## 1. Various Population Projections

## Source details

**Canonical URL:** [1. Various Population Projections](https://www.imf.org/-/media/files/publications/cr/2024/english/1bgrea2024002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1bgrea2024002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1bgrea2024002.pdf.json)

---

### Overview of the Bulgarian pension system
- Structure:
  - First pillar: mandatory, defined benefit, pay-as-you-go system administered by the National Social Security Institute (NSSI); financed by a contribution rate of 14.8 percent of the gross insurable income for individuals born after 1959 and 19.8 percent for individuals born in 1959 or before.
  - Second pillar: mandatory defined contribution system managed by licensed private pension providers for individuals born after 1959; funded by a contribution rate of 5 percent.
  - Third pillar: voluntary defined contribution system managed by licensed private pension providers (Republic of Bulgaria, 2020).
- Self-employed: contribute at the same rate of 14.8/19.8 percent to the first pillar, applied to declared covered earnings for the previous calendar year.

### Past reforms and quantitative effects
- 2015 reform:
  - Gradually increases the retirement age: from 63 and 8 months for men and 60 and 8 months for women to 65 in 2029 for men and in 2037 for women.
  - Raises the required minimum contribution period by two months every year until it reaches 40 years for men and 37 years for women by 2027.
  - Reduced pension spending by 1½ percent of GDP between 2014 and 2019.
- Coverage and adequacy impacts:
  - Coverage ratio (pensioners as a share of population 65+) continued to decline after the reform and is projected to remain below EU average by 2040 (2021 Ageing Report).
  - Expected years in retirement declined more rapidly than for the EU as a whole.
  - Pension support ratio declined markedly after the reform; benefit ratio continued to decline as pension increases lagged wage growth.

### Adequacy versus sustainability: key statistics
- Replacement rates and poverty:
  - The share of pensioners receiving the minimum pension increased from 27 percent in 2016 to 35 percent in 2019; the share reached 46 percent in 2022 and 2023.
  - The minimum pension fell from 57 percent of the minimum wage in 2010 to 38 percent in 2019 and remained well below the poverty line.
  - Until 2022, the elderly (65 and older) was the age group suffering the most from poverty; at-risk-of-poverty measures and median equivalized net incomes show elderly disposable income significantly lower than other age groups.
- Bulgaria’s pension spending:
  - On the eve of the pandemic, Bulgaria’s spending on pensions was more than 5 percent of GDP lower than the EU average.
  - Bulgaria is characterized as a low contribution - low benefit system resulting in relatively limited spending.

### COVID-19 pandemic measures and fiscal impact on pensions
- Pandemic-era increases:
  - More than half of the fiscal support provided to households in 2020–21 targeted pensioners.
  - Pandemic household support accounted for over 1 percent of GDP in 2021.
  - Support mainly took the form of pension increases and ad-hoc supplementary payments (bonuses); some measures planned as temporary were made permanent in 2022.
- Fiscal mismatch and indicators of imbalance:
  - Contribution rate and maximum insurable income were left unchanged during the pandemic.
  - Share of contributors subject to the cap on social contributions increased from 6.4 percent in 2019 to 14.2 percent in 2023.
  - Contributions at maximum social insurable income represent over 21 percent of total contributions (up from 13.6 percent in 2019).
  - Maximum pension relative to maximum insurable income:
    - Over 2009–19: maximum pension was set at 35 percent of maximum insurable income.
    - 2020: 40 percent.
    - 2021: 48 percent.
    - 2022: 62 percent.
    - 2023: 100 percent.
    - Increase in 2024 will bring this ratio to only 91 percent.
  - Share of pensioners receiving the maximum pension fell from 2 percent in 2020 to 0.1 percent in 2023.
- Net effect:
  - Contributions increasingly capped, constraining system revenue, while the cap on pensions has effectively disappeared, increasing spending.
  - In 2023 the full impact of pandemic-era measures is visible: old-age poverty dropped markedly but pension spending rebounded strongly.
- Additional 2024 supplement:
  - The Easter supplement of BGN 100 granted to pensioners whose pension is below the official poverty line of BGN 526 per month has an estimated cost of 0.03 percent of GDP; 2024 financial data do not include this impact.

### Trade-offs, risks, and projection drivers
- Trade-offs observed:
  - Reforms in the 2010s improved financial sustainability at the cost of reduced pension adequacy.
  - COVID-era measures improved adequacy but materially worsened financial sustainability and increased fiscal risk because the state is legally required to transfer funds to cover pension system deficits.
- Long-term pressure:
  - Recent measures compound long-term aging-related pressures on the pension system; demographic projections and coverage ratio trends point to continued stress on public pensions.
  - Since 2005, Bulgaria’s population has been shrinking due to a sharp decline in population below 65, while the population at retirement age (65 and older) increased.
  - Available population projections (EUROPOP 2019, EUROPOP 2023, and NSI revised projections) concur that the population will continue to shrink and continue to age, though the pace of aging is uncertain.
  - Old-age dependency (65+/20–64) is projected to increase markedly in the coming three decades before declining; old-age dependency will be the main driver in the projected increase in the public-pension-to-GDP ratio in the long run.

### Sustainability, fairness, and actuarial indicators
- Proportionality measure (PM):
  - PM = Present Value of Benefits ∕ Present Value of Contributions.
  - The PM in Bulgaria is significantly above 1; it is higher for younger cohorts than for older cohorts and higher than for the EU or Newer Member States peers.
  - Interpretation: a person born in 2000 can expect a pension that is low but is about twice larger than his/her contribution.
- Accrued rights and contingent liabilities:
  - Present value of pensions to be paid in the future based on accrued rights are among the lowest in Europe; accrued pension rights increased from 168 percent of GDP in 2015 to 186 percent of GDP in 2016 before declining to 178 percent in 2018.
  - Accrued pension rights are overwhelmingly related to the pay-as-you-go system; the share of private pension schemes accounted for less than 7 percent of accumulated entitlements in 2018.
  - Low contingent liabilities mitigate fiscal risks in the medium-term, in part due to low level of benefits; however, demographic pressures and recent pension measures increase medium- and long-term fiscal risks.

