## 9. Results: Impact of Leverage on Profitability

## Source details

**Canonical URL:** [9. Results: Impact of Leverage on Profitability](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022056-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022056-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022056-print-pdf.pdf.json)

---

### Introductory findings and motivation
- SOEs provide indispensable public goods, especially in natural monopolies, but often suffer poor governance, weak controls and operational inefficiencies that impose high costs on the sovereign via subsidies, liquidity provision, support payments and recapitalizations.
- SOEs frequently accumulate high debt loads that constitute contingent liabilities for government; many SOEs are not included in central government accounts and their losses and unsustainable debt can go unnoticed until large support is required.
- Literature gap: few cross-country studies assess SOE financial viability using firm-level financials; this paper fills the gap for Sub-Saharan Africa (SSA).
- Coverage context:
  - Dataset assembled covers SOEs in 35 of the 45 SSA countries.
  - Baum and others (2020) report that non-financial public enterprises are covered in less than 10 percent of the IMF’s 45 SSA countries (contextual citation in source).

### Data collection and sample composition
- Dataset scope and exclusions:
  - Financial data for 287 SOEs in which government holds majority shares.
  - Financial-sector SOEs (bank and non-bank) excluded.
- Country coverage and missing data:
  - SOE information captured for 35 SSA countries; requisite data not available for 10 countries: Democratic Republic of Congo, Equatorial Guinea, Eritrea, Eswatini, Gabon, Guinea, Malawi, Sierra Leone, South Sudan, and Togo.
- Data sources and variables collected:
  - Primary sources: SOE-published audited financial statements or country-authority reports.
  - Collected items include: revenue, expenses, depreciation and amortization, net interest expense, profit before and after taxes, net taxes paid, total assets, total liabilities, equity, current assets and current liabilities, cash and cash equivalents.
  - Derived ratios: EBITDA, return on (average) assets (ROA), leverage (debt-to-asset ratio), liquidity (current assets / current liabilities).
- Sample design to reduce bias:
  - Full sample is unbalanced and exhibits country and small-firm biases; a core sample was created with up to three largest SOEs per country (measured primarily by assets; in doubt by revenue).
  - The reduced (core) sample results in two-thirds of the countries having exactly 3 SOEs per country.

### Key descriptive statistics and performance metrics
- Data vintage:
  - In both full and core samples, about half of firm records are from 2017 and 2018; some records from earlier years remained the latest available.
- Sample composition (counts and shares):
  - Full sample: 287 SOEs; core sample: 89 SOEs.
  - PRIMARY: full sample 53 firms (18%); core sample 13 firms (15%).
  - ENERGY: full sample 49 firms (17%); core sample 30 firms (34%).
    - Electricity: full sample 33 firms (11%); core sample 23 firms (27%).
  - TRANSPORT: full sample 77 firms (27%); core sample 28 firms (31%).
    - Port authorities: full sample 29 firms (10%); core sample 12 firms (14%).
    - Air: full sample 27 firms (9%); core sample 13 firms (15%).
  - SERVICES: full sample 108 firms (38%); core sample 18 firms (20%).
- Profitability, liquidity, leverage summary:
  - About 40 percent of SOEs show negative profitability in both the full and the core sample.
  - Many firms have liquidity ratios well above 1 in both samples, but in the core sample half of the SOEs show ratios below the threshold of 1.
  - In the full sample many (small) firms display low leverage.
  - In the core sample, more than half of the larger firms have a capital-to-assets ratio of less than 20 percent (i.e., debt-to-assets ratio greater than 0.8) or are technically insolvent with debt exceeding assets.

### Correlation analysis among performance variables
- Leverage vs profitability:
  - Moderately negative correlation in both samples.
  - Firms with negative capital or a capital ratio < 20 percent (debt-to-assets > 0.8) tend to be loss-making (negative ROA).
- Liquidity vs profitability:
  - Moderately positive correlation in the full sample and stronger positive correlation in the core sample.
  - In the core sample, almost all firms with a liquidity ratio > 1 are profit-making.
- Leverage vs liquidity:
  - Moderately to strongly negative correlation in both samples.
  - Interpretation: illiquid enterprises resort to short-term trade and bank financing, leading to build-up of liabilities and leverage over time.

### Debt sustainability metric and core-sample econometric highlights
- Metric and thresholds:
  - Net debt-to-EBITDA ratio (DTE) used to measure viability; includes all SOE liabilities in debt (accounts payable included).
  - Typical thresholds: net debt not exceeding 5 times EBITDA commonly used; for capital-intensive industries a threshold of 7 times EBITDA may be considered.
- Core-sample DTE findings:
  - More than two-thirds of large SOEs in SSA display unsustainable debt per DTE5:
    - 64 of 89 firms, or 72 percent, display unsustainable debt defined as either DTE > 5 (40 cases) or DTE < 0 because of negative EBITDA (24 cases).
  - If critical value raised to 7 (DTE7):
    - Share of SOEs with unsustainable debt decreases to 64 percent: 57 out of 87 firms.
    - Seven firms have a DTE ratio between 5 and 7 (hence counted as unsustainable under DTE7 adjustment).
- Interest coverage ratio (ICR) not used due to insufficient coverage of interest-payment subcomponents.

