## sipea2025101

## Source details

**Canonical URL:** [sipea2025101](https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025101.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025101.pdf.md)
- [Structured JSON version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025101.pdf.json)

---

### Overview and purpose
- Revisits fiscal multipliers for Estonia to inform policy trade-offs and growth-friendly options for fiscal consolidation.
- Contextual drivers: a sharp and partly permanent increase in government spending triggered by the pandemic; intensifying climate- and ageing-related spending pressures; a sharp increase in defense spending due to geopolitical risks; two rounds of wide-ranging tax changes affecting PIT, CIT, VAT, and excises; and spending cuts based on comprehensive spending reviews.
- Role of granular multipliers: shed light on short-term output costs across specific fiscal instruments and underpin policy advice for fiscal consolidation.

### Headline multiplier estimates and comparative evidence
- First-year multiplier estimates reported in this SIP:
  - General fiscal shock: 0.85–1.4
  - Aggregate spending: 0.6–1.2
  - Revenue: about -0.2
- Literature summary:
  - Review of 37 papers (2002–2025) shows tax multipliers mostly negative and spending multipliers mostly positive.
  - Across 134 estimates of one-year tax/revenue multipliers, 110 show negative values; 95 of those are between zero and -1.
  - Across 56 estimates of spending multipliers, 46 are positive; 33 are between zero and 1.
  - Government consumption estimates: all but 2 out of 33 are positive.
  - Public investment estimates: 22 out of 29 are positive.

### IMF “Bucket Approach” (BA) applied to Estonia
- BA concept: classification of first-year “normal times” general fiscal multipliers based on six structural characteristics into low/medium/high buckets.
- Table 1 multiplier ranges by score category:
  - Score 0 - 3 (Low multiplier): 0.1 – 0.3
  - Score 3 - 4 (Medium multiplier): 0.4 – 0.6
  - Score 4 - 6 (High multiplier): 0.7 – 1.0
- BA scoring for Estonia as of 2025Q1:
  - Low trade openness: 0
  - High Labor market rigidity: 0.5
  - Weak automatic stabilizers: 1
  - Quasi-fixed exchange rate: 1
  - Low/safe public debt level: 1
  - Effective PFM and Revenue Administration: 1
- Resulting BA summary:
  - Aggregate score: 4.5 → implies a value in the high-multiplier range.
  - Mid-range BA estimate: 0.85.
- Business-cycle and monetary policy adjustments:
  - If economy close to lowest point in cycle, both bounds increased by 60 percent.
  - If economy close to peak, both bounds reduced by 40 percent.
  - If monetary policy close to zero-lower bound, multiplier range increased by up to 30 percent.
- Estonia-specific adjustment:
  - Staff estimate of the output gap in 2024: -2 percent of potential GDP; expected to remain broadly unchanged in 2025 and converge towards zero only after 2026.
  - Applying 60 percent upward adjustment: 0.85 x (1+0.6) = 1.36.
  - Adjusted lower bound example: 0.7 x 1.6 = 1.12.

### Static Keynesian fiscal multipliers (2024–2030 averages)
- Static multiplier depends on propensity to consume, import intensities, average tax rate, and savings rate.
- Estimated static multipliers and parameters:
  - Public Consumption (G) Static Fiscal Multiplier: 1.44
    - Import intensity (m_i): 0.12
      - direct: 0.04
      - indirect: 0.08
  - Public Investment (I) Static Fiscal Multiplier: 0.96
    - Import intensity (m_i): 0.41
      - direct: 0.25
      - indirect: 0.16
    - Note: uses the import intensity of total investment as proxy for public investment.
  - Combined G and I Static Fiscal Multiplier: 1.16
    - Combined import intensity: 0.29
      - direct: 0.16
      - indirect: 0.13
- Supporting parameter values:
  - Tax Revenue (ratio to GDP) A: 0.357
  - Transfers and Subsidies (ratio to GDP) B: 0.191
  - Net Tax Rate (A – B = t): 0.166
  - Private Savings Rate (s): 0.218
  - Propensity to consume (c = 1 − t − s): 0.617
  - Import intensity of private consumption (m_c): 0.372

