## _wp11175

## Source details

**Canonical URL:** [_wp11175](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11175.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11175.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11175.pdf.json)

---

### I. Introduction and scope
- MFIs provide financial services to low-income, economically active borrowers who seek relatively small amounts to finance businesses, manage emergencies, acquire assets, or smooth consumption (CGAP 2002).
- MFIs are heterogeneous by type (banks, rural banks, non-bank financial institutions – NBFIs, non-government organizations – NGOs, credit unions, cooperatives), legal status (regulated, unregulated), purpose (for profit, non profit), location, and age.
- The paper revisits microfinance systemic risk using a large sample of top MFIs over a period that includes the global financial crisis of late 2007, to assess robustness of earlier findings on weak correlations with domestic and international economic conditions.

### Impact of the global financial crisis and systemic risk
- Crisis effects and changing sensitivity:
  - The global financial crisis reduced boom growth rates in microcredit; MFI balance sheets were shocked on the liability side (drop in liquidity for fund providers) and the asset side (increases in loan delinquency and write offs alongside domestic macroeconomic deterioration).
  - As the crisis progressed, relatively high MFI lending rates and lack of transparency attracted public and political attention; in some countries this led to initiatives introducing interest rate ceilings or forced blanket restructurings of debt.
  - Correlations between microfinance performance and both domestic and international economic conditions are stronger when including the global financial crisis period.
- Subgroups most sensitive to domestic and international conditions:
  - Regions: Central America and the Caribbean (including Mexico), Eastern Europe, and the Middle East and Central Asia.
  - Institution types: banks and non-bank financial institutions (NBFIs).
- Possible drivers of increased sensitivity:
  - Emphasis on “sustainability”, diversification of funding towards more commercial sources, commercialization of portfolios, and increases in scale—making MFIs more similar to traditional financial institutions.
- Strengthening linkages over time:
  - Relationship between MFI performance and domestic economic conditions and international capital markets is stronger when the sample period is shrunk to 2002-2009 versus 1998-2009.
  - As MFIs mature, their linkages with domestic and international market conditions get stronger.

### Overall trends in microfinance (1998-2009) — median MFI performance
- Pre-crisis (1998-2006) median MFI:
  - Assets growth: 36.0 percent (median MFI).
  - Lending growth: 39.9 percent (median MFI).
  - Borrowing growth: 51.7 percent (median MFI).
  - Return on equity (ROE): 9.8 percent (median MFI).
  - Portfolio at Risk (PAR-30): 3.3 percent (median MFI).
  - Write-off ratio: 1.0 percent (median MFI).
- Crisis and post-crisis (2007-2009) median MFI:
  - Assets growth: 21.9 percent (median MFI).
  - Lending growth: 23.7 percent (median MFI).
  - Borrowing growth: 23.3 percent (median MFI).
  - ROE: 9.7 percent (median MFI); mean ROE reported about 5 percentage points lower than pre-crisis mean.
  - PAR-30: 3.9 percent (median MFI).
  - Write-off ratio: 1.2 percent (median MFI).
- Differences (2007-2009 minus 1998-2006) for median MFI:
  - Assets growth: -14.1 (percentage points).
  - Lending growth: -16.2 (percentage points).
  - Borrowing growth: -28.4 (percentage points).
  - PAR-30: 0.7 (percentage points).
  - Write-off ratio: 0.2 (percentage points).
  - ROE difference: 0.0 (percentage points).
- Regional and institutional heterogeneity (selected highlights):
  - Eastern Europe (EUR) and Middle East & Central Asia (MCE) median MFIs grew faster and had better asset quality in 1998-2006.
  - Profitability largest for Central America & Caribbean (CAC) and South America (SAC) pre-crisis.
  - Banks outperformed other institution types in expansion, asset quality, and profitability; NBFIs followed banks.
  - Regulated institutions outperformed non-regulated MFIs on expansion, asset quality, and profitability.
  - CAC and EUR were most strongly impacted by the crisis (2007-2009): both showed negative growth rates for lending and borrowing in 2009 for the median MFI, and sharp deteriorations in asset quality and profitability.
  - SAC and Asia-Pacific (ASP) appeared to weather the crisis better overall, though deterioration in ASP during 2010 (not in sample) was noted, particularly in India.
- Possible drivers of deterioration during 2008-2009:
  - Credit oversupply in some markets leading to lower-quality marginal lending in 2006-2007 (larger loans, lower quality).
  - Reduced financing opportunities and deteriorated financing terms (higher borrowing rates and shorter maturities).

### Lending rates, spreads, and lending fundamentals (1998-2009)
- Median lending rate and spreads:
  - Lending rate 1998-2006: 35.7 percent (median MFI lending rate).
  - Lending rate 2007-2009: 32.4 percent (median MFI lending rate).
  - Interest rate spread 1998-2006: 30.9 percent (median spread).
  - Interest rate spread 2007-2009: 26.7 percent (median spread).
- Loan size and productivity:
  - Median loan size 1998-2006: US$ 361 (median loan size).
  - Median loan size 2007-2009: US$ 631 (median loan size).
  - Borrowers/staff ratio: 115 (1998-2006) → 123 (2007-2009).
  - Operational cost/staff (US$ ,000): 1160 (1998-2006) → 1546 (2007-2009).
  - Capital/Asset ratio: 34.3 percent (1998-2006) → 21.7 percent (2007-2009).
- Dispersion and group differences:
  - Interest rates and spreads largest in CAC; lowest in EUR.
  - Loan sizes lowest in sub-Saharan Africa (AFR) and ASP; largest in EUR.
  - NGOs have lowest loan sizes; banks have largest loan sizes.
  - Unregulated MFIs have lower costs per staff than regulated ones.
  - NGOs and unregulated institutions expanded more strongly over the decade; MFIs in MCE, EUR, and CAC also showed stronger expansion.

