## _wp1626

## Source details

**Canonical URL:** [_wp1626](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1626.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1626.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1626.pdf.json)

---

### I. INTRODUCTION
- The global financial crisis highlighted the need for the IMF’s financial soundness indicators (FSIs) to capture build-up of systemic risks in a forward-looking manner.
- Current FSIs are contemporary sector average indicators that may hide variations within the population of financial institutions.
- Under the IMF/Financial Stability Board G-20 Data Gaps Initiative the Fund was asked “to investigate, develop, and encourage implementation of standard measures that can provide information on tail risks, concentrations, variations in distributions, and the volatility of indicators over time.”
- Discussions considered using Concentration and Distribution Measures (CDMs) to elaborate on the current set of FSIs; CDMs—calculated with due regard to confidentiality—could signal vulnerabilities better than simple averages.
- The IMF’s Statistics Department (STA) launched a pilot project in July 2014 with a group of 35 countries participating voluntarily.
- Primary objectives of the pilot:
  - Assess feasibility of calculating and reporting CDM data for selected FSIs for the Deposit-Takers’ (DTs’) sector.
  - Ascertain:
    - (1) effectiveness of the test set of CDMs in monitoring financial sector vulnerabilities;
    - (2) potential confidentiality concerns;
    - (3) extent of reporting burden; and
    - (4) procedures and resources the Fund would need to gather, compile, analyze, and disseminate CDMs alongside current FSI data and metadata.
- A report on the pilot was submitted to the Financial Soundness Indicators Reference Group (FSIRG); the majority of respondents were supportive and Fund staff intends to seek further feedback on analytical use and reporting burdens.

### II. MODALITIES OF THE PILOT PROJECT
- Indicators and reporting thresholds
  - CDMs were compiled for six FSIs of deposit takers (all deposit-taking institutions, except the central bank).
  - FSIs covered (as in Table 1):
    - Capital Adequacy: Regulatory Tier 1 capital to risk-weighted assets
    - Asset Quality: NPL to total gross loans
    - Profitability: Return on assets (ROA); Return on equity (ROE)
    - Liquidity: Liquid assets to short-term liabilities
    - Leverage: Capital to total assets
  - CDMs included:
    - (1) minimum, maximum, and mean;
    - (2) weighted standard deviations and skewnesses;
    - (3) quartiles and the asset share of the bottom quartile.
  - A concentration (Herfindahl) index was calculated.
  - Reporting minimum-number and frequency requirements (as in Table 2):
    - Mean (weighted by shares of assets in total assets) / Three / Monthly, quarterly, or annual
    - Median / Three / Monthly, quarterly, or annual
    - Minimum value / Three / Monthly, quarterly, or annual
    - Maximum value / Three / Monthly, quarterly, or annual
    - Standard deviation (weighted by shares of assets in total assets) / Five / Monthly, quarterly, or annual
    - Skewness (weighted by shares of assets in total assets) / Five / Monthly, quarterly, or annual
    - Average values by quartile / Twelve / Monthly, quarterly, or annual
    - Herfindahl index / Five / Annual
  - STA provided a standard Microsoft Excel template and a Guidance Note; template included worksheets for input/output, calculation of CDMs, and metadata.
  - Pilot participants were asked to report historical CDMs covering at least 2010–13 and encouraged to report 2007–13 if possible; use same reporting frequency and set of DTs as for regular FSI reporting.
  - To preserve confidentiality, participating countries were asked not to report underlying data for individual institutions; underlying data were not needed for the pilot objectives.
  - Minimum numbers of institutions were required to calculate CDMs to help preserve confidentiality; higher thresholds could be considered for data that could be publicly disseminated.

