## 1. Cost of Living and Fiscal Support, 2022–23

## Source details

**Canonical URL:** [1. Cost of Living and Fiscal Support, 2022–23](https://www.imf.org/-/media/files/publications/cr/2024/english/1grcea2024002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1grcea2024002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1grcea2024002.pdf.json)

---

### Motivation
- Average inflation shot up to almost 10 percent in 2022 from a marginally positive rate in 2021, with two thirds of the price increase accounted for by food, beverage, and energy.
- The government provided significant subsidies on electricity differentiated by usage, along with other budgetary measures to alleviate living and operating costs facing households and enterprises.
- Most of the measures are expected to be withdrawn by end-2023, except for some electricity subsidies for small users.
- The analysis uses household consumption data from the annual Household Budget Survey (HBS) (consumption patterns proxied by the 2021 HBS) to construct a household-level indicator of effective inflation (the loss in purchasing power).

### Consumption patterns and heterogeneity
- Lower-income households allocate a larger share of total consumption to basic goods:
  - Households in the bottom income quintile spend over 20 percent of their total expenditure on housing, water, electricity, gas, and other fuels — more than double the corresponding share for households in the top income quintile.
- Pandemic-related behavior:
  - 2020 saw substitution of services (travel, hotel, restaurants) for goods;
  - 2021 showed reversion to higher services share;
  - 2022 could interrupt that trend due to food and energy price jumps.
- Effective inflation differs across goods: food, utilities, and certain services played a major role in driving heterogeneity.

### Distributional impact of inflation (effective inflation)
- Definition and method:
  - Effective inflation approximates the compensating variation as the weighted sum of price changes for goods and services each household consumes, using expenditure shares from the old consumption bundle (proxied by 2021 HBS).
  - Constructed using detailed goods and services to the extent their prices are separately observed in the HICP and expenditure shares are calculable in HBS.
- Key findings:
  - On average, the effective inflation for households in the top three income deciles is 1½ percentage points below that for households in the bottom three income deciles.
  - Poorer households are hit harder mainly due to compositional factors (larger shares of expenditure on food and utilities).
  - Richer households in many cases experienced higher increases in utilities and transport (influenced by international energy prices); exceptions include food and non-alcoholic beverages, and health, where poorer households faced marginally higher increases in 2022.
- Correlations with household characteristics:
  - Large households on average spend a smaller share on utilities → less sharp increase in effective inflation.
  - Households with major bread-earners above age 65 or less educated spend more on utilities and food → more sizable effective inflation increases.
  - Households relying on secondary income (pensions, unemployment benefits, other non-wage, non-property income) face higher effective inflation.
  - Households living in sparsely populated areas face higher effective inflation due to higher food and some utilities shares.
  - No statistically significant difference by gender of major bread-earner in aggregate effective inflation; however, female-headed households tend to spend more on food, utilities, and health, while male-headed households spend more on alcoholic beverages and tobacco, transport, and restaurants and hotels.
- Within-group heterogeneity:
  - Among groups with higher average effective inflation, many households still experienced below-average effective inflation:
    - Close to 50 percent for households living in sparsely populated areas or with highest-paid member having education below tertiary.
    - Around 40 percent for households relying on secondary income sources or with highest-paid member above age 65.
  - Around 30 percent of households in each of these groups have above-average income, indicating that categorical targeting by group without income verification can include relatively well-off households.

### Methodology: compensating variation and data mapping
- Compensating variation (CV) framework:
  - CV0,1i = Σj s0i,j ⋅ π0,1j
  - s0i,j = c0i,j p0j / Σk c0i,k p0k
  - π0,1j = 100 ⋅ (p1j / p0j − 1)
  - CV measures minimal change in household expenditure—as a share of total expenditure under old prices—needed to make the old consumption bundle affordable under new prices; it is a measure of the effective inflation facing the household.
- Data sources and mapping:
  - Household consumption shares from 2021 Household Budget Survey (HBS)
  - Price changes (average 2022 over 2021) from HICP (Harmonized Index of Consumer Prices)
  - Mapping adjustments include excluding imputed rentals (042) and extrapolating prices for HBS items not covered in HICP using weighted averages of related HICP categories.
- Granularity and measurement:
  - Fine-grained HBS-HICP mapping (up to 355 major and minor categories) produces fatter tails in the distribution of household-specific inflation compared with using only 12 major categories.
  - Reconstructing HICP from 355 categories yields average inflation of 9.5 percent versus the official 9.3 percent.
  - Calculated compensating variation averages 10.3 percent in 2022, exceeding average HICP inflation (9.3 percent), which corresponds to the 32th percentile of the calculated compensating variation.
  - Rounding errors are pronounced for utilities (04), food (01), and transport (07).

### Empirical correlates and regression robustness
- Baseline regression coefficients (selected, significance noted):
  - Living in sparsely populated areas: baseline coefficient 0.5414*** (positive)
  - Above age 65: baseline coefficient 1.2321*** (positive)
  - Below tertiary education: baseline coefficient 0.4767*** (positive)
  - Main source of income is secondary: baseline coefficient 0.7035*** (positive)
  - Household monetary income (EUR 000): baseline coefficient -0.0150*** (negative)
  - Household equivalent size (OECD): baseline coefficient -0.1750** (negative)
- Robustness:
  - Coefficients remain robust to random sampling assumption (OLS), alternative income measures (total income), OECD modified scale for household size, different definitions of reference person, breakdown of income sources, and inclusion of additional household characteristics.
- Regression sample and fit:
  - Observations: 6,047
  - R-squared (baseline): 0.2039
  - Standard errors reported; significance notation: *** p<0.01, ** p<0.05, * p<0.1

### Policy implications from distributional analysis
- Targeted transfers are the most effective instrument to protect vulnerable households under a fixed fiscal envelope because they minimize benefit leakage to higher-income groups.
- Strengthen quasi-targeting features where feasible (e.g., energy subsidy tapering for large users, income criteria in food programs like “Market pass”).
- Build administrative capacity to implement targeted programs:
  - Create a centralized beneficiary registry
  - Improve income reporting and third-party verification
  - Close coverage and benefit shortfalls in existing programs such as the Guaranteed Minimum Income

---

### Policy simulations: design, benchmark, and comparative outcomes

### Simulation design and assumptions
- Simulations are hypothetical and not able to replicate actual government interventions during 2022–23 due to data limitations.
- Subsidies in simulations are defined as additional cash each recipient household receives proportional to its consumption of the subsidized products.
- Categorical transfers are cash payments to households belonging to predefined categories; if income is not observable, all eligible households receive equal amounts.
- For comparability, subsidies and categorical transfers are calibrated so the overall fiscal envelope equals the fiscal cost of the benchmark targeted transfer scenario.
- Energy subsidies in the simulation are implemented against HBS category 045 (electricity, gas, and other fuels); food subsidies are implemented against HBS category 011 (food).

