations 11,738; Mean 22.2691; S.D. 19.6288; Min 0.0000; Max 167.8800

### Heterogeneity: tradable vs non-tradable firms (section 4.3)
- Sample expansion: non-tradable firms included using regressions ∆Sales_it = σ_0 + σ_1i ∆USExports_kt + ε_ikt over 2017–2023 to construct TradeSensitivity_i; matched to Kim and Kung (2017) redeployability index; excluded financial sectors (SIC 600–699).
- Non-tradable sample summary (Table 8):
  - CRRLib_i [−1, +1]: Observations 11,956; Mean -0.0268; S.D. 0.0540; Min -0.2279; Max 0.1506
  - TradeSensitivity_i: Observations 11,956; Mean 0.1194; S.D. 0.9546; Min -1.8677; Max 2.4371
  - UncertaintySensitivity_j: Observations 11,956; Mean 0.6493; S.D. 0.0757; Min 0.5755; Max 0.8320
  - MTB_i: Observations 11,949; Mean 0.00740; S.D. 0.0283; Min -0.0157; Max 0.2053
  - Size_i: Observations 11,952; Mean 12.83062; S.D. 2.42025; Min 5.2067; Max 18.5589
  - ROA_i: Observations 11,912; Mean 1.245217; S.D. 17.5547; Min -185.1900; Max 34.1700
  - DA_i: Observations 11,944; Mean 25.4627; S.D. 23.7453; Min 0.0000; Max 166.5600
- Pooled split-slope estimates (Table 9) — key coefficients and statistical tests:
  - TradeSensitivity_i × Tradable_i:
    - April 2: -0.00756 (standard error 0.00129) ∗∗∗
    - April 9: 0.00443 (standard error 0.00129) ∗∗∗
    - Deal Days: 0.00164 (standard error 0.00123)
  - UncertaintySensitivity_j × Tradable_i:
    - April 2: -0.0510 (standard error 0.0220) ∗∗
    - April 9: 0.0368 (standard error 0.0211) ∗
    - Deal Days: 0.0171 (standard error 0.00885) ∗
  - TradeSensitivity_i × NonTradable_i:
    - April 2: -0.00128 (standard error 0.000522) ∗∗
    - April 9: -0.00111 (standard error 0.000526) ∗∗
    - Deal Days: 0.000342 (standard error 0.000547)
  - UncertaintySensitivity_j × NonTradable_i:
    - April 2: -0.00335 (standard error 0.0141)
    - April 9: -0.00641 (standard error 0.0164)
    - Deal Days: 0.00276 (standard error 0.00888)
  - NonTradable_i = 1 (baseline):
    - April 2: -0.0242 (standard error 0.0165)
    - April 9: 0.0309 (standard error 0.0152) ∗∗
    - Deal Days: 0.00642 (standard error 0.00850)
  - Observations: April 2 = 236,972; April 9 = 236,971; Deal Days = 15,463 (printed as 15463).
  - Formal p-values for equality of marginal effects across tradable and non-tradable:
    - Equality of Trade Sensitivity: April 2 p-value 0.000; April 9 p-value 0.000; Deal Days p-value 0.318
    - Equality of Uncertainty Sensitivity: April 2 p-value 0.056; April 9 p-value 0.060; Deal Days p-value 0.254
- Interpretation:
  - Tradable firms drive the bulk of valuation responses to April 2 and April 9 through both channels; non-tradable effects are much smaller.
  - During Deal Days, valuation changes for tradable and non-tradable firms are statistically indistinguishable (p = 0.318), implying broader investor-sentiment effects.

### Recent trade policy developments described
- January 2025: United States implemented a faster cadence of new trade measures including executive orders and proclamations introducing additional tariffs; actions applied to a broader set of economies relative to earlier phases of U.S.–China tensions.
- April 2, 2025 announcement:
  - Universal baseline tariff of 10% on all imports, to take effect on April 5.
  - Additional tariffs ranging from 10% to 50% on imports from a number of countries, with implementation beginning on April 9.
  - No additional tariff announced for the United Kingdom; the European Union was subject to an additional 20%.
  - Previously announced sector-specific tariffs on steel, aluminum, and automobiles remained in place.
  - Exemptions initially granted for semiconductors, pharmaceuticals, copper, and lumber.
- April 9, 2025 adjustment:
  - Implemented a 90-day pause of the country-specific tariffs announced earlier in the month from most trading partners but not China.
  - Maintained a uniform 10% ad valorem duty.
- Since May 2025: U.S. gradually achieved a series of trade agreements announced through end of July 2025; many deals not finalized or implemented and represent preliminary agreements.

