## _wp10247

## Source details

**Canonical URL:** [_wp10247](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10247.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10247.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10247.pdf.json)

---

### Core research focus and scope
- Objective: Stylized analysis of additional external capital that internationally active European banking groups might need if intra-group cross-border transfers are restricted (ring-fencing) following a credit shock in Central, Eastern, and Southern Europe (CESE).
- Data and scope:
  - Bank-level analysis of 25 European cross-border banking groups and 113 CESE subsidiaries in 18 CESE countries.
  - Inclusion criteria for CESE subsidiaries: total assets ≥ $1 billion and parent ownership stake ≥ 20 percent (majority exceed 50 percent).
  - Coverage: sample subsidiaries’ total assets represent, on average, about 50 percent of host country banking system assets; about 60 percent coverage within EU host countries.
- Mechanism analyzed: transfers of excess profits/capital as internal support. Other mechanisms (subordinated debt injections, asset shifting) noted but not empirically analyzed due to data limitations.
- Time horizon for shock: 2009–2010; simulations use actual 2009 data and 2010 projections assuming a slower recovery than April 2010 WEO forecasts.

### Ring-fencing scenarios analyzed
- Four scenarios used in simulations:
  - No ring-fencing: parent profits, subsidiaries’ excess profits, and excess capital buffers re-allocable to cover any subsidiary shortfalls.
  - Partial ring-fencing: parent profits and subsidiaries’ excess profits re-allocable; excess capital buffers not re-allocable.
  - Near-complete ring-fencing: only transfers from the parent to any subsidiary are allowed.
  - Full ring-fencing / Stand-Alone Subsidiarization (SAS): no transfers allowed between any affiliates, including parent to subsidiaries.
- Analysis accounts for parent ownership stakes; branch exposures treated as part of parent banks’ direct cross-border exposures and not estimated at subsidiary level.

### Calibration of shock, NPLs and ROAs
- Baseline scenario:
  - Uses actual 2009 NPLs and 2010 NPLs projected by the Arellano-Bover regression with WEO GDP, CPI and Libor six-month (in €) for 2010.
  - 2010 baseline NPLs broadly in line with GFSR (April 2010).
- Adverse scenario:
  - 2010 GDP growth assumed 2 percentage points lower than April 2010 WEO forecasts; 2010 interest rates assumed 200 basis points higher than in 2009.
  - Characterized as relatively “mild”; slow recovery and high NPLs concentrated in 2009.
- Country-specific NPL medians (Table 5):
  - Baltic countries: 2008 = 3.6; 2009 = 16.4; 2010 Baseline = 15.9; 2010 Adverse = 16.8
  - CEE-3: 2008 = 3.3; 2009 = 5.6; 2010 Baseline = 5.9; 2010 Adverse = 7.0
  - CIS: 2008 = 3.8; 2009 = 9.6; 2010 Baseline = 8.1; 2010 Adverse = 9.1
  - SEE: 2008 = 4.3; 2009 = 6.4; 2010 Baseline = 7.4; 2010 Adverse = 8.3
- ROA calibration and adjustments:
  - Subsidiary 2009–10 profits based on 2008 pre-provision profits scaled by country-level ROA changes from regression.
  - Adjustments to regression ROA estimates: +0.4 for Bulgaria, +0.7 for Romania; −0.5 for Belarus and Slovakia; −0.7 for Bosnia-Herzegovina and Slovenia.
- Country-specific ROA medians (Table 6):
  - Baltic countries: 2008 = 1.2; 2009 = -0.1; 2010 Baseline = 0.0; 2010 Adverse = -0.2
  - CEE-3: 2008 = 1.2; 2009 = 1.1; 2010 Baseline = 1.2; 2010 Adverse = 1.0
  - CIS: 2008 = 1.4; 2009 = 0.5; 2010 Baseline = 1.0; 2010 Adverse = 0.8
  - SEE: 2008 = 1.7; 2009 = 0.9; 2010 Baseline = 1.0; 2010 Adverse = 0.8
- Parent banks’ profits:
  - 2009 net profits are actual or market-consensus estimates.
  - 2010 profits assumed equal to 2009 profits if 2009 profits were positive, and zero otherwise.

### Methodology for assessing capital needs
- Loan loss reserve (LLR) after shock for subsidiary k in country i:
  - Post-shock LLRk,i = NPLk,i * E k,i * LGD i
    - NPLk,i: post-shock NPL ratio (percent of total exposure)
    - E k,i: total exposure (customer loans)
    - LGD i: loss given default for country i
  - Country-level end-2008 NPLs proxy pre-shock bank-level NPLs where bank-level data unavailable.
  - Country-specific LGDs from World Bank Doing Business; downturn LGDs derived with Downturn LGD = 0.08 + 0.92 * Long-term average LGD.
- CAR definitions:
  - Pre-shock CARk,i = Regulatory Capitalk,i / RWAk,i
  - Post-shock CARk,i = (Regulatory Capitalk,i − Additional Provisionsk,i) / (RWAk,i + ΔRWAk,i)
  - Additional Provisionsk,i = max( Post-shock LLRk,i − Pre-shock LLRk,i − Profitk,i, 0 )
  - Under standardized approach, ΔRWAk,i is negative (write-offs); IRB approach considered in sensitivity and yields broadly similar results with slightly higher recapitalization needs.
- Group-level capital need (CN) definitions under four ring-fencing scenarios (Table 7):
  - CN(1) No ring-fencing = sum of capital needs of all CESE subsidiaries − sum of excess profits and capital of all CESE subsidiaries − profits of the parent bank
  - CN(2) Partial ring-fencing = sum of capital needs of all CESE subsidiaries − sum of excess profits of all CESE subsidiaries − profits of the parent bank
  - CN(3) Near-complete ring-fencing = sum of capital needs of all CESE subsidiaries − profits of the parent bank
  - CN(4) SAS = sum of capital needs of all CESE subsidiaries
  - Excess profit: residual profit after a subsidiary covers its own capital needs; excess capital: cushion above regulatory minimum. Excesses and losses computed accounting for parent ownership stakes.
- Direct cross-border exposures:
  - 2009 cross-border losses estimated as direct cross-border exposures * additional NPLs * LGD.
  - Parent banks’ 2009 net profits adjusted by adding back 50 percent of estimated 2009 cross-border losses to actual/estimated 2009 net profits (50 percent provisioning assumption).

