## _wp05240

## Source details

**Canonical URL:** [_wp05240](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05240.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05240.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05240.pdf.json)

---

### I. INTRODUCTION
- Purpose:
  - Evaluate the impact of far-reaching financial sector reforms in a low-income country (Uganda) on banking competition and efficiency over the 1999–2004 period.
  - Draw lessons for the Ugandan banking system and LICs more generally.
- Context:
  - Concern that state-dominated, monopolistic, inefficient, and fragile banking systems in many LICs, especially in SSA, hinder economic development.
- Key summary findings:
  - "The level of competition has significantly increased following the privatization of UCB and has been associated with a rise in banking efficiency."
  - "Smaller banks seem to have recently fallen back in efficiency relative to larger banks."
  - "On average, larger banks and foreign-owned banks are more efficient than others."
  - Interpretation: Privatization of dominant state-owned commercial banks (including foreign-owned banks), combined with cleanup and strengthened supervision, can enhance financial stability, access, and efficiency—subject to caveats about sample size and market equilibrium assumptions.

### II. BANKING SECTOR DEVELOPMENTS
- Liberalization and supervisory timeline:
  - 1992: All interest rates market-determined, including treasury bill yields.
  - 1993: New financial institutions bill and central bank charter; clarified Bank of Uganda (BOU) role.
  - 1994: Rationalization of branches and personnel of two state-owned banks.
  - 1995: Two banks intervened and recapitalized with subsidized long-term loans from BOU.
  - 1995: Non-Performing Assets Recovery Trust (NPART) established to recover over U Sh 60 billion (US$34 million); U Sh 28 billion recovered by June 2003.
  - 1998–1999: Four banks accounting for 12 percent of total system deposits were intervened and closed.
- Deposit protection and failures:
  - Deposit Insurance Fund (DIF) protected deposits up to 3 million Uganda shillings (about US$2,000 in 1999); government fully paid deposits in some failed banks and later ceased payments.
  - UCB (accounting for 22 percent of total system deposits) declared insolvent; attempted privatization in 1990s failed due to irregularities.
- Post-reform principal actions:
  - Privatization of UCB in September 2002 identified as key reform.
  - Closure of distressed banks and strengthened supervision; introduction of risk-based supervision and passage of the Financial Institutions Act (FIA), 2004 (described as conforming to international standards).

- Balance-sheet and income structure (2002–2004; asset structure as of 2004/4Q):
  - Asset shares:
    - Treasury bills: 28%
    - Loans: 27%
    - Deposits abroad: 21%
    - Cash and balance with BOU: 11%
    - Fixed assets: 5%
    - Other assets: 8%
  - Maturity structure of assets:
    - Less than 6 months: 64%
    - More than 6 months: 36%
  - Sectoral loan distribution:
    - Other services: 40%
    - Manufacturing: 23%
    - Trade & commerce: 15%
    - Agriculture: 11%
    - Transport: 7%
    - Building & construction: 4%
  - Income growth from 2002 to 2004 (USH Mil% and percent change):
    - Total income + 43,155 + 57.0%
    - Interest on loans + 17,284 + 74.4%
    - Interest on T bills + 7,590 + 33.4%
    - Charges and fees + 19,795 + 408.7%
  - Income distribution, 2004/4Q:
    - Interest on loans: 34%
    - Interest on T bills: 26%
    - Charges and fees: 20%
    - Forex: 10%
    - Other interest income: 4%
    - Others: 6%

- Asset quality and profitability:
  - NPLs fell from 29 percent in 1998 to 12 percent in 1999 and further to 2.6 percent at end-September 2004.
  - Note: "The NPL ratio rose through end-2003 because of the default of a large trading company. Changes in NPLs can be relatively large because of the concentration of exposures."
  - High interest rate margins and reduction in NPLs underpinned banks’ profitability.

### III. BOX 1 — PRIVATIZATION OF UCB AND STANBIC’S PERFORMANCE
- Market position and outreach:
  - Stanbic acquired UCB and did not close any of the original 68 branches.
  - Geographic coverage: widest in Uganda.
  - Market shares (tier 1–3 institutions): 28 percent of loans and 29 percent of deposits.
  - Market shares in the < 3 million size category: 28 percent of loans and 29 percent of deposits.
  - Net increase of 150,000 deposit accounts (reversing negative trends of 1999–2002).
  - Minimum opening balance reduced to U Sh 10 thousand (only one other tier 1 bank has this: CERUDEB).
  - Introduced ATMs; reduced check clearing and transfer delays within network.
  - Established business development unit for SME loans and increased agricultural lending (commercial crops).
  - Reported growth of 55 percent from 2002 to 2004.
  - Capital increased in December 2003; profitability aligned with composite tier 1.
  - Stanbic’s share of tier 1 investments in T-bills declined from 46 percent to 39 percent as lending increased.
  - Perception: dynamism and modern techniques can have demonstration effect; concurrent concerns about substantial market dominance by a single bank (e.g., large share of T-bills).

- Institutional performance:
  - CAMEL1 ratings for Stanbic far superior to UCB; clear progress across Capital, Asset quality, Management, Earnings, Liquidity.

- Income structure and liquidity:
  - Shift: lower dependence on government securities income; higher income from fees and charges.
  - Sector asset allocation (September 2004): government securities 29 percent; placed abroad 22 percent.
  - Combined share of government securities and placed abroad in September 2004: roughly half of total sector assets.
  - Combined share in 2000: 63 percent.
  - Funds placed abroad mainly deposits in correspondent accounts of large foreign-owned banks.

- Maturity, concentration, and product-range constraints:
  - Maturities:
    - Only 12 percent of total loans
    - 35 percent of loan volume (text distinction preserved as supplied)
    - 17 percent of total deposits
    - Less than 0.4 percent of time deposits have maturity > 1 year.
  - Concentration:
    - Loans to top five borrowers (aggregate): about 24 percent of total loans.
    - Top five depositors: about 21 percent of total deposits.
  - Branch coverage: branches exist in "about of 50 of the 55 districts" (text as supplied); population per branch about 90 thousand.
  - Total deposit accounts in commercial banks in 2004: about 1.5 million (about 30 percent of households).
  - Share of bank credit to agriculture: about 11 percent.
  - Single-borrower capacity of system: U Sh 66 billion.
  - Single largest borrowing capacity of any one bank: U Sh 14 billion.
  - Development Finance Institutions: UDB insolvent; minor role.

