## 1estea2020002

## Source details

**Canonical URL:** [1estea2020002](https://www.imf.org/-/media/files/publications/cr/2020/english/1estea2020002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2020/english/1estea2020002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2020/english/1estea2020002.pdf.json)

---

### Introduction: purpose, scope, and methodology
- Purpose and scope:
  - Standard REER indexes assume trade is only in final goods; Estonia is highly integrated into global value chains (GVCs), so competitiveness assessments should account for trade in value added.
  - The paper uses a structural model (Bems and Johnson, 2017) to derive a value-added REER (VA-REER) that accounts for supply-chain linkages and maps trade in inputs and final goods.
  - The VA-REER is used to investigate the implications of over-valuation (and appreciation) for value-added exports and real GDP growth, and to assess costs from trade-tension induced tariff hikes.
- VA-REER methodology (model and data):
  - Derived from a theoretical framework distinguishing gross output and value-added by modeling production and trade in final goods and inputs.
  - Employs structural framework developed by Bems and Johnson (2017).
  - Data source: 2016 vintage of the World Input-Output Database (WIOD, Timmer et al., 2015).
  - Assumes weights remain constant from 2014 through 2018 due to high persistency.
  - A Leontief production function version assumes elasticities of substitution across inputs and between inputs and value added are zero to capture full global input-output linkages in the short run.

### Estonia’s GVC participation and bilateral linkages
- Key facts:
  - Estonia’s participation in GVCs is about 75 percent.
  - Baltic peers: Lithuania 64 percent, Latvia 60 percent.
  - Estonia’s participation exceeds the EU28 average.
- Backward participation (use of foreign intermediates in exports) important in:
  - electrical and machinery, textiles and apparel, petroleum and chemicals, transport, metal products, and wood and paper.
- Main sources of foreign intermediate imports: Finland, Russia, Germany, and Sweden.
- Forward participation (Estonia’s value added used as inputs abroad) predominantly with Finland and Sweden, concentrated in:
  - business and financial services, transport, petroleum and chemicals, woods and paper, and agricultural sectors.
- Services exhibit more forward linkages relative to manufacturing (Banh, Wingender and Gueye et al, 2019, forthcoming).

### Comparison with conventional REER and drivers of the gap
- VA-REER has appreciated more rapidly than conventional REER for Estonia post-adoption of the Euro.
- Correlation between VA-REER and unit labor costs: 0.8.
- Bilateral weight deviations (absolute percentage deviation):
  - Finland and Sweden: ranges from 36 percent to 79 percent, respectively.
  - China and USA: absolute percentage deviation averages 5 percent.
  - Deviations smaller for Russia and Germany and negligible for Poland.
- Decomposition of VA-REER vs conventional REER gap:
  - About 74 percent of the gap is explained by price differentials.
  - Remaining part explained by the weight component.
  - Half of the price gap is explained by Estonia’s own prices used in the VA-REER (the GDP deflator) compared to partner price differentials.
  - Value-added weights account for 29.7 percent of the gap, with elasticities accounting for 71.3 percent (and a small role for weights).

### Main implications and findings (overview)
- Incorporating GVCs into REER measurement (VA-REER) suggests potential competitiveness problems in Estonia that are less visible using conventional gross-trade-based REER measures.
- Close tracking of VA-REER with unit labor costs implies competitiveness in supplying domestic value added may be highly dependent on labor-cost developments.
- Observable effects of VA-REER overvaluation/appreciation on value-added export growth and real GDP growth are estimated (details below).
- Trade-tension induced tariff hikes could impose important costs on value added produced in Estonia.

---

### Price differential: GDP deflator versus CPI and drivers
- Price level developments:
  - The GDP deflator has grown cumulatively by 16.8 percent since 2012.
  - The CPI has increased cumulatively by 12.2 percent over the same period.
  - The cumulative difference between the GDP deflator and the CPI has increased substantially since 2012.
- Decomposition and drivers of the GDP deflator gap:
  - Early opening of the gap mainly accounted for by a rise in prices of capital goods.
  - Prices of capital goods have largely decreased since 2012 (with a pickup in 2017).
  - Using the income definition of GDP, unit labor costs (ULC) have been a steady driver of final output prices, increasing by about 5.3 percent annually since 2013.
  - The gap between the GDP deflator and CPI appears particularly large in Estonia compared with some euro area trading partners.

### VA-REER misalignment, export effects, and transmission to growth
- Empirical approach:
  - Local projection method à la Jordà (2005) applied to panel data of 27 European countries over 2003–13.
  - Controls: inflation, real GDP per capita, net foreign direct investment inflows, external demand; country and time fixed effects included.
- Impact of VA-REER over-valuation on value-added export growth:
  - A 10-percentage point over-valuation in VA-REER leads to a statistically significant reduction in value-added export growth by 0.8 percentage point in the first year, which cumulates to 1.5 percentage point by the third year.
  - Degree of GVC integration matters:
    - Sample median GVC participation index = 69.8.
    - In countries highly integrated into GVCs (index > 69.8), a 10-percentage point VA-REER over-valuation leads to a reduction in VA export growth by 1 percentage point in the first year and cumulates to 1.8 percentage point in the third year.
    - No statistically significant effect for countries weakly integrated into GVCs.
- Transmission to real GDP growth:
  - Estimated relationship: a 1 percentage point increase in real value-added exports is associated with a 0.3 percentage point increase in real GDP growth cumulatively over 4 years.
  - Using the trade–growth link, a 10 percent over-valuation in VA-REER could reduce growth rate by 0.5 percentage point (calculated as 0.3 * -1.5).
- Implied impact for Estonia:
  - Estonia is highly integrated into GVCs.
  - Average over-valuation in the VA-REER over the period = 4 percent.
  - Implied reduction in growth for Estonia = 0.2 percentage point given the average 4 percent over-valuation.
  - Growth would have been higher by 0.2 percentage point on average over the period in Estonia if there were no over-valuation (and thus no such rise in ULC), given the strong pass-through.

### Tariff shocks, GVCs, and short-run value-added effects (partial equilibrium estimates)
- Model and data:
  - Structural model: Bems and Johnson (2017) framework with WIOD 2016 vintage and World Economic Outlook data.
  - Estimates short-run partial-equilibrium effects on gross trade and value-added.
- Estonia’s exposure and simulated shocks:
  - VA weights for China, US, and UK are higher than gross trade weights, implying these partners are relatively more important for Estonia’s competitiveness once supply chain linkages are accounted for.
  - A 5.9 percent tariff imposed by the US on its imports (Layer 1), with retaliation from all countries using the same tariff, would lead to a reduction of 0.3 percent in Estonia’s value added (three times larger than the reduction in gross turnover flows).
  - A cumulated tariff shock (all three layers equivalent to a 14.7 percent tariff) would reduce value added produced in Estonia by 0.6 percent.
  - Largest exposure in Europe found in Germany where reduction in domestically produced value-added reaches 0.43 percent.

### Policy conclusions and implications (VA-REER and trade shocks)
- The VA-REER index that accounts for input-output linkages suggests more competitive problems for Estonia than a standard REER based on gross trade would imply.
- Recent rise in unit labor cost may have been a drag on Estonia’s ability to supply its domestic value added on world markets.
- Policy recommendations and scopes for improvement:
  - Prevent long-term misalignment between wage growth and productivity to preserve competitiveness.
  - For backward GVC participation: enhance sophistication of production by greater use of imported intermediate goods with high-technological content to benefit from knowledge transfers, diversify exports, and improve product quality.
  - For forward GVC participation: improve allocation and incentives for innovation—through better access to credit and skilled labor with knowledge of foreign markets—to yield productivity gains, particularly for firms operating in upstream GVCs.
  - To mitigate exposure to trade shocks in GVCs, pursue policies aimed at enhancing product sophistication or quality and export market diversification.

