## _wp09180

## Source details

**Canonical URL:** [_wp09180](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09180.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09180.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09180.pdf.json)

---

### I. Introduction: scope and structure
- Paper analyzes main drivers of credit growth in Sub-Saharan Africa (SSA) over the past decade and examines opportunities, challenges, and policy implications.
- Structure:
  - Document financial deepening during 1997–2007 and highlight post-2002 rapid credit growth in a select group of countries (Section II).
  - Examine contribution of demand and supply factors and catching-up effects to growth and level of bank credit to the private sector in SSA; use regression models to assess explanation of developments in rapid credit growth (RCG) countries (Section III).
  - Discuss prudential and macroeconomic risks of rapid credit expansion and policy choices to mitigate them (Section IV).
  - Evaluate likely impact of the ongoing global recession and financial crisis on credit growth in SSA (Section V).
  - Section VI summarizes main findings.

### II. Credit expansion in Sub-Saharan Africa: trends, RCG identification, and funding
- Stylized trends:
  - Median ratio of private sector credit to non-oil GDP in SSA increased from 11.5 to 18.6 percent over 1997–2007.
  - In 2003–07, a number of SSA countries experienced an average annual growth of real bank private sector credit of over 20 percent, leading to at least (near) doubling of the ratio of bank private sector credit to non-oil GDP.
  - Rapid credit growth (RCG) definition: average annual growth of real bank private sector credit of over 20 percent in 2003–07 and at least near doubling of the ratio of bank private sector credit to non-oil GDP.
- Countries identified as experiencing rapid credit growth:
  - Angola, Democratic Republic of Congo, Equatorial Guinea, Ghana, Guinea-Bissau, Liberia, Malawi, Nigeria, São Tomé and Príncipe, Sierra Leone, Swaziland, Tanzania, and Zambia.
- Context and common features:
  - RCG countries are geographically dispersed and include post-conflict countries, oil producers, heavy recipients of foreign aid, and macroeconomic success stories.
  - Common features: low initial level of financial development; acceleration of bank credit occurred after macroeconomic stabilization and with robust economic growth.
- Funding sources in RCG countries (ranked importance):
  - Primary: domestic deposits and their faster turnover into new loans (money multiplier).
  - Second: bank capital.
  - Third: borrowing from foreign banks.
  - Additional liquidity sources: significant external non-debt creating inflows such as oil revenue and aid.
- Balance-sheet decomposition (Table 1 highlights; percent contributions to growth of bank credit to the private sector, 2002–07):
  - Example country totals (Total = growth rate of credit to the private sector in 2002–07):
    - Angola: 412.0
    - Congo, Dem. Rep. of: -482.0
    - Equatorial Guinea: 5.9
    - Ghana: 12.8
    - Guinea-Bissau: 212.0
    - Liberia: -5.2
    - Malawi: -195.4
    - Nigeria: 26.6
    - São Tomé and Príncipe: -269.7
    - Sierra Leone: -106.6
    - Swaziland: -1.1
    - Tanzania: 8.8
    - Zambia: -11.6
  - Note: "..." indicates value of balance sheet item was zero in 2002.
  - Table algebra (as presented): Delta CRPS_t = sum_j [ (A_j,t-1 / CRPS_t-1) * Delta A_j,t ] - sum_i [ (L_i,t-1 / CRPS_t-1) * Delta L_i,t ].

### III. Drivers of credit growth — demand, supply, catching-up, and external factors
- Demand-side drivers:
  - Robust economic growth and falling nominal interest rates increased credit demand.
  - RCG SSA countries grew faster and experienced a bigger drop in interest rates than their peers.
- Supply-side drivers:
  - Banks’ high profitability and high system-wide capital adequacy ratios facilitated expansion without breaching prudential norms.
  - Improved liquidity conditions post-2002 lowered interest rates on bank liabilities through reduced country risk premia and low opportunity cost of funds for foreign parent banks through Q3 2007.
- Catching-up effects:
  - Low initial level of financial development (catching-up) contributed materially to high credit growth in RCG countries.
  - For every 1 percentage point difference between the initial ratio of bank credit to the private sector to non-oil GDP in two otherwise identical countries, the real growth of credit in the financially more developed country will be 1.1 percentage point lower than its peer (Model 7 long-run elasticity implication).
- Other contributors:
  - Sizable market shares of foreign banks with access to modern risk management and operational practices.

### IV. Regression approach, data, and estimation strategy
- Dependent variables:
  - Growth rate of real bank credit to the private sector (g1_rcrdprs, percent).
  - Ratio of bank credit to the private sector to non-oil GDP (crdprs_ngdp, percent).
- Key explanatory variables considered (Appendix II):
  - Macroeconomic demand factors: per capita GDP in thousand PPP USD (pcngdp_ppp) and the rate of interest (int_rat).
  - External financial flows: current account balance (b_ca_ngdp), current account transfers, various financial account inflows, claims of BIS-reporting foreign banks (frn_bnk_clm_nc).
  - Efficiency of financial intermediation: money multiplier (mmult).
  - Catching-up: lagged ratio of bank credit to the private sector to non-oil GDP.
  - Crowding-out: government balance (govbal_ngdp).
- Data sample and timeframe:
  - Sub-Saharan Africa excluding Guinea-Bissau, Liberia, and Zimbabwe due to data issues.
  - Time span: 1997-2007.
  - Number of countries: 43.
  - Observations: 430 (9 per country).
- Estimation strategy:
  - Exploratory pooled and within-group (fixed-effects) OLS initially.
  - Final estimation uses the System 2SLS estimator to address endogeneity and lagged dependent variable bias (instruments: second lags of the first two principal components of country-realizations of per capita GDP in PPP US$, the rate of interest, and the ratio of bank credit to the private sector to non-oil GDP — yielding 6 instruments per equation; instruments lagged by two periods).
  - Variables entered as ratios or in natural logs so coefficients can be interpreted as elasticities.
  - Preferred models arrived at by dropping variables insignificant at the 90 percent level; current account balance (in percent of non-oil GDP) dropped from final Models 7 and 19.