- Interest rate spreads and decomposition:
  - Interest rate spreads: about 20 percentage points.
  - Composition:
    - Operating costs explain about 9 percentage points.
    - Loan loss provisions explain about 2 percentage points.
    - Profit component: 30 percent of overall spread.
  - Table 2 (selected exact values):
    - Ugandan Banks:
      - Lending rate 21.7
      - Deposit rate 2.2
      - Total spread: 19.5
      - Overhead costs 9.0
      - Loan loss provisions 2.1
      - Reserve requirements + deposit insurance premium 0.4
      - Tax 2.3
      - Profit margin 5.7
    - Kenyan Private Banks:
      - Lending rate 17.5
      - Deposit rate 3.4
      - Total spread: 14.1
      - Overhead costs 5.1
      - Loan loss provisions 1.7
      - Reserve requirements + deposit insurance premium 0.4
      - Tax 2.1
      - Profit margin 4.8

### IV. CROSS-COUNTRY BENCHMARKS (TABLE 1 HIGHLIGHTS)
- Financial intermediation across countries, 2003 (In percent) — selected entries:
  - Uganda: Private Credit/GDP 7.0; Bank Deposits/GDP 19.6; Loan-Deposit Ratio 42.1; Overhead Costs 7.9
  - Tanzania: Private Credit/GDP 6.8; Bank Deposits/GDP 22.2; Loan-Deposit Ratio 40.9; Overhead Costs 7.0
  - Kenya: Private Credit/GDP 22.6; Bank Deposits/GDP 42.9; Loan-Deposit Ratio 60.1; Overhead Costs 6.1
  - SSA: Private Credit/GDP 19.1; Bank Deposits/GDP 31.3; Loan-Deposit Ratio 74.2; Overhead Costs 6.1
  - LICs: Private Credit/GDP 15.0; Bank Deposits/GDP 30.7; Loan-Deposit Ratio 70.0; Overhead Costs 5.9
- Definitions:
  - Private credit-to-GDP: total claims of financial institutions on the domestic private nonfinancial sector as a share of GDP.
  - Bank deposits/GDP: total deposits in deposit money banks as a share of GDP.
  - Loan-deposit ratio: aggregate lending to the private sector / total deposits for deposit money banks.
  - Overhead costs: banks’ operating costs relative to total earning assets.

### V. EMPIRICAL APPROACH, COMPETITION, AND CAUTIONS
- Empirical approach:
  - Panzar and Rosse (PR) methodology applied to a quarterly panel of 15 banks from March 1999 to June 2004.
  - Revenue equation specification (log-linear) with input price variables UPL, UPF, UPC and controls TA, RC1, RC2, dummies (FB, LB, SB, post-privatization), and macro controls NTBR and HINFL.
  - H-statistic defined as β1 + β2 + β3; H interpretation: < 0 collusive; < 1 monopolistic competition; =1 perfect competition.
- Caveats:
  - Small sample size and structural/regulatory changes complicate inference.
  - Equilibrium tests reported indicate system in equilibrium during most of period.

- Panzar and Rosse results (H-statistics and tests):
  - H-statistics (sample-wide and sub-periods):
    - Average entire sample (March 1999 to June 2004): 0.39
    - Pre-privatization period: 0.3
    - Post-privatization and consolidation period: 0.49
  - Table 3 reported Total Revenue specifications: 0.437, 0.311, 0.498; Total Revenue-Ratio: 0.359, 0.293, 0.505; Interest Revenue-Ratio: 0.376, 0.300, .460
  - Standard errors reported across specifications: .0457, 0.056, 0.052, 0.046, 0.0528, 0.050, 0.048, 0.060, .097
  - Statistical significance:
    - Increase in H-statistic post-privatization statistically significant at the 5 percent level in first two specifications and at the 10 percent level under interest-revenue specification.
    - Wald test rejects equality of pre- and post-privatization H-statistics.
  - Interpretation:
    - Ugandan banking sector characterized by monopolistic competition (H between 0 and 1).
    - Degree of competition increased after UCB privatization (September 2002).

- Selected coefficient findings (exact values reported across specifications):
  - UPL: 0.213**, 0.109*, 0.222**, 0.147**, 0.093*, 0.217**, 0.160**, 0.125**, 0.227**
  - UPF: 0.098**, 0.140**, 0.109*, 0.132**, 0.137**, 0.114*, 0.168**, 0.132**, 0.199**
  - UPC: 0.126**, 0.062**, 0.167*, 0.079**, 0.063**, 0.174**, 0.048*, 0.043, 0.033
  - TA: 0.964**, 1.067**, 0.940**
  - RC1: -0.022*, -0.051**, -.011, -0.023*, -0.053**, -0.010, -0.044**, -0.073**, -0.026
  - NTBR: 0.007**, 0.009**, 0.004, 0.007**, 0.009**, 0.003, 0.011**, 0.013**, 0.014**
  - HINFL: -0.001, 0.007, -.019*, -0.001, 0.005, -0.016*, -0.001, 0.006, -0.020**
  - SB (small bank dummy): -0.280**, -0.355*, -0.27**, -0.106, -0.050, -0.224
  - Sep2002 dummy: -0.001, -0.005, -0.122**

- Cross-country comparison (Table 4 excerpts):
  - Ghana: H statistic 0.56; Sample Period 1998 - 2003; Number of banks 13; Number of observations 65
  - Kenya: H statistic 0.58; Sample Period 1994 - 2001; Number of banks 34; Number of observations 106
  - Nigeria: H statistic 0.67; Sample Period 1994 - 2001; Number of banks 42; Number of observations 186
  - South Africa: H statistic 0.85; Sample Period 1994 - 2001; Number of banks 45; Number of observations 186
  - Uganda split:
    - 1999Q1 – 2002Q3: H statistic 0.30; Number of banks 15; Number of observations 196
    - 2002Q4 – 2004Q2: H statistic 0.49; Number of banks 15; Number of observations 97

- Authors’ interpretations from revenue functions:
  - UPL: personnel costs important; overhead costs relatively high.
  - UPF: higher impact on interest revenue post-privatization — better market responsiveness.
  - UPC: determinant of total revenue but not total interest revenue — reflects fee-based services like private money transfers and ATMs.
  - TA: size is a major determinant for total revenues; strong economies of scale exist.
  - RC1: negative effect mainly pre-privatization.
  - NTBR: positive effect pre-privatization; dissipated post-privatization.
  - SB: negative sign post-privatization — small banks vulnerable in a more competitive environment.
  - HINFL: negative effect on revenues post-privatization — linked to agriculture exposure and food-price-driven inflation.

