## 2. Climate Targets and Progress

## Source details

**Canonical URL:** [2. Climate Targets and Progress](https://www.imf.org/-/media/files/publications/cr/2022/english/1grcea2022002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2022/english/1grcea2022002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2022/english/1grcea2022002.pdf.json)

---

### Climate risks and vulnerability
- Greece is vulnerable to climate change: extreme weather events have caused a high number of fatalities and significant economic damage in recent decades.
- IMF INFORM risk index: climate risk in Greece is higher than the eurozone average (INFORM risk considers hazard & exposure, vulnerability, and lack of coping capacity using 54 indicators).
- Cost of climate change adaptation: estimated at 1.5 percent of GDP annually over the period 2025–50 according to the 2016 National Climate Change Adaptation Strategy.
- Exposure specifics:
  - Third longest coastline in Europe.
  - 80 percent of industrial activities and 90 percent of tourism infrastructure are currently in coastal areas and are at risk from sea level rise.
- Economic concentration at risk:
  - Shipping and tourism together account for about a quarter of GDP and could be affected by climate change and climate policies.

### Climate policy agenda and investment needs
- Key policy instruments:
  - National Energy and Climate Plan (NECP) introduced in 2019.
  - Ambitious new Climate Law in public consultation.
  - Climate Law initiatives include phasing-out the lignite plants by 2028, zero emissions for all new vehicles by 2030, and more than doubling the share of renewables in final energy consumption.
- Investment requirements:
  - Achieving NECP targets requires green investment estimated at €   43.8 billion (over 20 percent of 2021 GDP) during 2020-30.
  - NGEU provides part of the funding; a significant financing gap remains.
  - Mobilizing private green financing under the Recovery and Resilient Plan and effective implementation will be key.

### Distributional impacts of climate change and climate policies
- Differential impacts:
  - Communities: rural and coastal areas vulnerable.
  - Sectors: tourism and agriculture particularly exposed.
  - Households: low income and low skilled households more affected.
- Mitigation policy risks:
  - Policies such as higher carbon prices and phasing-out lignite could have regressive impacts, disproportionately affecting low-income households.
- Potential offsets:
  - Increased green investment could boost growth with potential positive effects on inequality.
- Policy approach:
  - Recommend a comprehensive strategy combining climate policies with strengthened social protection to protect vulnerable groups during the green transition.

### State of social protection and fiscal response
- System weaknesses:
  - Social protection described as weak: costly pensions, the largest in the EA, have crowded out critical social assistance (non-pension social spending).
- Pandemic response:
  - Pandemic highlighted social assistance gaps; government relied heavily on ad hoc discretionary measures with the largest budgetary measures in the EA.
  - Ensuring transparency and targeting of discretionary measures is challenging; publishing Covid-19 related public procurement contracts online noted as welcome.
- Reform priority:
  - Strengthening social assistance proposed as the basis for targeted support during adverse shocks.

### Income distribution: recent developments and disparities
- Historical and recent trends:
  - Entered the sovereign debt crisis with significantly higher-than-average income inequality (Gini coefficient and Income Quintile Share Ratio S80/S20).
  - Income inequality surged during the crisis and has been on a sharp declining trend since 2016, broadly converging to the euro area average, with a small uptick in 2020 (based on 2019 income data).
- 2020 distribution facts:
  - The richest quintile earned over 5 times more than the poorest quintile in 2020.
  - Preliminary Household Budget Survey data show a significant decline in the consumption-based inequality indicator (S80/S20 ratio) in 2020, possibly reflecting suppressed high-income consumption and sizable government support measures.
- 2021 developments:
  - Disposable income increased markedly in 2021 amid a considerable decline in the unemployment rate, which could contribute to a likely further drop in inequality; the full impact once support measures are withdrawn remains uncertain.

### Persistent income gaps by socio-economic group
- Education:
  - Greek workers with tertiary education recorded 66 percent higher median income than those with at most secondary education in 2020.
  - Income premium for higher education has been declining and below the EA average since 2015.
  - Contributing factors: Greece’s highest unemployment rate of high-education population (around 12 percent) in the eurozone and emigration during 2010-15.
- Age:
  - Age income gap remains elevated due to generous pensions and high youth unemployment.
  - 40 percent of youth unemployed have been jobless for at least 12 months.
  - Greece has the highest share of low-wage earners among young employees (aged under-30) at 47 percent in the eurozone; less than 10 percent of elderly employees (aged over-50) are low-wage earners.
- Urban vs Rural:
  - 2020 median incomes: urban median income was 27 percent higher than rural median income (largest gap since 2009).
  - Economic activity concentrated in cities; rural/intermediate regions specialize in natural resource-intensive industries and tourism-related sectors.

### Carbon pricing: methodology, scenarios, and distributional assessment
- CPAT (Carbon Pricing Assessment Tool):
  - Spreadsheet-based, country-parameterized; projects fossil fuel use and CO2 emissions, fiscal, economic, energy price, and distributional burden of carbon pricing and mitigation policies.
  - Distribution module uses 2019 HBS microdata from ELSTAT; calculates shares of direct and indirect energy consumption (8 fuels, 13 goods/service categories); estimates price changes using weighted average fuel intensities and sectoral Leontief relationships (GTAP10); microsimulation across income groups; revenue recycling options include uniform cash by decile, existing social assistance schemes (% increase; ASPIRE), infrastructure access provision.
- Baseline carbon tax assumption:
  - New carbon tax introduced to all non-ETS sectors, carbon price rising from initial level of 25 to 75 real$/tonCO2 by 2030 (the level recommended by the Fund to meet the 2oC global climate goal).
- Reform S1:
  - Recycle half of increases in carbon revenues to transfers to poor households (the bottom 40 percentile) and the other half to labor tax reductions for all taxpayers.
  - Coverage and targeting rates of social transfers used: 68 and 57 percent, respectively (calculated using 2019 HBS data), implying significant leakage of transfers to other groups.
  - Amount of labor tax reductions for each group depends on the size of their initial tax liability.
- Reform S2:
  - Improved targeting: half of carbon revenue increases directed to poor households (bottom 40 percentile) under coverage and targeted rates of 100 and 90 percent, respectively.
  - The other half used for scaling up public investment, notably green infrastructure.

### Environmental, fiscal, and growth outcomes of carbon tax scenarios
- Environmental co-benefits:
  - Reduced climate-related disaster risk, cleaner air, less traffic jams and accidents assessed to outweigh economic costs (deadweight losses) from the carbon tax before revenue recycling, implying net welfare gains for the whole society.
- GDP growth impacts (average over next 15 years):
  - Before revenue recycling, the new carbon tax is estimated to reduce GDP growth by about ⅓ percentage points on average.
  - Recycling carbon revenues for public investment has larger stimulus effects than transfers or tax cuts.
  - In reform S2, the negative growth impact of the carbon tax is fully offset by positive effects from public investment and transfers.

### Consumption and energy intensity by group (Consumption expenditure, 2020, ELSTAT Household Budget Survey 2020; values in percent of total consumption)
- Rich (Top25%): Energy 7.6, High energy intensive 51.2, Low energy intensive 41.2.
- Poor (Bottom25%): Energy 11.8, High energy intensive 53.6, Low energy intensive 34.6.
- Urban: Energy 9.5, High energy intensive 51.4, Low energy intensive 39.1.
- Rural: Energy 13.5, High energy intensive 58.7, Low energy intensive 27.9.
- Implication: Poor and rural households have higher shares of energy and high energy-intensive consumption and will be hit harder by higher carbon prices.

### Welfare impacts and distributional outcomes of reforms (summary)
- Poor households:
  - Direct and indirect consumption losses estimated at about 1–1.5 percent of total consumption, more than offset by positive gains from targeted transfers in both scenarios.
  - Gains more pronounced in reform S2 due to significantly improved targeting and coverage.
- Rich households:
  - Better off in reform S1 compared to S2 as they benefit more from labor tax reductions.
- Rural households:
  - Gain more than urban counterparts in both scenarios due to concentration of low-income groups in rural areas.
  - Poor rural households in reform S2 record the largest net welfare gains.
- Reform S2:
  - Combines a new carbon tax, better-targeted transfers, and increased green public investment to support growth, protect vulnerable households while also reducing emissions.
  - Stands out with less economic costs and more assistance to poor and rural groups.

### Policy recommendations and social protection priorities
- Combine climate-friendly policies with social protection reforms to assist the green transition.
- Improve coverage and targeting of social assistance schemes to protect vulnerable households against climate-related disasters and mitigate adverse impacts of higher carbon prices.
- Social protection reform priorities:
  - (a) expanding coverage of health care and housing benefits (where coverage is low);
  - (b) boosting spending on childcare and the GMI (where targeting accuracy is relatively higher);
  - (c) tackling evasions by the self-employed; and
  - (d) further simplification and consolidation of social assistance schemes.
- Fiscal instrument recommendation:
  - Introduce a new carbon tax and gradually increase it over time to finance targeted transfers and green investment.

*Source: IMF staff estimates and analysis in 1grcea2022002.*

### 2. Climate Targets and Progress __________________________________________________________ 6

### 2. Climate Targets and Progress

### Climate risks and vulnerability
- Greece is vulnerable to climate change: extreme weather events have caused a high number of fatalities and significant economic damage in recent decades.
- According to the IMF INFORM risk index, climate risk in Greece is higher than the eurozone average (INFORM risk considers hazard & exposure, vulnerability, and lack of coping capacity using 54 indicators).
- The cost of climate change adaptation is estimated at 1.5 percent of GDP annually over the period 2025–50 according to the 2016 National Climate Change Adaptation Strategy.
- With the third longest coastline in Europe, 80 percent of industrial activities and 90 percent of tourism infrastructure are currently in coastal areas and are at risk from sea level rise.
- Shipping and tourism together account for about a quarter of GDP and could be affected by climate change and climate policies.

### Climate policy agenda and investment needs
- The government introduced the National Energy and Climate Plan (NECP) in 2019 and an ambitious new Climate Law is in public consultation.
- Policy initiatives under the new Climate Law include phasing-out the lignite plants by 2028, zero emissions for all new vehicles by 2030, and more than doubling the share of renewables in final energy consumption.
- Achieving NECP targets requires a substantial boost in green investment, estimated at €   43.8 billion (over 20 percent of 2021 GDP) during 2020-30.
- While the NGEU provides part of the funding, a significant financing gap remains; mobilizing private green financing under the Recovery and Resilient Plan and effective implementation will be key.

### Distributional impacts of climate change and climate policies
- Climate change is likely to affect some communities (rural and coastal areas), sectors (tourism and agriculture), and households (low income and low skilled) more than others.
- Climate mitigation policies (e.g., higher carbon prices, phasing-out lignite) could have regressive impacts, disproportionately affecting low-income households.
- Increased green investment could boost growth with potential positive effects on inequality.
- A comprehensive strategy combining climate policies with strengthened social protection is recommended to protect vulnerable groups during the green transition.

### State of social protection and fiscal response
- The social protection system is described as weak: costly pensions, the largest in the EA, have crowded out critical social assistance (non-pension social spending).
- The pandemic highlighted social assistance gaps; the government relied heavily on ad hoc discretionary measures with the largest budgetary measures in the EA.
- Ensuring transparency and targeting of discretionary measures is challenging; publishing Covid-19 related public procurement contracts online is noted as a welcome step.
- Strengthening social assistance is proposed to form the basis for targeted support during adverse shocks.