### Data, sample, and variables
- Source: Microfinance Information Exchange (MIX) database covering 1995-2009.
- Full MIX dataset: 1774 MFIs in 107 countries.
- Analysis sample used in regressions:
  - 353 MFIs that report at least 7 years of data during 1998-2009 and three consecutive years for 2007-09.
  - Sample includes top 75 percent by asset size (to avoid volatility from very small MFIs).
- Groupings: region (AFR, ASP, CAC, EUR, MCE, SAC); institution type (banks, CUCs, NBFIs, NGOs, rural banks); legal status (regulated, non-regulated); purpose (for-profit, non-profit); age (young, middle-aged, mature; mature = in operation for 8 years or more by 2009).
- Key variable definitions:
  - Asset and lending growth: 2005 real annual end-of-period percentage terms.
  - ROE: annual percentage terms over values in US$.
  - PAR-30: percentage of loans in arrears for past 30 days.
  - Lending rates: total financial income divided by net lending portfolio, percent per year, US$ terms.
  - Funding costs: total financial costs divided by total assets, percent per year, US$ terms.
  - Spread: lending rate minus borrowing rate (absolute).
  - Loan size: total lending portfolio (US$) divided by number of borrowers.
  - Borrowers per staff: active borrowers divided by total staff.
  - Operational cost per staff: total operational cost (US$) divided by number of staff.
  - Capital-asset ratio: net worth divided by total assets (percent).
  - Age coding: young = <5 years (0), middle-aged = 5–8 years (1), mature = >8 years (2).
- Robustness checks:
  - Alternative cross-sections: all MFIs when possible; sample adding lowest quartile by asset size to main sample (390 MFIs).
  - Alternative time periods: 1995-2009 and 2002-2009.
  - Alternative borrowing rate definition for 2002-2009 regressions: total financial cost divided by total borrowings (percent per year, US$).

### Empirical analysis (systemic risk and lending determinants) — methodology and main results
- Methodology:
  - Regress MFI performance indicators (assets growth, lending growth, PAR-30, ROE) on domestic GDP growth and global market indices (S&P 500, MSCI World, MSCI Emerging Markets).
  - Fixed effects OLS regressions; White robust standard errors.
- Main empirical findings (group-level):
  - Asset growth: positive and statistically significant association with domestic GDP growth, S&P 500, MSCI World, and MSCI Emerging Markets.
  - Lending growth: positive and statistically significant association with domestic GDP growth; statistically significant association with MSCI Emerging Markets among capital market measures.
  - Asset quality (PAR-30): negative and statistically significant association with domestic GDP growth and with all (but one) capital market measures.
  - ROE: positive and statistically significant association with domestic GDP growth; mixed significance across capital markets measures.
- Regional and institutional heterogeneity (selected results):
  - CAC, EUR, and MCE show the most consistent and statistically significant associations between asset growth and both domestic and international markets.
  - AFR shows the weakest links to global markets; some association of lending growth and asset quality with domestic conditions.
  - Banks and NBFIs: consistent positive and statistically significant links of asset growth with domestic and international markets; asset quality regressions show expected negative and significant signs against domestic economic conditions.
  - NGOs: asset quality regressions show expected negative signs, statistically significant mainly against international market conditions.
  - CUCs and rural banks: no strong empirical evidence of links with domestic and international capital market conditions.
- Age, legal status, and purpose:
  - Non-regulated MFIs: consistent associations with domestic and international conditions; profitability and asset quality regressions statistically significant in all cases.
  - Regulated MFIs: asset and lending growth and asset quality regressions mostly show expected signs and statistical significance; profitability regressions largely not statistically significant except for MSCI Emerging Markets.
  - Non-profit MFIs: more consistently linked to domestic and international capital markets (asset growth and asset quality regressions significant).
  - For-profit MFIs: asset and lending growth regressions show positive and statistically significant differences relative to non-profit MFIs; asset quality and profitability regressions often not statistically significant.
  - Mature MFIs: most consistent associations with domestic and international conditions (asset growth and asset quality significant); however, asset growth of mature MFIs is less responsive to changes in the economic environment than that of younger MFIs.
  - Young MFIs: asset growth positive and statistically significant in all cases; lending growth significant mainly for domestic conditions and emerging markets; asset quality coefficients generally not statistically significant.

