## wpiea2022191-print-pdf

## Source details

**Canonical URL:** [wpiea2022191-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022191-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022191-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022191-print-pdf.pdf.json)

---

### Introduction and motivation
- Monetary policy in advanced economies, in particular in the United States, is identified as a key source of international spillovers.
- The new U.S. monetary policy tightening cycle is seen as a significant risk for the global economic recovery (IMF 2021).
- Spillover effects operate through multiple channels and depend on many factors, including domestic policy responses.
- The paper quantifies and compares the relative strength of these channels for 63 countries (29 AEs and 34 EMDEs) over the period 1996Q3 to 2016Q3 using a comprehensive quarterly firm-level dataset covering 23,482 firms.

### Key transmission channels (Mundell–Fleming framing)
- Balance sheet channel:
  - U.S. monetary tightening tightens foreign firms’ financial conditions, negatively affecting investment more for firms that are more leveraged and more dependent on external finance.
- Financial channel of the exchange rate:
  - Depreciation of the foreign currency vs. the U.S. dollar affects firm balance sheets when firms have large foreign-currency liabilities.
  - Statistically significant mainly for countries with flexible exchange rates.
- Trade channel:
  - Ambiguous a priori: “expenditure reducing effect” (lower U.S. import demand) vs. “expenditure switching effect” (foreign depreciation raises foreign demand for exports).
  - Short term: expenditure reducing channel dominates; medium term (about two years): expenditure switching can dominate.

### Data and identification
- Data source and sample:
  - S&P Capital IQ (CIQ) quarterly firm-level data; final sample after filtering: 23,482 firms; sample span 1996Q3 to 2016Q3; 63 countries; 20 industries (financial, insurance and utilities excluded).
- Main outcome:
  - Investment rate = Capital Expenditure/Net Property Plant and Equipment.
- Key firm-level exposures and covariates:
  - Leverage = Total Debt/Total Asset.
  - Foreign-currency liability ratio = Total foreign-currency Liabilities/Total Debt (CIQ aggregated by repayment currency; unspecified currency assumed USD in baseline).
  - Export Dependence Ratio = Exports/Output (baseline from WIOD; where WIOD unavailable, sectoral averages used).
  - Additional covariates: liquidity, size (log USD), collateral, bank debt ratio, age, dividend payment.
- Sample limitations and data handling:
  - Listed firms only; consolidated data; state-owned firms excluded; observations with negative total assets or debt dropped; winsorized at 5 percent.
- Identification of U.S. monetary policy shocks:
  - High-frequency surprises in Federal fund futures around FOMC announcements (30-minute window) used in a proxy-SVAR to instrument the one-year government bond yield.
  - Proxy-SVAR: one-year government bond yield instrumented together with industrial production, CPI, and excess bond premium; estimated 1973M1–2016M8 monthly; structural shocks aggregated to quarters.
  - Baseline instrument: three-month ahead Fed fund futures with one-year government bond yield (provides higher first-stage F-statistic and exceeds Stock-Yogo critical value).
  - Structural break allowed from November 2008 (post ZLB / QE).

### Empirical strategy
- Three-step approach:
  1. Average (unconditional) effect estimated with Jorda (2005) local projections:
     - log Y_{f,t+h} − log Y_{f,t−1} = β_h × i_t^{US} + γ_f + γ_{c s q} + ε_{f,t+h}
  2. Heterogeneous impacts via semi-parametric grouping (low/medium/high) per firm characteristic:
     - log Y_{f,t+h} − log Y_{f,t−1} = Σ_{g=1}^{G} β_{g h} × I[X_f ∈ g] × i_t^{US} + ρ_h Z_{f,t−1} + γ_f + γ_{c s t+h} + ε_{f,t+h}
     - Uses firm characteristic averages to reduce endogeneity; includes lagged firm controls Z_{f,t−1}.
  3. Two-dimension interaction specification to assess joint heterogeneity:
     - log Y_{f,t+h} − log Y_{f,t−1} = Σ_{g1=1}^{G1} Σ_{g2=1}^{G2} β_{g1 g2 h} × I[X_{f1} ∈ g1] × I[X_{f2} ∈ g2] × i_t^{US} + ρ_h Z_{f,t−1} + γ_f + γ_{c s t+h} + ε_{f,t+h}
- Estimation:
  - OLS with standard errors two-way clustered on firm and country-time (or clustered on firms where appropriate).
  - Fixed effects: firm and country-sector-quarter (or country-time where sector-country variation not applicable).

### Main empirical findings — average effects
- A 25-basis points U.S. monetary policy tightening leads to:
  - Peak investment decline of -1.3 percent after two quarters for foreign firms.
  - Peak firm revenue decline of about 0.6 percent after two quarters.
- The foreign firm investment response is comparable to, though slightly smaller than, U.S. firms in the sample.
- Robustness: results hold after removing large-country contributors, alternative lags, alternative shock measures, inclusion of time-varying firm characteristics, and alternative investment definitions.

### Spillover channels and heterogeneous firm responses
- Balance sheet channel (leverage):
  - Highly leveraged firms more responsive to U.S. shocks.
  - Difference in one-year-ahead investment response to a 25-basis points shock between top and bottom leverage quartiles ≈ 0.5 percentage point.
- Financial channel of the exchange rate (foreign-currency liability ratio):
  - Firms with higher foreign-currency liability ratios have larger investment declines.
  - Differential magnitude between top and bottom quartiles ≈ 0.5 percentage point.
  - More pronounced in countries with flexible exchange rate regimes.
- Trade channel (export dependence):
  - Short term: firms in sectors with higher export dependence are more negatively affected (expenditure reducing).
  - Medium term (≈ two years): expenditure switching leads firms in higher export-dependent sectors to fare relatively better.
  - Trade exposure is the most relevant transmission channel for firm revenue.
- Interactions and heterogeneity:
  - Channels can amplify each other: spillovers larger for firms that are both more leveraged and have higher foreign-currency liabilities.
  - Leverage role is larger for smaller and less liquid firms, consistent with tighter borrowing constraints.
  - Exchange-rate related adverse impacts on investment amplified under flexible exchange rate regimes.