---

### Inequality and poverty: headline outcomes and distributional dynamics
- Key inequality and poverty statistics (2017 and trends):
  - Income inequality (Gini index, equivalized disposable income):
    - Estonia: 31.6 points in 2017.
    - EA19 average: 30.6 points in 2017.
  - At-risk-of-poverty rate (share of population with disposable income lower than 60 percent of median equivalized disposable income):
    - Estonia: 21 percent in 2017.
    - EU27 average: 17 percent in 2017.
  - Persistent at-risk-of-poverty (share living below the 60 percent threshold for current year and at least two of the three preceding years):
    - Estonia: 16 percent in 2017.
    - EU28 average: 11 percent in 2017.
    - Estonia women: 18 percent in 2017.
    - EU27 women average: 11.6 percent in 2017.
  - Elderly vulnerability:
    - Proportion of elderly at-risk-of-poverty reached 41 percent (historical high).
    - Elderly poverty diverged from other groups particularly since 2013.
  - Broad income gains amid unequal distribution:
    - Real GDP per capita: from euros 7,600 (2000) to euros 15,100 (2018).
    - Average monthly wages growth: 5.9 percent (average over 2010–18).
    - Headcount ratio of extreme poverty (World Bank measure, consumption below $1.90/day 2011 PPP): decreased from 2.5 percent to 0.5 percent in 2015, an 80 percent reduction.
    - Household incomes (growth incidence, 2010–17): average growth 6.1 percent; bottom 10 percent grew 5.6 percent; seventh decile grew 7.2 percent.

### Decomposition methodology and drivers of inequality
- Theil index and decomposition:
  - TT = (1/N) ∑ (yi / ȳ) ln(yi / ȳ).
  - T = ∑ s_g T_g + ∑ s_g ln(s_g / p_g) where first term = within-group inequality and second term = between-group inequality.
- Subgroup classifications used (2010–17): residence (16 counties and rural/urban), gender, education (six ISCED categories), age cohorts (15–24, 25–54, 55–64, over 65), economic status (employee, self-employed, retired, other nonactive), economic sector of household head (six categories).
- Quantified contributions to recent trend in inequality:
  - Economic sector: accounts for 26.7 percent of the recent trend.
  - Economic status: contribution similar in magnitude to economic sector; between-group inequality of economic status grew by 43 percent between 2013 and 2010 and by 19 percent between 2010 and 2017.
  - Education: about 14.7 percent of the recent trend.
  - Age cohort: 13.5 percent of the recent trend.
  - County of residence: 8.4 percent of the recent trend.
  - Gender: 5 percent of the recent trend.
  - Rural versus urban difference: very insignificant.

### Gender pay gap (GPG): levels, trends, and drivers
- Unadjusted GPG at the mean (full-time workers):
  - 2010: 25.6%
  - 2012: 30.0%
  - 2013: 31.5%
  - 2014: 26.1%
  - 2015: 25.2%
  - 2016: 24.2%
  - 2017: 18.5%
- Unadjusted female/male log net monthly wage ratio:
  - 2010: 77.4%
  - 2012: 74.1%
  - 2013: 73.0%
  - 2014: 77.1%
  - 2015: 77.7%
  - 2016: 78.5%
  - 2017: 83.1%
- Dynamics and interpretation:
  - The unadjusted GPG (at the mean) decreased by 28 percent between 2010 and 2017.
  - Female to male net monthly wage ratio increased from 77 percent (2010) to 83 percent (2017).
  - GPG is on average lower at the bottom of the distribution compared to the top.
  - The rise in minimum income by 9 percent on average since 2012 may have contributed to GPG reduction at the bottom.
- Adjusted GPG (Oaxaca-Blinder and Mincer-type regressions):
  - Adjusted metrics are lower for GPG and higher for female/male wage ratio compared with unadjusted metrics.
  - Over 2010–16, once adjusted, female to male earnings ratio increased by 3 percent while the GPG decreased by 11 percent on average.
  - Estimation controls: education, experience, health, county and rural/urban location, managerial/supervisory position, and disability status.
- Unexplained component:
  - On average about 86 percent of the GPG cannot be explained over the period 2010–16.
  - The unexplained part may capture biased practices and/or unmeasured productivity differences.
- Labor market institutions and GPG:
  - High minimum wages can reduce the GPG because women are overrepresented among low-paid workers.
  - Collective bargaining coverage can set wage floors and raise wages of low-paid (often female) workers.
  - Low-wage incidence by gender (2014):
    - Share of women low wage earners: 29.4 percent.
    - Share of men low wage earners: 13.9 percent.
    - EU28 averages: 21.2 percent (women) and 13.5 percent (men).
  - Estonia characterized by relatively low minimum wage and low collective bargaining coverage.
  - Long parental leave may influence GPG by keeping women away from work for extended periods.

### Policies to reduce inequality: cross-country empirical estimates and counterfactuals
- Cross-country fixed effects model (equation (3)):
  - Gini_c,t = δ PolicyVars_{c,t-1} + β X_{c,t} + α_c + τ_t + ε_{c,t}.
  - Policy variables (lagged): public social protection expenditure as share of GDP and property tax revenues as share of GDP.
  - Data: 21 European countries over 2004–17.
- Predicted impacts if Estonia’s policy variables were at the EU28 average over the period:
  - Increasing property tax revenues to the EU28 average would have led to Gini index lower by 3 percent.
  - Increasing social protection spending to the EU28 average would have led to Gini index lower by 11.3 percent.
- Contextual policy numbers:
  - Property tax revenues as share of GDP: Estonia 0.3 percent; EU 1.7 percent.
  - Social protection spending as share of GDP: Estonia 16 percent; EU 23.9 percent.

---

### Estimation results and policy implications (model selection, coefficients, and recommendations)
- Model selection:
  - Two versions of equation (3) estimated; model 2 selected as it "captures better the trend in the Gini index (model 2)."
  - Model (2) accounts for nonlinearities of control variables and includes lagged policy variables.
- Selected coefficients and statistics (robust standard errors in parentheses; significance: * p<0.10, ** p<0.05, *** p<0.01):
  - Property Tax Revenues/GDP (t-1):
    - (1): -0.7709*** (0.2683)
    - (2): -0.5653** (0.2480)
  - Social Protection Spending/GDP (t-1):
    - (1): -0.3829*** (0.0789)
    - (2): -0.4574*** (0.0773)
  - Share of Employment in the Service Sector:
    - (1): -0.2980** (0.1235)
    - (2): 0.7099 (1.0683)
  - Share of Employment in Industry:
    - (1): -0.3834** (0.1520)
    - (2): -1.4193*** (0.4847)
  - Share of Urban Population:
    - (1): -0.2312** (0.1060)
    - (2): -1.3987*** (0.5219)
  - Share of the Population with Tertiary Education:
    - (1): -0.1202*** (0.0380)
    - (2): -0.2172 (0.1441)
  - Log Real GDP per Capita:
    - (1): -6.6083*** (2.3794)
    - (2): -7.6205*** (2.1407)
  - Unemployment Rate:
    - (1): 0.1242** (0.0533)
    - (2): 0.0432 (0.0603)
  - Output gap:
    - (1): 0.0526 (0.0576)
    - (2): -0.0148 (0.0609)
  - Nonlinear terms included in Model (2):
    - Share of Employment in the Service Sector^2: -0.0119 (0.0165)
    - Share of Employment in Industry^2: 0.0274** (0.0121)
    - Share of Urban Population^2: 0.0074** (0.0033)
    - Share of the Population with Tertiary Education^2: 0.0016 (0.0026)
- Model fit and sample:
  - Observations: 285 (for both models).
  - Adjusted R-squared:
    - Model (1): 0.904
    - Model (2): 0.913
  - Country Fixed Effects: Yes.
  - Year Fixed Effects: Yes.
- Policy implications and priorities:
  - Increase social protection spending (including pensions for the old) to strengthen the social safety net and reduce inequality.
  - Consider broadening the tax base via property taxation to reduce inequality; moving property tax revenues closer to EU28 average could yield measurable Gini reductions.
  - Enhance transparency and targeted measures to reduce the gender pay gap, including wider mandatory or encouraged reporting and measures to reduce occupational segregation and career interruptions for women.
  - Use renewed national and sectoral ML/TF risk assessment and data-driven monitoring to better target policy measures where they most reduce inequality.