### Income distribution: recent developments and disparities
- Greece entered the sovereign debt crisis with significantly higher-than-average income inequality (Gini coefficient and Income Quintile Share Ratio S80/S20).
- Income inequality surged during the crisis and has been on a sharp declining trend since 2016, broadly converging to the euro area average, with a small uptick in 2020 (based on 2019 income data).
- The richest quintile earned over 5 times more than the poorest quintile in 2020.
- Preliminary Household Budget Survey data show a significant decline in the consumption-based inequality indicator (S80/S20 ratio) in 2020, possibly reflecting suppressed high-income consumption and sizable government support measures.
- Disposable income increased markedly in 2021 amid a considerable decline in the unemployment rate, which could contribute to a likely further drop in inequality; the full impact once support measures are withdrawn remains uncertain.

### Persistent income gaps by socio-economic group
- Education:
  - Greek workers with tertiary education recorded 66 percent higher median income than those with at most secondary education in 2020.
  - The income premium for higher education has been declining and has been below the EA average since 2015.
  - Contributing factors include Greece’s highest unemployment rate of high-education population (around 12 percent) in the eurozone and emigration during 2010-15.
- Age:
  - The age income gap, although dropping from its crisis peak, remains elevated, reflecting generous pensions and high youth unemployment.
  - 40 percent of youth unemployed have been jobless for at least 12 months.
  - Greece has the highest share of low-wage earners among young employees (aged under-30) at 47 percent in the eurozone; less than 10 percent of elderly employees (aged over-50) are low-wage earners.
- Urban vs Rural:
  - In 2020, the median income of the urban population was 27 percent higher than that of the rural population, the largest gap since 2009.
  - Greece has a relatively higher share of population living in rural areas but economic activity is concentrated in cities; high value-added services are largely urban while rural/intermediate regions specialize in natural resource-intensive industries and tourism-related sectors.

### Key policy implication highlighted in this section
- Climate policies should be combined with social protection reforms to protect vulnerable groups during the green transition.
- Introducing a new carbon tax and gradually increasing it over time is recommended to finance targeted transfers and green investment while addressing social protection gaps.

*Source: EM-DAT; IMF, Climate Change Dashboard; IMF staff calculations; European Commission; Eurostat; Greece National Energy and Climate Plan; OECD.*

### 9.      Structural transformation caused by the pandemic and climate change could further

### 1grcea2022002 - 9.      Structural transformation caused by the pandemic and climate change could further

### Structural transformation and distributional vulnerabilities
- Pandemic-driven shift from high-contact to low-contact services and the green transition from high energy-intensive to low energy-intensive sectors will disproportionately hit workers in affected sectors and regions (notably rural areas), especially low-skilled workers.
- Young and low-educated workers are more disadvantaged due to limited working experience and lack of professional networks, exacerbated during structural transitions.

### Social transfers and their impact on inequality
- On average, social transfers reduced Greece’s income inequality by 42 percent from 2011 to 2020.
- Pensions alone contributed about 90 percent of the total reduction in inequality over that period, the highest in the region.
- Contributions from other social transfers are less than half of the Euro Area level.
- Although pensions reduce overall inequality, they increase the age income gap and crowd out social assistance, which is a more effective instrument for redistribution.

### Regional disparities and sectoral specialization (summary of figures and table)
- Distribution of population among regions provided (e.g., Attica 34.9%, Cent. Macedonia 17.5%, Thessaly 6.7%, etc.).
- Regional share of output and population, 2020: charts show notable concentration of population and GDP in Attica and Cent. Macedonia.
- Table of regional specialization in industries (2017) lists indices for sectors (Agriculture, Mining, Manufacturing, Construction, Distribution/trade/transport/accommodation/food services, Information and communication, Financial and insurance, Professional/scientific/technical, Administrative/support, Other services) for each region (values preserved as in source).

### State of social protection: spending, coverage, and targeting
- Spending on social assistance is one of the lowest in the eurozone, both as percent of GDP and in per capita terms.
- Major social assistance schemes show considerable gaps compared to the eurozone average: health care (sickness), education, housing, family and childcare, unemployment benefits, Guaranteed Minimum Income (GMI).
- Coverage: about 37.5 percent of the poorest quintile receives social assistance in Greece, compared with an average of 40.1 percent in the World Bank ASPIRE database (WB, 2018).
- Targeting: Greece’s social assistance targeting is on par with the high-income sample average in ASPIRE; the poorest quintile receives about ⅓ of total social assistance.
- Pensions have higher coverage but much worse targeting compared with social assistance.
- Heterogeneity across schemes:
  - Housing and education benefits: poor coverage (close to zero) and poor targeting.
  - Unemployment and disability benefits: weak coverage and targeting.
  - Health care benefits: low coverage but relatively good targeting.
  - Family/child benefits and other programs for socially excluded groups (including GMI): relatively better coverage and targeting.

### Administrative reforms and digitization
- Pre-2016: administration of social programs complex and fragmented, many small and poorly targeted benefits managed by multiple agencies.
- Reforms since 2016 include:
  - Feb-2017: Social Solidarity Income (SSI or GMI) scheme.
  - Mid-2017: Establishment and operation of community centers in most municipalities as one-stop-shop for all social programs.
  - May-2017: Abolishment of small family/child allowances.
  - Jan-2018: Consolidation of the two main family benefits into a single benefit.
  - Jan-2018: Modification of disability assessment system.
  - Feb-2018: Establishment of the single public payment authority for all welfare benefits (OPEKA) and IT reforms.
  - Jan-2019: Introduction of new mean-tested housing benefits.
- Despite improvements, significant fragmentation remains: major benefits operate under eight different payment platforms and some small benefits still operate manually.
- Authorities aim to establish a single payment platform for all social benefits by 2024.
- Digitization has facilitated eligibility checkups across platforms, but further data integration and automation is needed.

### Financing of social protection and labor tax wedge
- Government contributions account for about half of total financing of social protection, among the highest in the eurozone, reflecting large state transfers to cover pensions.
- Contributions from private employees and employers are in line with the regional average.
- Labor tax wedge (including personal income taxes and compulsory social contributions paid by both employees and employers) is close to the eurozone average; the most recent cut in SSC during 2021-22 is not captured in the chart.

### Effective tax rates and compliance issues
- Analysis of effective tax rates finds:
  - The poorest group has a relatively high effective tax rate at about 30 percent (income taxes and social contributions/gross income).
  - Differences in tax rates across income groups appear small, but this is biased by underreporting of income by the self-employed.
  - Excluding the self-employed, the tax burden for the poorest group becomes notably smaller and tax rates become more progressive.
  - The self-employed face higher effective tax rates compared to other employment groups (they also pay the employer’s share of social contributions).
  - The self-employed exhibit U-shaped tax rates, with the poorest and richest groups bearing the same high rates—likely reflecting underreporting by poorest self-employed and tax administration efforts to compensate.
- Overall finding: significant compliance gaps in income taxes and social contributions by the self-employed group.

### Distributional impact of climate policies and energy consumption patterns
- Climate-friendly policies that raise carbon prices will increase prices of fossil fuels and energy-intensive goods, affecting direct and indirect household energy consumption.
- Poor and rural households will be hit harder due to lower incomes and a higher consumption share on energy and energy-intensive goods.
- Consumption expenditure (2020, ELSTAT Household Budget Survey 2020):
  - Rich (Top25%): Energy 7.6, High energy intensive 51.2, Low energy intensive 41.2 (values in percent of total consumption).
  - Poor (Bottom25%): Energy 11.8, High energy intensive 53.6, Low energy intensive 34.6.
  - Urban: Energy 9.5, High energy intensive 51.4, Low energy intensive 39.1.
  - Rural: Energy 13.5, High energy intensive 58.7, Low energy intensive 27.9.

### Carbon Pricing Assessment Tool (CPAT) application and methodology
- CPAT: spreadsheet-based tool parameterized to individual countries, projects fossil fuel use and CO2 emissions, fiscal, economic, energy price, and distributional burden of carbon pricing and mitigation policies.
- Distribution module logic (applied to Greece):
  - Uses 2019 Household Budget Survey (HBS) microdata from ELSTAT to calculate shares of direct and indirect energy consumption (8 fuels, 13 goods/service categories).
  - Estimates price changes of goods and services in response to assumed carbon price increases using weighted average fuel intensities of 13 goods/services and sectoral Leontief relationships with respect to energy sections (GTAP10).
  - Calculates consumption effects across income groups via microsimulation; options include excluding cooking fuel by decile (WHO).
  - Allows revenue recycling options: uniform cash by decile; existing social assistance schemes (% increase; ASPIRE); infrastructure access provision.

### Carbon tax scenarios analyzed
- Common assumption: new carbon tax introduced to all non-ETS sectors, carbon price rising from initial level of 25 to 75 real$/tonCO2 by 2030 (the level recommended by the Fund to meet the 2oC global climate goal).
- Reform S1:
  - Recycle half of increases in carbon revenues to transfers to poor households (the bottom 40 percentile) and the other half to labor tax reductions for all taxpayers.
  - Coverage and targeting rates of social transfers used: 68 and 57 percent, respectively (calculated using 2019 HBS data), implying significant leakage of transfers to other groups.
  - Amount of labor tax reductions for each group depends on the size of their initial tax liability.
- Reform S2:
  - Assumes improved targeting: half of carbon revenue increases directed to poor households (bottom 40 percentile) under coverage and targeted rates of 100 and 90 percent, respectively.
  - The other half used for scaling up public investment, notably green infrastructure.

### Environmental, fiscal, and growth outcomes
- Environmental co-benefits (reduced climate-related disaster risk, cleaner air, less traffic jams and accidents) are assessed to outweigh economic costs (deadweight losses) from the carbon tax before revenue recycling, implying net welfare gains for the whole society.
- GDP growth impacts (average over next 15 years):
  - Before revenue recycling, the new carbon tax is estimated to reduce GDP growth by about ⅓ percentage points on average.
  - Recycling carbon revenues for public investment has larger stimulus effects than transfers or tax cuts.
  - In reform S2, the negative growth impact of the carbon tax is fully offset by positive effects from public investment and transfers.

*Source: IMF staff compilation and analysis from the document 1grcea2022002.*

### 21.      Welfare improves for low-income households, notably poor rural groups under the

### 1grcea2022002 - 21.      Welfare improves for low-income households, notably poor rural groups under the

### Welfare impacts of reform scenarios
- For poor households, in both scenarios, their direct and indirect consumption losses, estimated at about 1–1.5 percent of total consumption, are more than offset by the positive gains from targeted transfers.
- Gains are more pronounced in reform S2 due to significantly improved targeting and coverage of transfers.
- Richer households are better off in reform S1 compared to S2 as they benefit more from labor tax reductions.
- Rural households gain more than urban counterparts in both scenarios due to a higher share of low-income groups living in the rural area.
- Poor rural households in reform S2 record the largest net welfare gains.
- Reform S2 stands out with less economic costs and more assistance to poor and rural groups.
- Under reform S2, a new carbon tax combined with better-targeted social transfers and a boost in green investment could support growth, protect vulnerable households while also reducing emissions.