### Robustness checks and additional findings
- Joint estimation of all three channels yields qualitatively similar and statistically indistinguishable results from separate estimations—each channel plays an independent and significant role.
- Additional firm characteristics (size, bank debt share, dividend payments, age) interact with shocks: smaller, younger, lower-dividend-paying firms experience larger investment declines.
- Alternative classification methods (country-specific percentiles, median splits, first-year classification) yield similar results.
- Dropping large-country contributors (Canada, China, Hong Kong, India, Japan, Taiwan) one at a time does not materially change results.
- Trade channel more important for revenue than for investment; balance sheet and exchange-rate channels more important for investment.

### Back-of-the-envelope aggregation (investment and revenue)
- Method:
  - Use joint-specification estimates to compute channel contributions to aggregate investment response; cumulative responses sum impulse responses over 12 quarters to a 25-bps shock.
  - Assumptions impose lower-bound contributions by setting medium- and low-exposure group impacts to zero.
- Key figures (Table 2: cumulative responses over 12 quarters to a 25-bps U.S. monetary policy shock):
  - Unconditional: -5.86
  - Leverage: -4.33
  - Foreign-Currency Liability Ratio: -2.77
  - Export Dependence: -0.18
- Capital Expenditure Share of High Group:
  - Leverage: 34.2%
  - Foreign-Currency Liability Ratio: 28.4%
  - Export Dependence: 13.8%
- Contribution to the Average Investment Response:
  - Leverage: 25.5%
  - Foreign-Currency Liability Ratio: 13.4%
  - Export Dependence: 0.41%
- For firm revenue (Table A4.1 cumulative responses over 12 quarters):
  - Unconditional cumulative response: -3.77
  - Leverage cumulative response: -0.25
  - Foreign-Currency Liability Ratio cumulative response: -1.74
  - Export Dependence cumulative response: -2.99
- Revenue channel contributions (from Table A4.1):
  - Contribution to the Average Revenue Response:
    - Unconditional: 1.8%
    - Foreign-Currency Liability Ratio: 9.2%
    - Export Dependence: 13.5%
- Caveat: lower-bound assumption likely underestimates true general equilibrium contributions.

### Interactions across channels (two-dimension exercises)
- Balance sheet × financial (leverage × FX liability ratio):
  - Both dimensions have significant independent impacts; high leverage adverse effects amplified when FX exposure is also high.
- Balance sheet × trade:
  - Leverage remains important irrespective of export dependence.
  - Export dependence becomes statistically insignificant for investment once grouped by leverage, but remains significant for revenue.
- Financial × trade:
  - FX liability ratio matters more in high export-dependence sectors; insignificant in low export-dependence sectors for investment.
  - Trade remains important for revenue across FX liability groups.
- Leverage interactions with size and liquidity:
  - Leverage impact larger for smaller firms and for firms with lower liquidity.
- Exchange rate regime:
  - Foreign-currency borrowing has a more pronounced negative impact on investment in countries with more flexible exchange rate regimes.

### Empirical magnitudes and summary statistics (selected)
- Investment rate (log):
  - No. of Obs.: 887,901; Mean: -3.514; Std. Dev.: 1.346; 25th Pctile: -4.324; Median: -3.366; 75th Pctile: -2.571
- Revenue (log USD):
  - No. of Obs.: 1,156,564; Mean: 5.728; Std. Dev.: 3.182; 25th Pctile: 3.786; Median: 5.695; 75th Pctile: 7.856
- Leverage:
  - No. of Obs.: 1,250,768; Mean: 0.197; Std. Dev.: 0.182; 25th Pctile: 0.021; Median: 0.162; 75th Pctile: 0.324
- Foreign-currency liability ratio (%):
  - No. of Obs.: 708,800; Mean: 15.371; Std. Dev.: 32.335; 25th Pctile: 0; Median: 0; 75th Pctile: 1.964
- Export dependence ratio (%):
  - No. of Obs.: 2,298,087; Mean: 25.024; Std. Dev.: 23.36; 25th Pctile: 5.752; Median: 16.735; 75th Pctile: 43.686
- Size (log USD):
  - No. of Obs.: 1,258,486; Mean: 4.701; Std. Dev.: 2.253; 25th Pctile: 3.389; Median: 4.818; 75th Pctile: 6.200
- Bank debt ratio:
  - No. of Obs.: 542,363; Mean: 0.763; Std. Dev.: 0.292; 25th Pctile: 0.597; Median: 0.903; 75th Pctile: 1
- Liquidity:
  - No. of Obs.: 1,224,032; Mean: 0.176; Std. Dev.: 0.171; 25th Pctile: 0.050; Median: 0.121; 75th Pctile: 0.246
- Collateral:
  - No. of Obs.: 1,161,006; Mean: 0.513; Std. Dev.: 0.237; 25th Pctile: 0.325; Median: 0.502; 75th Pctile: 0.700
- Age:
  - No. of Obs.: 2,141,964; Mean: 40.593; Std. Dev.: 33.863; 25th Pctile: 20; Median: 30; 75th Pctile: 50

### Methodological and robustness notes (figures, clustering, inference)
- Impulse responses to a 25-bps U.S. monetary policy shock; Y-axis in percent.
- Graphical conventions:
  - Solid blue line: estimated impulse response.
  - Dashed blue and red lines: 90% and 68% confidence intervals.
- Standard errors: typically two-way clustered on firm and country-time; some specifications cluster on firms only.
- Fixed effects: firm and sector-country-time fixed effects commonly used; country-time fixed effects used when sector-country variation cannot be exploited.
- Robustness exercises include: alternative shock measures (ED4, FF4, MP3), alternative dependent-variable lags, inclusion of time-varying firm characteristics, alternative investment definitions, exclusion of large-country contributors.

### Theoretical framework (Annex 1: Mundell–Fleming formalization)
- Two-country setup with domestic tradable and non-tradable sectors; closed foreign economy (U.S.).
- Exchange rate and corporate borrowing wedge linked to U.S. policy rate R^* via parameters g, h, γ_c.
- Closed-form partial derivatives show three identifiable terms for ∂Y_c^T/∂R^*:
  - Balance Sheet Channel: − b h
  - Financial Channel of the Exchange Rate: − s (h + g(1 − h))
  - Trade Channel: [ d (h + g(1 − h)) − b f ] (ambiguous sign)
- Empirical proxies recommended:
  - b proxied by firm leverage.
  - s proxied by foreign-currency debt share.
  - Trade channel tested via sectoral export dependence.