### Blanchard-Perotti (BP) SVAR: data, identification, and controls
- Model scope and data:
  - Quarterly Estonian data on real GDP, total government spending, and total fiscal revenues from 2001Q1 to 2024Q2.
  - Variables normalized by the trend in GDP (HP-filtered).
  - Deterministic trends (linear and quadratic) included when statistically significant.
- Identification highlights:
  - Discretionary changes in government spending unlikely within a quarter; within-quarter variations attributed to automatic stabilizers or exogenous shocks.
  - Imposed elasticities of revenue and expenditure with respect to output to identify automatic stabilizers and isolate fiscal shock.
  - Spending shocks: government spending treated as pre-determined within the quarter (Cholesky with spending ordered first).
  - Tax shocks: taxes ordered first (tax decisions assumed to be taken first).
- Exogenous controls and dummies:
  - Seasonal dummies, indicators of known tax reforms in Estonia, ECB policy rate, a terms-of-trade index, and trade-weighted foreign partners’ GDP.
  - Seasonal dummies interacted with GDP, taxes, and spending.
- IRF normalization:
  - Impulse responses computed without quarterly dependence dummies and normalized to represent the response, in euros, to a one-euro shock in the selected fiscal instrument.

### BP SVAR empirical results (selected)
- Revenue shocks:
  - Real GDP falls for 10 quarters following exogenous increases in total net revenues (excludes interest income).
  - Non-cumulative peak response: -0.25 euro after 4 quarters (statistically significant at 5 percent for the shown horizon).
- Spending shocks:
  - Real GDP increases by about 0.7 euro immediately after a spending shock (excludes interest expenditures).
  - Non-cumulative response gradually decreases but remains statistically significant up to 4 quarters.
- Confidence bands reported at 95 percent.

### Granular fiscal multipliers (impact and cumulative) — key statistics (Table 4)
- Values reported as Impact (1st quarter), After 4 quarters, After 8 quarters, and Peak / Trough (quarter).
- Revenues:
  - Total Revenues: -0.07, -0.19, -0.52, -0.52 (8th quarter)
  - Direct Taxes: -0.62, -1.17, -1.21, -1.32 (7th quarter)
    - PIT: -0.54, -1.23, -1.47, -1.59 (7th quarter)
    - CIT: -1.07, -0.61, -0.84, -1.07 (1st quarter)
    - VAT: -0.35, -0.05, 0.19, -0.35 (2nd quarter)
- Spending:
  - Total Spending: 0.75, 0.64, 0.64, 0.75 (1st quarter)
  - Government Consumption: 1.05, 0.68, 0.35, 1.05 (1st quarter)
    - Wages and Salaries (includes employers’ social contributions): 1.89, 0.91, 0.38, 1.89 (1st quarter)
    - Intermediate consumption: 0.01, 0.00, 0.00, 0.01 (1st quarter)
  - Subsidies: 0.50, 1.12, 1.50, 1.50 (8th quarter)
  - Transfers: 0.69, -0.03, -0.10, 0.69 (1st quarter)
  - Gross Fixed Capital Formation: -0.02, 0.25, 0.17, 0.30 (2nd quarter)

### Dynamics, persistence, and interpretation
- General pattern:
  - Tax shocks: negative on impact and more persistent than spending multipliers.
  - Spending shocks: positive on impact, peak sooner (first or second quarter) and rapidly decay, except subsidies which strengthen over time.
- Specific persistence patterns:
  - Total net revenues, direct taxes, and PIT build up over time; within two years the negative multiplier effect of direct taxes on GDP doubles, driven by PIT which becomes three times as strong.
  - VAT and CIT multipliers decline after one year to almost zero (VAT) and by 40 percent (CIT); VAT peaks in the second quarter; CIT peaks in one or two quarters.
  - Subsidies: multiplier becomes three times stronger after two years, indicating a build-up as programs mature.
  - Wages and salaries: largest immediate impact (1.89 on impact) but fades by 80 percent in two years (0.38 after 8 quarters).
  - Intermediate consumption: almost null multiplier (0.01 on impact; 0.00 after 4 and 8 quarters).
  - Public investment: barely negative on impact (-0.02) but becomes positive and peaks after two quarters (0.30 at peak).