### Drivers of low contribution incentives and observed outcomes
- Benefit formula and incentive effects:
  - Current old-age benefit formula: B = AII × IC × IP × AR
    - B is the pension benefit,
    - AII is the national monthly average insurable income for 12 months preceding retirement,
    - IC is the ratio between the average insurable income of the person and the average insurable income for the country in the periods of insurance,
    - IP is the insurance period (contributory and non-contributory periods),
    - AR is the accrual rate.
  - Because benefits are based on length of contribution period and insurable income rather than on contributions actually paid, the formula provides individuals little incentive to contribute.
- Behavioral and fiscal outcomes:
  - The system encourages underreporting of wages and informal work; only about half of the working age population contributes to the mandatory pension system while the employment rate of working age population slightly exceeds 70 percent.
  - A retiree will receive the minimum pension even if the calculated pension is lower than the minimum pension; 46 percent of pensioners received the minimum pension in 2023.
  - Contributions-to-GDP ratio is low by EU standards; contribution rates are among the lowest in the EU.
  - Over 21 percent of contributions are now capped (benefiting 14 percent of contributors), while maximum pensions account for less than 0.5 percent of pension payments and 0.1 percent of pensioners.

### Fiscal implications and transfers
- Rising pension system deficit and transfers:
  - Transfers from the state to cover the deficit increased by 2.3 percentage points of GDP between 2019 and 2023 and are expected to further increase in the coming years.
  - The fiscal transfers required to cover the pension system deficit have jumped to over 5½ percent of GDP and are expected to remain at this level in the coming years.
  - Such an increase in transfers weighs on the general government’s fiscal balance and may crowd out other more productive spending and/or increase public debt.
- Revenue mismatch:
  - While the pension freeze and the 2015 reform had gradually increased the share of pension payments financed by social contributions, the share fell back to its 2015 level in 2023 and the MTBF expects only a limited recovery in the coming years.

### Policy implications and recommended focus for pensions
- Core policy direction:
  - Further reducing pension generosity to restore financial sustainability is not advisable: it would be socially undesirable, politically difficult, could be reversed in crises, and could increase incentives for contribution avoidance, eroding fiscal gains.
  - Given pensions remain low relative to wages and pensioner poverty is widespread, there is little scope for reducing generosity of pensions.
- Recommended reform emphasis:
  - Increase incentives to contribute (to raise revenue rather than cut benefits).
  - Revise the benefit formula to incorporate contributions actually paid to strengthen incentives to contribute and increase revenue for the pension system and the national budget (through better PIT reporting).
  - Increase the perception that contributions are a valuable saving.
  - Once incentives are reinforced, increase the contribution rate (currently among the lowest in the EU) and eliminate the cap on social contributions (maximum insurable income) in combination with elimination of the maximum pension.
  - In the meantime, link the maximum social insurance income to wage growth to revert the sharp increase in the share of contributors benefiting from the cap and to increase revenue.
- Expected effects of reforms:
  - Increased pension system revenue and reduced need for fiscal transfers.
  - Creation of fiscal space for more productive spending.
  - Improved actuarial fairness and sustainability without reducing pension adequacy.
  - Improved fiscal redistribution via social contributions over time.

*Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1bgrea2024002.pdf*

### 1. Various Population Projections _____________________________________________________ 12

### 1. Various Population Projections

### Overview of the Bulgarian pension system
- Structure:
  - First pillar: mandatory, defined benefit, pay-as-you-go system administered by the National Social Security Institute (NSSI); financed by a contribution rate of 14.8 percent of the gross insurable income for individuals born after 1959 and 19.8 percent for individuals born in 1959 or before.
  - Second pillar: mandatory defined contribution system managed by licensed private pension providers for individuals born after 1959; funded by a contribution rate of 5 percent.
  - Third pillar: voluntary defined contribution system managed by licensed private pension providers (Republic of Bulgaria, 2020).
- Self-employed: contribute at the same rate of 14.8/19.8 percent to the first pillar, applied to declared covered earnings for the previous calendar year.

### Past reforms and their quantitative effects
- 2015 reform:
  - Gradually increases the retirement age: from 63 and 8 months for men and 60 and 8 months for women to 65 in 2029 for men and in 2037 for women.
  - Raises the required minimum contribution period by two months every year until it reaches 40 years for men and 37 years for women by 2027.
  - Reduced pension spending by 1½ percent of GDP between 2014 and 2019.
- Coverage and adequacy impacts:
  - Coverage ratio (pensioners as a share of population 65+) continued to decline after the reform and is projected to remain below EU average by 2040 (2021 Ageing Report).
  - Expected years in retirement declined more rapidly than for the EU as a whole.
  - Generosity/support measures: the pension support ratio declined markedly after the reform; benefit ratio continued to decline as pension increases lagged wage growth.

### Adequacy versus sustainability: key statistics
- Replacement rates and poverty:
  - The share of pensioners receiving the minimum pension increased from 27 percent in 2016 to 35 percent in 2019; the share reached 46 percent in 2022 and 2023.
  - The minimum pension fell from 57 percent of the minimum wage in 2010 to 38 percent in 2019 and remained well below the poverty line.
  - Until 2022, the elderly (65 and older) was the age group suffering the most from poverty; at-risk-of-poverty measures and median equivalized net incomes show elderly disposable income significantly lower than other age groups.
- Bulgaria’s pension spending:
  - On the eve of the pandemic, Bulgaria’s spending on pensions was more than 5 percent of GDP lower than the EU average.
  - Bulgaria is characterized as a low contribution - low benefit system resulting in relatively limited spending.

### COVID-19 pandemic measures and fiscal impact
- Pandemic-era increases:
  - More than half of the fiscal support provided to households in 2020–21 targeted pensioners.
  - Pandemic household support accounted for over 1 percent of GDP in 2021.
  - Support mainly took the form of pension increases and ad-hoc supplementary payments (bonuses); some measures planned as temporary were made permanent in 2022.
- Fiscal mismatch and indicators of imbalance:
  - Contribution rate and maximum insurable income were left unchanged during the pandemic.
  - Share of contributors subject to the cap on social contributions increased from 6.4 percent in 2019 to 14.2 percent in 2023.
  - Contributions at maximum social insurable income represent over 21 percent of total contributions (up from 13.6 percent in 2019).
  - Maximum pension relative to maximum insurable income:
    - Over 2009–19: maximum pension was set at 35 percent of maximum insurable income.
    - 2020: 40 percent.
    - 2021: 48 percent.
    - 2022: 62 percent.
    - 2023: 100 percent.
    - Increase in 2024 will bring this ratio to only 91 percent.
  - Share of pensioners receiving the maximum pension fell from 2 percent in 2020 to 0.1 percent in 2023.
- Net effect:
  - Contributions increasingly capped, constraining system revenue, while the cap on pensions has effectively disappeared, increasing spending.
  - In 2023 the full impact of pandemic-era measures is visible: old-age poverty dropped markedly but pension spending rebounded strongly.

