## 1. DEA Estimation Results for 2007

## Source details

**Canonical URL:** [1. DEA Estimation Results for 2007](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1095.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1095.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1095.pdf.json)

---

### Context and key questions
- Region covered: five Central American countries, the Dominican Republic, and Panama; Panama treated separately due to financial-center status.
- Period of interest: foreign bank entry wave between 2005 and 2007; efficiency measures and productivity tracked over 2002-07 (Malmquist indices used for 2003-2007).
- Principal research questions:
  - Are foreign banks (global banks and acquired banks) more efficient than local incumbents?
  - How did takeovers affect efficiency of acquired banks and the rest of the banking sector?
  - How much do environment variables affect efficiency estimates?

### Principal findings (DEA and SFA summary)
- Overall conclusion: foreign banks are not necessarily superior in efficiency.
- Relative performance by bank type:
  - On average, local and regional banks have a higher efficiency than domestic banks both during and toward the end of the sample period.
  - Among foreign banks, the acquired banks can keep up with local banks; globally operating banks are clearly inferior.
  - Acquired banks lost ground in efficiency during the year of the takeover (evident in both DEA and SFA).
- Allocative and scale efficiency:
  - In all but one country, foreign-owned banks operate more efficiently on allocative (cost) efficiency; acquired banks often drive this result.
  - Scale efficiency of foreign banks is higher than that of local ones in all but one country; most scale efficiency scores lie between 90 and just under 100 percent relative to best practices.
- Importance of controls:
  - Including environment variables (country-, sector-, bank-specific) changes efficiency estimates considerably.
- Correlation with financial soundness:
  - Correlations between return on assets and SFA efficiency estimates range from -0.68 to 0.76, underlining conceptual differences between financial soundness indicators and productivity measures.

### DEA methodology and measurement
- DEA approach:
  - Linear non-parametric optimization constructing an efficient production frontier; efficiency score is the distance from the frontier.
  - Efficiency scores normalized between 0 and 1 (expressed in percentage terms relative to best practices).
- Three central DEA measures:
  - Technical Efficiency (TE): ability to minimize inputs (or maximize outputs) given targets.
  - Allocative Efficiency (AE): ability to use inputs in optimal proportions given input prices (cost efficiency).
  - Scale Efficiency (SE): portion of TE associated with operating at optimal firm size.
- Outputs and inputs:
  - Output: either sum of loans and securities, or total operational revenue from these assets.
  - Inputs: labor, physical capital, financing; costs: total personnel expenses, other administrative costs including depreciation, total interest paid on deposits.
  - Intermediation approach: deposits treated as an input.
- DEA models computed:
  - (1) Loans and securities as output related to labor/physical capital/financing.
  - (2) Total revenue as output with the three inputs.
  - (3) Revenue as output with expense inputs split into volume and cost to compute allocative efficiency.
  - Scale efficiency derived from model (1) by TE(CRS)/TE(VRS).
- Assumptions:
  - Constant returns to scale assumed in main DEA calculations because average scale efficiency scores are sufficiently close to 100 percent in most cases.

### Sample and classification
- Overall sample: 86 banks in six countries (Costa Rica, the Dominican Republic, El Salvador, Guatemala, Honduras, Nicaragua).
- Observations as of end-December 2007 (data supplied by national regulatory authorities).
- Bank-type definitions:
  - “Foreign” = globally operating banks headquartered outside the region.
  - “Acquired” = banks recently taken over.
  - “Regional” = domiciled in the region and present in at least two countries.
  - “Domestic” = operate only in their own jurisdiction.
  - “Large” vs. “small” determined by quartiles of total assets.
- Panama excluded from regional DEA comparisons; separate DEA for Panama provided in Appendix I.

### Country-level DEA highlights (year-end 2007 — key efficiency scores)
- Costa Rica (No. of banks: 15)
  - Technical Efficiency (Output: Loans + Sec.): 80.1%
  - Technical Efficiency (Output: Revenue): 81.4%
  - Allocative Efficiency: 64.1%
  - Scale Efficiency: 91.8%
- Dominican Republic (No. of banks: 11)
  - Technical Efficiency (Output: Loans + Sec.): 87.7%
  - Technical Efficiency (Output: Revenue): 88.2%
  - Allocative Efficiency: 88.9%
  - Scale Efficiency: 91.8%
- El Salvador (No. of banks: 13)
  - Technical Efficiency (Output: Loans + Sec.): 87.7%
  - Technical Efficiency (Output: Revenue): 82.8%
  - Allocative Efficiency: 60.5%
  - Scale Efficiency: 96.3%
- Guatemala (No. of banks: 20)
  - Technical Efficiency (Output: Loans + Sec.): 76.5%
  - Technical Efficiency (Output: Revenue): 78.7%
  - Allocative Efficiency: 69.5%
  - Scale Efficiency: 96.9%
- Honduras (No. of banks: 18)
  - Technical Efficiency (Output: Loans + Sec.): 85.2%
  - Technical Efficiency (Output: Revenue): 83.8%
  - Allocative Efficiency: 62.1%
  - Scale Efficiency: 96.7%
- Nicaragua (No. of banks: 9)
  - Technical Efficiency (Output: Loans + Sec.): 80.7%
  - Technical Efficiency (Output: Revenue): 93.4%
  - Allocative Efficiency: 89.5%
  - Scale Efficiency: 85.6%

### Key cross-country and bank-type findings (selected)
- Large vs. small banks:
  - Large banks generally more efficient than small banks in technical efficiency (assets) in five out of six countries.
  - Allocative efficiency advantage of large banks is inconsistent across countries.
  - Scale efficiency of large banks is superior except in Honduras and Nicaragua.
- Foreign vs. local/regional:
  - Foreign banks superior in technical efficiency only in the Dominican Republic, El Salvador, and partly in Guatemala (driven by acquired banks).
  - In Costa Rica and Honduras domestic banks outperform foreign and regional counterparts in technical efficiency.
  - In Nicaragua two regional banking groups of local ownership dominate in technical efficiency.
- Takeover effects:
  - Acquired banks often had higher allocative efficiency than global banks (notably El Salvador).
  - Acquired banks’ productivity tended to slow upon acquisition; takeover dummy in SFA is significantly negative.
- Heterogeneity:
  - Guatemala and Nicaragua show high homogeneity (low spread); El Salvador and Honduras show larger dispersion with outliers affecting country scores.

### Multi-year productivity (Malmquist indices, 2003-07 and 2006-07 takeover subperiod) — selected aggregate 2003-07 Δ TFP (loans & securities and revenue) and 2006-07 takeover-period changes
- Costa Rica (Overall 15): 2003-07 Δ TFP Loans&Sec.: +7.6% | 2003-07 Δ TFP Revenue: +3.1% | 2003-07 Δ Relative Efficiency: -0.1% | 2006-07 Δ TFP Loans&Sec.: +0.9% | 2006-07 Δ TFP Revenue: -0.8% | 2006-07 Δ Relative Efficiency: -5.6%
- Dominican Republic (Overall 8/10): +3.2% | -3.3% | +2.9% | +13.6% | -0.1% | +3.5%
- El Salvador (Overall 13): +1.9% | +1.0% | +2.4% | 0.0% | +1.9% | -1.5%
- Guatemala (Overall 18): +3.9% | +3.3% | +1.0% | -3.7% | +8.4% | -9.3%
- Honduras (Overall 15/16): +3.8% | -4.3% | +1.1% | +7.2% | -3.2% | -2.8%
- Nicaragua (Overall 7/9): +4.8% | +0.4% | -1.0% | -5.6% | -0.4% | +2.4%
- Interpretative summary:
  - Productivity in managing loans and securities increased in all countries (between 1.9 percent and 7.6 percent a year).
  - Productivity with respect to revenue is less clear: only three countries registered an annual productivity gain of 1 percent or more.
  - Foreign banks did not consistently achieve higher productivity gains than local institutions; global banks outperformed competitors only in Costa Rica, Nicaragua and the Dominican Republic.
  - During 2006-07, productivity in managing assets was weaker in four countries relative to the whole period; change in relative efficiency is on average positive for 2003-07 but negative for the takeover period.

