## _wp13177 - 4.      An overall relative health score for each bank at a particular point in time can be

## Source details

**Canonical URL:** [_wp13177 - 4.      An overall relative health score for each bank at a particular point in time can be](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13177.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13177.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13177.pdf.json)

---

### BHI definition and formula
- The overall relative health score for each bank at time t, denoted ݖ௕௔௡௞,௧, is estimated by summing the z-scores for each financial ratio:
  - ݖ௕௔௡௞,௧ = ݖ௖௔௣௜௧௔௟,௧ + ݖ௔௦௦௘௧,௧ + ݖ௘௔௥௡௜௡௚௦,௧ + ݖ௟௜௤௨௜ௗ௜௧௬,௧ + ݖ௟௘௩௘௥௔௚௘,௧
- Component definitions:
  - ݖ௖௔௣௜௧௔௟,௧ = z-score for capital adequacy at time t
  - ݖ௔௦௦௘௧,௧ = z-score for asset quality at time t
  - ݖ௘௔௥௡௜௡௚௦,௧ = z-score for earnings at time t
  - ݖ௟௜௤௨௜ௗ௜௧௬,௧ = z-score for liquidity at time t
  - ݖ௟௘௩௘௥௔௚௘,௧ = z-score for leverage at time t
- Interpretation:
  - ݖ௕௔௡௞ represents a relative overall measure of health for a particular bank in the defined peer sample.
  - The combination of ݖ௕௔௡௞ scores for a sample constitutes the BHI for that sample.
- Normalization:
  - Each financial ratio is normalized to a z-score.
  - Notation: z_{i,t} is the normalized (or z-score for) the financial ratio of bank i at time t; x_{i,t} is the financial ratio of bank i at time t; ߤ is the system mean over the three periods to time t;  is the system standard deviation over the three periods to time t.
  - System means and standard deviations are calculated over three periods to incorporate time and cross-sectional dimensions and to capture deterioration during the crisis period.
- Treatment of NPL ratio:
  - For the asset quality measure (NPL ratio), the NPL ratio z-score is multiplied by -1 so that increases in that score are represented as increasingly negative developments.

### HEAT! tool and visualization conventions
- HEAT! is an Excel-based tool that:
  - downloads requisite data for individual banks from BankScope, Bloomberg (for listed banks only) and SNL;
  - automatically calculates corresponding z-scores (Appendix I);
  - automatically generates heatmaps of the BHI to visually differentiate overall relative soundness and the individual constituent components for a period and over time.
- Heatmap color coding:
  - institutions in the top 90th percentile of soundness are denoted in green;
  - those in the bottom 10th percentile are denoted in red;
  - the rest are shown in various shades of yellow/orange.

### III. Application of the BHI and HEAT! — Key findings and demonstrations
- Purpose: Demonstrate BHI and HEAT! outputs using well-known peer groups and a domestic back-test; anonymize banks to avoid implicating specific institutions.
- A. International Peer Group Example (G-SIBs, data to December 2012)
  - Sample: global systemically important banks (G-SIBs) identified by BCBS (2011) and listed by the FSB (2012), consisting of 28 banks (anonymized; buckets 3 and 4 combined).
  - Illustrative findings:
    - G-SIBs have improved their earnings performance relative to the “trough” in 2008.
    - The majority of banks appear to have deleveraged since 2008 as evidenced by improved leverage ratios.
  - Bucket-differentiated observations:
    - A smattering of G-SIBs across buckets exhibit distinct overall weakness relative to peers.
    - Bucket 1: relatively lower capital adequacy, less liquid, more leveraged.
    - Bucket 2: relatively weaker earnings and remain more leveraged.
    - Bucket 3+4: asset quality and leverage are the main weaknesses.
- B. Back-testing: Spain Case Study (end-2011 data)
  - Context and reforms since the global financial crisis:
    - Mergers of savings banks into larger commercial banks and takeovers during the crisis; in some cases, several weak entities were merged into larger weak ones.
    - Authorities commissioned stress tests, audits of loan portfolios and asset valuations by third parties following the IMF’s FSAP Update assessment in 2012 H1.
    - Since 2012 Q3, per Spain’s Memorandum of Understanding (MoU) with the Eurogroup, restructuring included recapitalization and transfer of real estate loans to an asset management company.
  - Back-test findings (end-2011 BHI heatmap; results align with FSAP (IMF, 2012) and Oliver Wyman (2012) assessments):
    - The three largest institutions (Santander, BBVA, Caixa) are among the healthiest and most resilient within the Spanish system.
    - Smaller banks identified as not requiring additional capital even under severe stress (e.g., Kutxa and Unicaja) appear very sound according to the BHI.
    - The fourth largest bank (BFA-Bankia) undergoing restructuring and identified as requiring significant recapitalization is the result of a 2009 merger of several already-weak entities.
    - Banks considered non-viable and taken over by the Fund for Orderly Bank Restructuring (FROB) are among the weakest (examples: Unnim and CAM sold to other banks with financial support; Catalunya Caixa and Nova Galicia Caixa require recapitalization).
  - Conclusion from the back-test:
    - The BHI effectively differentiates relative soundness within a system—identifying problem banks prior to the MoU and highlighting healthier banks.
    - Relative nature emphasized: Spanish G-SIBs rank among the healthiest domestically but are about average relative to G-SIB peers.