### Economic and distributional channels highlighted
- Direct and indirect consumption losses for poor households: about 1–1.5 percent of total consumption.
- Distributional trade-offs:
  - Labor tax reductions: favor richer households (notably in reform S1).
  - Targeted transfers: favor poor households, especially under improved targeting/coverage (reform S2).
  - Rural vs. urban: rural households benefit more because of concentration of low-income groups in rural areas.
- Environmental and macro effects (scenario comparison):
  - Reform S2 combines a new carbon tax, better-targeted transfers, and increased green public investment to achieve stronger social protection and reduced emissions.

### Policy recommendations and social protection priorities
- Complement climate-friendly policies with social protection reforms to assist the green transition.
- Improve coverage and targeting of social assistance schemes to protect vulnerable households against climate-related disasters and mitigate adverse impacts of higher carbon prices.
- Social protection reform priorities (as listed):
  - (a) expanding coverage of health care and housing benefits (where coverage is low);
  - (b) boosting spending on childcare and the GMI (where targeting accuracy is relatively higher);
  - (c) tackling evasions by the self-employed; and
  - (d) further simplification and consolidation of social assistance schemes.
- Introduce a new carbon tax and gradually increase it over time to finance targeted transfers and green investment.

### Summary conclusion
- A just transition to the green economy requires more ambitious reform efforts. Reform S2—active reform with improved transfer coverage and targeting plus green investment—minimizes economic costs while maximizing protection for poor and rural groups and reducing emissions.

*Source: IMF staff estimates and analysis in 1grcea2022002.*

### 10.      Banks play a dominant role in Greece’s domestic financial system, but their resilience

### 10.      Banks play a dominant role in Greece’s domestic financial system, but their resilience remains weak

### Soundness of the banking system and structural vulnerabilities
- Banks account for around 90 percent of financial system assets in Greece; insurance companies account for around 5 percent; leasing and factoring companies, mutual funds, investment companies, and other financial institutions account for the remainder.
- The 2021 EBA stress test for Greek SIs suggests considerable losses under the adverse scenario assuming a prolonged and severe impact of the pandemic.
- Greece has the lowest bank capital adequacy among EU countries.
- The banking system is highly concentrated: the four major banks identified as systemically important institutions account for more than 95 percent of assets of the banking system, making it the most concentrated system in the euro area.
- Concentration implies significant structural systemic vulnerability from failure of a large and/or interconnected institution and associated moral hazard costs from direct support and implicit government guarantees.

### Macroprudential policy toolkit: current status
- The macroprudential policy toolkit in Greece is solely composed of capital buffers (CCoB, CCyB, and O-SII buffers).
- The framework for the CCyB has been implemented since 2016 but has never been activated.
- The framework for the O-SII buffer was adopted in 2014 and operationalized since 2016, with the phase-in period up until 2023.
- There is no policy framework in place to address household and real estate market risks via borrower-based measures, risk weights, or other capital measures; Greece is one of the very few EU countries hinging only on CCoB and O-SII buffers.

### Recent developments and international experience relevant for Greece
- A broader set of indicators beyond the Basel credit gap is being used across Europe to determine CCyB rates; several countries have introduced or announced positive CCyB rates.
- BoG has revised its methodology for cyclical systemic risk assessment (BoG, 2022a), retaining the credit gap but adding indicators covering private sector debt burden, real estate prices, soundness of credit institutions, and risk pricing.
- Ireland’s experience: activated CCyB in 2018 at 1 percent despite mixed signals from Basel credit gap; complemented by borrower-based measures (LTI and LTV); released buffer in April 2020 to cushion the pandemic shock; planned gradual rebuilding of CCyB in 2022 as conditions allowed.

### O-SII buffer: implementation and assessment
- BoG’s O-SII criteria follow Basel international standards and EU regulations; criteria include size, interconnectedness, substitutability, and complexity.
- Designated O-SIIs: Alpha Bank, Eurobank, National Bank of Greece, and Piraeus Bank.
- Banks are subject to ECB minimum floor requirements per bucket associated with systemic significance score.
- Buffer rates are subject to phase-in; buffer rates for Greek banks appear relatively low relative to their systemic importance scores compared to peer O-SII banks supervised by the SSM.
- Recommendation (short term): proceed with the envisaged phase-in period and ensure full phasing-in of the remaining buffer increase in 2023.
- Recommendation (medium term): review framework to ensure full adequacy of buffers for identified vulnerabilities and align with EBA harmonization work.

### CCyB: assessment and policy guidance
- Immediate activation of CCyB is judged suboptimal given relatively low private sector indebtedness and ongoing rapid clean-up of bank balance sheets.
- Emerging signs of systemic risk build-up warrant close monitoring and a plan to rebuild resilience over time.
- Staff recommend enhancing the methodology for cyclical systemic risk assessment and introducing a positive neutral CCyB rate gradually over the medium term.
- Suggested additional indicators to include in methodology:
  - (i) new lending flows,
  - (ii) real estate price valuations and supply-side constraints,
  - (iii) alternative measures of credit gaps.
- Preconditions for introduction of a positive neutral CCyB rate:
  - (a) a firm economic recovery;
  - (b) advanced clean-up of bank NPLs;
  - (c) favorable financing conditions.
- Activation of the CCyB would create releasable buffers to support credit provision and reduce output losses during stress.

### Conclusions and policy implications
- Signals of emerging systemic vulnerabilities:
  - After extended private sector deleveraging, indicators suggest some releveraging.
  - Basel credit gap remains negative, but alternative measures show positive gaps; all measures point to releveraging.
  - New lending to households has surpassed disposable income growth.
  - Residential real estate prices (and to a lesser extent commercial prices) have increased markedly since 2018, with a significant increase in the price-to-rent ratio and some overvaluation in the residential segment despite no discernible supply constraints.
  - Structural vulnerabilities stem from a highly concentrated and weakly capitalized banking system.
- Macroprudential framework revisions needed:
  - Review and revise methodologies underpinning CCyB and O-SII buffer rate determination in the short term.
  - Prepare conditions-based roadmaps to guide activation of CCyB and potential borrower-based measures over the medium term.
- Calibration work required:
  - For CCyB calibration: consider early warning models of banking crises, general equilibrium models with explicit bank capital role, regulatory and market-based stress tests.
  - For borrower-based measures: build on BoG work in line with ESRB Recommendation and advance granular data gathering before proceeding to calibration.

*Source: IMF staff analysis as presented in the chapter "Banks play a dominant role in Greece’s domestic financial system, but their resilience remains weak" (1grcea2022002).*

### References

### BANK PROFITABILITY DRIVERS AND CHALLENGES IN GREECE

### A. Introduction
- Greece’s bank profitability has sharply deteriorated recently; the aggregated measure of bank profitability, return on equity (RoE), has slipped into negative territory over the last two years, culminating in an almost 20 percent drop at end-2021 (EBA, 2022).
- The drop reflected primarily NPL disposal and the impact of the pandemic shock, although the net interest margin that constitutes the key income stream deteriorated earlier (BoG, 2021).
- Net fee and commission income has partially offset the decline, but profitability remains significantly below the average for euro area banks (BoG, 2022).
- Low profitability can pose significant financial stability risks:
  - Weak profitability reduces capacity to build buffers against unexpected shocks (ECB, 2019).
  - It can incentivize excessive risk taking to compensate for lower underlying profitability (Babihuga and Spaltro, 2014).
  - It may reduce attractiveness for investor capital (Gopinath and others, 2017).
  - Lower profitability can limit bank capacity to fund loan growth and, in Greece, bank losses could trigger Deferred Tax Credit conversion and ensuing capital dilution.
- Prudential responses and supervisory focus:
  - ICAAP and BMA form key elements of the SREP framework, assessing historical profitability and forward-looking stress tests (EBA 2010; 2018b).
  - ECB supervisory priorities have focused on profitability drivers within business model assessments (ECB 2018; 2019a; 2021).
- Strategic note:
  - Raising sustained profitability is key to strengthen resilience and to prepare banks to face competition from non-banks; banks need to enhance risk management frameworks and adapt business models.

### B. Bank Profitability Drivers
- Frameworks and models referenced:
  - Bank dealership model linking profitability to bank characteristics: Ho and Saunders (1981) and extensions (Angbazo 1997; Maudos and Fernandez de Guevara 2004).
  - Models linking profitability to macroeconomic environment: Gerali and others (2009; 2010); Coffinet and others (2009); Henri and Kok (2013); Dees and others (2017).
- Banking-variable mechanisms:
  - Cost efficiency, capitalization, and size affect bank margins.
  - Higher risk aversion -> better capitalization -> lower funding costs -> higher margins.
  - Operating costs tend to be passed on to clients, increasing margins.
  - Higher loan volume for a given total amount can raise administrative overhead and lower margins (Angbazo 1997; Maudos and Fernandez de Guevara 2004), though larger loan size can produce returns to scale (Engle and others 2014; Borio and others 2017).
  - Concentration effects are ambiguous and require control (Saunders and Schumacher, 2000; Berger, 1995).
- Macroeconomic-variable mechanisms:
  - Banks’ market power determines ability to pass on funding cost increases and to reprice loans.
  - Higher long-term rates allow loan repricing; lower long-term government bond rates can squeeze margins and shift focus to fee-earning activities (Albertazzi and Gambacorta 2009).
  - Improved economic conditions raise lending demand and borrower financial conditions, supporting profitability (DeYoung and Rice 2004; Coffinet and others 2009; Albertazzi and Gambacorta 2009).
  - Bank income tends to rise with better stock market performance (Lehmann and Manz 2006; Kok and others 2019; Gross and others, 2021).
- Empirical literature on Greece:
  - Earlier studies (Staikouras and Steliaros 1999; Eichengreen and Gibson 2001; Mamatzakis and Remoundos 2003; Athanasoulou et al 2008; Kosmidou 2008; Alexiou and Sofoklis 2009) find banking variables and the business cycle important for RoE.
  - Louzis and Vouldis (2015) provide a comprehensive analysis of net interest income and net fee and commission income for 2004-2011, but structural changes in the system followed that period.

### C. Empirical Analysis of Profitability Drivers
- Descriptive findings:
  - Bank profitability has suffered a series of substantial shocks over the last decade: large NPLs after the GFC, massive losses, and additional pressure from the pandemic.
  - Impairments from loan loss provisions for NPLs have often exceeded net interest income; cost of risk spiked while expenses remained supportive owing to staff and branch reductions.
  - Traditional banking intermediation (net interest income) is the dominant income source:
    - Net interest income and net fee and commission income historically accounted for 70 to 90 percent of operating income.
    - Net interest margin (net interest income scaled by assets) has shown a generally declining trend due to the low-rate environment, partially offset by yield curve steepening; subsequent yield curve flattening, the pandemic, and NPL sales further depressed margins.
    - Net fee and commission income is much lower in significance compared to the euro area average but has shown better performance for Greece when expressed in earning assets terms.
- Econometric (panel regression) insights:
  - Dataset: unbalanced panel of banks for 2007–2019; net interest income and net fee and commission income analyzed separately (data and methodology described in the Annex).
  - Net interest income determinants:
    - Banking variables: higher operating costs, higher risk aversion, and higher credit risk associate with higher margins; greater transaction size associates with lower margins.
    - Concentration evidence is inconclusive.
    - Macroeconomic variables: interest rate level, term spread (yield curve slope), and economic growth materially affect net interest income.
    - Relative importance: macroeconomic variables are substantially more significant than banking variables in their impact on net interest income.
  - Net fee and commission income determinants:
    - Banking variables: higher capitalization and higher cost efficiency associate with higher net fee and commission income; higher transaction size associates with lower net fee and commission income.
    - Macroeconomic variables: economic growth and stock market performance support higher fee and commission income; interest rates have an adverse impact on this income item.
    - Relative importance: macroeconomic variables are more important than banking variables, though the contrast is less stark than for net interest income.
- Empirical implications:
  - Profitability is strongly influenced by macroeconomic conditions (interest rates, yield curve slope, GDP growth, stock market performance) alongside bank-specific characteristics (capitalization, cost efficiency, operating costs, transaction size, credit risk).
  - Given macroeconomic dominance, profitability improvements will depend materially on macro-financial environment, while banks can affect outcomes via cost and capital management, fee income strategies, and risk management enhancements.