### Pre-/Post-Acquisition performance (acquired regional banks)
- Sample: five banking groups acquired during 2005-06 by institutions headquartered in the United States, the United Kingdom and Colombia.
- Absolute Δ Productivity (mean) regarding revenue for acquired banks:
  - T-1: +4.5% (Standard deviation: 12.1)
  - T: -0.5% (Standard deviation: 13.6)
  - T+1: +1.2% (Standard deviation: 20.1)
- Absolute Δ Productivity (median):
  - T-1: +3.5% | T: -1.1% | T+1: +2.6%
- Interpretive note: Productivity of acquired banks rose on average pre-acquisition but turned slightly negative in the takeover year; post-acquisition rebound is modest and often vanishes relative to industry-wide gains.

### Stochastic Frontier Analysis (SFA) — methodology and key results
- SFA model:
  - ln(y_i) = β x_i + v_i - u_i, with v_i ~ iid N(0, σ_v^2) and u_i non-negative (truncated normal).
  - Inefficiency contribution: γ = σ_u^2 / (σ_v^2 + σ_u^2).
  - Technical efficiency: TE = exp(-u_i); Battese and Coelli (1988) conditional expectation used to estimate E[u_i | v_i - u_i].
- Estimated models and regressions:
  - Two model structures: Model 1 (error components, Battese and Coelli (1992)) and Model 2 (technical efficiency effects, Battese and Coelli (1995)).
  - Six regressions: (1)–(3) cross section for 2007; (4)–(6) panel 2002-07. Regressions (3) and (6) include environmental correlates.
- Environmental correlates used (expected signs noted):
  - GDP per capita — positive expected sign.
  - Concentration (share of three largest banks) — ambiguous; empirical result negative.
  - Freedom (Heritage Foundation index, 0–100) — positive expected sign.
  - Bank-specific: NPL (negative), Liquidity (negative expected), Takeover dummy (negative).
- Hypothesis tests (likelihood ratio):
  - Translog preferred to linear in several regressions (rejected linear at 1%-level in regressions (1), (2), (4), (5)).
  - Inefficiency effects present in regressions (2) and (5); γ highly significant in panel regressions.
  - Environmental correlates jointly significant in specifications (3) and (6).
  - Time-invariant inefficiency (η = 0) cannot be rejected in panel regression (4).
- Selected estimation highlights (examples preserved exactly from source table):
  - Coefficient examples with standard errors and significance markers:
    - GDP/Capita: 0.287 (0.086)*** ; 2.067 (0.559)***
    - Concentration: -1.301 (0.354)*** ; -3.779 (0.771)***
    - Freedom: -0.020 (0.334) ; 4.906 (1.469)***
    - NPL: -0.904 (0.444)** ; -4.334 (1.167)***
    - Liquidity: -0.775 (0.265)*** ; 2.068 (0.740)***
    - Takeover: -0.181 (0.910)** ; -1.006 (0.353)***
  - γ examples: 0.573 (0.298)* ; 0.972 (0.066)*** ; 0.000 (0.000) ; 0.859 (0.026)*** ; 0.979 (0.008)*** ; 0.903 (0.022)***
  - Log-Likelihoods: -17.686, -16.640, 2.389, 52.120, -53.719, -46.335
- SFA confirmations relative to DEA:
  - Foreign banks are on average not more efficient than local/regional competitors; global banks consistently score lower than banks they acquired (with some regression-specific exceptions).
  - Efficiency scores tended to improve in latter years for Dominican Republic, El Salvador, Guatemala, Nicaragua by about ½ to 2 percentage points.
  - Takeover process impairs technical efficiency — negative takeover dummy.

### Selected SFA efficiency estimates (year-end 2007 and averages)
- Country-level (year-end 2007 and average 2002-07 in parentheses where provided):
  - Costa Rica: 77.5 82.9 56.3 66.7 83.1 (83.1) 79.6 (80.4)
  - Dominican Repub.: 80.5 85.9 80.9 74.4 86.3 (84.7) 86.3 (86.2)
  - El Salvador: 79.6 84.0 62.9 67.5 83.5 (81.8) 79.9 (79.7)
  - Guatemala: 87.5 91.1 74.1 84.7 92.0 (90.1) 91.4 (89.0)
  - Honduras: 75.6 80.3 64.3 62.5 78.7 (83.2) 78.5 (84.3)
  - Nicaragua: 85.9 89.9 83.7 80.4 90.3 (89.7) 93.2 (92.2)
- Bank-type averages (selected, 2007 and type averages where provided):
  - Foreign Banks: 80.1 84.1 68.8 71.4 83.6 (n/a) 82.5 (n/a)
    - o/w global: 76.5 79.6 66.7 61.7 77.7 (81.1) 74.9 (79.4)
    - o/w acquired: 82.2 86.8 70.0 77.2 87.2 (87.3) 87.0 (87.2)
  - Regional Banks: 79.6 84.9 70.4 67.5 84.1 (83.8) 83.2 (83.3)
  - Local Banks: 82.1 86.8 69.1 74.7 87.2 (86.0) 86.2 (85.8)
- Impact of environmental correlates:
  - Inclusion of correlates changes pure efficiency scores by up to 3.6 percentage points.
  - Example: Nicaragua’s average efficiency score rises by 2.5 and 2.9 points for 2007 and average 2002-07 scores, respectively, when correlates are included.

### Panama (Appendix I) — DEA and productivity notes
- Sector characteristics:
  - More than 70 licensed banks; foreign banks account for more than three fourths of all institutions.
  - Banks may operate under a general license or an international license.
- 2007 DEA efficiency (Panama) — Technical Efficiency (Output: Loans + Sec.), Technical Efficiency (Output: Revenue), Scale Efficiency:
  - Banks with general license (No. of banks 40): 34.3% ; 53.3% ; 59.0%
    - local public (2): 26.8% ; 52.7% ; 40.4%
    - local private (13): 22.3% ; 45.5% ; 50.0%
    - foreign-owned (25): 41.4% ; 57.3% ; 65.1%
  - Banks with international license (No. of banks 31): 51.3% ; 79.7% ; 72.1%
- Malmquist Δ TFP Revenue:
  - 2003-07, banks with general license 30: -2.1% ; international license 20: +1.0%
  - 2006-07, banks with general license 35: -3.5% ; international license 25: +1.0%
- Interpretation:
  - Low DEA efficiency estimates for Panama reflect a large gap between most banks and a few exceedingly efficient foreign-owned banks (statistical outliers); interpret with caution.