### IV. Main caveats when using the BHI and HEAT!
- Aggregation:
  - The BHI should be complemented by analysis of its individual constituent components because aggregated z-scores may hide valuable information about specific aspects of bank performance.
- Relativity:
  - The BHI and heatmap show only relative health within the chosen sample; a bank labeled “healthiest” is relative to peers in the sample, not necessarily in absolute terms.
  - Absolute comparisons require inclusion of a peer known to be a representative global benchmark for financial strength.
- Cross-sectional and inter-temporal comparisons:
  - The BHI does not adjust for heterogeneity across banks; HEAT! is best applied to homogeneous institutions (similar business models, same regulatory requirements).
  - Three-period rolling calculation of system means and standard deviations can affect comparability over longer time periods because z-scores for a period are based on different means and standard deviations than other periods.
- Definition of variables within and across databases:
  - Non-performing loans (NPLs): no single international standard; BankScope “impaired loans” used to calculate NPL ratio, but countries may define “impaired” or “NPL” differently.
  - Liquidity ratios: results may vary depending on whether the BankScope pre-defined liquidity ratio is used or if government securities are included to reflect Basel III HQLA requirements; inclusion of government securities can make banks holding high government securities appear very liquid.
  - Variable definitions may differ across BankScope, Bloomberg and SNL databases.

### V. Concluding remarks — utility and limitations
- Motivation: The global financial crisis highlighted the systemic importance of individual banks; surveillance and crisis management increasingly require analyses of individual (especially systemic) banks.
- Contribution: Development of the BHI for simple, preliminary analyses of individual banks and HEAT! to facilitate calculation and presentation.
- Back-test result: Spanish system back-test suggests the BHI accurately differentiates banks by financial soundness.
- Important caveats reiterated:
  - BHI is an aggregation of ratios—examine individual components as well.
  - Z-scores provide relative, not absolute, assessments; sample selection matters.
  - Differences in business models, changing bank activities over time, and definition choices for constituent components affect interpretation.
  - Users must be familiar with banking-system peculiarities and supplement BHI assessments with other quantitative and qualitative information.