### Determinants of lending rates and spreads — estimated model and quantitative impacts
- Conceptual decomposition:
  - Lending rates cover: cost of funds for on-lending, cost of risk (expected loan loss), and administrative costs (searching, appraising, processing, disbursing, monitoring, collecting, follow-up).
  - Administrative costs are not proportional to loan size; smaller average loan sizes imply higher rates to break even.
  - Productivity (borrowers per staff) and operational cost per staff matter; MFI age expected to reduce lending rates.
- Estimated specifications (fixed effects OLS):
  - Lending rate it_l r depends on: borrowing rate (it_br), average loan size (it_sz), borrowers per staff (it_bs), operating cost per staff (it_os), capital-asset ratio (it_ka), portfolio-at-risk lagged (1it_par−), real GDP growth (it_y), and MFI age (it_age).
  - Absolute spread (it_spr) specification identical.
- Key quantitative findings (preserved exactly as reported):
  - Borrowing rate (it_br)
    - Coefficient positive and highly statistically significant for all MFIs and most groupings.
    - In the aggregate regression, a Wald test cannot reject the null that the coefficient equals one (lending rate and spread regressions similar).
  - Average loan size (it_sz)
    - Coefficient negative and highly statistically significant for all MFIs and MFI groupings in lending rate regressions (spread regressions similar except NGOs).
    - Aggregate examples:
      - A US$100 increase in the average loan size would be associated with a decrease of lending rates (and spreads) of about 0.3 percentage points.
      - Depending on MFI purpose, legal status and age, the decrease would be in the range of 0.3 – 0.4 percentage points.
      - For MFIs in AFR: decrease of about 1.6 percentage points.
      - For MFIs in ASP: decrease of about 1.1 percent.
      - For MFIs in SAC: decrease of about 0.6 percentage points.
      - ASP region example: difference in loan sizes in 2009 between 80th and 20th percentiles was about US$ 500, which would explain a difference in lending rates of about 6 percentage points.
  - Borrowers per staff ratio (it_bs)
    - Coefficient negative and highly statistically significant for all MFIs and most groupings (not significant for most institution types except banks).
    - Quantitative examples:
      - An increase of 50 borrowers per staff would imply a decrease in lending rates (and spreads) of about 1 percentage points.
      - Depending on MFI purpose, legal status and age, the decrease would be in range between 1 – 2 percentage points.
      - For CAC and EUR regions, the decrease could be as high as 3.5 percentage points.
  - Operational cost per staff member (it_os)
    - Coefficient positive and highly statistically significant for all MFIs and most groupings (exceptions: banks and non-regulated MFIs).
    - Quantitative examples:
      - An US$1000 increase in operational expenditures per staff per year would cause an increase in lending rates (spreads) of about 0.3 percentage points.
      - Depending on MFI characteristics, the increase would be between 0.3 – 0.8 percentage points.
      - Coefficient larger for MFIs in AFR, MCE, and CAC regions.
  - Portfolio-at-Risk ratio (lagged 1it_PAR−)
    - Coefficient positive and statistically significant across MFIs in most regions (either in lending rate or spread regressions), but not for many other groupings—asset quality linked more to location.
    - Quantitative example:
      - An increase of 1 percentage point in the ratio would result in an increase in lending rates (spreads) of about 0.2 – 0.3 percentage points, depending on location.
  - GDP growth (it_y)
    - Coefficient negative and statistically significant only for EUR region; NBFIs; and regulated MFIs.
    - Quantitative example:
      - An increase of 1 percentage point in GDP growth would result in a decrease in spreads of about 0.1 – 0.3 percentage points depending on MFI characteristics.
    - Least consistent fundamental across MFI groupings.
  - Capital-Asset ratio (it_ka)
    - Coefficient positive and statistically significant for most MFI groupings (not significant for the sample of all MFIs).
    - Quantitative example:
      - A 1 percentage point additional capital-assets ratio would result in an increase of lending rates (spreads) of between 0.2 – 0.4 percent, depending on MFI characteristics.
  - Age (it_age)
    - Coefficient negative and statistically significant for all MFIs and most groupings.
    - Quantitative examples:
      - A MFI passing from “young” to “middle-aged” would result in a decrease of lending rates (spreads) of about 4 percentage points.
      - Going from “middle-aged” to “mature” would result in another 4 percentage point decrease.
      - The difference in lending rates (spreads) charged by a young MFI bank compared to a mature MFI bank would be of about 12 percentage points.

### Robustness
- Results broadly robust to:
  - Extending period to 1995-2009 and to 2002-2009.
  - Alternative cross-sections including additional smaller MFIs (sample of 390 MFIs).
  - Alternative definitions of borrowing rates for lending rate and spread regressions (total financial cost divided by total borrowings).
- Results stronger (more coefficients with expected sign and statistical significance) when period considered is 2002-2009, suggesting convergence of MFI behavior as they mature.
- Results somewhat weaker with larger MFI cross-sections due to higher variability from smaller MFIs.

### Policy implications and recommendations
- Systemic risk and regulatory stance:
  - The microfinance industry is more strongly linked to domestic economic conditions and international capital markets than previously thought; systemic risk has increased.
  - Systemic risk appears higher for MFIs operating in CAC and EUR regions; lower in ASP and SAC. Banks and NBFIs show higher systemic risk; NGOs lower. For-profit and younger MFIs show higher systemic risk than non-profit and mature MFIs.
- Interpretation of high lending rates:
  - High observed lending rates reflect structural industry fundamentals (small loan sizes, short durations, higher administrative costs), MFI productivity, portfolio quality, capital structure, and maturity rather than solely excessive profiteering.
- Risks of inappropriate regulation:
  - Inappropriate regulation (e.g., blanket loan restructurings at below-market interest rates, or interest rate caps using commercial bank benchmarks) risks withdrawal of funding, reduction in MFI scale, increases in average loan sizes, and increased effective interest rates for the poor due to disintermediation toward higher-rate informal lenders.
- Recommended regulatory and market measures:
  - If the policy objective is to reduce interest rates, regulation should:
    - Aim to create an enabling market structure for MFIs to operate, test lending technologies, innovate to increase productivity and efficiency, and exert competitive pressure.
    - Protect customers’ rights and ensure MFI transparency and truth in lending, while avoiding outright inappropriate policies such as interest rate caps that ignore MFI cost structures.
    - Choose prudential versus non-prudential measures appropriately according to local industry characteristics and avoid over-regulation given regulatory costs.
    - Strengthen legal frameworks: generalize use of credit bureaus, improve corporate governance principles, and enforce timely reporting on a standard basis to reduce boom-bust dynamics and limit lending-rate spikes as portfolios deteriorate.

*Source: _wp11175 - Section VI discusses the findings and concludes.*

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

### _wp11175 - References

### I. Introduction and scope
- MFIs provide financial services to low-income, economically active borrowers who seek relatively small amounts to finance businesses, manage emergencies, acquire assets, or smooth consumption (CGAP 2002).
- MFIs are heterogeneous by type (banks, rural banks, non-bank financial institutions – NBFIs, non-government organizations – NGOs, credit unions, cooperatives), legal status (regulated, unregulated), purpose (for profit, non profit), location, and age.
- The paper revisits microfinance systemic risk using a large sample of top MFIs over a period that includes the global financial crisis of late 2007, to assess robustness of earlier findings on weak correlations with domestic and international economic conditions.