### Correlation of return on assets with efficiency (regression (5))
- Correlations (return on assets vs. efficiency):
  - Costa Rica: correlation -0.36
  - Dom. Rep.: correlation 0.57
  - El Salvador: correlation 0.13
  - Guatemala: correlation 0.76
  - Honduras: correlation -0.68
  - Nicaragua: correlation -0.38
- Yearly ROA examples (annual values provided in source for 2002–2007).

### Conclusions and policy implications
- Main conclusions:
  - Foreign banks do not generally outperform local and regional banks in technical efficiency.
  - Acquisitions tended to target relatively efficient banks; acquired banks’ productivity often slowed post-acquisition.
  - Global banks were found to be considerably inferior in efficiency; acquired banks could match local institutions.
  - Correcting for environmental conditions is warranted as country-, sector-, and bank-specific correlates were significant in most regressions.
- Recommendations / further research:
  - Re-assess the regional case when substantially more post-acquisition data are available.
  - Examine latent characteristics driving efficiency differences between bank types and whether foreign-owned banks can catch up after merger processes settle.
  - Expectation: over time, higher competition and externally-driven investment in information technology, human capital, and better risk management should raise overall regional banking system efficiency.

*Source: "1. DEA Estimation Results for 2007" from the supplied IMF content unit.*

### 1. DEA Estimation Results for 2007................................................................

### 1. DEA Estimation Results for 2007

### Context and key questions
- Region covered: five Central American countries, the Dominican Republic, and Panama; Panama treated separately due to financial-center status.
- Period of interest: foreign bank entry wave between 2005 and 2007; efficiency measures and productivity tracked over 2002-07 (Malmquist indices used for 2003-2007).
- Principal research questions:
  - Are foreign banks (global banks and acquired banks) more efficient than local incumbents?
  - How did takeovers affect efficiency of acquired banks and the rest of the banking sector?
  - How much do environment variables affect efficiency estimates?

### Principal findings (DEA and SFA results summarized)
- Overall conclusion: foreign banks are not necessarily superior in efficiency.
- Relative performance by bank type:
  - On average, local and regional banks have a higher efficiency than domestic banks both during and toward the end of the sample period.
  - Among foreign banks, the acquired banks can keep up with local banks; globally operating banks are clearly inferior.
  - Even the acquired banks lost ground in efficiency during the year of the takeover (evident in both DEA and SFA).
- Country-specific highlights:
  - Degree of foreign bank entry varies by country, ranging from minor (Guatemala and the Dominican Republic) to very high (El Salvador, with close to 95 percent of bank assets now foreign-owned).
  - Foreign banks are superior in technical efficiency only in the Dominican Republic, El Salvador, and partly in Guatemala (driven by acquired banks).
  - In Costa Rica and Honduras domestic banks outperform foreign and regional counterparts in technical efficiency.
  - In Nicaragua two regional banking groups of local ownership dominate others in technical efficiency.
- Allocative (cost) efficiency:
  - In all but one country, foreign-owned banks operate more efficiently than local competitors; acquired banks often drive this result.
  - In four out of five countries with recent foreign entry, acquired banks have higher allocative efficiency than global banks (notably El Salvador).
- Scale efficiency:
  - Scale efficiency of foreign banks is higher than that of local ones in all but one country, but foreign banks do not uniformly outperform regional banks on scale efficiency.
  - Most scale efficiency scores lie between 90 and just under 100 percent relative to best practices.
- Importance of controls:
  - Including variables that control for the environment in which banks operate matters: efficiency estimates change considerably when accounting for country-, sector- and bank-specific factors.
- Correlation note:
  - Appendix II (bottom) reports correlations between return on assets and SFA efficiency estimates ranging from -0.68 to 0.76, underlining conceptual differences between financial soundness indicators and productivity measures.

### DEA methodology and measurement
- DEA approach:
  - Linear non-parametric optimization constructing an efficient production frontier; efficiency score is the distance from the frontier.
  - Efficiency scores normalized between 0 and 1 (expressed in percentage terms relative to best practices).
- Three central DEA measures:
  - Technical Efficiency (TE): ability to minimize inputs (or maximize outputs) given targets.
  - Allocative Efficiency (AE): ability to use inputs in optimal proportions given input prices (also referred to as cost efficiency).
  - Scale Efficiency (SE): portion of TE associated with operating at optimal firm size.
- Data and variable specification:
  - Output: either sum of loans and securities, or total operational revenue earned from these assets.
  - Inputs: labor, physical capital, and financing; costs correspond to total personnel expenses, other administrative costs including depreciation, and total interest paid on deposits.
  - Cost-efficiency specification decomposes expenses into volume (number of employees, number of bank offices, volume of deposits) and cost components (average compensation, administrative cost per office, interest paid on deposits) to compute allocative efficiency.
  - Intermediation approach: deposits treated as an input.
- DEA models computed:
  - Three DEA models: (1) loans and securities as output related to labor/physical capital/financing; (2) total revenue as output with the three inputs; (3) revenue as output with expense inputs split into volume and cost to compute allocative efficiency.
  - Scale efficiency scores derived from model (1) by dividing TE under constant returns to scale by TE under variable returns to scale.
- Assumptions and rationale:
  - Constant returns to scale assumed in main DEA calculations because average scale efficiency scores are sufficiently close to 100 percent in most cases (larger than 90 percent in 5 countries and larger than 95 percent in three of them).
  - Separate country-level DEA systems used to avoid cross-country comparability issues inherent to DEA.

### Sample and classification
- Sample size and scope:
  - Overall sample: 86 banks in six countries surveyed (Costa Rica, the Dominican Republic, El Salvador, Guatemala, Honduras, Nicaragua).
  - For each country a separate DEA was computed; largest individual country sample: 20 observations; smallest: 9 observations.
  - Observations are as of end-December 2007 (data supplied by national regulatory authorities).
- Bank-type definitions:
  - “Foreign” banks: globally operating banks headquartered outside the region.
  - “Acquired” banks: banks recently taken over.
  - “Regional” banks: domiciled in the region and present in at least two countries (home country + at least one other).
  - “Domestic” banks: operate only in their own jurisdiction.
  - “Large” vs. “small” banks: determined by quartiles of total assets.
- Panama:
  - Panama excluded from regional DEA comparisons due to financial-center status and specific economies of scale; separate DEA for Panama provided in Appendix I.

### DEA results interpretation and caveats
- DEA results are relative and country-specific; efficiency scores indicate distance to the most efficient bank(s) within a country and are not directly comparable across countries.
- Cost-efficiency DEA technical scores are not reported because they likely bias results in favor of banks with very small branch networks (typically foreign banks with only one subsidiary).
- Methodological note:
  - DEA is non-parametric and computationally simple; allows Malmquist indices to track productivity over time.
  - SFA (used elsewhere in the study) complements DEA by decomposing noise from inefficiency but requires functional form and distributional assumptions.