### Benchmark "targeted transfer" scenario
- Design:
  - All households in the bottom three income deciles that have experienced above-average effective inflation are given cash so that their loss of purchasing power is reduced to the rate of average (pre-intervention) effective inflation.
- Outcomes and calibration:
  - After intervention, the mean effective inflation within the bottom three deciles will be brought down to around 1½–2 percentage points below its original levels, or 1½ percentage points below the original sample average, comparable to the level of the top decile.
  - The intervention does not affect the top seven deciles.
  - The targeted transfer scenario has a fiscal cost of 0.1 percent of GDP.
  - The policy objective is modest: affected households continue to face a loss of purchasing power equal to the sample average (around 10 percent).

### Comparative point and caveats
- Targeted support tailored to recipients’ needs (income- and loss-targeted) would more effectively mitigate vulnerable households’ income loss than categorical programs without income criteria, although categorical programs can still alleviate cost-of-living pressures to some extent.
- Simulations are illustrative; sizes and specific designs are not policy prescriptions. Policymakers should calibrate interventions considering fiscal space and socioeconomic conditions.

---

### Distributional findings: subsidies vs categorical transfers vs targeted transfers

- Subsidies are regressive: benefits are proportional to consumption of subsidized goods/services, and higher-income households consume more.
- Share of subsidy benefits received by higher-income households:
  - Energy subsidies: 75 percent
  - Food subsidies: 80 percent
- Effect on bottom three deciles facing above-average effective inflation:
  - Pre-intervention share: 51.5 percent
  - After targeted transfer: drops to zero (by design)
  - After simulated categorical transfers: remains elevated (above 40 percent)
- Categorical transfers perform better than subsidies but still leak benefits to higher-income groups:
  - Around 60 percent of categorical-transfer benefits would be received by households in the top seven deciles
- Conclusion: targeted transfers are most effective at protecting vulnerable households within a fixed fiscal envelope because they reduce leakage of benefits to higher-income households.

### Practical considerations and implementation caveats
- Income misreporting risk: if household income is at high risk of being mis-reported, income-targeted support becomes less effective.
- Real-world programs contain quasi-targeting elements that improve progressive impact relative to the simple simulations:
  - Electricity subsidies often reduce unit-subsidy for large users (presumed richer households), producing a more progressive distributional impact than the simulated energy subsidy scenario.
  - The food program (“Market pass”) incorporates some income criteria and would be more effective than the simulated food subsidy.
- Implementation capacity gaps to be addressed:
  - Establish a centralized registry of beneficiaries
  - Improve reporting and verification of beneficiaries’ income leveraging available third-party information
  - Address gaps in coverage and benefit levels of existing targeted programs such as the Guaranteed Minimum Income

---

### Select systemic risk assessment and BBMs for Greece

### Real estate prices and household sector
- Residential real estate prices have increased significantly since their trough in 2017, exceeding 50 percent in nominal terms and 35 percent in real terms.
- Price-to-income and price-to-rent ratios exceeded their historical long-term averages by 6 and 29 percent, respectively; an econometric model suggests 8 percent.
- Demand drivers: strong employment and real disposable income growth, increased nonresident investment (including Golden Visa program), and structural features (low rooms per person, high share living in overcrowded homes).
- Supply response: housing investment as a share of GDP and building permits doubled since 2016 (from a low base).
- Household sector:
  - Deleveraging since the sovereign debt crisis left leverage much lower than other EA countries, but household debt service remained one of the highest in the EU (housing costs in disposable income and debt service ratio).
  - Greece records the highest share of loan-to-value (LTV) ratio exceeding 80 percent among EU countries (EBA, 2022).
  - Based on BoG data, the share of new loan disbursements with an LTV>80% stood at 5.7% in Q2 2023.
  - Bank exposure to real estate is below the EU average and less significant compared to pre-GFC levels.
  - Joint distribution of risk indicators does not suggest reasons for significant concern (BoG, 2023b).

### Macroprudential policy options and legal framework
- Legal framework amended in 2023 granting BoG power to enact Borrower-Based Measures (BBMs) for financial institutions established and operating in Greece and branches of foreign institutions in Greece, in connection with credit secured by real estate located in Greece.
- BoG may adopt BBMs covering caps on ratios related to the credit or the borrower or specific features of the credit; it may set types of BBMs, types of credit, cap percentages, terms and conditions, and reporting requirements.
- Modalities and trade-offs:
  - Income-based measures (primary instrument: caps on the DSTI ratio) aim to reduce probability of default (PD).
  - Collateral-based measures (primary instrument: cap on LTV) aim to reduce loss given default (LGD).
  - Income- and collateral-based measures are complementary and recommended to be considered jointly (ESRB, 2019).
  - Other complementary measures: maturity limits, loan amortization, risk weights, targeted CBMs.

### Timing, calibration, and distributional considerations
- Timing should consider the real estate market cycle and policy lags; early-stage activation addresses vulnerabilities while still building up.
- BBMs affect only the flow of new lending; CBMs affect lender resilience.
- Instruments can be calibrated as binding (taming cyclical risks) or nonbinding (curbing structural vulnerabilities), and can be static or dynamic.
- Speed limits and exceptions can target specific borrower segments; BBMs may restrict credit access for young low-wealth households but exceptions can allow a fraction of new lending to exceed limits.
- EU context:
  - BBMs widely implemented in the EU: 14 countries implemented DSTI and 20 countries LTV.
  - Greece is among the few EU countries that had not implemented BBMs as of the study period.

### Quantitative framework and simulation outputs
- Uses micro-level HFCS household data for Greece and macro-level quarterly data from the ECB and national sources.
- Captures dynamics of household debt service, consumption expenditure, labor income, and unemployment benefits at household-member level.
- Primary outputs: simulated PDs, LGDs, and loss rates (LRs) at household level; linked to bank mortgage portfolios to assess capital impact.
- Model accounts for macro-financial feedbacks from policy-induced credit demand drag.
- First-round effects: direct impact of BBM settings on PDs, LGDs, and LRs.
- Second-round effects: add impact of negative loan demand shock on PDs, LGDs, and LRs.
- Final step: computes impact on banks’ capital adequacy ratio via reduced loan losses and lower risk weights from lower PDs and LGDs.