### Econometric approach and main regression results (core sample unless noted)
- Estimation method and variables:
  - Probit estimation with robust standard errors; dependent variable DTE equals one if debt-to-EBITDA > 5 (alternatively, 7) or is negative, and zero otherwise.
  - Time averaging: two-year averages of explanatory variables; five-year averages used in robustness checks.
  - Firm-level explanatory variables: ROA, LIQUID (current assets / current liabilities); leverage excluded due to endogeneity concerns.
  - Macroeconomic variables: RGDPG, REER (change), CURACT, PSCRED, FISBAL, INFLTN.
  - Governance indicators (World Bank WGI): POLSTB, GOVEFF, REGQTY, RLAW, CORRPT.
- Table 2 summary statistics (selected exact entries):
  - DTE5 mean 0.72, Std. Dev. 0.45, Minimum 0, Maximum 1.
  - DTE7 mean 0.64, Std. Dev. 0.48, Minimum 0, Maximum 1.
  - ROA mean 0.68, Std. Dev. 23.65, Minimum -66.1, Maximum 181.5.
  - LIQUID mean 1.40, Std. Dev. 1.64, Minimum 0.10, Maximum 8.97.
  - RGDPG mean 3.52, Std. Dev. 2.52, Minimum -2.50, Maximum 10.35.
  - REER mean -0.18, Std. Dev. 5.53, Minimum -19.52, Maximum 9.06.
  - CURACT mean -8.10, Std. Dev. 9.12, Minimum -39.65, Maximum 7.75.
  - PSCRED mean 8.71, Std. Dev. 9.55, Minimum -16.04, Maximum 51.31.
  - FISBAL mean -6.83, Std. Dev. 4.54, Minimum -20.68, Maximum -0.33.
  - INFLTN mean 5.52, Std. Dev. 5.79, Minimum -1.98, Maximum 24.73.
  - POLSTB mean -0.36, Std. Dev. 0.79, Minimum -2.03, Maximum 1.02.
  - GOVEFF mean -0.61, Std. Dev. 0.65, Minimum -1.77, Maximum 0.93.
  - REGQTY mean -0.54, Std. Dev. 0.56, Minimum -1.78, Maximum 1.01.
  - RLAW mean -0.53, Std. Dev. 0.60, Minimum -1.77, Maximum 0.72.
  - CORRPT mean -0.49, Std. Dev. 0.69, Minimum -1.56, Maximum 0.91.
- Main regression findings (DTE5 dependent):
  - Firm-level:
    - ROA is highly significant and negatively associated with probability of unsustainable debt (regression (1): coefficient -0.190, significance at the 1 percent level).
    - LIQUID is significant and negatively associated (regression (1): coefficient -0.189, significance at the 5 percent level).
  - Macroeconomic covariates: generally non-significant or only marginally significant (change in REER and RGDPG close to the 10 percent level in regression (2)).
  - Governance: REGQTY significant at the 5 percent level in isolation (regression (2): REGQTY coefficient -0.633, significance at the 5 percent level); other governance indicators generally not significant in isolation.
  - Model fit and sample sizes (Table 4): No. firms ranges from 80 to 89; No. countries ranges from 32 to 35; Pseudo R2 reported across models: 0.38, 0.05, 0.02, 0.02, 0.06, 0.03, 0.02.

### Combined specifications, sector effects, and model fit (Table 5)
- REGQTY with ROA:
  - Regression (1): ROA coefficient -0.166***, Pseudo R2 0.37.
  - Regression (2): REGQTY coefficient -0.719**, Pseudo R2 0.41.
- Adding LIQUID improves fit:
  - Regression (3): ROA -0.179***, LIQUID -0.218**, REGQTY -0.811**, No. firms 80, Pseudo R2 0.44.
- Adding macro variables yields marginal improvement:
  - Regression (4): Pseudo R2 0.47.
- Sector dummies:
  - PRIMARY is strongly positive and significant (regressions (5)–(7): PRIMARY coefficient 5.836*** and 6.223*** in shown columns), indicating primary-sector firms more likely to be overindebted.
  - TRANSPT shows negative coefficients (e.g., -0.781*), suggesting transportation may be associated with lower probability of overindebtedness in some specifications.
- Overall Pseudo R2 in Table 5 ranges from 0.37 to 0.55 across specifications.

### Sensitivity checks and robustness (Table 6)
- Using higher threshold DTE7:
  - ROA and LIQUID remain negatively associated and generally significant though at somewhat lower significance levels.
  - REGQTY remains negative and significant in some specifications (e.g., -0.497, -0.517*, -0.517*).
  - Sector dummies lose robustness in some regressions for DTE7.
  - Pseudo R2 for DTE7 regressions: 0.37, 0.37, 0.38.
- Using five-year averages (with DTE5):
  - Macro variables remain generally non-significant overall.
  - In regression (6) with sector dummies, GDP growth and REER unexpectedly turn significant with the wrong sign, but dropping PRIMARY in (7) restores previous significance patterns.
  - Pseudo R2 for 5-yr avg specifications: 0.41, 0.48, 0.63, 0.52 (columns (4)–(7) in Table 6).

### Subsample analyses: income status (Table 7)
- Middle-income countries (MICs):
  - ROA remains robust and significant (e.g., -0.213***).
  - REGQTY and LIQUID lose significance in MIC subsample.
  - CORRPT becomes significant at the 5 percent level (e.g., CORRPT -0.784** in MICs).
  - REER and FISBAL gain significance in MICs (REER -0.252**, FISBAL -0.196**).
  - No. firms: MICs No. firms 45 (or 43 in some columns); No. countries 35 or 32 depending on column.
  - Pseudo R2: MICs 0.37–0.54 across models.
- Low-income countries (LICs):
  - LIQUID and REGQTY matter for LICs in some specifications.
  - PRIMARY sector dummy is significant for LICs (PRIMARY 7.659** when included).
  - No. firms: LICs No. firms 44 (or 37); No. countries 35, 32.
  - Pseudo R2: LICs 0.39–0.72 across models.