*Source — wpiea2022191-print-pdf (IMF Working Paper No. WP/2022/191).*

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

### wpiea2022191-print-pdf - References .............................................................................................................

### Introduction and motivation
- Monetary policy in advanced economies, and in particular in the United States, has been identified as a key source of international spillovers.
- The new monetary policy tightening cycle in the United States is seen as a significant risk for the global economic recovery (IMF 2021).
- Spillover effects depend on many factors—including the domestic policy response—and operate through multiple channels.
- The paper quantifies and compares the relative strength of these channels for a set of 63 advanced economies (AEs) and emerging market and developing economies (EMDEs) over a 20-year period using a comprehensive quarterly firm-level dataset.

### Key transmission channels (as framed in the Mundell-Fleming extension)
- Balance sheet channel:
  - U.S. monetary policy tightening leads to tighter financial conditions for firms in foreign countries, negatively impacting investment, especially for firms that are more leveraged and more dependent on external finance.
- Financial channel of the exchange rate:
  - Operates through depreciation of the foreign currency vis-à-vis the U.S. dollar affecting firms’ balance sheets when firms have large foreign-currency liabilities.
- Trade channel:
  - Ambiguous a priori effects: tighter U.S. monetary policy can lower U.S. import demand (“expenditure reducing effect”) but can increase foreign demand for imports via depreciation in foreign countries (“expenditure switching effect”).
- Annex 1 contains a formal exposition of the different channels.

### Data, identification, and empirical approach
- Dataset:
  - Quarterly firm-level dataset covering more than 20,000 firms in the estimation sample.
- Identification of firms’ exposures:
  - Observable measures used include debt-to-assets ratio (leverage) and foreign currency liabilities.
- Empirical strategy:
  - Difference-in-difference framework assigning firms into exposure groups (for example, low, medium, high) with country-sector-time fixed effects.
  - Use of semi-parametric approach building on Cloyne et al. (2019) and Duval et al. (2021) to estimate differential impulse responses for each group via local projections (Jorda 2005).
- Identification of exogenous U.S. monetary policy shocks:
  - Follows Duval et al. (2021): uses high-frequency movements in U.S. interest rate futures around Federal Open Market Committee (FOMC) meetings as instruments for the one-year bond yield in a proxy-Structural Vector Autoregression (SVAR) framework.
- Fixed effects:
  - Country-sector-time fixed effects control for domestic macroeconomic shocks and their differential sectoral effects, mitigating endogeneity concerns present in country-level analyses.

### Main empirical findings
- Independent and statistically significant roles for all three channels:
  - Balance sheet channel: U.S. monetary policy shocks have larger effects on investment for firms that are more leveraged.
  - Financial channel of the exchange rate: Larger effects for firms with a higher share of debt in foreign currency; statistically significant only for countries with flexible exchange rates.
  - Trade channel: Firms operating in sectors with higher trade linkages experience larger effects; trade exposure is the most relevant transmission channel for firm revenue.
- Interaction and heterogeneity:
  - Channels can amplify each other: spillovers are larger for more leveraged firms that also have higher foreign-currency liabilities.
  - The role of leverage is larger for smaller firms and firms with lower liquidity, consistent with tighter borrowing constraints for smaller and less liquid firms.
- Aggregate relevance:
  - Back-of-the-envelope calculations suggest the balance sheet channel contributes the most to the aggregate investment response to U.S. monetary policy shocks.
- Robustness:
  - Results are robust to alternative proxy variables for transmission channels, alternative classification approaches for firm exposure groups, different sets of controls, and other measures of U.S. monetary policy shocks.

### Contribution to the literature
- Adds to three strands:
  1. International macroeconomic spillovers of U.S. monetary policy (extends beyond country-level VARs and event studies by using firm-level data and analyzing real outcomes such as investment).
  2. Heterogeneous effects of monetary policy across firm characteristics (expands on Li et al. (2020) by including AEs, using quarterly dynamics, and incorporating multiple transmission channels).
  3. Effects of exchange rate fluctuations on firms (analyzes exchange rate channel simultaneously with the firm balance sheet channel and their interactions).
- Situates findings relative to prior studies:
  - Links to studies examining dependence of spillovers on income group, exchange rate regimes, macro fundamentals, and the state of the business cycle.
  - Relates to firm-level domestic studies finding roles for leverage, liquidity, distance-to-default, age, and markups in monetary policy transmission.

*IMF WORKING PAPERS U.S. Monetary Policy Shock Spillovers: Evidence from Firm-Level Data*