### V. Key empirical estimates and elasticities (preferred 2SLS Models)
- Growth of real bank credit (Model 7, 2SLS) — long-run elasticities:
  - A 1 percentage point increase in the growth rate of per capita GDP in PPP US$ is associated with a 0.6 percentage point increase in the growth rate of real bank credit to the private sector.
  - A 1 percentage point increase in the nominal interest rates is associated with a 0.6 percentage point decline in the growth rate of real bank credit to the private sector.
  - A 1 percentage point increase in the growth rate of the money multiplier is associated with a 0.09 percentage point increase in the growth rate of real bank credit to the private sector.
  - A 1 percentage point increase in the growth rate of claims of BIS-reporting foreign banks on domestic banks is associated with a 0.01 percentage point increase in the growth rate of real bank credit to the private sector.
  - For every 1 percentage point difference between initial credit ratios, the more developed country sees 1.1 percentage point lower real credit growth (catch-up effect).
- Ratio of bank credit to non-oil GDP (Model 19, 2SLS) — long-run elasticities:
  - A 1 percent increase in per capita GDP in PPP US$ is associated with a 9 percentage point increase in the ratio of bank credit to the private sector to non-oil GDP.
  - A 1 percentage point increase in the nominal interest rates is associated with a 0.08 percentage point decline in the ratio of bank credit to the private sector to non-oil GDP.
  - An increase in the money multiplier by 1 is associated with a 2 percentage point increase in the ratio of bank credit to the private sector to non-oil GDP.
  - A 1 percentage point increase in the ratio to non-oil GDP of claims of BIS-reporting foreign banks on domestic banks is associated with a 0.13 percentage point increase in the ratio of bank credit to the private sector to non-oil GDP.
  - Lagged dependent variable coefficient: 0.76 — implies slow adjustment, with approximately 24 percent of any gap closed in each subsequent year (long-run elasticities obtained by dividing 2SLS coefficients, excluding lagged dependent variable, by one minus the lag coefficient).
- Selected regression table highlights (System 2SLS coefficients and significance):
  - Growth regressions (Table 2):
    - Lagged bank credit to GDP: -1.076**, -1.084** (various specifications).
    - Growth rate of per capita GDP: 0.610*, 0.598*.
    - Interest rate: -0.586*, -0.591**.
    - Growth rate of money multiplier: 0.092, 0.092+.
    - Growth rate of claims of BIS-reporting foreign banks: 0.013+, 0.012*.
  - Level regressions (Table 3):
    - Lagged bank credit to GDP: 0.749**, 0.756**.
    - Per capita GDP in PPP USD (log): 2.526*, 2.214**.
    - Money multiplier: 0.492+, 0.505*.
    - Claims of BIS-reporting foreign banks in percent of non-oil GDP: 0.033, 0.032+.
  - Fit statistics for preferred models:
    - Models 7 (growth) and 19 (level) median R-squared: 0.21 and 0.50, respectively.
    - Within adjusted R-squared examples: growth model 0.21; level model variations 0.64/0.63.

### VI. Synthesis of empirical findings
- Primary drivers:
  - Bank credit to the private sector (both ratio to non-oil GDP and real growth) is primarily driven by macroeconomic factors: per capita GDP and nominal interest rates coefficients are sizable and signs align with theoretical priors.
- Catching-up:
  - Strong catching-up effect: countries with low initial financial development experience higher credit growth ceteris paribus; negative lagged-level coefficient in growth regressions supports this.
- Financial intermediation efficiency:
  - Money multiplier improvements have contributed materially to credit market developments; coefficients are statistically significant and non-negligible.
- External inflows and foreign bank claims:
  - Non-debt creating external financial inflows improve liquidity conditions but do not, by themselves, explain cross-country differences in credit market development (such inflow measures do not enter significantly in preferred regressions).
  - Claims of BIS-reporting foreign banks on domestic banks contributed to credit expansion but to a smaller extent than macro fundamentals and intermediation efficiency; coefficients are statistically significant but relatively small in magnitude.
- Equilibrium-consistent vs actual paths:
  - Fundamentals-consistent values (from long-run elasticities) explain much of post-2002 rapid credit growth, but excessive realizations in some countries pushed bank credit to non-oil GDP above levels supported by fundamentals; country-level analysis needed to explain deviations.

