## wpiea2022038-print-pdf — Section 4 concludes

## Source details

**Canonical URL:** [wpiea2022038-print-pdf — Section 4 concludes](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022038-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022038-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022038-print-pdf.pdf.json)

---

### Data description
- Data: annual firm-level database merging the Slovenian Business Register and Annual Reports of Corporate Entities (JOLP).
- Unique feature: firms operating abroad report BS and IS for foreign operations and liability positions separately, enabling calculation of the share of foreign equity in total foreign liabilities that closely resembles measures used in cross-country macro literature.
- Constructed firm-level foreign equity share:
  - Numerator: sum of foreign capital, long and short term loans, trade and consumption loans, and financial leasing from the rest of the world (where the foreign creditor owns more than 10 percent of the respective firm) that are encompassed under FDI in balance of payments statistics.
  - Denominator: total liabilities to the rest of the world.
- Limitations:
  - Portfolio equity and tradable debt securities not directly reported by firms due to lack of information on ultimate owner.
  - Rationale that omission is likely negligible for Slovenia because publicly-listed companies account for around 1 percent of firms and debt securities issuance is negligible; firms mainly use bank loans.

### Rationale for focusing on Slovenia and FDI
- At end of 2014:
  - 2,899 Slovenian firms with inward FDI in the form of direct affiliation.
  - Non-financial corporate sector accounted for 83 percent of total inward FDI in value terms.
  - Around 5 percent of all Slovenian firms had FDI liabilities.
  - Firms with FDI liabilities accounted for 19 percent of capital, 22 percent of assets, and 22 percent of employees in the corporate sector.
- FDI composition:
  - Out of 3,531 inward FDIs in Slovenia, 62 percent were new (greenfield) investment.

### Sample restrictions and robustness considerations
- Sample period: 2005 – 2014 (start chosen to avoid exchange rate confounding; Slovenia entered ERM II in July 2004; Euro adoption January 2007).
- Exclusions:
  - Firms in financial, insurance, and government sectors.
  - Firms with dependent variable trimmed at 0.1 – 99.9 percent on a year-by-year basis to remove extreme outliers.
  - Firms that went bankrupt (total excluded bankrupt firms: 5,106).
- Survival bias assessment:
  - Defaults very low pre-crisis, increased dramatically after the crisis for firms without foreign equity (Figure B.1). Inclusion of bankrupt firms would tend to increase the difference in sales growth between firms with and without foreign equity.
- Fire-sale FDI concern:
  - Firms that changed foreign equity status were few and stable over the period (Figure B.2).

### Summary statistics (key figures preserved)
- Table 1: Summary statistics for firms with and without positive foreign equity, split pre-crisis / post-crisis (post-crisis cut-off year: 2009; last year pre-crisis: 2008).
  - Sales growth:
    - Firms with positive foreign equity: Pre-crisis Mean 18.32, p50 10.24; Post-crisis Mean 9.73, p50 1.77
    - Firms without foreign equity: Pre-crisis Mean 15.15, p50 9.73; Post-crisis Mean 3.57, p50 0.10
  - Total liabilities/TA:
    - With equity: Pre-crisis Mean 67.74, p50 61.52; Post-crisis Mean 74.61, p50 59.38
    - Without equity: Pre-crisis Mean 77.14, p50 70.48; Post-crisis Mean 82.25, p50 68.37
  - Size - assets (1000€):
    - With equity: Pre-crisis Mean 11,541.91, p50 1,550.50; Post-crisis Mean 10,072.24, p50 3,84.00
    - Without equity: Pre-crisis Mean 3,872.38, p50 459.00; Post-crisis Mean 3,663.48, p50 451.00
  - Size - employment:
    - With equity: Pre-crisis Mean 86.33, p50 9.00; Post-crisis Mean 70.54, p50 8.00
    - Without equity: Pre-crisis Mean 29.89, p50 5.00; Post-crisis Mean 23.83, p50 5.00
  - Firm age:
    - With equity: Pre-crisis Mean 7.97, p50 7.00; Post-crisis Mean 8.49, p50 6.00
    - Without equity: Pre-crisis Mean 11.17, p50 13.00; Post-crisis Mean 12.62, p50 13.00
  - Notes: 736 observations dropped where sales growth > approximately 600 percent per year in absolute terms.
- Table 2 (entire sample 2005–2014):
  - Number of firms each year: 38,165; 39,782; 42,786; 45,856; 47,851; 49,086; 52,197; 54,577; 56,957; 59,856
  - % of firms with foreign liabilities each year: 23 23 23 22 22 22 22 22 22 21
  - o/w equity liabilities each year: 10 10 11 11 11 11 11 12 12 13
  - Average number of firms with a positive equity share: around one thousand.
- Table 3 (by firm size, 2005–2014):
  - % of firms with foreign liabilities:
    - Below median: 7 7 7 7 7 7 7 7 7 6
    - Above median: 39 38 39 38 38 37 37 36 36 36
  - o/w equity liabilities:
    - Below median: 7 9 8 9 10 9 10 12 14 16
    - Above median: 11 11 11 11 11 11 11 11 12 12
  - Mean equity share:
    - Below median: 52 47 49 52 54 50 49 47 48 47
    - Above median: 48 49 48 50 52 51 51 50 49 50