### Impact on household and bank resilience (key simulation findings)
- PD response to DSTI caps:
  - Initial PD starts to decline sharply from a DSTI cap around 90 percent, with an already sizeable impact for cap at 50 percent, and reaching maximum impact at the left end of the DSTI cap range.
- PD and LGD comparative effects:
  - DSTI caps have stronger potential to reduce PDs.
  - LTV caps have a stronger impact on reducing LGDs, with LGDs estimated to decline sharply from an LTV cap around 90 percent.
- Crowding-out and mortgage volume effects:
  - Mortgage volume reduction suggests a stronger impact of LTV caps compared to loss-rate-equivalent DSTI caps; DSTI caps would have to bind for a smaller share of the population to achieve the same loss rate as a given LTV cap.
- Mapping between LTV and DSTI:
  - For an LTV cap at 85 percent, the equivalent DSTI cap would equal around 40 percent.
- Benchmark calibration:
  - Primary calibration applied: DSTI = 40 percent and LTV = 85 percent (alternative calibration in Appendix II).
  - Joint implementation of DSTI and LTV caps yields stronger increases in borrower resilience than individual caps due to complementarities: DSTI acts primarily via PD channel and LTV via LGD channel.
- Impact on bank capital:
  - BBMs are estimated to have a positive impact on bank CET1 ratios for the mortgage loan portfolio, with stronger effects for DSTI caps compared to LTV caps.
  - Reduced loan losses from mortgage credit portfolios outweigh the benefit from lower risk weights due to lower PDs and LGDs.

### Conclusions and policy implications for BBMs
- Emerging vulnerabilities in real estate and household sectors warrant close monitoring: significant house price increases since 2017, moderate overvaluation signals, high household debt service despite low leverage.
- Enhancing the macroprudential toolkit with BBMs would help increase resilience of households and banks and contribute to systemic risk reduction.
- Legal framework amendment in 2023 enables BoG to activate both income- and collateral-based BBMs.
- Simulations indicate complementarities between DSTI and LTV caps: DSTI reduces PDs more; LTV reduces LGDs more; DSTI has stronger overall impact but comes at cost of lower lending.
- Joint implementation of DSTI and LTV caps produces stronger borrower resilience than individual caps.
- Caps could initially be set at less binding levels and be gradually tightened over time.

---

### Appendix II. Alternative Calibrations; Recent Trends of Informality in Greece

### Alternative calibration (Appendix II)
- DSTI = 45 percent and LTV = 90 percent
- Appendix II. Figure 1. Impact estimated based on the calibration assuming DSTI = 45 and LTV = 90

### Overview: informality trends and implications
- Informality has dropped significantly in Greece in recent years, with temporary increases during the sovereign debt crisis and the COVID-19 pandemic.
- Aggregate informality (weighted by regional gross value added) declined from a peak of about 30 percent of GDP in 2013 to a low of about 16 percent of GDP in 2021.
- Regional variation in informality declines ranges from 5 percent of GDP (Western Macedonia) to 18 percent of GDP (Crete).
- Sectoral and regional implied informality (from 2018 firm survey):
  - By region:
    - Northern Greece: 15%
    - Central Greece: 24%
    - Attica: 20%
    - Aegean Islands: 25%
  - By sector:
    - Food: 27%
    - Fabricated Metal Products: 17%
    - Other Manufacturing: 13%
    - Retail: 29%
    - Other Services: 19%
- Fiscal and growth implications:
  - The decline in informality from about 30 percent of GDP (2013) to about 16 percent of GDP (2021) is associated with a gain in tax revenue by about 4 percentage points of GDP.
  - European Commission estimates: Greece’s Value Added Tax gap decreased by over 16 percentage point of its tax base between 2013 and 2021, equivalent to about 2 percentage points of GDP.
  - Correlation evidence shows that lower informality is associated with higher subsequent per capita GDP growth.

### Methodology and data for informality analysis
- Multiple Indicators Multiple Causes (MIMIC) approach used to estimate a latent informality index and then calibrated to formal GDP statistics.
- Calibration uses World Bank (2018) Enterprise Survey for Greece.
- Subnational, quarterly, 2012–2021 data: satellite nighttime lights (VIIRS Day), Google search volume of subnational region name, labor participation rate (population aged 20-64), GDP (first differences).
- Causes included: agriculture employment share, self-employment share, unemployment rate, lagged income per capita, tourist arrivals per population, education attainment (share aged 25-64 with less than primary/primary/lower secondary).
- National causes: VAT gap, tax as percent of GDP, trade openness, government consumption as percent of GDP, World Bank governance indicators.

### Main findings and policy-relevant implications
- Drivers and mitigation:
  - Severe recessions, fiscal tightening, and weakened payment culture contributed to increases in informality during crisis periods.
  - Government support during the pandemic likely moderated informality increases relative to prior recessions.
  - Digitalization progress likely contributed to reductions in informality and explains part of subnational variation.
- Policy priorities:
  - Accelerate digitalization to further reduce informality.
  - Continue pro-growth structural reforms that improve labor market flexibility and business environment.
  - Strengthen tax administration and measures that narrow the VAT gap to convert informality reductions into sustained tax revenue gains.
  - Use subnational and alternative big-data indicators (satellite nightlight, google search) alongside surveys to monitor informality and tailor region-specific reforms.

---

### Digital infrastructure, usage, and informality: empirical correlations and policy priorities

### Empirical correlations (regional-level)
- Better digitalization indicators show consistent negative correlations with the share of the informal economy at the regional level in Greece.
- Key empirical associations (regional-level correlations):
  - A 10-percentage point increase in internet access is on average associated with a reduction of informality by about 3 percentage point.
  - A 10-percentage point reduction in the share of residents who have never used internet is associated with a reduction in informality by about 4 percentage points.
  - A 10-percentage point increase in the share of residents who used online purchase is associated with a reduction in informality by about 4 percentage points.
  - A 10-percentage point increase in the share of residents who used digital public services online purchase is associated with a reduction in informality by about 3 percentage points.
- These correlations align with prior findings that digital payments reduced informality and tax evasion and OECD (2023) findings that increased digital transactions raised tax compliance in Greece.
- Caveats:
  - Limits of correlation analysis; causality not established.
  - Lack of subnational data on tax collections and non-cash transactions constrains tighter links to regional economic activities.
  - Social structure of Greek regions may strongly affect informality and requires separate research.