*Source: IMF staff compilation from the provided content.*

### 13.      The scenario analysis hinges on models embedding only macroeconomic variables. The

### 13.      The scenario analysis hinges on models embedding only macroeconomic variables. The

### Model comparison and forecasting accuracy
- Models including only macroeconomic variables are compared to models including both macroeconomic and banking variables using the root mean squared error (RMSE) and the U-Theil ratio over different horizons.
- All models are also compared to the basic autoregressive model, AR(1), in terms of RMSE and the U-Theil ratio.
- Findings:
  - Macroeconomic variables dominate in explaining the main bank income sources.
  - Models accounting for banking variables provide valuable insights into drivers of profitability, consistent with theoretical and empirical literature, but their forecasting performance does not systematically appear superior and there are significant costs associated with those models.
  - Testing for forecasting accuracy suggests that models composed of macroeconomic variables do not generally perform worse compared to models composed of only banking variables or banking and macroeconomic variables.
- Practical cost noted:
  - Costs involve the need to make assumptions for banking variables across all banks over the projection horizon, requiring detailed information on bank management policy and potentially differing management responses across scenarios.

### Macroeconomic assumptions and scenarios
- Scenarios are based on a consistent set of macroeconomic and financial projections and are mainly illustrative.
- Scenarios used:
  - Benchmark scenario: baseline from the EU 2020 stress test.
  - Mid-adverse scenario: captures a degree of severity associated with the Covid-19 shock.
  - Severe-adverse scenario: captures a larger degree of severity associated with the Covid-19 shock.
- Scenarios are fed into the estimated benchmark models using macroeconomic and financial scenarios prepared by the ECB for Greece in the context of the 2020 Vulnerability Analysis.

### Results for key income sources
- Net interest income:
  - Benchmark scenario generates on average a steady improvement in the net interest margin over the projection horizon.
  - Adverse scenarios: net interest income declines significantly, especially in year 1 mirroring the projected V-shaped recovery profile, notwithstanding the assumed yield curve slope steepening.
- Net fee and commission income:
  - Benchmark scenario: generally stable with, on average, a mild trend upwards.
  - Severe-adverse scenario: envisages net fee and commission income plummeting in year 1 and remaining depressed in year 2 and 3.
  - Difference in rebound profiles between net interest income and net fee and commission income is likely driven by stock market variables for which only a limited recovery is assumed.

### Aggregated profitability (RoE) projections
- RoE requires additional assumptions for impairments and provisions, operating expenses, and other items.
- Assumptions noted:
  - Inorganic actions to reduce NPLs through the Hercules program with state guarantees are expected to come to an end soon, implying only a moderate contribution from securitizations going forward.
  - For operating expenses, an average based on the last five years is assumed.
  - Assets are assumed to grow with GDP.
- Simulation results:
  - Gradual recovery of aggregated profitability over the next three years under the baseline scenario.
  - In terms of RoE, the simulation suggests a return to the range of 7-9 percent, which is broadly in line with results recorded between 2001 and 2004, but well short of the boom years spanning the period 2005-2008.
  - The simulated range is in line with market analysts’ forecasts of RoE for the major Greek banks.

### Cost of Equity (CoE)
- CoE for banks equals the compensation market participants demand for investing and holding banks’ equity.
- Relevance:
  - A high CoE and ensuing limitations for raising new capital may prevent banks from enhancing buffers against negative shocks.
  - Supervisors may compare CoE to past and projected RoE; RoE should not be persistently below CoE.
- Empirical estimates:
  - EBA (2021) estimates: around 40 percent of EU banks estimated their CoE between 8 and 10 percent and another 30 percent between 10 and 12 percent, with the latter rising significantly.
  - Market-implied CoE for Greek banks estimated within the range of 12-17 percent.
- Implication:
  - Comparing simulated RoE to the estimated CoE range of 12-17 percent points to a gap between investors’ expectations and projected bank profitability performance.

### Loan pricing
- Adequate loan pricing is essential for business model sustainability; pricing should include cost of funding, credit risk cost, cost of capital, operating expenses, and market conditions.
- Evidence for Greece:
  - Lending margins in Greece have generally been well above the EA average, except for Ireland and Cyprus.
  - Despite high margins, calculations suggest loan pricing may not have covered all relevant costs and risks for both corporate and mortgage loans.
  - Expected losses associated with credit risk, capital charges, and other costs weigh heavily on margins; corporate lending especially may reflect a high degree of competition among banks.
- Policy references:
  - ECB (2017, 2018) and EBA (2020) methodologies and guidelines on loan pricing and loan origination/monitoring are relevant for assessing and improving pricing adequacy.

### Conclusions and policy implications
- Key conclusions:
  - The main bank income sources (net interest income and net fee and commission income) are determined by both macroeconomic and banking variables, with macroeconomic variables dominating in economic significance.
  - Models relying solely on macroeconomic variables are credible for the scenario analysis given their forecasting performance and the dominant role of macro variables.
  - Aggregated profitability is expected to rebound in the near term under the benchmark scenario, but sensitivity to macroeconomic performance warrants caution.
  - Even if the benchmark scenario materializes, estimated margins underpinning simulated RoE may fall short of the market-implied CoE and may not fully reflect Greece’s risk profile, indicating a need to adapt bank business models.
- Policy implications for supervisors:
  - Pay attention to managing interest rate risk via ICAAP components focused on interest rate risk of the banking book and stress testing.
  - Monitor weak capital position and significant credit risk and require banks to take additional actions, especially in the context of EBA guidance on loan pricing.
  - Investigate avenues to sustain increased income from fees and commissions, potentially in the context of the BMA.
  - Utilize SREP prerogatives to influence banks to adapt business models and restore sustainable profitability drivers, in line with recent strategic supervisory priorities (ECB, 2021).

*Source: IMF staff estimates and analysis as presented in the provided content.*

### References

### References

### Literature Base
- Key studies and sources cited:
  - Albertazzi and Gambacorta (2009); Alessandri and Nelson (2015); Angbazo (1997); Arellano and Bond (1991); Athanasoglou, Brissimis, and Delis (2008); Babihuga and Spaltro (2014); Berger (1995); Blundell and Bond (1998); Borio, Gambacorta, and Hofmann (2017); Bruno (2005); Claessens, Coleman, and Donnelly (2018); Coffinet, Lin, and Martin (2009); Dang, Kim, and Shin (2015); Dees, Henri, and Martin (2017); DeYoung and Rice (2004); Diebold (2017); Drakos (2002); EBA guidelines and dashboards (2010–2022); ECB publications and speeches (2017–2022); Engle et al. (2014); European Systemic Risk Board (2020); FED (1996); Gerali et al. (2010); Gross, Jarmuzek, and Pancaro (2021); Hahm (2008); Henry and Kok (2013); Ho and Saunders (1981); Jarmuzek and Lybek (2018, 2020); Kiviet (1995); KPMG (2021); Kok, Mirza, and Pancaro (2019); Kosmidou (2008); Lehmann and Manz (2006); Louzis and Vouldis (2015); Mamatzakis and Remoundos (2003); Maudos and Fernández de Guevara (2004); Oliveira and Elliott (2012); Saunders and Schumacher (2000); Staikouras and Steliaros (1999); Tennant and Sutherland (2014); Timmermann and Zhu (2019).
- Regulatory and supervisory sources: EBA guidelines (GL32, EBA/GL/2014/13, EBA/GL/2015/08, EBA/GL/2016/10, EBA/GL/2018/02), EBA Risk Dashboard (2021, data as of 4q 2021), ECB publications (including SSM priorities and vulnerability analyses), BCBS material referenced for capital-charge calculations.

### Thematic emphasis from cited literature (as used in the analysis)
- Bank profitability drivers: banking variables (net interest income, net fees and commission income, operating expenses, capital adequacy, size, asset riskiness, concentration) and macro/financial variables (short-term rate, yield curve slope, GDP growth, stock market returns).
- Econometric concerns and estimator choice: dynamic panel bias, endogeneity, slope homogeneity testing, multicollinearity, and forecasting performance.

### Annex I — Technical Aspects: Methodology
- Empirical specification (as presented):
  - Y_{i,t} = α_0 + Σ_{j=1}^{p} α_{1} Y_{i,t-j} + Σ_{k=1}^{q} Σ_{j=0}^{l} α_{2,k} BV_{k,i,t-j} + Σ_{k=1}^{r} Σ_{j=0}^{m} α_{3,k} MACV_{k,i,t-j} + ε_{i,t}
  - Y denotes bank profitability; BV denotes banking variables; MACV denotes macroeconomic and financial variables; model includes lagged dependent variable.
- Pre-estimation diagnostics:
  - Slope homogeneity tested using Pesaran and Yamagata (2008) and Blomquist and others (2013).
  - Multicollinearity tested via variance inflation factors.
- Estimators and rationale:
  - Fixed-effect (FE) estimator used as benchmark.
  - Bias-corrected FE estimator (Kiviet 1995; Bruno 2005) employed to address Nickell bias and shown to perform better than IV and GMM in presence of unobserved heterogeneity and residual serial correlation (citing Dang and others 2015; Beutler and others 2020).
  - GMM estimators (Arellano and Bond 1991; Blundell and Bond 1998) used as robustness checks to account for potential endogeneity.

### Annex I — Data
- Sample:
  - Annual bank-level data from Fitch and macroeconomic variables from ECB SDW.
  - Coverage: 100 banks located in euro area countries subject to direct supervision of the SSM.
  - Period: 2005-2019.
- Dependent variables:
  - Net interest income over assets (nim).
  - Net fees and commission income over assets (nfc).
- Bank variables and definitions (Table 1 excerpts — abbreviation, variable, definition, source):
  - nim: net interest income — net interest income over asset — Fitch.
  - nfc: net fees and commission income — net fees and commission income over assets — Fitch.
  - opex: operating expenses — administrative expenses over assets — Fitch.
  - capital: capital adequacy — common equity over assets — Fitch.
  - size: size — log of assets — Fitch.
  - riskiness: asset riskiness — risk-weighted assets over total assets — Fitch.
  - concentration: concentration — share of three largest banks in total assets — Fitch.
  - short-term rate: money market rate — money market rate — ECB SDW.
  - yield curve slope: yield curve slope — long-term rate minus short-term rate — ECB SDW.
  - GDP growth: GDP growth rate — annual growth rate of gross domestic product — ECB SDW.
  - stock market: stock market return — national stock market index change — ECB SDW.
- Data treatment:
  - Variables were winsorized to limit outlier effects.