### Policy implications and recommendations
- Maintain stable and predictable trade relationships to support investor confidence and medium-term growth.
- Enhance firms’ adaptability to external shocks:
  - Promote diversification of export markets.
  - Encourage investment in flexible production technologies.
  - Strengthen access to finance.
- Regional suggestion: deepen integration and broaden partnerships with other trading partners to cushion impact of future global policy shifts.

_Italic: Source: wpiea2026070-source-pdf - 4.1 Key findings; 4.3 Additional results — Trade Policy Shocks and Corporate Valuations, Working Paper No. WP/2026/070_

### 1.05 percentage points higher than those of a firm with low exposure on both margins.  This

### wpiea2026070-source-pdf - 1.05 percentage points higher than those of a firm with low exposure on both margins.  This

### Key empirical findings on firm-level stock returns
- A firm with high exposure on both trade-sensitivity and uncertainty-sensitivity experienced returns that were 1.05 percentage points higher (in magnitude) than those of a firm with low exposure on both margins around the initial escalation.
- After trade agreements and declining expected tariffs and trade policy uncertainty, comparing firms at the 90th and 10th percentiles:
  - Higher trade sensitivity is associated with a 0.33 percentage point additional increase in cumulative returns.
  - Higher uncertainty sensitivity is associated with a 0.34 percentage point additional increase in cumulative returns.
  - A firm with high exposure along both dimensions enjoys an additional cumulative return of about 0.67 percentage points from trade deals.
  - The 0.67 percentage point gain is close to 10 percent of the three-day gain of 8.5 percent around the deal dates for tradable firms over the same range.
- The trade and uncertainty measures are proxies constructed from historical data and thus subject to measurement error; the authors note the estimated effects are likely conservative and should be interpreted as lower-bound estimates of the underlying economic impact.

### Contributions to the literature
- Relates to stock market responses to trade policy shocks:
  - Builds on findings that tariff reductions and regional trade agreements generate heterogeneous valuation effects across firms and countries.
  - Consistent with literature showing protectionism and events like the Brexit referendum lowered stock prices by raising expected trade barriers and uncertainty.
  - Complements U.S.–China trade war studies emphasizing export exposure and supply-chain dependencies as key channels.
  - Links to structural work on trade and multinational production highlighting how firms’ foreign activities shape aggregate exposure to trade shocks.
- Complements macroeconomic studies of U.S. tariffs and geoeconomic fragmentation:
  - Connects micro (firm-level equity repricing) evidence to documented sizable effects of U.S. tariffs on prices, welfare, and regional labor markets.
  - Consistent with broader patterns of gradual fragmentation of trade and FDI along geopolitical lines.
- Connects to literature on policy uncertainty, global financial conditions, and firm behavior:
  - Confirms that higher uncertainty reduces investment and depresses firm values and that trade policy uncertainty can deter export entry and lower trade volumes.
  - Shows international spillovers where U.S. policy uncertainty raises risk premia and transmits to foreign stock markets.
  - Establishes that trade exposure and uncertainty sensitivity jointly shape global stock market responses to recent U.S. trade policy shocks.

### Recent trade policy developments described
- January 2025: United States implemented a faster cadence of new trade measures including executive orders and proclamations introducing additional tariffs; actions applied to a broader set of economies relative to earlier phases of U.S.–China tensions.
- April 2, 2025 announcement:
  - Universal baseline tariff of 10% on all imports, to take effect on April 5.
  - Additional tariffs ranging from 10% to 50% on imports from a number of countries, with implementation beginning on April 9.
  - No additional tariff announced for the United Kingdom; the European Union was subject to an additional 20%.
  - Previously announced sector-specific tariffs on steel, aluminum, and automobiles remained in place.
  - Exemptions initially granted for semiconductors, pharmaceuticals, copper, and lumber.
- April 9, 2025 adjustment:
  - Implemented a 90-day pause of the country-specific tariffs announced earlier in the month from most trading partners but not China.
  - Maintained a uniform 10% ad valorem duty.
- Since May 2025: U.S. gradually achieved a series of trade agreements announced through end of July 2025 (table provides dates and deal details for multiple partners). Note: Many deals not finalized or implemented and represent preliminary agreements; final parameters could deviate substantially.