*Source: "1. DEA Estimation Results for 2007" from the supplied IMF content unit.*

### chapter 6 of Coelli et al. (1998).

### _wp1095 - chapter 6 of Coelli et al. (1998)

### DEA Estimation Results for 2007 (Table 1 — country and bank-type efficiency scores)
- Costa Rica (No. of banks: 15)
  - Technical Efficiency (Output: Loans + Sec.): 80.1%
  - Technical Efficiency (Output: Revenue): 81.4%
  - Allocative Efficiency: 64.1%
  - Scale Efficiency: 91.8%
  - Foreign banks (6): 79.5% | 73.2% | 65.4% | 93.6%
    - o/w global banks (2): 88.8% | 81.4% | 60.5% | 91.8%
    - o/w acquired bks (4): 74.9% | 74.9% | 67.3% | 94.5%
  - Regional banks (2): 73.8% | 81.6% | 65.3% | 97.1%
  - Local banks (7): 82.4% | 88.4% | 62.3% | 89.1%
    - o/w public banks (4): 81.3% | 84.6% | 66.4% | 93.0%
  - Large banks (4 (1/0)): 82.4% | 80.8% | 59.5% | 89.7%
  - Small banks (4 (2/0)): 73.0% | 89.4% | 73.6% | 83.8%

- Dominican Republic (No. of banks: 11)
  - Technical Efficiency (Output: Loans + Sec.): 87.7%
  - Technical Efficiency (Output: Revenue): 88.2%
  - Allocative Efficiency: 88.9%
  - Scale Efficiency: 91.8%
  - Foreign banks (2): 97.1% | 90.8% | 91.0% | 94.2%
  - Local banks (9): 85.6% | 87.7% | 88.4% | 90.6%
  - Large banks (3 (0/0)): 98.1% | 96.0% | 83.9% | 96.0%
  - Small banks (3 (0/0)): 69.6% | 79.5% | 85.8% | 81.6%

- El Salvador (No. of banks: 13)
  - Technical Efficiency (Output: Loans + Sec.): 87.7%
  - Technical Efficiency (Output: Revenue): 82.8%
  - Allocative Efficiency: 60.5%
  - Scale Efficiency: 96.3%
  - Foreign banks (9): 88.9% | 87.0% | 63.0% | 96.4%
    - o/w global banks (4): 81.8% | 80.5% | 47.4% | 92.8%
    - o/w acquired bks (5): 94.6% | 92.3% | 75.6% | 99.2%
  - Regional banks (2): 85.3% | 77.3% | 56.6% | 98.7%
  - Local banks (2): 84.3% | 69.6% | 53.3% | 94.0%
  - Large banks (3 (3/0)): 98.4% | 92.9% | 68.5% | 96.5%
  - Small banks (3 (2/1)): 95.0% | 69.9% | 37.2% | 92.9%

- Guatemala (No. of banks: 20)
  - Technical Efficiency (Output: Loans + Sec.): 76.5%
  - Technical Efficiency (Output: Revenue): 78.7%
  - Allocative Efficiency: 69.5%
  - Scale Efficiency: 96.9%
  - Foreign banks (4): 84.9% | 85.9% | 83.2% | 99.2%
    - o/w global banks (1): 83.8% | 69.0% | 100.0% | 97.6%
    - o/w acquired bks (3): 85.3% | 91.6% | 77.6% | 99.7%
  - Regional banks (2): 97.1% | 90.5% | 47.2% | 97.1%
  - Domestic banks (14): 71.2% | 68.6% | 68.8% | 96.2%
  - Large banks (5 (0/0)): 97.0% | 80.5% | 72.0% | 97.0%
  - Small banks (5 (0/1)): 70.1% | 72.2% | 62.3% | 93.8%

- Honduras (No. of banks: 18)
  - Technical Efficiency (Output: Loans + Sec.): 85.2%
  - Technical Efficiency (Output: Revenue): 83.8%
  - Allocative Efficiency: 62.1%
  - Scale Efficiency: 96.7%
  - Foreign banks (7): 83.6% | 78.2% | 62.7% | 96.7%
    - o/w global banks (2): 99.4% | 78.5% | 59.4% | 99.7%
    - o/w acquired bks (5): 77.3% | 78.2% | 64.1% | 95.5%
  - Regional banks (3): 73.3% | 80.7% | 44.6% | 98.4%
  - Local banks (8): 90.9% | 89.8% | 68.0% | 96.0%
  - Large banks (4 (1/0)): 90.6% | 86.8% | 64.8% | 93.0%
  - Small banks (4 (1/1)): 88.8% | 87.0% | 36.2% | 99.0%

- Nicaragua (No. of banks: 9)
  - Technical Efficiency (Output: Loans + Sec.): 80.7%
  - Technical Efficiency (Output: Revenue): 93.4%
  - Allocative Efficiency: 89.5%
  - Scale Efficiency: 85.6%
  - Foreign banks (4): 69.9% | 92.5% | 91.5% | 78.5%
    - o/w global banks (1): 49.2% | 90.7% | 85.2% | 66.7%
    - o/w acquired bks (3): 76.8% | 93.1% | 93.6% | 82.4%
  - Regional banks (2): 98.9% | 94.3% | 86.7% | 98.9%
  - Local banks (3): 80.9% | 94.1% | 88.8% | 86.3%
  - Large banks (2 (0/2)): 98.9% | 94.3% | 86.7% | 98.9%
  - Small banks (2 (1/0)): 100.0% | 100.0% | 90.8% | 100.0%

### Key cross-country findings and market characteristics (text summary)
- Large banks are generally more efficient than small banks, with exceptions:
  - In five out of six countries large banks have superior technical efficiency in terms of assets.
  - When considering costs (allocative efficiency) this advantage is not consistent; large banks have much higher allocative efficiency particularly in El Salvador and Honduras but are outperformed by small banks in Costa Rica, the Dominican Republic and Nicaragua.
  - Scale efficiency of large banks is superior except in Honduras and Nicaragua.
  - Notable competitive concern: Dominican Republic shows a 15 percentage point difference (implying obstacles to competition).
- Country-specific notable characteristics:
  - Costa Rica: public banks hold a significant market share and keep up with other groups; allocative efficiency score is particularly competitive.
  - Guatemala: nine banks are 100 percent technically efficient in either category, implying a wide dispersion in the remaining banks.
  - Honduras: local banks define best practices in efficient input use, leading foreign and regional banks by about 10 percentage points when both categories of technical efficiency are jointly considered.
  - Nicaragua: striking gap between very high average cost efficiency and rather low scale efficiency; gap between technical efficiency in output and revenue terms is evident.

### Multi-year Changes in Factor Productivity (Malmquist indices, 2003-07 and 2006-07 takeover subperiod)
- Methodology notes:
  - Productivity measured using output-oriented Malmquist index for (i) aggregated portfolio of earning loans and securities and (ii) total revenue from such assets.
  - A "Δ Relative Efficiency" column records the catching-up to best-practice banks with 2002 as base year; productivity changes computed for 2003-2007 and 2006-07 isolates takeover-period effects.
  - Productivity time series are deflated using the consumer price index for each country and year.