- Definitions of CDMs
  - Concentration
    - The Herfindahl Index, H, was used as the reporting measure of concentration.
    - H is the sum of the squares of the asset shares (measured in percent) of all firms in a sector:
      - H = sum_{i=1}^N (a_i)^2
    - Values range from 0 to 1.00. Higher values indicate greater concentration.
    - If only one bank in a sector (perfect concentration), H = 1.00.
    - If 100 equal sized firms with perfectly even distribution, H = 0.01.
    - Rule of thumb: H below 0.10 indicates relatively limited concentration; H above 0.18 indicates significant concentration.
  - Measures of dispersion
    - Required reporting of measures of (1) central tendency (mean and median); (2) variability (minimum, maximum, and standard deviation); and (3) skewness.
    - Mean: weighted arithmetic average of FSI values:
      - Mean = sum_{i=1}^N (FSI_i * a_i)
    - Median: middle value of an FSI after ranking institutions by FSI from lowest to highest; median not weighted by assets.
    - Minimum and maximum: smallest and largest value of each FSI for any DT in the sector.
    - Standard deviation (σ): square root of the weighted variance (σ^2), where weighted variance = sum_{i=1}^N [ (FSI_i - Mean)^2 * a_i ], and σ = sqrt(weighted variance).
    - Skewness (μ_3):
      - μ_3 = sum_{i=1}^N [ (FSI_i - Mean)^3 * a_i ] / σ^3
      - Positive skewness = longer right-hand tail; negative skewness = longer left-hand tail.
    - Quartiles and asset shares:
      - Quartiles are values of FSIs for each of four quartiles of the DT sector, determined for each FSI by sorting institutions by the FSI from top to bottom then dividing into four equal sized groups.
      - Asset share is calculated for the bottom quartile only: assets of institutions in that quartile divided by total DT sector assets.

### III. RESULTS OF THE PILOT PROJECT
- Participation and coverage
  - IMF Staff contacted 95 authorities reporting FSIs at the time; 49 agreed to participate although data from only 35 was actually received.
  - Participation spanned all regions; Europe was more highly represented, with both advanced and emerging countries.
  - Good variation across income groups except low-income countries, which were less represented possibly due to capacity constraints.
  - Participant list (35 countries) included: Armenia, Bosnia and Herzegovina, Brazil, Canada, Chile, China, P.R.: Macao, Costa Rica, Cyprus, Czech Republic, Dominican Republic, El Salvador, France, Georgia, Germany, India, Ireland, Israel, Italy, Macedonia, FYR, Malta, Mauritius, Namibia, Netherlands, Nigeria, Norway, Panama, Paraguay, Romania, Slovak Republic, South Africa, Sri Lanka, Turkey, Uganda, Ukraine, Zambia.
  - Source note: CDM dataset and IMF staff calculations. AFR = African Department (Sub-Sahara African countries); APD = Asia and Pacific Department; EUR = European Department; MCD = Middle East and Central Asia Department; WHD = Western Hemisphere Department.

- Data completeness
  - Comprehensiveness of reporting varied across countries, indicators, and time periods.
  - Reporting was most complete for 2013, with less data for earlier years especially before 2010.
  - Reporting of the profitability CDMs was the most comprehensive, while capital to (text indicates lower completeness for some indicators).

### Pilot implementation, participation, and feedback
- Thirteen countries provided feedback on the pilot project.
- Twelve of the thirteen countries were broadly supportive of the project.
- One country that is not an FSI reporter indicated that it would not be able to participate in the compilation and dissemination of CDMs.
- None of the twelve broadly supportive countries reported any potential burden resource associated with the compilation of CDMs.
- Concerns and suggestions from participants:
  - A couple of countries questioned the inclusion of maximum and minimum values of FSIs and suggested percentiles be used instead.
  - Three countries raised confidentiality issues:
    - One country indicated using internal data suppression techniques beyond the thresholds suggested in the CDM report.
    - Another country suggested disseminating data on groups of countries (regional, according to level of development, etc.) rather than on an individual basis.
  - A couple of countries suggested improvements to CDM templates to address error messages when there is no data for a reporting entity for a specific CDM indicator.

### Data submission, compilation, and resource implications
- Many participants freely provided bank-by-bank input data even though the template clearly indicated such data was not requested.
- Some participants engaged IMF staff prior to submitting results to resolve methodological issues such as the computation of quartiles.
- IMF staff noted that most notes/comments related to country-specific adjustments for missing values (most often relating to short-term liabilities) or difficulties in providing historical values.
- Data compilation issues identified:
  - Several countries entered small amounts when institutions reported zeros so that the template would not return a blank for the CDM for the related indicator (most commonly for liquid liabilities); a rule could be adopted so a small number of missing or misleading values from one or two institutions do not cause an entire CDM to be reported as a blank or misleading value.
  - Submissions of different periodicities had to be annualized for all countries for comparison purposes.
  - Some countries submitted multiple output sheets for a single time period to capture variables reported by all but a small number of institutions; these multiple output sheets had to be combined into one, raising resource costs of compilation.
  - Output sheets required close validation for methodological errors, such as combining multiple periods in one output sheet.
- Compilation resource implications:
  - Submissions were received in Excel files and had to be transferred into Economic Outlook Suite (EcOS) software, a labor intensive process that could be automated if regular CDM reporting moves forward.
  - Notes and comments were compiled separately in Excel files; this process could also be automated with regular CDM reporting.
- Some issues could be solved by automating data submissions; others reveal conceptual issues with the data.