### Microcredit scale, growth, and heterogeneity
- For some countries microcredit represents a significant portion of GDP and of total credit to the private sector; examples cited include low-income countries (LICs): Bolivia, Cambodia, Kenya, Mongolia, Nicaragua, Tanzania, Vietnam; transition economies: Armenia, Bosnia & Herzegovina, Kyrgyz Republic, Tajikistan; emerging economies: Peru.
- Ratios in Figure 1 correspond to the average for 2008-09.
- Median MFI asset size is plotted in US$ million for 2000 through 2009 across regions and institution types (Figure 2).
- Lending growth (median MFI) is shown in percent over values in US$ for 2000 through 2009 across regions and institution types (Figure 3).

### Impact of the global financial crisis (late 2007 onward)
- The global financial crisis reduced the boom growth rates in microcredit; MFI balance sheets were shocked on the liability side (drop in liquidity for fund providers) and the asset side (increases in loan delinquency and write offs alongside domestic macroeconomic deterioration).
- As the crisis progressed, relatively high MFI lending rates and lack of transparency attracted public and political attention; in some countries this led to initiatives introducing interest rate ceilings or forced blanket restructurings of debt.
- Previously weak relationships between MFI performance and international/domestic markets (Krauss and Walter, 2006, 2009; Gonzalez, 2007; Ahlin, Lin and Maio, 2010) may reflect samples covering largely expansionary periods and an industry “diffusion” phase; inclusion of crisis-period data is necessary to reassess correlations.

### Empirical findings on systemic risk and sensitivity
- Correlations between microfinance performance and both domestic and international economic conditions are stronger when including the global financial crisis period.
- MFIs most sensitive to domestic and international economic environment: Central America and the Caribbean (including Mexico), Eastern Europe, and the Middle East and Central Asia.
- By institution type, banks and non-bank financial institutions show the closest links to macroeconomic and international conditions.
- Possible drivers of increased sensitivity include: emphasis on “sustainability”, diversification of funding towards more commercial sources, commercialization of portfolios, and increases in scale—making MFIs more similar to traditional financial institutions.

### Determinants of lending rates, spreads, and policy implications
- Empirical analysis identifies determinants of lending rates and absolute spreads; key variables and findings include:
  - Average loan size (used as a proxy for administrative costs) has the expected sign.
  - Productivity and efficiency measures (including ratio of borrowers per staff, and operational cost per staff) have the expected signs and are, in most cases, statistically significant regardless of MFI characteristics.
  - More mature MFIs charge lower lending rates.
  - Variables tracking asset quality and MFI funding structure also matter for lending rates and spreads.
  - There is no strong evidence linking lending rates (spreads) with domestic economic conditions.
- Policy implication: If the aim is to reduce lending rates, effective regulatory focus is to create an enabling environment for MFIs, promote competition, and foster innovation in lending technologies rather than imposing interest rate ceilings.

### Data, figures, and tables referenced
- Time spans and series referenced include: 1998-2009 (Tables 1–7), 2000–2009 (figures of median MFI assets and lending growth), 2008-09 averages (Figure 1).
- Figures referenced in the text: Figure 1 (Microcredit in Selected Countries), Figure 2 (Assets of the Median MFI), Figure 3 (Lending Growth), Figure 4 (Portfolio at Risk (PAR-30)), Figure 5 (Return on Equity (ROE)), Figure 6 (Lending Rate).
- Tables referenced include: Table 1. MFIs Performance Indicators, 1998-2009; Table 2. MFIs Lending Rate Fundamentals, 1998-2009; Table 3. Selected MFI Sample; Table 4. Performance Regressions (Selected Sample, 1998-2009); Table 5. Performance Regressions (Selected Sample, 1998-2009); Table 6. Lending Regressions (Selected Sample, 1998-2009); Table 7. Spread Regressions (Selected Sample, 1998-2009).

*Source: _wp11175 - References*

### Section VI discusses the findings and concludes.

### _wp11175 - Section VI discusses the findings and concludes.

### Overall trends in microfinance (1998-2009)
- Pre-crisis (1998-2006) median MFI performance:
  - Assets growth: 36.0 percent (median MFI).
  - Lending growth: 39.9 percent (median MFI).
  - Borrowing growth: 51.7 percent (median MFI).
  - Return on equity (ROE): 9.8 percent (median MFI).
  - Portfolio at Risk (PAR-30): 3.3 percent (median MFI).
  - Write-off ratio: 1.0 percent (median MFI).
- Crisis and post-crisis (2007-2009) median MFI performance:
  - Assets growth: 21.9 percent (median MFI).
  - Lending growth: 23.7 percent (median MFI).
  - Borrowing growth: 23.3 percent (median MFI).
  - ROE: 9.7 percent (median MFI); mean ROE reported about 5 percentage points lower than pre-crisis mean.
  - PAR-30: 3.9 percent (median MFI).
  - Write-off ratio: 1.2 percent (median MFI).
- Differences (2007-2009 minus 1998-2006) for median MFI:
  - Assets growth: -14.1 (percentage points).
  - Lending growth: -16.2 (percentage points).
  - Borrowing growth: -28.4 (percentage points).
  - PAR-30: 0.7 (percentage points).
  - Write-off ratio: 0.2 (percentage points).
  - ROE difference: 0.0 (percentage points).
- Regional and institutional heterogeneity:
  - Eastern Europe (EUR) and Middle East & Central Asia (MCE) median MFIs grew faster and had better asset quality in 1998-2006.
  - Profitability largest for Central America & Caribbean (CAC) and South America (SAC) pre-crisis.
  - Banks outperformed other institution types in expansion, asset quality, and profitability; NBFIs followed banks.
  - Regulated institutions outperformed non-regulated MFIs on expansion, asset quality, and profitability.
  - CAC and EUR were most strongly impacted by the crisis (2007-2009): both showed negative growth rates for lending and borrowing in 2009 for the median MFI, and sharp deteriorations in asset quality and profitability.
  - SAC and Asia-Pacific (ASP) appeared to weather the crisis better overall, though deterioration in ASP during 2010 (not in sample) was noted, particularly in India.
- Possible drivers of deterioration during 2008-2009:
  - Credit oversupply in some markets leading to lower-quality marginal lending in 2006-2007 (larger loans, lower quality).
  - Reduced financing opportunities and deteriorated financing terms (higher borrowing rates and shorter maturities).

