## ANNEX 2: ADDITIONAL RESULTS

## Source details

**Canonical URL:** [ANNEX 2: ADDITIONAL RESULTS](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022211-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022211-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022211-print-pdf.pdf.json)

---

### I. Introduction and motivation
- Major headwinds from the COVID-19 pandemic:
  - Cumulative per capita income losses over 2020–23, compared to pre-pandemic projections, are equivalent to about 2½ percent of 2019 per capita GDP for the world as a whole and over 4 percent in emerging markets and developing economies.
- Corporate debt trends and concern:
  - Corporate debt as a share of GDP increased significantly in the decade after the global financial crisis and rose sharply in 2020 with the pandemic.
  - Higher corporate debt may amplify scarring effects of recessions by constraining firms’ investment.
- Contribution:
  - Uses difference-in-difference setup and Jordà (2005) local projection methods to estimate dynamic scarring effects of recessions on firm-level investment conditional on corporate debt.
  - Employs a large quarterly firm-level sample across 75 advanced and emerging market and developing economies.
  - Explores channels, highlighting the role of financing constraints.

### II. Data and sample
- Source and coverage:
  - Source: S&P Capital IQ.
  - Sample restricted to 2001Q1 onwards; estimation panel covers 2001Q1 to 2020Q4.
  - Sample: 75 countries; panel size: over 24,000 firms and more than 800,000 observations.
  - Non-financial corporations only; real estate and insurance excluded.
- Variable construction and processing:
  - Investment measure: IQ_CAPEX-2021.
  - Leverage: IQ_TOTAL_DEBT-4173 divided by IQ_TOTAL_ASSETS - 1007.
  - High/low debt classification: firm-average leverage over the entire sample above/below industry median (baseline).
  - Winsorization: firm-level variables winsorized at the 1st and 99th percentiles; average firm age and leverage winsorized at the 5th and 95th percentiles (baseline); alternative winsorizations tested at 0.05 and 5 percent tails.
- Recession measures:
  - Baseline recession: start of a technical recession defined by two consecutive quarters of negative GDP growth using Haver Analytics and WEO where needed.
  - Total recessions identified: 231 for advanced economies and 336 for emerging market and developing economies.
  - Alternative recession measures: continuous GDP growth (sign inverted), peak-to-trough GDP using Harding-Pagan (2006), banking/financial crises dummies from Global Crises Data and Reinhart and Rogoff (2009).

### III. Empirical methodology
- Core estimation approach:
  - Impulse response functions estimated by Jordà (2005) local projection method for horizons k up to 12 quarters.
  - Step 1: unconditional average effect of recession dummy on log differences of firm capital expenditure with firm-quarter dummies and country-sector fixed effects; include leads of recession variable.
  - Step 2: heterogeneous effects by interacting recession dummy with a time-invariant firm high-debt dummy and including country-sector-time fixed effects.
  - Step 3: triple-interaction specifications to test dependence of the high-debt differential on firm characteristics X (lagged, averaged over four quarters), e.g., size, profitability, share of short-term debt.
- Estimation details:
  - Equations estimated by OLS.
  - Standard errors two-way clustered on firm and country-time.
  - Controls include leads and lags of dependent variables and rich fixed effects (firm-quarter and country-sector-time in interactive specifications).

### IV. Main results
- Unconditional scarring of recessions on investment:
  - The average recession in the sample is associated with a reduction in the level of firm investment by 30 percent four quarters after a recession and by about 15 percent 12 quarters after the recession.
- Heterogeneity by corporate debt:
  - Differential decline for high-debt firms (above-industry-median leverage) relative to medium-to-low-debt firms:
    - About 2 percent larger decline at four quarters after a recession.
    - About 5 percent larger decline at 12 quarters after a recession.
  - These differential effects are statistically significant and economically sizable.
- Contribution of leverage to aggregate scarring (back-of-the-envelope):
  - Firms’ leverage accounts for about 7 percent of the short-term (four quarters ahead) response of investment to recessions.
  - Firms’ leverage accounts for about 28 percent of the medium-term (12 quarters ahead) response of investment to recessions.
- Channels:
  - Debt amplifies scarring particularly for firms that are credit constrained: small firms, less profitable firms, and firms with a high share of short-term debt are more affected, consistent with difficulties rolling over or raising new funds.

### V. Robustness checks and sensitivity
- Sample sensitivity:
  - Excluding 2009 (Global Financial Crisis peak) and excluding 2020 (COVID-19): the differential response of investment for firms with high corporate debt is similar to, and not statistically different from, the baseline.
  - Leave-one-out country estimations (excluding one of the 10 countries with most observations at a time) and excluding one region at a time show baseline results are not driven by specific countries or regions.
  - Sector exclusions (excluding one 2-digit sector at a time) yield results almost unchanged relative to the full sample.
  - Winsorizing at 0.05 and 5 percent tails of the dependent variable distribution produces results similar to the baseline.
- Alternative dependent variables and recession definitions:
  - Using ratio of capital expenditure to lagged assets (IQ_CAPEX-2021 / IQ_NPPE - 1004) and log of revenue confirm that negative effects of recessions are amplified in firms with high corporate debt.
  - Using continuous GDP growth (sign inverted), peak-to-trough changes (Harding-Pagan), and financial/banking crisis dummies confirm main findings.
- Alternative high-debt definitions:
  - Tested alternatives: above the country-specific industry median; above the income-group-specific industry median; excluding recession years from average-debt computation; above the 4th quartile of industry distribution; two-dummy specification contrasting firms above 4th quartile and below 2nd quartile.
  - Results qualitatively similar: effect of recessions on investment is always larger for firms with higher debt.
- Additional controls and macro interactions:
  - Adding firm characteristics (total assets, ROA, liquidity, age), financial stress (Ahir et al. 2022), World Uncertainty Index, inflation, and fiscal policy controls — individually and jointly — does not change the main recession × corporate debt interaction result.
  - Financial stress and uncertainty have statistically significant unconditional effects on investment; the interaction of financial stress with corporate debt is not always statistically significant.
  - Interaction between inflation and corporate debt is negative and statistically significant for some horizons.
  - Time-varying fiscal policy countercyclicality shows no significant effect of fiscal policy on investment through corporate debt.
- Robustness to functional form:
  - Smooth transition function (Granger and Teräsvirta, 1993) transformation of size and ROA before estimating triple interactions yields robust and generally more precisely estimated results.