### VI. EFFICIENCY AND PRODUCTIVITY MEASURES
- Methodology:
  - Technical efficiency under variable returns to scale measured via DEA (Data Envelopment Analysis) using intermediation approach (Sealey and Lindley, 1977).
  - Outputs: Deposits (y1), securities (y2), loans (y3).
  - Inputs: deposits (x1), loans and contingent liabilities (x2).
  - Panel: 14 banks, 20 quarterly observations from June 2004 back to 1999 Q3; panel size for regressions: 270. All data in end-2003 prices.
- Descriptive statistics (Table 5, in Ugandan shillings):
  - Outputs means:
    - y1 (Deposits): 43,274,886
    - y2 (Securities): 33,102,894
    - y3 (Loans): 38,511,183
  - Inputs means:
    - x1 (Deposits): 99,755,085
    - x2 (Loans and contingent liabilities): 20,031,790
- DEA efficiency scores (20-quarter summary; 14-bank sample):
  - Average efficiency scores varied between a minimum of 95 percent and a maximum of 100 percent over the 20-quarter period.
  - Overall (unweighted) mean: 99 percent.
  - Weighted by market share in deposits: overall mean 99 percent.
  - Minimum efficiency across periods ranged between 69 and 100 percent.
  - Frontier formed by between 10 and 14 banks across periods.
  - Five of the 14 banks were on the frontier (fully efficient) in each of the 20 time periods.
- Selected quarterly DEA results (exact values preserved):
  - Sep-99: Mean 0.99; Mean weighted 0.99; S.D. 0.05; Maximum 1.00; Minimum 0.82; Number below 1.000 = 12
  - Dec-99: Mean 1.00; Mean weighted 1.00; S.D. 0.00; Maximum 1.00; Minimum 0.99; Number below 1.000 = 13
  - Mar-01: Mean 0.95; Mean weighted 0.98; S.D. 0.09; Maximum 1.00; Minimum 0.74; Number below 1.000 = 10
  - Jun-02: Mean 1.00; Mean weighted 1.00; S.D. 0.00; Maximum 1.00; Minimum 1.00; Number below 1.000 = 14
  - Mar-04: Mean 0.96; Mean weighted 0.99; S.D. 0.08; Maximum 1.00; Minimum 0.78; Number below 1.000 = 10
  - Jun-04: Mean 0.99; Mean weighted 0.92; S.D. 0.04; Maximum 1.00; Minimum 0.84; Number below 1.000 = 13
  - Total summary: Mean 0.99; Mean weighted 0.99; Standard Deviation 0.05; Maximum 1.00; Minimum 0.70
- Visual/interpretive patterns:
  - High efficiency scores indicate relatively homogenous bank efficiency distribution.
  - Weighted mean more stable than unweighted mean — smaller banks had more volatile scores.
  - Increasing divergence between weighted and unweighted means suggests smaller banks have recently fallen back relative to larger banks.
- Robustness checks:
  - Alternative input specifications yield similar results.
  - Splitting sample into two periods and establishing a frontier per period:
    - First part (with UCB included): average efficiency 91.6 percent.
    - Second part: average efficiency 92.9 percent.
    - Interpretation: slight improvement in efficiency aligned with increased competition after UCB privatization.
  - Excluding seven smallest banks (assets < U Sh 100 billion at end-June 2004):
    - Average efficiency first period 93.9 percent; second period 96.8 percent.
    - Interpretation: efficiency improvement robust; smaller banks are less efficient.

- Productivity change (Malmquist index; Table 7 summary):
  - 20-Period Geometric Mean of Changes:
    - Change Relative to t-1 Frontier (A): 1.006
    - Frontier Shift from t-1 and t (B): 1.005
    - Productivity (=A*B): 1.011
  - Interpretation: Productivity growth of 1.1 percent by quarter, split about half (0.6 percent and 0.5 percent on average, respectively) between efficiency change and frontier shift.

### VII. EXPLAINING EFFICIENCY DIFFERENCES
- Estimation approach:
  - Efficiency scores pooled over 20 quarters; Tobit model used (dependent variable bounded between zero and unity).
  - Heteroskedasticity accounted for with QML (Huber/White) standard errors.
  - Sample size: N=270.
- Descriptive covariates (Table 8 summaries):
  - SIZE (Deposit liabilities): Mean 92,543,643; S.D. 1.36e+08
  - SCOPE (Herfindahl index of three largest asset categories): Mean 0.41; S.D. 0.09
  - FOREIGN (foreign-owned dummy): Mean 0.86; S.D. 0.35
  - ADMIN (Administrative expenses % of total assets): Mean 3.0; S.D. 3.3
  - PUBLIC (Credit to public sector % of total credit): Mean 42.4; S.D. 20.3

- Tobit regression results (Table 9; exact coefficients and significance):
  - SIZE: 8.31E-10; Standard Error 3.07E-10; z-Statistic 2.709270; Probability 0.0067** (significant at the 1 percent level)
  - SCOPE: 2.539935; Standard Error 0.173531; z-Statistic 14.63677; Probability 0.0000** (significant at the 1 percent level)
  - FOREIGN: 0.083253; Standard Error 0.039836; z-Statistic 2.089887; Probability 0.0366* (significant at the 5 percent level)
  - ADMIN: -0.013141; Standard Error 0.005938; z-Statistic -2.212885; Probability 0.0269* (significant at the 5 percent level)
  - PUBLIC: 0.005845; Standard Error 0.001806; z-Statistic 3.237557; Probability 0.0012** (significant at the 1 percent level)

- Key empirical findings:
  - Efficiency increases with SIZE.
  - Efficiency increases with portfolio diversification (higher SCOPE).
  - FOREIGN-owned banks are on average more efficient than local banks.
  - Efficiency decreases with higher administrative expenses (ADMIN).
  - Efficiency increases with the share of credit to the public sector (PUBLIC).

- Robustness and excluded variables:
  - Other specifications and univariate regressions produced similar signs and significance.
  - Variables tried but not significant: share of deposits in total funds; share of foreign currency deposits; year of establishment; proxy for risk aversion; GDP growth.

### VIII. POLICY-ORIENTED INSIGHTS AND CONCLUSIONS
- Main conclusions:
  - The Ugandan banking system became more competitive and efficient following reforms: cleanup, consolidation, privatization of state-owned banks, and strengthened regulatory and supervisory framework.
  - System still characterized by monopolistic competition with scope to enhance competitive pressures, particularly in fee-based services.
- Distributional/market concerns:
  - Smaller banks have tended to fall back in efficiency relative to larger and foreign-owned banks — consolidation and privatization can have distributional consequences.
- Competition and foreign ownership:
  - Concerns that foreign ownership and concentration stifle competition and access appear unfounded in Uganda’s case.
- Scale and treasury-bill dependence:
  - Scale matters substantially; small banks may come under pressure as competition builds, especially if treasury-bill revenue sources decline.
  - Reliance on government securities potentially crowded out private lending; effect may have dissipated recently.
  - Stress testing: even a 50 percent reduction in government securities yield would not severely damage small banks (as noted in the paper).
- Policy recommendations / areas to address (authors’ suggestions):
  - Reduce net treasury bill issuance to lessen banks’ dependence on government securities and encourage lending to private sector and customer-base expansion.
  - Address high overhead, personnel, and loan loss provisioning costs by improving infrastructure, labor market flexibility, and commercial courts/insolvency regime to facilitate financial intermediation.
  - Continue strengthened regulation and supervision (risk-based approaches) and modernize legal/institutional frameworks to lower credit risk and transaction costs.

### IX. APPENDIX — EQUILIBRIUM TESTS (SPECIFICATION AND SELECTED RESULTS)
- Equilibrium test specification:
  - In(ROA_it) = α + β1 In UPL_it + β2 In UPF_it + β3 In UPC_it + γ1 In TA_it + γ2 In RC1_it + γ3 In RC2_it + γ4 DV + ε_it
  - ε_it = μ_i + ν_it
  - H = β1 + β2 + β3 = 0
  - ROA measured as ln(1+ROA) to handle negatives.
  - Wald-test used for H=0.
- Interpretation:
  - Market equilibrium condition cannot be rejected at the 5 percent level for the post-privatization period and at the 10 percent level for the pre-privatization period.
  - Input prices generally not statistically different from zero across sub-periods, with possible exception of fixed capital input price.
  - Returns on assets positively associated with loan ratio in both periods.
  - Nominal interest rate, inflation and NPL ratio important for profit pre-privatization but not post-privatization; foreign banks associated with larger profitability in post-privatization period.
- Selected reported coefficients (Table 1 excerpts):
  - UPC: (1) -0.019**, (2) -0.012**, (3) -0.008*
  - TA: (1) 0.004, (2) 0.050**, (3) 0.004
  - RC1: (1) -0.003*, (2) -0.005**, (3) -0.001
  - RC2: (1) 0.009, (2) 0.095**, (3) 0.062**
  - NTBR: (1) 0.001*, (2) 0.001**, (3) 0.000
  - HINFL: (1) 0.000, (2) 0.002**, (3) -0.000
  - FB: (1) 0.006, (3) 0.010**
  - Constant: (1) -0.148*, (2) -1.03**, (3) -0.002
- Goodness-of-fit and tests (selected):
  - R-squared: (1) 0.32, (2) 0.20, (3) 0.30
  - Wald chi2 Prob > chi2 (H test): (1) 0.08, (2) 0.11, (3) 0.26
  - No. of Obs.: 307, 196, 97

*Source: IMF Working Paper content unit: _wp05240 (authors’ calculations and reported tables).*

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

### _wp05240 - References..............................................................................................................