### Lending rate and lending fundamentals (1998-2009)
- Median lending rate:
  - 1998-2006: 35.7 percent (median MFI lending rate).
  - 2007-2009: 32.4 percent (median MFI lending rate).
- Interest rate spread:
  - 1998-2006: 30.9 percent (median spread).
  - 2007-2009: 26.7 percent (median spread).
- Median loan size:
  - 1998-2006: US$ 361 (median loan size).
  - 2007-2009: US$ 631 (median loan size).
- Productivity and costs:
  - Borrowers/staff ratio: 115 (1998-2006) → 123 (2007-2009).
  - Operational cost/staff (US$ ,000): 1160 (1998-2006) → 1546 (2007-2009).
  - Capital/Asset ratio: 34.3 percent (1998-2006) → 21.7 percent (2007-2009).
- Dispersion and group differences:
  - Interest rates and spreads largest in CAC; lowest in EUR.
  - Loan sizes lowest in sub-Saharan Africa (AFR) and ASP; largest in EUR.
  - NGOs have lowest loan sizes; banks have largest loan sizes.
  - Unregulated MFIs have lower costs per staff than regulated ones.
  - NGOs and unregulated institutions expanded more strongly over the decade; MFIs in MCE, EUR, and CAC also showed stronger expansion.

### Data, sample, and variables
- Source: Microfinance Information Exchange (MIX) database covering 1995-2009.
- Full MIX dataset: 1774 MFIs in 107 countries.
- Analysis sample used in regressions:
  - 353 MFIs that report at least 7 years of data during 1998-2009 and three consecutive years for 2007-09.
  - Sample includes top 75 percent by asset size (to avoid volatility from very small MFIs).
  - Groupings: region (AFR, ASP, CAC, EUR, MCE, SAC); institution type (banks, CUCs, NBFIs, NGOs, rural banks); legal status (regulated, non-regulated); purpose (for-profit, non-profit); age (young, middle-aged, mature; mature = in operation for 8 years or more by 2009).
- Key variable definitions:
  - Asset and lending growth: 2005 real annual end-of-period percentage terms.
  - ROE: annual percentage terms over values in US$.
  - PAR-30: percentage of loans in arrears for past 30 days.
  - Lending rates: total financial income divided by net lending portfolio, percent per year, US$ terms.
  - Funding costs: total financial costs divided by total assets, percent per year, US$ terms.
  - Spread: lending rate minus borrowing rate (absolute).
  - Loan size: total lending portfolio (US$) divided by number of borrowers.
  - Borrowers per staff: active borrowers divided by total staff.
  - Operational cost per staff: total operational cost (US$) divided by number of staff.
  - Capital-asset ratio: net worth divided by total assets (percent).
  - Age coding: young = <5 years (0), middle-aged = 5–8 years (1), mature = >8 years (2).
- Robustness checks:
  - Alternative cross-sections: all MFIs when possible; sample adding lowest quartile by asset size to main sample (390 MFIs).
  - Alternative time periods: 1995-2009 and 2002-2009.
  - Alternative borrowing rate definition for 2002-2009 regressions: total financial cost divided by total borrowings (percent per year, US$).

### Empirical analysis of systemic risk (1998-2009)
- Methodology:
  - Follows Krauss and Walter (2009) approach: regress MFI performance indicators (assets growth, lending growth, PAR-30, ROE) on domestic GDP growth and global market indices (S&P 500, MSCI World, MSCI Emerging Markets).
  - Fixed effects OLS regressions; White robust standard errors.
- Main findings for the whole sample (All MFIs):
  - Asset growth: positive and statistically significant association with domestic GDP growth, S&P 500, MSCI World, and MSCI Emerging Markets.
  - Lending growth: positive and statistically significant association with domestic GDP growth; statistically significant association with MSCI Emerging Markets among capital market measures.
  - Asset quality (PAR-30): negative and statistically significant association with domestic GDP growth and with all (but one) capital market measures.
  - ROE: positive and statistically significant association with domestic GDP growth; mixed significance across capital markets measures.
- Regional differences (selected results):
  - CAC, EUR, and MCE show the most consistent and statistically significant associations between asset growth and both domestic and international markets.
  - CAC and EUR asset quality regressions: negative and statistically significant coefficients for most market indicators.
  - ASP and SAC show some evidence of links (asset and lending growth, asset quality in ASP, profitability in SAC), but generally less pronounced than CAC/EUR/MCE.
  - AFR shows the weakest links to global markets; some association of lending growth and asset quality with domestic conditions.
- Institutional differences:
  - Banks and NBFIs: consistent positive and statistically significant links of asset growth with domestic and international markets; asset quality regressions show expected negative and significant signs against domestic economic conditions.
  - NGOs: expected negative signs for asset quality regressions, statistically significant mainly against international market conditions.
  - CUCs and rural banks: no strong empirical evidence of links with domestic and international capital market conditions (reasoning: mostly domestic funding and consumption-smoothing lending).
- Legal status, purpose, and age:
  - Non-regulated MFIs: consistent associations with domestic and international conditions; profitability and asset quality regressions statistically significant in all cases.
  - Regulated MFIs: asset and lending growth and asset quality regressions mostly show expected signs and statistical significance; profitability regressions largely not statistically significant except for MSCI Emerging Markets.
  - Non-profit MFIs: more consistently linked to domestic and international capital markets (asset growth and asset quality regressions significant).
  - For-profit MFIs: asset and lending growth regressions show positive and statistically significant differences relative to non-profit MFIs in Table 5; asset quality and profitability regressions often not statistically significant.
  - Mature MFIs: most consistent associations with domestic and international conditions (asset growth and asset quality significant); however, asset growth of mature MFIs is less responsive to changes in the economic environment than that of younger MFIs.
  - Young MFIs: asset growth positive and statistically significant in all cases; lending growth significant mainly for domestic conditions and emerging markets; asset quality coefficients generally not statistically significant.
- Interpretation and implications:
  - Microfinance activity is not detached from domestic economies and, in many subgroups, shows statistically significant correlations with international capital markets once the sample and period include the global financial crisis (1998-2009).
  - MFIs differ in their exposure: banks and NBFIs, CAC, EUR, and MCE MFIs exhibit higher sensitivity to domestic and international market conditions; AFR and some NGOs show lower sensitivity.
  - The extension of MFI funding and intermediation, lower capital-asset ratios, and larger loan sizes in some segments likely increase MFIs’ exposure to market conditions relative to earlier periods.