### Analytical value and characteristics of CDM data
- CDM data provide important information not revealed by averages and are useful for financial stability and performance assessments and monitoring financial sector vulnerabilities.
- Distributions of minimum values of CDMs (representing institutions with the most severe risks) show substantial variation across countries and over time within countries.
- Minimum values are consistently significantly lower than averages with notable outliers in several instances.
- Pilot figures illustrate:
  - Distribution of minimum values of Tier 1 Capital to Risk-Weighted Assets by country groups (using World Bank income-level classification).
  - Distribution of maximum values of NPLs to Total Gross Loans by country groups.
  - Evolution of minimum values of Tier 1 Capital to RWA and variation in standard deviations for Tier 1 Capital to RWA for different country groups.
- CDMs lend themselves to multiple analytical applications, including comparisons of individual countries with regional or income-level peers.

### Possible refinements to CDM reporting
- Expand the set of CDMs to include additional FSIs (selected entries shown exactly as in source):
  - Core FSIs for Deposit Takers: Solvency indicator (CET1 to RWA) — Yes; Net stable funding ratio; Provisions to NPLs — Yes
  - Additional FSIs for Deposit Takers: Credit growth to private sector
  - Additional FSIs for Other Financial Corporations: Capital adequacy ICs — Yes; Reinsurance issues ICs; Earnings and profitability ICs; Return on assets; Return on equity
  - Liquidity ratio PFs; Earnings and profitability PFs — Yes; Sectoral distribution of investments for MMFs; Maturity distribution of investments for MMFs
  - Additional FSIs for Nonfinancial Corporations: Return on assets; Earnings to interest expenses; Liquidity indicators; Current ratio; Liquidity ratio; NFC debt to GDP
  - Additional FSIs for HHs: Household debt to household disposable income
- Reporting format refinements:
  - Countries with large numbers of institutions could report decile averages in addition to quartile averages to provide additional information about tail risks without compromising confidentiality; countries with large numbers of institutions would report both decile averages and quartile averages.
  - For the ratio of liquid assets to short-term liabilities, report additionally the gap of short-term liabilities minus liquid assets, divided by capital or by total assets, in addition to reporting liquid assets divided by short-term liabilities. Both the gap and the ratio should be reported, and negative gaps as well as positive gaps should be reported.
  - Consider limits on the value of the liquid-assets-to-short-term-liabilities ratio:
    - Missing values should be thrown out.
    - Consider capping the value of this ratio at 100 percent (or some slightly higher level to take plausible reductions of liquid asset values into account).
  - Kurtosis will not be added as a CDM; it is noted as a descriptive statistic in the FSI Compilation Guide but its analytical value is not considered sufficient to merit additional loss of degrees of freedom.

### Confidentiality, reporting thresholds, and minimum number of institutions
- The required minimum numbers of reporting institutions (reporting threshold) should be increased for at least some CDMs.
- Example concerns:
  - The reporting threshold for the maximum, minimum, mean and median is only three institutions even though these four CDMs together would be sufficient to determine exact values for four individual institutions.
  - Some reporting thresholds should be increased at least to the point where values of individual institutions cannot be derived.
- Reporting thresholds considerations:
  - Introduce buffers and possibly a second set of stricter reporting thresholds for data to be reported publicly.
  - Some FSIs may be more sensitive than others and could have stricter reporting thresholds.
- Prioritization of CDMs:
  - The CDM with the lowest reporting threshold should be the minimum value (or for NPLs, the maximum), as this is the most important CDM for assessing tail risks.
  - The current threshold of three institutions might be sufficient for data reported to the IMF, but a higher number might be appropriate if these data were to be reported publicly.
  - The maximum value should have the highest reporting threshold because it normally has little bearing on tail risks.
  - The reporting threshold for the bottom quartile or decile averages should be lower than for the other quartile or decile averages; the bottom quartile and its share of total assets should be reported before other quartiles are reported.
- Data on the number of institutions covered in regular FSI reporting show that a number of countries have less than twenty institutions; reporting thresholds should account for country size.