### 2.  Data and Empirical Framework

### 2. Data and Empirical Framework

### Data
- Primary data source: S&P Capital IQ (CIQ), providing detailed firm balance sheet and income statement information at quarterly frequency.
- CIQ advantages highlighted:
  - Quarterly frequency suited to identify firm-level responses to high frequency shocks.
  - Contains information on foreign-currency liabilities (covers both bank debt and bond issuance).
- Key limitations and sample construction:
  - Coverage of non-listed firms insufficient in many countries; sample restricted to listed firms (both active and inactive).
  - Do not observe which financial institutions firms borrow from (cannot control for supply-side credit constraints).
  - Data collected on a consolidated basis; only ultimate corporate parents retained.
  - State-owned firms excluded when corporate parent identified as a government body.
  - Firms with negative total assets or total debt in a given year are entirely dropped.
  - Firm-observations with unexpected signs for capital expenditure, net property plant and equipment (NPPE), and revenue are excluded.
  - Observations filtered out if the difference between assets and liabilities is greater than USD 10,000, or if cash & cash equivalents and tangible assets exceed total assets.
  - All variables winsorized at 5 percent to exclude outliers.
- Final sample after filtering: 23,482 firms.
- Sample span and coverage:
  - Time span: 1996Q3 to 2016Q3 (20 years).
  - Countries: 63 countries (29 AEs and 34 EMDEs).
  - Industries: 20 CIQ-defined industries after filtering out the financial, insurance and utilities sectors.
  - Panel is highly unbalanced; country- and filing-requirement-driven disparities in firm coverage.
- Main outcome variable:
  - Investment rate defined as the ratio of capital expenditures to NPPE.
  - Cross-validation with macro data: aggregated firm-level log-investment highly correlated with macro log-investment (R-squared above 0.7); relation remains strong and significant with country fixed effects and when considering growth rates.
- Foreign-currency liability measure:
  - Foreign-currency liability ratio = foreign-currency liabilities to total liabilities.
  - Foreign-currency liabilities computed from CIQ’s Capital Structure module; CIQ collects liability data at the debt instrument level and the authors aggregate by repayment currency.
  - In cases where currency not specified, assume debt denominated in U.S. dollars (robustness check: assume European firms hold debt in euros when currency not available).
  - Kim et al. (2020) find CIQ-based foreign-currency liability ratio consistent with BIS aggregate data.
  - Do not observe firm hedging; note literature suggests foreign exchange hedging not widespread in EMDEs (Chui et al. 2014; Chuaprapaisilp et al. 2018).
- Trade/exposure measure:
  - Baseline uses World Input Output Database (WIOD): cross-country trade data for 56 sectors in 43 countries between 2000 and 2014.
  - Sector-country average export dependence computed as ExportDependence_cs = Exports_cs / Output_cs (average share of exports in sector s and country c over the period).
  - Only 35 out of 63 countries have WIOD data; for remaining 28 countries use average sectoral export dependence ratios from sample.
  - CIQ sector classification in Standard Industry Classification (SIC); WIOD uses NACE REV. 2—manual many-to-one matching applied.
  - Robustness: analysis repeated using firm-level export dependence where available; results qualitatively the same.
- Additional firm-level covariates included in regressions: leverage, liquidity, size, and collateral (definitions and summary statistics provided in Table 1).

### U.S. Monetary Policy Shocks
- Identification approach follows Duval et al. (2021) and Albrizio et al. (2021):
  - Use high-frequency monetary policy surprises: changes in Federal fund futures around FOMC announcements within a 30 minutes window (following Gürkaynak et al. 2005).
  - Identification assumption: responses in financial markets within this interval reflect exclusively monetary policy news.
- Instrument/proxy-SVAR setup:
  - Surprises used in a proxy-SVAR to instrument one-year government bond yield together with industrial production, consumer price index, and a measure of the excess bond premium from Gilchrist and Zakrajsek (2012).
  - Estimate 4-variable proxy-SVAR over 1973M1-2016M8 at monthly frequency to retrieve structural monetary policy shocks (Ramey 2016).
  - Use one-year government bond yield to better capture forward guidance effects compared with the Fed funds rate.
  - Allow for a structural break in VAR coefficients to account for the post zero-lower bound period and the beginning of quantitative easing in November 2008.
  - Aggregate series of structural shocks within each quarter to match firm-level quarterly data.
- Instrument strength and robustness (baseline specification):
  - Baseline uses one-year government bond yield as the policy rate and the three-month ahead Fed fund futures as instrument as this provides the higher first stage F-statistic and above the associated Stock-Yogo critical value for strong instruments.
  - Results robust to alternative combinations of Fed fund futures and government bond yields.

### Empirical Strategy
- Overall approach: three steps to quantify U.S. monetary spillovers on firm outcomes.
- Step 1 — Average (unconditional) effect:
  - Use Jorda (2005) local projections to estimate the average effect of U.S. monetary policy shocks on firm investment.
  - Baseline specification (symbols preserved from source):
    - log Y_{f,t+h} − log Y_{f,t−1} = β_h × i_t^{US} + γ_f + γ_{c s q} + ε_{f,t+h}
    - Dependent variable Y_{f,t} is the investment ratio or revenue of firm f at quarter t.
    - i_t^{US} denotes exogenous U.S. monetary policy shock at time t.
    - γ_f are firm fixed effects; γ_{c s q} are country-sector-quarter dummies.
- Step 2 — Heterogeneous impacts via channels:
  - Semi-parametric approach following Cloyne et al. (2019) and Duval et al. (2021) to flexibly estimate how spillovers vary with firm characteristics without strong functional-form assumptions.
  - Specification (symbols preserved):
    - log Y_{f,t+h} − log Y_{f,t−1} = Σ_{g=1}^{G} β_{g h} × I[X_f ∈ g] × i_t^{US} + ρ_h Z_{f,t−1} + γ_f + γ_{c s t+h} + ε_{f,t+h}
    - I[·] is an indicator equal to one if firm characteristic X_f falls in group g (e.g., average leverage above the 75th percentile = "high leverage"; below 25th percentile = "low leverage").
    - Use average over time of firm characteristics to reduce endogeneity from time-varying responses.
    - γ_{c s t+h} are country-sector-time fixed effects to account for macro shocks and their sectoral differential effects at country level.
    - Z_{f,t−1} are firm-specific characteristics (leverage, liquidity, size and collateral) lagged one period to reduce reverse causality concerns.
  - Estimate Equation (2) separately for each transmission channel and then with all three channels together.
- Step 3 — Interactions across channels:
  - Examine whether differential responses related to one channel vary with other firm characteristics via a two-dimension interaction specification:
    - log Y_{f,t+h} − log Y_{f,t−1} = Σ_{g1=1}^{G1} Σ_{g2=1}^{G2} β_{g1 g2 h} × I[X_{f1} ∈ g1] × I[X_{f2} ∈ g2] × i_t^{US} + ρ_h Z_{f,t−1} + γ_f + γ_{c s t+h} + ε_{f,t+h}
    - Allows assessment of whether, for example, leverage remains an important transmission mechanism conditional on foreign-currency liabilities.
- Estimation details:
  - Equations (1)-(3) estimated by OLS.
  - Standard errors two-way clustered on firm and country-time.  