### Annex I — Results: Tests and Key Coefficients
- Slope homogeneity tests (Table 2) — Delta and p-value:
  - Net interest income:
    - Persaran-Yamagata: Delta = -1.208, p-value = 0.227
    - Blomquist-Westerlund: Delta = -1.904, p-value = 0.057
  - Net fees and commission income:
    - Persaran-Yamagata: Delta = -1.209, p-value = 0.227
    - Blomquist-Westerlund: Delta = -0.862, p-value = 0.389
- Selected regression estimates for Net Interest Income (Table 3) — illustrative coefficients (FE, corrected FE, GMM where reported):
  - Lagged dependent variable:
    - 0.637*** (FE), 0.764*** (corrected FE), 0.732*** (GMM) in columns (1)-(3) reported set; other specifications show lagged coefficients ranging from 0.646*** to 0.928*** across columns (1)–(9).
  - Opex:
    - Examples: 0.0033 (FE), 0.000866 (corrected FE), 0.0471*** (GMM) in one block; elsewhere 0.0383**, 0.0373***, 0.0779*** in another block.
  - Capital:
    - 0.0274*** (FE), 0.0265*** (corrected FE), 0.0213*** (GMM) in one block; other blocks report 0.0309***, 0.0284***, 0.0149***.
  - Size:
    - -0.0621* (FE), -0.0576*** (corrected FE), -0.0118*** (GMM) in one block.
  - Asset riskiness:
    - 0.00461*** (FE), 0.00303*** (corrected FE), 0.00607*** (GMM) in one block.
  - Short-term rate:
    - Coefficients reported include 0.0151***, 0.00444***, 0.0239***, 0.0237***, 0.0183***, 0.00667*** across specifications.
  - Yield curve slope:
    - Examples: 0.0164*; 0.0102; 0.00301***; 0.015; 0.0111***; 0.00883*** across specifications.
  - GDP growth:
    - Examples: 0.0144*, 0.0177***, 0.0226***, 0.0153***, 0.0159***, 0.0214*** across specifications.
  - Estimation sample size for NIM models:
    - N of groups = 93; Observation = 968.
- Selected regression estimates for Net Fee and Commission Income (Table 4):
  - Lagged dependent variable:
    - 0.546*** (FE), 0.673*** (corrected FE), 0.786*** (GMM) in one block; other blocks show lagged coefficients ranging from 0.519*** to 0.869*** across columns (1)–(9).
  - Opex:
    - 0.0477*** (FE), 0.0465*** (corrected FE), 0.0545*** (GMM) in one block; other blocks show 0.0618***, 0.0610***, 0.0688***.
  - Capital:
    - 0.0134*** (FE), 0.0125*** (corrected FE), 0.00627*** (GMM) in one block.
  - Size:
    - -0.0409*** (FE), -0.0319* (corrected FE), -0.00567*** (GMM) in one block.
  - Concentration:
    - 0.000447, 0.000837, -0.00142*** across a block.
  - Long-term rate and Δlong-term rate:
    - Long-term rate examples: -0.00455*; -0.00324; 0.00391***; -0.00448*; -0.00274; -0.00899*** across specifications.
    - Δlong-term rate examples: -0.00205; -0.00559; -0.0172***; -0.00364; -0.00658; -0.00313*** across specifications.
  - GDP growth:
    - Coefficients include 0.00555***, 0.00519***, 0.00439***, 0.00932***, 0.00860***, 0.00679*** across specifications.
  - Stock market:
    - Coefficients include 0.000653***, 0.000712***, 0.000853***, 0.000547***, 0.000594***, 0.000539*** across specifications.
  - Estimation sample size for NFC models:
    - N of groups = 106; Observation = 1263.
- Homogeneity testing for Greek banks (Table 5) — interaction coefficients (examples):
  - Opex * Greek banks dummy: 0.0834 (NIM), -0.0409 (NFC).
  - Capital * Greek banks dummy: 0.0180 (NIM), -0.00815 (NFC).
  - Size * Greek banks dummy: 0.211 (NIM), -0.0203 (NFC).
  - Concentration * Greek banks dummy: -0.00877 (NIM), -0.00339* (NFC).
  - Yield curve slope * Greek banks dummy: -0.0344* (NIM).
  - Observations: NIM observation = 968; NFC observation = 1199.
- Relative strength of explanatory variables (Table 6) — implicit elasticities evaluated at sample means:
  - NIM:
    - Operating expenses: 1.5
    - Capital adequacy: 1.7
    - Size: -0.4
    - Concentration: 0.0
    - Asset riskiness: 0.1
    - Short-term rate: 3.4
    - Yield curve slope: 1.2
    - GDP growth: 2.5
  - NFC:
    - Operating expenses: 2.4
    - Capital adequacy: 0.5
    - Size: -0.2
    - Concentration: 0.0
    - Long-term rate: -0.2
    - Long-term rate change: 1.7
    - GDP growth: 1.4
    - Stock market return: 0.4

### Forecasting Performance (Tables 7–8) — model accuracy metrics (UTheil)
- Net Interest Income (Table 7) — selected UTheil values (Model vs Benchmark AR(1)):
  - FE model Horizons:
    - Horizon 1: Model = 0.3399, Benchmark = 0.2498, UTheil = 1.6325
    - Horizon 2: Model = 0.3442, Benchmark = 0.3439, UTheil = 1.0007
    - Horizon 3: Model = 0.3389, Benchmark = 0.3842, UTheil = 0.8820
  - Corrected FE (examples):
    - Horizon 1: Model = 0.2932, Benchmark = 0.2495, UTheil = 1.1755
    - Horizon 3: Model = 0.2743, Benchmark = 0.3842, UTheil = 0.7139
  - GMM shows larger UTheil values in reported cells (e.g., Horizon 1: UTheil = 3.0207 for one configuration).
- Net Fee and Commission Income (Table 8) — selected UTheil values:
  - FE model Horizons:
    - Horizon 1: Model = 0.2196, Benchmark = 0.1210, UTheil = 1.8158
    - Horizon 2: Model = 0.2181, Benchmark = 0.1756, UTheil = 1.2417
  - Corrected FE (examples):
    - Horizon 1: Model = 0.1964, Benchmark = 0.1210, UTheil = 1.6239
    - Horizon 3: Model = 0.1515, Benchmark = 0.1768, UTheil = 0.8570
  - GMM shows larger UTheil values (e.g., Horizon 1: UTheil = 4.2147 in one configuration).

### Loan Pricing (methodology and parameters)
- Standard components that lending rates should cover (building on FED (1996); Elliott and others (2012); ECB (2017); EBA (2020)):
  - (i) Funding cost — country-specific funding spreads depend on maturity, instruments, and funding mix (ECB).
  - (ii) Expected credit losses — estimated as probability of default × loss given default parameters (EBA).
  - (iii) Capital charge — risk weights from Basel formulae applied to exposures; assumed cost of equity equals the lower bound of CoE extracted from Bloomberg and equals 12 percent (BCBS, Bloomberg).
  - (iv) Operational cost — cost-to-income ratio assumed uniformly distributed across income streams (ECB).
  - Residual approximates excess margin obtained by banks.

### Contextual policy/analytical note from concluding text
- High-level finding (from following chapter opening included in the source):
  - "Greece needs to boost its savings in order to maintain external and debt sustainability while closing its significant investment gap."
  - Household savings are noted as "strikingly low over the last two decades," with some pandemic-period increase judged "at least partially temporary."
  - Structural impediments to raising household savings include "high unemployment and wide-spread informality."

*Source: IMF staff estimates.*

### 1.      Savings will need to increase significantly to simultaneously support external

### 1.      Savings will need to increase significantly to simultaneously support external sustainability and a strong investment recovery

### Key messages and macro targets
- Greece’s net international investment position (NIIP) stands at -180 percent of GDP.
- To converge to a more sustainable pre-crisis NIIP level (around -70 percent of GDP), the current account will need to decline from pandemic levels; applying the IMF external sustainability approach (IMF WP/19/65) a current account of about -3 percent of GDP is needed for the NIIP to stabilize around -70 percent of GDP (its 2000-2010 average).
- National savings need to increase from 8 percent of GDP in 2020 to an average of around 17 percent of GDP in the medium-term to simultaneously (i) support external sustainability and (ii) raise investment to close a sizable investment gap (IMF WP/22/13).

### Rationale for higher savings
- Greece has been underinvesting since the Sovereign Debt Crisis (SDC), accumulating a sizable investment gap.
- Leveraging Next Generation EU funding and raising investment is imperative to improve potential growth and living standards.
- Higher national savings would also aid the rollover of a large stock of official external loans that were transformed at the time of the SDC into ultra-long maturity, favorable-rate official external loans.

### Official loan rollover and investor base
- The ECB holds €38.5 billion of Greek securities and is highly exposed (around 65% of available eligible securities).
- As these official loans mature over the medium- to long-term, they will need to be rolled over to private holders; a recovery to investment grade will help widen the investor base.
- Higher domestic savings could allow the domestic non-financial private sector to hold more public debt, diversifying the investor base and supporting favorable and stable financing conditions.

### _Prepared By Shiqing Hua, Johanna Schauer and Wei Shi._

---

### Stylized facts on savings (recent decades)
- Greece’s gross saving in 2020 is estimated to be 7.1 percent of GDP, less than a third of the Euro Area average.
- Savings rebounded after the SDC from 4.7 percent of GDP in 2011 to 10.6 percent of GDP in 2016, but edged downward in recent years.
- Sectoral contributions:
  - Public saving: Historically the main driver of low savings; mostly negative and bottomed in 2009; improved fiscal position turned public saving positive from 2016 before falling again in 2020.
  - Financial corporations’ savings: Relatively stable and above the regional level for most of the last two decades.
  - Non-financial corporation savings: Close to the EA average until 2014, then declined in tandem with rising debt levels.
  - Household savings: The widest and most persistent gap—around 10.6 percentage point lower compared to the EA average—and the principal focus of the analysis going forward.

### Household financial asset trends
- Over €106 billion (31 percent) of household financial assets vanished from 2007 to 2012.
- Equity, investment fund shares and debt securities experienced rapid declines, while cash increased (shift to safer, more liquid assets).
- Household financial assets abroad initially dropped, then surged as the SDC deepened, with households moving savings abroad and storing them in cash, deposits and stocks.

### Distributional patterns (experimental and survey evidence)
- Experimental data (Eurostat 2020) and Household Budget Survey evidence indicate:
  - Close to 60 percent of Greek households were dis-saving in 2015, almost twice the average level in other EU countries.
  - Poorer and younger households drove the decline in household savings.
  - Greece had the highest Gini coefficient for household savings in the Euro Area.
- Caveats: Experimental joint distributions rely on strong assumptions; Household Budget Survey may under-report household income and thus under-estimate household savings.

---

### Household savings during the COVID-19 pandemic
- Greek households accumulated over €7 billion worth of savings from 2020:Q2 to 2021:Q2, with increases observed across a broad range of households.
- Factors behind the surge:
  - Proactive fiscal policy support that sustained household disposable income despite output declines.
  - Lower consumption, notably a drastic drop in services expenditures due to mobility restrictions and social distancing.
- Although households began to dissave in mid-2021, savings remained significantly above pre-pandemic levels.