### VII. Policy-relevant implications (introduced)
- Opportunities:
  - Rapid credit growth increases credit availability, enabling consumption smoothing and firm investment.
  - Banks play a key role in allocating capital and monitoring under adverse selection and moral hazard, particularly in SSA where legal/judicial institutions and corporate governance may be weak.
- Risks and considerations (context from Boxes and discussion):
  - Adverse selection and moral hazard raise borrowing costs through risk premia, reduce availability of finance to high-quality borrowers, and can prevent socially desirable projects from being undertaken.
  - Addressing asymmetric information and aligning private and social payoffs are central to improving credit allocation and financial market efficiency.

### VIII. Appendices — data definitions, visualization notes, and additional regression output
- Appendix I — Key variables and units (selected examples):
  - g1_rcrdprs — Growth rate of real bank credit to the private sector (percent). Series: Domestic money bank claims to private sector; Unit: Billion dom. currency units; CPI index 2000 =100.
  - crdprs_ngdp — Bank credit to the private sector in percent of non-oil GDP; Unit: Billion dom. currency units.
  - pcngdp_ppp — Per capita GDP in thousand PPP USD; Unit: Billion PPP USD.
  - mmult — Money multiplier (ratio).
  - int_rat — Interest rate; Unit: Percent.
  - frn_bnk_clm_nc — Consolidated international claims of reporting banks on individual countries; Unit: USD million; converted to local currency using average exchange rate.
  - remit_nc, aid_nc — Workers remittances and official development assistance; Unit: USD million; converted using average exchange rate.
- Appendix II — Methodological notes and robustness:
  - Within-group (fixed-effects) OLS preferred at exploratory stage; final estimates use System 2SLS.
  - Sample excludes Guinea-Bissau, Liberia, and Zimbabwe; time span 1997-2007; typical N = 430 observations.
  - Preferred interim specifications: Model 8 (growth) and Model 20 (level) in Appendix; coefficients consistent with theory and significant at 90 percent level.
  - Significance legend: + significant at 10%; * significant at 5%; ** significant at 1%.
- Stylized visualization note:
  - Appendix Figure 1 presents box plots of SSA external financial flows, 1997–2007, including foreign aid, remittances, claims of foreign banks, and flow of external financial liabilities (all in % of non-oil GDP). Box-plot interpretation as 25th/75th percentiles and medians noted.

*Source: Excerpt from IMF working paper _wp09180 (References and adjacent sections) contained in the provided PDF content.*

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

### _wp09180 - References

### I. Introduction: scope and structure
- Paper analyzes main drivers of credit growth in Sub-Saharan Africa (SSA) over the past decade and examines opportunities, challenges, and policy implications.
- Structure:
  - Document financial deepening during 1997–2007 and highlight post-2002 rapid credit growth in a select group of countries (Section II).
  - Examine contribution of demand and supply factors and catching-up effects to growth and level of bank credit to the private sector in SSA; use regression models to assess explanation of developments in rapid credit growth (RCG) countries (Section III).
  - Discuss prudential and macroeconomic risks of rapid credit expansion and policy choices to mitigate them (Section IV).
  - Evaluate likely impact of the ongoing global recession and financial crisis on credit growth in SSA (Section V).
  - Section VI summarizes main findings.

### II. Credit expansion in Sub-Saharan Africa: trends, factors, and funding

- Trends and financial deepening:
  - Between 1997 and 2007, the median ratio of private sector credit to non-oil GDP in SSA increased from 11.5 to 18.6 percent.
  - In 2003–07, a number of SSA countries experienced an average annual growth of real bank private sector credit of over 20 percent, which has led to at least (near) doubling of the ratio of bank private sector credit to non-oil GDP.
  - Rapid credit growth (RCG) countries identified as having experienced an average annual growth of real bank private sector credit of over 20 percent in 2003–07, and near doubling of the ratio of bank private sector credit to non-oil GDP.

- Countries identified as experiencing rapid credit growth:
  - Angola, Democratic Republic of Congo, Equatorial Guinea, Ghana, Guinea-Bissau, Liberia, Malawi, Nigeria, São Tomé and Príncipe, Sierra Leone, Swaziland, Tanzania, and Zambia.

- Stylized context:
  - RCG countries are geographically dispersed and include post-conflict countries, oil producers, heavy recipients of foreign aid, and macroeconomic success stories.
  - Common features: low initial level of financial development; acceleration of bank credit occurred after macroeconomic stabilization and with robust economic growth.

- Factors driving observed credit expansion:
  - Demand-side drivers:
    - Robust economic growth and falling nominal interest rates increased credit demand.
    - RCG SSA countries grew faster and experienced a bigger drop in interest rates than their peers.
  - Supply-side drivers:
    - Banks’ high profitability and high system-wide capital adequacy ratios facilitated expansion of credit portfolios without breaching prudential norms.
    - Indirect evidence suggests improved liquidity conditions post-2002, lowering interest rates on bank liabilities through reduced country risk premia and low opportunity cost of funds for foreign parent banks through Q3 2007.
  - Other contributing factors:
    - Low initial level of financial sector development (catching-up effect).
    - Sizable market shares of foreign banks with access to modern risk management and operational practices.
  - Box 2 (Determinants of Credit Availability) summarizes theoretical drivers:
    - Credit supply positively related to rate of return on equity, disposable income, and prices of assets usable as collateral.
    - Credit demand negatively related to borrowing costs, positively related to current income, wealth, expectations about future income and asset valuations, and expected returns on capital.
    - Initial state of financial development affects observed credit expansion; countries with low initial financial development tend to experience higher growth rates of credit.