### Stylized facts and definitions
- Foreign equity share (Equation 1):
  - ForeignEquityLiabShare = Equity_Foreign / (Equity_Foreign + Debt_Foreign)
- Overall equity share (Equation 2):
  - EquityLiabShare = (Equity_Home + Equity_Foreign) / (Equity_Home + Debt_Home + Equity_Foreign + Debt_Foreign)
- Observations:
  - The two ratios are highly correlated but not identical.
  - Distribution of firm-level foreign equity share is heterogeneous and relatively uniform with small spikes at extremes (Figure 1).
  - Aggregate equity share for Slovenia is slow-moving and relatively stable over time (Figure 2).
  - Firm-specific standard deviation of foreign equity share over 2005–2014 shows considerable temporal variation, not driven solely by larger firms (Figure 3 and 3b).

### 3.2 Firms’ crisis vulnerability and foreign capital structure — Empirical specification and identification
- Objective: assess whether composition of foreign liabilities (foreign equity share > 0) matters for susceptibility to the 2009 global financial crisis.
- Dependent variable: firms’ sales growth (annual percentage change in sales level).
- Main pooled difference-in-differences specification:
  - Y_igt = λ_t + D_g + γ_g t + δ(Post_t · D_g) + β X_igt + ε_igt
  - D_g = indicator if foreign equity share in foreign liabilities of a firm is larger than 0.
  - Post_t = indicator taking value 1 after 2008.
  - X_igt includes: size, openness, liquidity ratio, productivity, tangible assets, age, age squared, leverage, and dummy for publicly-listed firm.
- Two-way fixed effects (panel) specification includes α_i firm fixed effects.
- Time-varying effects specification estimates δ_t relative to 2009 (omitted year).
- Identification checks and mitigations:
  - Parallel trends assessed via pre-crisis trends (Figure 4).
  - Anticipation effects considered unlikely given annual data and time span.
  - Group-specific linear time trends and firm fixed effects used for sensitivity.
  - Selection bias addressed via firm fixed effects and entropy balancing re-weighting.
  - Pre-test limitations and trend extrapolation caveats noted.