### Trade-offs, risks, and projection drivers
- Trade-offs observed:
  - Reforms in the 2010s improved financial sustainability at the cost of reduced pension adequacy.
  - COVID-era measures improved adequacy but materially worsened financial sustainability and increased fiscal risk because the state is legally required to transfer funds to cover pension system deficits.
- Long-term pressure:
  - Recent measures compound long-term aging-related pressures on the pension system; demographic projections and coverage ratio trends point to continued stress on public pensions.

### Policy implications and recommended focus
- Primary policy direction:
  - Given pensions remain low relative to wages and pensioner poverty is widespread, there is little scope for reducing generosity of pensions.
  - Reforms should aim at increasing incentives to contribute to the pension system to boost revenues and improve financial sustainability.
- Specific reform emphasis:
  - Increase incentives to contribute (to raise revenue rather than cut benefits).
  - Consider revenue-side measures to match permanent spending increases.
  - Address the effective disappearance of the pension cap and the growing role of the contribution cap in concentrating revenue risk among capped contributors.

*Sources: NSSI, Eurostat, Republic of Bulgaria, and IMF staff calculations.*

### 13.      The COVID-19 measures have affected the financial sustainability of the pension

### 13.      The COVID-19 measures have affected the financial sustainability of the pension system.

### Impact of COVID-19 measures on pension spending and financing
- Pension spending reached a historical high in 2023.
- The 2024 financial data do not include the impact of the Easter supplement voted by Parliament in April 2024. The supplement of BGN 100 granted to pensioners whose pension is below the official poverty line of BGN 526 per month has an estimated cost of 0.03 percent of GDP.
- While the pension freeze and the 2015 reform had gradually increased the share of pension payments financed by social contributions, the share fell back to its 2015 level in 2023 and the MTBF expects only a limited recovery in the coming years.

### Fiscal implications and transfers
- The rising deficit of the pension system constitutes a fiscal burden because, by law, the pension system deficit is financed by transfers from the state.
- Transfers from the state to cover the deficit increased by 2.3 percentage points of GDP between 2019 and 2023 and are expected to further increase in the coming years.
- The fiscal transfers required to cover the pension system deficit have jumped to over 5½ percent of GDP and are expected to remain at this level in the coming years.
- Such an increase in transfers weighs on the general government’s fiscal balance and may crowd out other more productive spending and/or increase public debt.

### Long-term demographic pressures
- Present value of pensions to be paid in the future based on accrued rights are among the lowest in Europe; accrued pension rights increased from 168 percent of GDP in 2015 to 186 percent of GDP in 2016 before declining to 178 percent in 2018.
- Accrued pension rights are overwhelmingly related to the pay-as-you-go system; the share of private pension schemes accounted for less than 7 percent of accumulated entitlements in 2018.
- Since 2005, Bulgaria’s population has been shrinking due to a sharp decline in population below 65, while the population at retirement age (65 and older) increased.
- Available population projections (EUROPOP 2019, EUROPOP 2023, and NSI revised projections) concur that the population will continue to shrink and continue to age, though the pace of aging is uncertain.
- Old-age dependency (65+/20–64) is projected to increase markedly in the coming three decades before declining; old-age dependency will be the main driver in the projected increase in the public-pension-to-GDP ratio in the long run.

### Sustainability, fairness, and the proportionality measure (PM)
- The proportionality measure (PM) is defined as: PM = Present Value of Benefits ∕ Present Value of Contributions.
- The PM in Bulgaria is significantly above 1; it is higher for younger cohorts than for older cohorts and higher than for the EU or Newer Member States peers.
- In simple terms, a person born in 2000 can expect a pension that is low but is about twice larger than his/her contribution.
- Low contingent liabilities mitigate fiscal risks in the medium-term, in part due to low level of benefits; however, demographic pressures and recent pension measures increase medium- and long-term fiscal risks.

### Drivers of low contribution incentives and observed outcomes
- The current old-age benefit formula is: B = AII × IC × IP × AR
  - B is the pension benefit,
  - AII is the national monthly average insurable income for 12 months preceding retirement,
  - IC is the ratio between the average insurable income of the person and the average insurable income for the country in the periods of insurance,
  - IP is the insurance period (contributory and non-contributory periods),
  - AR is the accrual rate.
- Because benefits are based on the length of contribution period and insurable income rather than on contributions actually paid, the formula provides individuals little incentive to contribute.
- The system encourages underreporting of wages and informal work; only about half of the working age population contributes to the mandatory pension system while the employment rate of working age population slightly exceeds 70 percent.
- The disincentive is reinforced by the fact that a retiree will receive the minimum pension even if the calculated pension is lower than the minimum pension; 46 percent of pensioners received the minimum pension in 2023.
- Contributions-to-GDP ratio is low by EU standards; contribution rates are among the lowest in the EU.
- Over 21 percent of contributions are now capped (benefiting 14 percent of contributors), while maximum pensions account for less than 0.5 percent of pension payments and 0.1 percent of pensioners.

### Policy options to mitigate projected pension burden
- Further reducing pension generosity to restore financial sustainability is not advisable: it would be socially undesirable, politically difficult, could be reversed in crises, and could increase incentives for contribution avoidance, eroding fiscal gains.
- Policies should focus on increasing incentives to contribute to the pension system:
  - Revise the benefit formula to incorporate contributions actually paid to strengthen incentives to contribute and increase revenue for the pension system and the national budget (through better PIT reporting).
  - Increase the perception that contributions are a valuable saving.
  - Once incentives are reinforced, increase the contribution rate (currently among the lowest in the EU) and eliminate the cap on social contributions (maximum insurable income) in combination with elimination of the maximum pension.
  - In the meantime, link the maximum social insurance income to wage growth to revert the sharp increase in the share of contributors benefiting from the cap and to increase revenue.
- Expected effects of these reforms:
  - Increased pension system revenue and reduced need for fiscal transfers.
  - Creation of fiscal space for more productive spending.
  - Improved actuarial fairness and sustainability without reducing pension adequacy.
  - Improved fiscal redistribution via social contributions over time.

### Conclusion — fiscal risks and recommended direction
- Pension spending increased markedly in recent years as a key policy tool to support households during the COVID-19 pandemic; increases became permanent and improved adequacy.
- Additional spending was not matched by sufficient revenue measures, presenting fiscal risks; pension deficits risk crowding out more productive spending needed to boost productivity and income convergence.
- The last decade’s approach of reducing pension generosity helped financial sustainability but resulted in a system of low contribution and low benefit; further reductions in adequacy are no longer feasible given high old-age poverty and pensions often below minimum wage.
- Restoring fiscal sustainability requires changing the pension formula to increase revenue and incentivize contributing, thereby bringing pensions closer to a sustainable and actuarial fair level without reducing them.

*Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1bgrea2024002.pdf*

### 1.      State-owned enterprises play an

### 1.      State-owned enterprises play an

### Overview
- There are about 700 SOEs, defined as firms in which the central government or sub-national government levels own a minimum stake of 50.1 percent.
- SOEs are especially important in network industries, such as energy and transportation.
- SOEs’ value added is about 5 percent of GDP.
- SOEs account for about 4.1 percent of total employment.
- The paper’s purpose:
  - Consider fiscal risks from SOEs from an aggregate perspective based on state-guaranteed debt, contingent liabilities, and the budgetary impact of SOEs.
  - Use firm-level data to assess risks from important SOEs in the energy and transportation sectors by assessing financial performance (profitability, liquidity, and solvency) using the IMF’s SOE Health Check Tool (IMF, 2021).
  - Conclude and recommend policies.

### Aggregate perspective: guarantees, contingent liabilities, and budgetary impact
- State-guaranteed debt:
  - Explicit state-guaranteed debt of SOEs was only 0.5 percent of GDP on average over 2010–21.
  - This is far below the EU average of 9 percent and CESEE average of 3.5 percent of GDP (averages over 2010–21).
  - EU state guarantees of SOEs (percent of GDP) increased by almost 2 percentage points during 2019–21; the corresponding increase in Bulgaria was 0.3 percentage points.
- Guarantees issued by SOEs themselves:
  - Aggregated information on guarantees issued by SOEs is not available.
  - SOE issuance of guarantees does not require approval or monitoring of the Ministry of Finance.
- Contingent liabilities:
  - Total amount of liabilities of government-controlled entities classified outside general government was about 12 percent of GDP on average over 2013–21.
  - Historical restructuring since the 1990s reduced the share of SOEs in the Bulgarian economy.
- Fiscal support versus revenue contribution:
  - Over 2017–19, average government support to SOEs was about 1.5 percent of GDP annually (direct and indirect).
  - Annual contribution of SOEs to the budget was about 0.2 percent of GDP in 2017–19.
  - Net fiscal support over 2017–19 was about 1.3 percent of GDP.
  - In 2020, direct support increased to about 2.5 percent of GDP while revenue contribution fell to less than 0.1 percent, leading to a deficit of about 2.5 percent of GDP.
- Dividend policy:
  - Budget 2023 increased the dividend ratio to 100 percent from 50 percent, expected to boost budget revenues by 670 million leva (or 0.35 percent of GDP).
  - Budget 2024 maintains a 100 percent dividend policy, assumed in the 2025–26 medium-term budget framework.
  - Higher dividend payouts can improve fiscal revenue temporarily but may reduce SOEs’ incentives to improve productivity and profitability and curtail investment capacity by limiting retained earnings.

### Firm-level analysis: sample and sectors
- Sample:
  - Analysis focuses on 15 large SOEs in energy (8) and transportation (7).
  - These 15 SOEs account for about 70 percent of the total liabilities and assets of all SOEs over 2015–21.
  - Twelve of the considered SOEs are also in the top 15 largest SOEs by assets.
- Selected SOEs (by sector):
  - Energy Sector: National Electric Company (NEC); Kozloduy Nuclear Power Plan; Bulgargaz; TPP Maritsa Iztok 2; Electricity System Operator; Bulgarian Energy Holding; Mini Maritsa Iztok; Bulgartransgaz.
  - Transportation Sector: National Railway Infrastructure Company (NRIC); BDZ – Passenger Services, Ltd; Bulgarian Air Traffic Services Authority; Transport Construction and Recovery, TSV; BDZ – Cargo Services, Ltd; Port Varna; Bulgarian Port Infrastructure.
- Financial risk framework:
  - Three aspects assessed: profitability, solvency, and liquidity.
  - Table 1 links fiscal risks to main sources at SOE level and key financial indicators:
    - Lower dividends and taxes ← lower revenues or higher costs → deteriorating profitability indicators.
    - Higher subsidies ← higher cost of subsidized activities → deteriorating profitability indicators.
    - Equity injections ← losses eroding equity; unsustainably high debt; write-off/impairment of assets → deteriorating solvency indicators (debt to assets).
    - Increased borrowing needs ← weak internal cash generation; poor working capital management; inadequate access to market financing → deteriorating liquidity or solvency (interest coverage) indicators.
    - Materialization of contingent liabilities ← weak internal cash generation; inadequate market access → deteriorating liquidity or solvency (interest coverage) indicators.

### Profitability and efficiency findings
- Profitability:
  - Average ROA across the 15 SOEs over 2015–21 varies between -1 percent and 2 percent.
  - Average ROA for private firms is 10 percent.
  - In 2022, average ROA for SOEs was 9 percent vs. 11 percent for private firms, driven by high ROAs of National Electric Company, Kozloduy NPP, and TPP Maritsa Iztok 2; excluding these three, average SOA in 2022 was 2 percent.
  - Median ROA gap stable at 6 percentage points over the sample.
  - SOEs’ ROE was on average 20 percentage points lower than private firms’.
  - The ROE gap in Bulgaria is much larger than the 4 percentage-point gap documented in countries with better governance scores.
- Persistent losses:
  - Over 2015–22, about 6 SOEs made losses on average: TPP Maritsa Iztok 2; Transport Construction and Recovery; BDZ – Passenger Services; BDZ – Cargo Services; Bulgarian Port Infrastructure; and NRIC.
  - National Electric Company and other energy SOEs saw profits in 2021–22 due to high electricity prices; excluding 2021–22 NEC is loss-making over 2015–22.
- Capital and labor efficiency:
  - Return on capital employed (ROCE) shows SOE efficiency lags the private sector.
  - Average cost of employees is higher in SOEs than in private firms, though the gap is decreasing.
  - Cost of employees for SOEs is more than 20 percent of operating revenues, about double the corresponding value in private firms.