### Main quantitative results
- Aggregate and group-level outcomes (indirect exposures via subsidiaries, unless stated):
  - Eight out of 25 banking groups have no CESE-related capital needs (CN(4) = 0).
  - Five out of 25 banking groups have capital needs CN(4) > 10 percent of the banking group’s regulatory capital.
  - Under no ring-fencing, only five out of 25 groups face non-zero capital needs after re-allocation; under SAS, 17 out of 25 face non-zero needs.
  - Example: banking group #24 faces CESE-related capital needs over 20 percent of its regulatory capital under SAS (CN(4)) but zero capital needs under no ring-fencing (CN(1)).
  - When parent bank profits are zero/negative, CN(3) = CN(4).
- Including direct cross-border exposures and branches:
  - Number of banks needing extra capital ranges from 9 (no ring-fencing) to 22 (SAS) when direct exposures included.
- Aggregate impact:
  - Sample banks’ capital needs are 1.5–3 times higher in ring-fencing/SAS scenarios than under no ring-fencing.
  - Restoring post-shock CARs to pre-shock levels raises overall capital needs relative to restoring to regulatory minimums, but the main comparative result holds.
- Robustness:
  - Results robust to methodological variations including post-shock RWA adjustments and to IRB versus standardized approaches.

### Empirical models and key regression estimates
- Panel regression period: annual data for CESE countries for 1999–2009.
- NPL regression (Arellano-Bover preferred specification):
  - NPL (t-1): 0.7198***
  - GDP (t): -0.3576***
  - Interest (t): 0.0714**
  - Constant: 2.1161***
  - Number of Observations: 161
  - Number of Groups: 18
  - Wald Chi2: 424
- ROA regression (Arellano-Bover preferred specification):
  - ROA (t-1): 0.0834
  - GDP (t): 0.0768***
  - Interest (t): -0.0230***
  - NPLs (t): -0.0799***
  - Constant: 1.7181***
  - Number of Observations: 157
  - Number of Groups: 18
- Descriptive statistics (all in percent):
  - NPLs: N = 185; Mean = 7.6; StDev = 7.3; Min. = 0.2; Max. = 35.8
  - GDP growth: N = 198; Mean = 3.0; StDev = 4.9; Min. = -18.5; Max. = 12.2
  - Interest Rate: N = 187; Mean = 10.5; StDev = 12.9; Min. = 1.4; Max. = 93.2
  - ROA: N = 179; Mean = 1.2; StDev = 1.3; Min. = -6.1; Max. = 3.3
- Alternative panel estimates (fixed/random effects, Table A3.1) show:
  - GDP (t): Fixed Effect = -0.1864***; Random Effects = -0.1797***
  - Interest (t): Fixed Effect = 0.2325***; Random Effects = 0.2301***
  - Constant: Fixed Effect = 5.8369***; Random Effects = 5.7222***
  - Number of Observations: 178; Number of Groups: 18; R2 = 0.23

### Loss Given Default (LGD) estimates
- CESE long-term average LGDs range from 51 percent (Lithuania) to 90 percent (Ukraine).
- CESE average long-term LGD: 68.3 percent.
- CESE average downturn LGD: 70.7 percent.
- Downturn LGDs (selected exact values from Table A6.1):
  - Albania: Downturn LGDs 70.0
  - Belarus: Long-term Average LGDs 66.6; Downturn LGDs 69.3
  - Lithuania: Long-term Average LGDs 50.6; Downturn LGDs 54.6
  - Turkey: Long-term Average LGDs 79.8; Downturn LGDs 81.4
  - Ukraine: Long-term Average LGDs 90.9; Downturn LGDs 91.6
  - Average: Long-term Average LGDs 68.3; Downturn LGDs 70.7

### Capital adequacy context (Appendix Table A1.1 and regulatory minima)
- Cross-country average CARs:
  - CAR (sample subsidiaries) = 13.5; CAR (foreign-owned and domestic banks) = 14.7
- Selected country sample subsidiary vs. country average CARs (exact values):
  - Albania: CAR (sample subsidiaries) = 16.4; CAR (foreign-owned and domestic banks) = 17.2
  - Hungary: 9.4; 11.1
  - Poland: 10.9; 10.8
  - Slovenia: 11.7; 11.7
  - Ukraine: 16.1; 14.0
- Regulatory minimum CARs by country (percent):
  - Albania 12
  - Belarus 10
  - Bosnia 12
  - Bulgaria 12
  - Croatia 10
  - Czech Republic 8
  - Estonia 10
  - Hungary 8
  - Latvia 8
  - Lithuania 10
  - Poland 8
  - Romania 10
  - Russia 10
  - Serbia 12
  - Slovakia 8
  - Slovenia 8
  - Turkey 12
  - Ukraine 8