### Possible way forward and next steps
- CDM reporting can have global benefits as well as local benefits for reporting countries; participation by larger advanced economies and locally systemically important ones could be especially encouraged.
- The Pilot Project indicates that regular reporting of CDMs may be feasible but further feedback from participating countries and IMF users should be sought, particularly compilers and regulators about concerns with providing CDM data to the IMF on a regular basis and concerns if these data were to be made public.
- If data are to be made public, it may be necessary to develop a separate set of indicators with fewer CDMs and/or stricter reporting thresholds.
- If reporting countries are comfortable with reporting burden and data confidentiality protection, CDM reporting could be introduced and the current list of CDMs could be updated to include the new core FSIs for DTs introduced in the revised FSI Compilation Guide.
- Consideration should be given to introducing CDMs for OFCs’ FSIs, especially for insurance corporations and high-leveraged institutions such as investment banks and hedge funds, while weighing feasibility and burden for countries with many institutions.
- IMF staff envisaged the following next steps:
  - Reflect the FSIRG’s further feedback in the revised FSI Compilation Guide in the chapter on concentration and distribution measures.
  - Schedule a FSIRG meeting in 2016 to discuss a revised FSI Compilation Guide and seek any further feedback on the CDM project.
  - In parallel with the revision of the FSI Compilation Guide, consider the development of templates for regular CDM reporting.

*Source: CDM dataset and IMF staff calculations (pilot report content as provided).*

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

### _wp1626 - References .............................................................................................................

### I. INTRODUCTION
- The global financial crisis highlighted the need for the IMF’s financial soundness indicators (FSIs) to capture build-up of systemic risks in a forward-looking manner.
- Current FSIs are contemporary sector average indicators that may hide variations within the population of financial institutions.
- In the context of the IMF/Financial Stability Board G-20 Data Gaps Initiative, the Fund was called upon “to investigate, develop, and encourage implementation of standard measures that can provide information on tail risks, concentrations, variations in distributions, and the volatility of indicators over time.”1
- Discussions considered using Concentration and Distribution Measures (CDMs) to elaborate on the current set of FSIs; CDMs—calculated with due regard to confidentiality—could signal vulnerabilities better than simple averages.
- The IMF’s Statistics Department (STA) undertook a pilot project to test augmenting FSIs with a limited set of CDM data; the proposal was presented to the Financial Soundness Indicators Reference Group (FSIRG) and other experts.2
- The pilot project launched in July 2014 with a group of 35 countries participating voluntarily.
- Primary objectives of the pilot:
  - Assess feasibility of calculating and reporting CDM data for selected FSIs for the Deposit-Takers’ (DTs’) sector.
  - Ascertain:
    - (1) effectiveness of the test set of CDMs in monitoring financial sector vulnerabilities;
    - (2) potential confidentiality concerns;
    - (3) extent of reporting burden; and
    - (4) procedures and resources the Fund would need to gather, compile, analyze, and disseminate CDMs alongside current FSI data and metadata.
- A report on the pilot was submitted to the FSIRG; the majority of respondents were supportive and Fund staff intends to seek further feedback on analytical use and reporting burdens.