### Liquidity and solvency findings
- Short-term liquidity:
  - Current ratio (average 2015–22) indicates about six of the considered SOEs had current ratio < 1: NRIC; TPP Maritsa Iztok 2; National Electric Company; BDZ – Passenger Services; BDZ – Cargo Services; and Transport Construction and Recovery.
  - Consequence: accumulation of debt arrears to suppliers, sometimes paid by bridge loans and/or state aid.
  - Example: Bulgargaz faced a liquidity crisis in mid-2022 due to low collection of receivables and debt arrears from the Sofia district heating company.
- Debt and solvency:
  - Several major SOEs have high debt-to-asset ratios, raising solvency concerns.
  - An SOE with debt-to-assets > 100 percent has negative equity and is technically insolvent.
  - Debt-to-assets increased over time in Bulgartransgaz; National Railway Infrastructure Structure; Bulgaria Energy Holding; and Electricity System Operator.
  - Bulgarargaz’s debt-to-assets rose from about 45 percent in 2019 to above 90 percent in 2022.
  - Combining high debt with low profitability raises concerns about ability to service debt and fiscal risks.

### Assessment tool and indicators
- The IMF’s State-Owned Enterprise Health Check Tool (IMF, 2021) is applied to provide a comprehensive assessment at firm level.
- The tool presents risks associated with profitability, liquidity, and solvency across 12 indicators (as referenced in Table 3).

*Sources: Eurostat, Orbis database, WEO database, and IMF staff calculations.*

### 17.      The tool uses thresholds to define five categories of risks for each indicator. The risk

### 17.      The tool uses thresholds to define five categories of risks for each indicator. The risk

### Risk-categorization framework and thresholds
- The tool defines five risk categories for each indicator, increasing from low risk (Category 1) to high risk (Category 5).
- Thresholds are applied uniformly to all SOEs to guarantee consistent comparisons across companies, despite heterogeneity across industries.
- Most thresholds are common benchmarks (Halstead and others, 2021) with two exceptions:
  - For ROE, SOEs are classified in the lowest risk category where their ROE exceeds the median return of private firms, which is about 18 percent on average over 2015–22.
  - The low-to-moderate risk rating (Category 2) for ROE is set at a return of 5 percent — the average of (i) the return of the first quartile of private ROEs (about 7.2 percent) and (ii) the effective rate on government debt (2.9 percent) over 2015–21.
- Loss-making SOEs are included in the two highest-risk categories.
- For ROA, thresholds for Categories 1 and 2 are the median and the first quartile of private ROA.
- Risk thresholds derived from the ROE-related thresholds and balance sheet leverage produce similar values.

### Financial ratios and what they measure
- Liquidity
  - Current Ratio: ability to meet short-term liabilities (falling due within 12 months) from liquidating short-term assets.
  - Quick Ratio: ability to meet short-term liabilities with only the most liquid short-term assets.
  - Debtor Turnover Days: speed of collection from customers; a high ratio could indicate liquidity challenges.
  - Creditor Turnover Days: speed of payments to suppliers; a high ratio may indicate build up of arrears.
- Solvency
  - Debt to Assets: proportion of financing from liabilities; high ratio indicates greater reliance on debt.
  - Debt to Equity: liabilities relative to equity; high ratio indicates greater reliance on debt financing.
  - Debt to EBITDA: number of years at current cash generation to pay off all debt; higher indicates more indebted company and higher risk of inability to service debt.
  - Interest Coverage: operating profits relative to financing costs; higher indicates more capacity to absorb shocks.
  - Cash Interest Coverage: cash generation relative to financing costs; higher indicates more capacity to cover financing costs.
  - Debt Coverage: (described in Table 4 thresholds).
- Profitability
  - Return on Assets: allocative efficiency of assets in producing profits.
  - Return on Equity: ability to generate profits using shareholders’ capital.
- Cost Recovery
  - Measures ability to generate adequate revenue to cover operating expenses; a ratio < 1 indicates inability to cover operating expenses.

### Table 4 — Risk thresholds (Profitability, Liquidity, Solvency, Cost recovery)
- Profitability
  - Return on assets greater than: 7% (Low risk), 2% (Low-Moderate risk), 0% (Moderate risk), -5% (Moderate-High risk), [beyond Category 5 = High risk]
  - Return on equity greater than: 18% (Low risk), 5% (Low-Moderate risk), 0% (Moderate risk), -10% (Moderate-High risk), [beyond Category 5 = High risk]
  - Cost recovery greater than: 1.5 (Low risk), 1.3 (Low-Moderate risk), 1.0 (Moderate risk), 0.8 (Moderate-High risk), [beyond Category 5 = High risk]
- Liquidity
  - Current ratio greater than: 2.0 (Low risk), 1.5 (Low-Moderate risk), 1.3 (Moderate risk), 1.0 (Moderate-High risk), [Category 5 = High risk threshold exceeded]
  - Quick ratio greater than: 1.2 (Low risk), 1.0 (Low-Moderate risk), 0.8 (Moderate risk), 0.7 (Moderate-High risk)
  - Debtor turnover days less than: 30.0 (Low risk), 40.0 (Low-Moderate risk), 50.0 (Moderate risk), 75.0 (Moderate-High risk)
  - Creditor turnover days less than: 30.0 (Low risk), 60.0 (Low-Moderate risk), 90.0 (Moderate risk), 120.0 (Moderate-High risk)
- Solvency
  - Debt to assets less than: 30% (Low risk), 50% (Low-Moderate risk), 80% (Moderate risk), 100% (Moderate-High risk)
  - Debt to equity less than: 50% (Low risk), 100% (Low-Moderate risk), 150% (Moderate risk), 200% (Moderate-High risk)
  - Debt to EBITDA less than: 1.5 (Low risk), 2.0 (Low-Moderate risk), 3.0 (Moderate risk), 5.0 (Moderate-High risk)
  - Interest coverage greater than: 2.0 (Low risk), 1.5 (Low-Moderate risk), 1.2 (Moderate risk), 1.0 (Moderate-High risk)
  - Cash interest coverage greater than: 3.0 (Low risk), 2.0 (Low-Moderate risk), 1.5 (Moderate risk), 1.0 (Moderate-High risk)
  - Debt coverage greater than: 0.8 (Low risk), 0.6 (Low-Moderate risk), 0.4 (Moderate risk), 0.3 (Moderate-High risk)
- Notes on classification:
  - Threshold set for Category 2 means any indicator with a lower/higher value (depending on indicator) will be classified as Category 1.
  - Indicators lying between Category 2 and Category 3 thresholds, Category 3 and Category 4 thresholds, and Category 4 and Category 5 thresholds will be classified as Category 2, Category 3, and Category 4, respectively.
  - Indicators beyond the Category 5 threshold will be classified as Category 5 (high risk).