- Funding sources in RCG countries:
  - Primary funding source: domestic deposits and their faster turnover into new loans.
  - By end-2007, median money multiplier in RCG countries reversed from being lower than the rest of SSA at the start of the period to being higher.
  - Second most important funding source: bank capital.
  - Third: borrowing from foreign banks.
  - Liquidity further aided by significant external non-debt creating inflows, such as oil revenue and aid.

### III. Empirical / tabular evidence (selected findings)

- Median indicators and regional comparisons (as presented):
  - Median ratio of private sector bank credit to non-oil GDP (%) — increased from 11.5 to 18.6 percent over 1997–2007 in SSA.
  - Median growth of per capita GDP in PPP USD (%) — acceleration beginning in 2002.
  - Median Bank Return on Equity by Region (2003–2008) — Sub-Saharan Africa among highest regional medians.
  - Median Bank Regulatory Capital to Risk-Weighted Assets by Region (2003–2008) — Sub-Saharan Africa among higher regional medians.

- Table 1. Contribution of Other Balance Sheet Items to Growth of Bank Credit to the Private Sector in Rapid Credit Growth SSA Countries, 2002–07 (percent)
  - Table columns: Assets (Foreign assets; Cash and net claims on monetary authority; Claims on government), Liabilities (Foreign liabilities; Total deposits excl. government deposits; Government deposits; Capital accounts; Other Items (net)), and Total (equal to the growth rate of credit to the private sector in 2002–07).
  - Country rows with exact table entries as provided:
    - Angola: 1990.2 -659.8 -1438.5 -833.3 205.9 2969.0 573.5 761.4 412.0
    - Congo, Dem. Rep. of: 1492.3 -1525.2 -470.3 -223.6 341.7 3206.6 244.2 400.8 -482.0
    - Equatorial Guinea: 232.9 -129.9 -487.0 4.6 61.0 560.7 139.1 78.5 5.9
    - Ghana: 466.9 -78.5 -173.4 -101.5 88.3 577.3 39.1 102.9 12.8
    - Guinea-Bissau: 148.6 -111.4 -192.1 ... 63.4 223.6 160.8 ... 212.0
    - Liberia: 288.1 -297.2 -179.9 16.5 17.9 589.2 1.2 145.7 -5.2
    - Malawi: 361.6 -17.8 4.9 -201.5 64.2 587.1 -1.6 121.7 -195.4
    - Nigeria: 408.3 -57.2 -33.4 -167.9 23.1 412.9 21.7 182.5 26.6
    - São Tomé and Príncipe: 1286.0 -786.2 -524.8 ... 956.1 1149.0 27.6 734.1 -269.7
    - Sierra Leone: 415.8 -569.8 -82.3 -202.3 ... 1038.7 70.0 268.2 -106.6
    - Swaziland: 237.8 0.4 -19.8 -3.8 -0.5 173.9 11.4 77.3 -1.1
    - Tanzania: 379.2 -61.5 -91.3 -169.0 47.5 512.3 35.3 97.1 8.8
    - Zambia: 454.0 -109.2 -159.7 -151.8 105.2 627.2 72.8 81.1 -11.6
  - Source noted: IMF, International Finance Statistics and staff estimates.
  - Note in table: "..." indicates value of balance sheet item was zero in 2002.

- Methodological note (table algebra):
  - Table entries for assets and liabilities are computed using the expression provided:
    - Delta CRPS_t = sum_j [ (A_j,t-1 / CRPS_t-1) * Delta A_j,t ] - sum_i [ (L_i,t-1 / CRPS_t-1) * Delta L_i,t ]
    - (Expression appears in source in symbolic form; reproduced as presented.)

### IV. Identification of RCG episodes and context
- Box 1: Identifying episodes of rapid credit growth
  - Contrasts with credit boom literature which defines booms as extreme deviations from trend; this paper does not make an a priori distinction between unsustainable booms and fundamentals-driven expansions.
  - Selection rule: average annual growth of real bank private sector credit of over 20 percent in 2003–07 and at least near doubling of the ratio of bank private sector credit to non-oil GDP.
  - Notes synchronization of credit accelerations across countries; possible global triggers include rapid increase in international prices of SSA main export commodities and more than three-fold increase in private capital flows to SSA since 2003.
  - Source for data in box: IMF International Financial Statistics and WEO databases; commodity price indices shown with 2000=100.

### V. Transition to regression analysis
- Section III (headlined in source) begins regression analysis of credit availability and credit growth using developed models (details and additional regression outputs are provided in appendices referenced in the document).

*Source: Excerpt from the IMF paper (References and adjacent sections) contained in the provided PDF content.*

### 6.      We next examine the contribution of demand and supply factors and catch-up

### 6.      We next examine the contribution of demand and supply factors and catch-up effects to the level and growth of bank credit to the private sector in Sub-Saharan Africa