### Empirical approach and identification
- Event-study framework focusing on:
  - The April 2 universal tariff announcement.
  - The April 9 pause announcement.
  - Country-specific trade agreements (three-day windows centered on the first local trading day after the announcements).
- Outcome variable:
  - Y_i = CRR_i[−1, +1] = sum from t = −1 to +1 of R_it (cumulative raw returns over the three-day window).
- Cross-sectional regression estimated:
  - Y_i = ρ_0 + ρ_1 TradeSensitivity_i + ρ_2 UncertaintySensitivity_j + X_i + δ_k + ε_ijk
    - i: firm; j: sector; k: country.
    - X_i includes firm size (log of total assets in U.S. dollars), market-to-book ratio (MTB), debt-to-asset ratio (DA), return on assets (ROA).
    - δ_k is the country fixed effect to account for different tariff rates across countries.
- Interpretation:
  - For April 2 and 9 announcements, coefficients capture marginal effect of exposure intensity on firm valuations conditional on the universal shock.
  - For trade deals (sample restricted to countries reaching agreements), estimates capture heterogeneous response within the group of countries that reached agreement.

### Key channels and measurement
- TradeSensitivity_i:
  - Constructed by regressing percentage change in a firm’s sales on percentage change in its country’s exports to the U.S. at the SIC 3-digit level over 2017–2023:
    - ∆Sales_it = η_0 + η_1i ∆USExports_jkt + ε_ijkt
  - η_1i saved as TradeSensitivity_i; regression run firm-by-firm for tradable sectors.
  - Interpreted as a proxy for net trade exposure capturing market access shock and correlated intermediate input cost shock.
- UncertaintySensitivity_j:
  - Uses sectoral asset redeployability index by Kim and Kung (2017), computed from BEA capital flow table across 123 industries.
  - Index inverted so higher values indicate greater sensitivity to uncertainty; one added so index takes positive values for all industries.
  - Firms in sectors with low redeployability face higher adjustment costs under uncertainty.
- Distinctness of channels:
  - Scatter plot of average trade sensitivity (SIC3 level) vs uncertainty sensitivity shows low correlation of 0.0158; at firm level correlation coefficient is 0.1082.
  - Indicates trade and uncertainty are orthogonal channels.

### Data, sample, and summary statistics
- Data sources and coverage:
  - Firm-level data from Datastream (Worldscope): primary operating sector (SIC 4-digit), daily closing stock prices from February 28, 2025 to August 1, 2025, financial characteristics for year 2023, firm-level annual revenue data 2017–2023.
  - Sectoral U.S. export data 2017–2023 from World Bank WITS to obtain trade sensitivity index.
  - Merged with sectoral asset redeployability index by Kim and Kung (2017) at SIC 3-digit level.
- Sample:
  - 11,741 firms across 138 SIC 3-digit industries in 48 economies.
- Data cleaning:
  - Retained firms with identifiable operating sectors and non-missing data on sales and control variables.
  - Used firms with complete sales and sectoral U.S. exports data from 2017 to 2023 to run Equation (2).
  - Imputed missing industry redeployability values with SIC 2-digit average.
  - TradeSensitivity_i and UncertaintySensitivity_j are winsorized at the 5 percent level. All other variables winsorized at the 1 percent level.
- Selected summary statistics (from Table 2):
  - CRRLib_i[−1, +1]: Observations 11,741; Mean -0.0303; S.D. 0.0572; Min -0.2281; Max 0.1503
  - CRRPause_i[−1, +1]: Observations 11,741; Mean 0.0257; S.D. 0.0534; Min -0.1502; Max 0.2298
  - CRRDeal_i[−1, +1]: Observations 8,086; Mean 0.0126; S.D. 0.0399; Min -0.1025; Max 0.1797
  - TradeSensitivity_i: Observations 11,741; Mean 0.2281; S.D. 0.9388; Min -3.0913; Max 5.6148
  - UncertaintySensitivity_j: Observations 11,741; Mean 0.6767; S.D. 0.0598; Min 0.5755; Max 0.8320
  - MTB_i: Observations 11,737; Mean 0.0068; S.D. 0.0263; Min -0.0155; Max 0.2012
  - Size_i: Observations 11,740; Mean 12.9959; S.D. 2.1622; Min 5.2012; Max 18.5000
  - ROA_i: Observations 11,718; Mean 2.291515; S.D. 15.3350; Min -186.4500; Max 34.1600
  - DA_i: Observations 11,738; Mean 22.2691; S.D. 19.6288; Min 0.0000; Max 167.8800