### I. INTRODUCTION
- Purpose:
  - Evaluate the impact of far-reaching financial sector reforms in a low-income country (Uganda) on banking competition and efficiency over the 1999–2004 period.
  - Draw lessons for the Ugandan banking system and LICs more generally.
- Context and motivation:
  - Concern that state-dominated, monopolistic, inefficient, and fragile banking systems in many LICs, especially in SSA, hinder economic development.
  - Importance of identifying reforms and environments that promote competition and efficiency in LIC banking systems.
- Key summary findings:
  - "The level of competition has significantly increased following the privatization of UCB and has been associated with a rise in banking efficiency."
  - "Smaller banks seem to have recently fallen back in efficiency relative to larger banks."
  - "On average, larger banks and foreign-owned banks are more efficient than others."
  - Interpretation: Privatization of dominant state-owned commercial banks (including foreign-owned banks), combined with cleanup and strengthened supervision, can enhance financial stability, access, and efficiency—subject to caveats about sample size and market equilibrium assumptions.

### II. BANKING SECTOR DEVELOPMENTS
#### A. The Early Years
- Liberalization chronology and institutional change:
  - 1992: All interest rates allowed to become market-determined, including treasury bill yields.
  - 1993: New financial institutions bill and central bank charter enacted; clarified Bank of Uganda (BOU) role as regulator and supervisor.
- Supervisory capacity and outcomes:
  - Supervisory capacity initially weak due to understaffing of qualified bank inspectors; BOU made efforts to develop capabilities over time.
  - Growth in financial intermediation followed these measures.
- Sector cleanup and interventions:
  - 1994: Sharp rationalization of bank branches and personnel of two state-owned banks.
  - 1995: Two banks intervened and recapitalized with subsidized long-term loans from the BOU.
  - 1998–1999: Four more banks (accounting for 12 percent of total system deposits) were intervened and closed.
  - Deposit protection: deposits of up to a maximum of 3 million Uganda shillings (about US$2,000 in 1999) were protected by the Deposit Insurance Fund (DIF), but government fully paid deposits in some failed banks and later ceased payments.
  - UCB (accounting for 22 percent of total system deposits) was declared insolvent; an attempted privatization in the 1990s failed due to irregularities.
  - Non-Performing Assets Recovery Trust (NPART) established in 1995 to recover over U Sh 60 billion (US$34 million) in unpaid loans; U Sh 28 billion had been recovered by June 2003.

#### B. State of Play (Post-Reform)
- Principal reforms and their effects:
  - Privatization of UCB in September 2002 identified as the key reform to spur growth and reduce stability risks.
  - Closure of distressed banks and strengthened supervision, including introduction of a risk-based approach and passage of the Financial Institutions Act (FIA), 2004.
  - FIA, 2004 described as conforming to international standards.
- Balance-sheet and profitability developments (2002–2004):
  - Asset composition and growth (asset structure as of 2004/4Q):
    - Treasury bills: 28%
    - Loans: 27%
    - Deposits abroad: 21%
    - Cash and balance with BOU: 11%
    - Fixed assets: 5%
    - Other assets: 8%
  - Maturity structure of banks' assets:
    - Less than 6 months: 64%
    - More than 6 months: 36%
  - Sectoral breakdown of loans:
    - Other services: 40%
    - Manufacturing: 23%
    - Trade & commerce: 15%
    - Agriculture: 11%
    - Transport: 7%
    - Building & construction: 4%
  - Income growth from 2002 to 2004 (USH Mil%):
    - Total income + 43,155 + 57.0%
    - Interest on loans + 17,284 + 74.4%
    - Interest on T bills + 7,590 + 33.4%
    - Charges and fees + 19,795 + 408.7%
  - Income distribution, 2004/4Q:
    - Interest on loans: 34%
    - Interest on T bills: 26%
    - Charges and fees: 20%
    - Forex: 10%
    - Other interest income: 4%
    - Others: 6%
- Asset quality and profitability:
  - Nonperforming loans (NPLs):
    - NPLs fell from 29 percent of the portfolio in 1998 to 12 percent in 1999 and further to 2.6 percent at end-September 2004.
    - Note: "The NPL ratio rose through end-2003 because of the default of a large trading company. Changes in NPLs can be relatively large because of the concentration of exposures."
  - High interest rate margins and the marked reduction in NPLs have underpinned banks’ profitability.

### III. CROSS-COUNTRY BENCHMARKS (TABLE 1 HIGHLIGHTS)
- Financial intermediation across countries, 2003 (In percent) — selected entries exactly as presented:
  - Uganda: Private Credit/GDP 7.0; Bank Deposits/GDP 19.6; Loan-Deposit Ratio 42.1; Overhead Costs 7.9
  - Tanzania: Private Credit/GDP 6.8; Bank Deposits/GDP 22.2; Loan-Deposit Ratio 40.9; Overhead Costs 7.0
  - Kenya: Private Credit/GDP 22.6; Bank Deposits/GDP 42.9; Loan-Deposit Ratio 60.1; Overhead Costs 6.1
  - SSA: Private Credit/GDP 19.1; Bank Deposits/GDP 31.3; Loan-Deposit Ratio 74.2; Overhead Costs 6.1
  - LICs: Private Credit/GDP 15.0; Bank Deposits/GDP 30.7; Loan-Deposit Ratio 70.0; Overhead Costs 5.9
- Data sources: International Monetary Fund International Financial Statistics (IFS) and banks’ balance sheets.
- Definitions: Private credit-to-GDP is total claims of financial institutions on the domestic private nonfinancial sector as a share of GDP. Bank deposits/GDP is total deposits in deposit money banks as a share of GDP. Loan-deposit ratio is the aggregate ratio of lending to the private sector to total deposits for deposit money banks. Overhead costs are banks’ operating costs relative to total earning assets.

### IV. EMPIRICAL APPROACH AND CAUTIONS
- Empirical focus:
  - Systematically analyze the impact of reforms on banking competition and efficiency in a LIC (Uganda), applying models commonly used in industrial-country banking systems.
- Caveats and limitations:
  - Results must be read with caution because of the small sample size and market equilibrium conditions assumed (and tested).
  - Structural and regulatory changes in the sample period complicate pinning down empirical relationships.