### 3.2 Results on performance (sales growth)
- Main pooled and panel findings (Table 4; interaction term δ reported):
  - Column (1): δ = 1.992* (standard error (0.974)); Time FE: Yes; Firm FE: No; Group-specific time trend: No; N = 70,337; R^2 = 0.097.
  - Column (2): δ = 2.576* (standard error (1.405)); Time FE: Yes; Firm FE: No; Group-specific time trend: Yes; N = 70,337; R^2 = 0.097.
  - Column (3): δ = 3.576** (standard error (1.376)); Time FE: Yes; Firm FE: Yes; Group-specific time trend: No; N = 70,337; R^2 = 0.111.
  - Column (4): δ = 2.231 (standard error (1.628)); Time FE: Yes; Firm FE: Yes; Group-specific time trend: Yes; N = 70,337; R^2 = 0.111.
  - Column (5) (with matching / entropy balancing): δ = 2.688** (standard error (0.962)); Time FE: Yes; Firm FE: No; Group-specific time trend: No; N = 70,337; R^2 = 0.114.
  - Significance notes: *** 1% level; ** 5% level; * 10% level. Standard errors robust and clustered at sector level.
- Interpretation:
  - Positive and statistically significant interaction term in most specifications indicates firms with a positive foreign equity share tended to weather the global financial crisis better (higher sales growth) relative to firms without positive foreign equity share.
  - Including group-specific linear time trends with firm fixed effects can render the effect insignificant in some specifications.
- Time profile (Figure 5; model in (5), coefficients relative to 2009):
  - Effect materialized in the first year after the crisis.
  - Effect most pronounced in 2011, then slowly diminished.
  - Pre-crisis coefficients are insignificant, supporting parallel trends.
- Internal capital flows (Figure 6):
  - Intra-firm trade credit and intra-firm loans (as share of total liabilities) relatively stable before 2010, increased substantially in 2010, and continued growing thereafter — consistent with increased role of intra-firm funding after the crisis.
- Heterogeneity by external financial dependence (Table 5; triple-interaction):
  - Baseline coefficient δ reported:
    - Column (1): δ = 2.092*** (standard error (0.449)); N = 57,629; R^2 = 0.104.
    - Column (2): δ = 3.295** (standard error (1.168)); N = 57,629; R^2 = 0.104.
    - Column (3) (with matching): δ = 2.928*** (standard error (0.619)); N = 57,629; R^2 = 0.124.
  - Interaction with External Financial Dependence:
    - Column (1): δ·External Financial Dependence = 4.130** (standard error (1.595)).
    - Column (2): δ·External Financial Dependence = 4.153** (standard error (1.579)).
    - Column (3): δ·External Financial Dependence = 5.364** (standard error (1.874)).
  - All reported interaction coefficients with external financial dependence are positive and significant at the 5% level.
  - External financial dependence: industry-level, time-invariant measure — proportion of capital expenditures financed with external funds.
  - Excluding construction sector confirms larger effect in industries more dependent on external finance.
- Robustness:
  - Restricting sample to 2005–2011 leaves results virtually unchanged.
  - Narrower definition of foreign equity (excluding intra-firm trade credit and loans) yields broadly similar results.
  - Alternative outcomes (profitability ratio: EBIT/total assets; net investment rate) are broadly consistent.

### 3.2 Default probabilities after the crisis
- Models estimated:
  - Linear Probability Model and Logit Model predicting P(1[Default]_i,t>2008) using 2008 covariates.
- Controls (in 2008): size, leverage, openness, liquidity ratio, productivity, tangible assets, age, age squared, PLC dummy.
- Table 6 results (N = 7,599):
  - Linear Probability Model (column 1):
    - Foreign Equity Share Dummy = -0.0372*** (standard error (0.00864))
    - Leverage = 0.00148*** (standard error (0.000136))
    - Log size (assets) = 0.0175*** (standard error (0.00338))
    - Openness = -8.53e-05 (standard error (0.000188))
    - Liquidity Ratio = -1.78e-05 (standard error (1.12e-05))
    - Productivity = -8.30e-05** (standard error (3.01e-05))
    - Tangible assets = -0.000224 (standard error (0.000203))
    - Age = -0.00209 (standard error (0.00138))
    - Age squared = 4.29e-05 (standard error (4.16e-05))
    - PLC = 0.00645 (standard error (0.0159))
    - R^2 = 0.043
  - Logit Model (column 2):
    - Foreign Equity Share Dummy = -0.637*** (standard error (0.181))
    - Leverage = 0.0124*** (standard error (0.00165))
    - Log size (assets) = 0.239*** (standard error (0.0293))
    - Openness = -0.00111 (standard error (0.00285))
    - Liquidity Ratio = -0.00502*** (standard error (0.00135))
    - Productivity = -0.000745 (standard error (0.000671))
    - Tangible assets = -0.00448* (standard error (0.00237))
    - Age = -0.0302** (standard error (0.0128))
    - Age squared = 0.000534 (standard error (0.000410))
    - PLC = -0.0305 (standard error (0.267))
    - (Pseudo)R^2 = 0.0735
  - Standard errors robust and clustered at sector level.
- Interpretation:
  - Firms with a positive foreign equity share in 2008 were less likely to default after 2008; result robust across Linear Probability and Logit specifications.
  - Higher leverage increases default probability.
  - Larger firms (by assets) had higher conditional default probability.
  - Higher liquidity and productivity are associated with lower default probability.
- Robustness:
  - Restricting to defaults in 2009 and 2009–2010 leaves results virtually unchanged.

### 3.3 Determinants of firm’s foreign capital structure — Empirical specification
- Baseline: regress time-series mean of dependent variable on time-series means of explanatory variables (between estimator).
- Robustness: panel fixed effects to exploit time variation.
- Dependent variables: extensive margin (dummy for any foreign equity) and intensive margin (log share of foreign equity).
- Explanatory variables: firm size, productivity, tangibility, growth, capital intensity, age, profitability, openness.
- Controls: overall equity share in total liabilities, PLC dummy, sector fixed effects.