### Subsample analyses: resource intensity (Table 8)
- Resource-intensive countries:
  - REGQTY and INFLTN show strong linkages (e.g., REGQTY -1.745**, INFLTN -0.232***).
  - LIQUID turns insignificant in some specifications.
  - PRIMARY sector dummy significant (PRIMARY 7.274***).
  - No. firms 40; No. countries 17; Pseudo R2 up to 0.80 in some specifications.
- Non-resource-intensive countries:
  - ROA remains significant and negative (e.g., -0.130***).
  - REGQTY generally not significant for non-resource-intensive countries.
  - Pseudo R2 ranges 0.27–0.42.

### Impact of leverage on profitability (Table 9)
- Estimation method: cross-sectional OLS with robust standard errors; dependent variable DV: ROA.
- Key finding:
  - LEVER (average leverage ratio, debt-to-assets) is highly significant at the 1 percent level and negatively associated with ROA in core and full samples, confirming that highly leveraged firms tend to be less profitable.
    - Core sample regressions (1)–(3): LEVER coefficients -14.443***, -15.049***, -14.549***.
    - Full sample regressions (4)–(6): LEVER coefficients 2.691***, 2.630***, -14.041*** (coefficients vary across models and samples as reported).
  - LIQUID is not significant throughout.
- Occasional macro significance:
  - RGDPG sometimes significant (e.g., 2.023* in core sample; 1.626** in full sample).
  - INFLTN significant in full sample (0.364**).
- Sector TRANSPT shows negative association with ROA in some specifications (e.g., -8.089* and -10.653** in core regressions).
- Sample sizes and fit:
  - No. firms: core 84 (or 79); full sample 271 (or 218); No. countries 33 (core) or 32 (full).
  - R2 ranges: core 0.09–0.21; full 0.09–0.19.

### Macrofinancial implications
- Two macrofinancial linkage channels from SOE underperformance to the financial sector:
  - Direct linkage: loan exposures of banks and other lenders to SOEs; overindebted and illiquid SOEs risk defaulting and may trigger invocation of government guarantees.
  - Indirect linkage: illiquid SOEs accumulate arrears to suppliers, potentially causing suppliers to default on their bank loans, raising NPLs.
- Evidence on arrears and liquidity:
  - In three out of five countries reporting arrears from SOEs, large SOEs in the core sample had below-average liquidity ratios ranging from 74 to 105 percent against an average of 139 percent for the entire core sample.
  - Including countries where arrears to SOEs were recorded (assuming pass-through to suppliers), the average SOE had below average liquidity ratio in eight out of 13 countries, with a range of 37 to 115 percent; group means were almost identical.
- Visibility and provisioning:
  - Supervisory data such as NPLs to SOEs are rarely disclosed.
  - In stress tests, additional provisioning for SOE exposures may be reasonable because delayed payments lower banks' cash inflows and may affect liquidity and profitability.
  - Whether banks must provision depends on supervisory resolve and regulation; some frameworks may exempt provisioning due to SOE public sector identity or state guarantees.
  - Prior stress-test studies indicate explicit provisioning for SOE-related risks can significantly affect bank capitalization (examples cited in source).

### Stress testing, reporting, and transparency for SOEs (Box 1)
- Stress testing:
  - IMF stress-testing methodology can be applied to SOEs; the notion of long-run viability may need to be nuanced given state ownership.
- Transparency enhancements recommended:
  - Publication (not just submission to auditors) of SOEs’ annual reports, including detailed financial statements (balance sheet, income statement, cash flow statement).
  - Benchmarking outcomes to pre-defined budget targets or performance objectives and comprehensive reporting of SOE performance.
  - Application of new IMF benchmarking methodology for SOEs (benchmarks derived from quartile distributions of four financial indicators for about 22,000 SOEs worldwide) may increase accountability and help governments identify poor performance.
- Country examples of adverse macrofinancial linkages in SSA (selected cases summarized):
  - The Gambia: NAWEC losses, bank borrowing, restructure into state-guaranteed bond, default, guarantee called; trade arrears to NAWEC amounted to 0.6 percent of GDP in a 2019 audit.
  - São Tomé and Príncipe: SOE payment arrears emerged amounting to close to 20 percent of GDP.
  - Ghana (2013-14): energy-sector SOEs faced cash-flow difficulties; a quarter of bank loans to the energy sector became non-performing.
  - Mozambique: SOEs turned to bank financing; IMF simulation showed restructuring/provisioning of an SOE’s debt overhang significantly reduced banks’ profitability and capitalization.
  - Cross- and intra-sectoral claims and arrears can propagate macrofinancial stress by exposing banks to credit and liquidity risks.

### Key empirical findings, conclusions, and policy implications
- Sample and scope:
  - The study examined close to 300 firms in 35 of the 45 SSA countries, condensed to a core sample of about 90 large firms to reduce country bias.
- Performance outcomes:
  - Around 40 percent of SOEs are unprofitable.
  - The majority of the large firms in the core sample are illiquid and overleveraged.
- Drivers and interpretation:
  - Firm performance (ROA) and the liquidity ratio are the main drivers of debt sustainability; macroeconomic variables have marginal impact at best.
  - Institutional setting (quality of policies and regulations, REGQTY) affects SOE debt sustainability in several specifications.
  - For SOE profitability, high leverage is the main driver of low or negative profits; liquidity cannot be shown to have an effect.
  - Some macro factors matter for specific country groups (e.g., MICs and resource-intensive countries).
  - Regression fit improves when accounting for firm-level characteristics, but idiosyncratic or hard-to-measure determinants (restrictive rules, high overhead costs, dead-weight losses) are not captured.
- Policy implications and recommendations:
  - SOE reforms should address operational and governance deficiencies, pricing, payments, and performance discipline.
  - Governments should adopt comprehensive reporting on SOE performance and benchmark outcomes explicitly to pre-defined budgets or performance objectives.
  - Greater transparency and publication of SOE financials and aggregate SOE reports can help the public and policymakers assess sector performance and fiscal risks.