### Policy implications and recommendations
- Main policy conclusion:
  - Cross-border banking groups would require substantially higher capital buffers at parent and/or subsidiary levels if intra-group transfers are limited or prohibited.
- Two alternative policy paths:
  - (i) Establish a credible international resolution and burden-sharing mechanism to reduce incentives for ring-fencing and permit centralized capital and liquidity management aligned with the bank’s business model.
  - (ii) In absence of burden-sharing mechanisms, set higher minimum capital requirements for cross-border banking groups to ensure legal entities within groups hold self-sufficiency capital buffers (potentially favoring SAS-type or branch structures).
- Trade-offs and considerations:
  - Higher minimum capital to guard against ring-fencing could affect group structure, scope/scale of cross-border activities, credit supply, and competition between local and foreign banks.
  - Choice between approaches depends on benefits, costs, and political constraints; European experience indicates international consensus on cross-border resolution and burden-sharing is difficult but ongoing international work (e.g., Financial Stability Board) provides impetus.

*Source: _wp10247 - conclusions that such exercises might yield are likely to be quite similar to the results (IMF staff analysis, source PDF content).*

### References .............................................................................................................

### _wp10247 - References .............................................................................................................

### Tables and Figures (inventory)
- Tables (numbered 1–7) include:
  - 1. Sample Banking Groups and Their CESE Subsidiaries
  - 2. Descriptive Statistics
  - 3. The Dynamic Panel Regression Output for Nonperforming Loans
  - 4. The Dynamic Panel Regression Output for ROAs
  - 5. Country-Specific NPL Assumptions
  - 6. Country-Specific ROA Assumptions
  - 7. Definitions of Capital Needs Under Four Ring-Fencing Scenarios
- Figures (numbered 1–7) include stylized examples of cross-border banking group dynamics, asset and exposure aggregates for CESE subsidiaries and banking groups, and estimated capital needs resulting from a CESE shock (separately for indirect exposures via subsidiaries, for indirect plus direct cross-border exposures, and aggregate capital needs).
- Appendix Tables (A1.1, A2.1, A3.1, A4.1, A5.1, A6.1, A8.1) document country- and bank-level capital adequacy rates, sample description, panel regression outputs, interest rate data sources, regulatory minimum capital requirements, loss given default ratios, and private analysts’ estimates.

### Core research focus and scope
- The paper analyzes costs for internationally active banking groups associated with different restrictions on intra-group cross-border transfers (referred to as “ring-fencing”).
- Objective: to provide a stylized analysis of how much additional external capital might be needed if banking groups are restricted, to different degrees, in their ability to re-allocate funds across jurisdictions following a credit shock affecting lending in a given region (CESE).
- The analysis:
  - Is bank-level and uses data for European banking groups and their subsidiaries in Central, Eastern, and Southern Europe (CESE).
  - Does not estimate group-level potential recapitalization needs under an extreme scenario (as in a stress test); rather, it considers implications of adverse economic conditions under different forms of ring-fencing.
  - Focuses on transfers of excess profits/capital as the mechanism for internal support, while noting other possible mechanisms (capital injections via subordinated debt, shifting assets) but not empirically analyzing them due to lack of publicly available bank-level intra-group data.

### Background findings and empirical context
- Regulatory environment:
  - Several European countries have legal restrictions on intra-group cross-border asset transfers aimed at preventing undue influence by a foreign parent and protecting minority shareholders and creditors (European Commission, 2010).
  - Host countries can ring-fence subsidiaries by targeting subsidiaries’ ability to transfer funds abroad directly or indirectly (e.g., stopping distribution of dividends by all banks during a crisis).
- During 2008–09:
  - Many subsidiaries of European banking groups relied on foreign parents for capital and liquidity support.
  - Ex ante, CESE subsidiaries had expected parental support; evidence: “the average capitalization levels of the foreign-owned subsidiaries in most CESE countries were 1 to 2 percentage points lower than the banking system averages at the outset of the crisis (Appendix 1).”
  - Ex post, parental assistance validated these expectations.
- Regulatory responses during the crisis included recommendations or restrictions on profit distribution by subsidiaries in countries such as Croatia, Poland, and Turkey.

### Theoretical trade-offs and rationale
- Arguments against ring-fencing (favoring centralized structures):
  - Efficiency and financial stability benefits from diversification across country-specific shocks.
  - Centralized cross-border structures enable internal capital markets, providing subsidiaries better access to capital and liquidity than they might have on a stand-alone basis (De Haas and van Lelyveld, 2010).
  - Absence of ring-fencing facilitates diversification across home and host exposures, potentially stabilizing the group against home-country shocks.
- Arguments for ring-fencing:
  - Host regulators may impose ring-fencing to protect domestic banking systems from negative spillovers, or to increase domestic reserves during uncertainty about the magnitude of output collapse and bank losses.
  - Ring-fencing can limit contagion from parent to subsidiaries and supports selective resolution with minimal disruption to the rest of the group (stand-alone subsidiarization—SAS—arguments).
  - Difficulties in resolving cross-border banking groups and lack of burden-sharing agreements during the crisis motivated discussions about promoting greater self-sufficiency of affiliates.

### Analytical approach and limitations
- Cost metric:
  - Measured as the amount of external capital required to cover capital shortfalls faced by affiliates as a result of a credit shock, under varying restrictions on ability to re-allocate excess profits/capital.
- Scope and limitations:
  - Focuses on transfers of excess profits/capital as the internal support mechanism.
  - Does not empirically analyze intra-group lending or asset transfers due to lack of publicly available bank-level data on these flows.
  - Uses panel regression analysis (see Appendix 3) and country-specific assumptions on NPLs and ROAs (Tables 5 and 6) to inform scenarios.
  - Employs estimates of loss given default ratios by country (Appendix 6) and regulatory minimum capital requirements (Appendix 5) in defining capital needs under four ring-fencing scenarios (Table 7).