### SVECM estimates: effects of 1 percent permanent fiscal shocks on potential GDP
- Estimated effects of 1 percent permanent shocks to both revenues and spending:
  - Immediate (quarterly) decline in potential GDP growth: a little over 0.01 percentage points.
  - Compounded over 2 years: about 0.1 percentage point lower GDP.
  - Over 5 years: real GDP is about 0.2 percentage point lower following a permanent tax increase and 0.15 percentage point lower following a permanent spending increase.
  - Long run cumulative effect on GDP level: -0.25 percentage point for permanent tax shock and -0.22 percentage point for permanent spending shock.
- Interpretation:
  - Both permanent tax and spending increases have a small but persistent negative effect on potential GDP growth and permanently reduce the GDP level.
  - Negative output effect of permanent revenue shocks is larger (in absolute terms) than that of permanent spending shocks.

### Comparative consistency, identification notes, and robustness
- First-year aggregate spending multiplier larger in absolute terms than the revenue multiplier; estimates consistent with prior findings (Carnot and Castro (2015); Gersher (2015); Mineshima et al (2014); Hlaváček and Ismayilov (2022)) as discussed in the SIP.
- BP SVAR identification imposes c1 = 0 (spending does not change in response to GDP within the quarter); assumes b2 = 0 for tax shocks or c2 = 0 for spending shocks; b1 (elasticity of taxes to GDP) estimated as b1 = 0.8, with robustness check b1 = 1.

### Policy implications and trade-offs
- Short-term costs:
  - First-year multipliers indicate important short-term costs of fiscal consolidation: general fiscal variable first-year multiplier 1.1; aggregate spending 0.6−1.2; net fiscal revenues -0.2.
- Instrument-specific trade-offs:
  - Spending cuts (especially reductions in the wage bill) produce larger immediate negative output effects but are short-lived.
  - Revenue-based consolidations, especially via direct taxes (PIT and CIT), have lower immediate effects but output costs build over time and are more persistent.
  - VAT hikes do not exert a negative impact on GDP after four quarters in these estimates.
- Preferred consolidation strategies (abstracting from buoyancy, administrative burden, progressivity, income distribution, and tax avoidance risk):
  - Spending cuts unrelated to subsidies or cuts to public investment may be more growth-friendly over a two-year horizon given the high persistence of direct tax multipliers and the sizes of different multipliers.
  - Considering two-year multipliers, direct tax hikes are particularly persistent and costly over time, suggesting spending-based consolidation may be preferable where feasible.
- Long-run considerations:
  - Permanent revenue shocks entail larger negative long-term effects on potential GDP than permanent spending shocks, implying that short-run costs of spending cuts may be partially offset by relatively better long-run outcomes versus revenue-based consolidation.

*Source: IMF staff estimations and analyses in "Revisiting Fiscal Multipliers for Estonia", Selected Issues Paper (June 16, 2025).*

### 3. Fiscal Multipliers in Estonia: Total Spending ___________________________________________ 9

### 3. Fiscal Multipliers in Estonia: Total Spending

### A. Overview and purpose
- This Selected Issues Paper revisits fiscal multipliers for Estonia to inform policy trade-offs and growth-friendly options for fiscal consolidation.
- Contextual drivers cited: a sharp and partly permanent increase in government spending triggered by the pandemic; intensifying climate- and ageing-related spending pressures; a sharp increase in defense spending due to geopolitical risks; two rounds of wide-ranging tax changes affecting PIT, CIT, VAT, and excises; and spending cuts based on comprehensive spending reviews.
- Role of granular multipliers: shed light on short-term output costs across specific fiscal instruments and underpin policy advice for fiscal consolidation.

### B. Summary of headline multiplier estimates and comparative evidence
- First-year multiplier estimates reported in this SIP:
  - General fiscal shock: 0.85–1.4
  - Aggregate spending: 0.6–1.2
  - Revenue: about -0.2
- Literature overview:
  - Review of 37 papers (2002–2025) shows most estimates align with conventional Keynesian expectations: tax multipliers negative, spending multipliers positive.
  - Across 134 estimates of one-year tax/revenue multipliers, 110 show negative values; 95 of those are between zero and -1.
  - Across 56 estimates of spending multipliers, 46 are positive; 33 are between zero and 1.
  - Government consumption and investment multipliers: all but 2 out of 33 government consumption estimates are positive; 22 out of 29 public investment estimates are positive.