### Robustness of results and conclusions
- Results broadly robust to:
  - Extending period to 1995-2009 and to 2002-2009.
  - Alternative cross-sections including additional smaller MFIs (sample of 390 MFIs).
  - Alternative definitions of borrowing rates for lending rate and spread regressions (total financial cost divided by total borrowings).
- Overall conclusion:
  - For the period 1998-2009, including the global financial crisis, MFIs as a group show statistically significant links to domestic economic conditions and, for many groupings, to international capital markets; sensitivity varies by region, institution type, legal status, purpose, and age, with banks, NBFIs, CAC, EUR, and MCE being among the most exposed subgroups.

*Italic: Source — _wp11175 - Section VI discusses the findings and concludes.*

### 2009. However, the consistency of the relationship between MFI performance and domestic

### _wp11175 - 2009. However, the consistency of the relationship between MFI performance and domestic

### Strengthening linkages between MFIs and macro/international conditions
- Relationship between MFI performance and domestic economic conditions and international capital markets is stronger when the sample period is shrunk to 2002-2009 versus 1998-2009.
- Results are broadly similar but somewhat weaker when larger cross sections (including smaller MFIs) and longer time periods (1995-2009 and 1998-2009) are considered; weaker results reflect larger volatility introduced by smaller MFIs.
- As MFIs mature, their linkages with domestic and international market conditions get stronger.

### Determinants of lending rates: conceptual model
- Lending rates must cover: cost of funds for on-lending, cost of risk (expected loan loss), and administrative costs (searching, appraising, processing, disbursing, monitoring, collecting, follow-up).
- Administrative costs are not proportional to loan size; smaller average loan sizes imply a larger number of accounts and higher rates to break even.
- MFI portfolios are of significantly shorter duration and more volatile (mostly unsecured), so delinquency can be contagious and decapitalize MFIs more quickly.
- Productivity (borrowers per staff) and operational cost per staff matter: higher borrowers per staff reduce rates; higher operational cost per staff increases rates.
- MFI age is expected to reduce lending rates as lending technologies improve, staff gains experience, and access to funding and networks improves.
- Stable macroeconomic conditions (higher GDP growth) should, in principle, enable lower lending rates.

### Estimated models (fixed effects OLS)
- Two estimated specifications:
  - Lending rate it_l r depends on: borrowing rate (it_br), average loan size (it_sz), borrowers per staff (it_bs), operating cost per staff (it_os), capital-asset ratio (it_ka), portfolio-at-risk lagged (1it_par−), real GDP growth (it_y), and MFI age (it_age).
  - Absolute spread (it_spr) specification identical to above with 1iβ = 1.

### Key empirical findings (quantitative impacts preserved exactly as reported)
- Borrowing rate (it_br)
  - Coefficient positive and highly statistically significant for all MFIs and most groupings.
  - In the aggregate regression, a Wald test cannot reject the null that the coefficient equals one (lending rate and spread regressions similar).

- Average loan size (it_sz)
  - Coefficient negative and highly statistically significant for all MFIs and MFI groupings in lending rate regressions (spread regressions similar except NGOs).
  - Aggregate examples:
    - A US$100 increase in the average loan size would be associated with a decrease of lending rates (and spreads) of about 0.3 percentage points.
    - Depending on MFI purpose, legal status and age, the decrease would be in the range of 0.3 – 0.4 percentage points.
    - For MFIs in AFR: decrease of about 1.6 percentage points.
    - For MFIs in ASP: decrease of about 1.1 percent.
    - For MFIs in SAC: decrease of about 0.6 percentage points.
    - ASP region example: difference in loan sizes in 2009 between 80th and 20th percentiles was about US$ 500, which would explain a difference in lending rates of about 6 percentage points.

- Borrowers per staff ratio (it_bs)
  - Coefficient negative and highly statistically significant for all MFIs and most groupings (not significant for most institution types except banks).
  - Quantitative examples:
    - An increase of 50 borrowers per staff would imply a decrease in lending rates (and spreads) of about 1 percentage points.
    - Depending on MFI purpose, legal status and age, the decrease would be in range between 1 – 2 percentage points.
    - For CAC and EUR regions, the decrease could be as high as 3.5 percentage points.

- Operational cost per staff member (it_os)
  - Coefficient positive and highly statistically significant for all MFIs and most groupings (exceptions: banks and non-regulated MFIs).
  - Quantitative examples:
    - An US$1000 increase in operational expenditures per staff per year would cause an increase in lending rates (spreads) of about 0.3 percentage points.
    - Depending on MFI characteristics, the increase would be between 0.3 – 0.8 percentage points.
    - Coefficient larger for MFIs in AFR, MCE, and CAC regions.