### V. POLICY-ORIENTED INSIGHTS
- Reform elements associated with improved outcomes:
  - Cleanup of distressed banks and consolidation.
  - Privatization of dominant state-owned commercial banks.
  - Strengthened regulation and supervision, including risk-based supervision and modernized Financial Institutions Act (FIA), 2004.
- Expected benefits when combined:
  - Enhanced financial stability, increased access, and improved efficiency—conditional on effective regulatory and supervisory frameworks and the presence of competitive dynamics.
- Distributional/market concerns:
  - While overall efficiency and competition rose, smaller banks have tended to fall back in efficiency relative to larger and foreign-owned banks, suggesting possible distributional consequences of consolidation and privatization.

*IMF Working Paper content unit: _wp05240 - References..............................................................................................................*

### Box 1. Privatization of UCB and Stanbic’s Performance

### Box 1. Privatization of UCB and Stanbic’s Performance

### Stanbic’s market position and outreach
- Stanbic acquired UCB and has not closed any of the original 68 branches.  
- Geographic coverage: widest of any financial institution in Uganda.  
- Market shares (tier 1–3 institutions): 28 percent of loans and 29 percent of deposits.  
- Market shares in the < 3 million size category (lower end of banking market): 28 percent of loans and 29 percent of deposits.  
- Net increase of 150,000 deposit accounts (reversing negative trends of 1999–2002).  
- Minimum opening balance reduced to U Sh 10 thousand (a limit only one other tier 1 bank has: CERUDEB).  
- Introduced ATMs throughout branch network; reduced delays in check clearing and money transfers within Stanbic network.  
- Established a business development unit originating SME loans and increased agricultural lending (finance for growers of commercial crops).  
- Reported growth of 55 percent from 2002 to 2004.  
- Capital increased in December 2003 and profitability aligned with composite tier 1.  
- Stanbic’s share of tier 1 investments in T-bills declined from 46 percent to 39 percent as lending increased.  
- Perception: Stanbic’s dynamism and introduction of modern techniques/services have a demonstration effect that can enhance competition; concurrent concerns about substantial market dominance by a single bank (e.g., large share of T-bills).

### Institutional performance and CAMEL1 assessment
- Institutional performance of Stanbic, as measured by CAMEL1 ratings, is far superior to that of UCB; clear progress in all key CAMEL rating categories for the combined bank.  
- CAMEL1 definition: Capital, asset quality, management, earnings, and liquidity.

### Changes in banks’ income structure and liquidity
- Recent income-structure changes: lower dependence on government securities income and higher income from fees and charges.  
- Despite decline in income from government securities, profitability remained strong driven by growth in noninterest income and private sector lending.  
- Sector asset allocation (September 2004): government securities 29 percent; placed abroad 22 percent.  
- Combined share of funds in government securities and placed abroad in September 2004: roughly half of total sector assets.  
- Combined share in 2000: 63 percent.  
- Funds placed abroad largely comprise deposits in correspondent accounts of large foreign-owned banks.

### Maturity, concentration, and product-range constraints
- Maturities: only 12 percent of total loans, 35 percent of loan volume, 17 percent of total deposits, and less than 0.4 percent of time deposits have a maturity of more than one year.  
- Concentration: loans to the top five borrowers for each bank (aggregate) represent about 24 percent of total loans; top five depositors account for about 21 percent of total deposits.  
- Branch coverage: branches of commercial banks exist in about of 50 of the 55 districts (text supplied as "about of 50 of the 55 districts"), with a population per branch of about 90 thousand.  
- Total number of deposit accounts in commercial banks in 2004: about 1.5 million (about 30 percent of households).  
- Low use of loan services despite deposit coverage.  
- Share of bank credit allocated to agriculture: about 11 percent.  
- Single-borrower capacity of the Ugandan banking system: U Sh 66 billion.  
- Single largest borrowing capacity of any one bank: U Sh 14 billion.  
- Development Finance Institutions (DFIs) play a minor role; Uganda Development Bank (UDB) is insolvent with associated freeze on lending.

### Interest rate spreads and their decomposition
- Interest rate spreads (weighted average lending rate minus weighted average deposit rate): about 20 percentage points at present.  
- Composition of the spread: operating costs explain about 9 percentage points (about half of the spread).  
- Loan loss provisions explain another 2 percentage points.  
- High credit risk contributes to high overhead costs (evaluation, monitoring, enforcement). Causes of credit risk include lack of shared credit information, widespread fraud, dysfunctional land and company registries, and deficiencies in insolvency laws and administration.  
- Profit component: profits are the second largest component with 30 percent of the overall spread.  
- Table 2: Decomposition of Interest Spread (Ugandan Banks vs Kenyan Private Banks) 1/
  - Ugandan Banks:
    - Lending rate 21.7
    - Deposit rate 2.2
    - Total spread, of which: 19.5
    - Overhead costs 9.0
    - Loan loss provisions 2.1
    - Reserve requirements + deposit insurance premium 0.4
    - Tax 2.3
    - Profit margin 5.7
  - Kenyan Private Banks:
    - Lending rate 17.5
    - Deposit rate 3.4
    - Total spread, of which: 14.1
    - Overhead costs 5.1
    - Loan loss provisions 1.7
    - Reserve requirements + deposit insurance premium 0.4
    - Tax 2.1
    - Profit margin 4.8
  - Note: Source listed as World Bank and International Monetary Fund Financial Sector Assessment Program (FSAP) Update 2004. Data are from annual financial statements for 2003–04 in Uganda and 2002 in Kenya; calculations are averages across banks weighted by market share in the lending market.

### Competition analysis: methodology and data
- Purpose: measure and document competition in Uganda’s banking system and analyze whether privatization of UCB to Stanbic and 2002 consolidation changed competition; analyze role of foreign ownership and bank size.  
- Chosen approach: Panzar and Rosse (PR) methodology using bank-level data (H-statistic).  
- PR H-statistic interpretation: < 0 collusive (joint monopoly); < 1 monopolistic competition; 1 perfect competition.  
- Model: reduced-form revenue equation for quarterly panel of 15 banks from March 1999 to June 2004:
  - ln(Rit) = α + β1 ln(UPLit) + β2 ln(UPFit) + β3 ln(UPCit) + γ1 ln(TAit) + γ2 ln(RC1it) + γ3 ln(RC2it) + γ4 DV + εit
  - where R_it is ratio of gross interest revenue (or total revenue) to total assets; UPL is personnel expenses/total assets; UPF is interest expenses/total deposits; UPC is other operating/admin expenses/total assets.  
  - Exogenous/bank-specific controls: TA (total assets), RC1 (ratio of non-performing loans), RC2 (ratio of net loans to total assets). All in logs.  
  - Dummy variables (DV): foreign-owned banks (FB), large banks (LB), small banks (SB), and post-privatization period (September 2002 onwards).  
  - Macroeconomic controls: nominal treasury bill rate (NTBR) and headline inflation (HINFL).  
- H-statistic defined as β1 + β2 + β3; joint standard error reported.  
- Estimation choices: fixed effects or random effects based on Hausman test; pooling tested by F test that all μi = 0.  
- Equilibrium tests reported in Appendix indicate the Ugandan banking system was in equilibrium during most of the period.