### Annex I — full-sample takeaways (high-level)
- Full sample: 287 SOEs initially; regressions run on up to 276 firms depending on model due to missing firm-level data.
- Bias sources: (i) a few countries dominate (one country having close to 50 observations); (ii) services sector represents close to 40 percent of the sample; (iii) many smaller firms included.
- Selected full-sample regression notes:
  - REGQTY coefficients reported: -0.306 (0.152) **; -0.353 (0.169) **; -0.415 (0.218) * across models.
  - ROA reported coefficients: -0.019 (0.015); -0.020 (0.015); -0.162 (0.029) *** (ROA regains significance when liquidity is added in model (3)).
  - LIQUID reported as -0.008 (0.009) in one specification.
  - Macro variables generally insignificant.
  - ENERGY shows significance in some specifications (e.g., 0.387 (0.224) * and 0.600 (0.224) ***), indicating energy firms can be more overindebted.
  - No. firms across full-sample models: 276, 276, 219, 276, 276, 276, 276; No. countries: 35, 35, 32, 35, 35, 35, 35.
  - Pseudo R2 values for full-sample models: 0.09, 0.10, 0.38, 0.09, 0.11, 0.08, 0.10.
- Interpretation caution:
  - Swings in significance across specifications reflect sample bias and missing firm-level data; conclusions are more robust when drawn from the balanced core sample of mostly large, systemic enterprises.

*Source: IMF Working Paper chapter "9. Results: Impact of Leverage on Profitability" (dataset and analysis described within the chapter).*

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

### wpiea2022056-print-pdf - References .............................................................................................................

### Annexes and Boxes
- Annex I. Estimation Results for Debt Sustainability Using the Full Sample ................................................................................30
- Box 1. Country Examples of Adverse Macrofinancial Linkages in SSA ................................................................................25

### Figures
- Figure 1. Distribution of Number of SOEs per Country .......................................................................8
- Figure 2. Market Concentration and Small Firm Bias in Full Sample ..................................................9
- Figure 3. Distribution of Units Across Years and Sectors ....................................................................9
- Figure 4. Distribution of Return on Assets..........................................................................................10
- Figure 5. Distribution of Liquidity Ratio ............................................................................................11
- Figure 6. Distribution of Leverage Ratio ............................................................................................11
- Figure 7. Correlation between Leverage and Profitability ..................................................................12
- Figure 8. Correlation Between Liquidity and Profitability .................................................................13
- Figure 9. Correlation between Leverage and Liquidity ......................................................................13
- Figure 10. Distribution of Debt-to-EBITDA Ratio .............................................................................14

### Tables
- Table 1. Breakdown of Units by Sub-Sector .....................................................................................10
- Table 2. Summary Statistics, Data Sources and Definitions of Variables .........................................15
- Table 3. Correlation Matrix ...............................................................................................................16
- Table 4. Results: Determinants of SOE Debt Sustainability .............................................................17
- Table 5. Results: Determinants of SOE Debt Sustainability .............................................................18
- Table 6. Sensitivity Checks Using Alternative Variable Definitions ................................................19
- Table 7. Results by Income Status .....................................................................................................20
- Table 8. Results by Resource Intensity ..............................................................................................21

*Source: wpiea2022056-print-pdf - References.*

### 9. Results: Impact of Leverage on Profitability .......................................................................2

### 9. Results: Impact of Leverage on Profitability

### I. Introductory findings and motivation
- SOEs provide indispensable public goods, especially in natural monopolies, but often suffer poor governance, weak controls and operational inefficiencies that impose high costs on the sovereign via subsidies, liquidity provision, support payments and recapitalizations.
- SOEs frequently accumulate high debt loads that constitute contingent liabilities for government; many SOEs are not included in central government accounts and their losses and unsustainable debt can go unnoticed until large support is required.
- Literature gap: few cross-country studies assess SOE financial viability using firm-level financials; the paper fills this gap for Sub-Saharan Africa (SSA).
- Coverage context:
  - Dataset assembled covers SOEs in 35 of the 45 SSA countries.
  - Baum and others (2020) report that non-financial public enterprises are covered in less than 10 percent of the IMF’s 45 SSA countries (cited context from literature).

### II. Data collection and sample composition
- Dataset scope:
  - Financial data for 287 SOEs in which government holds majority shares.
  - SOEs excluded from dataset: bank and non-bank financial sectors (due to inherently high leverage and inapplicability of standard debt-sustainability concepts).
- Country coverage and data availability:
  - SOE information captured for 35 SSA countries; requisite data not available for 10 countries: Democratic Republic of Congo, Equatorial Guinea, Eritrea, Eswatini, Gabon, Guinea, Malawi, Sierra Leone, South Sudan, and Togo.
- Data sources and items:
  - Primary sources: SOE-published audited financial statements or country-authority reports (sometimes provided upon request).
  - Collected variables include: revenue, expenses, depreciation and amortization, net interest expense, profit before and after taxes, net taxes paid, total assets, total liabilities, equity, current assets and current liabilities, cash and cash equivalents.
  - Derived key performance ratios: EBITDA, return on (average) assets (ROA), leverage (debt-to-asset ratio), liquidity (current assets / current liabilities).
- Sample design to reduce bias:
  - Full sample is unbalanced and exhibits country and small-firm biases; created a core sample with up to three largest SOEs per country (measured primarily by assets; in doubt by revenue) to reduce bias.
  - The reduced (core) sample results in two-thirds of the countries having exactly 3 SOEs per country.