### Empirical patterns illustrated in figures (described)
- Figure 1:
  - Plots average trade sensitivity (SIC3) against uncertainty sensitivity; red dashed lines denote medians.
  - Reports correlation coefficient 0.0158 between the two measures; firm-level correlation coefficient 0.1082.
  - Identifies industries with highest trade sensitivity (e.g., Crude Petroleum SIC 131; Sand and Gravel SIC 144; Electronic Components SIC 367; Costume Jewelry SIC 396) and industries with highest uncertainty sensitivity (e.g., Textile Mills SIC 221-229; Coal Mining SIC 122; petroleum and electronics sectors).
- Figure 2:
  - Panel (A): Average cumulative returns for firms in top vs bottom quartiles of trade sensitivity show similar pre-April 2 trends, sharper decline for high trade exposure firms after April 2, and reversal after April 9 pause, with persistent divergence.
  - Panel (B): Similar pattern for uncertainty-sensitive sectors (top vs bottom quartiles of redeployability-based sensitivity).
  - Panel (C): Firms high on both channels (top-right quadrant of Figure 1) show the most striking persistent gap in cumulative returns relative to firms low on both channels.

*Italic source: Content extracted from the provided IMF PDF chapter/section.*

### 4.1    Key findings

### 4.1 Key findings

### April 2 Announcement
- Market reaction priced the universal tariff shock immediately into corporate valuations through two channels: TradeSensitivity_i and UncertaintySensitivity_j.
- Univariate and multivariate regressions show negative and statistically significant coefficients for both channels; significance persists after adding firm-level controls and when clustering standard errors at the industry level.
- Estimated coefficients (Table 3, April 2):
  - TradeSensitivity_i: -0.00815 (col (1)), -0.00764 (col (2)), -0.00789 (col (3)), -0.00789 (col (4)) — industry-clustered SE specification retains significance (col (5)).
  - UncertaintySensitivity_j: -0.0552 (col (1)), -0.0504 (col (2)), -0.0542 (col (3)), -0.0542 (col (4)) — remains significant in industry-clustered specification.
- Economic magnitudes:
  - Difference between 90th and 10th percentiles of TradeSensitivity_i is 1.383 → firms very sensitive to trade experienced an additional 1.09 percentage points decline in three-day cumulative returns compared to firms not very sensitive to trade.
  - Difference between 90th and 10th percentiles of UncertaintySensitivity_j is 0.186 → firms in low asset redeployability sectors underperformed peers by 1.01 percentage points.
- Interpretation:
  - Markets penalized firms for high U.S. export exposure and limited investment flexibility.
  - Negative valuation responses align with theoretical predictions that uncertainty shocks raise discount rates and delay investment.
  - Clustering at the industry level implies the TradeSensitivity coefficient exploits firm variation around industry averages.

### April 9 Announcement
- April 9 introduced a 90-day pause of reciprocal tariffs on imports from most trading partners, excluding China; produced positive market responses.
- Estimated coefficients (Table 3, April 9):
  - TradeSensitivity_i: 0.00410 (col (1)), 0.00380 (col (2)), 0.00374 (col (3)), 0.00374 (col (4)) — significance retained.
  - UncertaintySensitivity_j: 0.0319 (col (1)), 0.0295 (col (2)), 0.0283 (col (3)), 0.0283 (col (4)) — loses significance under industry clustering in col (5).
- Economic magnitudes:
  - Interdecile (90th–10th percentile) range in trade sensitivity → additional 0.52 percentage point return.
  - Equivalent range in uncertainty sensitivity → 0.53 percentage point increase.
- Interpretation:
  - Export-intensive firms and sectors with elevated uncertainty exposure realized larger gains as immediate policy risks diminished.
  - Residual policy uncertainty remains priced by markets when accounting for sector-wide correlation (industry-clustered SE).