---

### EFSA AML/CFT supervision: risk understanding, inspections, and recommendations (chapter 6)
- Data collection and off-site monitoring:
  - EFSA’s understanding of ML/TF risks is based largely on an annual questionnaire collecting hundreds of data points from supervised entities, focusing on non-residents and cross-border payments.
  - Since 2016 EFSA employs a tool to assess and quantify FI-specific ML/TF threats and vulnerabilities and to generate risk profiles; EFSA submitted 84 inquiries to supervised entities over the last 4 years.
- On-site inspections: scope, capacity, and outcomes:
  - EFSA conducts comprehensive/"full-scope" AML/CFT on-site inspections focusing almost exclusively on FIs it considers higher risk.
  - Typical on-site resource use:
    - EFSA usually spends from one to two months on-site.
    - On-site team involves, on average, around 3 full-time employees (FTEs).
    - Final report available at the latest six months after the start of a full-scope on-site inspection.
  - Full-scope inspections are resource-intensive and limit coverage:
    - At current staffing levels, full-scope inspections effectively limit the number of FIs that may be inspected to a maximum of five each year.
    - From 2014 to 2018, EFSA conducted 29 on-site inspections, including 7 on-site inspections of banks over the last 3 years.
  - Full-scope inspections have uncovered significant violations (examples: Estonian branch of Danske Bank and Versobank).
- Options to increase coverage and improve inspection model:
  - Two principal ways to increase coverage:
    - Substantial increase in AML/CFT Division staff.
    - Development of risk-based targeted and thematic inspections.
  - Benefits of targeted/thematic inspections:
    - Shorter, focused inspections targeting elevated-risk aspects within an FI or sector-wide categories.
    - Enables more dynamic response to evolving ML/TF risks; EFSA’s off-site monitoring and historic data facilitate strategic targeted inspections.
- Supervisory resources, scope, and workload:
  - AML/CFT Department operational in 2019 with 7 persons (7.5 percent of total staff), up from 4 in 2018.
  - EFSA supervises 110 entities for AML/CFT obligations, including:
    - 17 credit institutions (9 banks and 8 branches of foreign banks)
    - 15 fund management companies
    - 11 payment service providers
    - 5 life insurance companies
    - 5 investment firms
    - 57 consumer credit loan providers (considered lower ML/TF risk)
  - Financial sector asset statistics:
    - Total assets held by financial institutions: EUR 45 billion.
    - 83 percent of assets are held by the banking sector.
    - In the banking sector, three largest banks account for more than 84 percent of the assets.
  - EFSA notes scrutiny of recently revealed ML cases has taxed resources but believes it can conduct robust AML/CFT supervision with current staff, asserting ML/TF risks are decreasing while “legacy” cases are getting resolved.
- Procedures, sampling methodology, and enforcement powers:
  - Established routines for processing annual questionnaire data, classification of FIs by risk, and institutional risk profiles; AML/CFT inspection manual guides inspections.
  - Current sampling methodology focuses on customers presenting specific risk factors; recommended refinements include risk-based random transaction sampling and suspicious transaction reporting sample testing.
  - Enforcement framework and sanctions:
    - Range of sanctions: informal warnings, letters, administrative precepts, fines (via the misdemeanor process), and license withdrawals.
    - Penalty payments for non-compliance with precepts:
      - up to EUR 32 000 for the first specified time increment,
      - up to EUR 100 000 for each consecutive time increment,
      - and up to a total of 5 000 000.
  - Notable enforcement actions:
    - Precept to Estonian branch of Danske Bank to prohibit its non-resident business activity; February 2019 precept requiring Danske bank to terminate its activities in Estonia.
    - October 2019: EFSA opened a misdemeanor case after on-site inspection of Estonian subsidiary of Swedbank.
    - March 2018: European Central Bank, on request of EFSA, withdrew banking license of Versobank.
    - EFSA withdrew licenses of three payment service providers for AML/CFT breaches in 2019.
- Limitations of fines and reform needs:
  - Fines are available only under the misdemeanor framework, which requires establishing culpability of a specific natural person and a two-year limitations period between violation and final judgement.
  - Maximum fine per misdemeanor was increased in November 2017 from EUR 32 000 to EUR 400 000.
  - Legislation being drafted would increase the maximum fine to EUR 5 million or up to a certain percentage of revenue.
  - Recommendation: streamline and simplify the process of imposing fines by reintroducing an administrative sanctions regime for AML/CFT violations and/or establishing direct criminal liability of legal persons while extending the statute of limitations.
- Domestic and international cooperation:
  - MLTFPA established a governmental AML/CFT Committee chaired by the Minister of Finance with 14 other high-level members; committee meets quarterly.
  - Need for intensified continuous cooperation among EFIU, EFSA, Eesti Pank, Estonian Internal Security Service, Prosecutor General and law enforcement on operational issues and evolving ML/TF risks.
  - EFSA–EFIU cooperation generally ad hoc; recommendation to operationalize a Center for Strategic Analysis within EFIU (adopted October 2019, not yet operational).
- Selected recommendations (summarized):
  - Increase number and range of FIs subject to on-site inspections via risk-based targeted/thematic inspections or by increasing dedicated AML/CFT staff.
  - Under EFIU leadership, conduct comprehensive sectoral risk assessments regularly and incorporate results into inspection schedules.
  - Develop baseline internal procedures for addressing identified violations and following up on enforcement actions.
  - Develop suspicious transaction reporting sample testing.
  - Refine customer due diligence sampling methodology by introducing independent transaction sampling, randomness, risk-proportional sample sizes, and additional samples as necessary.
  - Streamline and simplify the process of imposing monetary penalties on FIs.
  - Increase the maximum monetary penalty for AML/CFT-related violations.