### II. MODALITIES OF THE PILOT PROJECT
- A. Indicators and Reporting Thresholds
  - CDMs were compiled for six FSIs of deposit takers (all deposit-taking institutions, except the central bank) (Table 1).
  - CDMs included:
    - (1) minimum, maximum, and mean;
    - (2) weighted standard deviations and skewnesses;
    - (3) quartiles and the asset share of the bottom quartile (Table 2).
  - A concentration (Herfindahl) index was calculated.
  - Table 1. Pilot Project: Subset of Financial Soundness Indicators
    - Capital Adequacy: Regulatory Tier 1 capital to risk-weighted assets
    - Asset Quality: NPL to total gross loans
    - Profitability: Return on assets (ROA); Return on equity (ROE)
    - Liquidity: Liquid assets to short-term liabilities
    - Leverage: Capital to total assets
  - Table 2. Pilot Project: Concentration and Distribution Measures (measure / required minimum number of financial institutions / required frequency and sample set of compilation)
    - Mean (weighted by shares of assets in total assets) / Three / Monthly, quarterly, or annual
    - Median / Three / Monthly, quarterly, or annual
    - Minimum value / Three / Monthly, quarterly, or annual
    - Maximum value / Three / Monthly, quarterly, or annual
    - Standard deviation (weighted by shares of assets in total assets) / Five / Monthly, quarterly, or annual
    - Skewness (weighted by shares of assets in total assets) / Five / Monthly, quarterly, or annual
    - Average values by quartile / Twelve / Monthly, quarterly, or annual
    - Herfindahl index / Five / Annual
  - STA developed a standard Microsoft Excel template and a Guidance Note to ensure uniform and consistent reporting; template included worksheets for input/output, calculation of CDMs, and metadata.
  - Pilot participants were asked to report historical CDMs covering at least 2010–13 and encouraged to report 2007–13 if possible; use same reporting frequency and set of DTs as for regular FSI reporting.
  - To preserve confidentiality, participating countries were asked not to report underlying data for individual institutions; underlying data were not needed for the pilot objectives.
  - Minimum numbers of institutions were required to calculate CDMs to help preserve confidentiality; higher thresholds could be considered for data that could be publicly disseminated.

- B. Definitions of CDMs
  - Concentration
    - The Herfindahl Index, H, was used as the reporting measure of concentration.
    - H is the sum of the squares of the asset shares (measured in percent) of all firms in a sector:
      - H = sum_{i=1}^N (a_i)^2
    - Values range from 0 to 1.00. Higher values indicate greater concentration.
    - If only one bank in a sector (perfect concentration), H = 1.00.3
    - If 100 equal sized firms with perfectly even distribution, H = 0.01.4
    - Rule of thumb: H below 0.10 indicates relatively limited concentration; H above 0.18 indicates significant concentration.
  - Measures of Dispersion
    - Required reporting of measures of (1) central tendency (mean and median); (2) variability (minimum, maximum, and standard deviation); and (3) skewness.
    - Mean: weighted arithmetic average of FSI values:
      - Mean = sum_{i=1}^N (FSI_i * a_i)
      - where FSI_i = value of the FSI for institution i; N = number of institutions in the DT sector.
    - Median: middle value of an FSI after ranking institutions by FSI from lowest to highest; median not weighted by assets.
    - Minimum and maximum: smallest and largest value of each FSI for any DT in the sector.
    - Standard deviation (σ): square root of the weighted variance (σ^2), where weighted variance = sum_{i=1}^N [ (FSI_i - Mean)^2 * a_i ], and σ = sqrt(weighted variance).
    - Skewness (μ_3):
      - μ_3 = sum_{i=1}^N [ (FSI_i - Mean)^3 * a_i ] / σ^3
      - Skewness indicates asymmetry around the mean: positive skewness = longer right-hand tail; negative skewness = longer left-hand tail.5
    - Quartiles and asset shares:
      - Quartiles are values of FSIs for each of four quartiles of the DT sector, determined for each FSI by sorting institutions by the FSI from top to bottom6 then dividing into four equal sized groups.7
      - Asset share is calculated for the bottom quartile only: assets of institutions in that quartile divided by total DT sector assets.