- Aggregate 2003-07 productivity findings (Table 2 — key summary figures per country)
  - Costa Rica (Overall 15): 2003-07 Δ TFP Loans&Sec.: +7.6% | 2003-07 Δ TFP Revenue: +3.1% | 2003-07 Δ Relative Efficiency: -0.1% | 2006-07 Δ TFP Loans&Sec.: +0.9% | 2006-07 Δ TFP Revenue: -0.8% | 2006-07 Δ Relative Efficiency: -5.6%
    - Global Banks (2): +17.3% | +12.3% | +1.0% | +15.6% | -3.1% | -7.7%
    - Acquired Banks (4): +1.3% | +2.0% | -1.5% | -9.9% | -4.3% | -10.6%
    - Regional Banks (2): +1.2% | +0.6% | +2.3% | +0.2% | +1.9% | +0.6%
    - Domestic Banks (7) o/w Public Banks (4): +10.3% (+12.8%) | +1.8% (+2.0%) | -0.2% (-0.9%) | +3.0% (+10.5%) | +1.4% (+0.3%) | -4.0% (-6.2%)
  - Dominican Republic (Overall 8/10): +3.2% | -3.3% | +2.9% | +13.6% | -0.1% | +3.5%
    - Global Banks (2): +5.9% | -2.9% | 0.0% | +15.3% | -2.0% | +0.6%
    - Domestic Banks (6/8): +2.3% | -3.4% | +3.9% | +13.1% | +0.3% | +4.2%
  - El Salvador (Overall 13): +1.9% | +1.0% | +2.4% | 0.0% | +1.9% | -1.5%
    - Global Banks (4): +1.2% | +0.1% | +1.9% | +1.7% | +6.2% | 0.0%
    - Acquired Banks (5): -0.1% | +0.8% | +3.1% | -7.5% | -0.8% | -2.1%
    - Regional Banks (2): +6.7% | +1.4% | +4.6% | +20.3% | +4.8% | +5.1%
    - Domestic Banks (2): +3.8% | +2.9% | +2.1% | -4.7% | -2.6% | -9.9%
  - Guatemala (Overall 18): +3.9% | +3.3% | +1.0% | -3.7% | +8.4% | -9.3%
    - Global Banks (1): +3.0% | -3.8% | -5.2% | -20.1% | -2.8% | -12.4%
    - Acquired Banks (3): -2.2% | +1.7% | +0.2% | +4.5% | +20.4% | -0.4%
    - Regional Banks (1): +9.0% | +9.4% | +7.5% | +1.2% | +12.1% | -0.1%
    - Domestic Banks (13): +5.0% | +3.8% | +1.2% | -4.8% | +6.3% | -11.9%
  - Honduras (Overall 15/16): +3.8% | -4.3% | +1.1% | +7.2% | -3.2% | -2.8%
    - Global Banks (1): -0.2% | +0.5% | +4.3% | -4.8% | -5.6% | 0.0%
    - Acquired Banks (4/5): +3.1% | -6.0% | +0.5% | +1.9% | -7.0% | -3.7%
    - Regional Banks (2): +2.3% | -11.9% | -2.2% | +9.3% | -7.0% | -4.6%
    - Domestic Banks (8): +5.0% | -2.2% | +1.8% | +11.5% | +0.5% | -2.2%
  - Nicaragua (Overall 7/9): +4.8% | +0.4% | -1.0% | -5.6% | -0.4% | +2.4%
    - Global Banks (1): +18.1% | +7.8% | +1.1% | -3.4% | -5.2% | -4.8%
    - Acquired Banks (2/3): +6.6% | +1.8% | -1.1% | -11.4% | -4.9% | +1.6%
    - Regional Banks (2): +1.5% | -1.1% | -1.2% | -6.3% | -0.2% | +4.0%
    - Domestic Banks (2/3): -0.3% | -3.3% | -1.6% | 0.0% | +5.5% | +4.5%

- Interpretative summary from chapter text:
  - Productivity in managing loans and securities increased in all countries (between 1.9 percent and 7.6 percent a year).
  - Productivity with respect to revenue is less clear: only three countries registered an annual productivity gain of 1 percent or more.
  - Costa Rica and Guatemala improved on both counts; Honduras and Dominican Republic experienced declining efficiency in generating revenue despite increasing productivity in managing assets.
  - Foreign banks did not consistently achieve higher productivity gains than local institutions; global banks outperformed competitors only in Costa Rica, Nicaragua and the Dominican Republic.
  - Domestic and regional banks achieved the strongest productivity gains in El Salvador and Guatemala.
  - The regional banking groups that were later acquired registered below-average productivity advances in all countries except Nicaragua.
  - During 2006-07 (bulk of cross-border acquisitions):
    - Productivity in managing assets was weaker in four countries relative to the whole period; in terms of revenues the reverse is true.
    - Change in relative efficiency is on average positive for the five-year period but negative for the takeover period, indicating some banks lost ground to market leaders.
    - Global banks saw productivity in generating revenue deteriorate in all but one country (El Salvador); declines pronounced in Costa Rica, Honduras and Nicaragua.
    - Domestic banks strengthened relative positions in some takeover countries; acquired banking groups improved markedly only in Guatemala.
    - Independent regional banks performed relatively well in the takeover period, with reasonably strong productivity growth in Costa Rica, El Salvador and Guatemala and catching up to best practices in several locations.

### Pre-/Post-Acquisition Performance (Acquired regional banks during takeover phase; Table 3)
- Sample and timing:
  - Five banking groups were acquired during 2005-06 by institutions headquartered in the United States, the United Kingdom and Colombia.
  - Acquisition year (T) defined as year takeover announced and integration begun.
- Absolute productivity change for acquired banks (mean and median, Δ Productivity regarding revenue)
  - Δ Productivity (mean)
    - T-1: +4.5% (Standard deviation: 12.1)
    - T: -0.5% (Standard deviation: 13.6)
    - T+1: +1.2% (Standard deviation: 20.1)
  - Δ Productivity (median)
    - T-1: +3.5%
    - T: -1.1%
    - T+1: +2.6%
- Interpretive note from chapter text:
  - Productivity of acquired banks rose by an average of 4.5 percent in the pre-acquisition year but turned slightly negative in the takeover year.

*Italic source attribution: chapter 6 of Coelli et al. (1998), _wp1095 - chapter 6 of Coelli et al. (1998).*

### 0.5 percent) and recovered somewhat to +1.2 percent thereafter. The rebound is stronger

### _wp1095 - 0.5 percent) and recovered somewhat to +1.2 percent thereafter. The rebound is stronger

### Relative Productivity Change — key findings
- Absolute Δ Productivity (mean) for Acquired Banks:
  - T-1: +4.5%
  - T: -0.6%
  - T+1: -0.2%
  - (Standard deviation): (9.0), (10.8), (16.2)
- Δ Productivity (median) for Acquired Banks:
  - T-1: +4.7%
  - T: 0.0%
  - T+1: -2.2%
- Interpretation:
  - Up until the takeover period absolute and relative changes (means) are virtually identical, indicating lack of cyclical bias.
  - In the post-acquisition year there is a difference of 1.4 percentage points between the two measures (+1.2 percent pre-acquisition vs. -0.2 percent post-acquisition).
  - Average total factor productivity of acquired banks improved in absolute terms, but this improvement vanishes relative to industry-wide gains.
  - The average post-merger performance of acquired banks is not different from the industry average.