### Key quantitative detail preserved from the source text
- “the average capitalization levels of the foreign-owned subsidiaries in most CESE countries were 1 to 2 percentage points lower than the banking system averages at the outset of the crisis (Appendix 1).”
- The paper’s inventory includes:
  - 7 main Tables and 7 Figures.
  - 8 Appendixes, with multiple appendix tables (A1.1, A2.1, A3.1, A4.1, A5.1, A6.1, A8.1).

* _wp10247 - References .............................................................................................................*

### conclusions that such exercises might yield are likely to be quite similar to the results

### _wp10247 - conclusions that such exercises might yield are likely to be quite similar to the results

### Ring-fencing scenarios analyzed
- Four ring-fencing scenarios defined and used in simulations:
  - No ring-fencing: parent bank’s profits, subsidiaries’ excess profits, and excess capital buffers can be used to cover capital shortfalls in any subsidiary.
  - Partial ring-fencing: parent bank’s profits and subsidiaries’ excess profits, but not excess capital, can be re-allocated within the group.
  - Near-complete ring-fencing: only transfers from the parent to any subsidiary are allowed.
  - Full ring-fencing / Stand-Alone Subsidiarization (SAS): no transfers between any of the group’s affiliates (including from the parent bank to subsidiaries) are allowed.
- Analysis takes into account parent banks’ ownership stakes in their subsidiaries.
- Branch-level data are not used in the estimation because branches are not stand-alone entities and are harder for host-country authorities to ring-fence; CESE exposures via branches are analyzed as part of total direct cross-border exposures of parent banks.

### Sample and scope
- Sample composition:
  - 25 European cross-border banking groups.
  - 113 CESE subsidiaries located in 18 countries in CESE.
- Inclusion criteria for CESE subsidiaries:
  - Total assets of at least $1 billion.
  - Parent bank ownership stake of at least 20 percent.
  - Note: for the majority of subsidiaries in the sample the ownership stake of the parent bank exceeds 50 percent.
- Coverage:
  - Total assets of the sample subsidiaries represent, on average, about 50 percent of the host country’s total banking system assets.
  - Significant coverage within EU host countries: about 60 percent.

### Calibration of the CESE regional credit shock and simulation design
- Time horizon for the shock: 2009–2010.
- Shock modeled as deterioration in macroeconomic conditions leading to:
  - Increase in nonperforming loans (NPLs).
  - Decrease in return on assets (ROAs) of CESE subsidiaries.
- Simulations rely largely on actual data for 2009 and on projections for 2010 that assume a slower economic recovery than envisaged in April 2010 WEO forecasts.
- Two-step estimation of capital needs:
  1. Subsidiary-level capital need: amount of capital required to bring post-shock CAR back to either the country-specific (Basel II) regulatory minimum or to the subsidiary-specific pre-shock level (the latter is conservative).
  2. Group-level total capital needs: sum of all subsidiaries’ capital needs (and losses on direct cross-border exposures of parent banks, in some simulations) offset by any other funds (excess profits and/or capital) re-allocable within the group; resulting total depends on availability of excess profits/capital and on degree of re-allocation permitted by ring-fencing.

### Main quantitative results
- Capital needs resulting from a simulated credit shock affecting CESE subsidiaries over the 2009–2010 period:
  - Sample banks’ capital needs are 1.5–3 times higher in the ring-fencing/SAS scenarios than under no ring-fencing.
- Robustness:
  - Results are robust to variations in methodology for computing capital needs, including post-shock adjustment in risk-weighted assets (standardized versus Basel II Internal Ratings Based (IRB) approach).

### Policy implications highlighted
- Establishment of a credible framework for resolution of cross-border banking groups would help avoid unilateral and likely more costly solutions (in terms of capital requirements) by reducing incentives for ring-fencing by home/host authorities.
- In absence of resolution and burden-sharing mechanisms, minimum capital requirements for cross-border banking groups should take into account potential presence of ring-fencing, especially in crisis times; such a possibility may push cross-border banks toward organizational structures more immune to ring-fencing (either SAS-type structures or branch structures).
- If regulators promote a SAS-like approach, potential benefits must be weighed against potential costs.

### Country-level data and regression-based shock modeling
- Data sources:
  - NPLs: IMF’s Global Financial Stability Reports (GFSR).
  - GDP: IMF’s World Economic Outlook (WEO) database.
  - Interest rates: T-bill rates where available; Money Market Rates or comparable rates for some countries (see Appendix references).
- Panel regression period: annual data for CESE countries for 1999–2009.
- Variables used in NPL regression: real annual GDP growth rates and short-term nominal interest rates (chosen for reliable predictive power).
- ROA regression explanatory variables: GDP growth, short-term interest rates, and NPL ratios.