### III. Key descriptive statistics and performance metrics
- Data vintage:
  - In both full and core samples, about half of firm records are from 2017 and 2018; some records from earlier years remained the latest available.
- Sectoral composition (summary of counts and % sample from Table 1):
  - Sum total: 287 SOEs (full sample); core sample: 89 SOEs.
  - Primary sector: full sample 53 firms (18%); core sample 13 firms (15%).
  - Energy sector: full sample 49 firms (17%); core sample 30 firms (34%).
    - Electricity: full sample 33 firms (11%); core sample 23 firms (27%).
  - Transport sector: full sample 77 firms (27%); core sample 28 firms (31%).
    - Port authorities: full sample 29 firms (10%); core sample 12 firms (14%).
    - Air: full sample 27 firms (9%); core sample 13 firms (15%).
  - Services sector: full sample 108 firms (38%); core sample 18 firms (20%).
  - Sum total reiterated: 287 firms = 100% (full sample); core sample 89 firms = 100%.
- Profitability (ROA):
  - About 40 percent of SOEs show negative profitability in both the full and the core sample.
- Liquidity (current ratio):
  - Many firms have ratios well above 1 in both samples, but in the core sample half of the SOEs show ratios below the threshold of 1.
- Leverage (debt-to-assets ratio):
  - In the full sample many (small) firms display low leverage.
  - In the core sample, more than half of the larger firms have a capital-to-assets ratio of less than 20 percent (i.e., debt-to-assets ratio greater than 0.8) or are technically insolvent with debt exceeding assets.

### IV. Correlation analysis among performance variables
- Pair-wise correlations (full and core samples):
  - Leverage vs profitability:
    - Moderately negative correlation in both samples.
    - Firms with negative capital or a capital ratio < 20 percent (debt-to-assets > 0.8) tend to be loss-making (negative ROA).
  - Liquidity vs profitability:
    - Moderately positive correlation in the full sample and stronger positive correlation in the core sample.
    - In the core sample, almost all firms with a liquidity ratio > 1 are profit-making.
  - Leverage vs liquidity:
    - Moderately to strongly negative correlation in both samples.
    - Interpretation: illiquid enterprises resort to short-term trade and bank financing, leading to build-up of liabilities and leverage over time.

### V. Debt sustainability metric and econometric findings (core sample highlights)
- Debt sustainability metric:
  - Net debt-to-EBITDA ratio (DTE) used to measure viability of SOE debt stock relative to operative earnings.
  - Definition note: includes all SOE liabilities in debt, including accounts payable (owing to prevalence of domestic arrears in SSA).
  - Typical thresholds:
    - Debt not exceeding 5 times EBITDA commonly used as critical threshold.
    - For capital-intensive industries a threshold of net debt 7 times EBITDA may be considered.
- Core-sample results:
  - More than two-thirds of large SOEs in SSA display unsustainable debt per DTE5:
    - 64 of 89 firms, or 72 percent, display unsustainable debt defined as either DTE > 5 (40 cases) or DTE < 0 because of negative EBITDA (24 cases).
  - If critical value raised to 7 (DTE7) to accommodate capital intensity and measurement error:
    - Share of SOEs with unsustainable debt decreases to 64 percent: 57 out of 87 firms.
    - Seven firms have a DTE ratio between 5 and 7 (hence counted as unsustainable under DTE7 adjustment).
- Interest coverage ratio (ICR) not used:
  - ICR (EBIT / net interest payments) typically has critical values around 1.5 to 1.0, but the study does not compute ICR due to insufficient coverage of interest-payment subcomponents in many reports.

### VI. Caveats and limitations (as stated in the source)
- Indicator coverage limitations:
  - Study focuses on a restricted set of financial performance indicators that are consistently published across SOEs in the region; efficiency-oriented indicators (operating costs relative to turnover, labor costs relative to revenue, turnover per employee) could not be compiled consistently and are excluded.
- Unobserved external supports:
  - Lack of data prevents controlling for external, largely unobserved factors influencing financial performance such as explicit government subsidies and transfers, recapitalizations, guarantees, or policies keeping input prices artificially low.
- Sample completeness:
  - Dataset is comprehensive but not claimed to cover the entire universe of SOEs in SSA; likely captures the largest or most important SOEs in reporting countries but completeness varies by country.

*Source: IMF Working Paper chapter "9. Results: Impact of Leverage on Profitability" (dataset and analysis described within the chapter).*

### Section IV).

### Section IV)

### Distribution of Debt-to-EBITDA
- Figure 10 shows the full distribution of debt-to-EBITDA ratios; extreme ratio readings (e.g., greater than 100 or smaller than -100) can arise from the denominator being relatively close to zero.
- DTE5 definition used as reference point; red bars indicate unsustainable debt-to-EBITDA ratios.
- Where EBITDA was positive but net debt was smaller than zero (cash holdings exceeding liabilities), the (net) debt-to-EBITDA ratio was manually set to zero, indicating sustainable debt.