### Trade Deals
- Signing of bilateral trade agreements reduced uncertainty and generated positive market responses.
- Estimated coefficients (Table 3, Trade Deals):
  - TradeSensitivity_i: 0.00256 (col (1)), 0.00244 (col (2)), 0.00237 (col (3)), 0.00237 (col (4)) — significance attenuates under clustering in col (5).
  - UncertaintySensitivity_j: 0.0180 (col (1)), 0.0169 (col (2)), 0.0180 (col (3)), 0.0180 (col (4)) — statistically significant.
- Economic magnitudes:
  - Firms at 90th percentile of trade sensitivity experienced a 0.33 percentage point excess return relative to 10th percentile.
  - Resolution of uncertainty generated a 0.34 percentage point relative gain for firms in high-rigidity sectors.
- Important asymmetry:
  - Trade deal recovery is roughly one-third the size of the initial April 2 announcement drop, indicating incomplete reversal and a continuing residual risk premium for global fragmentation.

### Robustness Checks

- Four-Day Window ([−1, +2])
  - Extending event window to four days captures delayed price discovery; effects generally strengthen for the April 2 announcement.
  - Table 4 estimates (April 2, col (5)):
    - TradeSensitivity_i: -0.0166 (col (4)/(5) consistent) → firms in top 10th percentile incurred an additional 2.30 percentage point decline over the four-day window relative to bottom 10th percentile.
    - UncertaintySensitivity_j: -0.0715 (col (4)/(5)) → firms in very uncertainty-sensitive sectors suffered an additional 1.33 percentage point decline over the four-day window.
  - April 9 four-day effects:
    - 90th percentile trade sensitivity → additional 0.47 percentage point increase in cumulative returns.
  - Trade deals four-day effects:
    - Additional gains for 90th percentile trade sensitivity = 0.12 percentage point; for 90th percentile uncertainty sensitivity = 0.36 percentage point.

- One Percent Winsorization
  - Baseline winsorization at 5 percent; robustness check uses 1 percent winsorization to retain more extreme values.
  - Table 5 estimates (col (5)):
    - April 2:
      - TradeSensitivity_i: -0.00253 (attenuated relative to baseline -0.00789) but remains statistically significant at 1 percent.
      - UncertaintySensitivity_j: -0.0570 (nearly identical to baseline -0.0542).
    - April 9:
      - TradeSensitivity_i: 0.00119 (col (3)/(4)/(5) reported as 0.00119) — small positive effect.
      - UncertaintySensitivity_j: 0.0235.
    - Trade Deals:
      - TradeSensitivity_i: 0.000837 (loses statistical significance).
      - UncertaintySensitivity_j: 0.0192 (remains significant).
  - Interpretation:
    - Trade sensitivity is susceptible to estimation noise in tails; attenuation under 1% winsorization is consistent with measurement error.
    - Uncertainty sensitivity is stable across winsorization thresholds, validating baseline 5% winsorization choice.

- All Tradable Firms (imputed sample)
  - Extended sample includes 16,267 firms across 142 SIC 3-digit industries in 48 economies with imputed trade sensitivity for firms with missing years.
  - Table 6 estimates:
    - April 2:
      - TradeSensitivity_i: -0.00506 (col (1)), -0.00470 (col (2)), -0.00504 (col (3)/(4)) — significant.
      - UncertaintySensitivity_j: -0.0516 (col (1)), -0.0489 (col (2)), -0.0557 (col (3)/(4)) — significant and strengthens with clustering.
      - Interdecile differences: trade = 1.807 → additional 0.91 percentage point decline for firms at top 10th percentile; uncertainty = 0.186 → additional 1.04 percent decline for 90th percentile industries.
    - April 9:
      - TradeSensitivity_i: 0.00834 (col (1)), 0.00799 (col (2)), 0.00762 (col (3)/(4)) — robust and significant.
      - UncertaintySensitivity_j: 0.0512 (col (1)), 0.0467 (col (2)), 0.0469 (col (3)/(4)) — weakens to weak significance in some specifications.
    - Trade Deals: results lose statistical significance (attenuation consistent with imputation measurement error).
  - Interpretation:
    - Core findings robust to extended sample; universal tariff shock produced systemic negative valuation effects across tradable firms while trade-deal recovery was more selective.