### Descriptive statistics for regression variables (all in percent)
- NPLs:
  - N: 185
  - Mean: 7.6
  - StDev: 7.3
  - Min.: 0.2
  - Max.: 35.8
- GDP growth:
  - N: 198
  - Mean: 3.0
  - StDev: 4.9
  - Min.: -18.5
  - Max.: 12.2
- Interest Rate:
  - N: 187
  - Mean: 10.5
  - StDev: 12.9
  - Min.: 1.4
  - Max.: 93.2
- ROA:
  - N: 179
  - Mean: 1.2
  - StDev: 1.3
  - Min.: -6.1
  - Max.: 3.3

### Dynamic panel regression results (preferred specifications)
- NPL model — Arellano-Bover specification:
  - NPL (t-1): 0.7198***
  - GDP (t): -0.3576***
  - Interest (t): 0.0714**
  - Constant: 2.1161***
  - Number of Observations: 161
  - Number of Groups: 18
  - R2: N/A
  - Wald Chi2: 424
- ROA model — Arellano-Bover specification:
  - ROA (t-1): 0.0834
  - GDP (t): 0.0768***
  - Interest (t): -0.0230***
  - NPLs (t): -0.0799***
  - Constant: 1.7181***
  - Number of Observations: 157
  - Number of Groups: 18

*Source: _wp10247 - conclusions that such exercises might yield are likely to be quite similar to the results (IMF staff analysis, source PDF content).*

### 0.52   N/A      N/A

### _wp10247 - 0.52   N/A      N/A

### Calibration of NPLs and ROAs
- Baseline scenario:
  - Uses actual 2009 NPL ratios for CESE banking sectors and 2010 NPL ratios projected using the Arellano-Bover regression estimation, the WEO forecast of GDP, CPI and Libor six-month (in €) for 2010.
  - 2010 baseline NPLs are broadly in line with estimates in the GFSR (April 2010, IMF 2010a).
- Adverse scenario:
  - Uses actual 2009 NPLs and 2010 NPL projections assuming for each CESE country: 2010 GDP growth is 2 percentage points lower than the 2010 April WEO GDP growth rate forecasts and 2010 interest rate is 200 basis points higher than in 2009.
  - Characterized as relatively “mild” among plausible adverse scenarios and features a slow recovery and high NPLs in both years, with most of the NPL increase taking place in 2009.
- Country-specific NPL medians (Table 5):
  - Baltic countries: 2008 = 3.6; 2009 = 16.4; 2010 Baseline = 15.9; 2010 Adverse = 16.8
  - CEE-3: 2008 = 3.3; 2009 = 5.6; 2010 Baseline = 5.9; 2010 Adverse = 7.0
  - CIS: 2008 = 3.8; 2009 = 9.6; 2010 Baseline = 8.1; 2010 Adverse = 9.1
  - SEE: 2008 = 4.3; 2009 = 6.4; 2010 Baseline = 7.4; 2010 Adverse = 8.3
- ROA calibration:
  - For subsidiaries, 2009–10 profits are calculated by taking actual 2008 pre-provision profits of individual subsidiaries as a base and applying the same rate of change as the country-level ROAs from the regression model.
  - Regression-based estimates of ROAs adjusted downward for Slovenia, Slovakia, Belarus and Bosnia-Herzegovina and upward for Romania and Bulgaria (adjustments: +0.4 for Bulgaria, +0.7 for Romania; −0.5 for Belarus and Slovakia, −0.7 for Bosnia-Herzegovina and Slovenia).
- Country-specific ROA medians (Table 6):
  - Baltic countries: 2008 = 1.2; 2009 = -0.1; 2010 Baseline = 0.0; 2010 Adverse = -0.2
  - CEE-3: 2008 = 1.2; 2009 = 1.1; 2010 Baseline = 1.2; 2010 Adverse = 1.0
  - CIS: 2008 = 1.4; 2009 = 0.5; 2010 Baseline = 1.0; 2010 Adverse = 0.8
  - SEE: 2008 = 1.7; 2009 = 0.9; 2010 Baseline = 1.0; 2010 Adverse = 0.8
- Parent banks:
  - 2009 net profits are actual numbers or estimates based on market consensus forecasts.
  - 2010 profits assumed equal to 2009 profits if 2009 profits were positive, and zero otherwise.

### Methodology for Assessing Bank Capital Needs
- Loan loss reserve (LLR) for subsidiary k in country i after credit shock:
  - Post-shock LLRk,i = NPLk,i * E k,i * LGD i
    - NPLk,i is post-shock NPL ratio (nonperforming loans in percent of total exposure).
    - E k,i is total exposure (customer loans).
    - LGD i is loss given default (same for all subsidiaries in country i).
  - Country-level end-2008 NPL ratios used to proxy pre-shock bank-level NPL ratios where bank-level data unavailable.
  - Country-specific LGDs from World Bank Doing Business; downturn LGDs derived using formula proposed by the Federal Reserve Board (2006).
- Capital Adequacy Ratio (CAR) definitions:
  - Pre-shock CARk,i = Regulatory Capitalk,i / RWAk,i
  - Post-shock CARk,i = (Regulatory Capitalk,i − Additional Provisionsk,i) / (RWAk,i + ΔRWAk,i)
  - Additional Provisionsk,i = max( Post-shock LLRk,i − Pre-shock LLRk,i − Profitk,i, 0 )
  - Under standardized approach, ΔRWAk,i is negative and corresponds to written-off losses; IRB approach considered in sensitivity and yields broadly similar results with slightly higher recapitalization needs.
- Capital need (CN) definitions at group level under four ring-fencing scenarios (Table 7):
  - No ring-fencing:
    - CN(1) = sum of capital needs of all CESE subsidiaries − sum of excess profits and capital of all CESE subsidiaries − profits of the parent bank
  - Partial ring-fencing:
    - CN(2) = sum of capital needs of all CESE subsidiaries − sum of excess profits of all CESE subsidiaries − profits of the parent bank
  - Near-complete ring-fencing:
    - CN(3) = sum of capital needs of all CESE subsidiaries − profits of the parent bank
  - Stand-alone subsidiarization (SAS):
    - CN(4) = sum of capital needs of all CESE subsidiaries
  - Excess profit: residual profit of a subsidiary after it covers its own capital needs; excess capital: cushion above regulatory minimum. Excess profits/capital and losses of subsidiaries computed accounting for parent ownership stakes.
- Cross-border direct exposures adjustment:
  - 2009 cross-border losses estimated as direct cross-border exposures * additional NPLs * LGD.
  - Parent banks’ 2009 net profits adjusted by adding back 50 percent of estimated 2009 cross-border losses to actual/estimated 2009 net profits (50 percent provisioning assumption), to avoid double counting due to unavailable provisioning data.