### Regression approach and data
- Dependent variables:
  - Growth rate of real bank credit to the private sector.
  - Ratio of bank credit to the private sector to non-oil GDP.
- Explanatory factors considered (exploratory stage, Appendix II):
  - Macroeconomic demand factors: per capita GDP in thousand PPP USD and the rate of interest.
  - External financial flows (proxies of liquidity and bank funding costs): current account balance, current account transfers, various financial account inflows.
  - Efficiency of financial intermediation: the money multiplier.
  - Catching-up effects: lagged ratio of bank credit to the private sector to non-oil GDP (in the credit growth regression).
  - Crowding-out effects: the government balance.
- Variable transformations:
  - All variables entered either as ratios or in natural logs so coefficients can be interpreted as elasticities.
- Estimation strategy:
  - Exploratory pooled and within-group (fixed-effects) OLS were used initially.
  - Final estimation uses the System 2SLS estimator (implemented as in Iossifov, Cihák, and Shanghavi (2008)) to address endogeneity and lagged dependent variable bias.
  - Instruments: second lags of the first two principal components of country-realizations of per capita GDP in PPP US$, the rate of interest, and the ratio of bank credit to the private sector to non-oil GDP (yielding 6 instruments per equation; there are 9 annual observations per country). Instruments lagged by two periods to ensure exogeneity.
- Sample and timeframe:
  - Sample: Sub-Saharan Africa excluding Guinea-Bissau, Liberia, and Zimbabwe due to data issues.
  - Time span: 1997-2007.
  - Number of countries: 43.
  - Observations: 430 (9 per country).

### Model fit and specification selection
- Preferred models arrived at by dropping variables insignificant at the 90 percent level; current account balance (in percent of non-oil GDP) was dropped from final Models 7 and 19.
- Fit statistics for preferred models:
  - Models 7 (growth regression) and 19 (level regression) have median R-squared of 0.21 and 0.50, respectively.
  - Within adjusted R-squared reported in tables (examples): within adj. R-squared 0.21 (growth model), within adj. R-squared 0.64/0.63 (level model variations).

### Key estimated long-run elasticities — Growth of real bank credit (Model 7, 2SLS)
- A 1 percentage point increase in the growth rate of per capita GDP in PPP US$ is associated with a 0.6 percentage point increase in the growth rate of real bank credit to the private sector.
- A 1 percentage point increase in the nominal interest rates is associated with a 0.6 percentage point decline in the growth rate of real bank credit to the private sector.
- A 1 percentage point increase in the growth rate of the money multiplier is associated with a 0.09 percentage point increase in the growth rate of real bank credit to the private sector.
- A 1 percentage point increase in the growth rate of claims of BIS-reporting foreign banks on domestic banks is associated with a 0.01 percentage point increase in the growth rate of real bank credit to the private sector.
- For every 1 percentage point difference between the initial ratio of bank credit to the private sector to non-oil GDP in two otherwise identical countries, the real growth of credit in the financially more developed country will be 1.1 percentage point lower than its peer.

### Key estimated long-run elasticities — Ratio of bank credit to non-oil GDP (Model 19, 2SLS)
- A 1 percent increase in per capita GDP in PPP US$ is associated with a 9 percentage point increase in the ratio of bank credit to the private sector to non-oil GDP.
- A 1 percentage point increase in the nominal interest rates is associated with a 0.08 percentage point decline in the ratio of bank credit to the private sector to non-oil GDP.
- An increase in the money multiplier by 1 is associated with a 2 percentage point increase in the ratio of bank credit to the private sector to non-oil GDP.
- A 1 percentage point increase in the ratio to non-oil GDP of claims of BIS-reporting foreign banks on domestic banks is associated with a 0.13 percentage point increase in the ratio of bank credit to the private sector to non-oil GDP.
- Lagged dependent variable coefficient: 0.76 — implies slow adjustment of actual ratio to its fundamentals-consistent value, with approximately 24 percent of any gap closed in each subsequent year (long-run elasticities obtained by dividing 2SLS coefficients, excluding lagged dependent variable, by one minus the lag coefficient).

### Regression table highlights (selected)
- Table 2 (growth regressions) reports:
  - Lagged bank credit to GDP coefficients in System 2SLS: -1.076**, -1.084** (various specifications).
  - Growth rate of per capita GDP coefficients in System 2SLS: 0.610*, 0.598*.
  - Interest rate coefficients in System 2SLS: -0.586*, -0.591**.
  - Growth rate of money multiplier in System 2SLS: 0.092, 0.092+.
  - Growth rate of claims of BIS-reporting foreign banks in System 2SLS: 0.013+, 0.012*.
- Table 3 (level regressions) reports:
  - Lagged bank credit to GDP coefficients in System 2SLS: 0.749**, 0.756**.
  - Per capita GDP in PPP USD (log) in System 2SLS: 2.526*, 2.214**.
  - Money multiplier in System 2SLS: 0.492+, 0.505*.
  - Claims of BIS-reporting foreign banks in percent of non-oil GDP in System 2SLS: 0.033, 0.032+.
- Notes from tables:
  - Legend: + - coefficient significant at 10%; * - significant at 5%; ** - significant at 1%.
  - Estimations performed using STATA (StataCorp, 2008).

### Synthesis of findings
- Primary drivers:
  - Bank credit to the private sector (both ratio to non-oil GDP and real growth) is primarily driven by macroeconomic factors. Coefficients on nominal interest rates and per capita GDP in PPP US$ are sizable and signs align with theoretical priors.
- Catching-up effects:
  - Strong catching-up: countries with low initial financial development experience higher credit growth ceteris paribus; lagged ratio coefficient in growth regression consistent with this hypothesis.
- Financial intermediation efficiency:
  - Improvements (measured by the money multiplier) have contributed materially to credit market developments; money multiplier coefficients are statistically significant and non-negligible.
- External inflows:
  - Non-debt creating external financial inflows (current and financial account) improve liquidity conditions but do not, by themselves, explain cross-country differences in credit market development — none of these inflow measures enter significantly in preferred regressions.
- Foreign bank lending:
  - Claims of BIS-reporting foreign banks on domestic banks contributed to credit expansion but to a smaller extent than other factors; coefficients are statistically significant but relatively small in magnitude.