### C. The IMF “Bucket Approach” (BA) applied to Estonia
- BA concept: back-of-the-envelope classification of first-year “normal times” general fiscal multipliers based on six structural characteristics; countries assigned to low/medium/high buckets with corresponding multiplier ranges (Table reproduced conceptually here).
- Table 1 multiplier ranges by score category:
  - Score 0 - 3 (Low multiplier): 0.1 – 0.3
  - Score 3 - 4 (Medium multiplier): 0.4 – 0.6
  - Score 4 - 6 (High multiplier): 0.7 – 1.0
- BA scoring for Estonia as of 2025Q1 (Table 2 factors and scores):
  - Low trade openness: 0
  - High Labor market rigidity: 0.5
  - Weak automatic stabilizers: 1
  - Quasi-fixed exchange rate: 1
  - Low/safe public debt level: 1
  - Effective PFM and Revenue Administration: 1
- Resulting BA summary:
  - Aggregate score: 4.5 → implies a value in the high-multiplier range.
  - Mid-range BA estimate: 0.85.
- Business-cycle and monetary policy adjustments under BA:
  - For an economy close to the lowest point in the cycle, both bounds are increased by 60 percent.
  - For an economy close to the peak, both bounds are reduced by 40 percent.
  - If monetary policy is close to the effective zero-lower bound, multiplier range is increased by up to 30 percent.
- Estonia-specific adjustment:
  - Staff estimate of the output gap in 2024: -2 percent of potential GDP; expected to remain broadly unchanged in 2025 and converge towards zero only after 2026.
  - Applying the 60 percent upward adjustment for a large negative output gap: 0.85 x (1+0.6) = 1.36.
  - Adjusted lower bound example: 0.7 x 1.6 = 1.12.

### D. Static Keynesian fiscal multipliers (2024–2030 averages)
- Static multiplier formula (textbook specification) depends on propensity to consume, import intensities, average tax rate, and savings rate (formula provided in source).
- Estimated static multipliers and parameters (Table 3):
  - Public Consumption (G) Static Fiscal Multiplier: 1.44
    - Import intensity (m_i): 0.12
      - direct: 0.04
      - indirect: 0.08
  - Public Investment (I) Static Fiscal Multiplier: 0.96
    - Import intensity (m_i): 0.41
      - direct: 0.25
      - indirect: 0.16
    - Note: uses the import intensity of total investment as proxy for public investment.
  - Combined G and I Static Fiscal Multiplier: 1.16
    - Combined import intensity: 0.29
      - direct: 0.16
      - indirect: 0.13
- Supporting parameter values:
  - Tax Revenue (ratio to GDP) A: 0.357
  - Transfers and Subsidies (ratio to GDP) B: 0.191
  - Net Tax Rate (A – B = t): 0.166
  - Private Savings Rate (s): 0.218
  - Propensity to consume (c = 1 − t − s): 0.617
  - Import intensity of private consumption (m_c): 0.372
- Combined import intensities lead to a static multiplier of 1.16 for combined public consumption and investment.

### E. Blanchard-Perotti (BP) SVAR approach: data, identification, and controls
- Model scope and data:
  - Estimation uses Estonian quarterly data on real GDP, total government spending, and total fiscal revenues from 2001Q1 to 2024Q2.
  - Variables normalized by the trend in GDP (HP-filtered).
  - Included deterministic trends (linear and quadratic) when statistically significant as exogenous regressors.
- Identification strategy highlights:
  - BP assumes discretionary changes in government spending are unlikely to be implemented within a quarter due to implementation lags; contemporaneous variations within a quarter attributed to automatic stabilizers or exogenous fiscal shocks.
  - Independently estimated or calibrated elasticities of revenue and expenditure with respect to output are imposed as non-zero restrictions to identify automatic stabilizers and isolate the fiscal shock.
  - For spending shocks: government spending is treated as pre-determined within the quarter (Cholesky with spending ordered first).
  - For tax shocks: taxes are ordered first (tax decisions assumed to be taken first, with spending responding).
- Exogenous controls and dummies included in VAR:
  - Seasonal dummies, indicators of known tax reforms in Estonia, ECB policy rate, a terms-of-trade index, and trade-weighted foreign partners’ GDP.
  - Seasonal dummies interacted with GDP, taxes, and spending series to capture quarter-specific patterns in response of taxes to economic activity.
- IRF normalization:
  - Impulse responses computed without quarterly dependence dummies and normalized to represent the response, in euros, to a one-euro shock in the selected fiscal instrument.