*IMF Working Paper — Section 2: Data and Empirical Framework*

### 3.  Results

### 3. Results

### 3.1. Average Effects
- A 25-basis points U.S. monetary policy tightening is followed by an economically and statistically significant decline in foreign firms’ investment, with a peak impact of -1.3 percent after two quarters.
- The estimated investment response of foreign firms is comparable to, albeit slightly smaller than, the response of the U.S. firms in the sample.
- A 25-basis points U.S. monetary policy tightening leads to a peak decline in firm revenue of about 0.6 percent after two quarters.
- Robustness highlights:
  - Results hold when removing countries with the highest number of firms among AEs and EMDEs, one at a time.
  - Results are robust to alternative lags of the dependent variable, alternative measures of monetary policy shocks, inclusion of time-varying firm characteristics (leverage, collateral, liquidity, size), and alternative measures of investment (change in capital expenditures; ratio of capital expenditures to total assets).

### 3.2. Spillover Channels and Firm Characteristics
- Method:
  - Firms are classified into low/medium/high exposure groups using the cross-country distribution; firms with an average exposure below (above) the 25th (75th) percentile are classified into the low (high) group.
- Balance sheet channel (leverage):
  - Highly leveraged firms are more responsive to U.S. monetary policy shocks than low-leverage firms.
  - The difference in the one-year-ahead response of investment to a 25-basis points monetary policy shock between the top and bottom quartiles of the firm leverage distribution is about 0.5 percentage point.
- Financial channel of the exchange rate (foreign-currency liability ratio):
  - Firms with higher foreign-currency liability ratios exhibit a larger decline in investment.
  - The magnitude of the differential response between the top and bottom quartiles is about 0.5 percentage point, similar to leverage.
- Trade channel (export dependence of sector):
  - Short term: firms in sectors with higher export dependence are more negatively affected (expenditure reducing channel dominates).
  - Medium term (about two years): the expenditure switching channel dominates and firms in higher export-dependent sectors fare relatively better.
  - Both expenditure reducing and expenditure switching channels are at play, consistent with lagged exchange rate effects on exports.

### 3.3. Robustness Checks
- Joint specification:
  - Estimating a specification including all three channels together yields qualitatively similar and not statistically different results from separate estimations, implying independent and significant roles for each channel.
- Additional firm characteristics:
  - Interactions of monetary policy shocks with firm size, bank debt to total debt ratio, dividend payment status, and firm age amplify spillovers (smaller, younger, lower-dividend-paying firms experience larger declines in investment).
  - Inclusion of these controls does not significantly affect differential impacts of leverage, foreign-currency liability, and trade exposure.
- Alternative classification methods:
  - Using country-specific distributions or median-based low/high groups yields similar results.
  - Classifying firms by characteristics from their first reporting year (instead of sample average) does not change significance of channels.
- Country exclusion:
  - Dropping, one at a time, the six countries with the highest number of firms (Canada, China, Hong Kong, India, Japan, Taiwan) produces results similar to baseline.
- Revenue channel:
  - Trade channel plays a larger role for revenue than for investment; the balance sheet and exchange rate channels play a larger role for investment than for revenue.

### 3.4. Back-of-the-Envelope Calculations
- Approach:
  - Use estimates from the joint specification to calculate each channel’s contribution to the average investment response, following Ciminelli et al. (2020).
  - Formula: aggregate investment response is the weighted average of group-specific investment responses; derive channel contributions assuming 휕K_t3/휕i_t_US = 0 and 휕K_t2/휕i_t_US − 휕K_t3/휕i_t_US = 0 (a lower-bound assumption).
- Findings:
  - The balance sheet channel accounts for the largest share of the total investment response.
  - The exchange rate channel is the next largest contributor.
  - The trade channel contributes less to the aggregate investment response.
  - For firm revenue, the trade channel accounts for the largest share of the total revenue response.
- Caveat:
  - By assuming zero contributions for medium- and low-exposure groups, these calculations likely under-estimate each channel’s true general equilibrium contribution; results should be considered a lower bound.

### 3.5. Interactions Between Channels
- Balance sheet × financial (foreign-currency liability ratio):
  - Both leverage and foreign-currency liability ratio have significant differential impacts on investment irrespective of each other.
  - The adverse effect of high leverage is amplified when foreign-currency exposure is also high.
- Balance sheet × trade:
  - Leverage retains an important role independent of firms’ export dependence.
  - The export dependence ratio is not statistically significant for investment once firms are grouped by leverage, suggesting trade exposure is a less robust investment channel.
  - For firm revenue, trade remains important and statistically significant irrespective of leverage groups.
- Financial × trade:
  - The financial channel (foreign-currency debt share) is significant for firms in sectors with high export dependence.
  - Foreign-currency debt share has no statistically significant role among firms in the low export dependence group.
  - Trade channel is not a significant differential driver of investment after grouping by foreign-currency liability ratio, but trade remains significant for revenue irrespective of foreign-currency liability group.

### 3.6. Extensions
- Leverage interactions with size and liquidity:
  - The impact of leverage is more pronounced for smaller and less liquid firms, consistent with stronger financial frictions for those firms.
- Exchange rate regime:
  - Foreign-currency borrowing has a more pronounced negative impact on investment in response to U.S. monetary policy shocks in countries with more flexible exchange rate regimes.

_Italic: Source — wpiea2022191-print-pdf, section 3. Results (IMF Working Paper)._

### References

### References (wpiea2022191-print-pdf)

### Scope and thematic coverage
- Bibliographic coverage focuses on empirical and theoretical literature on:
  - U.S. monetary policy spillovers and international transmission mechanisms.
  - Financial channels: leverage, foreign-currency liabilities, bank lending, credit spreads, balance-sheet effects.
  - Trade channel and export dependence.
  - Firm-level responses: investment, capital expenditure, liquidity, and corporate balance sheets.
  - Methodological references on impulse-response estimation and local projections.

### Empirical approach and identification (from figure notes)
- Main shock analyzed: a 25-bps U.S. monetary policy shock.
- Impulse responses are estimated from:
  - Equation (1) for average (unconditional) investment response.
  - Equation (2) for differential impulse responses across groups (leverage, foreign-currency liability ratio, export dependence).
  - Equation (3) for two-dimensional group separations (e.g., leverage × FX liability ratio).
  - Equation (4) for comparing multiple channels and for back-of-the-envelope decomposition.
- Graphical conventions and inference:
  - Y-axis expressed in percent.
  - Solid blue line: estimated differential or average impulse response to a 25-bps shock.
  - Dashed blue and red lines: 90% and 68% confidence intervals, respectively.
  - Standard errors: two-way clustered on firms and country-time for some specifications; clustered on firms for others.
  - Fixed effects: firm and sector-country-time fixed effects are commonly controlled; some exercises use country-time fixed effects when variation is only at sector-country level.
  - Time-varying firm controls included in many specifications: leverage, liquidity, size, collateral.