### VI. Non-linearity and heterogeneous effects (triple interactions)
- Triple-interaction findings (equation 3):
  - Firm size (log of total assets):
    - Triple interaction coefficient is positive and significant. Larger high-debt firms experience less scarring in investment after recessions than smaller high-debt firms.
  - Profitability (ROA):
    - Triple interaction coefficient is positive and significant at longer horizons. More profitable high-debt firms see less scarring than less profitable high-debt firms.
  - Debt maturity (share of short-term debt):
    - Triple interaction coefficient is negative. Firms with shorter maturity structures of debt experience more scarring compared to those with longer maturities.
  - Age and liquidity:
    - Triple interactions with age and liquidity have signs consistent with size and profitability results (larger differential for younger and less liquid firms) but are not always precisely estimated.

### VII. Key quantitative summary and statistics
- Sample and data processing:
  - Sample: 75 countries.
  - Panel for estimation: over 24,000 firms; period 2001Q1 to 2020Q4; more than 800,000 observations.
  - Winsorization: 1st and 99th percentiles for most variables; average firm age and leverage at 5th and 95th percentiles (baseline).
- Investment scarring magnitudes:
  - Average recession → investment down 30 percent at 4 quarters; down about 15 percent at 12 quarters.
- Debt-amplified differential:
  - High-debt vs medium-to-low-debt firms: additional decline ~2 percent at 4 quarters; ~5 percent at 12 quarters.
- Contribution of leverage to aggregate response (Table 1 selected horizons and values):
  - Horizon 1: Unconditional impact -0.071; Impact via leverage -0.006; Contribution 7.2%  
  - Horizon 2: Unconditional impact -0.188; Impact via leverage -0.010; Contribution 4.5%  
  - Horizon 3: Unconditional impact -0.248; Impact via leverage -0.025; Contribution 8.7%  
  - Horizon 4: Unconditional impact -0.304; Impact via leverage -0.024; Contribution 6.8%  
  - Horizon 5: Unconditional impact -0.300; Impact via leverage -0.039; Contribution 11.0%  
  - Horizon 6: Unconditional impact -0.374; Impact via leverage -0.042; Contribution 9.7%  
  - Horizon 7: Unconditional impact -0.344; Impact via leverage -0.042; Contribution 10.3%  
  - Horizon 8: Unconditional impact -0.303; Impact via leverage -0.041; Contribution 11.4%  
  - Horizon 9: Unconditional impact -0.235; Impact via leverage -0.067; Contribution 24.4%  
  - Horizon 10: Unconditional impact -0.224; Impact via leverage -0.047; Contribution 18.1%  
  - Horizon 11: Unconditional impact -0.208; Impact via leverage -0.059; Contribution 24.3%  
  - Horizon 12: Unconditional impact -0.153; Impact via leverage -0.051; Contribution 28.2%
- Recession counts:
  - 231 recessions in advanced economies; 336 recessions in emerging market and developing economies.
- Key variable identifiers:
  - Investment: IQ_CAPEX-2021.
  - Total debt: IQ_TOTAL_DEBT-4173.
  - Total assets: IQ_TOTAL_ASSETS - 1007.
  - Return on assets: IQ_RETURN_ASSETS - 4178.
  - Revenues: IQ_TOTAL_REV-Line - 28.
  - Net property, plant and equipment: IQ_NPPE - 1004.
  - Current assets: IQ_TOTAL_CA - 1008; current liabilities: IQ_TOTAL_CL - 1009.

### VIII. Conclusions and policy implications
- Main conclusions:
  - Recessionary shocks during 2001-2020 produced significant and persistent drops in investment, larger for firms with higher debt; the leverage effect on investment response to recessions is statistically significant and robust across sensitivity checks.
  - Effects are larger for smaller and less profitable firms; firms with higher shares of short-term debt face greater rollover difficulty and larger investment scarring.
  - Back-of-the-envelope estimate: firms’ debt accounts for at least 28 percent of the average medium-term response of investment to recessions.
- Policy recommendations:
  - Adopt improved policy regimes for restructuring unviable firms.
  - Avoid firms’ zombification.
  - Reduce overall leverage at the country-level.
  - Promote reallocation of capital and labor toward more productive firms.

*Source: ANNEX 2: ADDITIONAL RESULTS, wpiea2022211-print-pdf — https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022211-print-pdf.pdf*

### REFERENCES .............................................................................................................

### wpiea2022211-print-pdf - REFERENCES .............................................................................................................

### REFERENCES
- REFERENCES ................................................................................................................................................ 21

### TABLES AND FIGURES
- TABLES AND FIGURES ................................................................................................................................... 23

### ANNEX 1: DATA
- ANNEX 1: DATA ............................................................................................................................................ 31

*Source: wpiea2022211-print-pdf — https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022211-print-pdf.pdf*

### ANNEX 2: ADDITIONAL RESULTS ............................................................................................

### ANNEX 2: ADDITIONAL RESULTS

### I. Introduction and motivation
- Major headwinds from the COVID-19 pandemic have had large economic consequences:
  - Cumulative per capita income losses over 2020–23, compared to pre-pandemic projections, are equivalent to about 2½ percent of 2019 per capita GDP for the world as a whole and over 4 percent in emerging markets and developing economies.
- Corporate debt trends:
  - Corporate debt as a share of GDP increased significantly in the decade after the global financial crisis and rose sharply in 2020 with the pandemic.
  - Concern: higher corporate debt may amplify the scarring effects of recessions by constraining firms’ investment.
- Contribution of the paper:
  - Uses difference-in-difference setup and Jordà (2005) local projection methods to estimate dynamic scarring effects of recessions on firm-level investment conditional on corporate debt.
  - Employs a large quarterly firm-level sample across 75 advanced and emerging market and developing economies.
  - Explores channels, highlighting the role of financing constraints.