### Estimation results and interpretation
- Panzar and Rosse classification: Ugandan banking sector characterized by monopolistic competition.  
- H-statistics:
  - Average for entire sample period (March 1999 to June 2004): 0.39
  - Pre-privatization period: 0.3
  - Post-privatization and consolidation period: 0.49
- Results: H-statistic consistently between 0 and 1 across specifications.  
- Statistical significance: model includes a number of statistically significant variables; variance in H-statistics across specifications noted.  
- Comparison across sub-samples: equilibrium tests more robust for pre- and post-privatization sub-samples; emphasis placed on sub-sample estimates.  
- Change in competition: statistically significant increase in competition (H-statistic) between pre-privatization and post-privatization periods; Wald test rejects equality of pre- and post-privatization H-statistics.  
- Cross-country comparison caution: degree of competition in Uganda appears low compared to comparable countries in the region, especially in the pre-privatization period, but has caught up toward levels observed in Ghana and Kenya since privatization of UCB in September 2002.

*Source: Box 1, “Privatization of UCB and Stanbic’s Performance,” _wp05240 - Box 1. Privatization of UCB and Stanbic’s Performance*

### 2002. It appears that Uganda’s monopolistic market structure is slightly less competitive than

### _wp05240 - 2002. It appears that Uganda’s monopolistic market structure is slightly less competitive than

### Market structure and Panzar and Rosse results
- H-Statistic (overall reported in Table 3):
  - Total Revenue specifications: 0.437, 0.311, 0.498
  - Total Revenue-Ratio specifications: 0.359, 0.293, 0.505
  - Interest Revenue-Ratio specifications: 0.376, 0.300, .460
  - SE: .0457, 0.056, 0.052, 0.046, 0.0528, 0.050, 0.048, 0.060, .097 (reported across specifications)
- Statistical significance note:
  - The H-statistic in the post-privatization period is statistically significantly different from the pre-privatization level at the 5 percent significance level in the first two specifications and at the 10 percent significance level under the interest-revenue specification.
- Regression diagnostics and sample:
  - No. of Obs.: 307, 196, 97 (across reported specifications)
  - R-squared reported values include: 0.98, 0.97, 0.98, 0.30, 0.32, 0.56, 0.42, 0.44, 0.34
  - Test of significance statistics: 14386.6, 58.44, 1813.83, 11.82, 14.42, 70.01, 163.30, 109.41, 65.23 (Prob > stat 0.00 for all)
  - F-test reported values: 6.18, 4.52, 2.90, 4.62, 4.64, 2.98, 8.00, 5.50, 6.42 (Prob > F 0.00, 0.00, 0.001, etc.)
  - Hausman test values and Prob > chi2 reported per specification (examples): 5.48 (Prob > chi2 0.790), 23.24 (Prob > chi2 0.003), 7.34 (Prob > chi2 0.124), 85.16 (Prob > chi2 0.00), etc.
- Coefficients noted (selected, exact as reported):
  - UPL (unit price of labor): 0.213**, 0.109*, 0.222**, 0.147**, 0.093*, 0.217**, 0.160**, 0.125**, 0.227**
  - UPF (unit cost of funds): 0.098**, 0.140**, 0.109*, 0.132**, 0.137**, 0.114*, 0.168**, 0.132**, 0.199**
  - UPC (unit cost of fixed assets): 0.126**, 0.062**, 0.167*, 0.079**, 0.063**, 0.174**, 0.048*, 0.043, 0.033
  - TA (scale variable): 0.964**, 1.067**, 0.940**
  - RC1 (non-performing loans ratio): -0.022*, -0.051**, -.011, -0.023*, -0.053**, -0.010, -0.044**, -0.073**, -0.026
  - NTBR (nominal treasury bill rate): 0.007**, 0.009**, 0.004, 0.007**, 0.009**, 0.003, 0.011**, 0.013**, 0.014**
  - HINFL (inflation): -0.001, 0.007, -.019*, -0.001, 0.005, -0.016*, -0.001, 0.006, -0.020**
  - SB (small bank dummy): -0.280**, -0.355*, -0.27**, -0.106, -0.050, -0.224
  - Sep2002 dummy: -0.001, -0.005, -0.122**
  - Constants and other coefficients reported in Table 3 (see table for exact placement across specifications).

### Banking sector market structure in selected SSA countries (Table 4)
- Ghana: Sample Period 1998 - 2003; H statistic 0.56; Number of banks 13; Number of observations 65
- Kenya: Sample Period 1994 - 2001; H statistic 0.58; Number of banks 34; Number of observations 106
- Nigeria: Sample Period 1994 - 2001; H statistic 0.67; Number of banks 42; Number of observations 186
- South Africa: Sample Period 1994 - 2001; H statistic 0.85; Number of banks 45; Number of observations 186
- Uganda (pre/post split):
  - Uganda 1999Q1 – 2002Q3: H statistic 0.30; Number of banks 15; Number of observations 196
  - Uganda 2002Q4 – 2004Q2: H statistic 0.49; Number of banks 15; Number of observations 97

### Key findings from revenue functions (authors’ interpretations)
- UPL (unit price of labor):
  - Statistically significant in all specifications with comparable positive elasticities.
  - Interpretation: personnel costs are as important as overhead costs; overhead costs are relatively high in Uganda (see Section IV B).
- UPF (unit cost of funds):
  - Significant in all specifications.
  - Has a higher impact on interest revenue rather than other revenue in the post-privatization period, suggesting better market responsiveness post-privatization.
- UPC (unit cost of fixed assets):
  - Determinant of total revenue, but not of total interest revenue.
  - Possible explanation: importance of private money transfers and investment costs (e.g., in ATMs) with fee-based revenues; insignificance for interest revenue may indicate weak competition in those services.
- TA (scale variable):
  - Statistically significant and large.
  - Implication: size is a major determinant for total revenues; strong economies of scale exist.
  - Policy implication noted by authors: profitability is skewed toward larger banks, implying scope for greater consolidation, especially if government securities cease to be a relatively high-yield, risk-free income source.
- RC1 (non-performing loans ratio):
  - Statistically significant negative effect only in the pre-privatization period, not in the post-privatization period.
  - Interpretation: progress in cleaning up banking system toward negligible non-performing loans.
- NTBR (nominal treasury bill rate):
  - Significant positive effect on total and interest revenues in pre-privatization period; effect dissipated since UCB’s privatization.
  - Interpretation: potential private sector crowding-out effects due to government sterilization operations declined post-privatization.
- Small bank dummy (SB):
  - Significantly negative sign in the post-privatization period.
  - Interpretation: small banks are vulnerable in a more competitive environment, particularly if treasury-bill revenue sources decline.
- Inflation (HINFL):
  - Negative effect on revenues in the post-privatization period.
  - Interpretation: reflects banking system’s greater exposure to agriculture; lower rural economic activity and headline inflation rises driven by food-price surges due to crop failures.