### IV. BANK EFFICIENCY BASED ON STOCHASTIC FRONTIER ANALYSIS — A. Methodology
- Stochastic Frontier Analysis (SFA) estimated equation:
  - ln(y_i) = β x_i + v_i - u_i
  - v_i ~ iid N(0, σ_v^2); u_i ~ non-negative, truncated at zero of N(μ, σ_u^2)
- Inefficiency contribution measured by:
  - γ = σ_u^2 / (σ_v^2 + σ_u^2), expressed in percentage terms.
- Technical efficiency:
  - TE = exp(-u_i); best predictor E[u_i | (v_i - u_i)] via Battese and Coelli (1988) conditional expectation formula (explicit functional form provided in source).
- Functional form and distributional choices:
  - Common choices: Cobb-Douglas or translog; half-normal or truncated-normal inefficiency term.
- Two stochastic frontier models estimated:
  - Model 1 (error components model, Battese and Coelli (1992)): inefficiency term u_it = exp[η (t - T)] u_i, η ∈ [0,1]; η = 0 implies time-invariant inefficiency.
  - Model 2 (technical efficiency effects model, Battese and Coelli (1995)): truncation parameter μ_it = z_it δ incorporates environmental variables z_it directly into inefficiency distribution.

### IV. BANK EFFICIENCY BASED ON STOCHASTIC FRONTIER ANALYSIS — B. Data and Variables
- Output variable:
  - Total revenue.
- Inputs:
  - (i) personnel costs (L)
  - (ii) other administrative costs (K)
  - (iii) total interest paid (F)
- Functional forms:
  - Linear (Cobb-Douglas) and translog (with squared and cross terms; total of nine variables in translog).
- Regressions:
  - Six regressions run:
    - Regressions (1)–(3): cross section for 2007.
      - (1): Model 1 (error components)
      - (2): Model 2 (technical efficiency effects) without environmental variables
      - (3): Model 2 with environmental variables
    - Regressions (4)–(6): panel data for 2002-07, mirroring (1)–(3).
- Environmental variables (used in regressions (3) and (6)):
  - Country-specific correlates (expected signs noted):
    - GDP per capita — positive expected sign.
    - Concentration (share of three largest banks in total assets) — sign ambiguous; could be negative (barriers to entry) or positive (efficient structure).
    - Freedom (Heritage Foundation index of financial freedom, 0–100) — higher score expected to have positive effect.
  - Bank-specific correlates:
    - NPL: share of non-performing loans — negative expected sign.
    - Liquidity: share of liquid assets in total assets — negative expected sign (higher liquidity lowers contribution of earning assets).
    - Takeover: dummy = 1 in takeover year, 0 otherwise — negative expected sign.

### IV. BANK EFFICIENCY BASED ON STOCHASTIC FRONTIER ANALYSIS — C. Tests of Hypotheses
- Hypothesis testing approach:
  - Likelihood ratio tests used to test functional form, existence of inefficiency effects, significance of environmental variables, and time-invariant inefficiency.
- Key test outcomes (summary from Table 5):
  - Functional form:
    - Null of linear specification rejected in favor of translog at 1%-level in regressions (1), (2), (4), (5) (LR and critical values reported).
  - Inefficiency effects (γ = μ = 0):
    - In regressions (1) and (4) null not rejected (no inefficiency effects; μ omitted).
    - In regressions (2) and (5) null rejected, confirming inefficiency effects.
  - Significance of environmental correlates:
    - In specifications (3) and (6) null of joint insignificance rejected at least at 5% level.
  - Time-invariant inefficiency (η = 0):
    - In panel model of regression (4) hypothesis of no time variation cannot be rejected; η set to zero.

### IV. BANK EFFICIENCY BASED ON STOCHASTIC FRONTIER ANALYSIS — D. Results of the SFA Procedure
- Specification and significance:
  - Not all translog regressors significant in cross section for 2007 (regressions (1)–(3)), but translog form yields higher likelihood.
  - In panel regressions (4)–(6) most regressors highly significant with expected positive signs.
- Environmental variable results (regressions (3) and (6)):
  - GDP per-capita: benign positive influence where significant.
  - Freedom from government interference: positive in panel model (significant), insignificant in cross section (lack of cross-country variability).
  - NPL: negative, indicating cost burden of high loan delinquencies.
  - Takeover dummy: negative, indicating takeover process impairs technical efficiency.
  - Concentration: negative in both specifications, supporting barriers-to-entry hypothesis.
  - Liquidity: negative in cross section (as expected); in regression (6) liquidity has counterintuitive positive sign.
  - Note: γ significance disappears in cross section regression (3) (possible misspecification, small sample), but γ is highly significant in panel regressions (notably (6)).
- Confirmations relative to DEA findings:
  - Foreign banks are on average not more efficient than local or regional competitors; global banks consistently score lower than banks they acquired (exceptions in regressions (1) and (4) using time-invariant effects).
  - Efficiency scores tended to improve in latter years of sample: scores in Dominican Republic, El Salvador, Guatemala, Nicaragua rose by about ½ to 2 percentage points (comparing average 2002-07 scores listed in parentheses of regression (5) to the 2007 scores of regressions (2) or (5)).
  - Takeover process impairs technical efficiency—negative sign on takeover dummy confirms slump in efficiency in takeover year.

### Table 6 — Selected estimation outcomes and efficiency estimates (high-level)
- Regression-level highlights (selected coefficients with significance indicators preserved from source table):
  - Const. range across regressions: 3.505 (3.841), 3.991 (2.604), 2.403 (3.431), 1.937 (1.749), 5.097 (1.225)***, 4.654 (1.052)***
  - ß1 = Personnel: examples: 0.811 (0.947); 0.108 (0.830); 1.238 (0.339)***; 0.803 (0.308)**; 0.893 (0.326)***
  - ß2 = Admin. cost: examples: -1.323 (0.814); -1.446 (0.734)**; -0.663 (0.340)*; -1.071 (0.288)***; 1.014 (0.301)***
  - ß3 = Interest paid: examples: 1.268 (0.648)*; 1.061 (0.614)*; 1.416 (0.526)***; 0.446 (0.171)***; 0.840 (0.169)***; 0.754 (0.173)***
  - Selected translog terms (ß4..ß9) and squared terms reported with standard errors and significance in table.
  - Environmental variable coefficients (examples from regressions with correlates):
    - GDP/Capita: 0.287 (0.086)*** ; 2.067 (0.559)***
    - Concentration: -1.301 (0.354)*** ; -3.779 (0.771)***
    - Freedom: -0.020 (0.334) ; 4.906 (1.469)***
    - NPL: -0.904 (0.444)** ; -4.334 (1.167)***
    - Liquidity: -0.775 (0.265)*** ; 2.068 (0.740)***
    - Takeover: -0.181 (0.910)** ; -1.006 (0.353)***
  - μ or δ0 reported across regressions with standard errors; γ values reported, examples:
    - γ: 0.573 (0.298)* ; 0.972 (0.066)*** ; 0.000 (0.000) ; 0.859 (0.026)*** ; 0.979 (0.008)*** ; 0.903 (0.022)***
  - Log-Likelihoods: -17.686, -16.640, 2.389, 52.120, -53.719, -46.335
- Efficiency estimates by country / type of bank for year-end 2007 (averages for 2002-2007 in parentheses) — values preserved exactly:
  - Costa Rica: 77.5 82.9 56.3 66.7 83.1 (83.1) 79.6 (80.4)
  - Dominican Repub.: 80.5 85.9 80.9 74.4 86.3 (84.7) 86.3 (86.2)
  - El Salvador: 79.6 84.0 62.9 67.5 83.5 (81.8) 79.9 (79.7)
  - Guatemala: 87.5 91.1 74.1 84.7 92.0 (90.1) 91.4 (89.0)
  - Honduras: 75.6 80.3 64.3 62.5 78.7 (83.2) 78.5 (84.3)
  - Nicaragua: 85.9 89.9 83.7 80.4 90.3 (89.7) 93.2 (92.2)
  - Foreign Banks: 80.1 84.1 68.8 71.4 83.6 (n/a) 82.5 (n/a)
    - o/w global: 76.5 79.6 66.7 61.7 77.7 (81.1) 74.9 (79.4)
    - o/w acquired: 82.2 86.8 70.0 77.2 87.2 (87.3) 87.0 (87.2)
  - Regional Banks: 79.6 84.9 70.4 67.5 84.1 (83.8) 83.2 (83.3)
  - Local Banks: 82.1 86.8 69.1 74.7 87.2 (86.0) 86.2 (85.8)
- Impact of environmental correlates on pure efficiency scores:
  - Inclusion of correlates changes pure efficiency scores by up to 3.6 percentage points.
  - Example: Comparing regression (6) (with correlates) to regression (5) (pure efficiency), Nicaragua’s average efficiency score rises by 2.5 and 2.9 points for 2007 and average 2002-07 scores, respectively; Costa Rica and El Salvador scores drop by between 2.1 and (value truncated in source).