### Pre-COVID assessment (summary of findings)
- In the pre-COVID period:
  - About 30 percent of considered SOEs had risk rating above moderate.
  - Sustained low profitability was a concern in six or seven SOEs over the three selected years.
  - About half of SOEs were considered at high liquidity risks based on current and quick ratios, indicating limited ability to meet current liabilities.
  - Solvency risk was a concern for most SOEs, particularly in terms of Debt to EBITDA.
  - Several SOEs faced risks of insufficient operating profits to cover financing costs (low interest cover) or high debt-to-equity ratios.
  - The overall risk rating identified about five SOEs with moderate to high fiscal risk: National Electric Company (NEK), TPP Maritsa Iztok 2, National Railway Infrastructure Company (NRIC), BDZ – Cargo Services, and Transport Construction and Recovery (TSV).

### Post-COVID assessment and 2022 developments
- SOEs’ financial performance improved noticeably in 2022.
  - COVID-shock worsened profitability in the short run, but the effect was mitigated by fiscal measures.
  - SOEs with risk ranking from moderate-to-high for at least one year in 2020–21 are mainly the same as pre-COVID.
  - In 2022, almost all SOEs improved significantly, resulting in overall rankings of moderate or low-to-moderate risks driven by improved profitability.
  - Higher energy prices helped improve positions of loss-making SOEs in the energy sector.
  - Transportation sector SOEs showed improved profitability, suggesting they could pass costs to customers.
  - Bulgargaz was the only SOE with moderate-to-high risk ranking in 2022 due to deterioration of profitability caused by impairment of inventories and accrued losses from trade receivables.
- Highlight: Importance of buffers in case of unexpected shocks.

### Conclusion and policy recommendations
- Aggregate and firm-level analyses indicate potential SOE-related contributors to fiscal risks:
  - Despite low state-guaranteed debt, negative net budgetary flows from SOEs coupled with a contingent liability of 12 percent of GDP could lead to long-term fiscal challenges.
  - Profitability of major SOEs is far below that of private firms, partially due to resource allocation inefficiencies.
  - Several SOEs consistently incur heavy losses and have difficulties managing short-term debts (liquidity risks).
  - Some key SOEs have high liabilities relative to assets, raising concerns on long-term solvency.
- Recommended policy actions:
  - Closely monitor financial performance of SOEs and identify mitigation measures, including:
    - Establishing a (digital) unified database, publicly available and frequently updated, on the financial performance of SOEs.
    - Making budgetary flows between SOEs and the government more transparent.
    - Collecting and publishing information on guarantees issued by SOEs themselves.
  - Implement reforms to improve SOEs’ financial performance, with emphasis on SOE governance—management, oversight, and transparency—to increase worker productivity and lower costs, particularly in the electricity sector.
    - Such reforms can increase SOEs’ net fiscal contribution and boost overall economic competitiveness and productivity given SOEs’ role in the production network.
  - Set dividend policies that balance government fiscal interests and SOEs’ financial sustainability and productivity:
    - Recognize the government’s claim on SOE profits while allowing enterprises to retain earnings for reinvestment to achieve solid capital structures and long-term investments.
    - Ensure dividend policy is predictable to reduce uncertainty and increase firms’ incentives to invest.

*Sources: IMF (2021) and IMF staff calculations; Orbis database and IMF staff calculations; Public Enterprises and Control Agency annual report noted in text.*

### 7.      Exports are gradually shifting towards more technologically-intensive products,

### 7.      Exports are gradually shifting towards more technologically-intensive products,

### Export composition and sophistication
- Bulgaria’s exports are relatively diversified with higher shares in chemicals, refined fuels, non-ferrous metals, food, and machinery.
- Over the years, there has been a shift towards higher value-added, more high-tech products; exports of electrical and other machinery and chemicals have risen at the expense of textiles, apparel, and some other low-value-added manufactured goods.
- Bulgaria’s export content of productive knowledge, as measured by the Economic Complexity Index, still lags peers and has not increased over the last two decades.
- Note: The Economic Complexity Index measures the amount of productive knowledge in exports. It is a function of a country's export diversity (number of products exported) and ubiquity (number of countries exporting a product). Red squares are averages 1995-99; blue bars are averages 2015–19.

### Trade patterns and energy decoupling from Russia
- Imports from Russia are heavily concentrated in energy products, while exports to Russia are limited.
- The import of Russian gas has significantly declined in the last two years following Russia’s unilateral decision to stop pipeline gas supply to Bulgaria in April 2022.
- Import of Russian crude oil increased under the derogation to the EU embargo that the European Commission granted to Bulgaria.
- A more notable decline in the reliance on Russian oil is expected to be seen only starting from 2024, as the Bulgarian parliament revoked the derogation from March 1, 2024.
- Note: Data for the trade flows with Russia display high volatility due to price and accounting changes.

### Foreign Direct Investment (FDI)
- FDI remains substantially below its 2007 peak.
- After a boom in the leadup to EU accession and the GFC, foreign investment flows to Bulgaria have significantly slowed and settled to levels similar to other NMSs.
- In the last decade, FDI largely came to Bulgaria from EU partners and was mostly concentrated in the services sector, especially financial activities, including real estate, and wholesale and retail trade.
- Making Bulgaria’s economic environment more attractive to foreign investors would likely boost integration and allow the country to harness greater benefits from EU single market participation, including attracting multinational firms.
- Example: For instance, the Belgian battery manufacturer ABEE is reportedly planning to invest €1.1 billion in Bulgaria in three sites, a battery factory, a research and development center, and a recycling facility.

### Bulgaria’s participation in Global Value Chains (GVCs) — aggregate metrics
- In 2020, 35 percent of Bulgaria’s final demand was met by value added coming from abroad, about half of which from other EU countries.
- This foreign reliance remains below pre-GFC levels, with a decline observed in recent years.
- After peaking at almost 45 percent in 2017, Bulgarian value added going abroad declined to about 38 percent of total domestic value added in 2020, close to the EU average.
- Almost half of the value added sent abroad went to other EU countries; the share going to China has been steadily growing over time.
- About 25 percent of the domestic value added satisfying foreign final demand in 2020 came from the manufacturing sector (including construction and utilities), 13 percent from agriculture and mining, and the remaining 56 percent from services.
- The fraction of Bulgarian employment sustained by foreign demand rose from 35 percent during the GFC to about 50 percent by the end of the 2010’s, before slightly declining to 46 percent in 2020. Bulgaria’s share is the fourth highest in the EU behind Ireland, Luxembourg, and Malta.
- Bulgaria’s participation in GVCs declined in 2009 and only partially recovered; participation increased only marginally thereafter, leaving Bulgaria trailing most NMSs and the EU average.
- Integration in GVCs is the sum of backward linkages and forward linkages. Backward linkages are in line with peers and the EU average; forward linkages remain below the EU average and more contained than in peer countries.