- Portfolio-at-Risk ratio (lagged 1it_PAR−)
  - Coefficient positive and statistically significant across MFIs in most regions (either in lending rate or spread regressions), but not for many other groupings—asset quality linked more to location.
  - Quantitative example:
    - An increase of 1 percentage point in the ratio would result in an increase in lending rates (spreads) of about 0.2 – 0.3 percentage points, depending on location.

- GDP growth (it_y)
  - Coefficient negative and statistically significant only for EUR region; NBFIs; and regulated MFIs.
  - Quantitative example:
    - An increase of 1 percentage point in GDP growth would result in a decrease in spreads of about 0.1 – 0.3 percentage points depending on MFI characteristics.
  - Least consistent fundamental across MFI groupings.

- Capital-Asset ratio (it_ka)
  - Coefficient positive and statistically significant for most MFI groupings (not significant for the sample of all MFIs).
  - Quantitative example:
    - A 1 percentage point additional capital-assets ratio would result in an increase of lending rates (spreads) of between 0.2 – 0.4 percent, depending on MFI characteristics.

- Age (it_age)
  - Coefficient negative and statistically significant for all MFIs and most groupings.
  - Quantitative examples:
    - A MFI passing from “young” to “middle-aged” would result in a decrease of lending rates (spreads) of about 4 percentage points.
    - Going from “middle-aged” to “mature” would result in another 4 percentage point decrease.
    - The difference in lending rates (spreads) charged by a young MFI bank compared to a mature MFI bank would be of about 12 percentage points.

### Robustness
- Results robust to changes in sample period and to increases in cross-sections.
- Results stronger (more coefficients with expected sign and statistical significance) when period considered is 2002-2009, suggesting convergence of MFI behavior as they mature.
- Results somewhat weaker with larger MFI cross-sections due to higher variability from smaller MFIs.

### Summary implications for policy
- The microfinance industry is more strongly linked to domestic economic conditions and international capital markets than previously thought; systemic risk has increased.
- Systemic risk appears higher for MFIs operating in CAC and EUR regions; lower in ASP and SAC. Banks and NBFIs show higher systemic risk; NGOs lower. For-profit and younger MFIs show higher systemic risk than non-profit and mature MFIs.
- High observed lending rates reflect structural industry fundamentals (small loan sizes, short durations, higher administrative costs), MFI productivity, portfolio quality, capital structure, and maturity rather than solely excessive profiteering.
- Inappropriate regulation (e.g., blanket loan restructurings at below-market interest rates, or interest rate caps using commercial bank benchmarks) risks withdrawal of funding, reduction in MFI scale, increases in average loan sizes, and increased effective interest rates for the poor due to disintermediation toward higher-rate informal lenders.
- If the policy objective is to reduce interest rates, regulation should:
  - Aim to create an enabling market structure for MFIs to operate, test lending technologies, innovate to increase productivity and efficiency, and exert competitive pressure.
  - Protect customers’ rights and ensure MFI transparency and truth in lending, while avoiding outright inappropriate policies such as interest rate caps that ignore MFI cost structures.
  - Choose prudential versus non-prudential measures appropriately according to local industry characteristics and avoid over-regulation given regulatory costs.
  - Strengthen legal frameworks: generalize use of credit bureaus, improve corporate governance principles, and enforce timely reporting on a standard basis to reduce boom-bust dynamics and limit lending-rate spikes as portfolios deteriorate.

*Source: _wp11175 - 2009. However, the consistency of the relationship between MFI performance and domestic (PDF chapter/section).*

### REFERENCES

### _wp11175 - REFERENCES

### References
- Ahlin, C., J. Lin, and M. Maio, 2010, “Where does microfinance flourish? Microfinance Institution performance in macroeconomic context,” Journal of Development Economics (forthcoming).
- Banerjee, A., and E. Duflo, 2007, “The Economic Lives of the Poor,” Journal of Economic Perspectives, Vol. 21, No. 1, pp. 141-167.
- Caudill, S., D. Gropper and V. Hartarska, 2009, “Which Microfinance Institutions are Becoming More Cost Effective with Time? Evidence from a Mixture Model,” Journal of Money, Credit and Banking, Vol. 41, No. 4 (June).
- CGAP, 2000, “The Rush to Regulate,” Occasional Paper No.4, (April), Consultative Group to Assist the Poor, Washington, DC.
- CGAP, 2002, “Microfinance Consensus Guidelines. Guiding Principles on Regulation and Supervision of Microfinance,” Consultative Group to Assist the Poor, Washington, DC.
- CGAP, 2004, “Interest Rate Ceilings and Microfinance: The Story So Far,” Occasional Paper No. 9 (September), Consultative Group to Assist the Poor, Washington, DC.
- CGAP, 2009, ”The New Moneylenders: Are the Poor Being Exploited by High Microcredit Interest Rates?,” Occasional Paper No. 15 (February), Consultative Group to Assist the Poor, Washington, DC.
- CGAP, 2010, “Andhra Pradesh 2010: Global Implications of the Crisis in Indian Microfinance,” Focus Note No. 67 (November), Consultative Group to Assist the Poor, Washington, DC.
- Gonzalez, A., 2007, “Resilience of Microfinance Institutions to National Macroeconomic Events: An Econometric Analysis of MFI Asset Quality,” Working Paper, Microfinance Information Exchange (MIX), Washington, DC.
- Hermes, N., and R. Lensink, 2007, “The Empirics of Microfinance: What do we know?,” The Economic Journal, No. 117, (February).
- Krauss, N. and I. Walter, 2006, “Can Microfinance Reduce Portfolio Volatility?,” Working Paper, New York University.
- Krauss, N. and I. Walter, 2009, “Can Microfinance Reduce Portfolio Volatility?,” Economic Development and Cultural Change, Vol. 58, No. 1 (Oct., 2009), pp. 85-110.
- MIX, 2010a, “Is Microfinance Growing Too Fast?,” MIX Data Brief No. 5 (June).
- MIX, 2010b, “Analyzing Microcredit Interest Rates,” MIX Data Brief No. 4 (February).
- MIX, 2011, “Sacrificing Microcredit for Unrealistic Goals,” MicroBanking Bulletin (http://www.themix.org/publications/microbanking-bulletin/2011/01/sacrificing-microcredit-unrealistic-goals)
- Morduch, J., 1999, “The Microfinance Promise,” Journal of Economic Literature, Vol. XXXVII (December), pp. 1569-1614.
- Multilateral Investment Fund, 2010, “Microfinance Americas, The Top 100,” Interamerican Development Bank, Washington D.C.
- REDCAMIF, 2009, “Informe de Benchmarking the las Microfinanzas en Centroamerica,” Cuaderno No. 7, Red Centroamericana de Microfinanzas, Managua.
- United Nations, 2006, “Building Inclusive Financial Sectors for Development,” New York.