### Policy drivers behind recent declines and recommended digital priorities
- Factors contributing to decline in informality:
  - Tightened regulation and enforcement against financial crimes and tax evasion.
  - Increased labor market flexibilities.
  - Significant progress in digitalization: infrastructure improvements and digital public services increased tax morale and integrated the self-employed.
  - Modernized employment information system eased registration and increased monitoring.
  - Rationalization of tax policies, including reducing business income tax rates and marginal personal income tax.
- Recommended digitalization priorities:
  - Further improve digital skills and human capital relevant to industry needs.
  - Enhance integration of digital technology across public services and businesses to improve monitoring of tax compliance, ease business and employment registration, and better integrate self-employed into the formal economy.
  - Leverage the National Resilience and Recovery Plan and Next General EU funding for continued digitalization and skill training.
- Complementary structural reforms:
  - Improve regulatory quality and transparency to ease business entry and support productive competition.
  - Increase efficiency in spending on education and training to improve job matching and incentives for formal employment.
  - Targeted employment support for youth and women with disproportionately elevated unemployment rates.
  - Reforms linking firm and job formalization with access to finance to enhance incentives to formalize.

### MIMIC methodological summary and estimation highlights
- MIMIC links multiple observable indicators of the informal economy to multiple observable causes through a latent variable.
- Structural equation: y*_k = α′ x_k + η_k; measurement equation: y_k = β y*_k + ε_k.
- Two identification restrictions: Π = β α′ has rank one; error covariance Ω = σ^2 β β′ + Θ^2 (rank-one plus diagonal).
- Estimation via maximum-likelihood; data demeaned at country level for subnational analysis; index rescaled and calibrated to map to size of informal economy.
- Structural estimation results (selected coefficients, std.err., z, p-value):
  - agriculture employment share (%) 0.47 0.12 4.02 0.00
  - unemployment rate (%) 0.27 0.08 3.42 0.00
  - (lagged) GDP per capita 0.32 0.03 10.63 0.00
  - population aged 25-64 with secondary education or lower (%) -0.67 .09 -7.90 0.00
  - VAT gap 0.59 0.05 11.74 0.00
  - tax (% of GDP) -2.16 0.32 -6.74 0.00
  - trade openness -0.46 0.05 -9.35 0.00
  - government consumption (% of GDP) 1.14 0.25 4.52 0.00
  - political stability 0.16 0.02 7.74 0.00
- Measurement coefficients (selected):
  - labor participation rate growth Informal -0.14 .01 -9.79 0.00
  - Google Search Volume growth Informal -0.13 0.53 -0.24 0.81
  - Nighttime light growth Informal -0.07 0.10 -0.73 0.47

*Source: INTERNATIONAL MONETARY FUND — excerpts and staff calculations from 1grcea2024002.*

### 1. Cost of Living and Fiscal Support, 2022–23 _______________________________________________ 4

### 1. Cost of Living and Fiscal Support, 2022–23

### Motivation
- Average inflation shot up to almost 10 percent in 2022 from a marginally positive rate in 2021, with two thirds of the price increase accounted for by food, beverage, and energy.
- The government provided significant subsidies on electricity differentiated by usage, along with other budgetary measures to alleviate living and operating costs facing households and enterprises.
- Most of the measures are expected to be withdrawn by end-2023, except for some electricity subsidies for small users.
- The paper uses household consumption data from the annual Household Budget Survey (HBS) (consumption patterns proxied by the 2021 HBS) to construct a household-level indicator of effective inflation (the loss in purchasing power).

### Consumption patterns and heterogeneity
- Lower-income households allocate a larger share of total consumption to basic goods:
  - Households in the bottom income quintile spend over 20 percent of their total expenditure on housing, water, electricity, gas, and other fuels — more than double the corresponding share for households in the top income quintile.
- Pandemic-related behavior: 2020 saw substitution of services (travel, hotel, restaurants) for goods; 2021 showed reversion to higher services share; 2022 could interrupt that trend due to food and energy price jumps.
- Effective inflation differs across goods: food, utilities, and certain services played a major role in driving heterogeneity.

### Distributional impact of inflation (effective inflation)
- Definition and method:
  - Effective inflation approximates the compensating variation as the weighted sum of price changes for goods and services each household consumes, using expenditure shares from the old consumption bundle (proxied by 2021 HBS).
  - Constructed using detailed goods and services to the extent their prices are separately observed in the HICP and expenditure shares are calculable in HBS.
- Key findings:
  - On average, the effective inflation for households in the top three income deciles is 1½ percentage points below that for households in the bottom three income deciles.
  - Poorer households are hit harder mainly due to compositional factors (larger shares of expenditure on food and utilities).
  - Richer households in many cases experienced higher increases in utilities and transport (influenced by international energy prices); exceptions include food and non-alcoholic beverages, and health, where poorer households faced marginally higher increases in 2022.
- Correlations with household characteristics:
  - Large households on average spend a smaller share on utilities → less sharp increase in effective inflation.
  - Households with major bread-earners above age 65 or less educated spend more on utilities and food → more sizable effective inflation increases.
  - Households relying on secondary income (pensions, unemployment benefits, other non-wage, non-property income) face higher effective inflation.
  - Households living in sparsely populated areas face higher effective inflation due to higher food and some utilities shares.
  - No statistically significant difference by gender of major bread-earner in aggregate effective inflation; however, female-headed households tend to spend more on food, utilities, and health, while male-headed households spend more on alcoholic beverages and tobacco, transport, and restaurants and hotels.
- Within-group heterogeneity:
  - Among groups with higher average effective inflation, many households still experienced below-average effective inflation:
    - Close to 50 percent for households living in sparsely populated areas or with highest-paid member having education below tertiary.
    - Around 40 percent for households relying on secondary income sources or with highest-paid member above age 65.
  - Around 30 percent of households in each of these groups have above-average income, indicating that categorical targeting by group without income verification can include relatively well-off households.