### Empirical evaluation (panel framework and short-term determinants)
- Baseline specification follows Mody and others (2012): aggregate household savings rate (household gross savings / gross disposable income) related to household precautionary motives—risks of income loss, household wealth (proxied by house prices), relative price of savings (real short-term deposit rate), fiscal stance, and aggregate income uncertainty (output volatility).
- Data: Quarterly sample covering eleven eurozone countries during 2000:Q1–2021:Q3. Limited quarterly household wealth indicators prevented direct inclusion; house prices used as proxy.
- Key regression results (panel; significance and p-values preserved exactly as reported):
  - Expected employment coefficient: -4.6599*** (0.0004)
  - Gross disposable income, q/q, lead: -0.0800 (0.1481)
  - Real short-term deposit rate: -0.4148*** (0.0000)
  - GDP volatility: 2.2137*** (0.0000)
  - Fiscal balance (% of GDP), 4-quarter rolling average: -0.1569*** (0.0000)
  - Observations: 884
  - Country fixed effect: Yes
  - Number of countries: 11
  - Note: *, **, and *** indicate statistical significance at 10, 5, and 1 percent, respectively. P-values in parentheses. All series are seasonally adjusted.

### Interpretation of empirical results
- Supportive fiscal policy and elevated aggregate output volatility together explain over half of the increased household savings in 2020 relative to end-2019.
- The negative coefficient on expected employment implies higher precautionary savings when employment prospects worsen.
- The negative coefficient on the real short-term deposit rate indicates an income effect (higher real deposit rates induced more consumption in the current period) dominated any substitution effect.
- Employment risks and deposit rates explain a smaller share of the cumulated pandemic savings than aggregate factors like fiscal stance and output volatility.

### Forced savings and COVID-19 restrictions
- Around half of the increase in household savings rate in 2020–21 cannot be accounted for by the precautionary-savings model; this residual is positively correlated with the stringency of COVID-19 restrictions.
- The quarterly residual savings are lower in 2021 despite similar stringency levels, suggesting household adaptation of consumption behavior over time.
- Constructed forced savings are orthogonal to the precautionary-savings estimated in the first stage.

### Likely persistence of pandemic savings
- The high level of household savings during the pandemic is likely temporary because:
  - Output volatility is expected to subside as the economy recovers, reducing employment and income uncertainty.
  - Government extraordinary policy support is expected to be phased out, reverting to a tighter fiscal stance.
  - COVID-19 restrictions will be loosened and households will continue adapting consumption behavior.
- Structural policies and drivers that could make higher household savings more persistent are discussed in the subsequent section (not included in this excerpt).

---

*Source: IMF staff chapter excerpt prepared by Shiqing Hua, Johanna Schauer and Wei Shi.*

### 12.      A cross-country regression model aims to identify key savings determinants in the

### 1grcea2022002 - 12.      A cross-country regression model aims to identify key savings determinants in the

### Model design and data
- Panel of 16 European Union countries between 2005 and 2019.
- Dependent variable: aggregate household savings expressed as a share of GDP (household savings rate).
- Five groups of explanatory factors: general macro variables, demographic factors, fiscal variables, financial variables, structural policy variables.
- No fixed effects included to allow medium-term projections and capture slow-moving structural cross-country differences.
- Main specification selected based on explanatory power, consistency of variables, and fit for Greece.
- Methodology note: baseline regression may be subject to endogeneity; baseline appears largely robust to lagging independent variables.

### Regression results — broad findings
- Results align broadly with economic theory, reflecting income, substitution, and precautionary effects.
- Macro factors:
  - Real interest rate: negative (albeit insignificant) impact on savings.
  - Inflation: reduces savings; interpreted as households frontloading consumption expecting further price increases.
  - Expected employment: higher expected employment lowers savings (lower precautionary saving).
  - GDP volatility: does not show the usual positive relationship in this specification; likely reflects absence of country fixed effects and correlation of higher volatility with poorer countries that have lower saving rates.
  - Persistent crisis dummy (2008 onwards): strongly negative and significant coefficient, suggesting the global financial crisis shifted household behavior toward lower savings.
- Demographic factors:
  - Higher share of old age employment raises household savings, reflecting longer working lives lengthening the saving period.
- Fiscal factors:
  - General government balance (percent of GDP): negative and significant coefficient, consistent with some Ricardian offset.
  - Pension spending: inverse-U relationship — initially raises savings but beyond 13 percent of GDP substitution effect reduces precautionary saving.
  - Other social spending (excluding pensions): negative and significant impact, suggesting higher social spending reduces precautionary savings.
- Financial factors:
  - Private sector credit (% GDP): negatively associated with household savings, consistent with easier access to credit reducing precautionary motive.
- Structural factors:
  - Unemployment rate: negatively associated with household savings.
  - Self-employment (% 15+ employment): negatively associated with household savings; may proxy informality and underestimation of disposable income.
  - Gini index: small and insignificant impact.
  - Real effective exchange rate (Reer, 2010=100): positive relationship with savings rate.
  - Capital controls:
    - Overall inflow restriction index: higher controls on capital inflows reduce savings rate.
    - Overall outflow restriction index: associated with higher savings rate.

### Key regression coefficients and fit (Table 1 — Greece: Regression Results and Fit)
- Macro
  - Real short-term interest rate (%) -0.133
  - Inflation (%) -0.180**
  - Expected Employment -0.0551***
  - Persistent crisis dummy (2008 onwards=1) -0.898***
  - GDP Volatility -1.875
- Demographic
  - 65+ employment (% 15+ employment) 0.486***
- Fiscal
  - General Government Balance to GDP (%) -0.201***
  - Pension spending (% GDP) 0.845***
  - Pension spending (squared) -0.0321***
  - Other social spending (% GDP) -0.297***
- Financial
  - Private sector credit (% GDP) -0.00980*
- Structural
  - Unemployment rate (%) -0.212***
  - Self-employment (% 15+ employment) -0.312***
  - Gini index 0.00674
  - Reer (2010=100) 0.0785**
  - Overall inflow restriction index -12.34***
  - Overall outflow restriction index 1.988***
- Constant 9.688**
- Overall R-squared 0.756
- Number of Observations 240
- Significance notation: * p<0.10 ** p<0.05 ***p<0.01

### Model fit for Greece and historical drivers
- Model explains a large part of Greece’s lower savings rate relative to the sample average and the widening of the gap over time.
- Fitted values match key trends for Greece:
  - Higher pre-SDC savings rate,
  - Subsequent fall to negative levels,
  - Recent increase during the pandemic.
- Main structural contributors to Greece’s savings gap (2005–2019 averages and changes noted):
  - Self-employment: Greece average 30 percent (2005–2019) vs. sample average 16 percent.
  - Unemployment: Greece average 18 percent vs. sample average 9 percent.
  - Strict controls on capital inflows: index 0.23 vs. sample average 0.13.
- Contributors to the drop in savings rate between 2005 and 2013:
  - Unemployment rose from 10 percent to 28 percent (key contributor).
  - Self-employment increased from 30 percent to 32 percent.
  - Fiscal factors played a smaller role: government deficit declined and pension spending increased.
- Subsequent period: reductions in unemployment, self-employment, and pension spending would have pushed savings up but were largely offset by stricter controls on capital inflows, further improvements in the government balance, and increases in social spending.

### Projections and scenarios
- Baseline projection (2019–2027):
  - Removal of capital controls in 2019 and projected decline in unemployment from 17.5 percent in 2019 to 10.6 percent in 2027 expected to keep the savings rate above pre-pandemic levels.
  - Increases in old age employment and a slight decline in self-employment also supportive.
  - However, baseline is not sufficient to reach the paper’s objective of a household savings rate of about 7 percent in the medium-term.
- Active (stronger policy) scenario:
  - Simulating convergence of key policy variables toward the sample average could achieve the ~7 percent household savings rate.
  - Specific policy targets suggested:
    - Cut unemployment to 7 percent.
    - Reduce self-employment to 15 percent (proxying a reduction in informality and shift to higher-quality employment).
  - Assuming households invest 10 percent of their annual savings into government bonds could help roll over some external long-term loans, though other sectors would also need to increase their holdings.

### Quantitative objectives and gaps
- To ensure external and debt sustainability while closing the investment gap opened during the long recession, Greece needs additional savings of 9 percent of GDP compared to its 2020 level.
- Section A implication (contextual): overall savings rate would need to rise to about 17 percent of GDP; accounting for recovery in fiscal balance and corporations’ savings at 5-year averages implies a household savings gap of 7 percent of GDP that households would need to fill.

### Conclusion and policy implications
- Greece must boost savings to maintain external and debt sustainability and close the investment gap.
- The household sector is the main laggard relative to eurozone peers; non-financial corporates lag to a lesser extent.
- The COVID-19 driven surge in household savings is at least partially temporary as pandemic conditions normalize.
- Sustainable increase in household savings requires tackling structural constraints, notably:
  - High unemployment,
  - Widespread informality (as proxied by high self-employment rates),
  - Capital account restrictions that affect saving behavior.

*Source: IMF staff estimates.*

### 4.      The robust checks confirm the sign and statistical significance, and in many cases, the

### 1grcea2022002 - 4.      The robust checks confirm the sign and statistical significance, and in many cases, the

### Robust checks of baseline specification
- Adding a COVID-19 dummy:
  - Produces a significant positive coefficient: 7.7400*** (p-value (0.0000)).
  - Reduces the absolute value of the estimated coefficient on fiscal balance.
  - Switches the sign of the estimated coefficient on GDP volatility in some specifications.
- Replacing the COVID-19 dummy with an interaction COVID dummy * stringency index:
  - Coefficient on interaction: 0.1120*** (p-value (0.0000)).
  - To compare with COVID dummy (7.74), scale 0.1120 by mean of the stringency index (62.3) → 7.03.
  - Conclusion: the stringency index largely captures savings behavior in response to the COVID-19 shock.
- World uncertainty index:
  - Leads to an insignificant coefficient: 0.5354 (p-value (0.3806)).
  - Interpretation: the world uncertainty index captures policy uncertainty rather than income uncertainty embodied in GDP volatility.
- Consumer indicators:
  - Consumer confidence index coefficient: 0.1453*** (p-value (0.0000)).
  - Consumer financial condition, next 12 months coefficient: 0.1321*** (p-value (0.0000)).
  - These replace the lead of gross disposable income and influence expected employment and real interest rate coefficients, but do not materially change fiscal balance and output volatility coefficients.
- Household wealth proxy:
  - Change of nominal house prices scaled by household disposable income coefficient: -4.2500*** (p-value (0.0083)).
  - Interpretation: an increase in household wealth reduces households’ precautionary savings.