- Abnormal Returns (Cumulative Abnormal Returns, CAR)
  - Re-estimated using CAR from Market Model with firm-specific α_i and β_i estimated from monthly returns over five years preceding December 2023.
  - Table 7 estimates (CAR, three-day window [−1, +1]):
    - April 2:
      - TradeSensitivity_i: -0.00316 (col (1)), -0.00289 (col (2)), -0.00317 (col (3)/(4)/(5)) — significant.
      - UncertaintySensitivity_j: -0.0289 (col (1)), -0.0271 (col (2)), -0.0289 (col (3)/(4)/(5)) — negative and significant, though attenuated vs raw returns.
      - Economic magnitude: firms at 90th percentile trade sensitivity incurred an additional 0.44 percentage point decline in CAR relative to 10th percentile.
    - April 9:
      - TradeSensitivity_i: 0.00201 (col (1)), 0.00184 (col (2)), 0.00162 (col (3)/(4)/(5)) — positive but loses significance under industry clustering.
      - UncertaintySensitivity_j: 0.0184 (col (1)), 0.0172 (col (2)), 0.0145 (col (3)/(4)/(5)) — positive but loses significance under industry clustering.
    - Trade Deals:
      - TradeSensitivity_i: 0.00161 (col (1)), 0.00149 (col (2)), 0.00136 (col (3)/(4)/(5)) — positive effects become insignificant in clustered specification.
      - UncertaintySensitivity_j: 0.0177 (col (1)), 0.0170 (col (2)), 0.0167 (col (3)/(4)/(5)) — remains strongly robust; firms in sectors at 90th percentile of uncertainty sensitivity experienced a 0.31 percentage point excess CAR relative to 10th percentile.
  - Interpretation:
    - Negative trade exposure effects after April 2 are not driven by market beta; uncertainty effects partly reflect systematic risks but remain important and firm-specific after trade deals.

_Italic: Source: wpiea2026070-source-pdf - 4.1 Key findings_

### 4.3    Additional results

### 4.3    Additional results

### Firms in non-tradable sectors — sample construction and estimation
- Expanded sample to include firms in non-tradable sectors, constructing their trade sensitivity following Section 3.2.
- Rationale: non-tradable firms provide intermediate services to tradable firms within their domestic market; proxy indirect exposure to U.S. trade policy using changes in the firm’s home country aggregate U.S. exports.
- Regression estimated firm-by-firm over 2017–2023:
  - ∆Sales_it = σ_0 + σ_1i ∆USExports_kt + ε_ikt  (Equation (3))
  - Saved σ_1i as TradeSensitivity_i.
- Matched non-tradable firms to the Kim and Kung (2017) sectoral asset redeployability index at the 3-digit SIC level and merged with daily closing price data.
- Excluded firms in financial sectors (SIC 3-digit codes 600 to 699).

### Summary statistics for non-tradable sample (from Table 8)
- CRRLib_i [−1, +1]: Observations 11,956; Mean -0.0268; S.D. 0.0540; Min -0.2279; Max 0.1506
- CRRPause_i [−1, +1]: Observations 11,956; Mean 0.02100; S.D. 0.0512; Min -0.1502; Max 0.2288
- CRRDeal_i [−1, +1]: Observations 7,377; Mean 0.01270; S.D. 0.0384; Min -0.1025; Max 0.1797
- TradeSensitivity_i: Observations 11,956; Mean 0.1194; S.D. 0.9546; Min -1.8677; Max 2.4371
- UncertaintySensitivity_j: Observations 11,956; Mean 0.6493; S.D. 0.0757; Min 0.5755; Max 0.8320
- MTB_i: Observations 11,949; Mean 0.00740; S.D. 0.0283; Min -0.0157; Max 0.2053
- Size_i: Observations 11,952; Mean 12.83062; S.D. 2.42025; Min 5.2067; Max 18.5589
- ROA_i: Observations 11,912; Mean 1.245217; S.D. 17.5547; Min -185.1900; Max 34.1700
- DA_i: Observations 11,944; Mean 25.4627; S.D. 23.7453; Min 0.0000; Max 166.5600
- Notes: TradeSensitivity_i and UncertaintySensitivity_j are winsorized at 5% level. All other variables are winsorized at 1% level.

### Econometric specification for cross-sector comparison
- Pooled split-slope model estimated:
  - Y_i = ρ_0 + β_1 (TradeSensitivity_i × Tradable_i) + β_2 (TradeSensitivity_i × NonTradable_i)
    + γ_1 (UncertaintySensitivity_j × Tradable_i) + γ_2 (UncertaintySensitivity_j × NonTradable_i)
    + λ NonTradable_i + X_i + δ_k + ε_ijk  (Equation (4))
- Tradable_i and NonTradable_i are mutually exclusive sector indicators.
- β_1 and β_2 capture marginal effects of trade sensitivity for tradable and non-tradable firms.
- γ_1 and γ_2 capture sector-specific effects of uncertainty sensitivity.
- λ absorbs baseline difference in raw returns for non-tradable firms relative to tradable baseline.