### Policy simulations and design assumptions
- General simulation design:
  - Simulations are hypothetical and not able to replicate actual government interventions during 2022–23 due to data limitations.
  - Subsidies in simulations are defined as additional cash each recipient household receives proportional to its consumption of the subsidized products.
  - Categorical transfers are cash payments to households belonging to predefined categories; if income is not observable, all eligible households receive equal amounts.
  - For comparability, subsidies and categorical transfers are calibrated so the overall fiscal envelope equals the fiscal cost of the benchmark targeted transfer scenario.
  - Energy subsidies in the simulation are implemented against HBS category 045 (electricity, gas, and other fuels); food subsidies are implemented against HBS category 011 (food).
- Benchmark "targeted transfer" scenario:
  - All households in the bottom three income deciles that have experienced above-average effective inflation are given cash so that their loss of purchasing power is reduced to the rate of average (pre-intervention) effective inflation.
  - After intervention, the mean effective inflation within the bottom three deciles will be brought down to around 1½–2 percentage points below its original levels, or 1½ percentage points below the original sample average, comparable to the level of the top decile.
  - The intervention does not affect the top seven deciles.
  - The targeted transfer scenario has a fiscal cost of 0.1 percent of GDP.
  - The policy objective is modest: affected households continue to face a loss of purchasing power equal to the sample average (around 10 percent).
- Comparative point:
  - Targeted support tailored to recipients’ needs (income- and loss-targeted) would more effectively mitigate vulnerable households’ income loss than categorical programs without income criteria, although categorical programs can still alleviate cost-of-living pressures to some extent.
- Caveats:
  - Simulations are illustrative; sizes and specific designs are not policy prescriptions. Policymakers should calibrate interventions considering fiscal space and socioeconomic conditions.

*Source: INTERNATIONAL MONETARY FUND — 1. Cost of Living and Fiscal Support, 2022–23.*

### 10.      Both subsidies and categorical transfers to households help reduce the effective

### 10.      Both subsidies and categorical transfers to households help reduce the effective

### Distributional findings: subsidies vs categorical transfers vs targeted transfers
- Subsidies are regressive: benefits are proportional to consumption of subsidized goods/services, and higher-income households consume more.
- Share of subsidy benefits received by higher-income households:
  - Energy subsidies: 75 percent
  - Food subsidies: 80 percent
- Effect on bottom three deciles facing above-average effective inflation:
  - Pre-intervention share: 51.5 percent
  - After targeted transfer: drops to zero (by design)
  - After simulated categorical transfers: remains elevated (above 40 percent)
- Categorical transfers perform better than subsidies but still leak benefits to higher-income groups:
  - Around 60 percent of categorical-transfer benefits would be received by households in the top seven deciles
- Conclusion: targeted transfers are most effective at protecting vulnerable households within a fixed fiscal envelope because they reduce leakage of benefits to higher-income households.

### Practical considerations and caveats for targeting
- Income misreporting risk: if household income is at high risk of being mis-reported, income-targeted support becomes less effective.
- Real-world programs contain quasi-targeting elements that improve progressive impact relative to the simple simulations:
  - Electricity subsidies often reduce unit-subsidy for large users (presumed richer households), producing a more progressive distributional impact than the simulated energy subsidy scenario.
  - The food program (“Market pass”) incorporates some income criteria and would be more effective than the simulated food subsidy.
- Implementation capacity gaps to be addressed:
  - Establish a centralized registry of beneficiaries
  - Improve reporting and verification of beneficiaries’ income leveraging available third-party information
  - Address gaps in coverage and benefit levels of existing targeted programs such as the Guaranteed Minimum Income

### Simulation design and key simulation outputs
- Targeted transfers in simulations:
  - Designed to compensate households originally in the bottom three income deciles for the excess loss of real income benchmarked by the average effective inflation over the whole sample.
  - Distributional impact after policy interventions is calculated using the original income deciles; deviations presented are relative to the pre-intervention sample mean.
- Empirical context:
  - Observed inflation episode referenced: 2022
- Representative numerical outcomes highlighted:
  - Pre-intervention share of households in bottom three deciles facing above-average effective inflation: 51.5 percent
  - Targeted-transfer outcome for that group: 0 percent (by design)
  - Post-categorical-transfer outcome for that group: above 40 percent
  - Categorical-transfer leakage to top seven deciles: around 60 percent

### Methodology: compensating variation and data mapping
- Compensating variation (CV) framework (as in Causa and others (2022)):
  - CV0,1i = Σj s0i,j ⋅ π0,1j
  - s0i,j = c0i,j p0j / Σk c0i,k p0k
  - π0,1j = 100 ⋅ (p1j / p0j − 1)
  - Interpretation: CV measures minimal change in household expenditure—as a share of total expenditure under old prices—needed to make the old consumption bundle affordable under new prices; it is a measure of the effective inflation facing the household.
- Data sources and mapping:
  - Household consumption shares from 2021 Household Budget Survey (HBS)
  - Price changes (average 2022 over 2021) from HICP (Harmonized Index of Consumer Prices)
  - Mapping adjustments to align HBS and HICP concepts include excluding imputed rentals (042) and extrapolating prices for HBS items not covered in HICP using weighted averages of related HICP categories.

### Granularity, measurement issues, and aggregate comparisons
- Granular approach:
  - Using fine-grained HBS-HICP mapping (up to 355 major and minor categories) produces fatter tails in the distribution of household-specific inflation compared with using only 12 major categories.
  - Households on the right tail (high household-specific inflation) are more vulnerable, especially if income is modest.
- Measurement caveats and numeric divergences:
  - Reconstructing HICP from 355 categories yields average inflation of 9.5 percent versus the official 9.3 percent (rounding and accumulation errors).
  - Calculated compensating variation averages 10.3 percent in 2022, exceeding average HICP inflation (9.3 percent), which corresponds to the 32th percentile of the calculated compensating variation.
  - Rounding errors are pronounced for utilities (04), food (01), and transport (07).

### Empirical correlates of effective inflation and robustness
- Household characteristics associated with higher effective inflation (baseline regression and robustness checks):
  - Living in sparsely populated areas: positive and significant (baseline coefficient 0.5414***)
  - Above age 65: positive and significant (baseline coefficient 1.2321***)
  - Below tertiary education: positive and significant (baseline coefficient 0.4767***)
  - Main source of income is secondary: positive and significant (baseline coefficient 0.7035***)
  - Household monetary income (EUR 000): negative and significant (baseline coefficient -0.0150***)
  - Household equivalent size (OECD): negative and significant (baseline coefficient -0.1750**)
- Robustness:
  - Coefficients remain robust to random sampling assumption (OLS), alternative income measures (total income), OECD modified scale for household size, different definitions of reference person, breakdown of income sources, and inclusion of additional household characteristics.
- Regression sample and fit:
  - Observations: 6,047
  - R-squared (baseline): 0.2039
  - Standard errors reported; significance notation: *** p<0.01, ** p<0.05, * p<0.1