### III. RESULTS OF THE PILOT PROJECT
- Participation and coverage
  - IMF Staff contacted 95 authorities reporting FSIs at the time; 49 agreed to participate although data from only 35 was actually received.
  - Several countries reported capacity constraints preventing participation.
  - Participation spanned all regions; Europe was more highly represented, with both advanced and emerging countries.
  - Good variation across income groups except low-income countries, which were less represented possibly due to capacity constraints.
  - Participation figures and lists are summarized in Figure 1 (geographic coverage, participation by income level, and list of participating countries: 1 Armenia, Republic of; 2 Bosnia and Herzegovina; 3 Brazil; 4 Canada; 5 Chile; 6 China, P.R.: Macao; 7 Costa Rica; 8 Cyprus; 9 Czech Republic; 10 Dominican Republic; 11 El Salvador; 12 France; 13 Georgia; 14 Germany; 15 India; 16 Ireland; 17 Israel; 18 Italy; 19 Macedonia, FYR; 20 Malta; 21 Mauritius; 22 Namibia; 23 Netherlands; 24 Nigeria; 25 Norway; 26 Panama; 27 Paraguay; 28 Romania; 29 Slovak Republic; 30 South Africa; 31 Sri Lanka; 32 Turkey; 33 Uganda; 34 Ukraine; 35 Zambia.)
  - Source note: CDM dataset and IMF staff calculations. AFR = African Department (Sub-Sahara African countries); APD = Asia and Pacific Department; EUR = European Department; MCD = Middle East and Central Asia Department; WHD = Western Hemisphere Department.
- Data completeness
  - Comprehensiveness of reporting varied across countries, indicators, and time periods.
  - Reporting was most complete for 2013, with less data for earlier years especially before 2010.

*Source: _wp1626 - References .............................................................................................................*

### 2010. Reporting of the profitability CDMs was the most comprehensive, while capital to

### _wp1626 - 2010. Reporting of the profitability CDMs was the most comprehensive, while capital to

### Pilot implementation, participation, and feedback
- Thirteen countries provided feedback on the pilot project.
- Twelve of the thirteen countries were broadly supportive of the project.
- One country that is not an FSI reporter indicated that it would not be able to participate in the compilation and dissemination of CDMs.
- None of the twelve broadly supportive countries reported any potential burden resource associated with the compilation of CDMs.
- A couple of countries questioned the inclusion of maximum and minimum values of FSIs in the CDM dataset and suggested percentiles be used instead of minimum and maximum values.
- Three countries raised confidentiality issues:
  - One country indicated using internal data suppression techniques beyond the thresholds suggested in the CDM report.
  - Another country saw advantages of disseminating data on groups of countries (regional, according to level of development, etc.) rather than on an individual basis.
- A couple of countries suggested improvements to CDM templates to address error messages when there is no data for a reporting entity for a specific CDM indicator.

### Data submission, compilation, and resource implications
- Many participants freely provided bank-by-bank input data even though the template clearly indicated such data was not being requested under the pilot.
- Some participants engaged IMF staff prior to submitting results to resolve methodological issues such as the computation of quartiles.
- IMF staff noted that the majority of notes/comments related to country-specific adjustments made or that could have been made to accommodate missing values (most often relating to short-term liabilities) or difficulties in providing historical values.
- Data compilation issues identified:
  - Several countries entered small amounts when institutions reported zeros so that the template would not return a blank for the CDM for the related indicator (most commonly for liquid liabilities). A rule could be adopted so that a small number of missing or misleading values from one or two institutions do not cause an entire CDM to be reported as a blank or misleading value.
  - Submissions of different periodicities had to be annualized for all countries for comparison purposes.
  - Some countries submitted multiple output sheets for a single time period to capture variables reported by all but a small number of institutions; these multiple output sheets had to be combined into one, raising resource costs of compilation.
  - Output sheets required close validation for methodological errors, such as combining multiple periods in one output sheet.
- The compilation by IMF staff required considerable resources:
  - Submissions were received in Excel files and had to be transferred into Economic Outlook Suite (EcOS) software, a labor intensive process that could be automated if regular CDM reporting moves forward.
  - Notes and comments were compiled separately in Excel files; this process could also be automated with regular CDM reporting.
- Some issues could be solved by automating data submissions; others reveal conceptual issues with the data.

### Analytical value and characteristics of CDM data
- CDM data provide important information not revealed by averages and are useful for financial stability and performance assessments and monitoring financial sector vulnerabilities.
- Distributions of minimum values of CDMs (representing the institutions with the most severe risks) show substantial variation across countries and over time within countries.
- Minimum values are consistently significantly lower than averages with notable outliers in several instances.
- Figures in the pilot illustrate:
  - Distribution of minimum values of Tier 1 Capital to Risk-Weighted Assets by country groups (using World Bank income-level classification).
  - Distribution of maximum values of NPLs to Total Gross Loans by country groups.
  - Evolution of minimum values of Tier 1 Capital to RWA and variation in standard deviations for Tier 1 Capital to RWA for different country groups.
- CDMs lend themselves to multiple analytical applications, including comparisons of individual countries with regional or income-level peers.