*Source: content excerpts from the supplied PDF chapter/section.*

### 3.6 percentage points (with the other countries exhibiting changes of only up to 1½ percent).

### _wp1095 - 3.6 percentage points (with the other countries exhibiting changes of only up to 1½ percent).

### Regional and cross-country variation in efficiency
- The difference in scores between the most and the least efficient banking markets is generally 11 to 13 percentage points (except in regressions (3) and (4)).
- At year-end 2007:
  - Guatemala and Nicaragua featured the highest technical efficiency.
  - Honduras was the least efficient banking market—partly due to one particularly inefficient international bank.
  - The Dominican Republic was in third place, closely followed by El Salvador and Costa Rica.
- Scores fluctuate over time; some countries experienced dips (Dominican Republic, El Salvador, Honduras), one trended upward (Guatemala), and two stayed largely unchanged (Costa Rica, Nicaragua).

### Effects of environmental, country-, sector-, and bank-specific variables
- Adding environmental correlates to baseline specifications produced measurable, although not decisive, effects on individual and aggregate efficiency scores.
- Disaggregating by country and bank type shows that changes in efficiency caused by country and bank specific variables are generally smaller (up to 1 percentage point), except for global banks:
  - Global banks’ efficiency scores are corrected downward by 2.8 and 1.7 percentage points for 2007 and the 2002-07 average, respectively.
- Other exogenous, unaccounted-for factors likely influenced results: differences in market structure and business mix (e.g., state-owned banks in Costa Rica; corporate banking in Guatemala; cross-border investment in Panama), legal and accounting changes, and banking crises (Dominican Republic 2004; Guatemala 2007).

### Bank-type differences, takeovers, and the “home court advantage”
- Foreign-owned banks are not generally more efficient than local banks according to both DEA and SFA methods.
- Using DEA (year-end 2007):
  - Technical efficiency of foreign banks was lower in three of the six countries surveyed.
  - Foreign banks showed better cost efficiency, largely driven by the banks they acquired, indicating global banks chose to acquire relatively efficient banks.
  - Independent regional banks performed on average at least as well as foreign banks.
- Productivity indices for the acquisition period 2006-07:
  - Domestic and regional banks generally achieved higher productivity gains than international banks, particularly in revenue generation.
  - Banks taken over since 2005 tended to see productivity growth dwindle upon acquisition, likely due to reorganization costs; a takeover dummy in the SFA was significantly negative.
- SFA confirmed DEA findings: globally operating banks were found to be considerably inferior in efficiency, contrary to evidence from transition countries and several other emerging markets.
- The larger correction for global banks is consistent with the “home court advantage” hypothesis (Berger et al., 2000).

### Distribution and within-country heterogeneity
- Figure 3 (Distribution of Efficiency Estimates, 2007 from regression (5)) findings:
  - Banks in the Dominican Republic, Guatemala and Nicaragua are quite homogeneous (low spread between highest and lowest quartile).
  - Variation is larger in El Salvador and Honduras, where one bank each has a particularly low efficiency score.
  - In Honduras the extreme low value likely reflects an entrant bank in 2007 incurring significant set-up costs relative to recorded earnings; accounting for this raises Honduras’ efficiency in regression (5) from 78.7 to 81.5 percent but does not change the regional ordering.
- Correction for the inefficient entrant bank in 2007 raises acquired-bank group scores:
  - In regression (5): from 77.7 to 82.0 percent.
  - In regression (6): from 74.9 to 79.1 percent.
  - These corrected scores remain 1.5 and 3.6 percent lower, respectively, than in 2002.

### Panama (Appendix I): DEA findings and productivity
- Panamanian sector context:
  - More than 70 licensed banks (not counting representative offices).
  - Foreign banks account for more than three fourths of all institutions.
  - Banks may operate under a general license or an international license (lower equity capital, restricted business).
- Caveat: Low DEA efficiency estimates for Panama in 2007 reflect a large efficiency gap between most banks and a few exceedingly efficient foreign-owned banks (statistical outliers); interpret with caution.
- 2007 DEA efficiency (Panama) — Technical Efficiency (Output: Loans + Sec.), Technical Efficiency (Output: Revenue), Scale Efficiency:
  - Banks with general license (No. of banks 40): 34.3% ; 53.3% ; 59.0%
    - o/w local public (No. of banks 2): 26.8% ; 52.7% ; 40.4%
    - o/w local private (No. of banks 13): 22.3% ; 45.5% ; 50.0%
    - o/w foreign-owned (No. of banks 25): 41.4% ; 57.3% ; 65.1%
  - Banks with international license (No. of banks 31): 51.3% ; 79.7% ; 72.1%
  - Note: Allocative efficiency not computed due to incomplete data on inputs.
- Malmquist productivity indices (Δ TFP Revenue):
  - 2003-07, No. of banks:
    - Banks with general license 30: -2.1%
      - o/w local public 2: -1.7%
      - o/w local private 11: -2.0%
      - o/w foreign-owned 17: -2.2%
    - Banks with international license 20: +1.0%
  - 2006-07, No. of banks:
    - Banks with general license 35: -3.5%
      - o/w local public 2: +0.5%
      - o/w local private 12: -3.6%
      - o/w foreign-owned 21: -3.8%
    - Banks with international license 25: +1.0%
- Interpretation:
  - Revenue productivity for banks with general license was negative during both 2003-07 and 2006-07.
  - Foreign banks with international license advanced revenue productivity by 1 percent a year.
  - Among domestic institutions, public banks turned performance around recently while private domestic banks experienced continued erosion of revenue productivity.