### Policy implications and recommendations
- Targeted transfers are the most effective instrument to protect vulnerable households under a fixed fiscal envelope because they minimize benefit leakage to higher-income groups.
- Strengthen quasi-targeting features where feasible (e.g., energy subsidy tapering for large users, income criteria in food programs like “Market pass”).
- Build administrative capacity to implement targeted programs:
  - Create a centralized beneficiary registry
  - Improve income reporting and third-party verification
  - Close coverage and benefit shortfalls in existing programs such as the Guaranteed Minimum Income

*Source: ELSTAT, HBS; Eurostat; Haver Analytics; and IMF staff calculations (excerpts from the provided content).*

### 3.      The paper presents key trade-offs and some initial benchmark calibration of BBMs for

### 3.      The paper presents key trade-offs and some initial benchmark calibration of BBMs for Greece

### Select Systemic Risk Assessment
- Real estate prices and market dynamics
  - Residential real estate prices have increased significantly since their trough in 2017, exceeding 50 percent in nominal terms and 35 percent in real terms.
  - Price-to-income and price-to-rent ratios exceeded their historical long-term averages by 6 and 29 percent, respectively; an econometric model suggests 8 percent.
  - Demand drivers include strong employment and real disposable income growth, increased nonresident investment (including Golden Visa program), and structural features: Greece among the lowest number of rooms per person and highest share of population living in overcrowded homes in the EU.
  - Supply response: housing investment as a share of GDP and building permits doubled since 2016 (from a low base).

- Household sector and banking exposure
  - Household sector deleveraging since the sovereign debt crisis left leverage much lower than other EA countries, but household debt service remained one of the highest in the EU (housing costs in disposable income and debt service ratio).
  - Greece records the highest share of loan-to-value (LTV) ratio exceeding 80 percent among EU countries (EBA, 2022).
  - Based on BoG data, the share of new loan disbursements with an LTV>80% stood at 5.7% in Q2 2023.
  - Bank exposure to real estate is below the EU average and less significant compared to pre-GFC levels.
  - Joint distribution of risk indicators does not suggest reasons for significant concern (BoG, 2023b).

### Macroprudential Policy Options
- Policy toolkit and legal framework
  - No policy framework for BBMs in place until 2023; legal framework amended in 2023 granting BoG power to enact BBMs for financial institutions established and operating in Greece and branches of foreign institutions in Greece, in connection with credit secured by real estate located in Greece.
  - BoG may adopt BBMs covering caps on ratios related to the credit or the borrower or specific features of the credit; it may set types of BBMs, types of credit, cap percentages, terms and conditions, and reporting requirements.

- Modalities and trade-offs
  - Income-based measures (primary instrument: caps on the DSTI ratio) aim to reduce probability of default (PD) by relating loans to debt repayment/servicing capacity.
  - Collateral-based measures (primary instrument: cap on LTV) aim to reduce loss given default (LGD) by lowering unsecured loan portions.
  - Income- and collateral-based measures are complementary and are recommended to be considered jointly to increase borrower resilience and reduce bank defaults (ESRB, 2019).
  - Other complementary measures: maturity limits, loan amortization, risk weights, targeted CBMs.

- Timing, calibration, and distributional considerations
  - Timing should consider the real estate market cycle and policy lags; early-stage activation addresses vulnerabilities while still building up.
  - BBMs affect only the flow of new lending; CBMs affect lender resilience.
  - Instruments can be calibrated as binding (taming cyclical risks) or nonbinding (curbing structural vulnerabilities), and can be static or dynamic.
  - Speed limits and exceptions can target specific borrower segments; BBMs may restrict credit access for young low-wealth households but exceptions can allow a fraction of new lending to exceed limits.

- EU implementation context
  - BBMs widely implemented in the EU; DSTI and LTV caps prominent: 14 countries implemented DSTI and 20 countries LTV.
  - Greece is among the few EU countries that had not implemented BBMs as of the study period.

### Policy Calibration — Framework
- Quantitative framework
  - Uses micro-level HFCS household data for Greece and macro-level quarterly data from the ECB and national sources.
  - Captures dynamics of household debt service, consumption expenditure, labor income, and unemployment benefits at household-member level.
  - Primary outputs: simulated PDs, LGDs, and loss rates (LRs) at household level; linked to bank mortgage portfolios to assess capital impact.
  - Model accounts for macro-financial feedbacks from policy-induced credit demand drag.
  - First-round effects: direct impact of BBM settings on PDs, LGDs, and LRs.
  - Second-round effects: add impact of negative loan demand shock on PDs, LGDs, and LRs.
  - Final step: computes impact on banks’ capital adequacy ratio via reduced loan losses and lower risk weights from lower PDs and LGDs.

### Impact on Household Resilience
- Key simulation findings
  - PD response to DSTI caps:
    - Initial PD starts to decline sharply from a DSTI cap around 90 percent, with an already sizeable impact for cap at 50 percent, and reaching maximum impact at the left end of the DSTI cap range.
  - PD and LGD comparative effects:
    - DSTI caps have stronger potential to reduce PDs.
    - LTV caps have a stronger impact on reducing LGDs, with LGDs estimated to decline sharply from an LTV cap around 90 percent.
  - Crowding-out and mortgage volume effects:
    - Mortgage volume reduction suggests a stronger impact of LTV caps compared to loss-rate-equivalent DSTI caps, meaning DSTI caps would have to bind for a smaller share of the population to achieve the same loss rate as a given LTV cap.
  - Mapping between LTV and DSTI:
    - Estimated mapping suggests that for an LTV cap at 85 percent, the equivalent DSTI cap would equal around 40 percent.

- Benchmark calibration and joint implementation
  - Calibration applied: DSTI = 40 percent and LTV = 85 percent (primary calibration; alternative calibration in Appendix 2).
  - Joint implementation of DSTI and LTV caps yields stronger increases in borrower resilience than individual caps due to complementarities: DSTI acts primarily via PD channel and LTV via LGD channel.
  - Overall LRs estimated to be sizeable and higher for DSTI caps compared to LTV caps; second-round effects from negative loan demand shocks are relatively limited.

### Impact on Bank Resilience
- Capital effects
  - BBMs are estimated to have a positive impact on bank CET1 ratios for the mortgage loan portfolio, with stronger effects for DSTI caps compared to LTV caps.
  - Decomposition: reduced loan losses from mortgage credit portfolios outweigh the benefit from lower risk weights due to lower PDs and LGDs.