### II. Data: firm-level, recessions, and key variables
- Firm-level database and sample:
  - Source: S&P Capital IQ.
  - Sample restricted to 2001Q1 onwards and to advanced and emerging and developing economies, leaving 75 countries.
  - Data frequency: quarterly; panel described later covers the period 2001Q1 to 2020Q4 for estimation.
  - Panel size: over 24,000 firms and more than 800,000 observations.
  - Non-financial corporations only; real estate and insurance excluded.
  - Winsorization: all firm-level variables winsorized at the 1st and 99th percentiles; average firm age and leverage winsorized at the 5th and 95th percentiles.
- Investment and leverage measures:
  - Investment measure: Capital expenditures (IQ_CAPEX-2021).
  - Leverage: ratio of total debt (IQ_TOTAL_DEBT-4173) to total assets (IQ_TOTAL_ASSETS - 1007).
  - High/low debt classification: firm-average leverage over the entire sample above/below industry median.
- Other firm characteristics (Capital IQ codes):
  - Total Assets (IQ_TOTAL_ASSETS - 1007).
  - Return on Assets (IQ_RETURN_ASSETS - 4178).
  - Liquidity ratio: (IQ_TOTAL_CA - 1008 minus IQ_TOTAL_CL - 1009) divided by IQ_TOTAL_ASSETS - 1007.
  - Firm age (IQ_YEAR_FOUNDED).
  - Capital expenditure to lagged assets (IQ_CAPEX-2021 / IQ_NPPE - 1004).
  - Revenues (IQ_TOTAL_REV-Line - 28).
  - Firm size: logarithm of total assets.
- Recession measures and counts:
  - Baseline recession: start of a technical recession defined by two consecutive quarters of negative GDP growth using Haver Analytics and WEO where needed.
  - Total recessions identified: 231 for advanced economies and 336 for emerging market and developing economies.
  - Alternative recession dummies used: banking/financial crises dummies from Global Crises Data and peak-to-trough GDP using Harding-Pagan (2006) algorithm.

### III. Empirical methodology
- Core approach:
  - Uses Jordà (2005) local projection method to estimate short- and medium-term effects of recessions on log differences of firm capital expenditure.
  - Step 1: Estimate unconditional average effect of recession dummy (beginning of technical recession) on firm investment with firm-quarter dummies and country-sector fixed effects; include leads of the recession variable.
  - Step 2: Estimate heterogeneous effects by interacting recession dummy with a time-invariant firm high-debt dummy (above-industry-median average leverage) and include country-sector-time fixed effects to control for macro shocks and sector-specific country shocks.
  - Step 3: Triple-interaction specifications to test whether the differential impact for high-debt firms depends on firm characteristics X (lagged, averaged over four quarters), e.g., size, profitability, share of short-term debt.
- Estimation details:
  - Equations (1)-(3) estimated by OLS.
  - Standard errors two-way clustered on firm and country-time.
  - Controls include leads and lags of dependent variables and a rich set of fixed effects to mitigate confounding and reverse causality.

### IV. Main results
- Unconditional scarring of recessions on investment:
  - The average recession in the sample is associated with a reduction in the level of firm investment by 30 percent four quarters after a recession and by about 15 percent 12 quarters after the recession.
- Heterogeneity by corporate debt:
  - Differential decline for high-debt firms (above-industry-median leverage) relative to medium-to-low-debt firms:
    - About 2 percent larger decline at four quarters after a recession.
    - About 5 percent larger decline at 12 quarters after a recession.
  - These differential effects are statistically significant and economically sizable.
- Contribution of leverage to aggregate scarring (back-of-the-envelope):
  - Firms’ leverage accounts for about 7 percent of the short-term (four quarters ahead) response of investment to recessions.
  - Firms’ leverage accounts for about 28 percent of the medium-term (12 quarters ahead) response of investment to recessions.
- Channels and heterogeneity:
  - Debt amplifies scarring particularly for firms that are credit constrained:
    - Small firms, less profitable firms, and firms with a high share of short-term debt are more affected, consistent with difficulties rolling over or raising new funds.
- Robustness:
  - Results robust to alternative subsamples (across countries and times), alternative recession measures and debt definitions, different functional forms, additional firm controls (total assets, liquidity, ROA, age), and extensive fixed effects.

### V. Robustness checks (overview)
- Three classes of robustness checks:
  - i) Changing the sample (e.g., excluding major episodes such as 2009 GFC outcomes and 2020 COVID-19).
  - ii) Alternative definitions of key regressors and explanatory variables (recession and corporate debt dummies).
  - iii) Expanding the set of control variables and interactions to address potential confounding factors correlated with the recession × corporate debt interaction.
- Additional notes on cross-validation:
  - Aggregate validation: country-level aggregates of firm CAPEX (winsorized) correlate strongly with WEO private gross fixed capital formation (R-squared above 0.7) and remain strongly related when controlling for country fixed effects and when considering growth rates.

### Key statistics and exact measures (selected)
- Cumulative per capita income losses over 2020–23 vs pre-pandemic projections:
  - world: about 2½ percent of 2019 per capita GDP.
  - emerging markets and developing economies: over 4 percent of 2019 per capita GDP.
- Investment scarring magnitudes:
  - Average recession → investment down 30 percent at 4 quarters; down about 15 percent at 12 quarters.
- Debt-amplified differential:
  - High-debt vs medium-to-low-debt firms: additional decline ~2 percent at 4 quarters; ~5 percent at 12 quarters.
- Contribution of leverage to aggregate investment response:
  - 7 percent at 4 quarters; 28 percent at 12 quarters (back-of-the-envelope calculation).
- Sample and data processing:
  - Sample: 75 countries; panel for estimation: over 24,000 firms; period 2001Q1 to 2020Q4; more than 800,000 observations.
  - Winsorization: 1st and 99th percentiles (most variables); average firm age and leverage at 5th and 95th percentiles.
- Key variable identifiers:
  - Investment: IQ_CAPEX-2021.
  - Total debt: IQ_TOTAL_DEBT-4173.
  - Total assets: IQ_TOTAL_ASSETS - 1007.
  - Return on assets: IQ_RETURN_ASSETS - 4178.
  - Revenues: IQ_TOTAL_REV-Line - 28.
  - Net property, plant and equipment: IQ_NPPE - 1004.
  - Current assets: IQ_TOTAL_CA - 1008; current liabilities: IQ_TOTAL_CL - 1009.
- Recession counts:
  - 231 recessions in advanced economies; 336 recessions in emerging market and developing economies.

*Source: ANNEX 2: ADDITIONAL RESULTS, wpiea2022211-print-pdf*

### Annex 2 show that the differential response of investment for firm with high corporate

### Annex 2 show that the differential response of investment for firm with high corporate debt obtained excluding these two years is similar to, and not statistically different from, the baseline.