### 3.3 Results (Table 7: Cross-sectional estimates)
- Sample: N 15,392
- (Pseudo)R2: 0.138 (column (1)), 0.100 (column (2))
- Column (1): probit — dependent = 1 if firm has any foreign equity in foreign liabilities.
- Column (2): intensive-margin — log share of foreign equity.
- Coefficients (standard errors in parentheses):
  - Log size (assets): 0.140*** (0.017) in (1); 0.070*** (0.013) in (2)
  - Log size (employment): 0.107*** (0.019) in (1); 0.094*** (0.014) in (2)
  - Openness: 0.011*** (0.000) in (1); 0.008*** (0.000) in (2)
  - Productivity: 0.862*** (0.238) in (1); 1.168*** (0.203) in (2)
  - Tangible assets: -0.004*** (0.001) in (1); -0.002*** (0.000) in (2)
  - Growth: -0.148*** (0.051) in (1); -0.180*** (0.037) in (2)
  - Capital intensity: 0.000 (0.000) in (1); 0.000* (0.000) in (2)
  - Age: -0.028*** (0.002) in (1); -0.017*** (0.002) in (2)
  - Profitability: 0.139*** (0.035) in (1); 0.122*** (0.029) in (2)
  - PLC: 0.025 (0.082) in (1); -0.073 (0.069) in (2)
  - Sector FE: Yes (both columns)
- Main empirical findings:
  - Larger, more open, more productive, younger, and more profitable firms are more likely to have foreign equity and have higher foreign equity shares.
  - Firms with more tangible assets have lower probability and lower share of foreign equity.
  - Faster-growing firms have lower probability and lower share of foreign equity.
  - No significant effect for being publicly-listed (PLC) and only marginal effect for capital intensity.

### Additional considerations and aggregate context
- Concern that firm size/productivity correlate with choice of borrowing in foreign currency versus equity: after Euro adoption, share of external debt denominated in foreign currency is less than one percent for Slovenia (Figure B.4), indicating corporate short-term foreign-currency debt volume is small and unlikely to drive main results.

### Conclusion and policy implications
- A positive foreign equity share at the firm level is associated with greater resilience to external shocks:
  - Firms with positive foreign equity share performed better during the global financial crisis (higher sales growth) and were less likely to default.
- Micro-level determinants imply that countries with larger, more open, and more productive firms will tend to exhibit higher foreign equity shares in corporate external liabilities.
- Policy implication: assessing countries’ vulnerability to sudden changes in the financial account can be improved by incorporating information on corporate foreign funding structure. Where firm-level data are unavailable, readily available aggregate characteristics (size, productivity, age, openness) combined with aggregate net foreign asset statistics can provide useful indications of corporate-sector vulnerability.

*Determinants and Effects of Countries’ External Capital Structure: A Firm-Level Analysis — Working Paper No. WP/22/38 — Section 4 concludes*

### Section 4 concludes.

Section 4 concludes.

### Data description
- Data: annual firm-level database merging the Slovenian Business Register and Annual Reports of Corporate Entities (JOLP).
- Unique feature: firms operating abroad report BS and IS for foreign operations and liability positions separately, enabling calculation of the share of foreign equity in total foreign liabilities that closely resembles measures used in cross-country macro literature.
- Constructed firm-level foreign equity share:
  - Numerator: sum of foreign capital, long and short term loans, trade and consumption loans, and financial leasing from the rest of the world (where the foreign creditor owns more than 10 percent of the respective firm) that are encompassed under FDI in balance of payments statistics.
  - Denominator: total liabilities to the rest of the world.
- Limitations noted:
  - Portfolio equity and tradable debt securities not directly reported by firms due to lack of information on ultimate owner.
  - Rationale that omission is likely negligible for Slovenia because publicly-listed companies account for around 1 percent of firms and debt securities issuance is negligible; firms mainly use bank loans.

### Rationale for focusing on Slovenia and FDI
- At end of 2014:
  - 2,899 Slovenian firms with inward FDI in the form of direct affiliation.
  - Non-financial corporate sector accounted for 83 percent of total inward FDI in value terms.
  - Around 5 percent of all Slovenian firms had FDI liabilities.
  - Firms with FDI liabilities accounted for 19 percent of capital, 22 percent of assets, and 22 percent of employees in the corporate sector.
- FDI composition: Out of 3,531 inward FDIs in Slovenia, 62 percent were new (greenfield) investment.