### Equilibrium-consistent vs. actual credit paths
- Method:
  - Fundamentals-consistent values of the two credit measures obtained using long-run elasticities from 2SLS estimates of Models 7 and 19 and country-specific intercepts (predictions and observed values have same mean over 1997-2007).
- Result:
  - Post-2002 rapid credit growth is, to a large extent, explained by fundamentals, but excessive realizations in some countries have pushed bank credit to non-oil GDP above levels supported by fundamentals.
- Note:
  - Further country-level analysis (outside scope of paper) is needed to explore country-specific explanations for deviations.

### Policy-relevant implication introduced
- Rapid credit growth presents opportunities: credit availability enables consumption smoothing and firm investment; banks play a key role in allocating capital and monitoring under adverse selection and moral hazard, especially in SSA with weak legal/judicial institutions and corporate governance.

*Source: IMF working paper section titled "6.      We next examine the contribution of demand and supply factors and catch-up effects to the level and growth of bank credit to the private sector in Sub-Saharan Africa" (text and tables provided).*

### Box 3. Adverse Selection and Moral Hazard in Financial Markets

### Box 3. Adverse Selection and Moral Hazard in Financial Markets

### Adverse selection
- Borrowers typically know more about the quality of their investment opportunities than potential lenders due to the high cost of collecting information on the riskiness of projects by creditors.
- Potential lenders cannot distinguish high-risk from low-risk projects and they demand a risk premium above the risk-free rate of return on all their investments that corresponds to the average risk of all projects.
- The excess risk premium imposed on high-quality borrowers offsets the losses incurred in funding high-risk investment projects.
- As a result, the highest quality borrowers may be priced out of the market.
- Footnote 1: In the presence of asymmetric information on the side of borrowers (or shareholders) and creditors (or equity issuers), adverse selection could arise prior to the act of borrowing (or investing), whereas moral hazard could arise following the act of borrowing (or investing).

### Moral hazard
- Once external financing is secured, borrowers have an incentive to break the terms of their contracts with investors and use the funds for riskier projects than agreed upon.
- When the amount of external financing exceeds the value of firm’s own capital, there is a positive discrepancy between the expected private payoff from the project that accrues to the borrower and the expected social payoff that accrues to all parties involved.
- As a result, the debtor will always prefer to invest in projects with high expected private payoff, even though their social payoffs might be lower than the social payoff of an alternative low-risk project.
- Knowing this, lenders require an additional risk-premium on the cost of their funds to compensate them for the risks associated with moral hazard.
- The higher interest rates in the presence of moral hazard prevent firms from pursuing otherwise economically viable investment projects.
- Footnote 2: The expected private payoff is the weighted average (with weights equal to the expected probabilities of the two realizations) of the payoff in case of success and the loss incurred by the borrower, which is bounded by the value of his own capital, in case of failure. The expected social payoff is the weighted average of the payoff in case of success and the loss (up to the full amount invested) in case of failure.

### Key implications for financial markets and policy
- Adverse selection and moral hazard operate at different stages: adverse selection prior to borrowing/investing, moral hazard after borrowing/investing (see Footnote 1).
- Both problems raise borrowing costs through risk premia, reduce the availability of finance to high-quality borrowers, and can prevent socially desirable projects from being undertaken.
- Addressing asymmetric information and aligning private and social payoffs are central to improving credit allocation and financial market efficiency.

*Source: Sundararajan, Das, and Yossifov, 2003.*

### Chapter IV, (April), available via internet at:

### _wp09180 - Chapter IV, (April), available via internet at:

### Major references cited
- Chapter IV, (April), available via internet at: www.imf.org/external/pubs/ft/weo/2004/01/pdf/chapter4.pdf
- Iossifov, Plamen, Martin Cihák, and Shanghavi, Amar, 2008, “Interest Rate Elasticity of Residential Housing Prices,” IMF Working Paper, 08/247
- Loukoianova, Elena and Amar Shanghavi, 2007, “Growth in Credit to the Private Sector in Selected Sub-Saharan African Countries,” in Ghana: Selected Issues, IMF Country Report No. 07/208
- McDonald, Calvin and Schumacher, Liliana, 2007, “Financial Deepening in Sub-Saharan Africa: Empirical Evidence on the Role of Creditor Rights Protection and Information Sharing,” IMF Working Paper, 07/203
- Mendoza, Enrique and Terrones, Marco, 2008, “An Anatomy of Credit Booms: Evidence From Macro Aggregates and Micro Data,” IMF Working Paper, 08/226
- OECD, 2009, Aggregate Aid Statistics, http://stats.oecd.org/wbos/Index.aspx?DatasetCode=ODA_RECIPIENT
- Sacerdoti, Emilio, 2005, “Access to Bank Credit in Sub-Saharan Africa: Key Issues and Reform Strategies,” IMF Working Paper, 05/166
- Schadler, Susan, Paulo Drummond, Flavio Nacif, Louis Kuijs, Zuzana Murgasova, and van Elkan, Rachel, 2005, Adopting the Euro in Central Europe: Challenges of the Next Step in European Integration, Occasional Paper No. 234
- Singh, Raju, Kangni Kpodar, and Ghura, Dhaneshwar, 2009, “Financial Deepening in the CFA Franc Zone: The Role of Institutions,” IMF Working Paper No. 09/113
- StataCorp, 2008, STATA Statistical Software: Release 10
- Sundararajan, Vasudevan, Udaibir Das, and Yossifov, Plamen, 2003, “Cross-Country and Cross-Sector Analysis of Transparency of Monetary and Financial Policies,” IMF Working Paper, 03/94
- UNPD, 2007, World Population Prospects: The 2006 Revision
- WB, 2009, Remittances Data
- Yossifov, Plamen, 2002, “The Use of Credit Ceilings in the Presence of Indirect Monetary Instruments: An Analytical Framework,” IMF Working Paper, 02/206