---

### EFSA outreach, guidance to private sector, and cooperation (chapter 24)
- EFSA outreach and guidance:
  - November 2018 guidelines covered risk management, due diligence measures, record-keeping, refusal/termination of business relationships, and reporting of suspicious transactions; included information on ML/TF risk factors and methods relevant to Estonia.
  - Private sector feedback: banks report supervisory expectations—particularly regarding non-resident customers expected to have a valid connection to Estonia—are clearly communicated; FIs note EFSA’s proactive role.
  - EFSA engagement: regular bilateral meetings, trainings, involvement with Estonian Banking Association compliance meetings, and informing supervised entities of off-site monitoring conclusions.
- Framework and scope for AML/CFT cooperation:
  - All foreign banks operating in Estonia are EU-based; EFSA has AML/CFT-specific MOUs with Russia and Switzerland; additional MOUs could be useful but current arrangements are not judged insufficient.
  - EFSA founding member of Nordic-Baltic supervisors working group (May 2019); member of multinational supervisory colleges for four banks with significant presence in Estonia.
  - Options for formal integration of AML/CFT supervision at EU or Nordic-Baltic levels are met with ambivalence in Estonia.
- Recommendations from staff (selected):
  - EFIU should provide, upon EFSA request, assessments of the nature, quantity, quality, and pertinence of FIs’ suspicious transaction reporting.
  - Fully operationalize and staff the Center for Strategic Analysis within EFIU with necessary expertise for risk analysis.
  - When helpful and feasible, participate in foreign authorities’ on-site inspections of parent banks and enable foreign inspectors to join EFSA inspections as appropriate.
  - Consider supporting further integration/consolidation of AML/CFT supervision at the EU or Nordic-Baltic levels.
- Conclusion and priority actions:
  - Important progress achieved in recent years lays foundation for further enhancement of AML/CFT supervision.
  - Authorities should prioritize:
    - (i) streamlining and simplifying the process of imposing fines on financial institutions;
    - (ii) increasing the number and range of FIs subject to on-site inspections each year by increasing AML/CFT staff and/or developing risk-based targeted and thematic inspections;
    - (iii) consider supporting further integration/consolidation of AML/CFT supervision at the EU or Nordic-Baltic levels.

*International Monetary Fund — Republic of Estonia, Selected Issues (content unit 1estea2020002).*

### Introduction  _________________________________________________________________________ 3

### 1estea2020002 - Introduction

### Purpose and scope
- Standard REER indexes assume trade is only in final goods; Estonia is highly integrated into global value chains (GVCs), so competitiveness assessments should account for trade in value added.
- The paper uses a structural model (Bems and Johnson, 2017) to derive a value-added REER (VA-REER) that accounts for supply-chain linkages and maps trade in inputs and final goods.
- The VA-REER is used to investigate the implications of over-valuation (and appreciation) for value-added exports and real GDP growth, and to assess costs from trade-tension induced tariff hikes.

### Estonia’s GVC participation (key facts)
- Estonia’s participation in GVCs is about 75 percent.
- Baltic peers: Lithuania 64 percent, Latvia 60 percent.
- Regional comparison: Estonia’s participation exceeds the EU28 average.
- Backward participation (use of foreign intermediates in exports) has been important in: electrical and machinery, textiles and apparel, petroleum and chemicals, transport, metal products, and wood and paper.
- Main sources of foreign intermediate imports: Finland, Russia, Germany, and Sweden.
- Forward participation (Estonia’s value added used as inputs abroad) is predominantly with Finland and Sweden, and concentrated in business and financial services, transport, petroleum and chemicals, woods and paper, and agricultural sectors.
- Services exhibit more forward linkages relative to manufacturing (Banh, Wingender and Gueye et al, 2019, forthcoming).

### VA-REER methodology (model and data)
- The VA-REER is derived from a theoretical framework that distinguishes gross output and value-added by modeling production and trade in final goods and inputs.
- The paper employs the structural framework developed by Bems and Johnson (2017).
- Data source for computations: 2016 vintage of the World Input-Output Database (WIOD, Timmer et al., 2015).
- Given high persistency in the weights, it is assumed that they remain constant from 2014 through 2018.
- A version of VA-REER captures full global input-output linkages by assuming elasticities of substitution across inputs and between inputs and value added are zero (Leontief production function), reflecting rigid short-run production chains.

### Comparison with conventional REER and drivers of the gap
- VA-REER has appreciated more rapidly than conventional REER for Estonia post-adoption of the Euro.
- Correlation between VA-REER and unit labor costs: 0.8.
- Bilateral value-added weights are generally lower than conventional weights for most trading partners:
  - Absolute percentage deviation ranges from 36 percent to 79 percent for Finland and Sweden, respectively.
  - For China and USA the absolute percentage deviation averages 5 percent.
  - Deviations are smaller for Russia and Germany and negligible for Poland.
- Decomposition of the VA-REER vs conventional REER gap:
  - About 74 percent of the gap is explained by price differentials.
  - The remaining part is explained by the weight component.
  - Half of the price gap is explained by Estonia’s own prices used in the VA-REER (the GDP deflator) compared to partner price differentials.
  - Value-added weights account for 29.7 percent of the gap, with elasticities accounting for 71.3 percent (and a small role for weights).

### Main implications and findings
- Incorporating GVCs into REER measurement (VA-REER) suggests potential competitiveness problems in Estonia that are less visible using conventional gross-trade-based REER measures.
- The close tracking of VA-REER with unit labor costs implies competitiveness in supplying domestic value added may be highly dependent on labor-cost developments.
- The paper finds observable effects of VA-REER overvaluation/appreciation on value-added export growth and real GDP growth (details presented elsewhere in the paper).
- Trade-tension induced tariff hikes could impose important costs on value added produced in Estonia.

_ Source: 1estea2020002 - Introduction

### 14.      Estonia’s price differential shows large discrepancies between the GDP deflator and

### 1estea2020002 - 14.      Estonia’s price differential shows large discrepancies between the GDP deflator and

### Price differential: GDP deflator versus CPI
- The GDP deflator has grown cumulatively by 16.8 percent since 2012.
- The CPI has increased cumulatively by 12.2 percent over the same period.
- The cumulative difference between the GDP deflator and the CPI has increased substantially since 2012.

### Decomposition and drivers of the GDP deflator gap
- Early opening of the gap was mainly accounted for by a rise in prices of capital goods.
- Prices of capital goods have largely decreased since 2012 (with a pickup in 2017).
- Using the income definition of GDP, unit labor costs (ULC) have been a steady driver of final output prices, increasing by about 5.3 percent annually since 2013.
- The gap between the GDP deflator and CPI appears particularly large in Estonia compared with some euro area trading partners.

### VA-REER (value-added real effective exchange rate) misalignment and export effects
- Empirical approach: local projection method à la Jordà (2005) applied to panel data of 27 European countries over 2003–13 to estimate dynamic effects of VA-REER misalignment on real value-added export growth.
- Model controls include inflation, real GDP per capita, net foreign direct investment inflows and external demand; country and time fixed effects are included.
- A 10-percentage point over-valuation in VA-REER leads to:
  - a statistically significant reduction in value-added export growth by 0.8 percentage point in the first year, which cumulates to 1.5 percentage point by the third year.
- Degree of GVC integration matters:
  - Sample median GVC participation index = 69.8.
  - In countries highly integrated into GVCs (index > 69.8), a 10-percentage point VA-REER over-valuation leads to a reduction in VA export growth by 1 percentage point in the first year and cumulates to 1.8 percentage point in the third year.
  - No statistically significant effect is found for countries weakly integrated into GVCs.

### Transmission to real GDP growth
- Estimated relationship: a 1 percentage point increase in real value-added exports is associated with a 0.3 percentage point increase in real GDP growth cumulatively over 4 years.
- Using the estimated trade–growth link, a 10 percent over-valuation in VA-REER could reduce growth rate by 0.5 percentage point (calculated as 0.3 * -1.5).

### Implied impact for Estonia
- Estonia is highly integrated into GVCs.
- Average over-valuation in the VA-REER over the period = 4 percent.
- Implied reduction in growth for Estonia = 0.2 percentage point given the average 4 percent over-valuation.
- Growth would have been higher by 0.2 percentage point on average over the period in Estonia if there were no over-valuation (and thus no such rise in ULC), given the strong pass-through.