### Annex — Countries represented in each regional grouping
- Sub-Saharan Africa (Benin -5-, Burkina Faso -2-, Cameroon -3-, Congo (Democratic Republic) -1-, Republic of Congo -1-, Ethiopia -7-, Ghana -2-, Guinea -1-, Kenya -8-, Madagascar -5-, Mali -6-, Mozambique -1-, Malawi -2-, Niger -1-, Nigeria -2-, Rwanda -1-, Senegal -4-, Swaziland -1-, Chad -1-, Togo -2-, Tanzania -3-, Uganda -5-, South Africa -1-, Zimbabwe -2-).
- Asia-Pacific (Bangladesh -18-, Indonesia -3-, India -14-, Cambodia -10-, Sri Lanka -1-, Mongolia -2-, Nepal -9-, Philippines -22-, Thailand -1-, Vietnam -2-).
- Central America and the Caribbean (Costa Rica -2-, Dominican Republic -3-, Guatemala -4-, Honduras -5-, Haiti -4-, Mexico -3-, Nicaragua -13-).
- Eastern Europe (Albania -5-, Bulgaria -4-, Bosnia & Herzegovina -13-, Kosovo -5-, Former Yugoslav Republic of Macedonia -4-, Poland -2-, Romania -5-, Russia -5-, Serbia -4-, Slovenia -5-, Ukraine -2-).
- Middle East and Central Asia (Afghanistan -1-, Armenia -6-, Azerbaijan -5-, Egypt -6-, Georgia -5-, Jordan -4-, Kazakhstan -2-, Kyrgyz Republic -3-, Lebanon -1-, Morocco -7-, Pakistan -4-, West Bank & Gaza -3-, Tajikistan -2-, Tunisia -1-).
- South America (Argentina -1-, Bolivia -9-, Brazil -5-, Chile -1-, Colombia -9-, Ecuador -19-, Peru -18-, Paraguay -3-, Venezuela -1-).

### Table A1: Description of Data Set — Key panel statistics (selected entries)
- Overall time periods reported: 1998-2006 and 2007-2009.
- Metrics reported include: Assets (% increase), Lending (% increase), Constant Lending (% increase), Borrowing (% increase), PAR, Write-Offs (% portfolio), ROE.
- Example figures (aggregated blocks as presented):
  - 1998-2006: Assets (% increase) No. of observations 34; 4655.33 (Average No. of Observations); Median 4.8; Mean 38.8; 5 perc. -12.4; 95 perc. 105.2; SD 37.4.
  - 2007-2009: Assets (% increase) No. of observations 2006; 73.02 (Average No. of Observations); Median 22.2; Mean 27.2; 5 perc. -10.2; 95 perc. 81.; SD 430.9.
  - Additional region-specific blocks and institution-type breakdowns are presented throughout Table A1 with identical metric structure.

### Table A2: Description of Data Set — Operational and pricing indicators (selected entries)
- Indicators: Lending Rate (percent), Spread (percent), Loan Size (US$), Borrowers/staff ratio, Operational/staff (,000 US$), Capital/Asset ratio.
- Example figures (aggregated blocks as presented):
  - 1998-2006 (sample block): Lending Rate (percent) No. of observations 2333; Median 96.0; Mean 36.3; 5 perc. 42.8; 95 perc. 19.; SD 487.5; Spread (percent) No. of observations 2303; Median 95.9; Mean 29.2; 5 perc. 37.0; 95 perc. 16.1; SD 81.0.
  - 2007-2009 (sample block): Lending Rate (percent) No. of observations 1173; Median 93.0; Mean 27.7; 5 perc. 31.9; 95 perc. 18.9; SD 54.3.
  - Tables present parallel panels for Banks, Credit Unions/Cooperatives, Non-Bank Financial Institutions, Non-Government Organizations, Rural Banks with detailed counts and statistics.

### Table A3: Description of Data Set — Legal status (selected entries)
- Categories: Non-Regulated Institutions, Regulated Institutions.
- Example figures (aggregated blocks as presented):
  - Non-Regulated Institutions 1998-2006: Lending Rate (percent) No. of observations 66; Median 61.3; Mean 15.1; 5 perc. 38.1; 95 perc. 45.6; SD 21.5.
  - Regulated Institutions 1998-2006: Lending Rate (percent) No. of observations 1120; Median 21.3; Mean 15.3; 5 perc. 34.7; 95 perc. 38.; SD 417.0.
- All entries preserve the original No. of observations, No. of Institutions, Average No. of Observations, Median, Mean, 5 perc. (%), 95 perc. (%), and SD as reported.

*Source: Author's calculations and Microfinance Information Exchange (MIX).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11175.pdf_