### Key statistics (Table 1: Summary Statistics)
- Investment rate (log)
  - No. of Obs.: 887,901
  - Mean: -3.514
  - Std. Dev.: 1.346
  - 25th Pctile: -4.324
  - Median: -3.366
  - 75th Pctile: -2.571
- Revenue (log USD)
  - No. of Obs.: 1,156,564
  - Mean: 5.728
  - Std. Dev.: 3.182
  - 25th Pctile: 3.786
  - Median: 5.695
  - 75th Pctile: 7.856
- Leverage
  - No. of Obs.: 1,250,768
  - Mean: 0.197
  - Std. Dev.: 0.182
  - 25th Pctile: 0.021
  - Median: 0.162
  - 75th Pctile: 0.324
- Foreign-currency liability ratio (%)
  - No. of Obs.: 708,800
  - Mean: 15.371
  - Std. Dev.: 32.335
  - 25th Pctile: 0
  - Median: 0
  - 75th Pctile: 1.964
- Export dep. ratio (%)
  - No. of Obs.: 2,298,087
  - Mean: 25.024
  - Std. Dev.: 23.36
  - 25th Pctile: 5.752
  - Median: 16.735
  - 75th Pctile: 43.686
- Size (log USD)
  - No. of Obs.: 1,258,486
  - Mean: 4.701
  - Std. Dev.: 2.253
  - 25th Pctile: 3.389
  - Median: 4.818
  - 75th Pctile: 6.200
- Bank debt ratio
  - No. of Obs.: 542,363
  - Mean: 0.763
  - Std. Dev.: 0.292
  - 25th Pctile: 0.597
  - Median: 0.903
  - 75th Pctile: 1
- Liquidity
  - No. of Obs.: 1,224,032
  - Mean: 0.176
  - Std. Dev.: 0.171
  - 25th Pctile: 0.050
  - Median: 0.121
  - 75th Pctile: 0.246
- Collateral
  - No. of Obs.: 1,161,006
  - Mean: 0.513
  - Std. Dev.: 0.237
  - 25th Pctile: 0.325
  - Median: 0.502
  - 75th Pctile: 0.700
- Age
  - No. of Obs.: 2,141,964
  - Mean: 40.593
  - Std. Dev.: 33.863
  - 25th Pctile: 20
  - Median: 30
  - 75th Pctile: 50
- Dividend payment (USD)
  - No. of Obs.: 481,827
  - Mean: 23.201
  - Std. Dev.: 912.921
  - 25th Pctile: 0.027
  - Median: 0.830
  - 75th Pctile: 4.465
- Notes on definitions (as presented):
  - Investment rate = Capital Expenditure/Net Property Plant and Equipment
  - Revenue is in log USD
  - Leverage = Total Debt/Total Asset
  - Foreign-currency Liability Ratio = Total foreign-currency Liabilities/Total Debt
  - Export Dependence Ratio = Exports/Output
  - Size is the log of Total Assets
  - Bank Debt Ratio = Bank Debt/Total Debt
  - Liquidity = (Cash + Short-term Investment)/Total Assets
  - Collateral = Tangible Assets/Total Assets
  - Age = 2019 – Foundation year
  - Dividend payment is in USD

### Aggregate and channel contributions (Table 2: Back-of-the-Envelope Calculations)
- Cumulative Responses (%) to a 25-bps U.S. monetary policy shock (sum of impulse responses over 12 quarters)
  - Unconditional: -5.86
  - Leverage: -4.33
  - Foreign-Currency Liability Ratio: -2.77
  - Export Dependence: -0.18
- Capital Expenditure Share of High Group
  - Leverage: 34.2%
  - Foreign-Currency Liability Ratio: 28.4%
  - Export Dependence: 13.8%
- Contribution to the Average Response (흎1흏푲푡1흏풊풕푼푺)/(흏푲푡흏풊풕푼푺)
  - Leverage: 25.5%
  - Foreign-Currency Liability Ratio: 13.4%
  - Export Dependence: 0.41%
- Note: Cumulative responses are computed as the sum of impulse responses over 12 quarters to a 25- bps U.S. monetary policy shock. The first column (“unconditional”) refers to the estimation of Equation (1).

### Figures and robustness details
- Figures 1–8 present:
  - Figure 1: Average investment response to U.S. monetary policy shocks.
  - Figures 2–4: Role of Leverage (Balance Sheet Channel), Foreign-Currency Liability Ratio (Financial Channel of the Exchange Rate), and Export Dependence (Trade Channel), respectively—each showing differential impulse responses between high and low groups.
  - Figure 5: Comparison of the three channels using Equation (4).
  - Figures 6–8: Interaction exercises separating firms by pairs of dimensions (Leverage × FX liability ratio; Leverage × Export dependence; FX liability ratio × Export dependence) estimated via Equation (3).
- Common controls across figures: firm and sector-country-time fixed effects (or country-time fixed effects where appropriate), time-varying firm characteristics (leverage, liquidity, size, collateral). Standard errors clustered on firms unless otherwise stated.