### Tariff shocks, GVCs, and short-run value-added effects
- Tariff hikes propagate through global value chains, affecting countries and sectors beyond direct targets and can spill over into services.
- Structural model used: Bems and Johnson (2017) framework applied with WIOD 2016 vintage and World Economic Outlook data to estimate short-run partial-equilibrium effects on gross trade and value-added.
- Estonia’s exposure:
  - VA weights for China, US, and UK are higher than gross trade weights, implying these partners are relatively more important for Estonia’s competitiveness once supply chain linkages are accounted for.
  - A 5.9 percent tariff imposed by the US on its imports (Layer 1), with retaliation from all countries using the same tariff, would lead to a reduction of 0.3 percent in Estonia’s value added (three times larger than the reduction in gross turnover flows).
  - A cumulated tariff shock (all three layers equivalent to a 14.7 percent tariff) would reduce value added produced in Estonia by 0.6 percent.
  - Largest exposure in Europe found in Germany where reduction in domestically produced value-added reaches 0.43 percent.

### Conclusions and policy implications
- The VA-REER index that accounts for input-output linkages suggests more competitive problems for Estonia than a standard REER based on gross trade would imply.
- Recent rise in unit labor cost may have been a drag on Estonia’s ability to supply its domestic value added on world markets.
- Policy recommendations and scopes for improvement:
  - Prevent long-term misalignment between wage growth and productivity to preserve competitiveness.
  - Backward GVC participation: enhance sophistication of production by greater use of imported intermediate goods with high-technological content to benefit from knowledge transfers, diversify exports, and improve product quality.
  - Forward GVC participation: improve allocation and incentives for innovation—through better access to credit and skilled labor with knowledge of foreign markets—to yield productivity gains, particularly for firms operating in upstream GVCs.
  - To mitigate exposure to trade shocks in GVCs, pursue policies aimed at enhancing product sophistication or quality and export market diversification.

*International Monetary Fund — Republic of Estonia, Selected Issues (content unit 1estea2020002, section 14).*

### 1.      Despite progress over the last decade, income inequality and relative poverty remain

### 1.      Despite progress over the last decade, income inequality and relative poverty remain

### Summary of inequality and poverty outcomes
- Income inequality (Gini index, equivalized disposable income)
  - Estonia: 31.6 points in 2017.
  - EA19 average: 30.6 points in 2017.
- At-risk-of-poverty rate (share of population with disposable income lower than 60 percent of median equivalized disposable income)
  - Estonia: 21 percent in 2017.
  - EU27 average: 17 percent in 2017.
- Persistent at-risk-of-poverty (share living below the 60 percent threshold for current year and at least two of the three preceding years)
  - Estonia: 16 percent in 2017.
  - EU28 average: 11 percent in 2017.
  - Estonia women: 18 percent in 2017.
  - EU27 women average: 11.6 percent in 2017.
- Elderly vulnerability
  - Proportion of elderly at-risk-of-poverty reached 41 percent (historical high).
  - Elderly poverty diverged from other groups particularly since 2013.
- Broad income gains amid unequal distribution
  - Real GDP per capita: from euros 7,600 (2000) to euros 15,100 (2018).
  - Average monthly wages growth: 5.9 percent (average over 2010–18).
  - Headcount ratio of extreme poverty (World Bank measure, consumption below $1.90/day 2011 PPP): decreased from 2.5 percent to 0.5 percent in 2015, an 80 percent reduction.
  - Household incomes (growth incidence, 2010–17): average growth 6.1 percent; bottom 10 percent grew 5.6 percent; seventh decile grew 7.2 percent.

### Decomposition methodology and main drivers of inequality
- Theil index definition (used to measure inequality):
  - TT = (1/N) ∑ (yi / ȳ) ln(yi / ȳ)  (equation (1)), where yi and ȳ are equivalized disposable income of household i and the sample mean; N is number of households.
- Theil decomposition across subgroups:
  - T = ∑ s_g T_g + ∑ s_g ln(s_g / p_g)  (equation (2)), where s_g and p_g are income share and household share of subgroup g; first term = within-group inequality; second term = between-group inequality.
- Subgroup classifications used (2010–17):
  - Residence: 16 counties and rural/urban.
  - Gender.
  - Educational attainment: six ISCED-based categories (no formal education/below ISCED1; ISCED1; ISCED2; ISCED3; ISCED4; ISCED5–6).
  - Age cohorts: 15–24, 25–54, 55–64, over 65.
  - Economic status: employee, self-employed, retired, other nonactive.
  - Economic sector of household head: six categories (agriculture, forestry and fishing; industry including energy; construction; wholesale and retail trade; financial and real-estate; public sector and other services).
- Quantified contributions to the recent trend in inequality
  - Economic sector: accounts for 26.7 percent of the recent trend.
  - Economic status (employee, self-employed, retired, other nonactive): contribution similar in magnitude to economic sector; between-group inequality of economic status grew by 43 percent between 2013 and 2010 and by 19 percent between 2010 and 2017.
  - Education: about 14.7 percent of the recent trend.
  - Age cohort: 13.5 percent of the recent trend.
  - County of residence: 8.4 percent of the recent trend.
  - Gender: 5 percent of the recent trend.
  - Rural versus urban difference: very insignificant.

### Gender pay gap (GPG): levels, trends, and drivers
- Unadjusted GPG and female/male ratios (full-time workers, selected points)
  - Panel A: Unadjusted raw GPG at the mean
    - 2010: 25.6%
    - 2012: 30.0%
    - 2013: 31.5%
    - 2014: 26.1%
    - 2015: 25.2%
    - 2016: 24.2%
    - 2017: 18.5%
  - Panel B: Unadjusted female/male log net monthly wage ratio
    - 2010: 77.4%
    - 2012: 74.1%
    - 2013: 73.0%
    - 2014: 77.1%
    - 2015: 77.7%
    - 2016: 78.5%
    - 2017: 83.1%
- Key GPG dynamics and interpretation
  - The unadjusted GPG (at the mean) decreased by 28 percent between 2010 and 2017.
  - Female to male net monthly wage ratio increased from 77 percent (2010) to 83 percent (2017).
  - GPG is on average lower at the bottom of the distribution compared to the top.
  - The rise in minimum income by 9 percent on average since 2012 may have contributed to GPG reduction at the bottom.
- Adjusted GPG (Oaxaca-Blinder and Mincer-type wage regressions)
  - Adjusted metrics: adjusted GPG is lower and adjusted female/male log wage ratio is higher than unadjusted metrics.
  - Over 2010–16, once adjusted, female to male earnings ratio increased by 3 percent while the GPG decreased by 11 percent on average.
  - Estimation approach controls for human capital (education, experience, health), county and rural/urban location, managerial/supervisory position, and disability status.
- Unexplained component
  - On average about 86 percent of the GPG cannot be explained over the period 2010–16.
  - The unexplained part may capture biased practices and/or unmeasured productivity differences.
- Labor market institutions and GPG
  - Evidence consistent with a role for labor market institutions:
    - High minimum wages can reduce the GPG because women are overrepresented among low-paid workers.
    - Collective bargaining coverage can set wage floors and raise wages of low-paid (often female) workers.
  - Low-wage incidence by gender (2014)
    - Share of women low wage earners: 29.4 percent.
    - Share of men low wage earners: 13.9 percent.
    - EU28 averages: 21.2 percent (women) and 13.5 percent (men).
  - Estonia characterized by relatively low minimum wage and low collective bargaining coverage.
  - Long parental leave may influence GPG by keeping women away from work for extended periods.