### Efficiency: methodology and model
- Efficiency concept:
  - Technical efficiency under variable returns to scale is measured.
  - DEA (Data Envelopment Analysis) used to estimate empirical frontier; chosen for ability to describe frontiers rather than central tendencies.
- Intermediation approach:
  - Banks modeled as intermediating funds between savers and investors (Sealey and Lindley, 1977).
  - Inputs and outputs are typically balance-sheet stocks proportional to flows.
- Inputs and outputs modeling (Table 5 descriptive statistics in Ugandan shillings):
  - Outputs: Deposits (y1), securities (y2), loans (y3)
    - Mean: y1 43,274,886; y2 33,102,894; y3 38,511,183
    - S.D.: y1 64,870,772; y2 65,812,289; y3 48,508,621
    - Median: y1 15,215,724; y2 12,447,268; y3 16,825,453
    - Minimum: y1 460,391; y2 0; y3 0
    - Maximum: y1 394,754,996; y2 384,909,064; y3 203,289,125
  - Inputs: deposits (x1), loans and contingent liabilities (x2)
    - Mean: x1 99,755,085; x2 20,031,790
    - S.D.: x1 147,635,425; x2 34,702,396
    - Median: x1 39,509,113; x2 4,502,236
    - Minimum: x1 1,252,453; x2 0
    - Maximum: x1 782,394,619; x2 340,195,057
- Sample construction:
  - Sample covers Ugandan commercial banks existing at end-June 2004.
  - Balanced sample uses 20 quarterly observations from June 2004 back to third quarter of 1999.
  - Given 14 banks and 20 periods, panel size for regression analysis: 270.
  - All panel data are in end-2003 prices, deflated by the consumer price index.
  - One bank excluded as a “self-identifier” due to unusual input/output mix.
  - In cases of mergers and acquisitions during the 20 quarters, the larger of the combined banks is kept for pre-merger periods.

### DEA efficiency scores: results and evolution (Table 6 and figures)
- 20-quarter period summary (14-bank sample):
  - Average efficiency scores varied between a minimum of 95 percent and a maximum of 100 percent over the 20-quarter period.
  - Overall (unweighted) mean: 99 percent.
  - Weighted by market share in deposits: overall mean 99 percent.
  - Minimum efficiency across periods ranged between 69 and 100 percent.
  - The frontier was formed by between 10 and 14 banks across periods.
  - Five of the 14 banks were on the frontier (fully efficient) in each of the 20 time periods.
- Table 6 selected quarterly values (Mean, Mean weighted by customer deposits, Standard Deviation, Maximum, Minimum, Number of banks w/ score below 1.000):
  - Sep-99: Mean 0.99; Mean weighted 0.99; S.D. 0.05; Maximum 1.00; Minimum 0.82; Number below 1.000 = 12
  - Dec-99: Mean 1.00; Mean weighted 1.00; S.D. 0.00; Maximum 1.00; Minimum 0.99; Number below 1.000 = 13
  - Mar-00: Mean 0.99; Mean weighted 1.00; S.D. 0.03; Maximum 1.00; Minimum 0.89; Number below 1.000 = 12
  - Jun-00: Mean 0.99; Mean weighted 1.00; S.D. 0.04; Maximum 1.00; Minimum 0.85; Number below 1.000 = 13
  - Sep-00: Mean 0.99; Mean weighted 0.99; S.D. 0.03; Maximum 1.00; Minimum 0.88; Number below 1.000 = 12
  - Dec-00: Mean 0.97; Mean weighted 0.99; S.D. 0.08; Maximum 1.00; Minimum 0.69; Number below 1.000 = 12
  - Mar-01: Mean 0.95; Mean weighted 0.98; S.D. 0.09; Maximum 1.00; Minimum 0.74; Number below 1.000 = 10
  - Jun-01: Mean 0.99; Mean weighted 0.99; S.D. 0.03; Maximum 1.00; Minimum 0.90; Number below 1.000 = 12
  - Sep-01: Mean 0.98; Mean weighted 0.99; S.D. 0.06; Maximum 1.00; Minimum 0.79; Number below 1.000 = 12
  - Dec-01: Mean 0.99; Mean weighted 0.99; S.D. 0.02; Maximum 1.00; Minimum 0.95; Number below 1.000 = 11
  - Mar-02: Mean 1.00; Mean weighted 1.00; S.D. 0.01; Maximum 1.00; Minimum 0.95; Number below 1.000 = 13
  - Jun-02: Mean 1.00; Mean weighted 1.00; S.D. 0.00; Maximum 1.00; Minimum 1.00; Number below 1.000 = 14
  - Sep-02: Mean 0.99; Mean weighted 1.00; S.D. 0.02; Maximum 1.00; Minimum 0.94; Number below 1.000 = 12
  - Dec-02: Mean 0.99; Mean weighted 1.00; S.D. 0.02; Maximum 1.00; Minimum 0.92; Number below 1.000 = 13
  - Mar-03: Mean 0.99; Mean weighted 0.99; S.D. 0.04; Maximum 1.00; Minimum 0.83; Number below 1.000 = 12
  - Jun-03: Mean 0.99; Mean weighted 1.00; S.D. 0.02; Maximum 1.00; Minimum 0.93; Number below 1.000 = 13
  - Sep-03: Mean 0.99; Mean weighted 1.00; S.D. 0.02; Maximum 1.00; Minimum 0.92; Number below 1.000 = 12
  - Dec-03: Mean 0.98; Mean weighted 0.99; S.D. 0.06; Maximum 1.00; Minimum 0.84; Number below 1.000 = 12
  - Mar-04: Mean 0.96; Mean weighted 0.99; S.D. 0.08; Maximum 1.00; Minimum 0.78; Number below 1.000 = 10
  - Jun-04: Mean 0.99; Mean weighted 0.92; S.D. 0.04; Maximum 1.00; Minimum 0.84; Number below 1.000 = 13
  - Total (summary): Mean 0.99; Mean weighted 0.99; Standard Deviation 0.05; Maximum 1.00; Minimum 0.70
- Visual patterns reported:
  - Generally high efficiency scores indicate a relatively homogenous bank efficiency distribution.
  - Weighted mean is more stable than unweighted mean, suggesting smaller banks had more volatile scores.
  - Increasing divergence between weighted and unweighted means suggests smaller banks have recently fallen back relative to larger banks.
- Robustness checks:
  - Alternative model specification (treating loans and contingent liabilities as separate inputs) yields hardly changed results.
  - Splitting the sample into two periods (up to Q3 2002 and thereafter) and establishing a single frontier for each period:
    - For the first part (with UCB included), average efficiency 91.6 percent.
    - For the second part, average efficiency 92.9 percent.
    - Interpretation: slight improvement in efficiency aligned with increase in competition after privatization of UCB.
  - Splitting sample and excluding seven smallest banks (assets < U Sh 100 billion at end-June 2004):
    - Average efficiency first period 93.9 percent; second period 96.8 percent.
    - Interpretation: efficiency improvement robust to excluding smaller banks; smaller banks are less efficient.

### Additional interpretative points and links between competition and efficiency
- The authors note that increased competition is expected to lead to lower costs and enhanced efficiency as a first-order effect, but:
  - Recent research shows relationships between competition and banking performance, access to financing, stability, and growth are complex.
  - Market power in banking may to a degree be beneficial for access to financing (reference to Petersen and Rajan, 1995).
  - The view that competition is unambiguously good in banking is more naive than in other industries; vigorous rivalry may not be first-best for financial sector performance.
- The paper proceeds to analyze productivity change and the Malmquist index decomposition to separate efficiency change from frontier shifts (Malmquist results introduced as next topic).