### Appendix I excerpt — requisite software and HEAT! template instructions
- Requisite software:
  - BankScope: Excel Add-ins—BankScope (save HEAT! file in xls format for BankScope).
  - Bloomberg: Excel with linked access to a Bloomberg terminal (save HEAT! file in xlsx format for Bloomberg).
  - SNL: Excel Add-ins—SNLxl (save HEAT! file in xlsx format for SNL).
- Bank list definition and inclusion:
  - Define BankScope bank IDs, Bloomberg bank tickers, or SNL institution keys using the respective Wizard/Data Wizard or Help functions.
  - Update the Included? column with a “Y” against each bank to be included in the sample.
  - For BankScope use the Bank List spreadsheet; for Bloomberg and SNL follow analogous Bank List steps.
- Tool upload and refresh:
  - BankScope: Open BankScope Excel Add-ins, import the HEAT! file; the file will take several minutes to read and update.
  - Bloomberg: Open the HEAT! file in Excel with linked Bloomberg; the file will update automatically.
  - SNL: Open SNL Excel Add-ins, open the HEAT! file and use the Refresh Data function; ensure purple-tabbed spreadsheets do not have empty cells for SNL institution keys.
- Heatmap generation workflow (applies to BankScope, Bloomberg, SNL):
  - Purple spreadsheets: download data from the respective database.
  - Green spreadsheets: organize data and calculate means and standard deviations used for normalizing ratios.
  - Orange spreadsheets: compute the normalized ratios (z-scores).
  - Blue spreadsheets: generate the heatmaps.
- File update procedure (BankScope specific):
  - The file will update only after it is re-imported. After changes, save and close the Excel Add-in application; then re-open and re-import the file to update calculations.

*Source — content extracted from the provided PDF chapter/section.*

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

### References

### Key objectives and uses of the Bank Health Index (BHI) and HEAT!
- Proposes a simple, broadly-based measure of bank soundness (Bank Health Index, BHI) and automated spreadsheet templates (HEAT!) to facilitate preliminary, first-pass analysis of individual financial institutions and financial systems.
- Intended uses:
  - Enable early identification of vulnerabilities in global systemically important banks (G-SIBs) and domestic systemically important banks (D-SIBs).
  - Inform system-wide reform strategies by differentiating core, healthy banks from weak banks needing restructuring or resolution.
  - Inform restructuring decisions (mergers and acquisitions, recapitalization, liquidity support) by highlighting banks’ weaknesses.
- Limitations noted: BHI is a simple tool useful for initial identification and for highlighting specific vulnerability areas; analyses should be complemented by more rigorous quantitative (e.g., stress tests) and qualitative (e.g., supervisory and regulatory frameworks) assessments.

### Data requirements and databases used
- Preferred data: audited banks’ financial statements.
- Databases used in the paper and HEAT! template: BankScope, Bloomberg and SNL.

### Constituent financial ratios used to construct the BHI
- Five financial ratios calculated (definitions vary slightly by database):
  - Capital adequacy:
    - BankScope: Capital adequacy ratio, defined as total equity to RWA.
    - Bloomberg and SNL: Tier 1 capital to RWA.
  - Asset quality:
    - BankScope and SNL: Non-performing loans (NPL) ratio defined as impaired loans to gross loans, less ratio of provisions to gross loans.
    - Bloomberg and SNL: NPL to gross loans, less ratio of provisions to gross loans.
  - Earnings:
    - BankScope and SNL: Profitability ratio defined as the return on average assets.
    - Bloomberg: Return on total assets.
  - Liquidity:
    - BankScope and Bloomberg: Liquidity ratio defined as liquid assets to deposits and short-term borrowing.
    - SNL: Liquid assets to total liabilities.
  - Leverage:
    - BankScope, Bloomberg and SNL: Leverage ratio defined as tangible common equity to tangible assets.
- Note on SNL: availability of NPL or impaired loans depends on the accounting standard applied in a particular system.

### Normalization and construction of z-scores
- Each financial ratio is normalized (z-score) to facilitate comparability.
- Notation and definitions provided:
  - z_{i,t} is the normalized (or z-score for) the financial ratio of bank i at time t;
  - x_{i,t} is the financial ratio of bank i at time t;
  - ߤ is the system mean of a particular financial ratio over the three periods to time t;
  -  is the system standard deviation of a particular financial ratio over the three periods to time t.
- System means and standard deviations are calculated over three periods in order to:
  - provide a sufficiently large sample size;
  - incorporate both the time and cross-sectional dimensions;
  - ensure that any deterioration during the crisis period is adequately captured.

### Interpretation of z-scores and treatment of NPL ratio
- For all financial ratios except asset quality (NPL ratio), a positive z-score indicates that the bank’s financial ratio is better than the corresponding average across its peer group over three periods.
- For the asset quality measure (NPL ratio), the NPL ratio z-score is multiplied by -1 so that any increase in that score would be represented as an increasingly negative development.