### Conclusions and Policy Implications
- Risk assessment summary
  - Emerging vulnerabilities in real estate and household sectors warrant close monitoring: significant house price increases since 2017, moderate overvaluation signals, high household debt service despite low leverage.
- Policy recommendations and implications
  - Enhancing the macroprudential toolkit with BBMs would help increase resilience of households and banks and contribute to systemic risk reduction.
  - Legal framework amendment in 2023 enables BoG to activate both income- and collateral-based BBMs.
  - Simulations indicate complementarities between DSTI and LTV caps: DSTI reduces PDs more; LTV reduces LGDs more; DSTI has stronger overall impact but comes at cost of lower lending.
  - Joint implementation of DSTI and LTV caps produces stronger borrower resilience than individual caps.
  - Caps could initially be set at less binding levels and be gradually tightened over time.

### Appendix I. Data Description (selected)
- Household variables (HFCS codes and transformations) included: current value of house (DA1110), total financial assets (DA2100 - DA2104 - DA2107), outstanding balance of mortgage debt (DL1100), household income total quarterly gross DI2000 / 4, annuity for mortgage debt DL2100 * 3, household weight HW0010, year of 1st mortgage origination HB1301, duration of 1st mortgage HB1601, current interest rate on mortgage debt W.A. from mortgages outstanding (HB170x) and their interest rates (HB190x), among others.
- Macroeconomic anchors and sources: URX (Unemployment rate anchor point for the last year (2017)) — ECB Statistical Data Warehouse; IR (Short-term interest rate level anchor point for the last year) — ECB Statistical Data Warehouse; HPG (Annual house price growth in the last year) — ECB Statistical Data Warehouse; CP G, SPG, DEPR with similar sources.
- Banking inputs and sources: RWA and CET1 — ECB Statistical Data Warehouse; PiT PD and PiT LGD — EBA Risk Dashboard; mortgage loan stocks (performing and nonperforming) — ECB Statistical Data Warehouse.

*Source: IMF staff summary of the chapter titled "The paper presents key trade-offs and some initial benchmark calibration of BBMs for Greece" (PDF content unit: 1grcea2024002).*

### Appendix II. Alternative Calibrations

### Appendix II. Alternative Calibrations

### Alternative Calibration
- DSTI = 45 percent and LTV = 90 percent
- Appendix II. Figure 1. Greece: Impact of DSTI and LTV Caps on Households  
  - Sources: HFCS and IMF staff calculations.  
  - Notes: Impact estimated based on the calibration assuming DSTI = 45 and LTV = 90

### Recent Trends of Informality in Greece — Evidence from Subnational Data: Overview
- Paper focuses on informality as a structural impediment to fiscal capacity and sustainable growth in Greece.
- Main high-level findings summarized:
  - Informality has dropped significantly in Greece in recent years, with temporary increases during the sovereign debt crisis and the COVID-19 pandemic.
  - Lower informality is associated with higher subsequent per capita GDP growth and higher tax revenue.
  - Greece’s recent progress in digitalization appears to have helped reduce informality.
  - There remains scope to further reduce informality by accelerating digitalization and ongoing pro-growth structural reforms.

### Methodology and Data
- Estimation approach:
  - Multiple Indicators Multiple Causes (MIMIC) approach used to estimate a latent informality index and then calibrated to formal GDP statistics.
  - Calibration uses World Bank (2018) Enterprise Survey for Greece.
- Added data sources (subnational, quarterly, 2012–2021):
  - Satellite nighttime lights (VIIRS Day): sum of nighttime light intensity (radiance) within each region used as proxy for overall economic activity.
  - Google search volume of subnational region name used as proxy for tourism-related activity.
  - Conventional variables: labor participation rate (population aged 20-64), GDP (both transformed into first differences).
- Causes (subnational and national variables) included in analysis:
  - Subnational: agriculture employment as a share of total employment, self-employment share, unemployment rate, lagged income per capita, tourist arrivals as a share of total population, education attainment (share aged 25-64 with less than primary, primary, and lower secondary education).
  - National: Value Added Tax (VAT) gap, tax as percent of GDP, trade openness, government consumption as percent of GDP, World Bank governance indicators.
- Firm-survey calibration details:
  - Key survey questions used: “Does this establishment compete against unregistered or informal firms?” and “What is the number of permanent, full-time employees at the end of last fiscal year?”
  - Notation: e and l denote responses; e=0 indicates no competition with unregistered or informal firms and e=1 indicates competition.

### Main Findings — Magnitudes and Regional Patterns
- National and regional trends:
  - Aggregate informality (weighted by regional gross value added) declined from a peak of about 30 percent of GDP in 2013 to a low of about 16 percent of GDP in 2021.
  - During the sovereign debt crisis periods, most regions experienced increased informality by an average of 0.6 percentage point, peaking at about 30 percent of GDP at the national level.
  - Pandemic (2020) saw relatively small increases in informality in most regions, followed by continued declines in 2021.
  - Regional variation in informality declines ranges from 5 percent of GDP (Western Macedonia) to 18 percent of GDP (Crete).
  - Larger drops (above 15 percent of GDP) occurred in Attica and island regions (e.g., Crete and Ionia Islands); Western Greece saw a smaller decline at about 10 percent of GDP.
- Sectoral and regional implied informality (from 2018 firm survey):
  - By region:
    - Northern Greece: 15%
    - Central Greece: 24%
    - Attica: 20%
    - Aegean Islands: 25%
  - By sector:
    - Food: 27%
    - Fabricated Metal Products: 17%
    - Other Manufacturing: 13%
    - Retail: 29%
    - Other Services: 19%
- Fiscal and growth implications:
  - The decline in informality from about 30 percent of GDP (2013) to about 16 percent of GDP (2021) is associated with a gain in tax revenue by about 4 percentage points of GDP.
  - European Commission estimates: Greece’s Value Added Tax gap decreased by over 16 percentage point of its tax base between 2013 and 2021, equivalent to about 2 percentage points of GDP.
  - Correlation evidence shows that lower informality is associated with higher subsequent per capita GDP growth.

### Policy-relevant Insights and Implications
- Drivers and mitigation:
  - Severe recessions, fiscal tightening, and weakened payment culture contributed to increases in informality during crisis periods.
  - Government support during the pandemic likely moderated informality increases relative to prior recessions.
  - Digitalization progress likely contributed to reductions in informality and explains part of subnational variation.
- Policy priorities implied by findings:
  - Accelerate digitalization to further reduce informality.
  - Continue pro-growth structural reforms that improve labor market flexibility and business environment.
  - Strengthen tax administration and measures that narrow the VAT gap to convert informality reductions into sustained tax revenue gains.
  - Use subnational and alternative big-data indicators (satellite nightlight, google search) alongside surveys to monitor informality and tailor region-specific reforms.