### GVC composition and sophistication
- Backward linkages: the most significant involvement is in GVCs for basic metals and coke and refined petrol products, which accounted for almost a third of total foreign value added in Bulgaria’s exports in 2020.
- Forward linkages are well-balanced across manufacturing and services industries.
- Involvement in information GVCs (computers and electronics production and provision of communication, IT, and information services) is among the lowest in the EU.
- The sophistication of Bulgaria’s GVCs has been slowly increasing; recent FDI commitments have the potential to increase Bulgaria’s foothold in rapidly expanding GVCs, such as those for electric vehicles’ batteries.

### Integration with EU and other partners
- Bulgaria consistently lags peers in integration with other EU countries.
- On the buying side, relatively low backward linkages with the EU have been historically compensated by a significantly larger dependence on Russia and, to a lesser extent, Türkiye.
- In 2020, Russian inputs amounted to 8 percent of Bulgaria’s backward linkages compared to 12 percent for EU partners.
- On the selling side, Bulgaria’s forward linkages with the EU are the lowest among peers, amounting to 12.2 percent of exports in 2020.
- There is room for further increase in GVC integration with the EU on both the selling and, especially, the buying side.

### China’s growing role and foreign market reliance (FMR)
- In 2020, China became the largest GVC partner destination for Bulgaria’s value added outside of the EU.
- The dependence is particularly elevated in sectors such as basic metals and mining and quarrying of non-energy products.
- Foreign Market Reliance (FMR) measures the ratio of domestic output used in foreign production to total domestic production and captures “total” exposure of domestic activity to downstream disruptions in GVCs; FMR also considers exports consumed in destination countries and is based on gross trade flows.

### Drivers of GVC participation — identified correlates
- Commonly identified drivers include: (i) land and other natural resources endowments, labor and capital; (ii) geographic position with respect to large GVC hubs; (iii) domestic industrial capacity; (iv) openness to trade (tariffs and free trade agreements) and FDI; (v) institutional quality; (vi) transport connectivity; and (vii) macroeconomic factors such as cost competitiveness and degree of financial development.
- The analysis considers a parsimonious set of key GVC integration drivers for EU countries over the period 2002 to 2020.

*Sources: Eurostat; OECD TiVA 2023; (Hausmann, Hidalgo, Bustos, Coscia, & Simoes., 2014); (Salinas, 2021); and IMF staff calculations.*

### 19.      We estimate a panel regression, using machine learning (ML) techniques.

### 19.      We estimate a panel regression, using machine learning (ML) techniques.

### Methodology and Data
- ML models used: Linear regression, Elastic Net, Support Vector Regression (SVM), Random Forest Regression, Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (kNN).
- Data sources: OECD TiVA dataset (2023 vintage) for GVC data; Eurostat (as of April 17, 2024) for most drivers; World Bank World Development Indicators for share of private credit in GDP; World Governance Indicators for institutional quality; OECD transport infrastructure database for road infrastructure density.
- Sample and transformations:
  - Estimated with data for EU countries over 2002–20.
  - Variables are considered in percent of the EU average.
  - Data were centered and standardized; all explanatory variables calculated as a percentage of their EU average values.
  - Country dummies included to control for fixed effects.
- Training/testing split: 75 percent training, 25 percent testing.
- Hyperparameter tuning: cross-validated grid-search over a predefined parameter grid; cross validation through k-fold resampling.

### Interpretability and SHAP
- Interpretability approach: SHapley Additive exPlanations (SHAP) values (Lundberg & Lee, 2017) used to measure each driver’s contribution to predicted GVC integration.
- SHAP properties emphasized:
  - Efficiency — sum of feature contributions equals difference between instance prediction and average prediction.
  - Symmetry — equal contributors get equal Shapley values.
  - Dummy — feature with no effect has Shapley value of 0.
  - Additivity — Shapley value of aggregated object is sum of component Shapley values.
- Implementation: Kernel SHAP approximation used (perturbed samples with dropped features replaced by expected values; linear regression on derived dataset yields proxies for Shapley values).

### Model Forecasting Performance (testing subsample)
- Table of forecasting statistics (Mean Squared Error (MSE) and R2):
  - Linear regression: MSE 0.07; R2 0.93
  - Elastic net: MSE 0.07; R2 0.93
  - k-Nearest Neighbors: MSE 0.05; R2 0.95
  - Support vector machine: MSE 0.05; R2 0.95
  - Random forest: MSE 0.06; R2 0.94
  - Extreme gradient boosting: MSE 0.09; R2 0.91
- Diebold-Mariano forecast comparison (modified Harvey, Leybourne, & Newbold, 1997 test); positive statistic indicates model in column performs better than model in row. Selected statistics (significance levels noted: *p<0.1; **p<0.05; ***p<0.01):
  - Linear regression vs Support vector machine: 2.44**
  - Linear regression vs k-Nearest Neighbors: 2.79***
  - Elastic net vs Support vector machine: 2.43**
  - Elastic net vs k-Nearest Neighbors: 2.72***
  - Support vector machine vs Random forest: -2.68***
  - Random forest vs Extreme gradient boosting: -2.34**
  - Extreme gradient boosting vs k-Nearest Neighbors: 2.53**
- Summary interpretation:
  - All models perform similarly overall.
  - k-Nearest Neighbors, Support Vector Machine, and Random Forest do slightly better in forecasting accuracy.
  - Diebold-Mariano test indicates SVM and kNN have statistically significant better forecasting performance than other models; Random Forest performs better than Extreme Gradient Boosting.
- Decision on reporting: results cast in terms of average SHAP values across all models.

### Main Correlates of GVC Integration (ML results; weighted average of SHAP values)
- Major drivers identified (models generally confirm literature findings):
  - High transport connectivity (road infrastructure density).
  - Strong rule of law (governance).
  - Low wages (cost competitiveness; real wages).
  - Structure of the economy — share of manufacturing in GDP (industrial intensity).
  - Other contributing factors: education of the labor force (share of population with tertiary degree), level of investment in innovation (share of R&D in GDP), depth of financial markets (share of private credit in GDP), and internet connectivity (share of households using internet).
- Crisis effects:
  - Analysis confirms deceleration in GVC integration during the two most recent crises for which data are available — the GFC and the COVID pandemic.