*Source: Authors’ calculations (content unit _wp05240).*

### 1.1 percent (right column) by quarter, split about half (0.6 percent and 0.5 percent on

### _wp05240 - 1.1 percent (right column) by quarter, split about half (0.6 percent and 0.5 percent on

### Productivity change (Table 7 summary)
- 20-Period Geom. Mean of Changes:
  - Change Relative to t-1 Frontier (A): 1.006
  - Frontier Shift from t-1 and t (B): 1.005
  - Productivity (=A*B): 1.011
- Interpretation:
  - Productivity growth of 1.1 percent by quarter, split about half (0.6 percent and 0.5 percent on average, respectively) between changes in efficiency (relative to the frontier) and shifts in the frontier itself.

### Explaining efficiency differences (estimation approach and data)
- Methodology:
  - Efficiency scores pooled over 20 quarters.
  - Dependent variable bounded between zero and unity → Tobit model used.
  - Heteroskedasticity accounted for with QML (Huber/White) standard errors and covariances.
- Sample:
  - Note: N=270.
  - Sources: Bank of Uganda; authors’ calculations.
- Descriptive statistics (Table 8):
  - SIZE: Deposit liabilities, in Ugandan shillings — Mean 92,543,643, S.D. 1.36e+08
  - SCOPE: Herfindahl index of three largest asset categories (1 = max. specialization, 0.33 = max. diversification) — Mean 0.41, S.D. 0.09
  - FOREIGN: Dummy = 1 for banks that were (by majority) foreign-owned at the end of the sample period — Mean 0.86, S.D. 0.35
  - ADMIN: Administrative expenses in percent of total assets — Mean 3.0, S.D. 3.3
  - PUBLIC: Credit to the public sector in percent of total credit — Mean 42.4, S.D. 20.3

### Factors explaining efficiency differences (Table 9: regression results)
- Estimated coefficients (Coefficient, Standard Error, z-Statistic, Probability):
  - SIZE: 8.31E-10, 3.07E-10, 2.709270, 0.0067**  
    - Significance: ** significant at the 1 percent level.
  - SCOPE: 2.539935, 0.173531, 14.63677, 0.0000**  
    - Significance: ** significant at the 1 percent level.
  - FOREIGN: 0.083253, 0.039836, 2.089887, 0.0366*  
    - Significance: * significant at the 5 percent level.
  - ADMIN: -0.013141, 0.005938, -2.212885, 0.0269*  
    - Significance: * significant at the 5 percent level.
  - PUBLIC: 0.005845, 0.001806, 3.237557, 0.0012**  
    - Significance: ** significant at the 1 percent level.
- Key empirical findings:
  - Efficiency increases with SIZE.
  - Efficiency increases with portfolio diversification (higher SCOPE).
  - FOREIGN-owned banks are on average more efficient than local banks.
  - Efficiency decreases with higher administrative expenses (ADMIN).
  - Efficiency increases with the share of credit to the public sector (PUBLIC), a somewhat counterintuitive result.

### Robustness and excluded variables
- Other specifications and univariate regressions produced similar significance levels and coefficient signs.
- Variables tried but not significant in univariate or multivariate regressions:
  - Share of deposits in total funds.
  - Share of foreign currency deposits in total deposits.
  - Year of establishment.
  - Proxy for degree of a bank’s risk aversion.
  - GDP growth.

### Conclusions and policy-relevant findings
- Main finding:
  - The Ugandan banking system has become more competitive and efficient following reforms (cleanup, consolidation, privatization of state-owned banks, strengthened regulatory and supervisory framework).
- Competition and foreign ownership:
  - Concerns that foreign ownership and concentration stifle competition and access appear unfounded in Uganda’s case.
  - System still characterized by a monopolistic market structure with scope to enhance competitive pressures, particularly in fee-based services (nontransparent fee structure identified as an issue).
- Scale and risks to small banks:
  - Scale matters substantially; small banks may come under pressure as competitive pressures build, especially if the supply of treasury bills dries up.
  - Reliance on government securities has potentially crowded out private sector lending, though this effect may have dissipated recently.
  - Stress testing noted in the paper: even a 50 percent reduction in government securities yield would not severely damage small banks.
- Policy recommendations / areas to address:
  - Reduce net treasury bill issuance to lessen banks’ dependence on government securities, encouraging banks to seek new lending opportunities and expand customer bases.
  - Address high overhead, personnel, and loan loss provisioning costs by improving infrastructure, labor market flexibility, and commercial courts/insolvency regime to facilitate financial intermediation.

### Appendix — Equilibrium tests (specification and results)
- Specification (equilibrium test):
  - In(ROA_it) = α + β1 In UPL_it + β2 In UPF_it + β3 In UPC_it + γ1 In TA_it + γ2 In RC1_it + γ3 In RC2_it + γ4 DV + ε_it
  - ε_it = μ_i + ν_it
  - H = β1 + β2 + β3 = 0
  - ROA is pre-tax return on assets; dependent variable computed as ln(1+ROA) because ROA can be negative.
  - Wald-test used for H=0 hypothesis.
- Interpretation of results:
  - Market equilibrium condition cannot be rejected at the 5 percent level for the post-privatization period and at the 10 percent level for the pre-privatization period; close to rejection at the 10 percent level for the whole period.
  - With possible exception of input price of fixed capital, input prices generally not statistically different from zero across sub-periods.
  - Returns on assets positively associated with the loan ratio in both periods.
  - Nominal interest rate, inflation and NPL ratio were important for profit in the pre-privatization period but not in the post-privatization period, where foreign banks were associated with larger profitability.
- Table 1 — Results of the Equilibrium Tests for Uganda (selected reported values):
  - Columns: (1) Random effects (whole period, No. of Obs. 307), (2) Fixed effects (sub-period, No. of Obs. 196), (3) Random effects (sub-period, No. of Obs. 97)
  - Sample coefficient signs and significance (selected entries):
    - UPC: (1) -0.019**, (2) -0.012**, (3) -0.008*
    - TA: (1) 0.004, (2) 0.050**, (3) 0.004
    - RC1: (1) -.003*, (2) -0.005**, (3) -0.001
    - RC2: (1) 0.009, (2) 0.095**, (3) 0.062**
    - NTBR: (1) 0.001*, (2) 0.001**, (3) 0.000
    - HINFL: (1) 0.000, (2) 0.002**, (3) -0.000
    - FB: (1) 0.006, (3) 0.010**
    - Constant: (1) -0.148*, (2) -1.03**, (3) -0.002
  - Goodness-of-fit and tests (selected):
    - R-squared: (1) 0.32, (2) 0.20, (3) 0.30
    - Prob > stat (overall): 0.00 for all reported columns.
    - Wald chi2 Prob > chi2 (H test): (1) 0.08, (2) 0.11, (3) 0.26
    - No. of Obs.: 307, 196, 97
    - F-test Prob > F: (1) 0.00, (2) 0.00, (3) 0.001

*Source: Authors' calculations.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05240.pdf_