### Robustness checks and sample sensitivity
- Excluding the two specified years: the differential response of investment for firms with high corporate debt is similar to, and not statistically different from, the baseline.
- Leave-one-out country and region estimations: re-estimating equation (2) by excluding one country at a time and one region at a time shows baseline results are not driven by specific countries (Figure A2.2 and A2.3 of Annex 2).
- Sector exclusions: repeating the analysis by excluding one 2-digit sector at a time yields results almost unchanged relative to the full sample (Figure A2.4 of Annex 2).
- Influence of extreme observations: winsorizing 0.05 and 5 percent tails of the dependent variable distribution produces results similar to the baseline (Figure A2.5).

### Alternative dependent variables and recession definitions
- Alternative dependent variables:
  - Ratio of capital expenditure to total lagged assets.
  - Log of revenue.
  - These alternative specifications confirm that the negative effects of recessions tend to be amplified in firms with high corporate debt (Figure A2.6 of Annex 2).
- Alternative recession measures considered:
  - Continuous GDP growth (with the sign inverted).
  - Peak-to-trough changes in GDP growth.
  - Dummy for financial (banking, debt and currency) crises.
  - Dummy for banking crises.
  - Results with these alternatives confirm the main findings (Figure A2.7).

### Definitions and thresholds for high corporate debt
- Baseline high-debt definition: firm average level of debt over the entire sample above the industry median (prevents classification switching).
- Alternatives tested:
  - Above the country-specific industry median.
  - Above the income-group-specific industry median.
  - Excluding recession years from the computation of firms’ average corporate debt (Figure 4).
  - Above the 4th quartile of the industry distribution (results qualitatively similar; Figure A2.9).
  - Two-dummy specification: interaction terms for firms above the 4th quartile (high-debt) and firms below the 2nd quartile (low-debt); coefficients capture marginal effects relative to firms between the 2nd and 4th quartiles. Results show effect of recessions on investment is always larger for firms with higher debt (Figure A2.10).
- The baseline approach ensures treated and untreated firms remain the same over time—a key assumption for difference-in-difference validity.

### Additional control variables and macro interactions
- Adding firm characteristics (total assets, ROA, liquidity, age) to baseline regressions does not change baseline results (Figure A2.11).
- Baseline includes sector-country-time fixed effects and firm fixed effects to control macro shocks and time-invariant firm traits.
- Financial stress:
  - Augmenting equation (2) with interaction between Ahir et al. (2022) financial stress indicator and firms’ corporate debt dummy: effect of recessions on investment through corporate debt remains of expected sign and statistically significant, though point estimates are smaller in short term (top-left panel of Figure 6). Financial stress itself has a statistically significant effect on investment (Figure A2.12, top-left panel).
  - Interaction between financial stress and corporate debt dummy is not statistically significant (Figure A2.13, top-right panel).
- Uncertainty:
  - Including interaction between World Uncertainty Index (Ahir et al. 2022) and corporate debt dummy: recession effects on investment remain similar and not statistically different from baseline (top-left panel of Figure 6).
  - Investment falls after uncertainty shocks, more so for firms with higher debt (Figure A2.12, top-right panel; Figure A2.13, top-right panel).
- Inflation:
  - Including interaction between inflation and corporate debt dummy: recession effect via corporate debt is unchanged; interaction between inflation and corporate debt is negative and statistically significant for some horizons (middle-left panel of Figure 6; Figure A2.13, bottom-left panel).
- Fiscal policy:
  - Adding fiscal policy controls does not change results (middle-right panel of Figure 6); no significant effect of fiscal policy on investment through corporate debt (Figure A2.14, bottom-right panel).
- Joint inclusion:
  - Results robust when financial stress, uncertainty, inflation, and fiscal policy controls are included simultaneously (bottom-left panel of Figure 5).
- Firm-characteristic interactions:
  - Adding interactions between recessions and dummies for total assets, ROA, liquidity, and age does not alter the main finding that corporate debt significantly affects the investment response to recessions (Figure 6).
  - Effects of recessions on investment tend to be smaller for firms with higher total assets, ROA, liquidity, and older firms, though these are not always precisely estimated (Figure A2.14).

### Non-linearity and heterogeneous effects
- Sample and method: local projection methods (Jordà 2005) applied to a panel of more than 24,000 firms over 2001-2020.
- Triple-interaction tests (equation 3):
  - Firm size (log of total assets):
    - Triple interaction coefficient is positive and significant (Figure 7, Panel A). Larger high-debt firms experience less scarring in investment after recessions than smaller high-debt firms.
  - Profitability (ROA):
    - Triple interaction coefficient is positive and significant at longer horizons (Figure 7, Panel B). More profitable high-debt firms see less scarring than less profitable high-debt firms.
  - Debt maturity (share of short-term debt):
    - Triple interaction coefficient is negative (Figure 7, Panel C). Firms with shorter maturity structures of debt experience more scarring compared to those with longer maturities.
- Other variables:
  - Triple interactions with age and liquidity have signs consistent with size and profitability results (larger differential for younger and less liquid firms) but are not precisely estimated.
- Robustness: smooth transition function (Granger and Terävistra, 1993) transforming size and ROA before estimating triple interactions yields robust and generally more precisely estimated results (Annex Figure A2.15).

### Key quantitative summary and economic magnitude
- Panel and period: more than 24,000 firms, 2001-2020.
- Back-of-the-envelope estimate: firms’ debt accounts for at least 28 percent of the average medium-term response of investment to recessions.

### Conclusions and policy implications
- Main finding: recessionary shocks during 2001-2020 produced significant and persistent drops in investment, larger for firms with higher debt; the leverage effect on investment response to recessions is statistically significant and robust across sensitivity checks.
- Heterogeneity: effects are larger for smaller and less profitable firms; firms with higher shares of short-term debt face greater rollover difficulty and larger investment scarring.
- Policy recommendations:
  - Adopt improved policy regimes for restructuring unviable firms.
  - Avoid firms’ zombification.
  - Reduce overall leverage at the country-level.
  - Promote reallocation of capital and labor toward more productive firms.