### Possible refinements to CDM reporting
- Expand the set of CDMs to include additional FSIs listed in Table 3 (selected entries shown exactly as in source):
  - Core FSIs for Deposit Takers: Solvency indicator (CET1 to RWA) — Yes; Net stable funding ratio; Provisions to NPLs — Yes
  - Additional FSIs for Deposit Takers: Credit growth to private sector
  - Additional FSIs for Other Financial Corporations: Capital adequacy ICs — Yes; Reinsurance issues ICs; Earnings and profitability ICs; Return on assets; Return on equity
  - Liquidity ratio PFs; Earnings and profitability PFs — Yes; Sectoral distribution of investments for MMFs; Maturity distribution of investments for MMFs
  - Additional FSIs for Nonfinancial Corporations: Return on assets; Earnings to interest expenses; Liquidity indicators; Current ratio; Liquidity ratio; NFC debt to GDP
  - Additional FSIs for HHs: Household debt to household disposable income
- Countries with large numbers of institutions could report decile averages in addition to quartile averages to provide additional information about tail risks without compromising confidentiality; countries with large numbers of institutions would report both decile averages and quartile averages.
- Reporting of the ratio of liquid assets to short-term liabilities could be augmented by reporting the gap of short-term liabilities minus liquid assets, divided by capital or by total assets, in addition to reporting liquid assets divided by short-term liabilities. Both the gap and the ratio should be reported, and negative gaps as well as positive gaps should be reported.
- Consider limits on the value of the liquid-assets-to-short-term-liabilities ratio:
  - Missing values should be thrown out.
  - Consider capping the value of this ratio at 100 percent (or some slightly higher level to take plausible reductions of liquid asset values into account).
- Kurtosis will not be added as a CDM; it is noted as a descriptive statistic in the FSI Compilation Guide but its analytical value is not considered sufficient to merit additional loss of degrees of freedom.

### Confidentiality, reporting thresholds, and minimum number of institutions
- The required minimum numbers of reporting institutions (reporting threshold) should be increased for at least some CDMs.
- Example: the reporting threshold for the maximum, minimum, mean and median is only three institutions even though these four CDMs together would be sufficient to determine exact values for four individual institutions.
- Some reporting thresholds should be increased at least to the point where values of individual institutions cannot be derived.
- Reporting thresholds may need to include buffers, particularly for publicly reported data:
  - The minimum number of reporting institutions must be large enough that exact values cannot be calculated for each individual institution, but this may be insufficient to alleviate confidentiality concerns.
  - Consider introducing reporting thresholds in excess of what is needed to precisely calculate data for individual institutions, and introducing a second set of stricter reporting thresholds for data to be reported publicly.
  - Some FSIs may be more sensitive than others and could have stricter reporting thresholds.
- Prioritization of CDMs:
  - The CDM with the lowest reporting threshold should be the minimum value (or for NPLs, the maximum), as this is the most important CDM for assessing tail risks.
  - The current threshold of three institutions might be sufficient for data reported to the IMF, but a higher number might be appropriate if these data were to be reported publicly.
  - The maximum value should have the highest reporting threshold because it normally has little bearing on tail risks.
  - The reporting threshold for the bottom quartile or decile averages should be lower than for the other quartile or decile averages; the bottom quartile and its share of total assets should be reported before other quartiles are reported.
- Data on the number of institutions covered in regular FSI reporting show that a number of countries have less than twenty institutions; reporting thresholds should account for country size.

### Possible way forward and next steps
- CDM reporting can have global benefits as well as local benefits for reporting countries; participation by larger advanced economies and locally systemically important ones could be especially encouraged.
- The Pilot Project indicates that regular reporting of CDMs may be feasible but further feedback from participating countries and IMF users should be sought, particularly compilers and regulators about concerns with providing CDM data to the IMF on a regular basis and concerns if these data were to be made public.
- If data are to be made public, it may be necessary to develop a separate set of indicators with fewer CDMs and/or stricter reporting thresholds.
- If reporting countries are comfortable with reporting burden and data confidentiality protection, CDM reporting could be introduced and the current list of CDMs could be updated to include the new core FSIs for DTs introduced in the revised FSI Compilation Guide.
- Consideration should be given to introducing CDMs for OFCs’ FSIs, especially for insurance corporations and high-leveraged institutions such as investment banks and hedge funds, while weighing feasibility and burden for countries with many institutions.
- IMF staff envisaged the following next steps:
  - Reflect the FSIRG’s further feedback in the revised FSI Compilation Guide in the chapter on concentration and distribution measures.
  - Schedule a FSIRG meeting in 2016 to discuss a revised FSI Compilation Guide and seek any further feedback on the CDM project.
  - In parallel with the revision of the FSI Compilation Guide, consider the development of templates for regular CDM reporting.