### Structure of the paper and supplemental materials
- Paper structure (as described):
  - Section II: data requirements and construction of the BHI, and generation of the BHI heatmap using the HEAT! template.
  - Section III: example application and back-test of the BHI using a well-known country example.
  - Section IV: discussion of the main caveats associated with the BHI and its constituent components.
  - Section V: concluding remarks.
  - Appendix I: user guide for HEAT!
- Figures and Appendix listed (titles only):
  - Figures include heatmaps and consolidation visuals for G-SIBs and Spain, and liquidity ratio heatmaps by definition vis-à-vis government debt holdings.
  - Appendix I: User Guide: Operating Instructions for HEAT!

*Source: References section and extracted pages of the provided IMF working paper content.*

### 4.      An overall relative health score for each bank at a particular point in time can be

### _wp13177 - 4.      An overall relative health score for each bank at a particular point in time can be

### Summary of the BHI definition and formula
- The overall relative health score for each bank at time t, denoted ݖ௕௔௡௞,௧, is estimated by summing the z-scores for each financial ratio:
  - ݖ௕௔௡௞,௧ = ݖ௖௔௣௜௧௔௟,௧ + ݖ௔௦௦௘௧,௧ + ݖ௘௔௥௡௜௡௚௦,௧ + ݖ௟௜௤௨௜ௗ௜௧௬,௧ + ݖ௟௘௩௘௥௔௚௘,௧
- Component definitions:
  - ݖ௖௔௣௜௧௔௟,௧ = z-score for capital adequacy at time t
  - ݖ௔௦௦௘௧,௧ = z-score for asset quality at time t
  - ݖ௘௔௥௡௜௡௚௦,௧ = z-score for earnings at time t
  - ݖ௟௜௤௨௜ௗ௜௧௬,௧ = z-score for liquidity at time t
  - ݖ௟௘௩௘௥௔௚௘,௧ = z-score for leverage at time t
- Interpretation:
  - ݖ௕௔௡௞ essentially represents a relative overall measure of health for a particular bank in the defined peer sample.
  - The combination of ݖ௕௔௡௞ scores for a sample constitutes the BHI for that sample.

### HEAT! tool and data sources
- HEAT! is an Excel-based tool that:
  - downloads requisite data for individual banks from Bankscope, Bloomberg (for listed banks only) and SNL;
  - automatically calculates corresponding z-scores (Appendix I);
  - automatically generates heatmaps of the BHI to visually differentiate overall relative soundness and the individual constituent components for a period and over time.
- Heatmap color coding:
  - institutions in the top 90th percentile of soundness are denoted in green;
  - those in the bottom 10th percentile are denoted in red;
  - the rest are shown in various shades of yellow/orange.