*International Monetary Fund*

### REFERENCES

### REFERENCES

### Bibliographic sources
- Aghion, P., Hemous, D., and Kharroubi, E. (2014). “Cyclical fiscal policy, credit constraints, and industry growth”. Journal of Monetary Economics, 62, 41-58.  
- Ahir, H., Bloom, N., and Furceri, D., 2022. "The World Uncertainty Index," NBER Working Papers 29763, National Bureau of Economic Research.  
- Albuquerque, Bruno. (2021). “Corporate Debt Booms, Financial Constraints and the Investment Nexus.” Bank of England Working Paper, 935.  
- Arbatli-Saxegaard, E., Firat, M., Furceri, D., and Verrier, J., 2022. “U.S. monetary policy spillovers: Evidence from firm-level data”. International Monetary Fund Working Paper, 2022/191.  
- Blickle, K. S., and Santos, J. A. (2020). “The Costs of Corporate Debt Overhang Following the COVID-19 Outbreak”. Federal Reserve Bank of New York, 20201201.  
- Buera, F. and Karmakar, S. (2019), “Real Effects of Financial Distress: The Role of Heterogeneity“, Bank of England Working Papers, 814.  
- Campello, M., Graham, J. R. and Harvey, C. R. (2010), “The Real Effects of Financial Constraints: Evidence from a Financial Crisis”, Journal of Financial Economics, 97(3), 470–487.  
- Choi, S., Furceri, D., Huang, Y., & Loungani, P. (2018). “Aggregate uncertainty and sectoral productivity growth: The role of credit constraints”. Journal of International Money and Finance, 88, 314-330.  
- Choi, S., Furceri, D., Loungani, P., and Shim, M. (2022). “Inflation anchoring and growth: The role of credit constraints”. Journal of Economic Dynamics and Control, 134, 104279.  
- Ciminelli, G., Duval, R., and Furceri, D. (2018). “Employment protection deregulation and labor shares in advanced economies”. The Review of Economics and Statistics, 1-44.  
- Cloyne, J., Huber, K., Ilzetzki, E., and Kleven, H. (2019). “The effect of house prices on household borrowing: a new approach”. American Economic Review, 109(6), 2104-36.  
- Dell Ariccia, G., Detragiache, E., and Rajan, R., (2008). “The real effect of banking crises”. Journal of Financial Intermediation, 17(1), 89-112.  
- Dell”Ariccia, G., Igan, D., Laeven, L. and Tong, H. (2016), “Credit Booms and Macrofinancial Stability”, Economic Policy 31(86), 299–355.  
- Duval, R., Hong, G. H., and Timmer, Y. (2020). “Financial frictions and the great productivity slowdown”. The Review of Financial Studies, 33(2), 475-503.  
- Furceri, D., and Mourougane, A. (2012). “The effect of financial crises on potential output: New empirical evidence from OECD countries”. Journal of Macroeconomics, 34(3), 822-832.  
- Ghosal, V., and Loungani, P. (2000). “The differential impact of uncertainty on investment in small and large businesses”. Review of Economics and Statistics, 82(2), 338-343.  
- Granger C.W.J., Teräsvirta T. (1993). “Modelling Nonlinear Economic Relationships”. Oxford University Press New York.  
- Grigorian, D., and Raei, A. (2010). “Government Involvement in Corporate Debt Restructuring: Case Studies from the Great Recession”. IMF Working Paper No. 10/260, November.  
- Harding, D., and Pagan A. (2006). “Synchronization of cycles”. Journal of Econometrics, 132 (1), 59-79.  
- Hennessy, C. A., Levy, A. and Whited, T. M. (2007), “Testing Q Theory with Financing Frictions”, Journal of Financial Economics 83(3), 691–717.  
- International Monetary Fund. 2022. “World Economic Outlook: War Sets Back the Global Recovery”. Washington, DC, April.  
- Jordà, O. (2005). “Estimation and Inference of Impulse Responses by Local Projections” American Economic Review, vol. 95(1), pages 161-182, March.  
- Jordà, O., Schularick, M. and Taylor, A. M. (2013), “When Credit Bites Back”, Journal of Money, Credit and Banking 45(s2), 3–28.  
- Kalemli-Ozcan, S., Laeven, L. and Moreno, D. (2019), “Debt Overhang, Rollover Risk, and Corporate Investment: Evidence from the European Crisis”, ECB Working Paper Series 2241, European Central Bank.  
- Kim, M., Mano, R.C. and Mrkaic, M., 2020. “Do FX interventions lead to higher FX debt? Evidence from firm-level data”. International Monetary Fund Working Paper, 20/197.  
- Myers, S.C. (1977), “Determinants of Corporate Borrowing”, Journal of Financial Economics 5(2), 147–175.  
- Reinhart, C. M., and Rogoff, K.S. (2009). “The Aftermath of Financial Crises" American Economic Review, 99 (2): 466-72.  
- Teulings, C and Zubanov, N. 2014. “Is Economic Recovery a Myth? Robust Estimation of Impulse Responses”. Journal of Applied Econometrics. Vol 29(3).

### Notes on methodological sources cited
- Local projection methods following Jordà (2005) are repeatedly used to estimate impulse response functions for firm-level quarterly data across multiple figures and robustness checks.
- Harding and Pagan (2006) used for peak-to-trough period identification.
- Reinhart and Rogoff (2009) used for banking crisis dating (converted to quarterly).

---

### TABLES AND FIGURES: Key empirical findings and statistics

### Table 1 — Contribution of leverage to the average investment response following recessions
- Horizon, Unconditional impact, Impact via leverage, Contribution at each horizon:
  - 1: Unconditional impact -0.071; Impact via leverage -0.006; Contribution at each horizon 7.2%  
  - 2: Unconditional impact -0.188; Impact via leverage -0.010; Contribution at each horizon 4.5%  
  - 3: Unconditional impact -0.248; Impact via leverage -0.025; Contribution at each horizon 8.7%  
  - 4: Unconditional impact -0.304; Impact via leverage -0.024; Contribution at each horizon 6.8%  
  - 5: Unconditional impact -0.300; Impact via leverage -0.039; Contribution at each horizon 11.0%  
  - 6: Unconditional impact -0.374; Impact via leverage -0.042; Contribution at each horizon 9.7%  
  - 7: Unconditional impact -0.344; Impact via leverage -0.042; Contribution at each horizon 10.3%  
  - 8: Unconditional impact -0.303; Impact via leverage -0.041; Contribution at each horizon 11.4%  
  - 9: Unconditional impact -0.235; Impact via leverage -0.067; Contribution at each horizon 24.4%  
  - 10: Unconditional impact -0.224; Impact via leverage -0.047; Contribution at each horizon 18.1%  
  - 11: Unconditional impact -0.208; Impact via leverage -0.059; Contribution at each horizon 24.3%  
  - 12: Unconditional impact -0.153; Impact via leverage -0.051; Contribution at each horizon 28.2%  
- Share of Capital expenditure: Low leveraged firms 휔휔2 = 15%; High leverage firms 휔휔1 = 85%  
- Note: Coefficients from impulse response functions based on local projection methods using firm-level quarterly data from XX countries for the period 2000Q1 to 2020Q4. Column definitions and the formula for contribution are provided in the table note.