*References section and accompanying figures/tables from wpiea2022191-print-pdf*

### Annex 1: Mundell-Fleming Framework

### Annex 1: Mundell-Fleming Framework

### Framework and assumptions
- Two-country setup: a small domestic economy (subscript c) and a large foreign economy (United States, superscript *).
- Domestic output Y_c equals the sum of tradable Y_c^T and non-tradable Y_c^N sector outputs:
  - Y_c = Y_c^T + Y_c^N
  - Y_c^N = DD_c^N
  - Y_c^T = DD_c^T + NX
- Domestic demand in both sectors depends positively on demand shifters (ξ_T, ξ_N) and negatively on the domestic interest rate R_c and the nominal exchange rate E_c:
  - DD_c^N = ξ_N − b R_c − s E_c
  - DD_c^T = ξ_T − b R_c − s E_c
- Net exports:
  - NX = f(Y^* − Y_c^T) + d E_c
- U.S. (foreign) economy is closed with:
  - Y^* = DD^* = ξ^* − b R^*
- Exchange rate determination (following Gourinchas (2018)):
  - E_c = g(R^* − R_c) + χ_c, where χ_c = h R^* + γ_c
  - The interest rate differential captures UIP; χ_c captures country-specific risk premium and increases with R^*, amplifying depreciation when R^* rises.
- Wedge between corporate borrowing rate and government bond rate (following Kalemli-Özcan (2019)):
  - R_c = R_pc + χ_c = R_pc + h R^* + γ_c
  - R_pc is the government policy rate; χ_c is the corporate–government wedge; γ_c is domestic country-specific risk premium.
- Parameter roles:
  - b: sensitivity of domestic demand/output to interest rate (balance sheet channel).
  - s: sensitivity to exchange rate (financial channel of exchange rate).
  - d, f, g, h: parameters governing trade and exchange rate pass-through.
- Note: financial spillover vanishes when d = 0 (usual Mundell-Fleming case).

### Closed-form sectoral output (solved expressions)
- Tradable sector output:
  - Y_c^T = (1/(1+f)) [ (ξ_T + f ξ^*) − (b + g(d − s)) R_pc − (b + (s − d)(1 − g)) γ_c − [b h + b f + (s − d)(h + g(1 − h))] R^* ]
- Non-tradable sector output:
  - Y_c^N = [ ξ_N − (b h + s (h + g(1 − h))) R^* − (b − s g) R_pc − (b − s(g − 1)) γ_c ]

### Partial derivatives with respect to U.S. policy rate R^*
- Tradable sector:
  - ∂Y_c^T / ∂R^* = (1/(1+f)) [ − b h  (Balance Sheet Channel)
    − s (h + g(1 − h))  (Financial Channel of the Exchange Rate)
    + [ d (h + g(1 − h)) − b f ]  (Trade Channel) ]
- Non-tradable sector:
  - ∂Y_c^N / ∂R^* = − b h  (Balance Sheet Channel)
    − s (h + g(1 − h))  (Financial Channel of the Exchange Rate)

### Transmission channels and empirical implications
- Balance sheet channel
  - Effect: Negative for both sectors.
  - Heterogeneity: Adverse effects are greater for firms with higher vulnerability b to changing cost of finance.
  - Formal sign conditions:
    - ∂Y_f^T / ∂R^* ∂b < 0
    - ∂Y_f^N / ∂R^* ∂b < 0
  - Empirical proxy: firm leverage (debt-to-assets ratio). Rationale: firms with higher leverage face greater external finance premium during turbulence (higher R^*).

- Financial channel of the exchange rate
  - Effect: Negative for both sectors.
  - Heterogeneity: Adverse effects are greater for firms with higher vulnerability s to exchange rate fluctuations.
  - Formal sign conditions:
    - ∂Y_f^T / ∂R^* ∂s < 0
    - ∂Y_f^N / ∂R^* ∂s < 0
  - Empirical proxy: foreign-currency debt-to-total debt ratio. Rationale: firms with higher foreign-currency debt are more vulnerable to exchange rate depreciation induced by higher U.S. rates.

- Trade channel (additional effect on tradable sector)
  - Term: [ d (h + g(1 − h)) − b f ]  (labeled Trade Channel)
  - Ambiguity: The net sign is ambiguous. It depends on which sub-channel dominates:
    - d (h + g(1 − h))  (Expenditure Switching Channel)
    - b f  (Expenditure Reducing Channel)
  - Empirical test: compare differential effects of U.S. monetary policy shocks across firms/sectors by their trade dependence on the United States.

### Empirical focus and observable outcomes
- Primary outcome of interest: firm investment responses to U.S. monetary policy shocks (sectoral and firm-level heterogeneity).
- Secondary outcome: firm revenue responses (growth in total revenues).
- Identification strategy implied: interact U.S. monetary policy shocks with firm-level vulnerability measures (leverage, foreign-currency debt share) and sectoral trade exposure to test channel-specific predictions.

### Data highlights (sample coverage)
- Table A2.1: Number of firms and observations by country (selected top entries):
  - United States: Number of firms 4,740; Obs. 388,680
  - China: Number of firms 4,077; Obs. 334,314
  - Japan: Number of firms 3,085; Obs. 252,970
  - India: Number of firms 2,672; Obs. 219,104
  - Canada: Number of firms 2,213; Obs. 181,466
  - South Korea: Number of firms 1,747; Obs. 143,254
  - Taiwan: Number of firms 1,693; Obs. 138,826
  - Australia: Number of firms 1,356; Obs. 111,192
  - Hong Kong: Number of firms 1,106; Obs. 90,692
  - United Kingdom: Number of firms 870; Obs. 71,340

- Table A2.2: Number of firms and observations by sector (selected top entries):
  - Materials: Number of Firms 5,433; Obs. 445,506
  - Capital Goods: Number of Firms 4,888; Obs. 400,816
  - Technology Hardware and Equipment: Number of Firms 2,286; Obs. 187,452
  - Consumer Durables and Apparel: Number of Firms 2,032; Obs. 166,624
  - Software and Services: Number of Firms 2,027; Obs. 166,214
  - Pharmaceuticals and Biotechnology: Number of Firms 1,833; Obs. 150,306
  - Food, Beverage and Tobacco: Number of Firms 1,800; Obs. 147,600
  - Energy: Number of Firms 1,714; Obs. 140,548

### Empirical robustness and supplemental analyses reported (figures and tables)
- Correlation of investment between Capital IQ and World Economic Outlook data (Table A2.3):
  - Investment Growth (WEO) correlated with Log Investment USD (CIQ): coefficient 0.942*** (standard error 0.207) in column (2) and 0.867*** (0.211) in column (3).
  - Log Investment USD (WEO) coefficients: 1.173*** (0.0218) in column (1) and 1.350*** (0.0693) in column (4).
  - Observations: 1,107 in all reported columns.
  - R-squared values: 0.717, 0.925, 0.032, 0.101 across columns (1)-(4).
  - Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

- Figures and robustness checks summarized in Annex 3 (unconditional analysis):
  - Average investment responses to a 25-bps U.S. monetary policy shock are presented for: U.S. firms only; full sample dropping largest-country contributors; differential AE vs. EMDE responses (including and excluding China); inclusion of shock lags; alternative shock measures (ED4, FF4, MP3); adding time-varying firm characteristics (leverage, collateral, liquidity, size); alternative investment definitions (log change in capital expenditure; CapEx/Total Assets).
  - Average revenue response to a 25-bps U.S. monetary policy shock is reported (revenue = growth in total revenues).
  - Notes on figures: Y-axis in percent. Standard errors are two-way clustered on firm and country-time in reported estimations.