### Econometric approach and explanatory variables
- Estimation method: probit estimation with robust standard errors; dependent variable DTE assumes value one if debt-to-EBITDA ratio is either greater than 5 (alternatively, 7) or is negative, and zero otherwise.
- Time averaging: two-year averages of explanatory variables (year of balance sheet date and previous year); five-year averages used in robustness checks.
- Firm-specific variables:
  - ROA: Return on (average) assets (profitability indicator).
  - LIQUID: Current Assets to Current Liabilities (liquidity indicator).
  - Leverage ratio (debt-to-assets) excluded from probit regressions due to possible endogeneity.
- Macroeconomic variables: RGDPG, REER (change), CURACT, PSCRED, FISBAL, INFLTN.
- Governance indicators (World Bank WGI): POLSTB, GOVEFF, REGQTY, RLAW, CORRPT.
- Table 2 summary statistics (selected entries preserved exactly):
  - DTE5 mean 0.72, Std. Dev. 0.45, Minimum 0, Maximum 1 (Debt-to-EBITDA, Threshold 5 times).
  - DTE7 mean 0.64, Std. Dev. 0.48, Minimum 0, Maximum 1 (Debt-to-EBITDA, Threshold 7 times).
  - ROA mean 0.68, Std. Dev. 23.65, Minimum -66.1, Maximum 181.5 (Return on Average Assets).
  - LIQUID mean 1.40, Std. Dev. 1.64, Minimum 0.10, Maximum 8.97 (Current Assets to Current Liabilities).
  - RGDPG mean 3.52, Std. Dev. 2.52, Minimum -2.50, Maximum 10.35 (Real GDP growth).
  - REER mean -0.18, Std. Dev. 5.53, Minimum -19.52, Maximum 9.06 (Real Effective Exchange Rate, change).
  - CURACT mean -8.10, Std. Dev. 9.12, Minimum -39.65, Maximum 7.75 (Current Account, change).
  - PSCRED mean 8.71, Std. Dev. 9.55, Minimum -16.04, Maximum 51.31 (Credit to Private Sector, growth).
  - FISBAL mean -6.83, Std. Dev. 4.54, Minimum -20.68, Maximum -0.33 (Fiscal Balance, change).
  - INFLTN mean 5.52, Std. Dev. 5.79, Minimum -1.98, Maximum 24.73 (Inflation Rate).
  - POLSTB mean -0.36, Std. Dev. 0.79, Minimum -2.03, Maximum 1.02 (Perceived likelihood of political instability).
  - GOVEFF mean -0.61, Std. Dev. 0.65, Minimum -1.77, Maximum 0.93 (Perceived quality of public services).
  - REGQTY mean -0.54, Std. Dev. 0.56, Minimum -1.78, Maximum 1.01 (Ability to pursue sound policies/regulations).
  - RLAW mean -0.53, Std. Dev. 0.60, Minimum -1.77, Maximum 0.72 (Confidence in contracts/property rights).
  - CORRPT mean -0.49, Std. Dev. 0.69, Minimum -1.56, Maximum 0.91 (Use of public power for private gain).

### Main regression findings (DTE5 as dependent variable)
- Firm-level variables:
  - ROA is highly significant and negatively associated with probability of unsustainable debt (regression (1): coefficient -0.190, significance at the 1 percent level).
  - LIQUID is significant and negatively associated (regression (1): coefficient -0.189, significance at the 5 percent level).
- Macroeconomic covariates: generally non-significant or only marginally significant (change in REER and RGDPG close to the 10 percent level in regression (2)).
- Governance indicators: REGQTY is significant at the 5 percent level in isolation (regression (2): REGQTY coefficient -0.633, significance at the 5 percent level); other governance indicators generally not significant in isolation.
- Model fit and sample sizes (Table 4):
  - No. firms ranges from 80 to 89; No. countries ranges from 32 to 35.
  - Pseudo R2 reported across models: 0.38, 0.05, 0.02, 0.02, 0.06, 0.03, 0.02 (corresponding to regressions (1)–(7) in Table 4).

### Combined estimations and sector effects (Table 5)
- Inclusion of REGQTY with ROA:
  - Regression (1): ROA coefficient -0.166***, Pseudo R2 0.37.
  - Regression (2): REGQTY coefficient -0.719**, Pseudo R2 0.41.
- Adding LIQUID improves fit:
  - Regression (3): ROA -0.179***, LIQUID -0.218**, REGQTY -0.811**, No. firms 80, Pseudo R2 0.44.
- Adding macro variables yields marginal improvement:
  - Regression (4): Pseudo R2 0.47.
- Sector dummies:
  - PRIMARY is strongly positive and significant (regressions (5)–(7): PRIMARY coefficient 5.836*** and 6.223*** in shown columns), suggesting firms in the primary sector are more likely to be overindebted.
  - TRANSPT shows negative coefficients (e.g., -0.781*), suggesting transportation may be associated with lower probability of overindebtedness in some specifications.
- Overall Pseudo R2 in Table 5 ranges from 0.37 to 0.55 across specifications.

### Sensitivity checks (Table 6)
- Using higher threshold DTE7:
  - ROA and LIQUID remain negatively associated and generally significant though at somewhat lower significance levels.
  - REGQTY remains negative and significant in some specifications (e.g., -0.497, -0.517*, -0.517*).
  - Sector dummies lose robustness in some regressions (sector dummies lose significance in regression (3) for DTE7).
  - Pseudo R2 for DTE7 regressions: 0.37, 0.37, 0.38.
- Using five-year averages (while using DTE5):
  - Macro variables remain generally non-significant overall.
  - In regression (6) with sector dummies, GDP growth and REER unexpectedly turn significant with the wrong sign, but dropping PRIMARY in (7) restores previous significance patterns.
  - Pseudo R2 for 5-yr avg specifications: 0.41, 0.48, 0.63, 0.52 (columns (4)–(7) in Table 6).