### III. Application of the BHI and HEAT! — Key findings and demonstrations
- Purpose: Demonstrate BHI and HEAT! outputs using well-known peer groups and a domestic back-test; anonymize banks to avoid implicating specific institutions.
A. International Peer Group Example (G-SIBs, data to December 2012)
- Sample: global systemically important banks (G-SIBs) identified by BCBS (2011) and listed by the FSB (2012), consisting of 28 banks (anonymized; buckets 3 and 4 combined).
- Illustrative findings:
  - G-SIBs have improved their earnings performance relative to the “trough” in 2008.
  - The majority of banks appear to have deleveraged since 2008 as evidenced by improved leverage ratios.
- Bucket-differentiated observations:
  - A smattering of G-SIBs across buckets exhibit distinct overall weakness relative to peers.
  - Bucket 1: relatively lower capital adequacy, less liquid, more leveraged.
  - Bucket 2: relatively weaker earnings and remain more leveraged.
  - Bucket 3+4: asset quality and leverage are the main weaknesses.
B. Back-testing: Spain Case Study (end-2011 data)
- Context and reforms since the global financial crisis:
  - Mergers of savings banks into larger commercial banks and takeovers during the crisis; in some cases, several weak entities were merged into larger weak ones.
  - Authorities commissioned stress tests, audits of loan portfolios and asset valuations by third parties following the IMF’s FSAP Update assessment in 2012 H1.
  - Since 2012 Q3, per Spain’s Memorandum of Understanding (MoU) with the Eurogroup, restructuring included recapitalization and transfer of real estate loans to an asset management company.
- BHI back-test using end-2011 data produced a HEAT! heatmap showing clear differentiation across institutions and over time; results align with FSAP (IMF, 2012) and Oliver Wyman (2012) assessments:
  - The three largest institutions (Santander, BBVA, Caixa) are among the healthiest and most resilient within the Spanish system.
  - Smaller banks identified as not requiring additional capital even under severe stress (e.g., Kutxa and Unicaja) appear very sound according to the BHI.
  - The fourth largest bank (BFA-Bankia) undergoing restructuring and identified as requiring significant recapitalization is the result of a 2009 merger of several already-weak entities.
  - Banks considered non-viable and taken over by the Fund for Orderly Bank Restructuring (FROB) are among the weakest (examples: Unnim and CAM sold to other banks with financial support; Catalunya Caixa and Nova Galicia Caixa require recapitalization).
- Conclusion from the back-test:
  - The BHI effectively differentiates relative soundness within a system—identifying problem banks prior to the MoU and highlighting healthier banks.
  - Relative nature emphasized: Spanish G-SIBs rank among the healthiest domestically but are about average relative to G-SIB peers.

### IV. Main caveats when using the BHI and HEAT!
Aggregation
- The BHI should be complemented by analysis of its individual constituent components because aggregated z-scores may hide valuable information about specific aspects of bank performance.
Relativity
- The BHI and heatmap show only relative health within the chosen sample; a bank labeled “healthiest” is relative to peers in the sample, not necessarily in absolute terms.
- Absolute comparisons require inclusion of a peer known to be a representative global benchmark for financial strength.
Cross-sectional and inter-temporal comparisons
- The BHI does not adjust for heterogeneity across banks; HEAT! is best applied to homogeneous institutions (similar business models, same regulatory requirements).
- Three-period rolling calculation of system means and standard deviations can affect comparability over longer time periods because z-scores for a period are based on different means and standard deviations than other periods.
Definition of variables within and across databases
- Definitions matter; relative soundness may differ due to country peculiarities in definitions and variable definitions within and across databases:
  - Non-performing loans (NPLs): no single international standard; BankScope “impaired loans” used to calculate NPL ratio, but countries may define “impaired” or “NPL” differently (examples provided in text).
  - Liquidity ratios: results may vary depending on whether the BankScope pre-defined liquidity ratio is used or if government securities are included to reflect Basel III HQLA requirements; inclusion of government securities can make banks holding high government securities appear very liquid.
  - Variable definitions may differ across BankScope, Bloomberg and SNL databases (noted in Section II and Figure 5).

### V. Concluding remarks — utility and limitations
- Motivation: The global financial crisis highlighted the systemic importance of individual banks; surveillance and crisis management increasingly require analyses of individual (especially systemic) banks.
- Contribution: Development of the BHI for simple, preliminary analyses of individual banks and HEAT! to facilitate calculation and presentation.
- Back-test result: Spanish system back-test suggests the BHI accurately differentiates banks by financial soundness.
- Important caveats reiterated:
  - BHI is an aggregation of ratios—examine individual components as well.
  - Z-scores provide relative, not absolute, assessments; sample selection matters.
  - Differences in business models, changing bank activities over time, and definition choices for constituent components affect interpretation.
  - Users must be familiar with banking-system peculiarities and supplement BHI assessments with other quantitative and qualitative information.