### Key regression coefficients from Table 2 (selected, exact values preserved)
- Expected employment:
  - Range of estimated coefficients across columns: -12.6802*** to -2.6410** (examples include -4.6599***, -2.6410**, -4.4757***, -4.7903***, -4.7334***, -5.7622***, -12.6802***, -10.2289***, -5.7832***, -4.6971***). Associated p-values include (0.0004), (0.0339), (0.0007), (0.0000), (0.0000), (0.0000), (0.0000), (0.0000), (0.0002), (0.0003).
- Gross disposable income, q/q, lead:
  - Examples: -0.0800 (p-value (0.1481)), -0.0775 (0.1564), -0.0789 (0.1529), -0.2380*** (0.0000), -0.1738*** (0.0004), -0.0741 (0.1890), -0.0341 (0.5551).
- Real short-term deposit rate:
  - Examples: -0.4148*** (0.0000), -0.2563*** (0.0015), -0.2526*** (0.0015), -0.4015*** (0.0000), -0.2597*** (0.0022), -0.2935*** (0.0005), -0.4270*** (0.0000), -0.4408*** (0.0000).
- GDP volatility:
  - Examples: 2.2137*** (0.0000), 2.3025*** (0.0000), 2.3691*** (0.0000), -1.5770*** (0.0002), -1.1699*** (0.0028), 2.1008*** (0.0000), 1.9769*** (0.0000), 2.2389*** (0.0000), 2.2312*** (0.0000).
- Fiscal balance (% of GDP), 4-quarter rolling average:
  - Examples: -0.1569*** (0.0000), -0.1501*** (0.0000), -0.1660*** (0.0000), -0.0662** (0.0331), -0.0811*** (0.0078), -0.1818*** (0.0000), -0.2340*** (0.0000), -0.2370*** (0.0000), -0.1608*** (0.0000), -0.1463*** (0.0000).
- COVID dummy:
  - 7.7400*** (p-value (0.0000)).
- COVID dummy * stringency index:
  - 0.1120*** (p-value (0.0000)).
- World uncertainty index:
  - 0.5354 (p-value (0.3806)) — insignificant.
- Nominal house price, % of HH disposable income:
  - -4.2500*** (p-value (0.0083)).
- Constant terms across specifications (examples):
  - 15.6910***, 13.4997***, 15.0802***, 16.4418***, 16.1986***, 17.0210***, 25.2366***, 21.5052***, 16.8725***, 15.7334*** (all p-values (0.0000)).

- Data and sample details from Table 2:
  - Observations vary by specification: 884, 916, 884, 884, 884, 884, 895, 895, 895, 884.
  - Country fixed effect: Yes.
  - Number of countries: 11.
  - Note: * p<0.10 ** p<0.05 ***p<0.01. P-values in parentheses. All series are seasonally adjusted.

### Household Budget Survey 2019 — updated household income and savings profile
- Savings calculation:
  - Savings = monetary net income (“HH095”) − total expenditure on purchased goods and services (“HExxA” with xx from 01 to 12).
- Findings from Annex I Figure 1 and main text:
  - Older and richer households save while the rest dissave (consistent with Figure 5 based on 2015 data).
  - Pensioners, two-individual households, and households in less populated regions tend to save despite relatively moderate income levels, due to lower expenses.
  - No significant differences in savings behavior by gender of reference person.
- Figures present savings, income, and expenditure in thousands of euro by:
  - Gender of reference person (Male, Female).
  - Age groups (Under 25; Age 25-65; Older than 65).
  - Household income quintiles (1st to 5th).
  - Source of income (Wages or salary; Self-employment income; Property income; Pensions, retirement benefits; Unemployment benefits; Other).
  - Household size and residential area density.

### SILC 2020 — household-level savings behavior during the pandemic
- Survey question HC050: in a typical month, household saves / draws down savings or borrows / neither.
- Household grouping variables:
  - Gender (“RB090”) of household heads (proxied by most senior member).
  - Age of most senior household member (2020 − year of birth indicator “RB080”).
  - Major household activities: basic activity status (“RB210”) taken by at least half of household members; tie-break order: “at work” supersedes “retired,” which supersedes “unemployed.”
  - Household disposable income (“HY020”) divided into quintiles.
- Key finding:
  - A significant share of households increased savings in 2020 across almost all groups, except:
    - Young households (under age 25).
    - Households with majority members being unemployed.
    - Households in the bottom two income quintiles.
- Figures (percent of total households in each category) display Saved / Drew down savings / Borrowed / Other by:
  - Gender; Age; Major household activity; Income quintile.

### Annex II — Medium-term and long-term determinants of household savings: data and definitions
- Country set:
  - Austria, Belgium, Czech Republic, Germany, Spain, Finland, France, Greece, Hungary, Ireland, Italy, Netherlands, Poland, Portugal, Slovenia, and Sweden.
- Data:
  - 48 indicators in annual frequency from 1995 up to 2021 for all 27 EU countries gathered; sample restricted to construct strongly balanced dataset.
  - Panel unit root tests (Levin, Lin, and Chu Test) find savings rate and most explanatory variables are stationary (table A3).
- Savings rate definition:
  - Households sector includes all households and household firms plus NPISH.
  - Gross savings concept follows ESA 2010: portion of gross national disposable income not used for final consumption expenditure; gross disposable income adjusted for change in net equity of households in pension funds.
- Overview of key determinants (Table 1; variable definitions and sources include Eurostat, IMF WEO, World Bank, IMF staff estimates). Key variables listed include:
  - Real adjusted disposable income per capita growth rate (Eurostat).
  - Crisis dummy (1 for 2008 and onwards) (IMF staff).
  - Real short-term interest rate: Short-term deposit rate less HICP inflation (IMF, WEO/Eurostat).
  - Employment expectation indicator (European Commission).
  - GDP volatility (IMF staff).
  - 65+ employment (Eurostat).
  - General government balance (IMF, WEO).
  - Pension expenditure (Eurostat).
  - Other social protection expenditure, domestic private sector credit, Gini coefficient, unemployment rate, share of self-employment, size of shadow economy (IMF staff), REER, capital control indices.

### Panel unit root test results (Levin, Lin, and Chu Test) — selected exact values
- Saving rate (%GDP), Level, intercept: Adj. t-statistics -3.25, Prob. 0.00.
- Income per capita (%, yoy), Level, intercept: -8.13, 0.00.
- Crisis years (dummy), Level, intercept: -3.84, 0.00.
- Real short-term interest rate (%), Level, intercept: -1.63, 0.00.
- Inflation (%), Level, intercept: -5.16, 0.00.
- Expected Employment, Level, intercept & trend: -2.39, 0.00.
- GDP Volatility, Level, intercept: -5.66, 0.00.
- 65+ employment (% 15+ employment), Level, intercept: -3.84, 0.00.
- Capital control (index), Level, intercept & trend: 3.30E+15, Prob. 1.00 (indicating non-stationary series for capital control indices in this test).

### Robustness checks (Annex text and Table 3) — findings and exact coefficients (selected)
- Adding growth of real disposable income (regression 2) or life expectancy (regression 3):
  - Does not significantly alter r-squared and relationships are not statistically significant.
- Including NPL ratio (regression 3):
  - Raises explanatory power; NPL ratio coefficient example: 0.108*** and 0.0761** in specific models.
  - Interpretation: higher NPL ratio raises savings (precautionary motive).
  - NPL omitted from preferred regression due to data availability constraints; including it does not change fit, projections, and overall conclusions.
- Replacing self-employment with size of shadow economy:
  - Direction remains the same, coefficient much smaller, r-squared lower.
  - Interpretation: self-employment captures more than shadow economy size—likely lower income of self-employed limiting savings.
- Interaction terms with a Greece Dummy:
  - None of the interaction terms are statistically significant.
- Selected Table 3 coefficients (exact preserved examples):
  - Crisis dummy (2008 onwards=1): -0.898***, -0.960***, -0.970***, -1.194***, -0.610, -0.205 (across models).
  - Expected Employment: -0.0551***, -0.0526***, -0.0603***, -0.0629***, -0.0499***, -0.0379**.
  - 65+ employment (% 15+ employment): 0.486***, 0.484***, 0.489***, 0.438***, 0.344***, 0.450***.
  - General Government Balance to GDP (%): -0.201***, -0.200***, -0.171***, -0.209***, -0.161***, -0.109***.
  - Pension spending (% GDP): 0.845***, 0.842***, 0.876***, 1.225***, 2.312***, 0.747***; Pension spending (squared): -0.0321***, -0.0322***, -0.0347***, -0.0521***, -0.101***, -0.0298***.
  - NPL ratio (%): 0.108***, 0.0761** (in models where included).
  - Shadow economy size: -0.108*** (where included).
  - Self-employment (% 15+ employment): -0.312*** (repeated across models).
  - REER (2010=100): 0.0785**, 0.0776**, 0.0717**, 0.104***, 0.0294, 0.106***.
  - Overall inflow restriction index: -12.34***, -12.23***, -12.02***, -13.24***, -12.04***, -12.04***.
  - Overall outflow restriction index: 1.988***, 1.966***, 2.104***, 2.192***, 0.814, 1.550**.
  - Overall R-sq across models: 0.756, 0.756, 0.757, 0.780, 0.707, 0.753.
  - Between R-sq: 0.882, 0.883, 0.888, 0.896, 0.820, 0.884.
  - Within R-sq: 0.387, 0.384, 0.380, 0.394, 0.397, 0.318.
- Note: * p<0.10 ** p<0.05 ***p<0.01; model 6 uses first lag for all explanatory variables except for the crisis dummy.

### Financial risks and NPL legacy in Greece (background)
- Since 2016, major banks removed €85 billion non-performing loans (NPLs) from banks’ balance sheets, mainly using state-guaranteed securitization solutions.
- As of end-2018, NPL ratio for all Greek banks: 41.8 percent.
- By Q4:2021, NPL ratio remains well above the European Union (EU) average (7.0 percent against 2.0 percent).
- Credit servicers now manage more than €120 billion in exposures — supervision, market transparency, and financial disclosure on credit servicers and NPL workout are recommended to avoid residual credit risk on seller banks.

*Source: IMF staff estimates and the referenced tables and annexes in the provided content unit.*

### 2.      Public intervention was needed to address the systemic NPL issue. Given challenges to

### 2.      Public intervention was needed to address the systemic NPL issue. Given challenges to

### Background and public interventions
- Given challenges to resolve NPLs organically (restructuring, forbearance) and weak investor sentiment, the Greek government established the Hellenic Asset Protection Scheme (HAPS, or so-called Hercules) in 2019 (Law 4649/2019).
- Supervisory authorities (the SSM and the BoG) aided Hercules by assigning a zero-risk weight on the senior tranches of NPL securitizations.
- Corporate hive-downs allowed banks to dispose under-provisioned NPLs without subsequent losses triggering conversion of Deferred Tax Credits (DTCs).
- Government legal reforms and operational measures included:
  - Significant overhaul of the insolvency law in 2020.
  - Amendments to Hellenic Financial Stability Fund law.
  - Operationalizing frameworks to ease loan resolution, building on new digitalization platforms and processes.
  - Subsidizing loan repayments to distressed debtors (GEFYRA).
- The functioning of the early warning platform completed the electronic infrastructure of the new insolvency framework by end-2021.

### Key government schemes (Box 1)
- Hellenic Asset Protection Scheme (HAPS, Hercules I & II):
  - Securitization framework through SPVs issuing junior, mezzanine and senior securities tranches sold to investors.
  - Senior tranches are held by banks and guaranteed by the Greek government; guarantees priced to comply with EU state aid rules.
  - NPLs are securitized at market value triggering additional loan-loss provisioning if needed.
  - Hercules II is the extension of this scheme, ending by October 2022.
- GEFYRA, GEFYRA II, and "step-ups" (subsidized loans):
  - Temporary instalment subsidies granted to distressed debtors after loan moratoria were lifted throughout 2021.
  - GEFYRA (started in 2020) subsidizes mortgage repayments on primary residences; replaced the Katseli Law.
  - GEFYRA II (started in 2021) offers similar subsidies for corporate borrowers.
  - GEFYRA programs, largely unused since end 2021, will terminate in H1:2022.