### Results
- Simulations focus first on indirect CESE exposures via subsidiaries, then include direct cross-border exposures, and also test alternative deficit definition restoring pre-shock (end-2008) CARs.
- Key findings (only indirect exposures via subsidiaries; see Figures 5–7):
  - (i) Eight out of 25 banking groups have no capital needs related to their CESE subsidiaries (i.e., CN(4)=0).
  - (ii) Five out of 25 banking groups have significant capital needs related to their CESE subsidiaries (i.e., CN(4) > 10 percent of the banking group’s regulatory capital).
  - Ring-fencing impact:
    - Capital needs to recapitalize groups after shock are higher under near-complete/partial ring-fencing than under no ring-fencing; differences larger for more diversified groups.
    - Example: banking group #24 faces CESE-related capital needs over 20 percent of its regulatory capital under SAS (CN(4)) but has zero capital needs under no ring-fencing (CN(1)).
    - When parent bank profits are zero/negative, CN(3)=CN(4).
    - Under no ring-fencing, only five out of 25 groups face non-zero capital needs after re-allocation; under SAS, 17 out of 25 face non-zero needs.
  - Including direct cross-border exposures and lending via branches increases CN measures (Figure 6):
    - Range: 9 banks (in the no ring-fencing scenario) to 22 banks (in the SAS scenario) need extra capital when direct exposures included.
  - Restoring post-shock CARs to pre-shock levels raises overall capital needs compared to restoring to regulatory minimums, but main result remains: aggregate capital needs under ring-fencing/SAS are much higher than under no ring-fencing (Figure 7).
  - Aggregate capital needs of sample banks resulting from a CESE shock are 1.5–3 times higher in the case of ring-fencing/SAS compared with no ring-fencing.
  - These estimates are within the range of loss/recapitalization estimates provided by private analysts at end-2008 or early 2009 for Eastern Europe downturn scenarios in 2009–10.

### Conclusions and Policy Implications
- Main conclusion:
  - Simulation of capital needs of 25 large European banking groups shows these groups would need substantially higher capital buffers at parent and/or subsidiary levels if unable to transfer capital and/or profits across borders.
  - Material impact on individual groups depends on significance of CESE subsidiaries for each group.
- Two alternative policy paths:
  - (i) Establish a credible international resolution and burden-sharing mechanism to reduce need for ring-fencing and permit centralized capital and liquidity management consistent with the bank’s business model.
  - (ii) In absence of burden-sharing mechanisms, regulators should consider setting higher minimum capital requirements for cross-border banking groups, taking into account the possibility of ring-fencing—ensuring all legally independent parts of the group hold self-sufficiency capital buffers.
- Trade-offs:
  - Higher capital requirements to ensure resilience under ring-fencing could affect banking group structures, scope/scale of cross-border activities, supply of credit, and competition between local and foreign banks.
  - Choice between (i) and (ii) depends on balance of benefits and costs and political constraints; European experience suggests reaching international consensus on cross-border resolution and burden-sharing is challenging but ongoing international work (e.g., Financial Stability Board) provides impetus.

### Appendix — Capital Adequacy Rates by Country and Bank Type (Table A1.1)
- Average CARs of foreign-owned sample subsidiaries vs. country-level average CARs:
  - Albania: CAR (sample subsidiaries) = 16.4; CAR (foreign-owned and domestic banks) = 17.2
  - Belarus: 14.3; 21.8
  - Bosnia: 14.0; 16.3
  - Bulgaria: 13.6; 14.9
  - Croatia: 14.4; 15.4
  - Czech Republic: 11.0; 12.3
  - Estonia: 15.2; 13.3
  - Hungary: 9.4; 11.1
  - Latvia: 12.3; 11.8
  - Lithuania: 11.5; 12.9
  - Poland: 10.9; 10.8
  - Romania: 13.7; 13.8
  - Russia: 15.0; 16.8
  - Serbia: 18.6; 21.9
  - Slovakia: 10.0; 11.1
  - Slovenia: 11.7; 11.7
  - Turkey: 15.8; 18.0
  - Ukraine: 16.1; 14.0
  - Cross-country Average: CAR (sample subsidiaries) = 13.5; CAR (foreign-owned and domestic banks) = 14.7

*Source: Staff estimates as presented in the source document.*

### Appendix 2. Sample of Banking Groups and Their Subsidiaries

### Appendix 2. Sample of Banking Groups and Their Subsidiaries

### Sample composition
- The appendix provides a listing of sample banking groups and their subsidiaries, showing Parent Bank, Home country, Subsidiary, and Host country.
- The sample entries include (selected examples shown in original order):  
  - 1 RZB (Austria) — RAIFFEISEN BANKA DD (Slovenia)  
  - 2 Bank Austria (Austria) — UNICREDIT BANKA SLOVENIJA D.D.  
  - 3 Hypo Alpe Adria Group (Austria) — HYPO ALPE-ADRIA-BANK DD  
  - 4 KBC (Belgium) — NLB DD-NOVA LJUBLJANSKA BANKA  
  - 5 SocGen (France) — SKB BANKA DD  
  - ... (continued through entry 113, including subsidiaries in CESE countries and Russia, Turkey, Ukraine, Belarus, the Baltics, Hungary, Romania, Croatia, Serbia, Bosnia, Bulgaria, Poland, Czech Republic, Slovakia, Slovenia, Estonia, Latvia, Lithuania)  
- Sources for the sample: Bankscope, Bank reports, and Analysts’ reports.