### Figure 1 — Non-Financial Corporate Debt, Loans and Debt Securities (percent of GDP)
- Sample: Consistent sample of 70 countries across time: Advanced Economies (35) and Emerging Market and Developing Economies (35).  
- Visualization details: Dotted upper(lower) lines depict the 75th (25th) percentiles of the overall distribution. Solid lines depict the median.  
- Data source: IMF Global Debt Database and authors calculations.

### Figure 2 — Evolution of (log) Investment Following a Recession
- Method: Impulse response function based on local projection methods following Jordà (2005) using firm-level quarterly data from 75 countries for the period 2001Q1 to 2020Q4.  
- Regression specified for horizons k up to 12 quarters; variable of interest is a dummy taking value 1 at the start of a technical recession.  
- Presentation: Solid line shows point estimate for 훽0k for different horizons k; dotted lines are the 68 percent and 90 percent confidence intervals. Standard errors clustered two-way at firm and country-time level.  
- Graphical scale shown from -0.5 to 0 on the y-axis across horizons 0 to 12.

### Figure 3 — Differential Effect of Recession on (log) Investment for High-Debt Companies Relative to Low-Debt Companies
- Method: Local projections using firm-level quarterly data from 75 countries for 2001Q1 to 2020Q4; high-debt defined as above median leverage within industry.  
- Regression includes interaction of recession dummy and high-debt dummy; estimated separately for horizons up to 12 quarters.  
- Presentation: Solid line shows point estimate for 휇표k; dotted lines are the 68 percent and 90 percent confidence intervals. Y-axis range from -0.1 to 0.05 over horizons 0 to 12.

### Robustness figures (Figures 4–6) — Key robustness checks
- Figure 4: Robustness check excluding recession years when computing high-debt dummy (high debt defined by average leverage in non-recession years). Method and presentation analogous to Figure 3. Y-axis range shown -0.1 to 0 over horizons 0 to 12.
- Figure 5: Robustness by including interactions of macro variables with high-debt dummy:
  - Controls shown include: Financial Stress*Debt Dummy; Uncertainty*Debt Dummy; Inflation*Debt Dummy; Fiscal Countercyclicality*Debt; and Control for All Variables Together.  
  - Regression adds interactions 푀푀i,t−j * DebtDummy as controls. Presentation shows coefficient 휇표k with 68 percent and 90 percent confidence intervals; multiple panels with y-axis ranges including (-.3 to .1) and (-.15 to .05) depending on panel.
- Figure 6: Robustness including interaction of recession dummy with firm characteristics:
  - Controls shown include: Total Assets*Recession; Return on Assets*Recession; Liquidity*Recession; Age*Recession.  
  - Regression includes recession*FirmCharacteristic interactions (F_n). Presentation shows coefficient 휇표k with confidence intervals; y-axis ranges typically between -.15 and .05 or -.1 and .05 across panels.

### Figure 7 — Non-linear Effects with Triple Interaction
- Panels:
  - Panel A: Triple Interaction with (log) Assets. Y-axis range 0 to .1 over horizons 0 to 12.  
  - Panel B: Triple Interaction with ROA. Y-axis range approximately -.005 to .02 over horizons 0 to 12.  
  - Panel C: Triple Interaction with Share of Short-term Debt. Y-axis range approximately -.006 to .002 over horizons 0 to 12.  
- Method: Local projection estimates based on equation 3 for horizons k up to 12; solid line shows point estimate for 휈표k (coefficient on the triple interaction term) with 68 percent and 90 percent confidence intervals. Standard errors clustered two-way.

---

### ANNEX 1: DATA — sample composition and summary statistics

### Table A1.1 — Sample of 75 Countries by Region (country lists preserved)
- Africa – AFR (3): Botswana; Mauritius; South Africa.  
- Middle East and Central Asia - MCD (11): Bahrain; Egypt; Jordan; Kazakhstan; Kuwait; Oman; Pakistan; Qatar; Saudi Arabia; Tunisia; United Arab Emirates.  
- Western Hemisphere - WHD (10): Argentina; Brazil; Canada; Chile; Colombia; Jamaica; Mexico; Peru; Trinidad & Tobago; United States.  
- Asia & Pacific APD (17): Australia; Bangladesh; China; Hong Kong; India; Indonesia; Japan; Macau; Malaysia; New Zealand; Philippines; Singapore; South Korea; Sri Lanka; Taiwan; Thailand; Vietnam.  
- Europe - EUR (34): Austria; Belgium; Bulgaria; Croatia; Cyprus; Czech Republic; Estonia; Finland; France; Germany; Greece; Hungary; Iceland; Ireland; Israel; Italy; Latvia; Lithuania; Luxembourg; Malta; Netherlands; Norway; Poland; Portugal; Romania; Russia; Serbia; Slovakia; Spain; Sweden; Switzerland; Turkey; Ukraine; United Kingdom.

### Table A1.2 — Number of Firms and Observations by Country (selected 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 continues with country-level counts down to Slovakia: Number of firms 6; Obs. 492)

### Table A1.3 — Number of Firms and Observations by Sector (selected 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  
- (Table continues through Household and Personal Products: Number of Firms 361; Obs. 29,602)