### Main results — heterogeneity across tradable and non-tradable firms (from Table 9)
- Sample-wide evidence: tradable firms disproportionately affected by April 2 and April 9 announcements through both trade and uncertainty channels.
- Estimated coefficients (Table 9 columns labelled April 2, April 9, Deal Days):

  - TradeSensitivity_i × Tradable_i
    - April 2: -0.00756 ∗∗∗  (standard error 0.00129)
    - April 9: 0.00443 ∗∗∗  (standard error 0.00129)
    - Deal Days: 0.00164  (standard error 0.00123)

  - UncertaintySensitivity_j × Tradable_i
    - April 2: -0.0510 ∗∗  (standard error 0.0220)
    - April 9: 0.0368 ∗  (standard error 0.0211)
    - Deal Days: 0.0171 ∗  (standard error 0.00885)

  - TradeSensitivity_i × NonTradable_i
    - April 2: -0.00128 ∗∗  (standard error 0.000522)
    - April 9: -0.00111 ∗∗  (standard error 0.000526)
    - Deal Days: 0.000342  (standard error 0.000547)

  - UncertaintySensitivity_j × NonTradable_i
    - April 2: -0.00335  (standard error 0.0141)
    - April 9: -0.00641  (standard error 0.0164)
    - Deal Days: 0.00276  (standard error 0.00888)

  - NonTradable_i = 1 (baseline non-tradable effect)
    - April 2: -0.0242  (standard error 0.0165)
    - April 9: 0.0309 ∗∗  (standard error 0.0152)
    - Deal Days: 0.00642  (standard error 0.00850)

- Observations:
  - April 2: 236,972
  - April 9: 236,971
  - Deal Days: 154,63 (as printed: 15463)
- Fixed effects and controls:
  - Country FE: Y for all columns
  - Firm Controls: Y for all columns
  - Clustering: Industry for all columns

- Formal tests (p-values for equality of marginal effects across tradable and non-tradable):
  - p-value: Equality of Trade Sensitivity
    - April 2: 0.000
    - April 9: 0.000
    - Deal Days: 0.318
  - p-value: Equality of Uncertainty Sensitivity
    - April 2: 0.056
    - April 9: 0.060
    - Deal Days: 0.254

- Interpretation:
  - For April 2 and April 9 announcements, coefficients on TradeSensitivity × Tradable are consistent in magnitude with baseline estimates (reported elsewhere as −0.00789 for April 2 and 0.00374 for April 9).
  - The effect on non-tradable firms is much smaller for these announcement events; equality of coefficients across sectors is decisively rejected for the trade channel (p = 0.000 for both events).
  - During Deal Days, valuation changes for tradable and non-tradable firms are statistically indistinguishable (p = 0.318), implying trade deals generated broad shifts in investor sentiment lifting sectors relatively equally.

### Key takeaways (from section conclusions)
- Two channels shaped short-term market reactions to U.S. trade policy shocks: firms’ exposure to U.S. trade and firms’ sensitivity to investment uncertainty.
- After the initial tariff announcement:
  - Firms with greater export exposure to the United States experienced more pronounced stock price declines.
  - Firms operating in sectors with lower asset redeployability experienced more pronounced stock price declines.
  - The uncertainty channel was particularly relevant—investment inflexibility amplifies impact of policy shocks.
- Signing of trade agreements produced broad-based market gains, reflecting reduced uncertainty and greater clarity regarding future trade relations.

### Policy implications (as discussed)
- Maintain stable and predictable trade relationships to support investor confidence and medium-term growth.
- Policies to enhance firms’ adaptability to external shocks:
  - Promote diversification of export markets.
  - Encourage investment in flexible production technologies.
  - Strengthen access to finance.
- Regional policy suggestion:
  - Deepen integration and broaden partnerships with other trading partners to cushion impact of future global policy shifts.

*Source: 4.3 Additional results — Trade Policy Shocks and Corporate Valuations, Working Paper No. WP/2026/070*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026070-source-pdf.pdf_