*Source: CDM dataset and IMF staff calculations (pilot report content as provided).*

### REFERENCES

### _wp1626 - REFERENCES

### References
- Babihuga, R., 2007, “Macroeconomic and Financial Soundness Indicators: An Empirical, 
Investigation”, IMF Working Paper 07/115, Available at: 
http://www.imf.org/external/pubs/ft/wp/2007/wp07115.pdf  
- Costa Navajas, Matias and A. Thegeya, 2013, “Financial Soundness Indicators and Banking 
Crises”, IMF Working Paper No. 13/263.  
- Heath, Robert, 2013, “Why are the G-20 Data Gaps Initiative and the SDDS Plus Relevant 
for Financial Stability Analysis?”, IMF Working Paper WP/13/6.  
- IMF, 2006. Financial Soundness Indicators Compilation Guide, Available at: 
http://www.imf.org/external/pubs/ft/fsi/guide/2006/index.htm.  
- IMF, 2013, “Modifications to The Current List of Financial Soundness Indicators”, 
Available at: https://www.imf.org/external/np/pp/eng/2013/111313b.pdf.  
- IMF, 2013, “Modifications to the Current List of Financial Soundness Indicators (FSIs) 
Background Paper”; IMF Policy Paper; November 13, 2013, Available at: 
https://www.imf.org/external/np/pp/eng/2013/111313b.pdf.  
- IMF and FSB, 2009, “The Financial Crisis and Information Gaps”, Report to the G-20 
Finance Ministers and Central Bank Governors, Available at: 
http://www.imf.org/external/np/g20/pdf/102909.pdf.  

- 25

### Appendix I. FSIRG Composition
FSIRG Members
- ARGENTINA  
  Central Bank of Argentina  
- ARMENIA  
  Central Bank of Armenia  
- AUSTRALIA  
  Reserve Bank of Australia  
- BRAZIL  
  Banco Central do Brasil  
- CANADA  
  Bank of Canada  
  Statistics Canada  
- CHINA  
  People’s Bank of China  
  China Banking Regulatory Commission  
- CHILE  
  Central Bank  
- COLOMBIA  
  Superintendencia Financiera de 
  Colombia  
- DENMARK  
  Denmarks Nationalbank  
- FRANCE  
  Bank of France  
- GERMANY  
  Deutsche Bundesbank  
- INDIA  
  Reserve Bank of India  
- INDONESIA  
  Bank Indonesia  
- ITALY  
  Banca D’Italia  
- JAPAN  
  Bank of Japan  
- LEBANON  
  Central Bank of Lebanon  
- LUXEMBOURG  
  Central Bank of Luxembourg  
- MALAYSIA  
  Central Bank of Malaysia  
- MAURITIUS  
  Bank of Mauritius  
- MEXICO  
  Comisión Nacional de Valores of México  
- PHILIPPINES  
  Bangko Sentral ng Pilipinas  
- PORTUGAL  
  Banco de Portugal  
- ROMANIA  
  National Bank of Romania  
- RUSSIAN FEDERATION  
  Bank of Russia  
- SAUDI ARABIA  
  Saudi Arabian Monetary Agency  
- SOUTH AFRICA  
  South African Reserve Bank  
- SPAIN  
  Banco de España  
- SWITZERLAND  
  Swiss National Bank  
- TUNISIA  
  Banque Centrale de Tunisie  
- TURKEY  
  Bankacýlýk Düzenleme Ve Denetleme 
  Kurumu  
  Banking Regulation and Supervision Agency  
- UNITED KINGDOM  
  Bank of England  
- UNITED STATES  
  U.S. Federal Reserve Board

*Source: _wp1626 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1626.pdf_