### Table A1.4 — Summary Statistics by Leverage (High Debt vs Low Debt)
- High Debt (counts and summary):
  - Count (Log difference of investment) 578,167; Mean 0.02; Std 1.03; 25th -0.41; 50th 0.01; 75th 0.43.  
  - Log difference Capex to Lag Assets: Count 533,674; Mean 0.00; Std 0.09; 25th -0.01; 50th 0.00; 75th 0.01.  
  - Log difference of revenue: Count 843,137; Mean 0.02; Std 0.34; 25th -0.09; 50th 0.02; 75th 0.12.  
  - Return on assets: Count 760,256; Mean -2.41; Std 28.07; 25th -0.42; 50th 2.66; 75th 5.64.  
  - Log of assets (size): Count 794,191; Mean 1.61; Std 0.55; 25th 1.41; 50th 1.71; 75th 1.96.  
  - Firm average age: Count 1,209,254; Mean 34.59; Std 28.62; 25th 13.76; 50th 24.76; 75th 47.76.  
  - Net working capital ratio: Count 832,654; Mean 0.01; Std 0.63; 25th -0.03; 50th 0.09; 75th 0.23.
- Low Debt (counts and summary):
  - Count (Log difference of investment) 556,704; Mean 0.01; Std 1.10; 25th -0.47; 50th 0.00; 75th 0.49.  
  - Log difference Capex to Lag Assets: Count 514,915; Mean 0.00; Std 0.12; 25th -0.02; 50th 0.00; 75th 0.02.  
  - Log difference of revenue: Count 750,153; Mean 0.02; Std 0.37; 25th -0.10; 50th 0.02; 75th 0.13.  
  - Return on Assets: Count 773,571; Mean -4.84; Std 30.78; 25th -3.73; 50th 2.03; 75th 6.31.  
  - Log of Assets (Size): Count 780,024; Mean 1.36; Std 0.63; 25th 1.13; 50th 1.51; 75th 1.77.  
  - Firm average age: Count 1,191,542; Mean 28.44; Std 25.46; 25th 10.76; 50th 19.76; 75th 37.76.  
  - Net working capital ratio: Count 831,273; Mean 0.24; Std 0.56; 25th 0.12; 50th 0.29; 75th 0.47.

### Table A1.5 — Summary Statistics of Firm-Level Database
- Log difference of investment: Count 1,164,201; Mean 0.0; Std 1.1; 25th -0.4; 75th 0.5.  
- Log difference Capex to Lag Assets: Count 1,077,027; Mean 0.0; Std 0.1; 25th 0.0; 75th 0.0.  
- Log difference of revenue: Count 1,621,998; Mean 0.0; Std 0.4; 25th -0.1; 75th 0.1.  
- Return on Assets: Count 1,540,139; Mean -3.7; Std 29.7; 25th -1.8; 75th 5.9.  
- Log of Assets (Size): Count 1,574,488; Mean 1.5; Std 0.6; 25th 1.3; 75th 1.9.  
- Firm average age: Count 2,461,886; Mean 31.4; Std 27.2; 25th 11.8; 75th 42.8.  
- Net working capital ratio: Count 1,664,821; Mean 0.1; Std 0.6; 25th 0.0; 75th 0.4.

### Table A1.6 — Summary Statistics of Macroeconomic Variables
- Start of Technical Recession:
  - Source: Haver Analytics and World Economic Outlook. Countries 106. Coverage 1960Q2-2022Q4. Obs. 12,011; Mean 0.05; Std 0.21; Min 0; Max 1.  
- Peak-to-trough periods (converted to quarterly):
  - Source: Harding-Pagan (2006). Countries 93. Coverage 1960q2-2018q4. Obs. 1,108; Mean 0.34; Std 0.47; Min 0; Max 1.  
- Banking crises (converted to quarterly):
  - Source: Reinhart and Rogoff (2009). Countries 68. Coverage 1960q2-2014q4. Obs. 6,488; Mean 0.04; Std 0.20; Min 0; Max 1.  
- GDP growth (Q-o-Q):
  - Source: Haver Analytics and World Economic Outlook. Countries 106. Coverage 1960Q2-2022Q4. Obs. 12,011; Mean 0.83; Std 2.95; Min -38.02; Max 31.71.  
- Financial Stress:
  - Source: Ahir et al. (2022). Countries 110. Coverage 2000Q1-2018Q4. Obs. 8,360; Mean 0.03; Std 0.12; Min 0.00; Max 2.08.  
- World Uncertainty Index:
  - Source: Ahir, Bloom and Furceri 2022. Countries 143. Coverage 2000Q1-2021Q4. Obs. 12,441; Mean 0.19; Std 0.20; Min 0.00; Max 2.04.  
- Inflation (CPI growth):
  - Source: World Economic Outlook. Countries 71. Coverage 1996q1-2021q2. Obs. 6,782; Mean 1.17; Std 2.47; Min -4.67; Max 62.77.  
- Time-varying fiscal measure of fiscal policy countercyclicality (converted to quarterly):
  - Source: Choi, Furceri and Tovar-Jalles (2020). Countries 61. Coverage 2000Q1-2016Q4. Obs. 4,216; Mean 0.25; Std 0.34; Min -0.84; Max 2.17.

### Table A1.7 — Correlation of Investment between Capital IQ and World Economic Outlook Data
- Regression results (selected coefficients and significance):
  - Investment Growth (WEO) coefficient on Log Investment USD (CIQ): 0.942*** (standard error 0.207) and on Investment Growth (CIQ): 0.867*** (standard error 0.211) in columns reported.  
  - Log Investment USD (WEO) coefficients: 1.173*** (0.0218) and 1.350*** (0.0693) in respective specifications.  
  - Constant terms: 3.240*** (0.106), 2.556*** (0.276), 14.65*** (2.432), 15.29*** (2.588).  
  - Observations 1,107 in all columns.  
  - R-squared values: 0.717; 0.925; 0.032; 0.101 across columns.  
  - Country FE: NO/YES reported across columns.  
  - Robust standard errors in parentheses. Significance indicated: *** p<0.01, ** p<0.05, * p<0.1.

---

*Content unit: wpiea2022211-print-pdf - REFERENCES*

### ANNEX 2: ADDITIONAL RESULTS

### ANNEX 2: ADDITIONAL RESULTS

### Methodology and Data (common to figures)
- Sample: firm-level quarterly data from 75 countries for the period 2001Q1 to 2020Q4.
- Estimation method: Impulse response functions based on local projection methods following Jordà (2005).
- General regression specification (estimated separately for horizons k up to 12 quarters):
  - Dependent variable: 푦푦푛,푖,푡+푘 (log change in capital expenditure of firm n in country i at time t over the next k quarters), except where alternate dependent variables are used.
  - Key regressors: recession indicator 푅푅푖,푡−푗 (dummy = 1 at the start of a technical recession), high-debt dummy 퐷퐷푛 (various definitions across robustness checks), lags of dependent variable, and in some specifications macro variables 푀푀푖,푡−푗 or interactions.
  - Fixed effects: 훾훾푛푛푛푛푘 (firm-quarters) and 훼훼푖푖푖푖푡푡푘 (country-sector-time) in baseline interactive specifications; in unconditional macro responses country-sector fixed effects are used.
  - Standard errors: clustered two-way at the firm and country-time level.
- Horizons considered: up to 12 quarters.