### Key implication
- The sample aims to capture major European cross-border banking groups with significant presence in the CESE region (see Appendices 7 and 8 for treatment of remaining exposures and analysts’ estimates).

### Source attribution
*Sources: Bankscope, Bank reports, and Analysts’ reports.*

---

### Appendix 3. Panel Regression Analysis

### Model specification
- Basic panel regression model estimated with fixed-effects and random-effects:
  - npl_ti = alpha_i + beta_1 * GDPΔ_ti + beta_2 * interest_ti + epsilon_ti
  - npl stands for non-performing loans (as a share of total loans)
  - GDPΔ for real GDP growth on an annual basis
  - interest for the short-term nominal interest rates (T-bill rate)
  - i = 1,2,..N denotes country; t = 1,2,...T denotes period

### Estimation choice
- Based on the Hausman test, the random effects specification is more appropriate than the fixed-effects specification.

### Estimated coefficients (Table A3.1)
- GDP (t):
  - Fixed Effect: -0.1864***  
  - Random Effects: -0.1797***  
- Interest (t):
  - Fixed Effect: 0.2325***  
  - Random Effects: 0.2301***  
- Constant:
  - Fixed Effect: 5.8369***  
  - Random Effects: 5.7222***  
- Number of Observations: 178 (both specifications)
- Number of Groups: 18 (both specifications)
- R2: 0.23 (both specifications)
- Note: ***, **, * denote significance at the 1 percent, 5 percent, and 10 percent level respectively.
- Source: Staff estimates.

### Robustness checks and data issues
- Compared average level of bank provisions to NPLs over time to investigate differences in NPL definitions across countries:
  - For most of the 18 countries, the ratio of provisions to NPLs varies between 60 percent and 90 percent.
  - Estonia: provisions range between 150 percent and 300 percent of NPLs during 2003–06 (data available years), implying particularly low reported NPLs; Estonia excluded from some regressions as robustness check.
  - Russia: provisions fluctuate between 120 percent and 160 percent during 2003–08, suggesting relatively narrow NPL definition.
  - Romania and Ukraine: provisions indicate a fairly broad definition of potential losses.
- For countries with structural breaks (Latvia, Lithuania, Poland, Slovakia, Serbia, Ukraine), checked for major changes in NPL levels around timing of definition breaks:
  - No major changes detected; all NPL data from these countries included in regressions.
  - Regression results remained robust when excluding countries with structural breaks.

---

### Appendix 4. Interest Rates Used for Regression Analysis

### Data sources and construction rules (selected entries)
- Bosnia-Herzegovina: Interest rate constructed as 2/3 times Deposit Rate + 1/3 times Lending Rate, starting in 2002 (other data is not available). Source: MBTS.
- Czech Republic: Treasury Bill Rate. Source: MBTS.
- Latvia: Treasury Bill Rate. Source: MBTS.
- Bulgaria: Treasury Bill Rate when available (2 years), otherwise inferred from Interbank Rate (3 months, Bloomberg), otherwise from Lending Rate. Sources: MBTS and Bloomberg.
- Croatia: Treasury Bill Rate from Bloomberg (from 2002 onwards); before that: adjust Money Market Rate weighted by the relative portion of T-Bill rates on money market rates from 2002 to 2008. Sources: IFTSTSUB (for 1992 to 2001), Bloomberg (from 2002).
- Lithuania: Treasury Bill Rate (12 months) where possible, otherwise Government Bond Yield (12 months, from Bloomberg), recalculated to 12 months level, otherwise Interbank Rate, recalculated to 12-month T-Bill rate. Sources: MBTS and Bloomberg.
- Romania: Interbank short-term lending rate, similar to Treasury Bill Rate (91 days) (which is only available for some years). Source: MBTS.
- Slovakia: Average Deposit Rate (1993–1999), Government Bond Yield (2002-2006), Government Bond Yield (1 yr, from Bloomberg (2007 to 2009)). Sources: MBTS, Bloomberg.
- Slovenia: Treasury Bill Rate (from 1999), before that: Money Market Rate adjusted to Treasury Bill Rate. Source: MBTS.
- Russia: Treasury Bill Rate for 2000–2003, after 2003: avg of money market rate and interbank rate (3 months), before 1999: Money Market Rate. Sources: MBTS, Bloomberg.
- Turkey: Treasury Bill Rate, if available; (before 1999: Money Market Rate), after 2007: 1-yr government bond yield. Sources: IFTSTSUB, Bloomberg.
- Hungary: Treasury Bill Rate. Source: MBTS.
- Poland: Weighted Average Treasury Bill Rate. Source: MBTS.
- Serbia: Money Market Rate. Source: IFTS.
- Ukraine: Money Market Rate. Source: IFTS.
- Albania: Treasury Bill Rate. Source: IFTS.
- Estonia: Money Market Rate. Source: IFTS.
- Belarus: Data are scarce; adjusted deposit rates according to money market rates to deposit rates in Ukraine. Source: IFTS.

### Aggregate source note
- Sources: International Financial Statistics (IFTS), Money and Banking Statistics and Bloomberg.