### Appendix I — Variables and data sources: key definitions and units
- g1_rcrdprs — Growth rate of real bank credit to the private sector (percent). Series: Domestic money bank claims to private sector; Unit: Billion dom. currency units (Dom. curr.units); Source: IMF (2009b). CPI: Index (2000 =100); Source: IMF (2009a).
- crdprs_ngdp — Bank credit to the private sector in percent of non-oil GDP. Series: Domestic money bank claims to private sector; Unit: Billion dom. currency units; Source: IMF (2009b). Non-oil GDP current prices: Billion dom. currency units; Source: IMF (2009a).
- pcngdp_ppp — Per capita GDP in thousand PPP USD. Series: Gross Domestic Product in PPP USD; Unit: Billion PPP USD; Source: IMF (2009).
- Total population — Million inhabitants; Source: UNPD (2007).
- mmult — Money multiplier (ratio). Series: Broad money and Reserve money; Unit: Billion dom. curr. units; Sources: IMF (2009a), IMF (2009b).
- g1_cpi — Inflation rate (percent). Series: Consumer price index; Index (2000 =100); Source: IMF (2009a).
- int_rat — Interest rate. Series: Lending rate; Unit: Percent; Source: IMF (2009b).
- govbal_ngdp — Government balance in percent of non-oil GDP. Series: General government balance; Unit: Billion dom. curr. units; Source: IMF (2009a).
- b_ca_ngdp — Current account balance in percent of non-oil GDP. Series: Current account balance; Unit: Billion US dollars; Source: IMF (2009a).
- frn_bnk_clm_nc — Consolidated international claims of reporting banks on individual countries (immediate borrower basis). Unit: USD million; Source: BIS (2009). Converted to local currency using Average exchange rate Dom. curr. units/1US$; Source: IMF (2009a).
- remit_nc — Workers remittances, compensation of employees, and migrant transfers, credit. Unit: USD million; Source: WB (2009). Converted using Average exchange rate Dom. curr. units/1US$; Source: IMF (2009a).
- aid_nc — Official development assistance. Unit: USD million; Source: OECD (2009). Converted using Average exchange rate Dom. curr. units/1US$; Source: IMF (2009a).
- b_capinfl_lab_nc — Flow of external financial liabilities (billion dom. currency units). Conversion: Average exchange rate Dom. curr. units/1 US$; Source: IMF (2009a).

Notes:
- Lags denoted by suffixes “_#”. Annual growth rates denoted by prefix “g1_”. Log transformation denoted by prefix “ln_”.
- In case of missing observations, inflation data from the same source are used.

### Appendix I — Stylized data visualization note
- Appendix Figure 1 presents box plots of Sub-Saharan Africa external financial flows, 1997–2007, for:
  - foreign aid in % of non-oil gdp
  - inward remittances in % of non-oil gdp
  - claims of foreign banks in % of non-oil gdp
  - flow of external financial liabilities in % of non-oil gdp
- Note on reading box plots: lower and upper hinges are 25th and 75th percentiles; line in box indicates medians; whiskers mark next adjacent values. Source: See Appendix Table 1.

### Appendix II — Additional regression output: methodology and main findings
- Estimation approach:
  - Within-group (fixed-effects) OLS used at exploratory stage.
  - Within-group preferred over random effects because random effects are biased when country-specific error components correlate with explanatory variables.
  - Sample: Sub-Saharan Africa excluding Guinea-Bissau, Liberia, and Zimbabwe due to data issues.
  - Time span: 1997-2007.
  - Number of countries in tables: typically 43 (with some models listing 38 or 42 where noted).
- Model selection and specification:
  - Multivariate regression models built sequentially by adding fundamentals one at a time and keeping those with statistically significant coefficients (Appendix Tables 2 and 3).
  - Decision on entering interest rate: evidence favors entering interest rate in nominal terms rather than real terms because coefficients of inflation are negative and of lower magnitude than coefficients of the nominal interest rate; for a real interest rate specification the nominal interest rate and inflation coefficients would need similar magnitudes and opposite signs.
  - Models 8 and 20 are identified as preferred interim regression specifications for:
    - level relative to non-oil GDP (Model 20)
    - growth rate in real terms of bank credit to the private sector (Model 8)
  - All coefficients in preferred models have signs consistent with theoretical predictions and are statistically significant at the 90 percent level of confidence.
  - F-test of joint significance of all fixed effects (not shown) rejects a common intercept across countries, supporting longitudinal estimation methods.
  - Comment: First-difference transformation (growth rate regression) typically results in a drop in goodness-of-fit relative to level regressions.