### Policies to reduce inequality: empirical approach
- Cross-country fixed effects model specification (used to evaluate role of fiscal policy):
  - Gini_c,t = δ PolicyVars_{c,t-1} + β X_{c,t} + α_c + τ_t + ε_{c,t}  (equation (3)), where:
    - Dependent variable: Gini index of equivalized disposable income of country c in year t (Gini_c,t).
    - Policy variables (lagged): public social protection expenditure as share of GDP and property tax revenues as share of GDP.
    - X_{c,t}: control variables capturing structural characteristics and nonlinear forms.
    - α_c: country fixed effect.
    - τ_t: time fixed effect.
  - Data: 21 European countries over the period 2004–17.

*Prepared by Kodjovi Eklou; content based on Estonia Social Survey, Statistics Estonia, Eurostat, and IMF staff calculations.*

### 17.      We estimate the impact of policies

### 17.      We estimate the impact of policies

### Estimation approach and model selection
- Specification: nonlinear forms of structural variables to capture trend of inequality in Estonia.
- Two versions of equation (3) were estimated; model 2 selected as it "captures better the trend in the Gini index (model 2)."
- Model features:
  - Accounts for nonlinearities of control variables.
  - Includes lagged policy variables to mitigate potential endogeneity issues.
  - Sample size determined by data availability.
  - Dynamic panel models and models including lagged control variables were estimated, but the presented models performed better.
- Dependent variable: the Gini index in equivalized disposable income.

### Policy impact estimates (predicted effects on Gini index)
- Method: use estimated econometric model to predict the Gini index and estimate impact of "desirable" levels of policy variables (defined as values at the EU28 average), keeping other variables constant.
- Predicted impacts if Estonia’s policy variables were at the EU28 average over the period:
  - Increasing property tax revenues to the EU28 average would have led to Gini index lower by 3 percent.
  - Increasing social protection spending to the EU28 average would have led to Gini index lower by 11.3 percent.
- Contextual numbers cited:
  - Property tax revenues as share of GDP: Estonia 0.3 percent; EU 1.7 percent.
  - Social protection spending as share of GDP: Estonia 16 percent; EU 23.9 percent.
  - Note on property tax revenues: taken from the Global Revenue Statistics of the OECD and include components such as recurrent taxes on immovable property, on net wealth, estate inheritance and gift taxes.
  - Social protection spending data: from Eurostat.

### Main findings on inequality drivers and distributional issues
- Recent developments:
  - Estonian households experienced considerable income growth, especially at the upper part of the income distribution.
  - Income inequality remains elevated with two main aspects: the gender pay gap and old age poverty.
- Key drivers of recent inequality trend:
  - Income differences by economic sector of activity, economic status, age cohort, and level of education account mainly for the recent trend in inequality.
  - Income inequality between rural and urban areas has reduced recently.
- Gender pay gap (GPG):
  - Remains elevated and is mainly driven by the top of the wage distribution.
  - A significant portion of the GPG cannot be explained by the data.
  - Potential institutional contributors: low collective bargaining power, relatively low minimum wage, long parental leave that may keep women durably from the labor market.
  - Recommended policies to address GPG:
    - (i) Extend transparent reporting policy of gender pay gap (already applied by authorities in some public entities) to other sectors.
    - (ii) Reduce occupational gender biases and career interruptions by women to help reduce the GPG without raising female unemployment.
- Other policy options to reduce inequality:
  - Increasing social protection spending, including pensions for the old age population, to widen the social safety net.
  - Broadening the tax base through an "intelligent system of property tax" to contribute to reducing income inequality.
  - Given Estonia’s relatively low social protection spending compared with peers, increasing social protection bears potential to reduce income inequality.

### Appendix II — Estimation results (selected coefficients and statistics)
- Model (1) and Model (2) coefficient estimates (robust standard errors in parentheses; significance indicated):
  - Property Tax Revenues/GDP (t-1):
    - (1): -0.7709*** (0.2683)
    - (2): -0.5653** (0.2480)
  - Social Protection Spending/GDP (t-1):
    - (1): -0.3829*** (0.0789)
    - (2): -0.4574*** (0.0773)
  - Share of Employment in the Service Sector:
    - (1): -0.2980** (0.1235)
    - (2): 0.7099 (1.0683)
  - Share of Employment in Industry:
    - (1): -0.3834** (0.1520)
    - (2): -1.4193*** (0.4847)
  - Share of Urban Population:
    - (1): -0.2312** (0.1060)
    - (2): -1.3987*** (0.5219)
  - Share of the Population with Tertiary Education:
    - (1): -0.1202*** (0.0380)
    - (2): -0.2172 (0.1441)
  - Log Real GDP per Capita:
    - (1): -6.6083*** (2.3794)
    - (2): -7.6205*** (2.1407)
  - Unemployment Rate:
    - (1): 0.1242** (0.0533)
    - (2): 0.0432 (0.0603)
  - Output gap:
    - (1): 0.0526 (0.0576)
    - (2): -0.0148 (0.0609)
  - Nonlinear terms included in Model (2):
    - Share of Employment in the Service Sector^2: -0.0119 (0.0165)
    - Share of Employment in Industry^2: 0.0274** (0.0121)
    - Share of Urban Population^2: 0.0074** (0.0033)
    - Share of the Population with Tertiary Education^2: 0.0016 (0.0026)
- Model fit and sample:
  - Observations: 285 (for both models)
  - Adjusted R-squared:
    - Model (1): 0.904
    - Model (2): 0.913
  - Country Fixed Effects: Yes
  - Year Fixed Effects: Yes
  - Significance notation: * p<0.10, ** p<0.05, *** p<0.01

### Policy implications and priorities
- Short-term and medium-term priorities implied by findings:
  - Increase social protection spending (including pensions for the old) to strengthen the social safety net and reduce inequality.
  - Consider broadening the tax base via property taxation to reduce inequality; moving property tax revenues closer to EU28 average could yield measurable Gini reductions.
  - Enhance transparency and targeted measures to reduce the gender pay gap, including wider mandatory or encouraged reporting and measures to reduce occupational segregation and career interruptions for women.
  - Use renewed national and sectoral ML/TF risk assessment (mentioned elsewhere in the document) and data-driven monitoring to better target policy measures where they most reduce inequality.

*International Monetary Fund — Republic of Estonia (excerpt).*

### 6.     The EFSA’s understanding of the main ML/TF risks facing Estonia’s financial sector is

### 6. The EFSA’s understanding of the main ML/TF risks facing Estonia’s financial sector

### Data collection and off-site monitoring
- The EFSA’s understanding of ML/TF risks is based largely on information provided by supervised entities via an annual questionnaire that collects hundreds of data points relevant to ML/CFT threats and FIs’ AML/CFT systems and controls.
- The data primarily covers FIs’ engagement with countries and customers viewed as higher-risk, with a focus on non-residents and cross-border payments.
- Data on high risk customers is based on FIs’ own classification of risk; EFSA’s understanding can be refined by collecting data on types of customers EFSA perceives as higher risk (e.g. virtual asset service providers, payment service providers, Estonian e-residents).
- Annual reporting is supplemented with a monthly analysis of AML/CFT-relevant data from the EFSA’s Prudential Division, mostly on the composition of deposits of non-resident and offshore customers.
- Since 2016, EFSA has employed a tool to assess and quantify FI-specific ML/TF threats and vulnerabilities and to generate risk profiles using annual questionnaire information.
  - Vulnerability inputs: desk-based assessment of AML/CFT systems and controls, including corporate governance, risk management, internal controls, and suspicious transaction reporting procedures.
  - Threat inputs: focus on cross-border payments and deposits by offshore, politically exposed and other potentially higher-risk clients.
- The EFSA has a risk profile and quantification of total ML/TF risk for each FI and has submitted 84 inquiries to supervised entities over the last 4 years.