*Italic: Source — content extracted from the provided PDF chapter/section.*

### 1.      The requisite software is Excel Add-ins—BankScope.

### 1.      The requisite software is Excel Add-ins—BankScope.

### Bank list definition (BankScope)
- Save a copy of the HEAT! file in xls (not .xlsx) format.
- Go to the Bank List spreadsheet.
- Define the list of BankScope bank IDs, which can be obtained using the Wizard facility in BankScope Excel Add-ins. The Wizard facility has a search function, which allows the selection of banks based on various criteria.
- Go to the Soundness spreadsheet and update the Included? column with a “Y” against each bank that should be included in the sample. This feature enables the user to change the banks in a sample.
- Go to Formulas on the menu bar and select Calculate Now to update related links in the workbook, if necessary.

### Tool upload (BankScope)
- Open the BankScope Excel Add-ins program.
- Go to Add-ins on the menu bar and look for BankScope Add-ins.
- Import the HEAT! file. The file will take several minutes to read and update for the changes to the list of banks.

### Heatmap generation (BankScope)
- Ensure that the individual banks appear in the underlying purple-, green- and orange-tabbed spreadsheets:
  - The purple spreadsheets have the function of downloading data from BankScope.
  - The green spreadsheets have the function of organizing the data and calculating the values of means and standard deviations used for normalizing the ratios.
  - The orange spreadsheets have the function of computing the normalized ratios (or z-scores).
  - The blue spreadsheets generate the heatmaps.

### File update procedure (BankScope)
- The file will update only after it is re-imported. Every time changes are made to the workbook, save and close the Excel Add-in application; then re-open the Excel Add-in and re-import the file to update the calculations for the changes.

### Template instructions for Bloomberg — software requirement
- The requisite software is Excel with linked access to a Bloomberg terminal.

### Bank list definition (Bloomberg)
- Save a copy of the HEAT! file in xlsx format.
- Go to the Bank List spreadsheet.
- Define the list of Bloomberg bank tickers, which can be obtained by entering the bank name and using the [Help] function.
- Go to the Soundness spreadsheet and update the Included? column with a “Y” against each bank that should be included in the sample. This feature enables the user to change the banks in a sample.
- Go to Formulas on the menu bar and select Calculate Now to update related links in the workbook, if necessary.

### Tool upload (Bloomberg)
- Open the Excel program with linked access to a Bloomberg terminal.
- Open the HEAT! file. The file will update automatically.

### Heatmap generation (Bloomberg)
- Ensure that the individual banks appear in the underlying purple-, green- and orange-tabbed spreadsheets:
  - The purple spreadsheets have the function of downloading data from Bloomberg.
  - The green spreadsheets have the function of organizing the data and calculating the values of means and standard deviations used for normalizing the ratios.
  - The orange spreadsheets have the function of computing the normalized ratios (or z-scores).
  - The blue spreadsheets generate the heatmaps.

### Template instructions for SNL — software requirement
- The requisite software is Excel Add-ins—SNLxl.

### Bank list definition (SNL)
- Save a copy of the HEAT! file in xlsx format.
- Go to the Bank List spreadsheet.
- Define the list of SNL institution keys, which can be obtained using the Data Wizard facility in SNL Excel Add-ins. The Data Wizard facility has a search function, which allows the selection of banks based on various criteria.
- Go to the Soundness spreadsheet and update the Included? column with a “Y” against each bank that should be included in the sample. This feature enables the user to change the banks in a sample.
- Go to Formulas on the menu bar and select Calculate Now to update related links in the workbook, if necessary.

### Tool upload and data refresh (SNL)
- Open the SNL Excel Add-ins program.
- Open the HEAT! file. The file can be updated using the Refresh Data function.
- Ensure that all purple-tabbed spreadsheets do not have empty cells that are supposed to contain SNL institution keys—SNL treats empty cells as “0” and would not be able to obtain any data.

### Heatmap generation (SNL)
- Ensure that the individual banks appear in the underlying purple-, green- and orange-tabbed spreadsheets:
  - The purple spreadsheets have the function of downloading data from SNL.
  - The green spreadsheets have the function of organizing the data and calculating the values of means and standard deviations used for normalizing the ratios.
  - The orange spreadsheets have the function of computing the normalized ratios (or z-scores).
  - The blue spreadsheets generate the heatmaps.

*Source: _wp13177 - 1.      The requisite software is Excel Add-ins—BankScope.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13177.pdf_