### Appendix II — Key estimated coefficients and statistics (selected, as reported in Appendix Tables)
- Growth rate of real bank credit to the private sector regressions (selected coefficients from Appendix Table 2):
  - Bank credit to the private sector in percent of non-oil GDP (lagged): coefficients reported include -0.789**, -0.893**, -0.926**, -0.985**, -0.986**, -0.982**, -0.966**, -1.067**, -0.983**, -1.023**, -1.017** (standard errors reported in table).
  - Growth rate of per capita GDP in PPP USD (%): coefficients reported include 0.694**, 0.893**, 0.757**, 0.703**, 0.699**, 0.703**, 0.761**, 0.838**, 0.701**, 0.725**, 0.692** (standard errors reported in table).
  - Inflation (%): reported coefficient -0.094** (standard error in table).
  - Interest rate (%): reported coefficients include -0.401**, -0.535**, -0.557**, -0.536**, -1.140**, -0.551**, -0.557**, -0.543** (standard errors in table).
  - Growth rate of money multiplier (%): coefficients 0.074**, 0.072**, 0.070**, 0.087, 0.073**, 0.072**, 0.071** (standard errors in table).
  - Growth rate of claims of BIS-reporting foreign banks on country (%, in national currency): coefficients 0.013*, 0.014*, 0.011+, 0.013*, 0.014*, 0.014* (standard errors in table).
  - Common intercepts reported (examples): 24.858**, 8.078**, 22.591**, 24.919**, 32.856**, 35.171**, 35.069**, 34.788**, 44.675**, 35.029**, 35.715**, 35.487** (standard errors in table).
  - Observations across models: 430, 473, 430, 430, 430, 430, 430, 430, 346, 430, 420, 430 (as shown).
  - Within adjusted R-squared across sequential models: 0.03, 0.08, 0.10, 0.13, 0.16, 0.19, 0.20, 0.21, 0.20, 0.20, 0.20, 0.20.
- Bank credit to the private sector in percent of non-oil GDP regressions (selected coefficients from Appendix Table 3):
  - Bank credit (lagged): coefficient reported 0.818** and later model values 0.732**, 0.729**, 0.729**, 0.720**, 0.651**, 0.728**, 0.730**, 0.728** (standard errors reported).
  - Per capita GDP in PPP USD (log): reported coefficients include 8.721**, 3.191**, 3.104**, 2.908**, 3.566**, 4.927**, 2.854**, 2.963**, 2.717** (standard errors reported).
  - Money multiplier: coefficients reported 1.128**, 0.549**, 0.530**, 0.518**, 0.526**, 0.347*, 0.524**, 0.532**, 0.516** (standard errors reported).
  - Claims of BIS-reporting foreign banks on country in percent of non-oil GDP: coefficients reported 0.106**, 0.028**, 0.030**, 0.030**, 0.027**, 0.042**, 0.030**, 0.031**, 0.031** (standard errors reported).
  - Inflation (%): reported coefficient -0.01 (standard error in table).
  - Interest rate (%): reported coefficients -0.021+, -0.021+, -0.128**, -0.023+, -0.021+, -0.021+ (standard errors in table).
  - Current account balance in percent of non-oil GDP: reported coefficient -0.015+ (standard error in table).
  - Common intercepts reported (examples): 3.726**, 14.304**, 13.545**, 16.246**, 1.827*, 2.013**, 2.495**, 2.286**, 5.414**, 2.739**, 2.420**, 2.580** (standard errors in table).
  - Observations across models: 430, 473, 473, 473, 430, 430, 430, 430, 353, 430, 420, 430 (as shown).
  - Within adjusted R-squared examples: 0.60, 0.14, and other reported values corresponding to models in the table (table shows a sequence including 0.60, 0.14, 0.50, 0.10, 0.63, 0.63, 0.63, 0.64, 0.64, 0.63, 0.64, 0.63).
- Significance legend used in tables:
  - + coefficient significant at 10%
  - * significant at 5%
  - ** significant at 1%

### Key analytical takeaways
- Empirical regularities (from the sequential within-group regressions):
  - Lagged levels of bank credit to the private sector have strong negative association with subsequent real growth rates of credit (e.g., coefficients around -0.789** to -1.067** across specifications).
  - Real per capita GDP growth is positively associated with credit growth (e.g., coefficients around 0.694** to 0.893** across specifications).
  - Nominal interest rates enter significantly with negative coefficients in both growth and level regressions; evidence does not support transforming interest rates into real terms given coefficient magnitudes and signs for inflation and nominal interest rates.
  - Money multiplier growth and claims of foreign banks on country are positively associated with credit growth.
  - Cross-model goodness-of-fit is higher for level (relative to non-oil GDP) regressions than for first-difference (growth) regressions, consistent with first-difference reducing fit.
- Model robustness and estimation choice:
  - Fixed-effects (within-group) estimation strongly supported by F-tests (rejecting common intercept).
  - Preferred interim specifications: Model 8 for growth regression; Model 20 for level regression — both with coefficients consistent with theory and statistically significant at the 90 percent level.

*Source: _wp09180 - Chapter IV, Appendix Tables and Figures, as provided.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09180.pdf_