### Subsample analyses: Income status (Table 7)
- Middle-income countries (MICs):
  - ROA remains robust and significant (e.g., -0.213***).
  - REGQTY and LIQUID lose significance in MIC subsample.
  - CORRPT becomes significant at the 5 percent level (e.g., CORRPT -0.784** in MICs).
  - REER and FISBAL gain significance in MICs (REER -0.252**, FISBAL -0.196**), indicating sensitivity to real depreciations and fiscal deficits.
- Low-income countries (LICs):
  - LIQUID and REGQTY matter for LICs (liquidity significant in some specifications).
  - PRIMARY sector dummy is significant for LICs (PRIMARY 7.659** when included).
- No. firms and No. countries: MICs No. firms 45 (or 43 in some columns), LICs No. firms 44 (or 37 in some columns); No. countries 35, 32 respectively.
- Pseudo R2: MICs 0.37–0.54; LICs 0.39–0.72 across models.

### Subsample analyses: Resource intensity (Table 8)
- Resource-intensive countries:
  - Strong linkages to REGQTY (e.g., REGQTY -1.745**) and to inflation (INFLTN -0.232***), with REGQTY and INFLTN highly significant.
  - LIQUID turns insignificant in some specifications for resource-intensive sample.
  - PRIMARY sector dummy significant (PRIMARY 7.274***).
  - No. firms 40; No. countries 17; Pseudo R2 up to 0.80 in some specifications.
- Non-resource-intensive countries:
  - ROA remains significant and negative (e.g., -0.130***).
  - REGQTY generally not significant for non-resource-intensive countries.
  - Pseudo R2 for non-resource-intensive subsample ranges 0.27–0.42.

### Impact of leverage on profitability (Table 9)
- Estimation method: cross-sectional OLS with robust standard errors; dependent variable DV: ROA.
- Key result: LEVER (average leverage ratio, debt-to-assets) is highly significant at the 1 percent level and negatively associated with ROA in core and full samples, confirming that highly leveraged firms tend to be less profitable.
  - Core sample regressions (1)–(3): LEVER coefficients -14.443***, -15.049***, -14.549***.
  - Full sample regressions (4)–(6): LEVER coefficients 2.691***, 2.630***, -14.041*** (note: coefficients vary across models and samples as reported).
- LIQUID is not significant throughout, implying short-term debt overhang does not necessarily affect profitability in these regressions.
- Some macro variables show occasional significance:
  - RGDPG sometimes significant (e.g., 2.023* in core sample; 1.626** in full sample).
  - INFLTN significant in full sample (0.364**).
- Sector dummy TRANSPT shows negative and sometimes significant association with ROA (e.g., -8.089* and -10.653** in core regressions).
- No. firms: core 84 (or 79 in some columns); full sample 271 (or 218); No. countries 33 (core) or 32 (full).
- Pseudo R2 (reported as R2 here) ranges: core 0.09–0.21; full 0.09–0.19.

### Macrofinancial implications
- Two macrofinancial linkage channels from SOE underperformance to the financial sector:
  - Direct linkage: loan exposures of banks and other lenders to SOEs; overindebted and illiquid SOEs risk defaulting and may trigger invocation of government guarantees.
  - Indirect linkage: illiquid SOEs accumulate arrears to suppliers, potentially causing suppliers to default on their bank loans, raising NPLs.
- Evidence from dataset on liquidity and arrears:
  - In three out of five countries reporting arrears from SOEs, large SOEs in the core sample had below-average liquidity ratios ranging from 74 to 105 percent against an average of 139 percent for the entire core sample.
  - Including countries where arrears to SOEs were recorded (assuming pass-through to suppliers), the average SOE had below average liquidity ratio in eight out of 13 countries, with a range of 37 to 115 percent; group means were almost identical.
- Visibility and provisioning:
  - Supervisory data such as NPLs to SOEs are rarely disclosed, so macrofinancial impact on banks is not always visible.
  - In stress tests, additional provisioning for SOE exposures may be reasonable because delayed payments lower banks' cash inflows and may affect liquidity and profitability.
  - Whether banks must provision depends on supervisory resolve and regulation (some frameworks may exempt provisioning due to SOE public sector identity or state guarantees).
- Prior stress-test studies cited indicate explicit provisioning for SOE-related risks can significantly affect bank capitalization (examples referenced: Wezel 2018a; Mansilla et al., 2018).

*Source: Authors’ calculations based on SOE reports (Section IV).*

### Box 1). Similarly, corporate stress tests of SOEs themselves may render vulnerabilities more

### wpiea2022056-print-pdf - Box 1). Similarly, corporate stress tests of SOEs themselves may render vulnerabilities more evident.

### Stress testing, reporting, and transparency for SOEs
- Stress testing methodology developed by the IMF can be applied to SOEs; the notion of long-run viability may need to be nuanced given state ownership.
- Transparency enhancements recommended:
  - Publication (not just submission to auditors) of SOEs’ annual reports, including detailed financial statements (balance sheet, income statement, cash flow statement).
  - Benchmarking outcomes to pre-defined budget targets or performance objectives and comprehensive reporting of SOE performance.
- Application of new IMF benchmarking methodology for SOEs (benchmarks derived from quartile distributions of four financial indicators for about 22,000 SOEs worldwide) may increase accountability of SOE managers and help governments identify lackluster performance.
- Several SSA countries have published aggregate SOE reports (examples listed in source). These reports disclose summary financial statements, ratios, subsidies received, and governance-related information (e.g., auditor name or lack of audit, board composition, CEO name and picture), but typically do not benchmark results against explicit performance goals.