---

### Appendix 5. Regulatory Minimum Capital Requirements by Country

### Minimum Capital Adequacy Requirements (CARs) (in percent)
- Albania 12  
- Belarus 10  
- Bosnia 12  
- Bulgaria 12  
- Croatia 10  
- Czech Republic 8  
- Estonia 10  
- Hungary 8  
- Latvia 8  
- Lithuania 10  
- Poland 8  
- Romania 10  
- Russia 10  
- Serbia 12  
- Slovakia 8  
- Slovenia 8  
- Turkey 12  
- Ukraine 8

### Source
- Source: IMF country desks.

---

### Appendix 6. Loss Given Default Ratios in the CESE Countries

### Methodology
- Country-specific Loss Given Default Ratios (LGDs) taken from the World Bank Doing Business webpage based on Djankov et al. (2008).
- Downturn LGD computed using formula from the Federal Reserve Board (2006):  
  - Downturn LGD = 0.08 + 0.92 * Long-term average LGD
- CESE long-term average LGDs range from 51 percent (Lithuania) to 90 percent (Ukraine).
- CESE average long-term LGD: 68.3 percent.
- CESE average downturn LGD: 70.7 percent.
- CESE downturn LGDs range from 55 percent (Lithuania) to 92 percent (Ukraine).

### Table A6.1 — Long-term Average LGDs and Downturn LGDs (exact values)
- Albania 1/: Long-term Average LGDs N/A; Downturn LGDs 70.0  
- Belarus: Long-term Average LGDs 66.6; Downturn LGDs 69.3  
- Bosnia: Long-term Average LGDs 64.1; Downturn LGDs 67.0  
- Bulgaria: Long-term Average LGDs 67.9; Downturn LGDs 70.5  
- Croatia: Long-term Average LGDs 69.5; Downturn LGDs 71.9  
- Czech Republic 2/: Long-term Average LGDs 41.3; Downturn LGDs 46.0  
- Estonia: Long-term Average LGDs 62.5; Downturn LGDs 65.5  
- Hungary: Long-term Average LGDs 61.6; Downturn LGDs 64.7  
- Latvia: Long-term Average LGDs 71.0; Downturn LGDs 73.3  
- Lithuania: Long-term Average LGDs 50.6; Downturn LGDs 54.6  
- Poland: Long-term Average LGDs 70.2; Downturn LGDs 72.6  
- Romania: Long-term Average LGDs 71.5; Downturn LGDs 73.8  
- Russia: Long-term Average LGDs 71.8; Downturn LGDs 74.1  
- Serbia: Long-term Average LGDs 74.6; Downturn LGDs 76.6  
- Slovakia: Long-term Average LGDs 54.1; Downturn LGDs 57.8  
- Slovenia: Long-term Average LGDs 54.5; Downturn LGDs 58.1  
- Turkey: Long-term Average LGDs 79.8; Downturn LGDs 81.4  
- Ukraine: Long-term Average LGDs 90.9; Downturn LGDs 91.6  
- Average: Long-term Average LGDs 68.3; Downturn LGDs 70.7

### Notes
- 1/ For Albania, the downturn LGD is assumed to be equal to its peer country average (70 percent).  
- 2/ For the Czech Republic, the average LGD is the one for corporate credit published by the Czech National Bank (CNB 2009, p.81).

### Comparison and implication
- The estimated CESE downturn LGDs are substantially higher than those observed for senior bank loans in OECD countries, which range from 30 percent and 45 percent (see Moody’s 2010, Doing Business database).

---

### Appendix 7. Using BIS Data to Capture Remaining Banking Groups’ Exposures to the CESE Countries

### Purpose and approach
- Subsidiaries’ balance sheet data do not capture full CESE exposures of sample banking groups because they miss:
  - Direct cross-border claims of parents
  - Lending through branches in host countries
- To capture these exposures, combine subsidiaries’ balance sheet data with consolidated BIS international banking statistics (country-level consolidated claims of BIS reporting banks on each CESE country).

### Adjustment method
- BIS total foreign claims (immediate borrower’s basis) include cross-border claims and all foreign affiliates’ claims.
- To separate direct cross-border and branches’ claims:
  - Subtract domestic lending to customers by sample subsidiaries from BIS total foreign claims (example: subtract customer lending by Greek-controlled subsidiaries operating in Turkey from consolidated foreign claims of BIS reporting Greek banks on Turkish residents).

### Caveats and data limitations
- (i) Residual labeled as “direct cross-border and branches claims” could include non-lending claims by subsidiaries (e.g., government bonds held by a subsidiary). Potential recapitalization needs were calculated based on lending portfolio; treating non-lending claims as parent-linked claims is deemed appropriate for this purpose.
- (ii) Discrepancies exist between BIS data and other sources:
  - Potential small downward bias in BIS total claims of Italian banks on Ukraine and Swedish banks on Latvia, based on subsidiaries balance sheet data.
  - Discrepancies may arise from differences in consolidation methods and group structure classification.
- (iii) Residual exposure at a parent bank’s home country level is distributed across sample banking groups of each home country proportionally to the group’s assets.
  - Given the sample was constructed to capture all major European cross-border banking groups with significant presence in CESE, it is unlikely BIS includes large cross-border banks with CESE operations not already in the sample.

---

### Appendix 8. Analysts’ Estimates of Losses and Capital Needs

### Content note
- Appendix contains Table A8.1 titled: Private Analysts’ Estimates of Losses and Capital Needs of Western European Banks on their CESE Exposures.
- (Table content referenced but not reproduced in this extract.)

---

*Sources: Bankscope, Bank reports, and Analysts’ reports.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10247.pdf_