*Source: Appendix II. Alternative Calibrations; “Recent Trends of Informality in Greece – Evidence from Subnational Data” (prepared by Larry Qiang Cui and Jiaxiong Yao, with research assistance by Shiqing Hua and Katherine Dai). PDF content provided.*

### 12.      Furthermore, regional-level correlations show that digital infrastructure and its usages

### 1grcea2024002 - 12.      Furthermore, regional-level correlations show that digital infrastructure and its usages

### Digital infrastructure, usage, and informality: empirical findings
- Better digitalization indicators show consistent negative correlations with the share of the informal economy at the regional level in Greece.
- Key empirical associations (regional-level correlations):
  - A 10-percentage point increase in internet access is on average associated with a reduction of informality by about 3 percentage point (Figure 6).
  - A 10-percentage point reduction in the share of residents who have never used internet is associated with a reduction in informality by about 4 percentage points (Figure 7).
  - A 10percentage point increase in the share of residents who used online purchase is associated with a reduction in informality by about 4 percentage points (Figure 8).
  - A 10-percentage point increase in the share of residents who used digital public services online purchase is associated with a reduction in informality by about 3 percentage points (Figure 9).
- These correlations are consistent with earlier findings that digital payments reduced informality and tax evasion (IOBE(2018); Hondroyannis and Papaoikonomou (2017)) and OECD (2023) findings that increased digital transactions raised tax compliance in Greece.
- The results suggest positive effects of:
  - digital infrastructure expansion,
  - digital training,
  - digital public services.
- Caveats noted:
  - Limits of correlation analysis; causality not established.
  - Potential diverging trends between variables (e.g., satellite nightlight intensity or google search) and formal/informal activities are not fully controlled.
  - Lack of subnational data on tax collections and non-cash transactions constrains tighter links to regional economic activities.
  - Social structure of Greek regions may strongly affect informality and requires separate research.

### Policy drivers behind recent declines in informality
- Factors contributing to the significant decline in informality:
  - Tightened regulation and enforcement against financial crimes and tax evasion.
  - Increased labor market flexibilities (e.g., working hours) that made formalization more attractive while preserving flexibility.
  - Significant progress in digitalization: infrastructure improvements and digital public services increased tax morale and integrated the self-employed.
  - Modernized employment information system eased registration and increased monitoring.
  - Policy continuity with political stability and reduced corruption as potential drivers of higher tax morale.
  - Rationalization of tax policies, including reducing business income tax rates and marginal personal income tax, which likely reduced incentives for tax evasion.

### Remaining gaps, potential, and recommended digitalization priorities
- Digital skills and DESI findings:
  - Greece recorded above average increase in digital skills (Figure 10) but the level remains on the low end among euro area peers (Figure 11).
  - Remaining gaps: integration of digital technology and digital public services, particularly for Small and Mediums Enterprises.
- Policy priorities to strengthen digitalization and reduce informality:
  - Further improve digital skills and human capital relevant to industry needs.
  - Enhance integration of digital technology across public services and businesses to:
    - improve monitoring of tax compliance,
    - ease business and employment registration,
    - better integrate self-employed into the formal economy.
  - Leverage the National Resilience and Recovery Plan and Next General EU funding for continued digitalization and skill training.
- Complementary structural reforms to reduce informality:
  - Improve regulatory quality and transparency to ease business entry and support productive competition.
  - Increase efficiency in spending on education and training to improve job matching and incentives for formal employment.
  - Targeted employment support for youth and women with disproportionately elevated unemployment rates.
  - Reforms linking firm and job formalization with access to finance to enhance incentives to formalize.
  - Concerted efforts to integrate high levels of self-employment into the formal economy.

### Methodological approach: MIMIC model (summary)
- The Multiple Indicator Multiple Cause (MIMIC) model links multiple observable indicators of the informal economy to multiple observable causes through a latent variable (index of the informal economy).
- Structural equation: y*_k = α′ x_k + η_k, where η_k is a structural disturbance.
- Measurement equation: y_k = β y*_k + ε_k.
- Two identification restrictions in the reduced-form regression y_k = β α′ x_k + (β η_k + ε_k):
  - The coefficient matrix Π = β α′ has rank one.
  - Error covariance Ω = σ^2 β β′ + Θ^2 (rank-one plus diagonal).
- Normalization: first indicator set equal to latent index (y_1k = y*_k + ε_1k).
- Estimation via maximum-likelihood.
- Data transformation for subnational analysis: all variables demeaned at country level to account for institutional and cultural differences.
- From index to size: once α̂ estimated, ŷ*_k = α̂′ x_k. Rescaling and recalibration needed to map the index to the size of the informal economy:
  - Level index s_k = ∑_k ŷ*_k.
  - Calibration z_k = λ s_k + δ, with λ and δ computed from known size points (or estimated via OLS when more points available).

### MIMIC estimation results (Appendix II: structural and measurement coefficients)
- Structural: Informal — coefficient std.err. z p-value
  - agriculture employment share (%) 0.47 0.12 4.02 0.00
  - self employment share (%) 0.16 . 111.51 0.13
  - unemployment rate (%) 0.27 0.08 3.42 0.00
  - tourist arrivals as mutiple of population -0.01 0.00 -6.24 0.00
  - (lagged) GDP per capita 0.32 0.03 10.63 0.00
  - population aged 25-64 with secondary education or lower  (%) -0.67 .09 -7.90 0.00
  - VAT gap 0.59 0.05 11.74 0.00
  - tax (% of GDP) -2.16 0.32 -6.74 0.00
  - trade openness -0.46 0.05 -9.35 0.00
  - government consumption (% of GDP) 1.14 0.25 4.52 0.00
  - rule of law -0.08 0.02 -3.76 0.00
  - control of corruption 0.02 .02 1.31 0.19
  - government effectivenss -0.03 0.02 -1.35 0.18
  - political stability 0.16 0.02 7.74 0.00
- Measurement:
  - (-) GDP growth Informal 1 (constrained)
  - labor participation rate growth Informal -0.14 .01 -9.79 0.00
  - Google Search Volume growth Informal -0.13 0.53 -0.24 0.81
  - Nighttime light growth Informal -0.07 0.10 -0.73 0.47

*Sources: Greece authorities, Eurostat, and IMF staff estimates.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1grcea2024002.pdf_