### Country examples of adverse macrofinancial linkages in SSA (selected cases)
- The Gambia
  - NAWEC (state-owned electricity company) accumulated losses, borrowed from banks, loans were restructured into a state-guaranteed bond, NAWEC defaulted, guarantee was called, and NAWEC’s long-term debt was transferred to the government.
  - A 2019 special audit revealed trade arrears to NAWEC amounted to 0.6 percent of GDP, equivalent to the current share of its long-term borrowings and half its trade payables.
- São Tomé and Príncipe
  - SOE payment arrears emerged amounting to close to 20 percent of GDP; banks were exposed directly to SOEs or indirectly through private suppliers; large arrears constitute significant contingent liabilities for the government.
- Ghana (2013-14)
  - Energy-sector SOEs suffered cash flow difficulties from exchange rate depreciation, higher oil prices, and a drought-related switch to higher-cost thermal generation.
  - SOEs borrowed short-term from banks and postponed payments to fuel suppliers; eventually a quarter of bank loans to the energy sector became non-performing.
- Mozambique
  - Cash-strapped SOEs increasingly turned to bank financing to cover operational costs; debt and interest loads ballooned.
  - An IMF simulation showed that a hypothetical restructuring/provisioning of an SOE’s debt overhang (obtained via the excess of the debt-to-EBITDA ratio) led to a significant reduction in banks’ profitability and capitalization.
- Cross- and intra-sectoral claims and arrears can propagate macrofinancial stress by exposing banks to credit and liquidity risks, potentially leading to loan defaults and weakened bank soundness.

### Key empirical findings and conclusions
- Sample and scope:
  - The study examined close to 300 firms in 35 of the 45 SSA countries, condensed to a core sample of about 90 large firms to reduce country bias.
- Performance outcomes:
  - Around 40 percent of SOEs are unprofitable.
  - The majority of the large firms in the core sample are illiquid and overleveraged.
- Econometric analysis on drivers of debt sustainability and profitability:
  - Firm performance, represented by return on assets (ROA) and the liquidity ratio, is the main driver of debt sustainability, largely independent of macroeconomic conditions (macro variables have marginal impact at best).
  - Institutional setting (quality of policies and regulations) affects SOE debt sustainability.
  - For SOE profitability, high leverage is the main driver of low or negative profits; liquidity cannot be shown to have an effect.
  - Some macro factors matter for certain country groups (e.g., middle-income and resource-intensive countries).
  - Some SOEs operate efficiently under difficult economic conditions generating profits and cash flows that preempt excessive borrowing; others perform poorly despite favorable environments.
  - Regression fit improves when accounting for firm-level characteristics, but idiosyncratic or hard-to-measure determinants (restrictive rules, high overhead costs, dead-weight losses) are not captured.
- Policy implications and recommendations:
  - SOE reforms should address operational and governance deficiencies, pricing, payments, and performance discipline.
  - Governments should adopt comprehensive reporting on SOE performance and benchmark outcomes explicitly to pre-defined budgets or performance objectives.
  - Greater transparency and publication of SOE financials and aggregate SOE reports can help the public and policymakers assess sector performance and fiscal risks.

### Annex I — Estimation results for debt sustainability using the full sample (high-level takeaways)
- Full sample composition and biases:
  - Full sample: 287 SOEs initially; regressions run on up to 276 firms depending on model due to missing firm-level data.
  - Bias sources: (i) a few countries dominate (one country having close to 50 observations); (ii) services sector represents close to 40 percent of the sample; (iii) many smaller firms included.
- Selected regression findings (Authors’ calculations; robust standard errors in parentheses; ***, **, * denote significance at the 1, 5, 10 percent level):
  - In parsimonious models for DTE5:
    - REGQTY coefficients reported: -0.306 (0.152) **; -0.353 (0.169) **; -0.415 (0.218) * across models.
    - ROA reported coefficients: -0.019 (0.015); -0.020 (0.015); -0.162 (0.029) *** (ROA regains significance when liquidity is added in model (3)).
    - LIQUID reported as -0.008 (0.009) in one specification (loses significance in full-sample compared to core sample).
  - Macro variables (RGDPG, REER, CURACT, PSCRED, FISBAL, INFLTN) are generally insignificant in these full-sample regressions.
  - Sector dummies:
    - ENERGY shows significance in some specifications: 0.387 (0.224) * and 0.600 (0.224) *** in selected regressions; firms in the energy sector (about one-fifth of the full sample) tend to be more overindebted.
    - PRIMARY loses robustness in the full sample relative to the core sample.
  - Regression sample sizes and fit:
    - No. firms: 276, 276, 219, 276, 276, 276, 276 across models (1)–(7).
    - No. countries: 35, 35, 32, 35, 35, 35, 35 across models (1)–(7).
    - Pseudo R2 values: 0.09, 0.10, 0.38, 0.09, 0.11, 0.08, 0.10 for models (1)–(7).
- Interpretation cautions:
  - Swings in significance (e.g., ROA) when moving between specifications reflect sample bias and missing firm-level data (nearly 60 SOEs drop out in one specification due to missing current assets/liabilities).
  - Results from the “random” full sample should be taken with caution; conclusions are more robust when drawn from the balanced core sample of mostly large, systemic enterprises.

*The Financial Performance and Macrofinancial Implications of Large State-Owned Enterprises in Sub-Saharan Africa — Working Paper No. WP/22/56*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022056-print-pdf.pdf_