### Sample restrictions and robustness considerations
- Sample period: 2005 – 2014 (start chosen to avoid exchange rate confounding; Slovenia entered ERM II in July 2004; Euro adoption January 2007).
- Exclusions:
  - Firms in financial, insurance, and government sectors.
  - Firms with dependent variable trimmed at 0.1 – 99.9 percent on a year-by-year basis to remove extreme outliers.
  - Firms that went bankrupt (total excluded bankrupt firms: 5,106).
- Survival bias assessment:
  - Figure B.1 (appendix B) shows defaults across firm types; defaults very low pre-crisis, increased dramatically after the crisis for firms without foreign equity.
  - Inclusion of bankrupt firms would tend to increase the difference in sales growth between firms with and without foreign equity.
- Fire-sale FDI concern:
  - Figure B.2 (appendix B) plots firms that changed foreign equity status; number of changes is small and stable over period, including during and immediately after the global financial crisis.

### Summary statistics (key figures preserved)
- Table 1: Summary statistics for firms with and without positive foreign equity, split pre-crisis / post-crisis (post-crisis cut-off year: 2009; last year pre-crisis: 2008).
  - Sales growth:
    - Firms with positive foreign equity: Pre-crisis Mean 18.32, p50 10.24; Post-crisis Mean 9.73, p50 1.77
    - Firms without foreign equity: Pre-crisis Mean 15.15, p50 9.73; Post-crisis Mean 3.57, p50 0.10
  - Total liabilities/TA:
    - With equity: Pre-crisis Mean 67.74, p50 61.52; Post-crisis Mean 74.61, p50 59.38
    - Without equity: Pre-crisis Mean 77.14, p50 70.48; Post-crisis Mean 82.25, p50 68.37
  - Size - assets (1000€):
    - With equity: Pre-crisis Mean 11,541.91, p50 1,550.50; Post-crisis Mean 10,072.24, p50 3,84.00
    - Without equity: Pre-crisis Mean 3,872.38, p50 459.00; Post-crisis Mean 3,663.48, p50 451.00
  - Size - employment:
    - With equity: Pre-crisis Mean 86.33, p50 9.00; Post-crisis Mean 70.54, p50 8.00
    - Without equity: Pre-crisis Mean 29.89, p50 5.00; Post-crisis Mean 23.83, p50 5.00
  - Firm age:
    - With equity: Pre-crisis Mean 7.97, p50 7.00; Post-crisis Mean 8.49, p50 6.00
    - Without equity: Pre-crisis Mean 11.17, p50 13.00; Post-crisis Mean 12.62, p50 13.00
  - Tangibility, firm openness, productivity, liquidity ratio, capital intensity, PLC share, and N reported in Table 1 (see table for exact numbers).
  - Notes: 736 observations dropped where sales growth > approximately 600 percent per year in absolute terms.
- Table 2 (entire sample 2005–2014):
  - Number of firms each year: 38,165; 39,782; 42,786; 45,856; 47,851; 49,086; 52,197; 54,577; 56,957; 59,856
  - % of firms with foreign liabilities each year: 23 23 23 22 22 22 22 22 22 21
  - o/w equity liabilities each year: 10 10 11 11 11 11 11 12 12 13
  - Average number of firms with a positive equity share: around one thousand.
- Table 3 (by firm size, 2005–2014):
  - % of firms with foreign liabilities:
    - Below median: 7 7 7 7 7 7 7 7 7 6
    - Above median: 39 38 39 38 38 37 37 36 36 36
  - o/w equity liabilities:
    - Below median: 7 9 8 9 10 9 10 12 14 16
    - Above median: 11 11 11 11 11 11 11 11 12 12
  - Mean equity share:
    - Below median: 52 47 49 52 54 50 49 47 48 47
    - Above median: 48 49 48 50 52 51 51 50 49 50

### Stylized facts and definitions
- Definition of firm foreign equity share (Equation 1):
  - ForeignEquityLiabShare = Equity_Foreign / (Equity_Foreign + Debt_Foreign)
- Definition of firm overall equity share (Equation 2):
  - EquityLiabShare = (Equity_Home + Equity_Foreign) / (Equity_Home + Debt_Home + Equity_Foreign + Debt_Foreign)
- Key observations:
  - The two ratios (foreign equity share and overall equity share) are highly correlated but not identical.
  - Distribution of firm-level foreign equity share (Figure 1): heterogeneous and relatively uniform with small spikes at extremes.
  - Aggregate equity share for Slovenia (Figure 2, Lane and Milesi-Ferretti (2018) data): slow-moving, relatively stable over time.
  - Micro-level variation: firm-specific standard deviation of foreign equity share over 2005–2014 shows considerable temporal variation (Figure 3); this variation is not driven solely by larger firms (Figure 3b).