### Outcomes and scale of NPL transfers
- From March 2016 to Q2:2021, €78 billion of NPLs have been removed from Greek banks’ balance sheets.
- According to the BoG, by end-2021 credit servicers were managing €123 billion of loans:
  - €85 billion from NPL purchasers.
  - €38 billion from Greek banks, including €12 billion off-balance sheet claims.
- 84 percent of loans managed by credit servicers were NPLs of relatively poor quality, mainly denounced loans.
- NPL sales have often been to external purchasers (international investors specialized in distressed debt restructuring); credit servicers became systemically important managers of distressed debt.

### Credit servicers: structure, roles, and risks
- Role and relationships:
  - Credit servicers act on behalf of NPL purchasers independently from originating banks; a servicing agreement is required for Hercules sales.
  - Credit servicers are Greek regulated companies; banks have provided operational support (staffing, customer data sharing).
  - Three major banks maintained a 20 percent minority shareholder stake in each credit servicer they partnered with.
- Risk allocation:
  - Credit servicers do not assume credit risk on their balance sheets; they earn servicing fees per long-term servicing agreements validated by authorities and monitored by the BoG.
  - Under Hercules, the Greek government guarantees senior tranche notes held by banks, which must be rated by at least one credit rating agency not lower than BB-; activation of the state guarantee would create residual risk for the government and ultimately for banks if maximum amounts were reached.
- Market concentration and scale:
  - By end-2021, 24 credit servicers were licensed by the BoG to operate in Greece, of which 16 were active.
  - The four major players (Cepal, DoValue, Intrum and QQuant) managed 88 percent market share of total loans assigned to credit servicers; most of their capital is foreign-owned.
- Business model constraints:
  - Credit servicers must meet conflicting stakeholder expectations: implement approved business plans, maximize recovery rates, avoid additional losses/re-defaults, optimize recovery delays, process with limited staff/budgets, leverage digitalization and big data, and comply with laws and customer protection.

### Legal and regulatory framework
- EU and Greek laws:
  - EU Directive 2021/2167 (COD) dated November 24, 2021, on credit servicers and credit purchasers requires review of Greek legislation; transposition impact not expected to be substantial given Greece’s existing framework.
  - Credit servicers are regulated by the BoG under Greek Law 4354/2015 and Executive Committee Act (ECA) 118/19.5.2017 ("Framework of establishment and operation of credit servicing firms").
- Licensing and permitted activities:
  - Law 4354/2015 and ECA 118/2017 cover authorization, submission of servicing agreements, assessment of qualifying holdings and fit & proper requirements, internal systems and controls, reporting, enforcement, and a Code of Conduct governing relations with borrowers.
  - Credit servicers may request license extensions to engage in debt refinancing, but to date no servicer has applied.
- Prudential supervision:
  - Credit servicers must comply with minimum capital requirements, but no risk-based quantitative prudential ratios are in place for solvency, liquidity, or concentration risks; supervision focuses on qualitative rules.
  - The BoG’s supervisory strategy includes monitoring reporting (portfolio evolution, collateral coverage, auctions), financial performance, business plans, servicing agreements, code of conduct compliance, customer complaints, and fit & proper assessments.
  - On-site inspections were delayed to 2022 because of the pandemic.

### Operational and legal constraints to NPL workout (including Box 2 points)
- Available workout mechanisms:
  - Initiation of proceedings for adjudication or payment orders (unsecured claims).
  - Foreclosure and auction of assets (secured claims).
  - Participation in judicial or extra-judicial insolvency procedures including bankruptcy, rehabilitation, and Out-of-Court Workout of the new Insolvency Code.
- Practical impediments and recommendations:
  - NPL workout processing by credit servicers proceeded more slowly than expected as of end-2021 due to legal obstacles.
  - Authorities could help by: (i) collecting and analyzing data on system usage and restructuring solutions reached; and (ii) increasing capacity and efficiency of the court system (digitalization, remote hearings, eliminating backlogs), especially for household insolvency cases.
- Specific legal issues highlighted:
  - Institutional capacity: court processes are slow with a backlog (exacerbated by COVID); around 48,000 unresolved cases as of Q1:2022; court modernization is in train but progress expected to be gradual.
  - Sale and Lease Back (SLBO): provisions protect vulnerable debtors’ primary residence, but delays in operationalization (now expected in mid-2023 at the earliest) prevent resolution for these debtors; linking subsidy schemes with sustainable restructuring and accelerating SLBO operationalization are suggested.
  - Enforcement procedures (E-Auctions): resumed as of September 2021 after suspension; numbers of successful auctions have been relatively low with most real estate acquired by the bank leading the auction, though signs of gradual improvement since Q1:2022.
  - Data collection and monitoring: authorities have access to significant data via electronic platforms (e.g., OCW, “second chance” procedure) and should develop a comprehensive framework to evaluate procedures, bottlenecks, and depth of restructurings; authorities should actively monitor credit servicers’ performance and debtors’ capacity to repay.
- Information gaps:
  - No public data provided on NPL recovery performance by credit servicers.

### Supervisory and policy implications
- Supervision:
  - Credit servicers require close supervision to assess whether financial risks from distressed debt management could affect financial stability.
  - A clear entity-based and activity-based mapping of the credit servicer landscape is needed to apply the regulatory and prudential framework effectively.
  - Monitoring should include effectiveness and efficiency of collection and restructuring, residual credit risk, and implications for bank lending.
- Regulatory adjustments and support:
  - Ensure effective implementation of the revised legal framework for debt resolution.
  - Enhance court capacity and accelerate digitalization and remote hearings to reduce backlog and improve enforcement credibility.
  - Develop comprehensive data collection and analysis frameworks to monitor outcomes, bottlenecks, and the depth of restructuring solutions.
  - Monitor potential residual fiscal risk from state guarantees under Hercules and exposure stemming from activation of guarantees.

*Source: INTERNATIONAL MONETARY FUND (content unit 1grcea2022002).*

### 16.      From a macroprudential perspective, financial stability risks from credit servicers

### 16.      From a macroprudential perspective, financial stability risks from credit servicers

### Assessment of financial stability risk
- Financial stability risks from credit servicers appear modest.
- Financial stability has been enhanced from the banking sector’s perspective with the reduction of NPL-related credit risk.
- Credit servicers may be less impactful on financial stability because these non-bank financial intermediaries are less interconnected with the financial system and payment system infrastructure.
- Unexpected additional losses on securitized NPL portfolios would have a fiscal impact on the government through the state guarantees on senior notes held by banks; only losses beyond the maximum amount of state guarantees might be impactful on banks.

### Micro-prudential risks for credit servicers
- Credit servicers are non-bank financial institutions that collect no customer deposits and are exposed to micro-prudential risks.
- Based on loan servicing agreements, credit servicers are supposed to be immunized against direct credit risk, which is ultimately borne by bearers of senior notes.
- Credit risk has not disappeared due to possible further deterioration of debtors’ financial situation, depreciation of real estate collateral, and/or ineffectiveness of other available guarantees; this credit risk is not borne by credit servicers.
- Credit servicers are exposed to other risks, including operational risk, fraud, cyber, legal and compliance risks, potentially stemming from poor data quality on NPLs, adverse events undermining workout execution, customer complaints, and similar issues.
- NPL servicing is a higher risk activity than regular loan servicing.
- No public rating of credit servicers is available.
- If a credit servicer faced severe financial difficulty or default, the Bank of Greece (BoG) could make use of existing legal powers at national level.

### Regulatory and supervisory context
- A regulatory framework applicable to credit servicers has been implemented in Greece since 2015, to be aligned with the EU Directive 2021/2167 (COD) of November 24, 2021, on credit servicers and credit purchasers.
- The BoG is monitoring credit servicers through regular prudential supervision, including licensing, watching their business plans and effective NPL recovery performance.
- The BoG is planning to resume onsite inspections of credit servicers in 2022 to supplement remote surveillance based on regulatory reporting.

### Identified shortcomings affecting NPL workouts and financial stability
- Market transparency is low, with few available and reliable public information and data on the Greek NPL market, NPL investors, and credit servicers.
- Financial disclosure on NPL recovery performance and risks is insufficient, making it difficult for outsiders to monitor effectiveness and efficiency of NPL workouts.
- Supervision of credit servicers could be stepped up, with more thorough on-site inspections to be performed once the pandemic recedes.
- Recovery performance of distressed debt by credit servicers has been weak so far, raising uncertainty about their capacity to comply with business plans in the long run, adding to overall uncertainty that has steadily increased since 2022.

### Policy recommendations to support efficient NPL workout and financial stability
- Ensure full implementation of legal reforms aimed at easing NPL workouts and increase the pace and effectiveness of amicable and judicial NPL recovery.
  - Focus efforts on: (i) data collection and analysis regarding the use of the system and the restructuring solutions reached to identify possible bottlenecks and impediments; and (ii) increasing capacity and efficiency of the court system, including by accelerating, where possible, digitalization and remote hearings, and by eliminating case backlogs, particularly of household insolvency cases.
  - Aim to reduce moral hazard and ensure debtors have incentives to agree to restructuring solutions offered.
  - Data collection and monitoring should focus on analysis of remaining NPL borrowers and solutions offered by credit servicers to ensure realistic, sustainable solutions based on reasonable capacity-to-pay indicators.
- Improve market transparency, enrich financial disclosure, and systematize data collection and analysis relating to credit servicers and NPL purchasers.
  - Upgrade financial disclosure standards applicable to credit servicers.
  - Publish periodic surveys including risk-based indicators by the Bank of Greece together with relevant legal and analytical information.
  - Disclose relevant statistics on distressed debts due by the private sector, whether still bank NPLs or transferred to credit servicers through securitization or outright sales, to enable global monitoring of the resolution of NPL legacy, including post-pandemic NPLs.
  - Consider rating credit servicers as an incentive for transparency and performance.
- Fine-tune supervisory expectations and implement thorough onsite inspections.
  - Monitor a risk-based set of early warning indicators to ensure diligent identification of, and reaction to, material risks that might undermine credit servicers’ business plans and lead to additional loan losses requiring use of state guarantees or impacting banks.
  - Use the implementation of the EU Directive in Greece as an opportunity to review and fine-tune the regulatory and prudential framework applicable to credit servicers.
  - Consider publishing prudential guidelines to credit servicers to specify risk-based supervisory expectations relating to appropriate management of NPL legacy.
- Encourage credit servicers to apply for extending their license to provide refinancing to support restructurings of viable debtors.
  - Recognize that credit servicers would become more high-risk profile regulated institutions and that their exposure to financial risks would need close monitoring and more thorough and adapted prudential requirements on solvency, liquidity, risk management, and governance.
- Coordinate authorities’ action plans to reduce the distressed debt overhang in Greece.
  - A coordinated response is required beyond financial stability considerations, given broader economic and social implications.
  - While debt is gradually being offloaded from the banks, it remains in the real economy with attendant economic effects; a strategy to deal with the distressed debt as opposed to non-performing loans should be put in place.
  - Debt should be restructured, or written off, depending on the financial condition of the debtor, using legal tools available so that viable debtors may get access to bank credit again after their debt repayment obligations are cleared in a reasonable timeframe.

*Source: 1grcea2022002 - 16.      From a macroprudential perspective, financial stability risks from credit servicers*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2022/english/1grcea2022002.pdf_