### F. BP SVAR empirical results (selected)
- Revenue shocks:
  - Real GDP falls for 10 quarters following exogenous increases in total net revenues (excludes interest income).
  - Non-cumulative peak response: -0.25 euro after 4 quarters (statistically significant at 5 percent for the shown horizon).
- Spending shocks:
  - Real GDP increases by about 0.7 euro immediately after a spending shock (excludes interest expenditures).
  - Non-cumulative response gradually decreases but remains statistically significant up to 4 quarters.
- Confidence bands:
  - Figures reported with 95% confidence bands (red lines in source figures).

### G. Key interpretive findings highlighted by the SIP
- Multiplier effects in Estonia are not negligible.
- Granular multipliers suggest initially larger but less persistent output costs of spending cuts relative to tax increases.
- Static and SVAR-based estimates are broadly consistent with the BA’s implication of medium-to-high multipliers for Estonia, particularly once the recent negative output gap is accounted for.

*Source: IMF staff estimations and analyses in "Revisiting Fiscal Multipliers for Estonia", Selected Issues Paper (June 16, 2025).*

### 16. To estimate granular fiscal multipliers for different categories of revenues and

### 16. To estimate granular fiscal multipliers for different categories of revenues and 

### Methodology
- Applied a slightly modified version of the Blanchard-Perotti (BP) SVAR approach, including the fiscal instrument of interest as an endogenous variable and its aggregated variable as an exogenous control.
- Cumulative multiplier k_i(h) for fiscal instrument i at horizon h calculated following Ramey and Zubairy (2018).
- Only statistically significant responses at the 5 percent level are considered in the computation of multipliers.
- For long-run/permanent vs temporary effects on potential GDP, estimated a Structural Vector Error Correction Model (SVECM) with Bayesian techniques using quarterly Estonian data on real GDP, total spending and total fiscal revenues from 2001Q1 to 2024Q2.
- SVECM setup:
  - Contains 10 endogenous variables in three blocks: trend and cycle for real GDP, fiscal revenues, and total spending.
  - Temporary shocks affect cycles only; permanent shocks affect both cycles and trends (permanent on level, temporary on growth).
  - Model calibrated so immediate effects of temporary fiscal shocks match BP SVAR results (when converted to euros).
  - Composite variable Ω_a_y includes detrended series of capacity utilization, unemployment rate, goods imports, industrial production, retail sales, and an economic confidence indicator to identify the business cycle.

### Granular fiscal multipliers (impact and cumulative) — key statistics from Table 4
- Note: entries are reported as Impact (1st quarter), After 4 quarters, After 8 quarters, and Peak / Trough (quarter).
- Total Revenues: -0.07, -0.19, -0.52, -0.52 (8th quarter)
- Direct Taxes: -0.62, -1.17, -1.21, -1.32 (7th quarter)
  - PIT: -0.54, -1.23, -1.47, -1.59 (7th quarter)
  - CIT: -1.07, -0.61, -0.84, -1.07 (1st quarter)
  - VAT: -0.35, -0.05, 0.19, -0.35 (2nd quarter)
- Total Spending: 0.75, 0.64, 0.64, 0.75 (1st quarter)
  - Government Consumption: 1.05, 0.68, 0.35, 1.05 (1st quarter)
    - Wages and Salaries (includes employers’ social contributions): 1.89, 0.91, 0.38, 1.89 (1st quarter)
    - Intermediate consumption: 0.01, 0.00, 0.00, 0.01 (1st quarter)
  - Subsidies: 0.50, 1.12, 1.50, 1.50 (8th quarter)
  - Transfers: 0.69, -0.03, -0.10, 0.69 (1st quarter)
  - Gross Fixed Capital Formation: -0.02, 0.25, 0.17, 0.30 (2nd quarter)