*Source: wpiea2022038-print-pdf - Section 4 concludes.*

### 3.2    Firms’ crisis vulnerability and foreign capital structure

### 3.2    Firms’ crisis vulnerability and foreign capital structure

### 3.2.1    Empirical specification and identification
- Objective:
  - Investigate whether the composition of foreign liabilities (presence of foreign equity share > 0) matters for a firm’s susceptibility to the 2009 global financial crisis.
  - Dependent variable: firms’ sales growth (annual percentage change in sales level).
- Main difference-in-differences specification (pooled):
  - Y_igt = λ_t + D_g + γ_g t + δ(Post_t · D_g) + β X_igt + ε_igt
  - D_g = indicator if foreign equity share in foreign liabilities of a firm is larger than 0.
  - Post_t = indicator taking value 1 after 2008.
  - X_igt includes: size, openness, liquidity ratio, productivity, tangible assets, age, age squared, leverage, and dummy for publicly-listed firm.
- Two-way fixed effects specification (panel):
  - Y_igt = α_i + λ_t + D_g + γ_g t + δ(Post_t · D_g) + β X_igt + ε_igt
  - α_i captures firm fixed effects.
- Time-varying effects specification:
  - Y_igt = (α_i) + λ_t + D_g + Σ_{t=2005, t≠2009}^{2015} δ_t · D_gt + β X_igt + ε_igt
  - Coefficients δ_t measure effect relative to 2009 (omitted year).
- Threats to identification and mitigation:
  - Parallel trends assumption assessed via plotted pre-crisis sales growth trends (Figure 4) — trends similar before the crisis.
  - Address anticipation effects: annual data and focus on long time span make early anticipation unlikely; crisis effects should be incorporated by end-2009.
  - Group-specific linear time trends included in some specifications to test sensitivity to trend differences.
  - Selection bias concern: foreign investors may sort into more resilient firms — addressed by firm fixed effects and entropy balancing re-weighting.
  - Composition changes over time addressed by entropy balancing (Hainmueller (2012); Hainmueller and Xu (2013)) to re-weight control group to match treatment group first moments.
  - Note on pre-testing limitations: pre-tests have low power (Roth (2020)); linear trend extrapolation may be simplistic (Wolfers (2006), Lee and Solon (2011), Rambachan and Roth (2020)).

### 3.2.2    Results on performance (sales growth)
- Main pooled and panel findings (Table 4; interaction term δ reported):
  - Column (1): δ = 1.992* (standard error (0.974)); Time FE: Yes; Firm FE: No; Group-specific time trend: No; N = 70,337; R^2 = 0.097.
  - Column (2): δ = 2.576* (standard error (1.405)); Time FE: Yes; Firm FE: No; Group-specific time trend: Yes; N = 70,337; R^2 = 0.097.
  - Column (3): δ = 3.576** (standard error (1.376)); Time FE: Yes; Firm FE: Yes; Group-specific time trend: No; N = 70,337; R^2 = 0.111.
  - Column (4): δ = 2.231 (standard error (1.628)); Time FE: Yes; Firm FE: Yes; Group-specific time trend: Yes; N = 70,337; R^2 = 0.111.
  - Column (5) (with matching / entropy balancing): δ = 2.688** (standard error (0.962)); Time FE: Yes; Firm FE: No; Group-specific time trend: No; N = 70,337; R^2 = 0.114.
  - Notes on significance: *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level.
  - Standard errors are robust and clustered at sector level.
- Interpretation:
  - Positive and statistically significant interaction term in most specifications indicates firms with a positive foreign equity share tended to weather the global financial crisis better (higher sales growth) relative to firms without positive foreign equity share.
  - When including group-specific linear time trend with firm fixed effects the effect becomes insignificant in some specifications, possibly due to reduced precision or overfitting.
- Time profile of effect (Figure 5; model in (5), coefficients relative to 2009):
  - Effect materialized in the first year after the crisis.
  - Effect was most pronounced in 2011, then slowly diminished.
  - Pre-crisis coefficients are insignificant, supporting parallel trends.
  - Figure 5 shows point estimates with 90 percent confidence intervals.
- Additional evidence on internal capital flows (Figure 6):
  - Intra-firm trade credit and intra-firm loans (as share of total liabilities) were relatively stable before 2010, increased substantially in 2010, and continued growing thereafter — consistent with increased role of intra-firm funding after the crisis.
- Heterogeneity by external financial dependence (Table 5; triple-interaction):
  - Baseline coefficient δ reported:
    - Column (1): δ = 2.092*** (standard error (0.449)); N = 57,629; R^2 = 0.104.
    - Column (2): δ = 3.295** (standard error (1.168)); N = 57,629; R^2 = 0.104.
    - Column (3) (with matching): δ = 2.928*** (standard error (0.619)); N = 57,629; R^2 = 0.124.
  - Interaction with External Financial Dependence:
    - Column (1): δ·External Financial Dependence = 4.130** (standard error (1.595)).
    - Column (2): δ·External Financial Dependence = 4.153** (standard error (1.579)).
    - Column (3): δ·External Financial Dependence = 5.364** (standard error (1.874)).
  - All reported interaction coefficients with external financial dependence are positive and significant at the 5% level.
  - External financial dependence measure: proportion of capital expenditures financed with external funds (Duygan-Bump, Levkov, and Montoriol-Garriga (2015), following Cetorelli and Strahan (2006)); industry-level, time-invariant.
  - Excluding construction sector (high external dependence and large recession hit) confirms the conjecture that effect is larger in industries more dependent on external finance.
- Robustness checks:
  - Restricting sample to 2005–2011 (avoid confounding from Slovenia banking crisis in 2012) leaves results virtually unchanged.
  - Narrower definition of foreign equity (excluding intra-firm trade credit and loans) yields broadly similar results.
  - Alternative outcomes (profitability ratio: EBIT/total assets; net investment rate) are consistent, though some specifications render effect insignificant.