### Feature-level SHAP insights and heterogeneity
- SHAP distributions indicate:
  - Road infrastructure, real wages, and rule of law have high contribution to explanation of GVC participation.
  - Internet connectivity and share of population with higher education typically rank lower and show less homogeneous impact across models.
- Color-coded SHAP interpretation rule used:
  - Blue for high feature values, red for low feature values.
  - Example: positive SHAP values that are blue imply higher feature values are associated with positive deviations of the forecast (e.g., rule of law); positive SHAP values that are red indicate higher feature values associated with negative deviations (e.g., real wages).

### Bulgaria — position relative to EU peers and clusters
- Bulgaria’s cluster membership and characteristics:
  - Cluster: green cluster — Bulgaria, Croatia, Estonia, Italy, Latvia, and Lithuania.
  - Cluster profile: low wages and low levels of infrastructure, human capital and innovation, governance, and financial integration; relatively manufacturing-oriented.
- Two broad categories of EU economies identified via cluster analysis:
  - Group 1: large non-price competitiveness advantages — infrastructure, governance, human capital, innovation, and financial depth.
  - Group 2: high cost-competitiveness and higher share of manufacturing in GDP.
- Clustering method: Ward’s hierarchical clustering minimizing within-cluster variance using Euclidean distance; five clusters identified based on within group sum of squares; averages for 2010–20 used; data centralized and standardized.

### Bulgaria-specific findings and policy implications
- Current strengths and weaknesses:
  - Strength: relative cost competitiveness — Bulgaria has the lowest real wage in the EU.
  - Weaknesses: deficiencies in infrastructure, financial intermediation, and rule of law.
  - Additional gaps: R&D investment and internet connectivity lag notable gaps with other EU countries.
  - Relatively better: industrial intensity (share of manufacturing in GDP), share of labor force with a tertiary degree, and financial intermediation development — though quality of education and other improvements remain needed.
- Policy recommendations to boost GVC integration as cost advantage erodes:
  - Step up investment in transport infrastructure (density of road infrastructure).
  - Strengthen governance (rule of law).
  - Increase investment in R&D.
  - Improve internet connectivity.
  - Improve the business environment.
  - Enhance workforce skills and the quality of education.
- Strategic context and opportunities:
  - Prospects of joining the euro and Schengen areas provide opportunities to forge stronger ties with regional partners.
  - Deeper integration requires filling gaps with EU peers to become more competitive and attractive for foreign investors.
  - Trade pattern shifts: links with Russia weakening since Russia’s invasion of Ukraine in 2022 (decoupling especially in energy inputs); China’s role as a buyer in Bulgaria’s global production chains has risen since the GFC, with China becoming the largest non-EU buying partner in recent years.

*Source: IMF staff calculations and analysis as presented in the chapter on machine learning estimation of GVC integration drivers.*

### References

### 1bgrea2024002 - References

### Geoeconomic fragmentation and multilateralism
- Aiyar, S., Chen, J., Ebeke, C., Garcia-Saltos, R., Gudmundsson, T., Ilyina, A., Trevino, J. P. (2023). Geoeconomic Fragmentation and the Future of Multilateralism. IMF Staff Discussion Note SDN/2023/001.
- Baba, C., Lan, T., Mineshima, A., Misch, F., Pinat, M., Shahmoradi, A., van Elkan, R. (2023). Geoeconomic Fragmentation: What's At Stake For the EU. IMF Working Paper 23/245.
- Aiyar, S., Malacrino, D., & Presbitero, A. F. (2024). Investing in friends: The role of geopolitical alignment in fdi flows. European Journal of Political Economy.

### Global Value Chains (GVCs), participation, and development
- Antràs, P. (2020). Conceptual aspects of global value chains. The World Bank Economic Review, 34(3), 551–574.
- Baldwin, R., & Freeman, R. (2022). Risks and global supply chains: What we know and what we need to know. Annual Review of Economics, 153–180.
- Buelens, C., & Tirpák, M. (2017). Reading the footprints: How foreign investors shape countries’ participation in global value chains. Comparative Economic Studies, 59, 561–584.
- Constantinescu, C., Mattoo, A., & Ruta, M. (2019). Does vertical specialisation increase productivity? The World Economy 42.8, 2385–2402.
- Fernandes, A. M., Kee, H. L., & Winkler, a. D. (2022). Determinants of Global Value Chain Participation, Cross-Country Evidence. The World Bank Economic Review, 36(2), 329–360.
- Ignatenko, A., Raei, F., & Mircheva, B. (2019). Global Value Chains: What are the Benefits and Why Do Countries Participate? IMF Working Paper WP/19/18.
- Kummritz, V., Taglioni, D., & Winkler., D. E. (2017). Economic upgrading through global value chain participation: which policies increase the value added gains? World Bank Policy Research Working Paper 8007.
- Pahl, S., & Timmer, M. P. (2020). Do global value chains enhance economic upgrading? A long view. The journal of development studies, 56(9), 1683–1705.
- Salinas, G. (2021). Proximity and Horizontal Policies: The Backbone of Export Diversification. IMF Working Paper No. 2021/064.
- Schwellnus, C., A. H., Samek, L., Pechansky, R. C., & Cadestin, C. (2023). Global value chain dependencies under the magnifying glass. OECD Science, Technology and Industry Policy Papers No. 142.
- Taglioni, D., & Winkler, D. (2016). Making Global Value Chains Work For Development. World Bank Publications.
- Urata, S., & Baek, Y. (2020). “The Determinants Of Participation In Global Value Chains: A Cross-Country, Firm-Level Analysis”. Asian Development Bank Institute Working Paper 1116.
- World Bank. (2020). Trading for Development in the Age of Global Value Chains. World Development Report.

### Methods, forecasting, and model interpretation
- Diebold, F. X., & Mariano, R. S. (1995). "Comparing Predictive Accuracy". Journal of Business and Economic Statistics, 20(1), 134–144.
- Harvey, D., Leybourne, S., & Newbold., P. (1997). "Testing the equality of prediction mean squared errors". International Journal of Forecasting, 13, 281–291.
- Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in neural information processing systems, 30.
- Ward, J. H. (1963). "Hierarchical Grouping to Optimize an Objective Function", Journal of the American Statistical Association, 58, 236–244. Journal of the American statistical association, 58(301), 236–244.

### Bulgaria-specific studies and country analysis
- Ivanov, E., & Ivanova, N. (2021). Determinants of Bulgarian Exports: The Role of Price and Non-Price Competitiveness. Bulgarian National Bank Discussion Paper 118/2021.
- Ivanova, N., & Ivanov, E. (2017). The Role of Bulgaria in Global Value Chains. Bulgarian National Bank Discussion Paper 105/2017.

*Source: 1bgrea2024002 - References*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1bgrea2024002.pdf_