*Source: Annex 1 and Annex 2 extracts from the PDF chapter "wpiea2022191-print-pdf - Annex 1: Mundell-Fleming Framework".*

### Annex 4: Conditional Analysis

### Annex 4: Conditional Analysis

### Role of Firm Characteristics in Investment and Revenue Responses
- Figures A4.1–A4.2 investigate how monetary policy shocks interact with firm characteristics (size, age, bank debt, leverage, liquidity, collateral, foreign-currency liability ratio, export dependence).
- Firms are typically split into three categories (low/medium/high) using the 25th and 75th percentiles of the distribution across countries, except where noted otherwise.
- Alternative splits used: two-category split based on the median; country-specific thresholds using the 25th and 75th percentiles within each country; classification based on initial levels (first year observed).
- Key methodological features across figures:
  - Shock size referenced: 25-bps U.S. monetary policy shock.
  - Differential impulse responses are typically presented as (high − low) group responses (e.g., βHigh − βLow).
  - Confidence intervals displayed: 90% and 68%.
  - Standard errors are clustered on firms.
  - Fixed effects and controls vary by exercise:
    - Firm and sector-country-time fixed effects (baseline for many figures).
    - Country-time fixed effects only (used where sector-country variation cannot be exploited, e.g., some trade-channel and initial-level exercises).
    - Time-varying firm characteristics controlled for in many exercises: leverage, liquidity, size, collateral.

### Channel-Specific and Robustness Analyses
- Figures A4.2–A4.6 show the role of each channel controlling for other characteristics and performing robustness checks:
  - Control for interactions of size, age and bank debt with the monetary policy shock.
  - Use country-specific thresholds (25th and 75th percentiles within countries) in Figure A4.3.
  - Use median-based two-group splits in Figure A4.4.
  - Classification based on initial levels of leverage, foreign-currency liability and export dependence in Figure A4.5 (uses only country-time fixed effects and controls for time-varying firm characteristics).
  - Dropping each large country separately to assess sensitivity in Figure A4.6.
- Across these exercises, the presentation emphasizes the differential responses between high and low groups, and confidence intervals (90% and 68%) around the impulse responses.

### Combined and Interaction Exercises (Revenue)
- Figure A4.7: All channels jointly estimated using Equation (4) with firm revenue growth as dependent variable.
  - Uses country-time fixed effects (instead of country-sector-time) because trade-channel variation is at sector-country level.
  - Differential impulse responses (βHigh − βLow) against a 25-bps U.S. monetary policy shock are shown with 90% and 68% confidence intervals.
  - Controls include time-varying firm characteristics (leverage, liquidity, size, collateral); standard errors clustered on firms.

- Pairwise interaction exercises using Equation (3) with log change in revenues:
  - Figure A4.8: Leverage × Foreign-Currency Liability Ratio (low/medium/high for each dimension).
    - Upper row: differential responses between high and low leverage within high and low FX liability-ratio firms.
    - Lower row: differential responses between high and low FX liability-ratio within high and low leverage firms.
  - Figure A4.9: Leverage × Export Dependence (low/medium/high for each dimension).
    - Upper row: differential responses between high and low leverage within high and low export-dependence firms.
    - Lower row: differential responses between high and low export dependence within high and low leverage firms.
  - Figure A4.10: Foreign-Currency Liability Ratio × Export Dependence.
    - Upper row: differential responses between high and low FX-liability groups within high and low export-dependence firms.
    - Lower row: differential responses between high and low export-dependence groups within high and low FX-liability firms.
  - Figure A4.11: Leverage × Size and Leverage × Liquidity interactions.
    - Upper row: differential leverage responses within smaller and larger firms.
    - Lower row: differential leverage responses among low and high liquidity firms.
  - Figure A4.12: Foreign-Currency Liability Ratio × Exchange Rate Regime.
    - Firms grouped by FX liability ratio and country exchange rate regime (information from IMF ARAER Database).
    - Differential responses between high and low FX-liability firms within fixed and flexible exchange rate regimes.
  - All pairwise interaction exercises:
    - Use 90% and 68% confidence intervals.
    - Control for time-varying firm characteristics (leverage, liquidity, size, collateral).
    - Standard errors clustered on firms.
    - Results are expressed as responses to a 25-bps U.S. monetary policy shock.

### Back-of-the-Envelope Calculations (Revenue) — Table A4.1
- Calculation inputs and results (from estimation of Equation (4); cumulative responses are the sum of impulse responses over 12 quarters):
  - Columns/rows reported in table:
    - Unconditional cumulative response: -3.77
    - Leverage cumulative response: -0.25
    - Foreign-Currency Liability Ratio cumulative response: -1.74
    - Export Dependence cumulative response: -2.99
  - Capital Expenditure Share of High Group:
    - Unconditional: 27.9%
    - Foreign-Currency Liability Ratio: 19.8%
    - Export Dependence: 17.1%
  - Contribution to the Average Response ((τ1 Kt_i_US)/(τ Kt_i_US)):
    - Unconditional: 1.8%
    - Foreign-Currency Liability Ratio: 9.2%
    - Export Dependence: 13.5%
- Notes:
  - The “Unconditional” column refers to the estimation from Equation (1).
  - Cumulative responses sum impulse responses over 12 quarters.

*U.S. Monetary Policy Shock Spillovers: Evidence from Firm-Level Data — Annex 4: Conditional Analysis (Working Paper No. WP/2022/191)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022191-print-pdf.pdf_