### On-site inspections: approach, scope, and capacity constraints
- EFSA conducts comprehensive/"full-scope" AML/CFT on-site inspections focusing almost exclusively on FIs it considers higher risk.
- Full-scope inspection scope includes:
  - analysis of FIs’ risk appetites and understanding of ML/TF risks;
  - review of structure and functioning of AML/CFT systems and controls;
  - sampling of customer files and testing of customer due diligence and other AML/CFT measures;
  - examination of suspicious transaction or activity reporting to the FIU;
  - interviews with staff at various levels.
- Typical resource and timing characteristics:
  - EFSA usually spends from one to two months on-site.
  - On-site team involves, on average, around 3 full-time employees (FTEs).
  - Final report available at the latest six months after the start of a full-scope on-site inspection.
- Full-scope inspections have uncovered significant violations (examples include the Estonian branch of Danske Bank and Versobank) but are resource-intensive.
  - At current staffing levels, full-scope inspections effectively limit the number of FIs that may be inspected to a maximum of five each year.
- Annual number of on-site inspections is low:
  - From 2014 to 2018, EFSA conducted 29 on-site inspections, including 7 on-site inspections of banks over the last 3 years.
- Consequences of exclusive reliance on full-scope inspections:
  - Lower-risk FIs are rarely if ever subject to on-site inspection.
  - Medium-risk FIs, including large banks, may have years-long supervisory cycles.
  - Preference for full-scope inspections is partly driven by large gaps between inspections, which are themselves a result of conducting only full-scope inspections.

### Options to increase coverage and improve inspection model
- Two principal ways to increase coverage of FIs:
  - Substantial increase in AML/CFT Division staff.
  - Development of risk-based targeted and thematic inspections.
- Benefits of targeted and thematic inspections:
  - Shorter, focused inspections targeting elevated-risk aspects within an FI (specific customers, transactions, services, or compliance gaps) or sector-wide categories.
  - Enables more dynamic response to uncovered or rapidly evolving ML/TF risks, trends and methods.
  - EFSA’s well-developed off-site monitoring and historic full-scope inspection data would facilitate strategic, efficient targeted/thematic inspections.

### Supervisory resources, scope, and workload
- The EFSA established a dedicated Department for AML/CFT supervision, operational in 2019.
  - AML/CFT Department currently consists of 7 persons (7.5 percent of the total staff), up from 4 persons in 2018.
  - Department mix: lawyers and analysts.
  - AML/CFT Department also responsible for non-AML/CFT supervision of payment service providers with an EFSA-estimated workload of 0.5 FTE.
- EFSA supervision coverage:
  - EFSA supervises 110 entities for compliance with AML/CFT obligations, including:
    - 17 credit institutions (9 banks and 8 branches of foreign banks)
    - 15 fund management companies
    - 11 payment service providers
    - 5 life insurance companies
    - 5 investment firms
    - 57 consumer credit loan providers (considered by EFSA to pose lower ML/TF risk)
- Financial sector asset statistics:
  - Total assets held by financial institutions: EUR 45 billion
  - 83 percent of assets are held by the banking sector.
  - In the banking sector, three largest banks account for more than 84 percent of the assets.
- Resource implications:
  - AML Division would need additional human resources to continue prioritizing full-scope inspections while shortening supervisory cycles for low- and medium-risk institutions and responding to emerging risks.
  - EFSA noted scrutiny of recently revealed ML cases has taxed resources but believes it can conduct robust AML/CFT supervision with current staff, asserting ML/TF risks are decreasing while “legacy” cases are getting resolved.

### Procedures, sampling methodology, and enforcement powers
- Off-site monitoring and on-site inspection routines:
  - Established routines for processing and analysis of annual questionnaire data, classification of FIs by risk level, and generation of institutional risk profiles.
  - AML/CFT inspection manual guides inspectors on preparation, staff interviews, customer file sampling, and grounds for additional scrutiny.
  - Staff assignments ensure a range of experience levels and competencies on each on-site team.
  - EFSA can conduct unannounced inspections and has done so where there were grounds to believe a supervised entity might obscure or hamper collection of critical information.
- Current sampling methodology:
  - AML/CFT Department tests implementation by sampling customer files prior to and during on-site inspections, examining customer due diligence at onboarding and ongoing monitoring vs. transaction patterns.
  - EFSA will request additional customer files if it detects unusual transactions or transactions with no economic or legal justification.
  - Methodology would benefit from introducing risk-based random transaction sampling and suspicious transaction reporting sample testing.
- Sampling methodology refinements recommended:
  - Current sampling focuses only on customers presenting certain specific risk factors.
  - Recommendations include introducing random selection of additional customer files; adjusting number of files sampled depending on total number in category to achieve consistent confidence levels; and collecting additional samples during on-site inspections as necessary.
- Enforcement framework and sanctions:
  - Range of sanctions available: informal warnings, letters, administrative precepts, fines (via the misdemeanor process), and license withdrawals.
  - Prudential supervisors incorporate ML/TF risk considerations in Supervisory Review and Evaluation Process.
  - EFSA has used administrative precepts to restrict business activities or force remediation by a certain date; penalty payments for non-compliance with precepts:
    - up to EUR 32 000 for the first specified time increment (e.g. a day) past the due date,
    - up to EUR 100 000 for each consecutive time increment,
    - and up to a total of 5 000 000.
  - Notable enforcement actions cited:
    - Precept to the Estonian branch of Danske bank to prohibit its non-resident business activity following the 2014 on-site inspection; February 2019 precept requiring Danske bank to terminate its activities in Estonia.
    - October 2019: EFSA opened a misdemeanor case following an on-site inspection of the Estonian subsidiary of Swedbank.
    - March 2018: European Central Bank, on request of the EFSA, withdrew the banking license of Versobank due to longstanding and serious AML/CFT breaches.
    - EFSA withdrew licenses of three payment service providers for AML/CFT breaches in 2019.
- Limitations of fines and recommendation for reform:
  - Fines are available only under the misdemeanor framework regulated by the Penal and Criminal Procedure Codes, which:
    - requires EFSA to establish culpability of a specific natural person for the act and prove it was committed in interest of the legal person;
    - includes a two-year limitations period between the violation and entry of final judgement after which a misdemeanor expires (this differs from other jurisdictions where the period may end with filing of a case).
  - These procedural and evidentiary requirements make timely and effective imposition of fines nearly insurmountable in complex cases.
  - Maximum fine per misdemeanor was increased in November 2017 from EUR 32 000 to EUR 400 000.
  - Legislation being drafted would increase the maximum fine to EUR 5 million or up to a certain percentage of revenue.
  - Estonia should streamline and simplify the process of imposing fines by reintroducing an administrative sanctions regime for AML/CFT violations and/or establishing the direct criminal liability of legal persons while extending the applicable statute of limitations.