### Robustness: Excluding Major Episodes and Observations
- Figure A2.1 — Excluding key crisis years:
  - Left panel excludes 2009 (Global Financial Crisis peak).
  - Right panel excludes 2020 (COVID-19).
  - Purpose: check sensitivity of interaction coefficients 휇̂푘 to omission of these years.
- Figure A2.2 — Excluding countries with most observations:
  - Each impulse response omits one of the 10 countries with the most observations, excluded one at a time (labels include AUS, CAN, CHN, GBR, HKG, IND, JPN, KOR, TWN, USA).
  - Purpose: assess influence of dominant-country samples on 휇̂푘.
- Figure A2.3 — Excluding one region at a time:
  - Regions excluded one at a time: AFR, APD, EUR, MCD, WHD.
  - Purpose: test regional concentration effects on estimates.

### Robustness: Sectoral and Data Processing Variations
- Figure A2.4 — Excluding one 2-digit sector at a time:
  - Sectors excluded in separate specifications include Automobiles and Components; Capital Goods; Consumer Durables and Apparel; Consumer Services; Energy; Food and Staples Retailing; Food, Beverage and Tobacco; Health Care Equipment and Services; Household and Personal Products; Materials; Media and Entertainment; Pharmaceuticals and Biotechnology; Professional Services; Retailing; Semiconductors; Software and Services; Technology Hardware and Equipment; Telecommunication Services; Transportation; Utilities.
  - Purpose: ensure results are not driven by any single sector.
- Figure A2.5 — Different winsorizing schemes:
  - Left panel: winsorize 0.05 percent tails of the dependent variable.
  - Right panel: winsorize 5 percent tails of the dependent variable.
  - Baseline winsorization: 1 percent tails.
  - Purpose: test sensitivity to extreme dependent-variable observations.

### Robustness: Alternate Dependent Variables and Recession Definitions
- Figure A2.6 — Different dependent variables:
  - Left panel: Capex to Lag Assets (log).
  - Right panel: Revenues (log).
  - Purpose: verify whether results hold when using capex normalized by lagged assets or revenues as outcome.
- Figure A2.7 — Alternate recession variables:
  - Panels include: continuous growth variable (inverted), peaks and troughs in economic activity (Harding-Pagan algorithm), start of financial crises, start of banking crises, and periods of negative quarterly real GDP growth.
  - Notes: Periods of banking crises identified as quarters during which a banking crisis-led recession occurred; peaks and troughs identified with Harding-Pagan.
  - Purpose: assess robustness to different definitions of recession/timing.

### Robustness: Alternate Debt Definitions and Specifications
- Figure A2.8 — Alternate debt dummies:
  - Left panel: high-debt defined as above median leverage within country.
  - Right panel: high-debt defined as above median leverage within income group (advanced economies or EMDEs) and industry.
  - Purpose: test sensitivity to within-country versus within-income-group definitions.
- Figure A2.9 — Alternate high-debt specifications II:
  - Left panel: highest quartile (or tercile) leverage within industry.
  - Right panel: mean of leverage over time (continuous variable).
  - Purpose: evaluate robustness to identifying high-debt firms via top quartile or continuous average leverage.
- Figure A2.10 — Companies in the 2nd quartiles vs Companies in the 4th quartile:
  - Comparison of high-debt (4th quartile) versus low-debt (2nd quartile) firms.
  - Purpose: contrast responses across debt quartiles.

### Robustness: Additional Firm-Level Controls
- Figure A2.11 — Additional controls:
  - Left panel: adds lags of firm size (ln assets), return on assets (ROA), and liquidity (current assets minus liabilities as a share of assets) as additional controls.
  - Right panel: further adds log revenues and firm age as controls.
  - Purpose: check whether inclusion of firm-level controls alters the interaction coefficient 휇̂푘 relative to baseline.

### Unconditional and Conditional Macro Variable Effects
- Figure A2.12 — Unconditional response of log(Investment) to macro variables (regression with macro variable M only):
  - Macro variables considered in separate panels: Change in Financial Stress; Change in Uncertainty; Change in Inflation; Fiscal countercyclicality.
  - Specification: 푦푦푛,푖,푡+푘 = 훼푖푖푘 + 훾푛푛푘 + ∑훽푗푘 M𝑖,𝑡−𝑗 + ∑휃푗푘 y𝑛,𝑖,𝑡−𝑗 + ε.
  - Purpose: estimate 훽̂0𝑘, the unconditional effect of macro variables on firm-level investment over horizons.

- Figure A2.13 — Coefficients on interaction of macro variables with high-debt dummy as controls:
  - Interaction terms included as controls: Financial Stress*Debt Dummy; Change in Uncertainty*Debt Dummy; Inflation*Debt Dummy; Fiscal countercyclicality*Debt Dummy.
  - Specification adds ∑휈푗푘 M𝑖,𝑡−𝑗 * D𝑛 to the baseline interactive regression.
  - Purpose: account for macro-variable differential effects by debt status when estimating recession*debt interaction.

### Robustness: Interaction of Recession with Firm Characteristics
- Figure A2.14 — Coefficients on interaction of recession dummy with firm-characteristic dummies:
  - Interaction coefficients estimated for: Total Assets*Recession; Return on Assets*Recession; Liquidity*Recession; Age*Recession.
  - Firm-characteristic dummies defined analogously to high-debt dummy: 1 if above median of characteristic averaged across the sample.
  - Purpose: control for heterogeneous recession responses by firm characteristics.

### Non-linear Effects: Triple Interaction with Smooth Transition Functions
- Figure A2.15 — Triple interaction using smooth transition functions:
  - Equation 3-type specification where firm-level characteristic X𝑛,𝑖,𝑡 is transformed via G(X) = 1 / (1 + exp(−X)) before interacting with debt and recession dummies.
  - Panels:
    - Panel A: Triple Interaction with (log) Assets.
    - Panel B: Triple Interaction with ROA.
    - Panel C: Triple Interaction with Share of Short-term Debt.
  - The coefficient of interest: 휈̂푘, the triple interaction coefficient, shown across horizons with 68 percent and 90 percent confidence intervals.
  - Purpose: explore non-linear heterogeneous effects of firm characteristics on the recession*debt interaction.

*Scarring and Corporate Debt — Working Paper No. WP/2022/211*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022211-print-pdf.pdf_