### Appendix II — Yearly efficiency estimates (regressions (5) and (6)) by country and by type of bank
- Efficiency Estimates By Country — Regression (5) (each entry is the country-year efficiency):
  - Costa Rica: 2002 84.3% ; 2003 82.2% ; 2004 83.2% ; 2005 82.9% ; 2006 83.1% ; 2007 83.1% ; Country Avg. 83.1%
  - Dom. Rep.: 2002 85.2% ; 2003 84.8% ; 2004 90.2% ; 2005 75.8% ; 2006 85.8% ; 2007 86.3% ; Country Avg. 84.7%
  - El Salvador: 2002 82.6% ; 2003 83.4% ; 2004 79.5% ; 2005 80.8% ; 2006 81.1% ; 2007 83.5% ; Country Avg. 81.8%
  - Guatemala: 2002 88.2% ; 2003 89.1% ; 2004 88.6% ; 2005 91.2% ; 2006 91.6% ; 2007 92.0% ; Country Avg. 90.1%
  - Honduras: 2002 86.3% ; 2003 84.6% ; 2004 81.8% ; 2005 83.2% ; 2006 84.6% ; 2007 78.7% ; Country Avg. 83.2%
  - Nicaragua: 2002 91.1% ; 2003 90.4% ; 2004 91.0% ; 2005 85.8% ; 2006 90.0% ; 2007 90.3% ; Country Avg. 89.7%
  - Yearly Avg.: 2002 86.3% ; 2003 85.8% ; 2004 85.7% ; 2005 83.3% ; 2006 86.0% ; 2007 85.7% ; Overall Avg. 85.4%
- Efficiency Estimates By Country — Regression (6):
  - Costa Rica: 2002 82.5% ; 2003 79.2% ; 2004 80.4% ; 2005 79.7% ; 2006 80.8% ; 2007 79.6% ; Country Avg. 80.4%
  - Dom. Rep.: 2002 86.6% ; 2003 89.5% ; 2004 92.4% ; 2005 75.4% ; 2006 86.7% ; 2007 86.3% ; Country Avg. 86.2%
  - El Salvador: 2002 81.6% ; 2003 82.0% ; 2004 77.5% ; 2005 78.7% ; 2006 78.6% ; 2007 79.9% ; Country Avg. 79.7%
  - Guatemala: 2002 87.2% ; 2003 87.9% ; 2004 86.3% ; 2005 90.9% ; 2006 90.2% ; 2007 91.4% ; Country Avg. 89.0%
  - Honduras: 2002 89.5% ; 2003 88.2% ; 2004 82.6% ; 2005 83.4% ; 2006 83.9% ; 2007 78.5% ; Country Avg. 84.3%
  - Nicaragua: 2002 91.1% ; 2003 92.8% ; 2004 93.1% ; 2005 90.2% ; 2006 92.6% ; 2007 93.2% ; Country Avg. 92.2%
  - Yearly Avg.: 2002 86.4% ; 2003 86.6% ; 2004 85.4% ; 2005 83.1% ; 2006 85.5% ; 2007 84.8% ; Overall Avg. 84.9%
- Efficiency Estimates By Type of Bank — Regression (5):
  - Global: 2002 83.5% ; 2003 83.2% ; 2004 79.1% ; 2005 80.7% ; 2006 82.1% ; 2007 77.7% ; Type Avg. 81.1%
  - Acquired: 2002 88.4% ; 2003 87.5% ; 2004 86.2% ; 2005 86.0% ; 2006 88.3% ; 2007 87.2% ; Type Avg. 87.3%
  - Regional: 2002 85.3% ; 2003 85.2% ; 2004 84.2% ; 2005 82.6% ; 2006 81.4% ; 2007 84.1% ; Type Avg. 83.8%
  - Local: 2002 85.5% ; 2003 85.5% ; 2004 86.5% ; 2005 84.3% ; 2006 87.1% ; 2007 87.2% ; Type Avg. 86.0%
  - Yearly Avg.: 2002 86.3% ; 2003 85.8% ; 2004 85.7% ; 2005 83.3% ; 2006 86.0% ; 2007 85.7% ; Overall Avg. 85.4%
- Efficiency Estimates By Type of Bank — Regression (6):
  - Global: 2002 82.7% ; 2003 82.5% ; 2004 77.5% ; 2005 78.8% ; 2006 80.3% ; 2007 74.9% ; Type Avg. 79.4%
  - Acquired: 2002 88.5% ; 2003 87.4% ; 2004 85.4% ; 2005 86.9% ; 2006 87.7% ; 2007 87.0% ; Type Avg. 87.2%
  - Regional: 2002 85.7% ; 2003 85.0% ; 2004 83.6% ; 2005 81.5% ; 2006 80.8% ; 2007 83.2% ; Type Avg. 83.3%
  - Local: 2002 86.1% ; 2003 86.5% ; 2004 86.1% ; 2005 83.5% ; 2006 86.3% ; 2007 86.2% ; Type Avg. 85.8%
  - Yearly Avg.: 2002 86.4% ; 2003 86.6% ; 2004 85.4% ; 2005 83.1% ; 2006 85.5% ; 2007 84.8% ; Overall Avg. 84.9%

### Return on Assets correlation with Efficiency Estimates of Regression (5)
- Return on Assets (annual values) and correlation with regression (5) efficiencies:
  - Costa Rica: 2002 1.6 ; 2003 1.9 ; 2004 1.7 ; 2005 2.1 ; 2006 2.0 ; 2007 2.6 ; Correlation -0.36
  - Dom. Rep.: 2002 2.8 ; 2003 2.0 ; 2004 2.6 ; 2005 2.0 ; 2006 2.2 ; 2007 2.3 ; Correlation 0.57
  - El Salvador: 2002 1.1 ; 2003 1.1 ; 2004 1.0 ; 2005 1.2 ; 2006 1.5 ; 2007 1.3 ; Correlation 0.13
  - Guatemala: 2002 0.8 ; 2003 1.1 ; 2004 1.3 ; 2005 1.6 ; 2006 1.2 ; 2007 1.8 ; Correlation 0.76
  - Honduras: 2002 1.1 ; 2003 1.6 ; 2004 1.6 ; 2005 2.0 ; 2006 2.2 ; 2007 2.4 ; Correlation -0.68
  - Nicaragua: 2002 1.9 ; 2003 2.1 ; 2004 2.8 ; 2005 2.6 ; 2006 2.4 ; 2007 2.2 ; Correlation -0.38
- Source of return on assets data: IMF Global Financial Stability Report April 2008 and authorities’ websites.

### Conclusions and policy implications
- The study is the first bank efficiency analysis for this region using DEA and SFA on a sample of 86 banks for 2002-07, encompassing a wave of foreign bank entry in the latter part.
- Key findings:
  - Foreign banks do not generally outperform local and regional banks in technical efficiency.
  - Acquisitions tended to target relatively efficient banks; acquired banks’ productivity often slowed post-acquisition.
  - Global banks were found to be considerably inferior in efficiency; acquired banks could match local institutions.
  - Correcting for environmental conditions is warranted as country-, sector-, and bank-specific correlates were significant in most regressions.
- Recommendations / further research:
  - Re-assess the regional case when substantially more post-acquisition data are available.
  - Examine latent characteristics driving efficiency differences between bank types and whether foreign-owned banks can catch up after merger processes settle.
  - Expectation: over time, higher competition and externally-driven investment in information technology, human capital, and better risk management should raise overall regional banking system efficiency.

*Source: _wp1095 - 3.6 percentage points (with the other countries exhibiting changes of only up to 1½ percent).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1095.pdf_