### 3.2.3    Default probabilities after the crisis
- Models estimated:
  - Linear Probability Model: P(1[Default]_i,t>2008) = 1[Foreign Equity Share Dummy]_i,−1 + X_i,−1
  - Logit Model: P(1[Default]_i,t>2008) = G(1[Foreign Equity Share Dummy]_i,−1 + X_i,−1)
  - Controls (in 2008): size, leverage, openness, liquidity ratio, productivity, tangible assets, age, age squared, PLC dummy (publicly-listed).
- Table 6 results (N = 7,599):
  - Linear Probability Model (column 1):
    - Foreign Equity Share Dummy = -0.0372*** (standard error (0.00864)); (*** Significant at the 1% level)
    - Leverage = 0.00148*** (standard error (0.000136))
    - Log size (assets) = 0.0175*** (standard error (0.00338))
    - Openness = -8.53e-05 (standard error (0.000188))
    - Liquidity Ratio = -1.78e-05 (standard error (1.12e-05))
    - Productivity = -8.30e-05** (standard error (3.01e-05))
    - Tangible assets = -0.000224 (standard error (0.000203))
    - Age = -0.00209 (standard error (0.00138))
    - Age squared = 4.29e-05 (standard error (4.16e-05))
    - PLC = 0.00645 (standard error (0.0159))
    - R^2 = 0.043
  - Logit Model (column 2):
    - Foreign Equity Share Dummy = -0.637*** (standard error (0.181))
    - Leverage = 0.0124*** (standard error (0.00165))
    - Log size (assets) = 0.239*** (standard error (0.0293))
    - Openness = -0.00111 (standard error (0.00285))
    - Liquidity Ratio = -0.00502*** (standard error (0.00135))
    - Productivity = -0.000745 (standard error (0.000671))
    - Tangible assets = -0.00448* (standard error (0.00237))
    - Age = -0.0302** (standard error (0.0128))
    - Age squared = 0.000534 (standard error (0.000410))
    - PLC = -0.0305 (standard error (0.267))
    - (Pseudo)R^2 = 0.0735
  - Notes: Robust and clustered (at sector level) standard errors in parentheses.
- Interpretation:
  - Firms with a positive foreign equity share in 2008 were less likely to default after 2008; result robust across Linear Probability and Logit models.
  - Leverage increases default probability; more leveraged firms pre-crisis more likely to default.
  - Larger firms (by assets) had higher probability of default conditional on controls.
  - More productive, more liquid, younger, and firms with more tangible assets were less likely to default.
- Robustness:
  - Restricting sample to firms that defaulted in 2009 and 2009–2010 leaves results virtually unchanged.
  - In all specifications, positive foreign equity share in 2008 reduces probability of default after the crisis.

*Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022038-print-pdf.pdf*

### 3.3    Determinants of firm’s foreign capital structure

### 3.3    Determinants of firm’s foreign capital structure

### 3.3.1    Empirical specification
- Baseline approach: regress time-series mean of dependent variable on time-series means of explanatory variables (between estimator regression).  
- Robustness check: panel fixed effects regression to exploit time-variation in foreign equity share.  
- Dependent variables considered: extensive margin (dummy for any foreign equity in foreign liabilities) and intensive margin (log share of foreign equity).  
- Explanatory variables follow corporate finance literature: firm size, productivity, tangibility of assets, growth, capital intensity, firm age, profitability, openness.  
- Controls included: overall equity share in firms’ total liabilities, dummy for publicly-listed companies (PLC), and sector-level fixed effects.

### 3.3.2    Results (Table 7: Cross-sectional estimates for existing foreign capital structure)
- Sample size and fit:
  - N 15,392
  - (Pseudo)R2 0.138 (column (1)), 0.100 (column (2))
- Estimation details:
  - Column (1): probit regression — dependent variable equals one if firm has any foreign equity in its foreign liabilities.
  - Column (2): intensive-margin regression — log share of foreign equity.
  - Robust and clustered (at sector level) standard errors reported in parentheses.
  - All regressions control for the overall equity share in firms’ total liabilities.

- Estimated coefficients (standard errors in parentheses):
  - Log size (assets): 0.140*** (0.017) in (1); 0.070*** (0.013) in (2)
  - Log size (employment): 0.107*** (0.019) in (1); 0.094*** (0.014) in (2)
  - Openness: 0.011*** (0.000) in (1); 0.008*** (0.000) in (2)
  - Productivity: 0.862*** (0.238) in (1); 1.168*** (0.203) in (2)
  - Tangible assets: -0.004*** (0.001) in (1); -0.002*** (0.000) in (2)
  - Growth: -0.148*** (0.051) in (1); -0.180*** (0.037) in (2)
  - Capital intensity: 0.000 (0.000) in (1); 0.000* (0.000) in (2)
  - Age: -0.028*** (0.002) in (1); -0.017*** (0.002) in (2)
  - Profitability: 0.139*** (0.035) in (1); 0.122*** (0.029) in (2)
  - PLC: 0.025 (0.082) in (1); -0.073 (0.069) in (2)
  - Sector FE: Yes (both columns)

- Main empirical findings:
  - Larger firms (assets and employment) have higher probability of having foreign equity and higher foreign equity share.
  - More open firms have higher probability and higher share of foreign equity.
  - More productive firms have higher probability and higher share of foreign equity.
  - Younger firms (lower age) are more likely to have foreign equity and have higher foreign equity share.
  - More profitable firms have higher probability and higher share of foreign equity.
  - Firms with more tangible assets have lower probability of having foreign equity and lower foreign equity share (consistent with tangible assets enabling higher leverage via collateral).
  - Firms with higher growth exhibit lower probability and lower share of foreign equity (consistent with literature arguing fast-growing firms accumulate more debt).
  - No significant effect found for capital intensity (except a marginal 0.000* in column (2)) or being publicly-listed (PLC not significant).

- Additional considerations:
  - Potential concern: correlation between firm size/productivity and decisions to borrow in foreign currency could confound interpretation (debt currency composition vs. debt vs. equity financing).
  - Aggregate evidence for Slovenia: share of external debt denominated in foreign currency (after Euro introduction) is less than one percent (Figure B.4), suggesting corporate sector’s short-term foreign-currency debt volume is relatively small and unlikely to drive main results.

### Conclusion (implications from section)
- At the firm level, a positive foreign equity share is associated with greater resilience to external shocks (firms with positive foreign equity share performed better during the global financial crisis and were more likely to survive — discussed in the paper’s broader findings).  
- Micro-level determinants imply that economies with larger, more open, and more productive firms are likely to exhibit higher foreign equity shares in corporate external liabilities.  
- Policy implication: assessing countries’ vulnerability to sudden changes in the financial account can benefit from information on the foreign funding structure of the corporate sector. Readily available aggregate characteristics (size, productivity, age, openness) combined with aggregate net foreign asset statistics can provide useful indications of corporate-sector vulnerability where firm-level foreign-liability data are unavailable.

*Determinants and Effects of Countries’ External Capital Structure: A Firm-Level Analysis — Working Paper No. WP/22/38*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022038-print-pdf.pdf_