### Domestic and international cooperation
- AML/CFT Committee:
  - MLTFPA established a governmental AML/CFT Committee, chaired by the Minister of Finance and consisting of 14 other high-level members.
  - Core functions: coordination of the NRA, preparation and monitoring of implementation of an action plan to address findings of the NRA, and developing AML/CFT policies.
  - Committee usually holds high-level quarterly meetings and plays an important role in AML/CFT-related legislative initiatives.
  - Two ad hoc expert-level committees analyzed the AML/CFT institutional framework and supervisory and sanctioning powers; they were dissolved in June 2019 after completing their work.
- Need for intensified, continuous domestic cooperation:
  - Little infrastructure exists for continuous cooperation between EFIU, EFSA, Eesti Pank, Estonian Internal Security Service, Prosecutor General and other law enforcement agencies on operational issues and generation of information on new and evolving ML/TF risks.
  - EFIU also has responsibilities in AML/CFT supervision of the financial sector (licensing and supervision of virtual asset service providers, supervision of currency exchange services, and implementation of targeted financial sanctions).
  - Successful completion of the new NRA will require close cooperation among relevant authorities and continuation/extension of cooperation to the operational level.
- EFSA–EFIU cooperation:
  - Bilateral lines of communication established but generally ad hoc.
  - Potential for closer cooperation, notably in advance of on-site inspections, so EFSA could benefit from EFIU input on the nature, relevance and quality of suspicious transaction reports and notifications about changes in supervised entities’ risk appetite or reporting patterns.
  - EFIU generally responds to EFSA requests and on some occasions proactively communicated concerns.
  - Government initiative to create a Center for Strategic Analysis within the EFIU (adopted in October 2019 and not yet operational) could produce useful risk analysis on evolving ML/TF trends, methods and risks for key authorities, including the EFSA, and for the private sector.

### Recommendations (selected)
- Increase the number and range of financial institutions subject to on-site inspections each year by developing risk-based targeted and thematic on-site inspections or considering an increase in the number of dedicated AML/CFT staff in the EFSA. Review the current on-site inspection model to ensure an appropriate distribution of resources across all sectors, commensurate with the associated levels of ML/TF risk.
- Under the leadership of the EFIU, conduct comprehensive sectoral risk assessments on a regular basis and incorporate the results into on-site inspection schedules and dialogues with the private sector.
- Develop baseline internal procedures for addressing identified violations and following up on previous enforcement actions.
- Develop suspicious transaction reporting sample testing.
- Refine the EFSA’s customer due diligence sampling methodology by:
  - (i) developing independent transaction sampling;
  - (ii) introducing an element of randomness (sampling of additional, randomly selected customer files);
  - (iii) varying the number of records requested in accordance with relative risk and the number of total records in the relevant category; and
  - (iv) requesting, as necessary and appropriate, additional samples during the inspection.
- Streamline and simplify the process of imposing monetary penalties on financial institutions.
- Increase the maximum monetary penalty that may be imposed for violations of AML/CFT-related requirements.

*IMF staff summary of chapter 6 from the referenced EFSA assessment.*

### 24.     The EFSA has intensified its outreach and provision of guidance to the private sector.

### 1estea2020002 - 24.     The EFSA has intensified its outreach and provision of guidance to the private sector.

### EFSA outreach and guidance to the private sector
- The EFSA issued detailed guidelines to its supervised entities in November 2018 covering:
  - risk management, due diligence measures, record-keeping, the refusal to establish (and the termination of) business relationships, the reporting of suspicious transactions.
  - information on ML/TF risk factors and methods relevant to Estonia.
- Private sector feedback:
  - Banks report supervisory expectations—particularly regarding non-resident customers expected to have a valid connection to Estonia—are clearly communicated.
  - Financial institutions have noted the proactive role of the EFSA and its contribution to AML/CFT discussions.
- EFSA engagement activities:
  - Regular bilateral meetings with the private sector.
  - Organization of trainings.
  - Active involvement in the work of the Estonian Banking Association, particularly in meetings for compliance personnel.
  - Informing supervised entities of conclusions from off-site monitoring, including main risks and trends observed.

### Framework and scope for AML/CFT cooperation
- Legal and institutional cooperation framework:
  - All foreign banks operating in Estonia are EU-based and all branches and subsidiaries of Estonian banks are currently located within the EU, giving the EFSA a sufficient legal and institutional framework to cooperate and exchange information with primary foreign counterparts.
  - The EFSA has concluded AML/CFT-specific memoranda of understanding (MOUs) with two non-EU supervisory authorities – in Russia and Switzerland.
  - While additional AML/CFT MOUs with non-EU financial sector supervisors could be useful, there is no indication that current formal cooperative arrangements are insufficient.
- Nordic-Baltic cooperation:
  - The EFSA was a founding member of the permanent working group of Nordic-Baltic financial sector supervisors created in May 2019 to strengthen exchange of information and coordination of AML/CFT supervisory activities.
  - The EFSA is a member of multinational supervisory colleges for four banks with significant presence in Estonia and operating throughout the Nordic-Baltic region.
  - Ad hoc cooperation and coordination with counterparts is pursued as appropriate.
- Scope for greater international cooperation:
  - EFSA and Swedish FSA already share inspection information for banks operating in both countries.
  - Participation in foreign authorities’ on-site inspections of parent banks with branches or subsidiaries in Estonia (e.g., Finland, Latvia, and Sweden) could yield insights into corporate compliance culture, management awareness, and reporting lines on ML/TF risk.
  - Foreign inspector participation in key EFSA on-site inspections could provide similar value.
  - Options for formal integration of AML/CFT supervision at the EU or Nordic-Baltic levels (e.g., consolidation of off-site supervision, integration under the EU’s Enhanced Cooperation procedure, or full integration within the EU) are met with ambivalence in Estonia, driven by a perception that EFSA’s AML/CFT supervision is increasingly sophisticated and effective and concern that a European-level supervisor might be less attuned to Estonia’s specific risk and context.

### Recommendations (from staff)
- To promote effective international and domestic cooperation on AML/CFT supervision of the banking sector, staff recommends that the Estonian authorities:
  - Upon request of the EFSA, the EFIU should provide assessments of the nature, quantity, quality, and pertinence of financial institutions’ suspicious transaction reporting.
  - Fully operationalize and staff the recently approved Center for Strategic Analysis within the EFIU, ensuring that it has the range of expertise and information necessary to produce risk analysis for key authorities and the private sector.
  - When helpful and feasible, participate in foreign authorities’ on-site inspections of parent banks with branches or subsidiaries in Estonia. As appropriate, enable foreign inspectors to join EFSA on-site inspections of their banks’ local branches or subsidiaries.
  - Consider supporting further integration/consolidation of AML/CFT supervision at the EU or Nordic-Baltic levels.

### Conclusion and priority actions
- Recent progress:
  - Important progress achieved in recent years lays the foundation for further enhancement of AML/CFT supervision of the financial sector.
  - Estonia has taken several significant steps to bolster its AML/CFT supervision and to strengthen its AML/CFT regime more generally over the last 5 years.
- Authorities should prioritize:
  - (i) streamlining and simplifying the process of imposing fines on financial institutions;
  - (ii) increasing the number and range of financial institutions subject to on-site inspections each year by increasing the number of dedicated AML/CFT staff in the EFSA and/or developing risk-based targeted and thematic on-site inspections;
  - (iii) consider supporting further integration/consolidation of AML/CFT supervision at the EU or Nordic-Baltic levels.

*Source: 1estea2020002 - 24.     The EFSA has intensified its outreach and provision of guidance to the private sector. (https://www.imf.org/-/media/files/publications/cr/2020/english/1estea2020002.pdf).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2020/english/1estea2020002.pdf_