### Dynamics, persistence, and interpretation
- General pattern:
  - Tax shocks: consistently negative on impact (within a quarter) and tend to be more persistent than spending multipliers.
  - Spending shocks: positive on impact, tend to peak sooner (first or second quarter) and rapidly decay, except for subsidies which strengthen over time.
- Specific persistence patterns:
  - Total net revenues, direct taxes, and PIT build up over time; within two years the negative multiplier effect of direct taxes on GDP doubles, driven by PIT which becomes three times as strong.
  - VAT and CIT multipliers decline (in absolute terms) after one year to almost zero (VAT) and by 40 percent (CIT), respectively; VAT peaks in the second quarter; CIT peaks in one or two quarters.
  - Subsidies: multiplier becomes three times stronger after two years, indicating a build-up as programs mature.
  - Wages and salaries: largest immediate impact (1.89 on impact) but fades by 80 percent in two years (becomes 0.38 after 8 quarters).
  - Intermediate consumption: almost null multiplier (0.01 on impact; 0.00 after 4 and 8 quarters).
  - Public investment: barely negative on impact (-0.02) but becomes positive and peaks after two quarters (0.30 at peak), generally smaller than government consumption multipliers but larger and more persistent than intermediate consumption.

### SVECM estimates: effects of 1 percent permanent fiscal shocks on potential GDP
- Estimated effect of 1 percent permanent shocks to both revenues and spending:
  - Immediate (quarterly) decline in potential GDP growth: a little over 0.01 percentage points.
  - Compounded over 2 years: about 0.1 percentage point lower GDP.
  - Over 5 years: real GDP is about 0.2 percentage point lower following a permanent tax increase and 0.15 percentage point lower following a permanent spending increase.
  - Long run cumulative effect on GDP level: -0.25 percentage point for permanent tax shock and -0.22 percentage point for permanent spending shock.
- Interpretation:
  - Both permanent tax and spending increases have a small but persistent negative effect on potential GDP growth and permanently reduce the GDP level.
  - Negative output effect of permanent revenue shocks is larger (in absolute terms) than that of permanent spending shocks.

### Comparative consistency with literature and auxiliary results
- First-year aggregate spending multiplier larger in absolute terms than the revenue multiplier; estimates consistent with:
  - Carnot and Castro (2015): average one-year aggregate spending multiplier between two to three times as large (in absolute terms) as revenue multipliers for the European Union.
  - Gersher (2015) meta-regression: spending multipliers exceed tax multipliers by 0.3−0.4 unit.
  - Mineshima et al (2014) survey: first-year multipliers average 0.75 for government spending and -0.25 for government revenues in advanced economies (our estimates somewhat smaller but broadly consistent).
  - Hlaváček and Ismayilov (2022) meta-analysis: spending multipliers in the range of 0.75−0.82.
- Notes on identification and robustness:
  - BP SVAR identification imposes c1 = 0 (spending does not change in response to GDP within the quarter); assumes b2 = 0 for tax shocks or c2 = 0 for spending shocks; b1 (elasticity of taxes to GDP) estimated as b1 = 0.8, with robustness check b1 = 1.

### Policy implications and trade-offs
- Short-term costs:
  - First-year multipliers indicate important short-term costs of fiscal consolidation: general fiscal variable first-year multiplier 1.1; aggregate spending 0.6−1.2; net fiscal revenues -0.2.
- Instrument-specific trade-offs:
  - Spending cuts (especially reductions in the wage bill) produce larger immediate negative output effects but are short-lived.
  - Revenue-based consolidations, especially via direct taxes (PIT and CIT), have lower immediate effects but output costs build over time and are more persistent.
  - VAT hikes do not exert a negative impact on GDP after four quarters in these estimates.
- Preferred consolidation strategies (abstraction made from other considerations such as buoyancy, administrative burden, progressivity, income distribution, and tax avoidance risk):
  - Spending cuts unrelated to subsidies or cuts to public investment may be more growth-friendly over a two-year horizon given the high persistence of direct tax multipliers and the sizes of different multipliers.
  - Considering two-year multipliers (Ramey and Zubairy (2018) recommendation), direct tax hikes are particularly persistent and costly over time, suggesting spending-based consolidation may be preferable where feasible.
- Long-run considerations:
  - Permanent revenue shocks entail larger negative long-term effects on potential GDP than permanent spending shocks, implying that short-run costs of spending cuts may be partially offset by permanent positive effects on potential GDP relative to revenue-based consolidation.

*Source: IMF staff estimations.*

---


_Source: https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025101.pdf_